diffusional change in the english system of complementation

KATHOLIEKE UNIVERSITEIT LEUVEN

FACULTEIT LETTEREN

SUBFACULTEIT TAALKUNDE

Diffusional change in the English system of

complementation

Gerunds, participles and for...to-infinitives

Proefschrift ingediend tot het behalen van de graad van Doctor in de

Taal- en Letterkunde: Germaanse Talen

door

Hendrik De Smet

Promotor: Prof. dr. Hubert Cuyckens

Co-promotor: Prof. dr. Olga Fischer

2008

ACKNOWLEDGEMENTS

If the completing of a PhD thesis is immensely rewarding and is bound to lead the

author into a momentary state of euphoria, the actual writing and the research that

precedes it are likely to have had their ups and downs. On looking back, however,

and resisting the temptation (in my present state) to romanticise, I feel I have been

fortunate to know many more ups than downs. In no small measure, this is due to the

people that have been surrounding me over the past five years. To those people I

now wish to express my gratitude for the opportunities they have offered me, and for

the help and support they have given me along the way.

In the first place I thank my promotor Hubert Cuyckens. Inexplicably, I did

not think complementation all that interesting when it was first proposed to me as a

possible topic for my MA thesis – that is, until Hubert made me see that the English

system of complementation combined an intricate internal organisation with all

kinds of puzzling historical developments. After I had completed my MA thesis,

Hubert had to conquer yet another unaccountable prejudice in me, using the gentlest

means of persuasion first to get me to accept a job as research assistant and then to

convince me to apply for a PhD grant. In retrospect, I cannot but thank him for his

persistence and for the confidence he implicitly showed in me. In doing so, he has

opened an exciting new world to me and given me the chance and freedom to make

my own small discoveries in it. In addition to creating those vital opportunities,

moreover, I thank him for his ever-careful judgement, which has been a touchstone

for almost anything I did – indeed, his critical comments, accompanied by constant

demands for clarity and precision, have considerably improved the contents and

form of this thesis. On top of the many hours of linguistic discussion we must have

spent over the years, I am finally grateful for the hours of pleasant non-linguistic

conversation, spent in between Hubert’s classes, meetings, editing work, reviewing

and many other academic duties.

I thank my co-promotor, Olga Fischer, whose scholarly wisdom and warm-

hearted generosity I fully came to appreciate when she gave me the chance to come

and stay in Amsterdam, where she not only made sure that I should have shelter and

pleasant company (animal as well as human), but also provided me with a whole set

of new ideas. The time in Amsterdam has been very inspiring and productive, and

various central ideas in this thesis have first taken root as I reflected on our discus-

sions on language structure and change. Additionally, in the past few months, her

comments on various draft chapters – always sent to me with remarkable speed –

ii – Acknowledgments

have at several points guarded me against oversimplification, inconsistency or

downright unintelligibility.

The length of this thesis must be a measure of the gratitude I wish to express

to the members of my doctoral examination board, Teresa Fanego, Liesbet Heyvaert

and Hans Smessaert, for being prepared to read and judge my work. Liesbet Hey-

vaert I am further indebted to for reading and commenting on my papers on the ger-

und and for introducing me to her own ideas on the English gerund, which have

been a significant source of inspiration.

Among international colleagues, I wish to thank Svenja Kranich, for reading

an earlier version of Chapter 8 and making a number of critical but justified re-

marks. I thank David Denison and several anonymous reviewers of Journal of Eng-

lish Language and Linguistics, who commented on the subsequent versions of the

same text. I am also grateful to Magnus Huber and Robert Shoemaker, who did me a

great favour in giving me access to their digitised version of the proceedings of the

Old Bailey as well as to the as yet unpublished proceedings of the Criminal Court.

To Evelien Keizer I am grateful for our discussions on the fuzziness of grammar.

Anni Sairio is to be thanked for providing me with a bike when I needed one.

Among my colleagues at the Linguistics Department in Leuven, I would like

to thank Kristin, whose encouraging comments on my MA thesis have at the time

meant a great deal to me; Jean-Christophe, who it has been a pleasure and privilege

to work with on a topic of mutual concern, and who has generally been one of those

people that make the Linguistics Department an interesting place; An, who has

taught me how to set up a tent and has joined me in many an adventure involving

North American fauna or Italian policemen. For their friendship, encouragement,

enlightening lunch time discussions, conference-buddieship, diversion, and practical

assistance I thank Barbara, Bert, Eva, Freek, Geert, Gert, Lot, Peter, Tine and many

other colleagues who have livened up visits to our coffee room or Alma lunches and

have encouraged me on an almost daily basis during the final months of thesis writ-

ing. Thanks must go also to Catharina, Kristel and Peter, who have been happy to

assist me on issues of French vocabulary, grammar and corpora. I am grateful to the

colleagues who endured physical hardship (and frequent defeat) with me on the

football ground, the many non-colleagues tacitly recruited to fill our team’s ranks,

and Arne for his organisational efforts in setting up his prestigious football competi-

tion. Finally, I should thank Isolde, for her companionship when sharing an office

with me, and her predecessor, Nele, not only an excellent colleague but also one of

the best friends I could ever have wished for, who has patiently listened to many

exasperated complaints about unyielding corpus data and disproved theories but has

Acknowledgments – iii

equally shared and allowed me to share many happy moments, inside and outside

our office.

I thank my former house-mates, David and Maria, for two pleasant years in

our tiny house in the Rapengang; and my close friends, Bram, Charlotte, Griet, Joeri,

Karolien, Peter, Stefan and Yves, for the distraction they gave me, for the memora-

ble trips to Spain, and for their healthy scepticism about my linguistic endeavours.

I thank Lieve and Johan, as well as Els and Jan, for providing me with a sec-

ond home and family in Torhout and for their interest and concern about my work.

I cannot begin to sum up the debt of gratitude I owe to my parents. As for this

thesis, however, it is they that first stimulated my curiosity, who without ever pres-

suring me encouraged me in my studies, who allowed me to come to Leuven and

study languages, and who have continued to support, clad and lavishly feed me

throughout my years as a PhD student. Though in the last few months, they have

either seen little of me, or have seen me at their computer, I am infinitely grateful to

them.

Finally, I thank Lies for her love and care, which I have had the great happi-

ness to receive, and to be allowed to return.

TABLE OF CONTENTS

Acknowledgements i

Table of contents v

Part I: Preliminaries 1

Chapter 1: Introduction 3

Chapter 2: The corpus data 13

1. Middle English 13

2. Early Modern English 14

3. Late Modern English 17

4. Present-Day English 20

Part II: Complementation and change 45

Chapter 3: Complementation 47

1. One form one meaning 48

2. A multiplicity of factors 58

3. Constructions 63

3.1. Construction grammar 64

3.2. Construction grammar and complementation 65

4. Conclusions 74

Chapter 4: Diffusional change 77

1. General perspectives on diffusion 79

1.1. Reanalysis and actualisation 80

1.2. Lexical diffusion in sound change and elsewhere 87

1.3. The two-sided effect of frequency: A case-study 90

2. Diffusional change in the system of complementation 96

2.1. Blocking 98

2.2. Analogy 102

2.2.1. Semantic analogy 103

2.2.2. Paradigmatic analogy 118

2.3. Complicating factors 127

vi – Table of contents

3. Conclusions 128

Part III: For...to-infinitives 129

Chapter 5: A brief history of for...to-infinitives 131

1. Methodology 132

2. Functions of the for...to-infinitive 134

3. Historical developments 142

3.1. The origins 143

3.2. Developments in Modern and Present-Day English 154

3.2.1. Diffusion in predicand environments 155

3.2.2. Diffusion in non-predicand environments 161

4. Conclusions 165

Chapter 6: For...to-infinitives as verb complements 167

1. Methodology 168

2. For...to-infinitives as verb complements in Present-Day English 169

3. The diffusion of for...to-infinitives as verb complements 174

3.1. For...to-infinitives as verb complements before 1850 176

3.2. For...to-infinitives as verb complements after 1850 184

4. Synchrony and diachrony reconciled 189

5. Conclusions 191

Part IV: Ing-complements 193

Chapter 7: Ing-clauses and their problems 195

1. Ing-complements and the matching-problem 197

2. Gerunds vs. participles 214

3. Conclusions 231

Chapter 8: The rise of verbal gerunds 233

1. The verbalisation of the gerund 235

2. Quantitative developments in nominal and verbal gerunds 240

3. Functional differentiation 245

3.1. Discourse-functional differences 246

3.2. Distributional differences 260

3.2.1. Bare nominal and verbal gerunds: the rivals 263

Table of contents – vii

3.2.2. Definite nominal gerunds: watching from the side? 273

3.3. Syntactic differences 285

4. Functional and other motivations in the development of the gerund 292

5. Conclusions 297

Chapter 9: The diffusion of ing-complements I: Gerunds as verb com-

plements

305

1. Methodology 308

2. Two synchronic slices: 1640-1710 and 1850-1920 309

3. Innovations and their causes 323

3.1. Stage I: narrow paradigmatic analogy 327

3.2. Stage II: semantic analogy 343

3.3. Stage III: indirect paradigmatic analogy 370

3.4. Stage IV: broad paradigmatic analogy 399

3.5. A bird’s-eye view of the diffusion of ing-complements 427

4. Conclusions 437

Chapter 10: The diffusion of ing-complements II: Integrated participle

clauses

439

1. Synchronic characterisation 441

1.1. Complements and adverbial clauses 441

1.2. Between complements and adverbial clauses 450

2. Historical developments 462

2.1. Diffusion 462

2.2. Mechanisms of change 468

2.2.1. Reinterpretation 468

2.2.2. Analogical extension 481

2.3. Diachronic overview 484

3. Conclusions 487

Part V: Conclusions 489

Chapter 11: Conclusions 492

References 499

PART I Preliminaries

“language change takes place in the present”

(Brian Joseph)

CHAPTER 1: Introduction

The topic of this study is a set of linguistic changes that occurred in the history of

the English language. To start with an illustration, one of these changes is the

change that gave rise to examples such as the following.

(1) a. The cat loves being stroked, absolutely loves it! (BNC)

b. Will the monsters even consider helping us? (CB)

In (1), gerundial ing-clauses function as complements to verbs (love, consider),

which means that they function as a ‘participant’ involved in the process denoted by

the verb. In fact, in the case of gerundial ing-clauses, this complement function is

particularly evident from the equivalence between the ing-clauses and ordinary di-

rect objects with the same transitive matrix verbs. That is, the ing-clauses fill a syn-

tactic slot that is normally filled by a noun phrase functioning as direct object to the

verb (e.g. milk in the cat loves milk).

Although the use of gerundial ing-clauses to complement transitive verbs is

well-established in Present-Day English, it represents a relatively recent phenome-

non. We could think of the innovations giving rise to the examples in (1) as isolated

lexical changes, but that would be a misrepresentation of what happened histori-

cally. The fact is that the use of gerundial ing-complements has been on the increase

since the Middle English period, with ing-complements continually growing in dis-

course frequency, and associating with an ever-increasing number of matrix predi-

cates. In other words, there is a long-standing historical trend in English for ing-

complements to become more numerous and to spread to more and more new lexical

environments. For example, around 1500, ing-complements combined with only a

handful of matrix verbs, including cease, love, forbear and some others; by the be-

ginning of the eighteenth century, they could be used with a larger set of verbal

predicates, including avoid, finish, propose, bear, etc. and at least one adjectival

predicate, be worth; by now, in Present-Day English, ing-complements are highly

productive and must be counted as one of the main complement types of the lan-

guage. Importantly, few of the innovations in ing-complementation that appeared in

the course of the last five centuries have again disappeared. This makes the histori-

cal trend stable and, by that token, unidirectional. This also means that the appear-

ance of ing-complements with specific predicates fits the larger pattern of a major

grammatical change, with serious consequences for the structure of English.

4 – CHAPTER 1: Introduction

It is the higher-order pattern of change that this study is concerned with: the

unidirectional diffusion of a new grammatical pattern over a growing number of

environments. The primary focus is on the diffusion of new complement types over

the set of possible complement-taking matrix predicates. This type of change turns

out to be common: although the diffusion of gerundial ing-complements as in (1)

above is the most dramatic change of its kind to have occurred in the recent history

of English, it is not an isolated case. At least two other complement types have been

going through a very similar development. These are illustrated in (2) and (3). The

structure illustrated in (2) is similar to that in (1) above in showing a non-finite

clause that denotes an argument of a verbal predicate, but the non-finite clause com-

plementing in question is a for...to-infinitive – that is, an infinitive marked by to and

with its own explicit subject introduced by for. The examples in (3) again look simi-

lar to those in (1), showing subject-controlled ing-clauses that function as comple-

ment to some matrix predicate (be late, have trouble), but unlike gerundial ing-

complements they do not fill a nominal slot in the matrix clause. Ing-complements

as in (3) will here be referred to as ‘integrated participle clauses’.

(2) a. You yearn for them to walk and they walk straight into trouble (CB)

b. We couldn’t afford for it to go wrong. We’ve got to have someone in

there. (BNC)

(3) a. Another man, again in an official car, was a bit late spotting the pho-

tographer (BNC)

b. Dauntless found he was having trouble remembering how to speak.

(BNC)

Just like the gerundial ing-complements in (1), for...to-infinitives and integrated par-

ticiple clauses diffused over the inventory of possible complement-taking predicates,

gradually growing more frequent and occurring in an increasing number of envi-

ronments over the course of several centuries. Together, the diffusional changes

undergone by gerundial ing-complements, for...to-infinitives and integrated partici-

ple clauses form the central empirical phenomenon to be accounted for in this study.

The phenomenon of diffusion derives its interest from the theoretical issues it

raises. One issue is how diffusional change works and why it occurs at all; another is

what diffusional change tells us about the structure of grammar. The interconnected-

ness of these two issues reflects the central claim in this study, namely that diffusion

is driven by the same mechanisms that organise grammar synchronically. The under-

lying idea in this is that there is no such thing as a diachronic grammar. Language

CHAPTER 1: Introduction – 5

users have no internalised rules of language change, only the strategies by which

they handle their linguistic resources as they produce or interpret linguistic output

(Joseph 1992). In other words, it can only be through language use that grammar

changes. This implies that an understanding of synchronic grammar is necessary to

understand grammatical change. Conversely, of course, the only realistic view of

synchronic grammar must be one that accommodates the grammatical changes his-

torically attested.

With respect to the working of synchronic grammar, the domain that is of

greatest interest here is the system of complementation. By the system of comple-

mentation I understand the set of clausal complement types language users have

available and the principles that determine how they put them to use. For Present-

Day English, the set of complement types includes, next to the gerundial ing-

complements, for...to-infinitives, and integrated participle clauses already intro-

duced, also that-clauses, to-infinitives, and bare infinitives, among other clause

types. It is characteristic of the system of complementation that there are specific

restrictions on the use of each of these complement types, and particularly on which

complement type can combine with which predicates. In (4), for instance, three dif-

ferent complement types (an ing-complement, a bare infinitive and a to-infinitive)

are used with three different verbs, each time in a causative object-control construc-

tion, yet none are interchangeable – witness the ungrammatical examples in (5).

(4) a. The examples here should set you thinking. (CB)

b. It made Euphrasia think. (CB)

c. If there are any defects likely to cause the house to fall down around

your ears, they are not the inspector’s concern. (CB)

(5) a. *The examples here should set you think / to think.

b. *It made Euphrasia thinking / to think.

c. *any defects likely to cause the house falling / fall down around your

ears

Puzzling though variation of this kind looks, it is generally assumed that the use of

the different complement types is organised in a principled way, non-arbitrary and

motivated. That is, there are reasons why a speaker uses or can use a particular com-

plement type with a given predicate (Noonan 1985).

I concur with this view, although at the same time I believe the organisation

of the system of complementation is more complex than is usually recognised and

less neat than is sometimes imagined. The use of a given complement type rarely


answers to a single principle. First, language users have the ability to infer multiple

regularities from usage. However, they are likely to seek regularities locally and in

doing so may disregard global consistency. For example, a specific combination of a

predicate and a complement type may be sanctioned as long as it corresponds to

some uses of the complement type, without necessarily being fully consistent with

the complement type’s entire range of use. By the same token, though, it is also pos-

sible (even likely) for a given use to be sanctioned by more than one regularity. Sec-

ond, independent of the regularities language users might infer from their store of

linguistic experience, language production and linguistic choices are determined by

a variety of pragmatic and functional-cognitive constraints (including constraints on

information structure, processing constraints, stylistic constraints, etc.) that may cut

across perceived regularities. In sum, the system of complementation shows organi-

sation, but the organising principles are not mapped perfectly to the set of comple-

ment types. Instead, they interact and compete in the selection of a complement

type, and the choices between complement types often involve a weighing off of

very different considerations with no predictable outcome.

Turning to diffusional change, I argue in this study that diffusional change

typically consists in a bearing out of the synchronic regularities of usage, and that to

a certain extent it derives its diffusional character precisely from the nature of those

regularities. First, diffusion can proceed along analogical chains, with speakers’ rec-

ognition of some regularity triggering changes that give rise to new regularities and

so on. Second, the fact that the use of a pattern in a given environment can corre-

spond to more than one regularity also means that environments can differ in the

degree to which they license the pattern (the more regularities a use instantiates, the

more strongly it is sanctioned). Third, regularities can be suspended or reinforced by

independent cognitive-functional factors pertaining to a specific environment. In

brief, it is the multiplicity and interconnectedness of the principles organising the

system of complementation that cause different environments to respond differen-

tially to the appearance of a new complement type.

There is a further complication, however. The development of any pattern

builds on its own history, so to understand how some clause type diffuses as a com-

plement we have to understand where that clause type comes from in the first place.

Indeed, gerundial ing-complements, for...to-infinitival complements and integrated

participle clauses all derive from clause types that, before they came to be used as

clausal complements, fulfilled functions outside the system of complementation

(which implies that their very use as complement types instantiates another diffu-

sional change). Thus, gerundial ing-complements historically derive from nominal


derivations in -ing as in (6a). Even though historical developments have contributed

to weakening the tie between gerundial ing-clauses and their nominal origins, it has

never been completely broken and the ‘nouniness’ of gerundial ing-clauses (Ross

1973) has continued to be reflected in their interpretation and distribution, playing a

central part in how they spread as a new complement type in the system of comple-

mentation. To give just one example, nominal ing-derivations as in (6a) were espe-

cially common following prepositions, which helps to explain why the ing-clauses to

which they gave rise were also used with prepositions, as in (6b), before they ap-

peared as complements to verbal predicates (Fanego 2004a).

(6) a. in a flexing of its corporate muscles, it recently outbid several rivals

(CB)

b. The damage could be minimised by removing the staghorn beetles by

hand (CB)

For…to-infinitives likewise emerged and developed outside the system of

complementation before turning into a complement type. The use in which for…to-

infinitives first gained currency (if not, strictly speaking, the environment where

they first appeared) was as extraposed subject, as illustrated in (7).

(7) a. the countless millions who evidently believe it is proper for a tiger to

be slaughtered for rheumatism or nausea cure (CB)

b. It is “tabu” for men to walk under a washing line of women’s clothes.

(CB)

Characteristic of this use is that, as (7b) shows, it often allows an alternative reading,

with the for-NP functioning as a benefactive to the preceding predicate, and the to-

infinitive alone taking the role of extraposed subject (the ambiguity is between ‘for

men to walk … is tabu’ and ‘for men it is tabu to walk …’). In subsequent develop-

ments, this ambiguous pattern has acted as an important catalysing factor in the dif-

fusion of for…to-infinitives both outside and inside the system of complementation,

in that environments where the ambiguous structure could occur accommodated

unambiguous for…to-infinitives more easily as well.

Integrated participle clauses, finally, derive from adverbial participle clauses,

as illustrated in (8).


(8) a. Wearing his air-conditioned boots laced with chickenwire, Max treats

his extended family of animals to virtuoso cello performances. (CB)

b. I mean I c I came out of that book feeling enriched despite the fact that

er I’d er been expecting to r read a trashy bestseller you know. (CB)

The process that gave rise to integrated participle clauses is linked to the appearance

of adverbial participle clauses, consisting in the syntactic reinterpretation of adver-

bial clauses as complements in ambiguous contexts. Which predicates first com-

bined with integrated participle clauses, then, has been determined by which con-

texts could trigger syntactic reinterpretation.

In addition to explaining aspects of diffusion, these findings feed back into

our view of the synchronic system of complementation. As one gets to understand

the peculiarities of the histories of different complement types, it also becomes clear

that the architecture of the system of complementation is not entirely shaped by

principles specific to that system. In part, the development of gerundial ing-

complements, for…to-infinitival complements and integrated participle clauses has

been steered by factors that have little to do with clausal complementation as such

(the properties of noun phrases, ambiguities involving benefactive participants, the

use of adverbial participles). Consequently, the (provisional) outcomes of these de-

velopments cannot be understood without reference to complement types’ specific

histories. This puts in perspective any overly neat view of complementation as a

functionally organised system whose internal coherence is the raison d’être of its

diversified set of complement types.

Still, not all questions have been answered yet. The rise of new complement

types is diffusional because different environments sanction the new pattern more

strongly or earlier than others, for various reasons, partly dependent on the nature of

the system of complementation, and partly on the specific histories of clause types.

But why do new complement types diffuse at all? Why can there be no status quo?

While I cannot give a fully satisfactory answer to this question, I believe the solution

is likely to contain the following elements: Where there is variation between two

variants that are roughly equivalent, functional pressures may systematically favour

one of the variants, which will eventually lead to the favoured variant ousting its

competitor (Haspelmath 1999). As I will argue, this is probably what happened in

the competition between gerundial ing-clauses and the nominal constructions from

which they derive.

However, where there is no variation between an old and a new variant, as in

the majority of diffusional changes considered in this study, there can be no consis-


tent selective pressure, and even where there is variation, functional considerations

may just as well pull in different directions. In such situations, the driving force be-

hind diffusion has to be sought elsewhere – and I suggest here it be sought in an

analogical snowballing effect (Ogura & Wang 1996). Note that most innovations in

diffusion appear to arise through analogy (i.e. as a result of the bearing out of syn-

chronic regularities). If we assume that analogy is sensitive to frequency, in that

analogical pressure grows as the analogical model becomes more frequent, it follows

that as diffusion proceeds analogy grows stronger and more and more environments

will yield under its pressure. In this sense, diffusion is a self-feeding process, keep-

ing itself going, once it has been triggered. As to potential triggers, I see (at least)

the following two possibilities: First, because linguistic choices are determined by

multiple and variable factors, it is inevitable that language users produce occasional

innovations, which may help breaking down resistance to change. Second, because

the construction from which a new spreading pattern derives usually occurs in a

number of different environments, the early steps in the diffusional process tend to

be very easy to take, thus giving the little boost needed for analogical change to set

in.

As any investigation, this study has its limitations. It is useful at the outset to

point out the most obvious of these and to defend the choices that have been made.

First, only a limited number of constructions have been subjected to close investiga-

tion. The three complement types on which I focus – gerundial ing-complements,

for...to-infinitives and integrated participle clauses – are, in terms of discourse fre-

quency, a comparatively small subset of the complement types that make up the

English system of complementation. Among gerundial ing-complements, moreover,

I have primarily investigated subject-controlled complements to verbal predicates

not introduced by a preposition. However, I see no reason why the changes investi-

gated would not be representative of change in the system of complementation, and

the methodological practice of using very exhaustive corpus searches (targeting

more constructions than are actually needed) may to some extent have guarded

against significant data being overlooked.

Second, this study faces the two perennial problems in historical linguistics.

The first problem is the paucity of evidence. The recorded evidence on which a lin-

guistic pattern’s history has to be construed is a rather unrepresentative sample of

the linguistic output of poorly identified language users that is, moreover, infinitely

small in comparison to what has been lost. As the problem is insurmountable, how-

ever, it is probably best to assume that it only makes the study of language change

the more challenging. To this it may be added that the historical record for English is


in fact still exceptionally rich, to the extent that for most historical periods and for

most practical purposes it is inexhaustible. The second problem involves the funda-

mental impossibility of explanation (Lass 1980, 1998). The mechanisms invoked to

explain changes in language consist in psychological processes that, taking place in

the minds of language users, are at best indirectly accessible. Thus, the historical

record only shows the changes, while the causes have to be inferred. The best solu-

tion to this problem, I believe, is to have historical explanations encompass as many

historical facts as possible. The more findings can be fitted into a hypothesis, the

stronger the hypothesis gets, even if it remains a hypothesis.

Third, one crucial aspect of language change has been entirely neglected in

this study, namely the social embeddedness of change. Innovations only become

part of the grammar of a language by virtue of their being taken up by a community

of language users and getting socially accepted. Still, while I have no wish to deny

the relevance of the social propagation of change, I refrain from taking social factors

into account here. In this respect, it is appropriate to react against a possible view of

language change that the social embeddedness of change can be taken to imply, but

to which I cannot ascribe. Since social selection is driven by social mechanisms, the

acceptance or rejection of an innovation might appear to be independent of consid-

erations regarding the internal make-up of the language. Long-term unidirectional

developments such as diffusional change, however, indicate that innovations that

catch on do fit a larger picture of change, whereby social selection must operate on a

constrained set of options, pre-determined by which innovations are likely at a given

time from a purely language-internal perspective. It is not true, then, that all innova-

tions are equally plausible at any given time and that social selection quasi arbitrar-

ily decides which innovations can ‘stick on’ and which must return to oblivion. In

other words, I believe in the sensibility of studying language change from a lan-

guage-internal perspective only.

A final remark should be made regarding the linguistic framework underlying

the present study. The framework implicit in much of what follows is construction

grammar (Langacker 1987; Goldberg 1995, 2006; Croft 2001). Thus, I adopt here

the view that grammar is organised around networks of meaningful constructions,

which language users infer from their linguistic experience. The topic studied is in

part responsible for this choice of framework. Construction grammar happens to be

particularly apt at describing certain phenomena characteristic of the synchronic

system of complementation and it is also evident that the gradualness of diffusional

changes, which typically extend over various centuries, is more easily described by a

theory of language that has room for small regularities and lexical idiosyncrasies


than by historical frameworks that work with dramatic and essentially abrupt syntac-

tic reanalyses (Lightfoot 1979) or that would have to postulate a proliferation of

separate grammars simultaneously present in language users’ minds (Kroch 1989).

This being said, there are many aspects of language that are not covered by

construction grammar (which is not to say that they must be incompatible with it).

Most relevantly here, construction grammar does not encourage us to think of the

system of complementation as a system of functional choices (in the systemic-

functional sense; cf. Halliday 1994), yet this view is probably appropriate to some

extent (e.g. with respect to the choices between finite and non-finite complements).

Further, construction grammar only captures grammatical knowledge, which is at-

tached to constructions as complex form-meaning pairings, but other factors im-

pinge on language production and processing than the symbolic resources available

(pragmatics, processing constraints, etc.). Finally, construction grammar is a very

vague model of language: while it defines the abstract processes by which speakers

construct and exploit their internalised linguistic knowledge (e.g. schematisation,

routinisation), the interaction between those processes is still poorly understood.

Moreover, relatively few linguistic phenomena have been described in construction

grammatical terms and there is as yet no such thing as a construction grammar of a

specific language. The present study presents one attempt to make a construction

grammar solve actual linguistic problems.

To conclude, let us briefly consider the organisation of the chapters to follow.

Chapter 2 describes and discusses the corpora that have provided the data for the

analyses. Chapters 3 and 4 introduce from a general perspective the topics of com-

plementation and diffusional change, giving a first outline of how the system of

complementation is organised, and of how this system interacts with diffusional

change. At the same time, the constructional framework is introduced. Chapters 5

and 6 deal with for…to-infinitives, addressing first their general development, both

inside and outside the system of complementation (Chapter 5), then in particular the

use of for…to-infinitives as complements to verbal predicates (Chapter 6). Chapters

7 through 10 deal with ing-clauses. Chapter 7 gives an overview of the use of ing-

complements and ing-clauses in general, discussing from a synchronic point of view

the problems of the distribution of ing-complements, and the relation between ge-

rundial and participial ing-clauses. Chapter 8 turns to the historical development of

the gerund and describes the rise of gerundial ing-clauses, unearthing some of the

functional motivations in this development by contrasting gerund clauses with their

nominal alternatives. Chapter 9 then addresses the history of gerundial ing-

complements, analysing the mechanisms driving diffusional change and how these


shaped the use of ing-complements. Chapter 10 considers the status of integrated

participle clauses as another type of ing-complement and describes their historical

development. Chapter 11, finally, draws together the findings and fits them in a gen-

eral picture of diffusional change in the system of complementation.

CHAPTER 2: The corpus data

Since the history of a language is only accessible through its written records, histori-

cal linguistics is corpus-based by necessity. This means that in any historical inves-

tigation, gathering data is the first hurdle to take. Students of the history of English

are particularly fortunate, however, to have at their command a great and growing

number of digitised text corpora, which have made data-gathering considerably less

painstaking a task. Indeed, the present study is based almost entirely on such elec-

tronic corpora and the purpose of the present chapter is to describe and discuss what

is available. The corpora are here discussed in the order of the historical sub-periods

they represent.

1. Middle English

The Middle English period is fairly well-represented in electronic corpora. The Hel-

sinki corpus (HC) covers the whole period with a balanced sample of texts, repre-

sentative of as many genres and registers available, divided over four sub-periods, as

represented in Table 2.1 (Kytö 1996).

Sub-period Corpus size

HC 1150-1250 113,010

HC 1250-1350 97,480

HC 1350-1420 184,230

HC 1420-1500 213,850

Table 2.1. The Middle English section of the Helsinki corpus.

Because HC is relatively small, a number of additional sources have been drawn on.

The second edition of the Penn-Helsinki parsed corpus of Middle English (PPCME)

partly overlaps with HC as it samples a larger portion of a subset of the texts con-

tained in HC (Kroch & Taylor 2000). The sampler of the Innsbruck Middle English

prose corpus (IMEPC) has served here as another addition to the data (Markus

1999). PPCME contains 1.16 million words of text, while the IMEPC sampler con-

tains 3.55 million words. Because PPCME and IMEPC too partly overlap, it is un-

clear how many words of Middle English text are available in total. With the addi-

tion of the quotation databases of the Oxford English dictionary (OED) and the

Middle English dictionary (MED) (which also largely overlap) it is clear, though,

14 – CHAPTER 2: The corpus data

that there is a substantial amount of data available in digitised form for the Middle

English period.

2. Early Modern English

The Early Modern period is covered in its entirety by the Penn-Helsinki parsed cor-

pus of Early Modern English (PPCEME) (Kroch, Santorini & Delfs 2004), which

contains the well-balanced sample of texts from the Early Modern section of HC, to

which are added two entirely parallel samples (Penn 1 and Penn 2). The corpus is

represented in Table 2.2.

HC Penn 1 Penn 2 Total

1500-1570 196,754 194,018 185,423 576,195

1570-1640 196,742 223,064 232,993 652,799

1640-1710 179,477 197,908 187,631 565,016

Table 2.2. The Penn-Helsinki parsed corpus of Early Modern English.

Two corpora supplement the last sub-periods of PPCEME. The first is the Lampeter

corpus (LC), which runs from 1640 to 1740 (Claridge s.d.). The section 1640-1710,

corresponding to the last sub-period of PPCEME, contains 0.85 million words of

text. The second corpus is the Corpus of Early Modern English texts (CEMET), a

corpus compiled by myself on the basis of electronic texts made available by archiv-

ing projects such as the Gutenberg Project or the Oxford Text Archive. CEMET

contains two sub-periods, 1570-1640 and 1640-1710. The sub-period 1570-1640

contains 1.09 million words, some 800,000 of which, however, are taken up by

Shakespeare’s First folio.1 The second sub-period is more representative (at least in

being less biased to the use of a single author), and contains 1.94 million words of

text. The contents of CEMET are detailed in Table 2.3.

The advantages and disadvantages of using online editions to build a histori-

cal corpus have been discussed in some detail in De Smet (2005). The main advan-

tages, of course, are that the corpus data are free and that a lot is available, especially

for the more recent historical periods. As to disadvantages, the texts collected by

major archiving projects are obviously not collected for their linguistic

1 As a minor inconsistency between CEMET and PPCEME, Roger Ascham’s The scholemaster is contained in the sub-period 1570-1640 in CEMET, but in the sub-period 1500-1570 in PPCEME.

DATE AUTHOR TEXT SIZE

CEMET 1570-1710: 3,037,741

CEMET 1570-1640: 1,094,349

1570 Roger Ascham The scholemaster 52,769

1596 Walter Raleigh The discovery of Guyana 34,487

1602 Richard Carew The survey of Cornwall 85,176

1605 Francis Bacon The advancement of learning 82,310

1623 William Shakespeare The first folio 815,143

1658 Nathaniel Brook The compleat cook 24,464

CEMET 1640-1710: 1,943,392

1643 Thomas Browne Religio Medici 32,885

1644 John Milton Areopagitica 17,999

1651 Thomas Hobbes Leviathan 212,871

1653 Izaak Walton The complete angler 65,486

1656 James Harrington Oceana 110,818

1658 Thomas Browne Hydriotaphia 14,926

1661 Samuel Pepys Diary 1661-1662 172,197

1668 Abraham Cowley Essays 28,965

1675 Izaak Walton Lives 75,879

1676 Anne Fanshawe Memoirs of Lady Fanshawe 65,180

1678 John Dryden All for love 28,248

1678-84 John Bunyan Pilgrim’s progress 58,070

1680 John Bunyan The life and death of Mr. Badman 72,544

1682 John Bunyan The holy war 98,723

1682-87 William Petty Essays 28,113

1683 Thomas Ellwood The history of Thomas Ellwood written by himself 70,756

1684 Aphra Behn Love-letters between a nobleman and his sister 179,902

1688 John Dryden Life of Saint Francis Xavier 166,867

1689 John Locke Essay concerning human understanding (Vol.1) 149,718

1690 John Locke Two treatises of government 56,203

1690 Thomas Browne Letter to a friend 7,701

1692 William Congreve Incognita 22,407

1692-7 John Dryden Discourses on satire and on epic poetry 61,860

1693 William Congreve The old bachelor 25,243

1694 William Congreve The double dealer 27,592

1695 William Congreve Love for love 32,764

1700 William Congreve The way of the world 28,735

1705 Daniel Defoe The consolidator 63,625

Table 2.3. The Corpus of Early Modern English texts.

CHAPTER 2: The corpus data – 17

interest, but because they have some historical and literary value. In practice this

unfortunately reinforces the bias already prevalent in historical data towards the

writings of literate men. Another problem is the editions used and what has hap-

pened to those editions in the process of digitising the text. More often than not, it is

unclear which editions of texts have been drawn on as a source for the texts put

online and to what extent the secondary ‘editor’ who digitised the text has respected

the original. All that is certain is that editions must be in the public domain, which

means that they generally predate 1920 – itself a reason for being somewhat suspi-

cious about their general quality. Nonetheless, the corpora compiled in this way

have proven extremely useful to the research reported in this study, which in the end

is the best argument for compiling and using them.

Finally, where further examples were required (especially in Chapter 9), I

have drawn on a number of additional sources. These include the quotation database

of the OED, the works of Philip Sidney and Edmund Spencer (the omission of

which in CEMET is, in retrospect, an oversight), as well as some further works of

John Dryden and the complete diary of Samuel Pepys (both of which had been ex-

cluded from CEMET to keep the corpus reasonably balanced). These sources are

referred to as SIDNEY, SPENCER, DRYDEN and PEPYS respectively. Finally, the

Proceedings of the Old Bailey corpus (POB), discussed below, cover the final dec-

ades of the Early Modern period, running from 1674 to 1710. The size of this sec-

tion of the corpus is unknown, however, yet it is evident that the earlier records of

the Old Bailey are far less extensive then the records from the eighteenth and nine-

teenth centuries.

3. Late Modern English

At the outset of this study, no sizeable corpus was available to represent the Late

Modern English period, which is probably the most neglected period of the English

language (Rydén 1984; Denison 1998). The only exception is the two relatively

small corpora compiled under the direction of David Denison, the Corpus of Late

Modern English prose (CLMEP) and the Corpus of late-eighteenth-century prose

(CLECP), together about 400,000 words of text (Denison 1994). Since most of the

developments studied here take place during the Late Modern period, the need for a

corpus was therefore acute. To fill this need, I have compiled first the Corpus of

Late Modern English texts (CLMET), then the Corpus of Late Modern English texts

(extended version) (CLMETEV), and finally the Corpus of English novels (CEN).

The data collecting procedure has been the same as for CEMET (see Section 2


above), namely to mine the resources that are freely available online. This means

that the corpora are likely to suffer from the same shortcomings as CEMET. At the

same time, for the more recent periods more material is available, which makes it

easier to compensate at least to some extent for the bias towards the writings of lit-

erate men. Also note that the reliability of the editions used becomes less of a prob-

lem as the texts at issue become more recent, younger texts causing fewer editorial

problems (in the rule).

As CLMET, described in full detail in De Smet (2005), is entirely contained

within CLMETEV, it needs no detailed discussion here. Only note that the research

reported in Chapters 5 and 6 (on the for…to-infinitive) was largely conducted on the

basis of CLMET rather than CLMETEV; and further that at the time of data-

gathering, the original first sub-period, running from 1710-1780, had already been

extended somewhat in comparison to the version of the corpus described in De Smet

(2005). This preliminary extension had been achieved by lifting a restriction on the

amount of text that could be sampled per author (200,000 words in the original ver-

sion of the corpus). Thus, the corpus used in Chapters 5 and 6 looks as in Table 2.4.

Sub-period Corpus size

1710-1780 3,390,000

1780-1850 3,830,000

1850-1920 3,980,000

Table 2.4. The Corpus of Late Modern English texts (as used in Chapters 5 and 6).

The more systematic extension of CLMET that gave rise to CLMETEV

largely maintained the principles underlying the compilation of CLMET. No more

than 200,000 words per author were selected for inclusion in the corpus (for which

reason many texts are not taken up in their totality), and authors whose works are

taken up in a sub-period had to be born within a corresponding sub-period, as repre-

sented in Figure 2.1. Archiving projects drawn on included, next to the Project

Gutenberg and the Oxford Text Archive also the Victorian Women Writer’s Project,

which substantially increased the share of women writers in the corpus data (if not

the share of non-literary texts). CLMETEV deviates from the original CLMET in

only one more way, namely the inclusion of Irish and Scottish authors, next to Eng-

lish authors. Incidentally, it is evident that the data available online were still far

from exhausted even at the time of compilation, especially as regards the second and

CHAPTER 2: The corpus data – 19

Figure 2.1. Sub-periodisation in CLMET and CLMETEV.

third sub-periods. This means that further extensions of the corpus must be possible

in future. The structure and contents of CEN are outlined in Table 2.5; as can be

seen, CLMETEV contains about 15 million words of text.

CEN, finally, is a large corpus made up entirely of novels. The novels in-

cluded were all written between 1881 and 1922 by 25 different authors, all born be-

tween 1848 and 1862, thus approximately representing the writings of a single gen-

eration of authors in a single text genre. The authors whose work has been taken up

are not exclusively British, but also include North American novelists. The main

merit of CEN is its size – some 26 million words – which has made it especially

useful for the analyses of infrequent constructions in Chapters 9 and 10. Its contents

are detailed in Table 2.6.

Apart from CLMET, CLMETEV and CEN, some additional resources have

been made use of. Most prominent among these is the Proceedings of the Old Bailey

Corpus (POB) (Huber forthc.). POB collects all the trial proceedings published by

the Old Bailey in London (from 1674 to 1834), which was later reformed to the Cen-

tral Criminal Court (from 1834 to 1912). In size, the corpus is immense, with about

57 million words from between 1674 and 1834, and 87 million words from between

1834 and 1912. What is more, text genre and register are also quite unique, since the

major part of the texts consists of transcriptions of orally delivered testimonies,

cross-examinations, etc. Of course, transcriptions were not meant to reflect the spo-

ken language as accurately as possible but they are probably as close as we can get

to spoken historical data, with speakers coming from all social classes, moreover,

and including both men and women, young and old, and so on. On the downside,

there is a certain monotony in the contents of the texts, most of which are narrative

in character and invariably report about crimes committed or uncommitted.

Finally, to make up for lack of data representative of the beginning of the

Late Modern period, I have drawn on the OED, the remainder of LC (covering the

1680

1710 1780 1850 1920

1750 1820 1890

Part 1 Part 2 Part 3

Year of pu b-lication

Author’s year of birth


period 1710-1740, with about 320,000 words of text) and occasionally on the writ-

ings of Daniel Defoe (DEFOE), Fanny Burney (BURNEY) as well as the material

from Addison and Steele’s complete Spectator (SPECTATOR) and Steele’s Tatler

(TATLER).

4. Present-Day English

For Present-Day English, finally, a large variety of corpora, both written and spoken,

are available. The corpora made use of in this study are primarily the Collins

Cobuild corpus (CB) and the British national corpus (BNC). These are both very

large – CB contains about 57 million words, BNC 100 million words – and represent

a variety of genres and registers, including spoken language. BNC is exclusively

British English, while CB also contains some American and Australian English.

Another major corpus has recently been added to this list: TIME contains all vol-

umes of Time Magazine published from 1926 until 2007. With more than 100 mil-

lion words of text, it is large enough to allow tracking of short-term lexico-

grammatical changes in the language. Smaller corpora that have also proven useful

are the four parallel corpora LOB, FLOB, BROWN and FROWN, and the ICE-GB.

Each of these corpora is exactly one million words in size. LOB, FLOB, BROWN

and FROWN, constructed in parallel, separately represent British and American

English for 1960/61 and the 1990s, which makes them a useful basis for quick com-

parisons between the two major varieties of English. ICE-GB, on the other hand,

allows searches on syntactic trees. COLT has been useful as a small corpus repre-

sentative of the very informal spoken language of London teenagers. Finally, I have

occasionally taken up examples encountered in reading, or found on the world wide

web.


CLMETEV 1710-1920: 14,970,622

CLMETEV 1710-1780: 3,037,607

1726 Butler, Joseph (1692-1752) Human nature and other sermons 42,537

1728 Gay, John (1685-1732) The beggar’s opera 17,427

1732 Bradley, Richard (1688-1732) The country housewife and lady’s director 90,007

1733-34 Pope, Alexander (1688-1744) An essay on man 46,995

1735-48 Walpole, Horace (1717-97) Letters (Vol. 1) (s) 162,799

1739-40 Hume, David (1711-76) A treatise of human nature (s) 113,935

1740 Richardson, Samuel (1689-1761) Pamela (s) 200,225

1740-41 Johnson, Samuel (1709-84) Parliamentary debates (Vol. 1) (s) 163,695

1744 Haywood, Eliza Fowler (1693-1756) The fortunate foundlings 102,644

1746-71 Chesterfield, Philip Dormer Stanhope (1694-1773) Letters to his son (s) 199,819

1749 Fielding, Henry (1707-54) The history of Tom Jones, a foundling (s) 100,242

1749 Fielding, Sarah (1710-68) The governess; or, the little female academy 50,708

c1750 Doddridge, Philip (1701-1752) The life of Col. James Gardiner 48,995

1751 Fielding, Henry (1707-54) Amelia (s) 99,569

1751 Hume, David (1711-76) An enquiry concerning the principles of morals 48,245

1751 Smollett, Tobias George (1721-71) The adventures of Peregrine Pickle (s) 99,421

1753 Cibber, Theophilus (1703-1758) The lives of the poets of Great Britain and Ireland

(Vol. 1; Vol. 3)

200,636

1759 Johnson, Samuel (1709-84) Rasselas, prince of Abyssinia 37,070

1759-67 Sterne, Laurence (1713-68) The life and opinions of Tristram Shandy (s) 158,135

1764 Walpole, Horace (1717-97) The castle of Otranto 36,171

1766 Smith, Adam (1723-90) An inquiry into the nature and causes of the wealth of

nations (s)

200,667

1766 Goldsmith, Oliver (1728-74) The vicar of Wakefield 63,730

1768 Sterne, Laurence (1713-68) A sentimental journey through France and Italy 42,249

1768-71 Cook, James (1728-1779) Captain Cook’s journal during the first voyage round

the world (s)

201,095

1769-76 Reynolds, Joshua (1723-92) Seven discourses on art 39,563

1770 Burke, Edmund (1729-97) Thoughts on the present discontents 30,386

1771 Smollett, Tobias George (1721-71) The expedition of Humphrey Clinker (s) 100,675

1773 Goldsmith, Oliver (1728-74) She stoops to conquer 22,962

1775 Burke, Edmund (1729-97) On conciliation with America 26,883

1776 Gibbon, Edward (1737-94) The decline and fall of the Roman Empire (Vol. 1) (s) 199,087

1777 Reeve, Clara (1729-1807) The old English baron 55,063

1779 Hume, David (1711-76) Dialogues concerning natural religion 35,972

CLMETEV 1780-1850: 5,723,988

1780-96 Burns, Robert (1759-96) The letters of Robert Burns 124,247

1782 Burney, Fanny (1752-1840) Cecilia (Vol. 1-2) (s) 198,671

1783 Kilner, Dorothy (1755-1838) The life and perambulations of a mouse 30,153

1783 Beckford, William (1760-1844) Dreams, waking thoughts, and incidents 80,746

1783-84 Godwin, William (1756-1836) Four early pamphlets 45,459

1792 Wollstonecraft, Mary (1759-97) Vindication on the rights of woman 86,670

1792 Carey, William (1761-1834) An enquiry into the obligations of Christians 14,901

1794 Godwin, William (1756-1836) The adventures of Caleb Williams (s) 36,311

1796 Inchbald, Elisabeth (1753-1821) Nature and art 47,126

1796 Wollstonecraft, Mary (1759-97) Letters on Norway, Sweden, and Denmark 48,219

1796-1817 Austen, Jane (1775-1817) Letters to her sister Cassandra and others (s) 77,989

1796-1801 Edgeworth, Maria (1767-1849) The parent’s assistant 168,182

1798 Inchbald, Elisabeth (1753-1821) Lover’s vows 17,701

1798 Wollstonecraft, Mary (1759-97) Maria 45,428

1798 Malthus, Thomas (1766-1834) An essay on the principle of population 54,451

1807 Lamb, Charles (1775-1834) Tales from Shakespeare 100,349

1808 Lamb, Charles (1775-1834) Adventures of Ulysses 33,727

1810-13 Byron, George Gordon (1788-1824) Letters 1810-1813 110,243

1811 Austen, Jane (1775-1817) Sense and sensibility (s) 61,546

1812 Smith, James (1775-1839), and Horace Smith (1779-

1849)

Rejected addresses 28,759

1813 Owen, Robert (1771-1858) A new view of society 34,124

1813 Southey, Robert (1774-1843) Life of Horatio Lord Nelson 96,781

1813 Austen, Jane (1775-1817) Pride and Prejudice (s) 60,141

1817 Coleridge, Samuel Taylor (1772-1834) Biographia Literaria 138,354

1818 Shelly, Mary Wollstonecraft (1797-1851) Frankenstein 75,082

1820-22 Hunt, Henry (1773-1835) Memoirs of Henry Hunt (Vol. 1) 130,079

1821 Foster, John (1770-1843) An essay on the evils of popular ignorance 92,695

1821 Galt, John (1779-1839) The Ayrshire Legatees 50,072

1821 Galt, John (1779-1839) Annals of the Parish 65,613

1821-22 Hazlitt, William (1778-1830) Table talk 160,700

1822 De Quincey, Thomas (1785-1859) Confessions of an English opium-eater 38,839

1823 Hazlitt, William (1778-1830) Liber Amoris 30,911

1823 Galt, John (1779-1839) The provost 55,353

1824 Hogg, James (1770-1835) The private memoirs and confessions of a justified sin-

ner

84,166

1824 Gore, Catherine Grace Frances (1799-1861) Theresa Marchmont, or the maid of honour 15,965

1826 Disraeli, Benjamin (1804-81) Vivian Grey (s) 100,147

1829 Southey, Robert (1774-1843) Sir Thomas More 39,124

1830 Babbage, Charles (1792-1871) Reflections on the decline of science in England 50,169

1831 Godwin, William (1756-1836) Thoughts on man 116,208

1832 Babbage, Charles (1792-1871) The economy of machinery and manufactures 105,468

1834 Bulwer-Lytton, Edward (1803-73) The last days of Pompeii 151,692

1837 Carlyle, Thomas (1795-1881) The French revolution (s) 200,251

1837 Disraeli, Benjamin (1804-81) Venetia (s) 99,263

1838 Gillman, James (1782-1839) Life of Samuel Taylor Coleridge 85,331

1839 Clarkson, Thomas (1760-1846) The history of the abolition of the African slave-trade

(s)

200,018

1839 Norton, Caroline (1808-77) A plain letter to the Lord Chancellor on the Infant Cus- 33,451

tody Bill

1839 Darwin, Charles (1809-82) The voyage of the Beagle (s) 199,777

1839 Ellis, Sarah Stickney (1812-72) The women of England, their social duties, and domes-

tic habits

75,614

1841 Marryat, Frederick (1792-1848) Masterman Ready 99,705

1841 Dickens, Charles (1812-70) Barnaby Rudge (s) 78,226

1842 Borrow, George Henry (1803-81) The Bible in Spain (s) 199,199

1843 Ainsworth, William Harrison (1805-82) Windsor Castle 117,072

1843 Dickens, Charles (1812-70) A Christmas carol in prose 28,673

1844 Kinglake, William (1809-91) Eothen, or Traces of travel brought home from the East 89,058

1846 Cary, Henry Francis (1772-1844) Lives of the English poets 97,740

1846 Hunt, Leigh (1784-1859) Stories from the Italian poets (Vol. 1) 95,971

1847 Cottle, Joseph (1770-1853) Reminiscences of Samuel Taylor Coleridge and Robert

Southey

149,309

1847 Brontë, Emily (1818-48) Wuthering Heights 116,760

1847 Brontë, Anne (1820-49) Agnes Grey (s) 50,133

1847-48 Thackeray, William Makepeace (1811-63) Vanity Fair (s) 200,907

1848 Gaskell, Elizabeth (1810-65) Mary Barton 160,888

1848 Dickens, Charles (1812-70) Dombey and son (s) 93,352

1848 Brontë, Anne (1820-49) The tenant of Wildfell Hall (s) 150,730

CLMETEV 1850-1920: 6,251,564

1850 Craik, Dinah Maria Mulock (1826-1887) Olive (Vol. 1-3) 152,605

1854 Baker, Samuel White (1821-93) The rifle and the hound in Ceylon 90,467

1855 Baker, Samuel White (1821-93) Eight years’ wanderings in Ceylon 89,221

1856 Bird, Isabella Lucy (1831-1904) The Englishwoman in America 127,423

1857 Hughes, Thomas (1822-96) Tom Brown’s schooldays 105,982

1859-60 Collins, William Wilkie (1824-89) The woman in white (s) 96,398

1865 Yonge, Charlotte Mary (1823-1901) The clever woman of the family (s) 74,807

1865 Carroll, Lewis (1832-98) Alice’s adventures in Wonderland 26,699

1867 Bagehot, Walter (1826-77) The English constitution 97,933

1868 Collins, William Wilkie (1824-89) The moonstone (s) 101,932

1869 Blackmore, Richard Doddridge (1825-1900) Lorna Doone, a romance of Exmoor (s) 202,593

1869 Bagehot, Walter (1826-77) Physics and politics 56,554

1870 Yonge, Charlotte Mary (1823-1901) The caged lion (s) 77,241

1870 Meredith, George (1828-1909) The adventures of Harry Richmond (s) 97,677

1871 Carroll, Lewis (1832-98) Through the looking glass 29,639

1873 Yonge, Charlotte Mary (1823-1901) Young folk’s History of England (s) 51,339

1873 Bagehot, Walter (1826-77) Lombard Street 48,440

1873 Hardy, Thomas (1840-1928) A pair of blue eyes (s) 101,665

1874 Hardy, Thomas (1840-1928) Far from the madding crowd (s) 100,100

1879 Booth, Catherine Mumford (1829-90) Papers on practical religion 53,616

1879 Buckley, Arabella Burton (1840-1929) The fairy-land of science 61,653

1880 Booth, Catherine Mumford (1829-90) Papers on aggressive Christianity 54,668

1880 Butler, Samuel (1835-1902) Unconcious memory (s) 51,231

1883 Blind, Mathilde (1841-96) George Eliot 60,685

1884 Webster, Augusta (1837-1894) Daffodil and the Croäxaxicans 107,442

1884 Abbott, Edwin (1838-1926) Flatland 33,805

1885 Linton, E. Lynn (1822-1898) The autobiography of Christopher Kirkland (Vol. 1-3) 147,209

1885 Pater, Walter Horatio (1839-94) Marius the Epicurean (Vol. 1) 56,847

1885 Blind, Mathilde (1841-96) Tarantella (Vol. 1) 62,835

1886-90 Pater, Walter Horatio (1839-94) Essays from ‘The Guardian’ 24,020

1887 Haggard, Henry Rider (1856-1925) She 111,944

1888 Freeman, Edward Augustus (1823-92) William the Conqueror 57,067

1888 Wells, Herbert George (1866-1946) The time machine 32,507

1889 Cassels, Walter (1826-1907) A reply to Dr. Lightfoot’s essays 59,949

1889 Carroll, Lewis (1832-98) Sylvie and Bruno 65,579

1889 Jerome, Jerome K. (1859-1927) Three men in a boat 67,445

1890 Booth, William (1829-1912) In darkest England and the way out 126,065

1891 Gissing, George (1857-1903) New Grub Street (s) 94,810

1893 Rutherford, Mark (1831-1913) Catherine Furze 67,367

1893 Gissing, George (1857-1903) The odd woman (s) 101,691

1894 Huxley, Thomas Henry (1825-95) Discourses 95,883

1894 Grossmith, George (1847-1912), and Weedon Gross-

mith (1852-1919)

The diary of a nobody 42,276

1894 Ward, Humphry (1851-1920) Marcella (Vol. 1) 119,519

1894 Hope, Anthony (1863-1933) The prisoner of Zenda 54,157

1894 Kipling, Rudyard (1865-1936) The jungle book 51,162

1895 Meredith, George (1828-1909) The amazing marriage (s) 98,235

1896 Rutherford, Mark (1831-1913) Clara Hopgood 48,987

1896 Pater, Walter Horatio (1839-94) Gaston de Latour, an unfinished romance 38,212

1897 Caine, Hall (1853-1931) The Christian 200,268

1897 Kipling, Rudyard (1865-1936) Captains courageous 53,452

1897 Wells, Herbert George (1866-1946) The war of the worlds 60,308

1898 Hope, Anthony (1863-1933) Rupert of Hentzau 83,351

1899 Churchill, Winston (1874-1965) The river war, an account of the reconquest of the Su-

dan

126,807

1899-1902 Bridge, Cyprian (1839-1924) Sea-power and other studies 79,267

1901 Malet, Lucas (1852-1931) The history of Sir Richard Calmady (s) 99,115

1902 Bacon, John Mackenzie (1846-1904) The dominion of the air 89,946

1902 Nesbit, Edith (1858-1924) The children and it 54,379

1902 Bennett, Arnold (1867-1931) The grand Babylon Hotel (s) 51,852

1902-03 Wells, Herbert George (1866-1946) Mankind in the making 103,549

1903 Butler, Samuel (1835-1902) The way of all flesh (s) 74,069

1904 Radford, Dollie (1858-1920) Sea-thrift: A fairy-tale 29,135

1904 Galsworthy, John (1867-1933) The island Pharisees 70,492

1905 Forster, Edward Morgan (1879-1970) Where angels fear to tread 49,988

1906 Galsworthy, John (1867-1933) The man of property 110,623

1907 Gosse, William Edmund (1849-1928) Father and son, a study of two temperaments 79,185

1908 Bennett, Arnold (1867-1931) The old wives’ tale (s) 149,599

1908 Forster, Edward Morgan (1879-1970) A room with a view (s) 49,518

1909 Jerome, Jerome K. (1859-1927) They and I 70,125

1910 Brebner, Percy James (1864-1922) The brown mask 94,713

1910 Blackwood, Algernon (1869-1951) The human chord 58,957

1910 Chambers, Dorothea Lambert (1878-1960) Lawn tennis for ladies 22,335

1910 Forster, Edward Morgan (1879-1970) Howards end (s) 100,510

1912 Butler, Samuel (1835-1902) Note-books (s) 76,734

1912 Chesterton, Gilbert Keith (1874-1936) What’s wrong with the world 60,318

1912 Beesley, Lawrence (1877-1967) The loss of the SS Titanic 49,917

1914 Cheyne, Thomas Kelly (1841-1915) The reconciliation of races and religions 49,597

1914 Chesterton, Gilbert Keith (1874-1936) The wisdom of father Brown 71,935

1915 Carpenter, Edward (1844-1929) The healing of nations and the hidden sources of their

strife

47,476

1915 Blackwood, Algernon (1869-1951) The extra day 95,143

1920 Bagnold, Enid (1889-1981) The happy foreigner 63,560

Table 2.5. The Corpus of Late Modern English texts (extended version) ((s) = sample).


CEN 1884-1922: 26,237,428

1881 Burnett, Frances (1849-1924) A Fair Barbarian 41,718

1881 Ward, Humphrey [Mary Augusta Ward] (1851-1920) Milly and Olly 47,903

1881 Stevenson, Robert Louis (1850-1894) Treasure Island 68,319

1882 Crawford, Francis Marion (1854-1909) Mr. Isaacs 82,513

1882 Stevenson, Robert Louis (1850-1894) The New Arabian Nights 102,366

1883 Crawford, Francis Marion (1854-1909) Doctor Claudius 101,791

1884 Crawford, Francis Marion (1854-1909) A Roman Singer 96,604

1884 Ward, Humphrey [Mary Augusta Ward] (1851-1920) Miss Bretherton 57,324

1884 Allen, Grant (1848-1899) Philistia 143,694

1884 Stevenson, Robert Louis (1850-1894) The Black Arrow 79,850

1884 Gissing, George (1857-1903) The Unclassed 124,922

1885 Moore, George Augustus (1852-1933) A Mummer’s Wife 142,687

1885 Haggard, Henry Rider (1856-1925) King Solomon’s Mines 82,423

1885 Burnett, Frances (1849-1924) Little Lord Fauntleroy 58,669

1885 Stevenson, Robert Louis (1850-1894) Prince Otto 62,680

1885 Crawford, Francis Marion (1854-1909) The Children of the King 61,120

1885 Caine, Hall (1853-1931) The Shadow of a Crime 123,977

1885 Crawford, Francis Marion (1854-1909) Zoroaster 77,310

1886 Crawford, Francis Marion (1854-1909) A Tale of a Lonely Parish 107,246

1886 Gissing, George (1857-1903) Demos 195,675

1886 Jerome, Jerome Kapla (1859-1927) Idle Thoughts of an Idle Fellow 42,106

1886 Stevenson, Robert Louis (1850-1894) Kidnapped 82,164

1886 Moore, George Augustus (1852-1933) Muslin 94,086

1886 Corelli, Marie (1855-1924) The Romance of Two Worlds 99,558

1886 Stevenson, Robert Louis (1850-1894) The Strange Case of Dr. Jekyll and Mr. Hyde 25,763

1886 Corelli, Marie (1855-1924) Vendetta 144,414

1887 Moore, George Augustus (1852-1933) A Mere Accident 47,695

1887 Crawford, Francis Marion (1854-1909) Marzio’s Crucifix 57,522

1887 Burnett, Frances (1849-1924) Sara Crowe 66,881

1887 Crawford, Francis Marion (1854-1909) Saracinesca 154,520

1887 Haggard, Henry Rider (1856-1925) She 113,137

1887 Stevenson, Robert Louis (1850-1894) The Merry Men 80,332

1887 Corelli, Marie (1855-1924) Thelma 205,564

1887 Gissing, George (1857-1903) Thyrza 195,303

1888 Gissing, George (1857-1903) A Life’s Morning 141,730

1888 Doyle, Arthur Conan (1859-1930) A Study in Scarlet 43,542

1888 Moore, George Augustus (1852-1933) Confessions of a Young Man 59,560

1888 Haggard, Henry Rider (1856-1925) Mr Meeson’s Will 67,617

1888 Ward, Humphrey [Mary Augusta Ward] (1851-1920) Robert Elsmere 287,990

1888 Moore, George Augustus (1852-1933) Spring Days 99,088

1888 Atherton, Gertrude (1857-1935) What Dreams May Come 44,409

1889 Corelli, Marie (1855-1924) Ardath 216,533

1889 Haggard, Henry Rider (1856-1925) Cleopatra 102,931

1889 Crawford, Francis Marion (1854-1909) Greifenstein 153,533

1889 Doyle, Arthur Conan (1859-1930) Micah Clarke 178,167

1889 Moore, George Augustus (1852-1933) Mike Fletcher 91,214

1889 Stevenson, Robert Louis (1850-1894) The Master of Ballantrae 87,499

1889 Doyle, Arthur Conan (1859-1930) The Mystery of Cloomber 48,880

1889 Gissing, George (1857-1903) The Nether World 160,882

1889 Jerome, Jerome Kapla (1859-1927) Three Men in a Boat 67,445

1890 Haggard, Henry Rider (1856-1925) Beatrice 112,327

1890 Doyle, Arthur Conan (1859-1930) The Captain of the Polestar 84,491

1890 Gissing, George (1857-1903) The Emancipated 143,716

1890 Doyle, Arthur Conan (1859-1930) The Firm of Girdlestone 137,297

1890 Weyman, Stanley John (1855-1928) The House of the Wolf 56,735

1890 Caine, Hall (1853-1931) The Scapegoat 101,245

1891 Crawford, Francis Marion (1854-1909) Don Orsino 151,630

1891 Haggard, Henry Rider (1856-1925) Eric Brighteyes 108,189

1891 Gissing, George (1857-1903) New Grub Street 186,718

1891 Jerome, Jerome Kapla (1859-1927) The Diary of a Pilgrimage 41,869

1891 Allen, Grant (1848-1899) The Great Taboo 75,648

1891 Doyle, Arthur Conan (1859-1930) The White Company 150,866

1891 Crawford, Francis Marion (1854-1909) The Witch of Prague 135,647

1891 Allen, Grant (1848-1899) What’s Bred in the Bone 98,820

1892 Doyle, Arthur Conan (1859-1930) Beyond the City 39,536

1892 Gissing, George (1857-1903) Born in Exile 158,445

1892 Gissing, George (1857-1903) Denzil Quarrier 80,265

1892 Merriman, Henry Seton (1862-1903) From one Generation to Another 63,22

1892 Haggard, Henry Rider (1856-1925) Nada the Lily 117,77

1892 Doyle, Arthur Conan (1859-1930) The Adventures of Sherlock Holmes 104,05

1892 Doyle, Arthur Conan (1859-1930) The Doing of Raffles Haw 38,35

1892 Bacheller, Irving (1859-1950) The Master of Silence 29,81

1892 Merriman, Henry Seton (1862-1903) The Slave of the Lamp 74,04

1893 Weyman, Stanley John (1855-1928) A Gentleman of France 147,23

1893 Stevenson, Robert Louis (1850-1894) An Island Night’s Entertainment 49,07

1893 Barr, Robert (1850-1912) From Whose Bourne 29,85

1893 Allen, Grant (1848-1899) Michael’s Crag 34,69

1893 Haggard, Henry Rider (1856-1925) Montezuma’s Daughter 151,75

1893 Parker, Gilbert (1862-1932) Mrs. Falchion 91,87

1893 Jerome, Jerome Kapla (1859-1927) Novel Notes 69,45

1893 Wiggin, Kate Douglas (1856-1923) Penelope’s English Experiences 31,73

1893 Wiggin, Kate Douglas (1856-1923) Polly Oliver’s Problem 40,38

1893 Doyle, Arthur Conan (1859-1930) The Memoirs of Sherlock Holmes 87,48

1893 Gissing, George (1857-1903) The Odd Woman 139,84

1893 Doyle, Arthur Conan (1859-1930) The Refugees 123,33

1893 Parker, Gilbert (1862-1932) The Trail of the Sword 55,18

1893 Parker, Gilbert (1862-1932) The Trespasser 61,01

1894 Moore, George Augustus (1852-1933) Esther Waters 111,41

1894 Barr, Robert (1850-1912) In the Midst of Alarms 77,78

1894 Gissing, George (1857-1903) In the Year of Jubilee 131,47

1894 Ward, Humphrey [Mary Augusta Ward] (1851-1920) Marcella 249,03

1894 Doyle, Arthur Conan (1859-1930) Round the Red Lamp 68,57

1894 Barr, Robert (1850-1912) The Face and the Mask 71,57

1894 Haggard, Henry Rider (1856-1925) The People of the Mist 146,15

1894 Weyman, Stanley John (1855-1928) Under the Red Robe 64,97

1894 Merriman, Henry Seton (1862-1903) With Edged Tools 105,24

1895 Gissing, George (1857-1903) Eve’s Ransom 54,86

1895 Weyman, Stanley John (1855-1928) From the Memoirs of a Minister of France 73,03

1895 Allen, Grant (1848-1899) The British Barbarians 36,61

1895 Merriman, Henry Seton (1862-1903) The Grey Lady 68,55

1895 Gissing, George (1857-1903) The Paying Guest 27,70

1895 Merriman, Henry Seton (1862-1903) The Sowers 106,74

1895 Doyle, Arthur Conan (1859-1930) The Stark Munro Letters 77,37

1895 Allen, Grant (1848-1899) The Woman who Did 45,56

1895 Moore, George Augustus (1852-1933) Vain Fortune 57,34

1895 Parker, Gilbert (1862-1932) When Valmond Came to Pontiac 50,80

1896 Barr, Robert (1850-1912) One Day’s Courtship 36,73

1896 Barr, Robert (1850-1912) Revenge! 81,58

1896 Doyle, Arthur Conan (1859-1930) Rodney Stone 91,94

1896 Ward, Humphrey [Mary Augusta Ward] (1851-1920) Sir George Tressady (Vol. 1-2) 171,63

1896 Parker, Gilbert (1862-1932) The Pomp of the Lavilettes 38,89

1896 Parker, Gilbert (1862-1932) The Seats of the Mighty 127,07

1896 Wiggin, Kate Douglas (1856-1923) The Village Watchtower 41,37

1896 Haggard, Henry Rider (1856-1925) The Wizard 60,06

1896 Corelli, Marie (1855-1924) Ziska 59,54

1897 Allen, Grant (1848-1899) An African Millionaire 65,48

1897 Merriman, Henry Seton (1862-1903) In Kedar’s Tents 71,73

1897 Wiggin, Kate Douglas (1856-1923) Marm Lisa 38,19

1897 Caine, Hall (1853-1931) The Christian 200,68

1897 Parker, Gilbert (1862-1932) The Translation of a Savage 50,47

1897 Gissing, George (1857-1903) The Whirlpool 164,46

1897 Doyle, Arthur Conan (1859-1930) Uncle Burnac 57,75

1898 Connor, Ralph (1860-1937) Black Rock 56,24

1898 Haggard, Henry Rider (1856-1925) Doctor Therne 47,98

1898 Moore, George Augustus (1852-1933) Evelyn Innes 178,37

1898 Ward, Humphrey [Mary Augusta Ward] (1851-1920) Helbeck of Bannisdale (Vol. 1-2) 137,55

1898 Wiggin, Kate Douglas (1856-1923) Penelope’s Experience in Scotland 58,86

1898 Haggard, Henry Rider (1856-1925) Swallow 107,90

1898 Parker, Gilbert (1862-1932) The Battle of the Strong 125,33

1898 Weyman, Stanley John (1855-1928) The Castle Inn 110,87

1898 Jerome, Jerome Kapla (1859-1927) The Second Thoughts of an Idle Fellow 64,46

1898 Gissing, George (1857-1903) The Town Traveller 64,53

1898 Atherton, Gertrude (1857-1935) The Valiant Runaways 46,23

1899 Barr, Robert (1850-1912) A Woman Intervenes 97,96

1899 Allen, Grant (1848-1899) Hilda Wade 89,09

1899 Barr, Robert (1850-1912) Jennie Baxter, Journalist 68,36

1899 Gissing, George (1857-1903) The Crown of Life 120,42

1899 Connor, Ralph (1860-1937) The Sky Pilot 44,35

1899 Nesbit, Edith (1858-1924) The Story of the Treasure Seekers 53,64

1899 Barr, Robert (1850-1912) The Strong Arm 96,85

1900 Parker, Gilbert (1862-1932) An Unpardonable Liar 22,32

1900 Haggard, Henry Rider (1856-1925) Black Heart and White Heart 21,44

1900 Bacheller, Irving (1859-1950) Eben Holden 89,45

1900 Ward, Humphrey [Mary Augusta Ward] (1851-1920) Eleanor 148,09

1900 Atherton, Gertrude (1857-1935) Senator North 102,89

1900 Atherton, Gertrude (1857-1935) The Doomswoman 50,21

1900 Hough, Emerson (1857-1923) The Girl at the Halfway House 84,02

1900 Merriman, Henry Seton (1862-1903) The Isle of Unrest 74,82

1900 Corelli, Marie (1855-1924) The Master Christian 226,36

1900 Wharton, Edith (1862-1937) The Touchstone 27,57

1900 Baum, Lyman Frank (1856-1919) The Wonderful Wizard of Oz 39,67

1900 Jerome, Jerome Kapla (1859-1927) Three Men on the Bummel 67,10

1901 Baum, Lyman Frank (1856-1919) American Fairy Tales 31,53

1901 Gissing, George (1857-1903) By the Ionian Sea 40,75

1901 Weyman, Stanley John (1855-1928) Count Hannibal 110,73

1901 Bacheller, Irving (1859-1950) D’ri and I 66,65

1901 Burnett, Frances (1849-1924) Emily Fox-Setton 78,63

1901 Haggard, Henry Rider (1856-1925) Lysbeth 157,97

1901 Crawford, Francis Marion (1854-1909) Marietta 117,55

1901 Gissing, George (1857-1903) Our Friend the Charlatan 132,61

1901 Wiggin, Kate Douglas (1856-1923) Penelope’s Irish Experiences 71,34

1901 Moore, George Augustus (1852-1933) Sister Teresa 114,40

1901 Connor, Ralph (1860-1937) The Man from Glengarry 117,85

1901 Parker, Gilbert (1862-1932) The Right of Way 119,75

1901 Merriman, Henry Seton (1862-1903) The Velvet Glove 73,39

1901 Nesbit, Edith (1858-1924) The Wouldbegoods 78,18

1902 Nesbit, Edith (1858-1924) Five Children and It 53,34

1902 Connor, Ralph (1860-1937) Glengarry Schooldays 59,58

1902 Jerome, Jerome Kapla (1859-1927) Paul Kelver 136,54

1902 Corelli, Marie (1855-1924) Temporal Power 202,67

1902 Wiggin, Kate Douglas (1856-1923) The Diary of a Goose Girl 18,28

1902 Baum, Lyman Frank (1856-1919) The Surprising Adventures of the Magical Monarch of

Mo and his People

32,56

1902 Hough, Emerson (1857-1923) The Mississippi Bubble 93,03

1902 Atherton, Gertrude (1857-1935) The Splendid Idle Forties 86,59

1902 Wharton, Edith (1862-1937) The Valley of Decision 153,78

1902 Merriman, Henry Seton (1862-1903) The Vultures 92,58

1903 Merriman, Henry Seton (1862-1903) Barlasch of the Guard 75,38

1903 Bacheller, Irving (1859-1950) Darrel of the Blessed Isles 70,07

1903 Ward, Humphrey [Mary Augusta Ward] (1851-1920) Lady Rose’s Daughter 130,40

1903 Haggard, Henry Rider (1856-1925) Pearl-Maiden 133,86

1903 Wiggin, Kate Douglas (1856-1923) Rebecca of Sunnybrook Farm 73,49

1903 Wharton, Edith (1862-1937) Sanctuary 27,10

1903 Baum, Lyman Frank (1856-1919) The Enchanted Island of Yew 37,34

1903 Crawford, Francis Marion (1854-1909) The Heart of Rome 99,32

1903 Adams, Andy (1859-1935) The Log of a Cowboy 99,28

1903 Gissing, George (1857-1903) The Private Papers of Henry Ryecroft 61,51

1904 Adams, Andy (1859-1935) A Texas Matchmaker 90,90

1904 Corelli, Marie (1855-1924) God’s Good Man 212,23

1904 Haggard, Henry Rider (1856-1925) Stella Fregelius 99,54

1904 Haggard, Henry Rider (1856-1925) The Brethren 128,38

1904 Merriman, Henry Seton (1862-1903) The Last Hope 96,69

1904 Hough, Emerson (1857-1923) The Law of the Land 88,87

1904 Nesbit, Edith (1858-1924) The Phoenix and the Carpet 62,17

1904 Connor, Ralph (1860-1937) The Prospector 95,11

1904 Jerome, Jerome Kapla (1859-1927) Tommy and Co. 55,68

1904 Bacheller, Irving (1859-1950) Vergilius 41,77

1905 Haggard, Henry Rider (1856-1925) Ayesha 117,19

1905 Hough, Emerson (1857-1923) Heart’s Desire 85,52

1905 Wharton, Edith (1862-1937) The House of Mirth 130,13

1905 Ward, Humphrey [Mary Augusta Ward] (1851-1920) The Marriage of William Ashe 144,94

1905 Adams, Andy (1859-1935) The Outlet 94,20

1905 Gissing, George (1857-1903) Will Warburton 85,96

1906 Barr, Robert (1850-1912) A Rock in the Baltic 60,29

1906 Baum, Lyman Frank (1856-1919) Aunt Jane’s Nieces 51,44

1906 Baum, Lyman Frank (1856-1919) Aunt Jane’s Nieces Abroad 57,02

1906 Haggard, Henry Rider (1856-1925) Benita 77,37

1906 Ward, Humphrey [Mary Augusta Ward] (1851-1920) Fenwick’s Career 97,19

1906 Atherton, Gertrude (1857-1935) Rezanov 64,36

1906 Doyle, Arthur Conan (1859-1930) Sir Nigel 130,20

1906 Burnett, Frances (1849-1924) The Dawn of a Tomorrow 18,39

1906 Connor, Ralph (1860-1937) The Doctor 95,79

1906 Nesbit, Edith (1858-1924) The Incomplete Amorist 92,58

1906 Nesbit, Edith (1858-1924) The Railway Children 59,95

1906 Burnett, Frances (1849-1924) The Shuttle 214,62

1906 Nesbit, Edith (1858-1924) The Story of the Amulet 69,25

1907 Haggard, Henry Rider (1856-1925) Fair Margeret 105,68

1907 Wharton, Edith (1862-1937) Madame de Treymes 19,73

1907 Wiggin, Kate Douglas (1856-1923) New Chronicles of Rebecca 61,24

1907 Baum, Lyman Frank (1856-1919) Ozma of Oz 38,46

1907 Adams, Andy (1859-1935) Reed Anthony, Cowman 92,86

1907 Nesbit, Edith (1858-1924) The Enchanted Castle 68,58

1907 Wiggin, Kate Douglas (1856-1923) The Old Peobody Pew 12,88

1907 Hough, Emerson (1857-1923) The Way of a Man 97,38

1907 Parker, Gilbert (1862-1932) The Weavers 160,83

1908 Baum, Lyman Frank (1856-1919) Dorothy and the Wizard of Oz 41,29

1908 Haggard, Henry Rider (1856-1925) The Ghost Kings 122,99

1908 Ward, Humphrey [Mary Augusta Ward] (1851-1920) The Testing of Diana Mallory 151,78

1908 Haggard, Henry Rider (1856-1925) The Yellow God 92,07

1909 Hough, Emerson (1857-1923) 54-40 or Fight! 89,17

1909 Wiggin, Kate Douglas (1856-1923) Homespun Tales 72,08

1909 Parker, Gilbert (1862-1932) Northern Lights 112,78

1909 Connor, Ralph (1860-1937) The Foreigner 86,16

1909 Haggard, Henry Rider (1856-1925) The Lady of Blossholme 95,77

1909 Burnett, Frances (1849-1924) The Secret Garden 81,19

1909 Jerome, Jerome Kapla (1859-1927) They and I 70,11

1910 Haggard, Henry Rider (1856-1925) Morning Star 85,99

1910 Haggard, Henry Rider (1856-1925) Queen Sheba’s Ring 102,11

1910 Baum, Lyman Frank (1856-1919) The Emerald City of Oz 53,15

1910 Hough, Emerson (1857-1923) The Purchase Price 89,91

1910 Barr, Robert (1850-1912) The Sword Maker 112,32

1911 Baum, Lyman Frank (1856-1919) Aunt Jane’s Nieces and Uncle John 43,63

1911 Wharton, Edith (1862-1937) Ethan Forme 34,00

1911 Bacheller, Irving (1859-1950) Keeping up with Lizzie 21,61

1911 Wiggin, Kate Douglas (1856-1923) Mother Carey’s Chickens 72,12

1911 Ward, Humphrey [Mary Augusta Ward] (1851-1920) The Case of Richard Meynell 154,19

1911 Corelli, Marie (1855-1924) The Life Everlasting 131,36

1911 Adams, Andy (1859-1935) Wells Brothers 74,40

1912 Baum, Lyman Frank (1856-1919) Aunt Jane’s Nieces on Vacation 48,26

1912 Baum, Lyman Frank (1856-1919) Sky Island 54,28

1912 Connor, Ralph (1860-1937) Corporal Cameron of the North West Mounted Police 124,59

1912 Doyle, Arthur Conan (1859-1930) The Lost World 75,95

1912 Wharton, Edith (1862-1937) The Reef 96,13

1913 Merriman, Henry Seton (1862-1903) Roden’s Corner 84,09

1913 Burnett, Frances (1849-1924) T. Tembarom 181,25

1913 Ward, Humphrey [Mary Augusta Ward] (1851-1920) The Coryston Family 86,78

1913 Wharton, Edith (1862-1937) The Custom of the Country 140,08

1913 Parker, Gilbert (1862-1932) The Judgment House 149,03

1913 Ward, Humphrey [Mary Augusta Ward] (1851-1920) The Mating of Lydia 132,82

1913 Connor, Ralph (1860-1937) The Patrol of the Sundance Trail 94,39

1913 Doyle, Arthur Conan (1859-1930) The Poison Belt 29,09

1913 Caine, Hall (1853-1931) The Woman thou gavest me 240,47

1914 Wiggin, Kate Douglas (1856-1923) A Summer in a Canyon 54,77

1914 Ward, Humphrey [Mary Augusta Ward] (1851-1920) Delia Blanchflower 109,89

1914 Corelli, Marie (1855-1924) Innocent 127,62

1914 Burnett, Frances (1849-1924) The Lost Prince 99,15

1915 Ward, Humphrey [Mary Augusta Ward] (1851-1920) A Great Success 32,05

1915 Wiggin, Kate Douglas (1856-1923) Penelope’s Postscripts 30,83

1915 Baum, Lyman Frank (1856-1919) The Scarecrow of Oz 46,35

1916 Ward, Humphrey [Mary Augusta Ward] (1851-1920) Lady Connie 115,72

1916 Baum, Lyman Frank (1856-1919) Mary Louise 46,02

1917 Wharton, Edith (1862-1937) Summer 57,21

1917 Bacheller, Irving (1859-1950) The Light in the Clearing 91,03

1917 Connor, Ralph (1860-1937) The Major 118,78

1918 Hough, Emerson (1857-1923) The Passing of the Frontier 35,81

1918 Atherton, Gertrude (1857-1935) The White Morning 32,55

1918 Parker, Gilbert (1862-1932) The World for Sale 103,99

1918 Parker, Gilbert (1862-1932) Wild Youth 46,42

1919 Jerome, Jerome Kapla (1859-1927) All Roads Lead to Calvary 92,35

1919 Atherton, Gertrude (1857-1935) The Avalanche 41,24

1920 Wharton, Edith (1862-1937) The Age of Innocence 102,24

1920 Burnett, Frances (1849-1924) The White People 21,77

1921 Parker, Gilbert (1862-1932) No Defence 84,81

1921 Corelli, Marie (1855-1924) The Secret Power 93,82

1921 Atherton, Gertrude (1857-1935) The Sisters-in-Law 112,86

1921 Connor, Ralph (1860-1937) To Him that Hath 80,56

1922 Bacheller, Irving (1859-1950) In the Days of Poor Richard 101,25

1922 Atherton, Gertrude (1857-1935) Sleeping Fires 53,62

1922 Hough, Emerson (1857-1923) The Covered Wagon 87,44

1922 Wharton, Edith (1862-1937) The Glimpses of the Moon 83,27

1922 Burnett, Frances (1849-1924) The Head of the House of Coombe 111,80

Table 2.6. The Corpus of English novels.

PART II Complementation and change

“Signs are small measurable things,

but interpretations are illimitable”

(George Eliot)

CHAPTER 3: Complementation

In his typological study of complementation, Noonan (1985: 88) writes that “Com-

plementation is basically a matter of matching a particular complement type to a

particular complement-taking predicate”. This matching-problem turns out to be the

single issue on which most functionally and cognitively oriented research in the area

of complementation has centred, the question of why the different complement types

in a language distribute differently over the inventory of complement-taking predi-

cates and how language users decide which complement type to combine with which

predicate. The ultimate goal of such research is to explain, for instance, why in Eng-

lish the verb want combines with a to-infinitive, while the verb enjoy selects an ing-

complement, and not the other way around; or why some verbs, such as wish, can

take more than one complement type – a to-infinitive and a that-clause, in this case –

and what accounts for this variation.

(1) a. “Who is spreading these rumours?” they wanted to know. (BNC)

b. it is natural for a dog to enjoy chewing a good bone (CB)

(2) a. just dial in your card number, your PIN number, and the number you

wish to call. (CB)

b. you may well wish that you had followed your ‘gut reaction’ (CB)

The theoretical problem posed by complementation, then, is essentially a synchronic

one, and what is sought after are the linguistic principles or determinants that organ-

ise speakers’ usage.

Studies of complementation nevertheless take different forms. For one thing,

the matching-problem can be approached from slightly different perspectives. It can

be approached from a complement-oriented perspective, taking a particular com-

plement type and asking which predicates it combines with and why; or, it can be

approached from a predicate-oriented perspective, taking a specific predicate and

looking at which complement-types it selects, with the goal of finding out why it

sometimes takes one and sometimes takes the other. Moreover, the matching-

problem can be treated as a distributional problem pertaining to the grammar of the

language (asking which combinations the grammar allows and why), or as a prob-

lem of choices applying to discourse and speakers’ lexico-grammatical decisions on

specific occasions of use (asking why a language user selects one specific comple-

ment type on a given occasion). These different perspectives can be expected to bal-

48 – CHAPTER 3: Complementation

ance one another out and in the end produce a consistent picture of the system of

complementation. Put differently, a valid solution to the matching-problem will

automatically incorporate the different perspectives on the issue.

More fundamental differences arise in the various solutions that are proposed

to solve the matching-problem. On the one hand, there is a tension between ap-

proaches that try to formulate highly general principles and those that look for more

local explanations. On the other hand, there is a tension between approaches that

take a strictly semantic perspective on complementation, and those that allow for a

multiplicity of factors, including semantics but also other determinants of variation.

It is the purpose of this chapter to summarise and discuss the major currents in solv-

ing the matching-problem. The focus is on English and (although diachrony is never

far away) on the synchronic organisation of the system of complementation. Further,

the aim is not so much to provide a full explanation of any linguistic phenomenon,

as to examine what such an explanation might have to consist of. In this spirit, Sec-

tion 1 focuses on the initial cognitive-functionalist attempt to formulate general se-

mantic principles in order to account for the use and distribution of different com-

plement types. Section 2 presents a first reaction against this view, inspired by varia-

tionist and corpus-based research, that consists in highlighting the multiplicity of

factors involved in complement choice. Section 3 represents a second reaction, in-

spired by constructional models of language, which consists in drawing attention to

the locally organised character of the system of complementation. As I will argue,

the last approach can reconcile the demands of both the semanticist and the varia-

tionist approaches.

1. One form one meaning

As in other areas of language, the original impetus for functionally and cognitively

oriented research in complementation was a dissatisfaction with the generative mod-

els of language promoted in the 1960s and 1970s. It consisted, more in particular, in

a drive to compensate for the almost total neglect of semantics in generative descrip-

tions of syntax. In his seminal paper on complementation, Bolinger (1968) quite

explicitly attacks generative thinking on language. Bolinger names no particular

source, but Rosenbaum (1967) was probably his main target. This very study has

remained a scapegoat of later functionally inspired work on complementation (e.g.

Wierzbicka 1988), despite the fact that some of its shortcomings had been corrected

early on in the generative literature as well (see for instance Bresnan 1979).

CHAPTER 3: Complementation – 49

In essence, Rosenbaum (1967), taking it for granted that all complement

clauses derive from simple S nodes in deep structure, sets out to describe in trans-

formational terms how complementisers, such as ‘that’, ‘to’, or ‘ing’, are inserted in

complement clauses and how they determine the subsequent transformations that

complement clauses can or must undergo. From the functionalist point of view, his

main offence in this is the assumption – tacit for the most part – that “complemen-

tizers seem, in general, not to affect the semantic interpretation of the complement

sentence” (1967: 28). The corollary of this assumption is that although Rosenbaum

is aware of the distributional restrictions on complement types (or in his view, com-

plementisers), he fails rather dramatically to provide a non-tautological explanation

of those restrictions; instead to resort to postulating lexically-specified features

(unrevealingly named [±C], [±D] and [±E]) that are marked in the complement-

taking predicates and from there determine the predicate’s selectional potential

(1967: 26). With hindsight, this is of course a travesty of an explanation, although in

part it certainly stems from the fact that, from the early generativist point of view,

the variation between complementisers is simply not very interesting.

The reaction that follows primarily restores the relevance of semantics to the

organisation of the system of complementation, setting itself as a general goal to

demonstrate that language structure reflects the main function it implements, com-

municating meaning (Bolinger 1977; Langacker 1987). Inspiration is drawn mostly

from applications of the isomorphic principle (Bolinger 1968; Dixon 1984) and from

typological studies (Givón 1980; Noonan 1985). As such, this type of approach en-

courages a view of complementation as a tightly organised system, with a function-

ally motivated division of labour between different complement types. Under the

assumption that “a difference in syntactic form always spells a difference in mean-

ing” (Bolinger 1968: 127), most studies have naturally tended to define the semantic

contrasts between complement types as the principal topic of investigation. The dis-

tributional and discursive variation between complement constructions is subse-

quently accounted for as the motivated outcome of semantic principles, the assump-

tion being that the semantics of a complement type must be compatible with the se-

mantics of the predicates it combines with and vice versa. In the extreme case,

grammar, including complementation, is seen as uniquely predictable from meaning,

as in the following programmatic statement from Wierzbicka (1988):

Grammar is not semantically arbitrary. On the contrary, grammatical distinc-

tions are motivated (in the synchronic sense) by semantic distinctions; every

grammatical construction is a vehicle of a certain semantic structure; and this


is its raison d’être, and the criterion determining its range of use. (Wierzbicka

1988: 3)

In other words, the key to unravelling the matching-problem is to get to the seman-

tics of different complement types. Interestingly, the kind of evidence used in this

central quest fully coincides with the phenomena that have to be explained – viz. the

distribution of complement types and the contexts where different complement types

are in variation.

In fact, a convincing case can be made that the semantics of a complement

type may constrain its combinatorial potential and thus determine its distribution.

This is relatively plain to see for the typological tendencies described by Givón

(1980, 1990) and Noonan (1985). Both authors demonstrate that a higher degree of

syntactic integration between matrix and complement clause correlates with a higher

degree of conceptual integration (cf. Haiman 1983). An obvious parameter of formal

integration is the choice between a finite clause, which (prototypically) expresses its

own subject and has its own tense and modality marking on the verb, and a non-

finite clause, which (prototypically) cannot express its own subject and does not

allow for independent temporal and modal grounding. On the semantic side, formal

distinctions of this kind can correspond to various degrees of ‘binding’, which ex-

presses the strength of the influence exerted by the agent of the matrix clause over

the agent of the complement clause (Givón 1980, 1990). Or it may correspond to

Noonan’s (1985) distinction between ‘dependent time reference’ and ‘independent

time reference’, which reflects the varying degrees to which aspects of the meaning

or interpretation of a complement clause, for instance with respect to modality or

temporal orientation, follow from the information conveyed by the matrix predicate.

For example, a verb such as know underdetermines the state of affairs that is known:

the situation in the complement clause may be anterior, simultaneous or posterior to

the state of knowing; it may be plausible or implausible; and it may have the knower

as its agent or have an independent agent. By contrast, a verb such as strive reliably

predicts that the action strived for is temporally posterior to the striving and that it is

to be carried out by the person striving. Accordingly, know selects finite that-

clauses,2 whereas strive can select a subject-controlled to-infinitive, as shown in (3)

(cf. Noonan 1985: 103).3

2 I use the term that-clause here (and throughout this study) also to refer to the for-mal variant without explicit that, as in Do you think you are battered unfairly? (CB). 3 For a much more detailed example of the same correspondence between formal and functional integration, see Fischer (1995) on the distinction between bare and to-infinitives in Middle English.


(3) a. I know I’ll love Ireland (CB)

b. But otherwise I think that you should really strive to get her to your

doctor and I think you should be looking for a marital counsellor. (CB)

Although they provide an excellent demonstration of the semantic basis of

syntactic organisation, the principles formulated by Givón (1980, 1990) and Noonan

(1985) are too general to account for the specificities of the English system of com-

plementation. The most evident problem – and the one that is of most concern here –

is that English has different types of non-finite clauses that cannot easily be distin-

guished in terms of syntactic integration in the matrix clause. It is at this point that

the need for more language-specific semantic descriptions becomes particularly out-

spoken. A straightforward case is the variation between infinitival and participial

complements following perception verbs, as in (4). The semantic contrast typically

attributed to this pattern of variation relies on an aspectual distinction, with particip-

ial ing-complements construing the process in the complement clause as ongoing at

the moment of perception, and bare infinitives construing it holistically – that is,

perceived in its entirety (see e.g. Declerck 1981, 1991). Note that characterisations

of this kind, too, may have the potential of predicting distributions – for example,

punctual perception verbs may be less compatible with the holistic construal im-

posed by the bare infinitive, because the act of perception cannot encompass the

entire action perceived (e.g. ?she noticed him steal the chairs from the house).

(4) a. When she heard the door open far below her and someone start up, her

breathing tightened, as though she were the one mounting the steps.

(CB)

b. You can hear the bottles being smashed all the way up the road. (CB)

The major challenge in the English system of complementation, however, is

posed by the two most frequent non-finite complement types, to-infinitives and ing-

complements. Not surprisingly, the relation between these two complement types is

frequently commented on in general grammars (Quirk et al. 1985; Declerck 1991;

Langacker 1991; Huddleston & Pullum 2002) and has been the topic of a number of

detailed studies (Wood 1956; Bladon 1968; Bolinger 1968; Ney 1981; Wierzbicka

1988; Rudanko 1989; Duffley 1992, 1999, 2000, 2004; Smith & Escobedo 2001).

As I will discuss the distribution of ing-complements in detail in Chapters 7, 9 and

10, I will here primarily focus on the way these and other studies have dealt with the


to-infinitive, and evaluate semantically oriented approaches to complementation in

this light.

As it happens, the semantic characterisation of the to-infinitive can be

counted a success, and serves as another excellent example of the benefits of a se-

mantic approach to complementation. It is commonly agreed that the to-infinitive

expresses some kind of potentiality (alternatively labelled as ‘irrealis’, ‘purpose’,

‘volition’, ‘future’, etc.), and that its distribution as a complement clause is effec-

tively restricted or at least strongly skewed to verbs more or less compatible with

such semantics (see, in addition to the references cited above, Curme 1931; Jesper-

sen 1940; Bresnan 1979; Mair 1990; Langacker 1992; Fischer 2000; Los 2005).4

The arguments in support of this view are considerable. First, there is the distribu-

tional fact, first observed by Bolinger (1968), that the verbs taking to-infinitival

complements typically denote the subject’s orientation to some goal. For example,

drawing on a survey of corpus data, Mair (1990: 102-3), lists the following as typi-

cal verbs taking subject-controlled to-infinitival complements: ache, agree, aim,

aspire, arrange, beg, condescend, decide, expect, intend, long, plan, try, struggle,

volunteer, wait, want, wish, yearn. Second, where this distributional principle is

breached, the to-infinitival complement nevertheless tends to retain its semantics of

potentiality, as is illustrated in (5), where say is interpreted as denoting a directive (if

figurative) speech-act, so that to take the Honda still denotes a posterior event.

(5) Logic says to take the Honda. You get so much more car. (CB)

Third, that to-infinitival complements carry meaning is evident from the ‘minimal

pairs’ found with a number of complement-taking predicates. For example, with the

verb remember, ing-complements and to-infinitives contrast in expressing anteri-

ority and posteriority respectively, as is illustrated in (6) (see also Bolinger 1968:

123; Quirk et al. 1985: 1193; Wierzbicka 1988: 23; Fanego 1996b). Another exam-

ple is the contrast found with start, where the to-infinitive, unlike the ing-

complement, can be used when the subject is on the brink of initiating the action in

the complement clause but does not actually do so, as illustrated in (7) (Duffley

1999; Mair 2002). Once more, it is clear that these observations fit comfortably with

the general semantic characterisation of the to-infinitive as marking potentiality.

4 Indeed, this view has appealed even to more generatively inspired syntacticians, who have come to treat, under various guises, infinitival to as a non-finite modal element (e.g. Pullum 1982).


(6) a. I can remember finding a woman in my office. We struggled, then

nothing. (BNC)

b. Someone will remember to wake me up early tomorrow, won’t they?

(BNC)

(7) a. Cramer started to say something to Lucinda, but she spoke first (CB)

b. *Cramer started saying something to Lucinda, but she spoke first

Nevertheless, continuing to take the English to-infinitive as an example, we

can also see how and where the semantic approach – or an oversimplified concep-

tion of it – may run into difficulties. Several of these difficulties relate to the behav-

iour of predicate-complement combinations as constructions, so that the observa-

tions made here will be taken up again in Section 3 below. For a start, consider again

(5) above, rendered here anew as (8).

(8) Logic says to take the Honda. You get so much more car. (CB)

To be sure, the meaning of (8) does not contradict the characterisation of the to-

infinitive as marking potentiality or posteriority. Yet strictly speaking, the semantics

of posteriority/potentiality that (supposedly) attach to the to-infinitive underdeter-

mine the eventual interpretation of say as denoting a directive speech-act. For exam-

ple, mere posteriority would also be compatible with an interpretation of say as de-

noting an act of prediction. The question then is where exactly the directive sense in

say comes from. The easiest solution is to assume that say is used in analogy to

other directive speech act verbs also used with the to-infinitive, such as ask, order or

demand, deriving its directive sense from those recurrent combinations. But that of

course means that something else is brought into play next to the semantics of a

complement type, namely the semantic characteristics of entire predicate-

complement constructions, as will be argued in Section 3 below.

More seriously, certain to-infinitival complements simply do not seem to ex-

press potentiality at all. De Smet & Cuyckens (2005), for instance, have described

how, next to its original volitional meaning illustrated in (9a), the combination of

like with to-infinitive acquired the habitual interpretation evident in (9b). This habit-

ual meaning is certainly less closely related to the notion of potentiality, but even if

it is assumed, with Duffley (2004), that the habitual semantics still contain traces of

volition, and hence of the expected semantics of the to-infinitive, there is no denying

that from a historical point of view the semantic interplay between matrix verb and

to-infinitive no longer yields the same outcome as it used to. Where the combination


of an emotive verb and a to-infinitive used to be interpreted as denoting the matrix

subject’s desire with respect to a posterior action, it now denotes a habitual action, at

best compatible with the matrix subject’s desires. Again, then, the assumed seman-

tics of the complement type underdetermine the meaning of the combination of

predicate and complement. But what is more, the example of like with to-infinitive

suggests that the original semantics of a complement-type may begin to disintegrate

as a result of the semantic changes affecting the constructions in which it figures.

(9) a. I dare say we shall sleep here too, for I told Mr. Seagrave positively

not to expect us back to-night. I did not like to say so before your

mother, she is so anxious about you. (1841, CLMETEV)

b. “Substitute Prosecutor Lapo Bardi, terrorism’s most predatory enemy”

as the newspapers liked to describe him. (CB)

Some development of this kind may have to account for the to-infinitival

complements in exceptional case-marking constructions found with verbs such as

say, believe, or find, as in (10a). Of these even a strong proponent of the isomorphic

principle as Wierzbicka (1988) feels obliged to admit that they are best classified as

having a different semantic import from the to-infinitives with other verbs. Equally

inconsistent with the semantics of potentiality is the use of the to-infinitive with

emotive adjectives like sorry, happy or glad, as in (10b), its use in a so-called sub-

ject-to-subject raising construction with happen in (10c), or its use in another habit-

ual construction with wont, as in (10d). In light of examples like these, the isomor-

phic principle in its purest form becomes hard to maintain. What we find instead is

that the various uses of the to-infinitive are not necessarily semantically homogene-

ous.

(10) a. Norwich rain has been found to be twice as acid as rain at a rural site

nearby. (BNC)

b. The guard walked round to the passenger window, glad to get away

from the stench of Whitlock’s clothing. (CB)

c. And you just happened to have some ether in your car? (CB)

d. Parliamentary democracy is untidy difficult to operate and frequently

makes mistakes, but, as Sir Winston was wont to remark, the alterna-

tive is far worse. (CB)


Yet another problem is that some semantic distinctions turn out to be some-

what uninformative in practice, leaving us to wonder whether semantics is really the

decisive factor in explaining certain patterns of variation. One problem-case is the

variation between the to-infinitives in exceptional case-marking constructions, as in

(10a) above, and the extraposed that-clauses found with exactly the same class of

verbs, as illustrated in (11).

(11) Rail’s stables were quarantined when it was found that both he and his

race horses were suffering from the same mysterious symptoms. (CB)

Checking previous claims against corpus data, Noël (2003) convincingly argues that

the semantic distinction drawn by Wierzbicka (1988) between subjective opinion

(expressed by the to-infinitive) and objective truth (expressed by the extraposed

that-clause) has no more bearing on actual usage than the apparently contrary dis-

tinction drawn by Verspoor (1993) between strong commitment (for the to-

infinitive) and weak commitment (for the that-clause). Following this rebuttal, Noël

goes on to show that the choice is rather determined by the demands of information

structure and the discursive prominence attributed to different clausal participants.

On a broad conception of semantics, considerations of this kind may still be re-

garded as constructional meanings – indeed, Noël has to recognise that his explana-

tion in pragmatic terms parallels Langacker’s (1991), who nevertheless continues to

see his as based in semantics. However, regardless of whether they are pragmatic or

semantic in nature, there can be little doubt that the principles accounting for the

choice between to-infinitive and that-clause are far from compelling and can at best

predict tendencies of use. As such, Noël’s findings certainly question the claim that

all choices are semantically determined.

Looking beyond the to-infinitive, similar problems can be found for other

complement types. A good example is the ing-complements with and without from

following verbs such as prevent, as illustrated in (9). For these, Dixon (1991) – fol-

lowed by Rudanko (2003), albeit with less conviction – has postulated a semantic

distinction revolving around the notions of prevention by direct means (expressed by

the form without from) and prevention by indirect means (expressed by the form

with from). The corpus examples in (12a-b), however, would seem to behave con-

trary to what Dixon’s principle predicts (surely, covering a plant with a brick is a

more direct means of preventing its growth than is putting a label on magazines to

keep teenage girls from reading their contents).


(12) a. If you place a brick on the surface of the soil it will prevent the stem

from springing out. (CB)

b. Teen magazines face a crackdown from MPs. Some want a new health

warning on the covers to prevent young girls reading sexually explicit

advice. (CB)

It is unfair to discard Dixon’s distinction on the basis of two examples – there may

be a tendency observable over larger amounts of data – but even so it should be clear

that his claim is unconvincing when formulated in its strongest sense, namely that

the presence or absence of from semantically encodes indirect or direct causation,

committing a speaker to having stated as much the moment they make the choice

between either variant. It seems equally unlikely, moreover, that the distinction be-

tween direct and indirect causation, which more often than not is difficult to make in

the first place, is the raison d’être for the variation attested, as Wierzbicka (1988)

would have had it (see above).5 In the words of Noël (2003: 371): “the use of one or

the other construction does not seem a conscious choice by the speaker to convey a

particular meaning”.

To be sure, if it can be argued that the semantic characterisations of comple-

ment types do not necessarily explain why one is chosen over the other in a given

environment, this is still not the same as saying that the combinatorial potential of

predicates and complement types is semantically unmotivated. For example, the

verb intend combines both with to-infinitives and with ing-complements. In this

particular case, it is usually impossible to come up with a semantic explanation of

the choices speakers make. Fanego (1996a: 50), for instance, provides the two ex-

amples in (13), from the writings of Lady Mary Wortley Montagu, adding that there

does not seem to be “any apparent difference as regards the dependent verb in-

volved, the type of subject, or the status of the clause as dependent or nondepen-

dent”.

(13) a. I intend staying here the remainder of this Month. ‘Tis impossible to be

in a more agreeable place (1749, Fanego 1996a: 50)

5 In this respect, Kempson & Quirk (1971) conduct a series of elicitation tests whose results give a good illustration of the kind of situation we are facing, suggesting that semantic considerations play a role in complement choice but are not compelling. For a detailed discussion, see De Smet (2004).


b. I am still at Louvere, thô the high Season for drinking the Waters is

over, but my Health is so much mended by them that I intend to stay

some time longer. (1749, Fanego 1996a: 50)

The indeterminacy of choice is to be situated on the level of usage, however. On the

more abstract distributional level, both the to-infinitive and the ing-complement are

motivated. The to-infinitive is motivated because intend is a verb of volition and

therefore readily combines with a complement type expressing potentiality (cf. Ru-

danko 1989); the ing-complement is motivated because, as we will see in Chapters 7

and 9, intend is a transitive verb that is semantically compatible with actions con-

trolled by the matrix subject.

The question arises therefore whether semantic accounts also face problems

on the distributional level. The fact that the semantics of complement types are not

always consistent, pointed out above, is to some extent a distributional problem,

since overly strict semantic principles will underdetermine a complement’s actual

range of occurrence. Conversely, it also happens that a complement’s range of oc-

currence contains unexpected gaps – in such cases, its range is semantically overde-

termined. In the following chapters, I will argue that this is so for for…to-infinitives

(Chapter 6), gerundial ing-complements (Chapters 7 and 9) and especially for inte-

grated participle clauses (Chapter 10). For now, it suffices to point out that asymme-

tries as in (14) certainly look suspicious: if one transitive verb of volition, intend,

takes ing-complements, than why not another transitive verb of volition such as

wish?

(14) a. You seem to be under the misapprehension that I intend making love to

you. (BNC)

b. *You seem to be under the misapprehension that I wish making love to

you.

Similar slight but hard-to-ignore anomalies are found for the to-infinitive. For ex-

ample, while the majority of volitional predicates are more than happy to sanction

to-infinitives, a few verbs that appear to have volitional semantics exclusively com-

bine with ing-complements. Compare the behaviour of the verb mind, which ex-

presses objection or negative volition, to that of the roughly synonymous adjectival

predicate be unwilling in (15).


(15) a. It wasn’t that she minded working in the shop, just that there were a

million things she really wanted to get done. (CB)

b. *It wasn’t that she minded to work in the shop, just that there were a

million things she really wanted to get done.

c. If they are unwilling to work, they must be beaten with the whip or the

stick. (CB)

In summary, then, strictly semantic approaches to complementation face four

different problems. First, the meaning attributed to complement types may under-

specify the meaning of particular predicate-complement combinations. Second, the

meaning attributed to a complement type is not necessarily attested across all the

different uses of that complement type. Third, in the event of variation between

complement types in a specific lexical environment, semantic principles may be

insufficiently capable of predicting speakers’ actual choices. Fourth, the distribution

of complement types over the inventory of complement-taking predicates may con-

tain (small) anomalies. This being said, it is clear that the contribution of semanti-

cists to complementation research is invaluable and that semantics plays a crucial

role in shaping the system of complementation.

2. A multiplicity of factors

Despite the obvious merits of semantic approaches, it is clear that semantics cannot

offer a complete explanation of complementation phenomena, and various linguists

have begun to look for other determining principles. They have concentrated on one

problem area in particular, the variation between complement types in specific lexi-

cal environments, showing that such variation is determined by a multiplicity of

factors. The impetus for this kind of research comes, on the one hand, from the idea

that the functional principles of grammar must interact with one another and that

discourse and even language structure emerge from a continual striving to maintain

a balance between various and sometimes opposing functional demands (Haiman

1980, 1983; Hopper & Thompson 1980; Givón 1984; Halliday 1985; Hawkins 2004;

Croft 2001). On the other, it comes from the growing awareness of subtle tendencies

of use that are not apparent to the naked eye of linguistic intuition but that do surface

in the ever more fine-grained empirical analyses of corpus data (e.g. Gries 2003;

Rohdenburg & Mondorf 2003; De Sutter 2005).

In the area of complementation, the corpus-based method in particular has

thrown up a variety of factors that influence in one way or another the on-line


choices speakers make. One example is the role of information structure in the

choice between to-infinitives and that-clauses after verbs such as find, believe, say

and so on (Noël 2003) (cf. Section 1). In all, a number of non-semantic principles

have been put forward that appear to influence numerous instances of variation

across the system of complementation. The simplest illustration is the principle of

horror æqui, which predicts that speakers will avoid the repetition of structures.

Thus, Rudanko (2000: 111-2) finds that the verb avoid, which normally selects ing-

complements, can combine with the to-infinitive when it is itself in the ing-form, as

illustrated in (16). Similar examples are discussed by Rohdenburg (1995), Fanego

(1996: 59, 2007: 176-8), Rudanko (2000: 116) and Vosberg (2003), and some more

will crop up in the discussion in Chapter 9. Schlüter (2005: 324), finally, gives a

more general overview of the phenomenon.

(16) in avoiding to be thought a rake, he hardly escaped being considered as

a fool (1770, Rudanko 2000: 111-2)

A second principle is that of ‘cognitive complexity’, formulated by Rohden-

burg (1995, 1996) and further substantiated in Rohdenburg (2002, 2003) and

Schlüter (2005), among others. The principle states that “In the case of more or less

explicit grammatical options the more explicit one(s) will tend to be favored in cog-

nitively more complex environments” (1996: 151). Grammatical explicitness is un-

derstood by Rohdenburg in terms of formal coding: more formal coding (i.e. bulkier

means of expression) means greater grammatical explicitness. Cognitively complex

environments, on the other hand, are defined by the hypothesised effort involved in

processing: for example, elements intervening between dependents or great constitu-

ent length make for increased cognitive complexity. Accordingly, in his 1995 article

on that-clause complements and to-infinitival complements, Rohdenburg expresses

the relationship between explicitness and complexity as follows:

The less directly the dependent clause is linked to its superordinate clause, or

the more complex the dependent clause turns out to be, the greater is the need

to make its sentential status more explicit (Rohdenburg 1995: 368)

Using historical data on the variation between that-clauses and to-infinitives, Roh-

denburg (1995) can, among other things, demonstrate that the formally more explicit

that-clauses are preferred when material intervenes between matrix verb and com-

plement clause, as is illustrated by the contrast in (17a-b), or that that-clauses are


favoured when the syntax and semantics of the complement clause is complicated by

a negative element, as in (17c).

(17) a. I therefore often begged his Honour to let me go among the herds of

the Yahoos in the neighbourhood (1726, Rohdenburg 1995: 376)

b. I told him who we were, and begged him in consideration of being

Christians and Protestants, of neighbouring countries, in strict alliance,

that he would move the captains to take some pity on us. (1726, Roh-

denburg 1995: 376)

c. and withal commanded them that they should not ordayne for him any

more but so small a competence, as might euen scarcely maintain na-

ture (1608, Rohdenburg 1995: 378)

A third principle is that of ‘rhythmic alternation’, whose operation has been

worked out in great depth by Schlüter (2005). The principle of rhythmic alternation

demands that stressed syllables alternate with unstressed syllables, so that the imme-

diate consecution of two stressed or unstressed syllables is avoided. Dealing with

clausal complementation, Schlüter examines the variation between bare and to-

infinitives and between ing-complements with and without the a-prefix. She finds

robust quantitative evidence that infinitival to and prefixal a-, both unstressed ele-

ments, serve as ‘stress-clash buffers’, dividing stressed syllables that would other-

wise collide. The pair of examples in (18) illustrates how sentence rhythm can be

optimised by the presence or absence of prefixal a-, with accents marking the rele-

vant stressed syllables. The same is illustrated for infinitival to in (19).

(18) a. I never should have sét it rólling. (1879, Schlüter 2005: 218)

b. Sometimes the faint sounds of the soldier’s songs would reach me [...]

and sét my héart a béating to be with them once more. (1844, Schlüter

2005: 215)

(19) a. He is a coward and dáres survíve his honour; let him live. (1830,

Schlüter 2005: 207)

b. The man who dáres to lóve her ought at any rate to be something in the

world (1880, Schlüter 2005: 207)

In addition to these principles that are primarily grounded in the way lan-

guage is processed (see especially Schlüter 2005: ch.6), the variation between com-

plement types also appears to be sensitive to the better-known variables of variation-


ist research, in particular, social and geographical stratification and register – al-

though it must immediately be added that these variables have so far received little

attention in complementation research. Of course, there are the clearly dialectal pat-

terns. An example is the infinitival reinforcer for before the to-infinitive, surviving

from Middle English in Northern Ireland and in the south of the United States. An

attested instance is given in (20) (from the lyrics of Our mother the mountain by the

Texan singer-songwriter Townes Van Zandt).

(20) I watch her, I love her, I long for to touch her

Less ad hoc is Mair’s (2002, 2003) fairly detailed case-study, which points to re-

gional influence on the variation between to-infinitives and ing-complements fol-

lowing the inceptives start and begin, with American English more strongly favour-

ing ing-complements than British English. Data from Mindt (2000) (quoted in Mair

2002: 116) further show that ing-complements with start are proportionately more

common in conversation than in expository writing and fiction. If, in addition, we

take into account that ing-complements have been spreading at the expense of to-

infinitives with start and begin over the last one or two centuries, we have quite a

few of the ingredients for a language situation in flux that might also respond to the

more fine-grained categories of sociolinguistic analysis – this, however, is as yet to

be demonstrated.

In all events, from this brief overview it is certainly clear that not all in syntax

is semantics (Noël 2003). Even though we have to accept the idea that syntactic

variation is constrained by the different meanings expressed by the variants in-

volved, we also have to acknowledge that semantic determination is sometimes

overridden by clearly a-semantic principles and by sociolinguistic variables such as

dialect and register. This in turn suggests that some of the supposed semantic dis-

tinctions cannot be all that critical in the first place. Variationist studies thus show

that the use of different complement types is determined by a multiplicity of func-

tional and contextual considerations.

At the same time, the multiplication of relevant principles and variables has

not led to a better understanding of the organisation of the system of complementa-

tion as a whole. The principles and variables posited all operate on the variation that

exists but do not explain the variation per se. As a consequence, while much light

has been cast on the choice between complement types in specific lexical environ-

ments, the prior question why a complement type can occur in a specific environ-

ment remains unresolved and to a large extent even unaddressed. The problem is


that, on the whole, the variationist factors examined are sensitive to more or less

incidental features of the context of use, so that, in the majority of cases, they seem

to have no lasting effect on grammatical structure and the issue of distribution. The

principle of horror æqui, for instance, can trigger an occasional to-infinitive or ing-

complement where we would not expect one, but unless certain matrix verbs were to

occur exclusively either as to-infinitives or as ing-complements it cannot determine

which verbs generally prefer to-infinitives and which verbs prefer ing-

complements.6 And yet these combinatorial preferences too are a linguistic fact that

requires an explanation. Analogous arguments can be formulated for the principle of

cognitive complexity and that of rhythmic alternation, and obviously for the socio-

linguistic variables of variation as well.

We can argue this point ex negativo by considering a counter-example.

Schlüter (2005: 185-203) describes the historical development of to-infinitives and

bare infinitives with causative make. She shows that in the initial situation, illus-

trated in (21), causative make allowed both types of infinitive in both its active use

and in its passive use.

(21) a. Thus she spoke with a fierceness that máde the Lóver trémble with fear

of losing her (1687, Schlüter 2005: 194)

b. she acted the part of Zantippe, and máde the Hóuse to ríng with her

scolding (1668, Schlüter 2005: 190)

c. they were most certain, their purses should be máde páy for the

contention (1661, Schlüter 2005: 199)

d. Upon that the Fire was máde to bróil his Flesh, he was stript naked and

tyed to the Tree (1693, Schlüter 2005: 199)

Subsequently, however, bare infinitives disappeared after passive make, while to-

infinitives disappeared after active make, resulting in the strikingly incongruous sys-

tem found today, where active and passive make obligatorily take different comple-

ments. As a reason for this remarkable development, Schlüter (2005: 202) proposes

that both the cognitive complexity principle and the principle of rhythmic alternation

strongly favoured the use of the to-infinitive following passive make, whereas the

active form of make, being cognitively less complex, and being typically followed

by its direct object, which pre-empts a systematic stress clash, made no such func-

6 Some such cases do seem to occur, though – see the discussion of attempt in Chap-ter 9, or of bother in Chapter 10.


tional demands.7 Although the direct causal link between cognitive principles and

historical development is hard to prove, Schlüter certainly manages to demonstrate

the effects of rhythmic alternation and cognitive complexity on the variation that

existed in the seventeenth and eighteenth centuries, so that the least we can say is

that in this case a functionally motivated tendency has ended up grammaticalised, in

the sense of being fully regularised (cf. Keller 1990; Haspelmath 1999). It is clear

that lasting distributional effects of this type may occur only locally and have only a

small impact on a complement type’s overall distribution. For some factor to have a

constant, distributional effect, it needs to be sensitive to a constant feature of the

linguistic items involved in the matching-problem. This finding is of some impor-

tance when it comes to integrating the variationist factors in a more general model of

the system of complementation, as I will argue in the following section.

3. Constructions

We are now left with two problems. The first problem is that we know semantics to

be relevant to complementation but, it seems, not quite in the way suggested by the

isomorphic principle. That is to say, we still have no satisfactory account of how

exactly semantics interact with the matching problem (cf. Section 1). The second

problem is that semantics is clearly not the only determinant of variation and that we

ultimately have to incorporate, at least in principle, a multiplicity of different factors

in our account of complementation (cf. Section 2). Without wishing to hallow one

particular framework, I want to argue here that constructional approaches to lan-

guage (as developed by Hopper 1987, 1988; Langacker 1987, 2003; Goldberg 1995,

2006; Croft 2000, 2001, forthc.; Tomasello 2006) can to a large extent solve the first

problem, and seem compatible with a solution to the second. In what follows, I will

therefore give a brief sketch of construction grammar (Section 3.1) and then attend

in more detail to its problem-solving potential in dealing with complementation

(Section 3.2).

7 Of course, what Schlüter (2005) does not explain is why to-infinitives disappeared with the active form of causative make. The explanation here may lie in semantic analogy between make and other causative verbs such as do, have and let, which also take bare infinitives only.


3.1. Construction grammar

Construction grammar is a usage-based theory of language. This means that gram-

mar is thought to be “the cognitive organization of one’s experience with language”

(Bybee 2006: 711). Put differently, language users are assumed to generalise pat-

terns of language from their everyday experience, picking out and storing the regu-

larities they come across. It is stored regularities of this kind that are known as ‘con-

structions’. In construction grammar, these constructions are believed to be pairings

of meaning and form which, through their multiple interconnectedness, make up the

fabric of speakers’ linguistic knowledge and the basis of their linguistic output.

In practice, every utterance manifests multiple constructions. The utterance in

(22), for instance, contains lexical items, such as today or yoghurt, which are con-

ventionalised pairings of meaning and form, and therefore in a sense constructions;

it contains the transitive construction, which (sticking to traditional terminology)

consists of the schematic slots for a subject, a verb and an object, and conveys the

abstract meaning that an agent does something to affect a patient; it contains the

perfect construction, consisting of a form of the verb have and a past participle, and

conveying perfect aspect, as well as the progressive construction, consisting of the

form be and a present participle, conveying progressive aspect; and so forth.

(22) Today I’ve been mostly eating yoghurt!

By comparing an utterance like (22) to other utterances and by analysing its form

and meaning, language users are able to infer schematic constructions like these,

store them and eventually exploit them in their own speech production.

In addition to pairing a particular form to a particular meaning, constructions

are also the carriers of other linguistic knowledge. Most importantly, constructions

carry information specifying how they relate to other constructions. Two such rela-

tions are generally acknowledged by construction grammarians (though, as I will

suggest in Chapter 4, a third may have to be added). One is the compositional rela-

tion of integration. Constructions specify which other constructions they may con-

tain and in which other constructions they themselves may be contained. For exam-

ple, the ‘noun phrase construction’, as realised in the faintly whispered words, has a

slot for a ‘nominal premodifier construction’, in this case filled by faintly whispered.

In turn, noun phrase constructions can themselves integrate in a clause, as in barely

had the faintly whispered words passed her lips. The other relation type is the hier-

archic relation of instantiation. Speakers may recognise the sequence baked beans as


a recurrent symbolic unit and, given its slightly unpredictable meaning, they are

likely to separately store it as a ready-made chunk of language, yet they will also

realise that it still instantiates the more abstract ‘classifier-noun construction’ and

treat it as such, for instance, by integrating it in other constructions, as in the full

noun phrase the last can of baked beans. It is the infinitely complex network of inte-

grating and instantiating relations that constitutes grammar – or more precisely, as

knowing constructions at once means knowing how they interconnect, constructions

are at once the basic units of grammar and its principle of organisation.

3.2. Construction grammar and complementation

There are certainly a number of important advantages to looking at complementation

through constructional eyes. The least we can say is that construction grammar can

solve many of the problems encountered by other semantically oriented approaches

and has the additional advantage of handling some further features of the system of

complementation not even considered so far. Further, construction grammar may be

capable of integrating into a single account not only semantic principles but also the

functional and sociolinguistics factors considered in the foregoing discussion.

Let us start by looking at how construction grammar deals with the semantics

of complementation. Given that constructions are symbolic units, there is ample

space for semantic description, but there are a number of added advantages. First,

the meanings significant to the matching problem can be located at different levels

of constructional assembly. A given sequence of a predicate and a clausal comple-

ment integrates at least two different constructions: the complement clause and the

combination of a complement-clause with a complement-taking predicate (for con-

venience, we may speak of these as the ‘complement construction’ and the ‘predi-

cate-complement construction’ respectively). Each of these two constructions can

carry its own meaning and, in addition, can specify (semantic) restrictions on the

elements it may contain. This is in contrast to the earlier semantic approaches to

complementation where meaning had to be located in the complement type or the

complementiser as the sole potential carrier of semantic information (e.g. Wierz-

bicka 1988; Duffley 1999).

For example, comparing (23a) to (23b), lest-complements do not look strik-

ingly different from the that-clauses found with the same complement-taking predi-

cates (in fact, the two have been in competition historically; see López-Couso

forthc.). However, as shown by (24), the combinatorial potential of lest-

complements is much more restricted than that of that-clauses. At the same time the


restrictions on the use of lest-complements are obviously semantically motivated, as

lest-clauses only appear with predicates denoting fear or apprehension. While these

facts are hard to capture with a semantic characterisation of lest-complements as

such, they are easy to incorporate in the description of the predicate-complement

construction as a whole.

(23) a. Afraid lest she had revealed too much, she added quickly, “Tell me

about your visit to Maythorpe House.” (BNC)

b. He tried to quell the rising joy within, afraid that he’d do something

foolish (CB)

(24) a. Rik is making good progress now. We are very hopeful that he will

return to full health. (CB)

b. He had even imagined that she mocked him with her smile, though he

could not be certain that she had actually seen him there. (BNC)

c. *Rik is making good progress now. We are very hopeful lest he will

return to full health.

d. *He had even imagined lest she mocked him with her smile, though he

could not be certain that she had actually seen him there.

Of course, this alternative view of semantic motivation fits in with the tradition of

dividing complement-taking predicates into groups of semantically related verbs

(see e.g. Visser 1963-73; Rudanko 1989, 1998; Fanego 1996a). The extra a con-

structional approach has to offer is its theoretical framework, which can account for

the different groupings without necessarily having to recur to the semantics of com-

plement types and their compatibility or incompatibility with the meanings of matrix

predicates. This is especially an advantage when it turns out to be difficult to isolate

the semantics of a complement type.

Various similar phenomena become less troubling when the isolated comple-

ment clause is no longer taken as the only possible carrier of semantic information.

The to-infinitive in exceptional case marking constructions discussed above is a case

in point (cf. Section 1). While it is difficult to say exactly what the to-infinitive se-

mantically contributes to the construction (when compared, for instance, to an extra-

posed that-clause), it is clear that the predicate-complement construction as a whole

is found only with a specific class of semantically related matrix verbs and also that

the construction is used with a particular communicative intent. Specifically, the

matrix verbs found in the construction are all verbs of information-handling, and as

shown by Noël (2001), the construction tends to be used to evaluate information in


terms of its evidential value, highlighting how certain information is socially repre-

sented or transmitted, as is illustrated in (25). At least from this point of view, the

use and distribution of the to-infinitive can by no means be claimed to be semanti-

cally unmotivated.

(25) a. The Prince is known to be very interested in the fate of the rain forests.

(CB)

b. Joseph Truman, “based on papers now lost”, is said to have set up

business as a brewer in 1666, the year of the Great Fire of London.

(CB)

Second, speakers, constantly analysing and re-analysing the regularities of

language, may in the course of time reinterpret the location of meaning in a con-

struction (Croft 2000). This may explain various breaches of the isomorphic princi-

ple. Take lest-complements again. If we compare lest-complements, as in (26a), to

adverbial lest-clauses, as in (26b), we find that the latter express an idea of negative

purpose that is absent in the former (i.e. (26b) can be paraphrased, ‘so that he would

not burst into tears’, but (26a) cannot be paraphrased as ‘frightened that he would

not see her’).

(26) a. She drew back from the window, frightened lest he should see her

(CB)

b. Was he not like Pinocchio, obliged to laugh lest he burst into tears?

(CB)

From a strictly isomorphic perspective this is a problem, but it is easy enough to see

what may have happened historically. Presumably, adverbial lest-clauses were rein-

terpreted as complements in ambiguous environments such as (27a) (López-Couso

forthc.). The moment this reinterpretation took place, the apprehension pragmati-

cally associated with negative purpose clauses came to be explicitly encoded in the

complement-taking predicate, while the negative semantics of the lest-clause be-

came semantically redundant, since it is pragmatically evident that what is feared is

wished not to happen. Superficially, lest then becomes equivalent to that, as in

(27b). But in actual fact, meaning here does not disappear; it is reinterpreted and

relocated – in this case, to the combinatorial (or even, as in (27b), collocational)

restrictions on the use of lest-clauses.


(27) a. Se ðonne se ðe fundige wislice to sprecanne, ondræde he him suiðlice,

ðylæs his spræc gescynde ða anmodnesse ðara ðe ðærto hlystað (a900,

López-Couso forthc.).

‘he, therefore, who intends to speak wisely ought to fear greatly, so

that his words do not confound the unity of those who listen / ought to

be afraid that his words might confound the unity of those who listen’

(after López-Couso forthc.)

b. All the danger is least we take too much liberty herein. (1657, OED)

Similar mechanisms may account for some of the incongruities found in the

semantics of the to-infinitive pointed out earlier (cf. Section 1). As described above,

the strict semantic approach would have it that to-infinitives express potentiality or

posteriority (Wierzbicka 1988; Duffley 1999, 2000). This works fine for the adver-

bial to-infinitive in (28a), but it is problematic for the to-infinitival complement to

sorry in (28b).

(28) a. She ran to meet her father. (1890, OED)

b. We are sorry to learn from the local police that you have recently been

the victim of crime (CB)

Nevertheless, the use of to-infinitives can be seen to be semantically motivated in

both of these examples, if only we are willing to accept that speakers may extract

and apply different semantic generalisations that only partly overlap. Thus, even

from a purely synchronic point of view, (28a) and (28b) are linked by a series of

related constructions and, in this light, simply represent different ends of a cluster of

perceived regularities. As shown in (29a), to-infinitival complements pattern with

predicates that are compatible with a complement conveying potentiality. This com-

plement use is still in parallel to the adverbial use of the to-infinitive in (28a). The

same complement use may bring speakers to decide that to-infinitives must always

denote a potential situation and that matrix complements must be interpreted accord-

ingly, as in (29b), where the otherwise emotive adjective happy has to be interpreted

as volitional (‘be prepared, willing’).

(29) a. It was supposed to take three to five days, but we aimed to do it in two.

(BNC)

b. Illich was happy to take an oath declaring he would tell the truth. (CB)


However, given the predominance of matrix predicates that imply potentiality any-

way, speakers may also decide that the meaning of potentiality really resides in the

matrix predicate and that the to-infinitive is semantically more neutral. To help such

an alternative interpretation, there are cases where potentiality can be attributed to

yet other elements of the context than the matrix predicate, such as the modal would

in (30a), as well as a host of semantically ambiguous instances, such as (30b).

(30) a. Would you be happy to have that as your eventual epitaph? (CB)

b. The real mother seems more than happy to feed the pups (BNC)

In light of such examples, uses like (31), where happy is purely emotive, are not

unmotivated. The final link in the chain is the recognition of a construction that

sanctions semantically neutral to-infinitives with non-volitional emotive adjectives,

such as sorry in (28b) above.

(31) We could never have believed that one day he would go back to

school. He is just so happy to be there again. (CB)

In other words, apparently incongruous examples such as (28a) and (28b) above are

nonetheless linked semantically by a series of intermediate and ambiguous uses.

Third, a constructional approach allows for ‘constructional splits’. As indi-

cated above, concrete linguistic sequences may simply instantiate more abstract con-

structions, but it is also possible for the sequence to be stored separately as a ready-

made chunk of language. The reason for this may be that the concrete sequence has

something unpredictable about it that cannot be derived from a higher-order schema

(e.g. the semantically idiosyncratic phrasal verb look up) (Goldberg 1995), but it

may also be stored separately simply because it is itself highly frequent (e.g. the

plural form peas) (Langacker 1987; Goldberg 2006). The latter phenomenon gives

rise to the former: because certain sequences are stored separately from their more

abstract parent constructions, they may undergo independent semantic or even for-

mal changes that further differentiate them from their parent constructions (Bybee

1985: 88, 2006: 721-3; Croft 2003: 58-9; Hay 2006; De Smet & Cuyckens 2007:

195-8). This gradational process is constructional split. The process is fairly com-

mon in the system of complementation, if only because complement constructions

are regularly recruited as input to the development of auxiliaries and pragmatic par-

ticles (see e.g. Thompson & Mulac 1991; Heine 1993; Hopper 2001; Kuteva 2001;

see also Hopper’s 1991 notion of ‘divergence’). What happens in such cases is that a


predicate-complement combination develops new meanings and functions inde-

pendently of the lexical use of the matrix predicate or the semantics normally asso-

ciated with the complement type.

Thus, constructional split sometimes interferes with semantic generalisations.

An example already pointed out above is the development of habitual semantics in

the combination of like with a to-infinitive (cf. Section 1). Consider again the exam-

ples given under (9) above, rendered here as (32):

(32) a. I dare say we shall sleep here too, for I told Mr. Seagrave positively

not to expect us back to-night. I did not like to say so before your

mother, she is so anxious about you. (1841, CLMETEV)

b. “Substitute Prosecutor Lapo Bardi, terrorism’s most predatory enemy”

as the newspapers liked to describe him. (CB)

The link between (32a) and (32b) is found in examples like (33), which allow habit

to be inferred from volition as a pragmatic implicature (by the reasoning that desir-

able behaviour is common behaviour). Crucially, however, for the new meaning to

attach itself, language users have to be able to associate the combination of like with

the to-infinitive with the recurrent pragmatic implicatures it triggers, and this is only

possible if the combination of like and to-infinitive is already stored as a partly inde-

pendent unit of language use. In turn, once the semantic change has taken place, it

adds further to the somewhat independent status of the concrete combination of like

and to-infinitive in relation to the more abstract pattern of to-infinitival complemen-

tation.

(33) It’s not my way, you see, to receive my friends with my back to the

fire. I like to give them a hearty reception in the old style at my gate.

(1773, CLMETEV)

Apart from the semantic problems encountered by the isomorphic approach,

there are at least two other phenomena that a constructional framework is more apt

at describing, and that are therefore worth briefly drawing attention to. First, con-

structional approaches allow us to deal with patterns of complementation that are

only semi-productive. That such patterns exist can be shown beyond doubt. For ex-

ample, the verb hear combines with non-controlled bare infinitives without preced-

ing subject, but the verb appearing in the infinitive form must be either say or tell, as

in (34a). The phrasal verb burst out combines with ing-complements and could be


regarded as an aspectual, but the ing-forms it occurs with are almost exclusively

laughing or crying, as in (34b). At one point in the history of English, retrospective

verbs such as remember or recollect, could combine with perfect to-infinitives mark-

ing an anterior situation, but the infinitival verb would in the great majority of cases

be see or some other verb of perception, as in (34c). All of these patterns are charac-

terised by deficient syntax, yet none of them can be described purely as fixed

phrases, for they all have some limited degree of productivity.

(34) a. Had this village really not heard tell of the miracle-working Gabriel?

(BNC)

b. A PAIR of TV newsreaders have been suspended from duty after they

burst out laughing during the evening news. (BNC)

c. I scarcely recollect to have heard one grunting swine or snarling mas-

tiff during my whole progress. (1783, CLMETEV)

The restrictions on productivity are all semantically determined, since in all cases,

the verbs that are allowed to occur in the complement clause belong to the same

lexical field. Consequently, it is again clear that the semantic motivation can hardly

reside in the meaning of the isolated complement types: as the restrictions are de-

pendent on the verb the complement type combines with, the semantic restriction

must lie at the level of the predicate-complement construction. Note further that the

semi-productive nature of the patterns illustrated in (34) will pose problems to tradi-

tional grammars that preferentially handle maximally abstract syntactic structure.

Usage-based theories of language, on the other hand, endorse a view of grammar as

an inventory of expressions of various degrees of specificity and productivity, and

will readily welcome semi-productive patterns of the kind illustrated in (34) (cf.

Pawley & Syder 1983; Bybee 2006: 712-4).

Second, when combining with certain complement-types, some matrix predi-

cates adopt new meanings or show unexpected argument structures. For example,

the verb argue, as shown in (36a), can exceptionally combine with animate objects

when it takes an into...-ing-complement, and in this use, its interpretation also

changes from a verb of communication to a manipulative verb, meaning ‘persuade

someone by the use of arguments’. Similarly, the verb laugh, not normally a verb

that takes objects at all, can combine with a that-clause, and then means ‘say in a

laughing manner’, as shown in (36b).


(36) a. “You found someone up there,” she said, seeing his face. He cleared

his throat. Better not tell her the truth. Better just argue her into going

back where she belonged. (BNC)

b. Molly often laughed that her life had been quiet, restful, and well-

organized until a wee babe was put into her arms. (CB)

Since these special combinatorial and semantic properties of argue and laugh are

unique to one specific construction, it can be assumed that the properties are in fact

constructionally imposed (Goldberg 1995; Israel 1996; Michaelis 2003). The

mechanism needed to explain this again draws on speakers extracting different regu-

larities from usage. For example, the ordinary use of into...-ing-complements is with

manipulative verbs that take animate direct objects, such as cajole, confuse and trick

in (37) (Rudanko 2000: ch.5). It appears, however, that the combination of such

verbs with into...-ing-complements is sufficiently well-established for the manipulat-

ive semantics of the verb to be taken for granted just on the basis of the presence of

an animate direct object and an object-controlled into...-ing-complement. The pat-

tern can therefore extend to other verbs, which are then automatically taken to de-

note some manipulative action. A similar account may be set up for laugh in (36b)

above, where the relevant construction involved is the combination of verbs of com-

munication with that-clauses, as illustrated in (38).

(37) a. he hoped to cajole her into selling him her house (BNC)

b. Today’s voters are not morons; neither are we prepared to follow

blindly an advertising campaign that is meant to confuse and trick peo-

ple into believing what are clearly half-truths. (CB)

(38) Gorbachev says that President Bush is taking credit for winning the

Cold War (CB)

In summary then, there is a variety of semantic and other phenomena in the

area of complementation that a constructional framework can describe or integrate

with relative ease. These include the different ways in which semantic motivation

may manifest itself, the semantic shifts witnessed in the use of specific complement

types, the semantic changes in specific predicate-complement constructions, the

semi-productive nature of some predicate-complement constructions, and the unex-

pected syntax and semantics of some complement-taking predicates. We could say

that, while taking semantic approaches seriously, construction grammar brings in

something of the debunking attitude of variationist research, in the form of an


awareness that semantic regularities are mutable, open-ended, numerous and poten-

tially contradictory, translating the multiplicity of factors that characterises the

variationist approach into a multiplicity of constructions.

However, can construction grammar integrate the more concrete demands of

variationist approaches as well – that is, the actual functional and sociolinguistic

factors that have been shown by variationist research also to affect the matching

problem? There is certainly a degree of theoretical compatibility. The fact that con-

struction grammar is made to run on people’s general cognitive abilities (handling

symbols, inferring patterns, establishing categories, etc.) implies that it does not

need to posit an autonomous and modular language faculty and that, presumably, it

does not in principle resist the various functional and cognitive constraints interfer-

ing with language production (see e.g. Croft 2000). At the same time, functional and

cognitive constraints are not constructions. That is, constraints such as horror æqui,

rhythmic alternation or cognitive complexity, unlike semantic regularities, do not

represent linguistic knowledge stored in language users’ long-term memory. This

means that they are not part of what a construction grammar describes. To fit them

into our picture of the system of complementation, we have to think of the process

of language production as drawing on a set of stored regularities but crucially in-

volving some selectional procedure as well that evaluates possible constructions for

output. Visualising the process of producing a predicate-complement construction,

we can think of speakers first drawing on their store of constructional knowledge to

make possible matches between a complement type and a predicate, and then evalu-

ating those matches by running through a mental checklist, balancing off a great

number of different considerations, operative at various levels of linguistic aware-

ness but all having to do with the appropriateness of expressing a given content in a

particular way. These considerations may involve any of the factors discussed previ-

ously, so that the eventual choice will be determined by the linguistic means of ex-

pression available on the one hand, and by grammatical and semantic well-

formedness, social acceptability, ease of processing, etc. on the other.8

Notice that in this scheme of things the linguistic knowledge stored as con-

structions in the grammar comes into play twice: once as a repository of possible

expressions and once as a measure of the grammatical acceptability of a given lin-

guistic sequence for output. This double role might be artefactual. For example, in a

8 This view of language production is roughly similar to that promoted by optimality theory (Prince & Smolensky 1993; Kager 1999), with at least one important differ-ence: I assume no fixed ranking of constraints, so that variable outcomes are possi-ble (cf. Schlüter 2005: ch.6).


spreading activation model (as applied in Berg 1998; Schlüter 2005; Hudson 2007),

evaluation might be conceived of as a race between competing means of expression

whose selection is determined by the ease with which they are activated. Ease of

activation is in turn a consequence of the degree to which a given expression is sanc-

tioned by other grammatical structures (i.e. the constructions it instantiates) and by

the degree to which it corresponds to the actual content the speaker wants to convey.

In such a view, constructional coinage and evaluation are a single process. Inciden-

tally, some of the functional and cognitive constraints, too, might be fitted into a

spreading activation model (cf. Schlüter 2005).

Be that as it may, the distinction between stored knowledge (i.e. grammar)

and evaluation in language production is useful in as far as it fits in with the applica-

bility of different factors to different levels of language, namely discursive choices

vs. grammatical organisation, corresponding to the different questions that are perti-

nent to the matching problem: what is grammatically possible vs. what form is used

in a specific discourse context. At the same time, all these considerations are taken

into the task of linguistic decision-making, so if it is assumed that linguistic deci-

sions shape usage and usage shapes grammar, even the cognitive and functional

constraints are fed into the grammar and may become (ever so slightly) grammati-

calised, in the sense of being constructionally stored.

4. Conclusions

This chapter has discussed the kind of elements that must be contained in a satisfac-

tory description of complementation in English. The central issue in complementa-

tion studies is how to resolve the matching-problem, that is, how to account for the

fact that certain predicates combine with certain complement types. It has been

shown that matching is determined by a variety of factors, including sociolinguistic

tendencies, processing constraints, and semantic considerations, which may influ-

ence the choices speakers make in specific lexical environments, as well as deter-

mine the grammatical options that speakers have available to choose from in the first

place. Grammatical knowledge can be pictured as consisting of a great network of

regularities, or constructions, which determine usage options. Regularities are not

stable, however, but provisional and erratic, continually branching off in different

directions. The view that thus emerges of complementation as a synchronic system

is one of an intricate network of local interrelated and overlapping regularities, a

patchwork of constructions rather than a tightly-organised system with a neat divi-

sion of labour between different complement-types. As I will argue in Chapter 4,


this view of complementation is also well-suited for describing the processes of

lexical diffusion that appear to characterise the system of complementation when

looked at from a diachronic perspective.

CHAPTER 4: Diffusional change

Synchronically, the matching problem manifests itself as a quest for the principles

that determine which predicates combine with which complements (Chapter 3). Tak-

ing a diachronic point of view, a further complication is added: the matches between

predicates and complement-types turn out to be historically unstable. Looking at the

history of English, various new complement-types have been introduced in the lan-

guage and have subsequently taken a strong foothold in the system of complementa-

tion. In some cases this has happened at the cost of older patterns of complementa-

tion, but in other cases old and new patterns continue to exist side-by-side and the

best we can say is that the system of complementation has become increasingly

complicated.

There are two aspects to the emergence of new complement-types. One is the

complement-type’s previous history – what other construction it derives from and

how it came to be used as a complement-type. The other is the further development

of the complement-type as a complement-type. The historical sources of comple-

ment-types are diverse. For example, the histories of different clause-types elabo-

rated in the following chapters show that complements can derive from adverbial

clauses (this is most clearly the case for the integrated participle clauses discussed in

Chapter 10) or arise from nominalised clauses historically stemming from action

nominals (as is the case for the gerundial ing-complements discussed in Chapters 7

through 9). Once a clause type has established itself as a complement clause, it typi-

cally begins to extend its range of use. The way this commonly happens is through

processes of lexical diffusion. This means that, through time, complement-types are

selected by an ever-growing number of complement-taking predicates. Even a cur-

sory look at the history of gerundial ing-complements, for example, will tell us that

controlled ing-complements occurred with avoid before they occurred with like, and

that they occurred with like before they occurred with try (Visser 1963-73; Fanego

1996a, 2007; see further Chapter 9). The use in (1a) is thus older than that in (1b)

which in turn is older than that in (1c). What is more, the differences in age are re-

markably big: ing-complements appeared with avoid at the end of the sixteenth cen-

tury; with try they appeared at the beginning of the nineteenth, more than two centu-

ries later (see Chapter 9).

(1) a. Complacently relying on an outdated self-image is a great way to avoid

moving on (CB)

78 – CHAPTER 4: Diffusional change

b. Lucky Louie liked licking lollipops lazily. (CB)

c. We tried cold compresses on it but they didn’t help, so then he tried

soaking it in hot water. (CB)

The purpose of this chapter, then, is to examine how complement types de-

velop. The primary focus is on the diffusional changes that take place within the

system of complementation, with the goal of formulating hypotheses about how a

complement-type spreads from one predicate to another. The questions to be re-

solved are, first, what principles and mechanisms of change can account for diffu-

sion in the domain of complementation; and, second, how these mechanisms are

related to the synchronic model of complementation developed in the previous chap-

ter. Note that detailed discussions of specific diffusional changes are reserved for the

chapters in Parts III and IV (see especially Chapters 6, 9 and 10); the present chapter

is only to highlight the key mechanisms of diffusional change in the system of com-

plementation. In as far as attaining this goal depends on an understanding of diffu-

sional change in general, the present chapter also bears on the issue of how clause-

types come to be used as complements in the first place – indeed, this latter type of

development, too, is often part of a larger diffusional change. No attempt is made,

however, at formulating generalisations about the different possible sources of com-

plement-types (which would require a typological perspective). Discussions of how

specific clause types came to function as complements are again reserved for the

chapters in Parts III and IV (see especially Chapters 5, 8 and 10).

In view of its obvious relevance to linguistic theory in general and to histori-

cal linguistics in particular, the issue of diffusion has received relatively little ex-

plicit attention in the literature. Nevertheless, the existing literature proposes a num-

ber of generalisations and it is only natural to take these as a starting-point for the

discussion of diffusion in the system of complementation. I will therefore begin by

discussing these general proposals in Section 1. Section 2 next addresses the contri-

bution the proposals can make to understanding historical developments in the area

of complementation and, drawing on a number of case-studies, sets up a simple

model of change that can be applied specifically to diffusion in the system of com-

plementation.

1. General perspectives on diffusion

If diffusion is broadly understood as the extension of a linguistic pattern to new lin-

guistic environments, with environments including both lexical and syntactic envi-

CHAPTER 4: Diffusional change – 79

ronments, most changes in syntax are diffusional in one way or another. Aitchison

(1991: 98) even seems to suggest that all syntactic changes involve diffusion, when

she writes that “Syntactic changes [...] start out slowly, then, like a snowball bound-

ing down a hill under its own impetus, they suddenly gather up numerous other en-

vironments. Then they slow down.”

It is certainly true that a great variety of changes contain some element of

diffusion. Hopper & Traugott (1993: 56-9), for instance, argue that grammaticalisa-

tion typically involves ‘generalisation’, whereby constraints on the occurrence of a

grammaticalising item are lifted or loosened, sometimes in progressive stages. For

example, modal verbs, in the course of their development, are likely to sanction a

broadening range of subject types, which implies some form of distributional gener-

alisation (Heine 1993); similarly, newly emerging negative elements may initially

combine with only a few lexical verbs and then extend to the entire range of verbs

(Hopper & Traugott 1993: 58). On a higher level of abstraction, separate grammati-

calisations may, when considered in unison, constitute a diffusional change: in Eng-

lish, for instance, the adoption of modal features that marks the development of the

class of modal auxiliaries occurred at different times for different verbs (Plank

1984), suggesting that a modal construction has been diffusing over a set of verbs.

Outside grammaticalisation examples of diffusion abound as well. It is found, for

instance, that subjunctive marking diffuses gradually over different clause-types

(Hengeveld & Wanders 1997); or that word-order changes may be implemented

differentially depending on clause type (see e.g. Traugott 1992). It is further telling

that the diffusion of new complement types is itself often embedded in some larger

pattern of diffusion. English to-infinitives, for instance, first appeared as adverbial

clauses and later spread to subject and object positions (Los 2005) – a pattern of

change that can be witnessed in many Indo-European languages (Haspelmath 1989)

– and the spread of ing-complements and for...to-infinitival complements is simi-

larly set against the background of more extensive diffusional changes (see Chapters

5, 8 and 10). Finally, even the disappearance of grammatical constructions may pro-

ceed at a different pace in different lexical or grammatical environments (Joseph

1983; García 1999).

In view of the immense diversity of changes to be covered, it is understand-

able that attempts at extracting general principles have had a way of ending up

vague and somewhat tautological, or else can only handle part of the phenomena to

be accounted for. This being said, it is still useful to consider the generalisations that

have been made more closely, for despite their problems they offer a number of im-

portant insights into the nature of diffusional change. Two different – though not


incompatible – perspectives have been taken on diffusion, one inspired by the no-

tions of reanalysis and actualisation, the other inspired by the study of lexical diffu-

sion in sound change. In Sections 1.1 and 1.2, each of these approaches is discussed

separately. Section 1.3 presents a case-study on morphological productivity which,

though not dealing directly with diffusion, builds on the insights from the previous

two sections and as such takes some steps towards integrating the different ap-

proaches.

1.1. Reanalysis and actualisation

Changes in syntax are often believed to start with reanalysis, causing a specific sur-

face sequence to receive a new underlying semantic or syntactic representation

(Timberlake 1977; Langacker 1977). Because the surface sequence that is the locus

of reanalysis does not change, reanalysis itself is considered invisible, but its conse-

quences are not. These make their appearance the moment the new representation

inspires innovative uses that were impossible under the old analysis. What happens,

typically, is that some element of the original surface sequence suddenly expands its

range of use. For example, the English word fun, clearly a noun in (2a), can be re-

analysed as an adjective in sequences like (2b), where its grammatical status as a

noun or adjective makes no difference to the surface form of the sentence. Once fun

is thought of as an adjective, it may also occur in adjectival uses that are not superfi-

cially shared by nouns, as in (2c-d) (Denison 2001).

(2) a. that’s when the real fun of being a cowboy begins. (CB)

b. It should be fun trying to complete this one. (CB)

c. if one twin was having a really fun time, the other would immediately

trip and break its leg. (CB)

d. Herp Herp Hooray! The Funnest Reptile & Amphibian site! (Google,

30 October 2007)

The actualisation phase of reanalysis, during which reanalysis verifiably manifests

itself, entails a kind of diffusion, as an item spreads from an old environment, the

locus of reanalysis, to new environments. The diffusional character of actualisation

is especially apparent if some new uses are slower to be implemented than others. In

the case of fun, for instance, the superlative form funnest in (2d) is still less accept-

able than the combination with adjectival modifier really in (2c), indicating that the

consequences of the recategorisation of fun are only worked out gradually.


The question that is of most interest here is what principles guide the diffu-

sional changes accompanying actualisation.9 A first generalisation, proposed ini-

tially by Timberlake (1977) and taken over by Harris & Campbell (1995), is that

changes spread from unmarked environments to marked environments (see also

Warner 1982 on ‘salience’). It is clear that both Timberlake and Harris & Campbell

mean for markedness to be defined with specific reference to the changes in ques-

tion. There is a certain justification in defining markedness in this way, as attempts

to define it independently seem unworkable or at least inapplicable to diffusional

change.10 However, a change-dependent understanding of markedness also has the

obvious disadvantage of inviting circular reasoning, as it might tempt us to define

markedness by observing the order of changes and then explain the order of changes

by invoking markedness. Although a more careful formulation may thus be called

for, one must admit that for certain historical changes, the markedness principle has

some intuitive appeal as an explanation for the sequence of events observed histori-

cally. What we find in such cases is that the diffusion of a new pattern from one en-

vironment to another is at least in part determined by the linguistic features of the

environments and of the spreading pattern itself, in such a way that each diffusional

step implements a minimal change to the existing system and maximally avoids

clashes between the new pattern and the expectations a given environment raises or

the functional constraints it imposes. A few examples may suffice to clarify this

point.

Naro (1981), investigates the loss of verbal agreement for the third person

plural inflections in Brazilian Portuguese. He finds that the change is favoured in

environments where it is least noticeable. That is, the chance of agreement loss is

increased by low ‘phonic salience’ in the singular/plural distinction. For example,

phonic salience is enhanced when the syllable containing the plural marking is

stressed rather than unstressed, or when it undergoes a change in vowel quality con-

comitant to plural marking. Naro concludes that “The change thus sets in at the zero

point of surface differentiation between the old and the new systems, and spreads to

other points along the path of least surface differentiation” (1981: 97). In terms of

markedness, we could say that singular/plural distinctions with high phonic salience

are marked with respect to a change that undermines the distinction and therefore

maximally resist the change.

In another case-study, Fanego (2004a) examines the spread of gerund phrases

with verbal syntax at the expense of gerund phrases with purely nominal syntax (a

9 The related question of how reanalysis itself works is addressed in De Smet (2007). 10 See Fanego (2004: 36) on Andersen (2001).


change to be discussed at length in Chapter 8). She finds that verbal syntax is

adopted more slowly in those environments that show more explicit nominal mark-

ing in the form of a definite article or a possessive pronoun preceding the gerund

phrase.11 Thus, using prepositional patient-marking by of as a diagnostic for nominal

syntax and its absence as a feature of verbal syntax, we may say that the replacement

of the construction type in (3a) by that in (3b) preceded the replacement of the type

in (4a) by that in (4b), which in turn preceded the replacement of the type in (5a) by

that in (5b). It may further be noted that so far only the first process of replacement

(in (3)) has been completed, while the second (in (4)) appears to have stranded mid-

way and the third (in (5)) has even been reversed.

(3) a. Most diseases are healed either by letting of bloud, by taking vp of

vaines, by purgation, or else by cauterisation. (1565, OED)

b. I haue lost my edifice, by mistaking the place, where I erected it.

(1598, OED)

(4) a. We know the Aspworme by his stinging of vs. (1587, OED)

b. As my purpose was also to say something to you concerning my taking

the Seals from the Chancellor. (1667, OED)

(5) a. Either of these ways Satan makes use of for the befooling of men.

(1677, OED)

b. I went to this Newcastle in Staffordshire to see the makeing the fine

tea-potts cups and saucers of the fine red earth. (c1702, OED)

Fanego (2004a: 38) thus concludes that “the more noun-phrase-like a sequence was,

the slower was it to acquire verbal traits”, and speaks in this respect of a ‘hierarchy

of relative nominality’. In terms of markedness, a change that implements a verbal

structure is most strongly resisted in those environments that show clear non-verbal

(i.e. nominal) traits and which are therefore marked with respect to the change.

As a final example, García (1999) discusses the loss of the Italian enclitic

adposition -co used in combinations with pronouns in meco (‘with me’), teco (‘with

you’) and seco (‘with 3rd person self’). These were replaced by the propositional

constructions con me, con te and con se, albeit not at the same time, teco being the

first combination to disappear, followed by meco and later still by seco. García ar-

gues that the early disappearance of teco can be explained by a clash in pragmatic

values in teco which arose as on the one hand the enclitic construction attached

11 The reason why possessive determiners are less explicitly nominal than definite determiners is that the former semantically correspond to clausal subjects.


greater salience to the pronominal referent, while on the other the tu-paradigm, in-

cluding the form te, was reserved for social inferiors (voi being neutral and Ella po-

lite/deferential). Conversely, the prepositional variant with con reduces the referen-

tial prominence of the pronoun and is therefore compatible with a pronoun marking

social inferiority. In this case, a change that reduces pronouns’ referential promi-

nence is less marked with a pronoun that pragmatically already signals lower promi-

nence. The change-dependent understanding of markedness in this case is particu-

larly striking, since te is unmarked with respect to the change, but probably marked

as a second person pronoun (as opposed to voi and Ella).

According to a second generalisation, proposed by Harris & Campbell

(1995), diffusional changes also create natural classes. This means that the environ-

ments over which a new pattern extends naturally belong together. Thus, writing on

rule-extension, Harris & Campbell (1995: 102) formulate this principle negatively

by stating that “The kind of extension which seems not to exist, but which is logi-

cally possible, would generalize to categories that fail to form a natural class with

the categories in which the rule applied before extension.” In other words, no exten-

sion creates unnatural classes. What exactly constitutes a natural class is not ex-

plained by Harris & Campbell, whose claim thus leaves more room for interpreta-

tion than is probably desirable. However, like the markedness principle, the natural-

ness principle has some explanatory appeal as a constraint on certain diffusional

changes. For example, the principle seems to account for cases of analogical level-

ling, which essentially remove exceptions and thus restore natural classes. The level-

ling out of the Middle English plural form shoen to shoes, for instance, serves to

establish the natural class of nouns as the class of items relevant to the application of

the plural -s-suffix. In grammaticalisation, too, an item’s distribution before and

after change is typically delimited by what seem to be natural classes. For example,

when the French noun pas meaning ‘step’ develops into a clausal negator, it shifts

from occurring exclusively with verbs of motion (one natural class, determined by

the lexical semantics of pas) to combining with any verb (another natural class, de-

termined by the new grammatical function of pas) (Hopper & Traugott 1993: 58).

There seems to be a straightforward relationship between the naturalness

principle and the markedness principle. Indeed, if we take the markedness principle

(somewhat narrowly) to predict that changes spread to environments that are maxi-

mally similar to the environment where the change started, and if we assume that

natural categories are based on similarity, the two principles coincide. Harris &

Campbell (1995) themselves seem to treat the naturalness principle simply as a re-

formulation of the markedness principle. In actual fact, they are not the same thing.


Markedness is defined with respect to a kind of change, whereas naturalness is de-

fined with respect to the source and target environments of diffusion. In practice, the

two do not necessarily coincide. For example, at one point the spread of the Italian

preposition con at the expense of enclitic -co, discussed above, clearly cuts across

the syntactic behaviour of the pronominal forms te and se, yet it is hard to see what

natural class would include te but exclude se. As argued above, however, the mark-

edness principle can, in this case, explain the discrepancy.

The example of Italian enclitic -co at once reveals an important weakness of

the naturalness principle. Cases of analogical levelling nearly always leave some of

the exceptions intact. In English, for instance, weak past tense forms have been en-

croaching upon the less regular strong forms ever since Old English times. We could

therefore rephrase the problem in more general terms, and say that the naturalness

principle may tell us about the expected outcome of diffusion but reveals little about

its intermediate stages, as it fails to tell us why exceptions are not removed all at

once. Harris & Campbell (1995) recognise the difficulty posed by partial restora-

tions of regularity but dismiss it as unproblematic:

In situations such as this, extension has not yet established a natural class as

the domain in which the rules apply; nevertheless, after the extension the

natural class is more complete than it was before. (1995: 106)

Despite Harris & Campbell’s optimism, however, it cannot be denied that the natu-

ralness principles – in its present formulation – loses some of its explanatory poten-

tial.

The same problem also surfaces in a subtler form. Consider Brems’ (2003)

analysis of the development of (a) bunch (of) from a lexical noun to a quantifier.

Brems finds that in its original lexical use the items with which bunch co-occurs are

strictly determined by its lexical semantics: bunch denotes “a somewhat unruly clus-

ter of things fastened at one point” and is accordingly found with nouns such as

flowers, bananas, grapes, carrots or herbs, as in (6a) (2003: 294), which look like

forming a natural class. In its grammaticalised use, however, bunch – or rather, a

bunch of – functions as a quantifier and occurs with almost any kind of noun, includ-

ing inanimate count nouns, inanimate noncount nouns, and animate count nouns, as

in (6b-d). These, too, would make a natural class, were it not for the fact that, when

combining with inanimate noncount nouns and animate count nouns, bunch also

carries a negative semantic prosody, meaning that it almost exclusively combines

with nouns bearing negative connotations (witness (6c-d)) (Brems 2003: 298).


(6) a. The fox, unable to reach a bunch of grapes that hangs too high, decides

that they were sour anyway. (Brems 2003: 295)

b. There’s now a whole bunch of studies from different cities that show

the same thing. (Brems 2003: 297)

c. Trouble was, the funds were able to neatly hide all but the most con-

spicuous of their charges in a bunch of legalese. (Brems 2003: 297)

d. Who said Americans were a jingoistic bunch of rednecks who know or

care nothing about what happens beyond their shores? (Brems 2003:

297)

These restrictions are not unmotivated: meanings have simply moved around in the

construction, from a slightly negative connotation on bunch to a collocational re-

striction on the nouns it can quantify (see the relocation of meanings in predicate-

complement constructions discussed in Chapter 3). Yet from the point of view of

diffusion, it is difficult to think of inanimate count nouns on the one hand, and nega-

tively connotated inanimate uncount and animate count nouns on the other, as form-

ing a natural class. If a bunch of ever becomes a general-purpose quantifier, like a

lot of, its environments of use may again constitute a natural class, but at the present

stage of its development this is not yet the case. Note though that the markedness

principle might help us out, since the environments imposing the strongest con-

straints (i.e. animate count nouns, inanimate uncount nouns) seem to be those most

dissimilar from the environment in which bunch was first reinterpreted as a quanti-

fier (i.e. the small set of inanimate count nouns containing flowers, bananas, etc.).

Alternatively, it could be argued that regularisation proceeds along lower-level regu-

larities, whereby the distribution of bunch splits off into a number of uses that may

fail collectively to form a natural class but that do cluster into internally coherent

sub-classes (which, arguably, are also natural classes).

What is clear from this discussion is that, despite their intuitive appeal, both

the naturalness principle and the markedness principle are still too elusive. On the

positive side, it is probably true and non-circular to say that diffusional change will

be sensitive to the constraints characteristic of the linguistic sub-system in which it

operates, in such a way as to be delayed if features of the spreading construction are

in conflict with the constraints and expectations associated with a specific environ-

ment (markedness principle). Likewise, it seems true that the environments over

which a construction diffuses will tend to form natural classes and it may even be

the case that the restoration of natural classes is sometimes a driving force of change


– even if, in view of the evidence, we also have to recognise that the establishment

of natural classes may be gradual and that distributional organisation may fall apart

into smaller natural sub-classes rather than large natural super-classes (naturalness

principle). However, the notions of markedness and naturalness as such remain ill-

defined, in part because they presuppose a theory of grammar which unfortunately

remains implicit.

Therefore, while the analyses in the coming chapters are indebted to the no-

tions of markedness and naturalness, no explicit reference is made to them. Instead, I

reformulate the principles here as a single principle of ‘sanction’, which links back

to the constructional framework developed in Chapter 3 (cf. Langacker 1987). It is

assumed that every utterance, including every historical innovation, is sanctioned,

both by the regularities that constitute grammar and by more general functional,

cognitive and pragmatic factors and constraints. Diffusional change derives its diffu-

sional character, first, from the fact that a new pattern is sanctioned to different de-

grees in different environments and, second, from the fact that the degree to which

patterns are sanctioned can change as diffusion proceeds and the grammar underly-

ing sanctioning changes. This is the main principle explaining why diffusion follows

the course it does.

I believe sanctioning can accommodate both the notions of markedness and

naturalness. On the one hand, sanctioning emulates the notion of naturalness, as it

can be defined with respect to the existing linguistic structures of the language in-

stead of natural classes whose existence is actually prospective or teleological. A

natural class is seen as the instantiation of a grammatical regularity and arises as

some provisional regularity is perceived and subsequently borne out. On the other,

as sanctioning is sensitive to multiple regularities as well as to constraints that cut

across those regularities (cf. Chapter 3), it can incorporate aspects of change ex-

plained by the principle of markedness but incompatible with naturalness. For ex-

ample, a pragmatic motivation cuts across grammatical regularity in the case of the

loss of the Italian -co-suffix.

At the same time, as a diachronic principle, sanctioning remains vague as

long as our understanding of synchronic grammar remains vague. Ultimately, it is

the model of synchronic grammar – with its multiple regularities and constraints –

that fills in how sanctioning operates. Indeed, how this sanctioning principle mani-

fests itself in actual changes will be illustrated at length in Section 2 below. To get a

firmer grip on sanctioning, however, it is first necessary to consider another perspec-

tive on diffusional change, whose insights stem from the study of lexical diffusion in

sound change.


1.2. Lexical diffusion in sound change and elsewhere

A second perspective on diffusional change is inspired by the study of processes of

lexical diffusion in sound change. Because lexical diffusion goes against the original

Neo-Grammarian assumptions of regularity and simultaneity in sound change, the

possibility of its occurrence has sparked off some debate in historical phonology and

it is partly thanks to this discussion that a number of new insights have been arrived

at concerning the mechanisms underlying diffusional change. It is now commonly

accepted that lexical diffusion does occur in sound change, even if the conditions

under which it occurs and its relation to regular Neo-Grammarian sound change are

still not fully agreed on (Wang 1969; Fidelholz 1975; Hooper 1976; Labov 1994:

ch.15; Phillips 1984, 2001; Bybee 2000, forthc.). Most interesting for our purposes

is that studies of lexical diffusion in sound change have also thrown light on cases of

lexical diffusion outside phonology, involving the spread of morphosyntactic pat-

terns (Ogura 1993; Ogura & Wang 1996).

Lexical diffusion in sound change implies that sound changes are imple-

mented in a lexically gradual fashion, affecting some words earlier than others. By

way of example, Labov (1994: 429-37) reviews the literature on the tensing of short

/æ/ from [æ] to [e:ə] in an intricately conditioned set of environments (e.g. the

change is conditioned in words such as ham but not in words such as hammer) in a

number of varieties of English. He finds that when all phonetic and grammatical

determinants of change are factored out, evidence remains indicating that the change

is in part lexically gradual. For instance, in Philadelphia, where the change has run

its full course, mad, bad, and glad are consistently pronounced tense but sad is not.

This state of affairs presents “a clear case of lexical diffusion, arrested in mid-career

at some point in the past” (Labov 1994: 431).

In the realm of morphosyntax, Ogura (1993), using data from Ellegård

(1953), discusses the diffusion of do-support in English in environments as those

italicised in (7), where present-day speakers of English would preferentially use the

auxiliary do (do you think for (7a); do not trifle for (7b)).

(7) a. Would he like it, think you? (1849, Ogura 1993: 71)

b. Trifle not; answer me. (1835, Ogura 1993: 71)

Ogura shows that the spread of the pattern is sensitive to syntactic environments

(e.g. negative questions are quicker to adopt do-support than affirmative questions,


which in turn are quicker than negative declaratives, etc.) and within those environ-

ments is sensitive to lexical items. She demonstrates, for example, that a group of

verbs including say, think, mean and do are about a century behind on other verbs in

adopting the shift to using do-support.

When it comes to explanation, it is to be noted that not in all changes involv-

ing lexical diffusion have the lexical irregularities been accounted for (the tensing of

short /æ/, for instance, is a notable problem-case), so our understanding of lexical

diffusion can be partial at best. Where an explanation has been found, however, a

major determining factor appears to be word frequency, with highly frequent words

being either behind or ahead of less frequent words with respect to the change. Thus,

in diffusional sound-change highly frequent words tend to lead the change if the

change is physiologically-induced, as in assimilation or reduction (Phillips 2001:

125-6). For example, in the case of schwa-reduction, more frequent words such as

memory show further-advanced reduction than less frequent words like armory –

indeed the former is generally pronounced [′memri], the latter [′A:məri] (Hooper

1976). By contrast, low-frequency words tend to lead in non-physiologically-

induced sound changes, such as typological changes that reorder the phonological

system of a language (Phillips 2001: 127). For example, Phillips (1984) discusses

the merger of /ø(:)/ with /e(:)/ in Middle English, which apart from being condi-

tioned grammatically is also conditioned lexically, with infrequent words running

ahead of frequent ones.

The frequency effects in physiologically-induced sound changes appear to be

well-understood and are commonly related to the same mechanism that produces the

well-know tendency for more frequent words to be shorter (Zipf 1949), namely the

fact that high-frequency sequences get automated so that they are stored, accessed

and produced holistically, with a lower degree of awareness of and attention to indi-

vidual segments (Phillips 1984, 2001; Bybee forthc.). On the other hand, changes

that proceed from less frequent to more frequent words are linked – though rather

more tentatively – to the assumption that low-frequency words are more sensitive to

changes that involve imperfect learning (Hooper 1976) or that act on underlying

forms and therefore require access to segments (Phillips 1984, 2001).

The diffusional effects observed in morphosyntactic changes are likewise

sensitive to word frequency (Ogura 1993; Ogura & Wang 1996). Clearly, word fre-

quency nicely accounts for the lag of highly frequent lexical items such as do, say

and think in adopting do-support, as discussed above (Ogura 1993). Similarly, Ogura

& Wang (1996: 131), studying the replacement of third person singular -th by -s in

the present indicative in English, find that among verbs not ending in a sibilant, -s


first appeared with the highly frequent verbs, but was then adopted by low-

frequency verbs at a higher rate of change, so that eventually it was also high-

frequency verbs such as do, have and say that were last to abandon the old -th-

ending.

Krug (2003: 20-2) goes to some length to explain the findings in Ogura &

Wang (1996), arguing that, phonologically, the shift from -th to -s is a ‘leniting’ (i.e.

physiologically-induced) change, /θ/ and /ð/ being more marked than /s/ and /z/, and

therefore affects a few high-frequency items first, yet once the change starts spread-

ing, it becomes an analogical change and is quicker to affect low-frequency items

than high-frequency items, because the latter resist analogical levelling. The first

part of Krug’s explanation echoes the accounts of lexical diffusion in physiologi-

cally-induced sound change. The second goes back on other findings indicating that

frequent forms are conservative with respect to non-reductive changes. As regards

the latter tendency, the best-known example is the levelling in the past tense forms

of English irregular verbs, where it is found that low-frequency irregular verbs may

show variation between an irregular form and a regularised form (e.g. weep –

wept/weeped), whereas high-frequency irregular verbs tend to allow no such varia-

tion, retaining their irregular past tense form (e.g. keep – kept/*keeped) (Hooper

1976; Bybee & Slobin 1982; Bybee 1985). The reason for this is once more to be

sought in holistic storage and automation of high-frequency items; that is, “fre-

quency strengthens the memory representations of words or phrases, making them

easier to access whole and thus less likely to be subject to analogical reformation”

(Bybee 2006: 715; see also Warner 1993). If we assume competition between an

irregular and a regularised form in cognitive retrieval, it may be further assumed that

repeated use of the irregular form causes it to get increasingly entrenched and

thereby makes it easier to activate (Langacker 1987). In a spreading-activation

model of language production (e.g. McClelland & Rumelhart 1981; Berg 1998;

MacWhinney 2000; Schlüter 2005; Hudson 2007), this adds further to frequent

forms’ resistance to change, as greater ease of activation will make it easier for the

irregular form to win out from its regularised competitor.12

12 It is to be noted that automation and ease of activation are not the same thing, though both are side-effects of high frequency and entrenchment and both may give rise to conservative effects. For example, automation, but not ease of activation, is the process that accounts for the fact that once we have learned how to drive a car we lose conscious access to the separate segments of certain action sequences and may in the end even find that we are unable to explain them to someone else first driving a car. By contrast, ease of activation, but not automation, is responsible for the fact that when we are confronted with a different dialect of our own language we


For our purposes, what discussions of lexical diffusion give us is a mecha-

nism that accounts for the exceptions and their different degrees of persistence to

analogical levelling. As a conservative force in morphosyntactic change, frequency

and entrenchment help us explain some of the problems encountered by the natural-

ness principle discussed earlier in this chapter, as the gradual levelling out of mor-

phological exceptions can be seen to arise from the interaction between the regular-

ising force underlying the naturalness principle and the resistance to change that

results from automation. Importantly, this account seriously weakens one of the

classical objections against analogy as an explanatory principle of change, namely

the fact that analogy always leaves exceptions (Lass 1980, 1998). One class of ex-

ceptions can now be accounted for.

Conservative effects can be understood in terms of sanctioning. As a sanc-

tioning factor, frequency is indirectly responsible for a construction’s sanctioning

‘strength’. In the examples of competition between an old and a new form illustrated

above, both the old and new form are sanctioned, the former by its high degree of

entrenchment and direct accessibility, the latter by its regularity (i.e. conformity to

other forms or an abstract schema). The new form replaces the old form when the

balance between sanctioning factors tips in favour of the new form. The diffusional

character of analogical levelling derives from the fact that different environments

resist innovations to different degrees.

1.3. The two-sided effect of frequency: A case-study

What the studies of lexical diffusion above pay less attention to is that frequency and

entrenchment also interact with the sanctioning principle at the schematic end of the

constructional hierarchy. Apart from being a conservative force at the level of spe-

cific instances, frequency also underlies the productivity of abstract patterns, and

thus, at the level of more schematic representations, it constitutes one of the forces

that push analogical levelling forward. Taking again English past tense formation as

an example, it is no coincidence that weak past tense formation, which is the most

frequent pattern of past tense formation, is the pattern most enthusiastically substi-

tuting for other ways of expressing the past tense. In addition, the weak pattern is

not the only one to have been expanding historically. Some vowel alternation pat-

may, as activation becomes easier with prolonged exposure, at some point take over features of the other dialect. Ease of activation effects could perhaps also be inter-preted in terms of priming, although this could only work for high-frequency items, as priming is a short-term effect (cf. Szmrecsanyi 2005).


terns, too, have extended their range of application, as happened for instance in the

recent emergence of snuck as a past tense form for the originally weak verb sneak

(Hogg 1988). The reason must be that speakers still extract weakly productive sche-

mas from existing alternations such as strike/struck or stick/stuck and that it becomes

increasingly easy for speakers to do this if the alternation is associated with other

recurrent features, such as the initial /s/ and final /k/ of the stem. In this case, then, it

is the repetition of feature combinations that allows speakers to extract a schema that

can subsequently be extended to new past tense formations (Bybee & Slobin 1982).

To further demonstrate this two-sided effect of frequency and entrenchment,

the rest of this section is devoted to an analysis of morphological productivity and

transparency in the English derivational suffix -ness, which is used to form abstract

nouns from adjectives, as in darkness, greatness, laziness, etc. (reported more fully

in De Smet in prep.). Although the case-study does not directly address diffusion, it

further substantiates the double role of frequency. On the basis of the foregoing dis-

cussion, it may be hypothesised that high type and token frequency of an abstract

derivational schema will lead to high productivity for the schema, while high token

frequency of a particular instantiation of the schema may bring about automation

and morphological opacity of that instantiation. Since entrenchment is a property of

a cognitive structure in the mind of an individual language user, one way to test the

hypotheses is to examine the behaviour of different language users. More specifi-

cally, it is possible to estimate the degree to which a schematic pattern is entrenched

in the mind of a particular language user and then link that estimate to the productiv-

ity of some more specific (less schematic) sub-pattern in the language user’s output.

Over a large population of language users, it can then be investigated how en-

trenchment of the schematic pattern correlates with productivity on the level of its

instances.

To do this I have compiled a corpus of 348 present-day English novels drawn

from the BNC. The novels are all more or less equal in size (between 31,000 and

48,000 words, totalling about 13.8 million words for the whole corpus) and each

represent the use of one individual in one specific text genre. Even a casual glance at

idiolectal variation in the corpus tells us that language users differ enormously in the

extent to which they rely on -ness-derivation. Looking at type frequencies across

individual users, an author at one extreme is found to produce 65 different deriva-

tions with -ness in a 37,000-word-text, whereas at the other extreme an author pro-

duces only one derivation with -ness in a text of almost 41,000 words. We may as-

sume that for the former author, ness-derivation is easy to activate and well-

entrenched, while for the latter it is harder to activate and less firmly entrenched.


Using this kind of information, language users can therefore be given an entrench-

ment score for -ness-derivation by counting the type frequency of -ness-derivations

in their output and weighing it by text size.13 If language users are then ranked ac-

cording to their entrenchment score, the corpus can be subdivided into segments

each containing language users with comparable scores. These segments can then be

examined with respect to the properties we wish to correlate to the degree of en-

trenchment for -ness-derivation. For this case-study, use was made of a division in

five equally-sized segments, I, II, III, IV and V, whereby the authors with weakly-

entrenched -ness-derivation are put in segment I and so on.

There is a catch, of course. If the estimate of entrenchment for a pattern is

based on a language user’s entire output of the pattern, it will correlate with the

same user’s productivity in a purely circular fashion. For this reason, the specific

patterns whose relation to entrenchment I wished to test could not be taken along in

the calculation of entrenchment scores. The patterns chosen to be investigated were

two specific derivation types, happiness and business, as well as all happax le-

gomena in the corpus. In raw figures, this meant the exclusion of 449 tokens of hap-

piness, 3497 tokens for business and 440 single occurrences of items such as age-

lessness, evenness, succinctness, upfrontness, zestfulness, and so on. In practice, the

corpus has been segmented for entrenchment scores three times, once based on

speaker’s entire output excepting business, once based on their output excepting

business and happiness, and once based on their output excepting business and all

happax legomena. The first segmentation could be used in the analysis of business,

the second for that of happiness and the third for that of the happax legomena. That

business has never been included in the calculations of entrenchment scores is justi-

fied by the results discussed below.

The next step, then, is to compare the behaviour of language users in different

segments of the corpus with respect to their use of the patterns to be investigated.

This is done in Figures 4.1 and 4.2, with Figure 4.1 giving the relative frequency

(per 10,000 words) of the patterns investigated for each segment and Figure 4.2 giv-

13 The formula used in this case is simple if somewhat inaccurate: type frequency is divided by text size. The inaccuracy of this entrenchment measure lies in the fact that the incidence of new types decreases as text length increases (texts get saturated for new types, so to speak). As far as this case-study is concerned, however, it has been assumed that the effect of text length is compensated for by the fact that all texts are roughly equal in size, that the texts are relatively short, and that -ness-derivation is extremely productive in English, so that even the most prolific -ness-users in the corpus are still very far from exhausting the supply of -ness-derivations attested even in the limited data set examined. However, for a more sophisticated approach to measuring entrenchment, see De Smet (in prep.).


ing the percentage of authors per segment that is found to use the patterns investi-

gated at least once.14 The results are straightforward and conform to the hypotheses

formulated above. First, there is a positive correlation between the degree to which -

ness-derivation is entrenched and the incidence of happax legomena. In general, this

means that the higher the (type) frequency of -ness-derivation in a given idiolect (or,

presumably, any other lect), the more likely -ness-derivation is to spawn off new or

unique instances. Second, the frequency of the individual item happiness, too, is

sensitive to the productivity of -ness-derivation. This suggests that although happi-

ness is itself relatively frequent, it is still morphologically transparent, as it is easier

to activate if the morphological pattern on which it is based is readily accessible as

well. If it is still surprising that a highly frequent item such as happiness appears to

be equally sensitive to entrenchment as the set of happax legomena, it must be borne

in mind that automation appears to get considerably stronger if the complex word is

more frequent than its morphological base (Hay 2006), which is definitely not the

case for happiness. Third, business is not sensitive to the degree of entrenchment of

the morphological pattern on which it is formed – or if it is, the correlation is, rather

unaccountably, inverse. This is to be expected, as business is highly frequent (about

eight times as frequent as happiness) and more frequent than its base (busy is about

ten times less frequent than business), and as it already shows evidence of formal

reduction alienating it from its historical source.

The case-study on -ness-derivation is of interest because it links differences in

frequency to differences in productivity. The link is established on the basis of syn-

chronic idiolectal variation, but it is not implausible to assume that the same link is

also valid diachronically. Especially interesting in diachronic terms are the findings

regarding happax legomena. Although it is to be granted that a happax legomenon

and an historical innovation are not exactly the same thing, the fact that higher en-

trenchment leads to a higher incidence of happax legomena does suggest that higher

entrenchment may also make it easier for a construction to extend into new envi-

ronments (cf. Krug’s 2000 gravitational model). In turn, this translates into a mecha-

nism that accounts for historical snowballing effects (Aitchison 1991; Ogura &

Wang 1996). In processes of analogical levelling, for example, each batch of excep-

tions that is removed can be assumed to raise the pressure on the exceptions that

14 Notice that for the results to be comparable, it is crucial that the token frequencies of the specific patterns investigated are roughly equal. This holds for happiness and the set of happax legomena, but for business the counts in Figures 4.1 and 4.2 are based on a random 12.5%-sample, bringing its token frequency in line with that of the other two patterns.

0

1

2

3

4

5

6

7

I II III IV V

Rel. freq. 'business'

Rel. freq. 'happiness'

Rel. freq. happaxlegomena

Figure 4.1. Relative frequencies (per 10,000 words) of business, happiness and happax

legomena according to five entrenchment segments in the BNC-novel corpus.

0

20

40

60

80

100

I II III IV V

% 'business' users

% 'happiness' users

% happax legomenausers

Figure 4.2. Percentage of users per entrenchment segment that show at least one instance

of business, happiness or one happax legomenon in the BNC-novel corpus.


remain, until there are so few exceptions left that the removal of an odd irregularity

no longer makes any great difference to the overall productivity of the spreading

pattern. This explains the S-shaped curve alleged to be typical of processes of re-

placement and accounts for the fact that, in changes involving competition between

variants, the rate of turnover seems to be higher in environments that are affected

later (Aitchison 1991; Ogura 1993). Once again, in terms of sanctioning, the degree

to which a given use is sanctioned is in part determined by frequency. The use of

highly frequent constructions – whether schematic or specific – receive stronger

sanction than the use of low-frequency patterns.

2. Diffusional change in the system of complementation

The discussion so far has largely disregarded the fact that diffusional changes fall

into a number of subtypes. These subtypes are defined by whether the gradualness of

diffusional changes is conditioned lexically or syntactically; whether or not they are

driven by functional reinterpretations of the spreading pattern; whether or not they

involve semantics; and whether they are replacive or (partly) non-replacive. Thus,

diffusional change in the system of complementation is lexically gradual, as the dif-

fusing pattern spreads from one lexical environment to another (rather than from one

major syntactic function to another). This means at once that it is not driven by func-

tional reinterpretations of the spreading pattern, as further appears from the fact that

it mostly does not result in clearly homonymous constructions or synchronic layer-

ing (as in the case of grammaticalisation).15 Further, as is to be expected from the

discussion in Chapter 3, diffusional change in the system of complementation is

sensitive to semantics, meaning being involved far more directly than in some other

diffusional changes, such as sound change.16 Finally, diffusional change in the sys-

15 Strictly speaking, as a predicate-complement construction recruits new members to fill the predicate slot (which is another way of saying that the complement dif-fuses), the predicate-complement construction also changes, for assuming that a construction specifies the lexical elements it can accommodate, a change in combi-natorial potential is a constructional change. However, the spreading pattern as such does not change and in retrospect what changes have taken place are invisible, in that they are not synchronically reflected in a layering of different functions of the element that diffused (as opposed to grammaticalisation; see Hopper 1991). 16 In sound change, especially, semantic effects are fairly inconspicuous and can at best account for occasional exceptions to otherwise regular changes (e.g. the contro-versy on homonymy avoidance; see Labov 1994: ch.20). In the diffusion of new complement types, by contrast, meaning leaves its mark in several ways. Most obvi-ously, semantically related predicates tend to behave alike in their syntactic pattern-ing and therefore can influence one another in the course of diffusion (see Section


tem of complementation is not necessarily replacive, which means that it need not

involve the systematic substitution of an older variant by a new variant.17

If anything, this typological classification underscores the difficulty of formu-

lating general principles of diffusion. The purpose of this section, therefore, is not to

further pursue this general goal, but to work out in more detail how the principles of

2.2.1). On a more abstract level, semantics also set limits to which predicates have a chance of getting involved in the change, as a great number of predicates are more or less excluded on semantic grounds from taking complement clauses at all – for instance, because they combine with material objects only (e.g. drink, borrow, fling, pierce etc.). Finally, it is worth pointing out that, from a semantic point of view, the diffusion of a new complement-type may in fact be an onomasiological change that gives speakers access to a new realm of semantic possibilities. This is most evident when in the course of change predicates begin to be complemented that never or hardly took clausal complements before (for instance, in the case of ing-complementation, this is true for avoid, enjoy, risk, or stop; see Chapter 9). In the great majority of cases, the introduction of new meaning potential is not particularly dramatic or spectacular (presumably, new meanings are simply variations on mean-ings that could already be expressed through a score of periphrastic alternatives), but is nonetheless hard to deny or ignore, and as such underscores the difference be-tween diffusional change in complementation and diffusional sound change. 17 This also means that diffusional change in the system of complementation may not have a predefinable end-point. The natural end-point of replacive changes is reached when the change is evidenced in all susceptible types and in all tokens of each type. For English such ‘replacive’ changes include the spread of the third per-son singular s-ending (Ogura & Wang 1996; Nevalainen & Raumolin-Brunberg 2003) and the spread of do-support (Ogura 1993). In the diffusion of a complement type, however, no end-point can be defined. We do not know how many predicates are in fact susceptible to complementation by a given complement. Even for a given predicate we cannot sensibly think of a change as having run its full course, as there is no sense in postulating a point at which that predicate makes full use of its new constructional potential. The fact that new complements do not necessarily substi-tute for a single older variant is the main reason for this, but there are other reasons as well, such as the variable degree of susceptibility of different predicates to com-plementation, which is a consequence of the involvement of semantics, and the fact that change is partly onomasiological and thereby open-ended. Because the progres-sion of change cannot be measured against a theoretical point of completion, the spread of a new complement is more difficult for the linguist to interpret quantita-tively. This methodological problem apart, it is not unreasonable to assume that the difference matters to speakers as well. By hypothesis, the lack of systematic substi-tutability might cause speakers to be less conscious of making a choice between variants. Further, in the absence of an obvious older variant calling out to be re-placed, the syntactic innovation may less clearly suggest itself to the speaker (on whatever level of linguistic awareness), and as a result, the progression of change may be to a higher degree dependent on speakers’ linguistic creativity. At the same time, once a complement type is introduced with a given predicate it will be less sensitive to competition from older variants and may gain ground more quickly.


diffusion apply specifically to diffusional change in the system of complementation

(introducing along the way the mechanisms that figure most prominently in the fol-

lowing chapters). Each of the mechanisms to be discussed in this section relates to

the general principle of sanctioning and can be seen as one of its manifestations.

In particular, two aspects of sanctioning are addressed in more detail: its rela-

tion to frequency (uses are sanctioned on the basis of the frequency of the pattern

they instantiate) and its relation to grammatical regularities (uses are sanctioned on

the basis of their instantiating some grammatical pattern in the first place). Thus,

Section 2.1 goes back to the conservative effects of frequency, applying the notion

of entrenchment to complementation in the form of a blocking mechanism. Section

2.2 turns to grammatical regularities, which manifest themselves diachronically in

the form of analogical extension. Analogy itself is analysed into sub-types, the two

most important of which are semantic analogy and paradigmatic analogy. Together,

blocking, semantic analogy and paradigmatic analogy are the major recurrent

mechanisms in diffusional changes in the system of complementation.

It is to be recognised that a complete general model of diffusional change in

the system of complementation is unattainable, as diffusion is also sensitive to fac-

tors that are incidental to the specific spreading construction, as well as to other his-

torical developments, taking place independently of the diffusional change. For that

reason, Section 2.3 briefly addresses a number of factors that can further complicate

diffusional changes in the system of complementation.

2.1. Blocking

Although diffusional change in complementation is not necessarily replacive, fre-

quency effects do seem to occur that are reminiscent of the conservative effects of

high frequency in other changes involving lexical diffusion (see Section 1.2). It is

unlikely that the simple token frequencies of predicates would have any effect on the

diffusion of complement types. The conservative behaviour that correlates with high

frequency affects frequent combinations, not one element of the combination (e.g. in

the competition between -s and -th-endings in Early Modern English, it is not the

frequency of the stem do- but that of doth that slows down the emergence of does).

Following this reasoning, the only frequency effect conceivable is that certain predi-

cate-complement constructions are sufficiently frequent as a whole to be stored in-

dependently and resist change when a new complement-type emerges – assuming of

course that the new complement happens to compete with an old one. This phe-

nomenon will be referred to here as ‘blocking’.


The necessary conditions for blocking are certainly met in the system of

complementation. The discussion in Chapter 3 above shows that certain specific

predicate-complement constructions can undergo independent semantic changes,

indicating that they must be cognitively represented in such a way as to allow some

degree of independent access. This suggests that to some extent they form separately

stored units. Further evidence to this effect comes from examples of phonological

reduction. In Present-Day English, the to-infinitive, in particular, can be seen to un-

dergo partial univerbation with the preceding matrix verb if the combination of ma-

trix verb and to-infinitive is sufficiently common, as is illustrated in the examples in

(8). Changes of this kind of course strongly suggest that these sequences of matrix

verb and to-infinitives are stored and regarded as single units, no longer analysed

into their constituent parts.

(8) a. we don’t wanna be no symbol (CB)

b. don’t mean I’m tryna diss you, just tryna tell you the facts! (Google, 25

October 2007)

c. Eh…hiya! Me, myself, and I, huh? Well…I like a guy who’s…nah j/k.

I liketa talk to people! come talk to me!!:) I liketa party…liketa

race…liketa sing and dance…liketa play the piano and the drums.

um…i listen to almost every kina music out there… (Google, 6 April

2003)

Looking at historical changes, then, some genuine instances of blocking do

seem to exist in the area of complementation. An interesting case is discussed by

Harris & Campbell (1995: 104-6). In Finnish, non-finite complements of non-

motion verbs are usually formed on the basis of the so-called ‘first infinitive’, as in

yrittää tehdä ‘try to do’. There is a class of exceptions, however, consisting of verbs

that combine with the ‘third infinitive’, for instance pyrkiä tekemään ‘strive to do’

(instead of *pyrkiä tehdä).18 Normally, third infinitives combine with verbs of mo-

tion to denote a goal, as in mennä tekemään ‘go to do’. As it turns out, the excep-

tional non-motion verbs with third-infinitive complements can all be traced back

historically to motion verbs – pyrkiä ‘strive’, for instance, originally meant ‘hurry to

do’. What must have happened, then, is that a number of motion verbs, including

18 The Finnish first infinitive is formed by the addition of the suffix -ta / -tä to the stem of the verb; the third infinitive by the addition of the participle suffix -ma / -mä in combination with the illative case ending Vn. In the case of tehdä and tekemään, the first and third infinitive forms are the result of further stem changes to the stem tek- that are conditioned by the suffixes added.


pyrkiä, in combination with the third infinitive underwent semantic change and lost

their motion sense. The constructions that resulted were anomalous from the per-

spective of synchronic grammar but withstood regularisation, at least in a number of

dialects and with a number of verbs. Both the semantic developments and the subse-

quent resistance to regularisation can only be explained as the result of entrenchment

and autonomous storage, and as such appear to constitute an excellent example of

blocking in the system of complementation.

De Smet & Cuyckens (2007: 198-206) discuss another case of blocking, in

which the diffusion of ing-complements is delayed in environments where the to-

infinitive is strongly entrenched. Thus, in the course of the Modern English period,

emotive verbs such as like, love and hate began increasingly to combine with ing-

complements. Up to a certain point, these ing-complements were semantically

equivalent to to-infinitives – or equivalent enough, in any case, to compete with the

to-infinitive (cf. Chapter 3). That the semantic difference is sometimes negligible

appears from examples such as the following, where ing-complements and (to-

)infinitives appear side-by-side. As the examples show, the interchangeability of the

two complement types is especially evident when the emotive verb is used primarily

to denote pleasure or displeasure at something (see further De Smet 2004).

(9) a. She really hates to campaign and getting out there on the trail, on the

hustings. (Google, 4 January 2008)

b. He does not like to be in a frenzied atmosphere and hates being hurried

into making a quick decision. (CB)

There are some environments, however, where ing-complements are numerically

underrepresented compared to other environments, despite semantic compatibility.

Typically, these environments instantiate predicate-complement constructions that

may be assumed to be represented independently in the mind of the language user so

that the choice for the to-infinitive does not follow from semantically controlled

matching but is an almost automatic consequence of the choice for the predicate-

complement construction as a whole. One such environment is a construction in

which the infinitive to see combines with a predicate expressing the emotive state

that arises upon becoming aware of some situation. As the examples show, the con-

struction is found with emotive verbs, as in (10a), but also with other predicates de-

noting an emotive state, as in (10b), and may even be related to uses with a causal

infinitival adjunct, as in (10c). Presumably, if it is the predicate-complement con-

struction (i.e. ‘emotive predicate + to see’) that determines the choice of comple-


ment-type, the selectional potential of like or hate as such may be bypassed, explain-

ing why the verb see in the complement clauses following like and hate usually

takes the form of a to-infinitive (for figures see De Smet & Cuyckens 2007: 201).

(10) a. I hate to see birds in cages. (CB)

b. I WAS amused to see a sign advertising “Race-horse manure for sale”.

(CB)

c. They occasionally looked over the multitude of squabbling children,

[...] smiling to see their own proclivities reproduced in their offspring

(CB)

Another environment is the would-like-to-construction, where the emotive

verb, most commonly like, but in recent times also hate or love, is preceded by mo-

dal would or should. The construction has a number of specialised uses, where the

verb no longer clearly denotes an emotive sensation of (dis)pleasure, as in (11a), and

where ing-complements are probably ruled out on semantic grounds. However,

where the construction simply expresses hypothetical enjoyment, as in (11b-d), ing-

complements are licensed semantically, and attested very occasionally, but are strik-

ingly underrepresented compared to other environments.

(11) a. I would like to thank you all very much for coming along this evening

(CB)

b. she made me feel at home in the Middle Ages and persuaded me that I

would have liked to live in the twelfth century. (CB)

c. I would hate to have to give up beef completely. (CB)

d. Would she get much bigger physically? Would delivery be very pain-

ful? Would she like breastfeeding her infant? (CB)

The reticence of ing-complements following would/should like may be accounted

for by the high degree of entrenchment of the to-infinitive, for in addition to the con-

struction’s frequency, there is an extremely strong collocational tie between

would/should like and the following to-infinitive. Corpus data show that in Present-

Day English about 72% of all instances of would/should like are followed by a to-

infinitive, making it plausible that the presence of would or should before like is

enough for language users to call up the to-infinitive as the complement most likely

to follow (cf. Krug 2003 on ‘transitional probability’). Tellingly, around the begin-

ning of the eighteenth century, when the chance for a to-infinitive to follow on


would/should like was considerably lower and the would-like-to-construction was

also much less frequent, ing-complements were more readily sanctioned following

would/should like than they are now, even though they were much less common

with like in general (for figures, see again De Smet & Cuyckens 2007: 206). This is

illustrated in the examples in (12).

(12) a. How say you, mistress, would you like going to sea? (1695, CEMET)

b. I can’t tell how they would like travelling into Italy, when there is a

prospect of the rest of their race returning from thence (1745,

CLMETEV)

In brief, it is plausible that the present-day near-absence of ing-complements follow-

ing would/should like is the result of blocking induced by strong entrenchment.

2.2. Analogy

Sanctioning is not only determined by the frequency of constructions but also by the

regularities those constructions embody. As has been argued in Chapter 3, the

matching of a complement-type to a given predicate, including any innovative

match, crucially depends on a weighing of various motivating factors, whereby con-

structional regularities play a key role. In diffusion, these regularities manifest them-

selves as analogical extensions. Diffusion is possible because of the open-endedness

of grammatical organisation (i.e. the provisional nature of some regularities), and

proceeds as regularities are born out, or as one regularity is taken up where another

has been left off. Thus, from the perspective of analogical sanctioning, diffusional

change derives its diffusional character from the emergence of new regularities that

trigger new changes, and from the fact that different possible innovative uses, being

more or less similar to already established usage, resonate differently to the analogi-

cal pressure exerted by established uses.

The notion of analogy, though highly compatible with usage-based models of

language (Langacker 2000: 59-60; Itkonen 2005), is not unproblematic as an ex-

planans of historical changes (Lass 1980; 1998). One objection is that analogy char-

acteristically removes some exceptions while leaving others untouched. This is true

of course, but as was argued above, the resistance of some exceptions to analogical

change is not necessarily unpredictable or inexplicable (Section 1.2). The other ob-

jection is that analogy as a process is invisible and is therefore always postulated

post hoc – and by the same token, rather too often ad hoc. This too may be true, but


invoking analogy obviously becomes less gratuitous if there is corroborating evi-

dence. In this section, specific cases will be highlighted where the role of analogy

happens to be particularly plausible because of additional chronological and distri-

butional evidence.

The purpose of this section, then, is twofold. On the one hand, it introduces

two sub-types of analogy that will play an important role in the following chapters:

‘semantic analogy’ and ‘paradigmatic analogy’. These two types of analogy work on

different kinds of regularities that can be extracted from usage, and the distinction

between them serves as a first illustration of the multiplicitous interaction of analo-

gies that drives diffusion and which results in the complex synchronic situations

discussed earlier (Chapter 3). On the other hand, this section aims to defend the no-

tion of analogy against potential criticism and draws on case-studies to show the

relevance of the analogical mechanisms to some specific instances of diffusional

change in the system of complementation. In what follows, Section 2.2.1 introduces

and exemplifies semantic analogy, while Section 2.2.2 discusses paradigmatic anal-

ogy.

2.2.1. Semantic analogy

Semantic analogy is a mechanism of analogical extension that allows a construction

to extend its range of application on the basis of semantic similarity between the

source environment and the target environment of extension. It is easy to see how

the diachronic mechanism of semantic analogy is linked to the regularities that or-

ganise the system of complementation synchronically. It may be assumed that one is

the reflex of the other, in that what appear to be semantic regularities from a syn-

chronic point of view are often constructional extensions motivated by semantic

analogy from a diachronic point of view (cf. Israel 1996). Semantic analogy readily

relates to the sanctioning principle, in that change proceeds along a path of least

resistance and, by recruiting new environments on the basis of (semantic) similarity,

guarantees the formation of natural classes.

Notice that semantic analogy rarely occurs in its purest form (i.e. extension

based exclusively on semantic resemblance between environments). Semantically-

based extensions are often restricted to environments that are structurally similar as

well (e.g. from one verbal predicate to another verbal predicate Fischer p.c.).19

19 Indeed, the examples in Chapters 9 and 10 suggests that semantic analogy typi-cally works across structural boundaries only in the case of additional formal or


Moreover, as will be shown in Section 2.2.2, semantic analogy often converges with

paradigmatic analogy. Nevertheless, its influence can be isolated as a factor of

change.

As a particularly clear example of semantic analogy in the area of comple-

mentation, I will here discuss at some length the shift from a transitive to an intransi-

tive use for worth, and the shift from an intransitive use to a transitive use for

worthwhile. In Present-Day English these changes are reflected in the layered syn-

chronic patterning of both adjectives. Thus, in all but one of its uses, worth can be

described as a ‘transitive’ adjective,20 which means that next to a subject it selects

another argument, which can be conveniently referred to as an object (even if this

object is hardly the prototypical patient of transitive verbs), as is illustrated in (13).

Corresponding to the syntactic roles of subject and object are the semantic roles of

that which is valued (the subject) and the value (the object), between which a rela-

tion of fair or profitable exchangeability is predicated. Note here that omission of the

object is clearly ungrammatical (*a touchdown is worth).

(13) A touchdown is worth six points. (CB)

When worth takes an ing-clause as its object, the syntax gets more complicated be-

cause the subject of worth controls a secondary participant of the ing-clause, with

the controlled participant ‘appearing’ in the ing-clause as a missing object or as a

stranded preposition, as in (14). Still, the relation expressed by the adjectival predi-

cate can be interpreted in the same way as when worth is used with an ordinary ob-

ject: the subject (the restaurant) is now valued in relation to an activity, typically in

terms of the effort that activity requires (visiting).

(14) a. the restaurant was crook and therefore not worth visiting. (CB)

morphological relatedness between source and target environments (see the cases of fear and intend in Chapter 9, or hesitate and succeed in Chapter 10). 20 This is of course unusual for English, and worth has been alternatively analysed as a preposition (Maling 1983). Unlike other prepositions, however, worth is not used to introduce adverbial phrases, and, like adjectives, worth still has comparative and superlative uses with more and most and is even occasionally modified by very and other intensifiers (e.g. another winter salad very worth growing (CB)). In their ref-erence grammar, Quirk et al. (1985) seem to be undecided as to whether worth is an adjective (as implied in 16.83) or a preposition (as implied in 9.8). Huddleston & Pullum (2002: 607), by contrast, are convinced worth is an adjective rather than a preposition.


b. Now was not this heroic lover worth running after? (1742,

CLMETEV)

What is unexpected are instances as those in (15): the structure in which worth is

attested has only one participant, an extraposed ing-clause (emigrating).21 Syntacti-

cally, worth now seems to behave like other intransitive adjectives (for instance,

easy in it hadn’t been easy breaking the news to Nicole (CB)), while semantically,

the gerund clause following worth no longer specifies an exchange value for a given

subject, but is itself implicitly valued as being ‘of positive value, worthwhile’. Ac-

cordingly, the use of worth in examples as in (15) can be characterised as intransi-

tive – even if some intransitive operations are clearly disallowed (e.g. fronting of the

gerund clause, as in *emigrating is worth with anthems like that, isn’t it).

(15) With anthems like that it’s worth emigrating isn’t it. (CB)

The adjective worthwhile shows a roughly similar discrepancy in its use.

Worthwhile is an intransitive adjective as the example in (16), with its single partici-

pant, illustrates. This state of affairs makes sense historically: as worthwhile derives

from the semi-fixed phrase worth one’s while, an object is already incorporated

morphologically and semantically, and thereby pre-empted syntactically. Apart from

being intransitive, worthwhile also differs from worth, in that it can be used both

predicatively (the investment is worthwhile) and attributively (a worthwhile invest-

ment) – another point in which it more closely resembles ordinary adjectives.

(16) all the work has been worthwhile (CB)

In light of worthwhile’s intransitivity, there is nothing very remarkable about struc-

tures with extraposed ing-clauses, as illustrated in (17): worthwhile is used like other

adjectives such as difficult, funny, great, etc. More problematic are the examples in

21 The examples in (3) are obviously hard to analyse syntactically. Hantson (1987: 264-5) feels compelled to postulate a unique movement rule to accommodate the pattern. Further, the referential status of dummy-it in extraposition structures re-mains a disputed issue. According to Kaltenböck (2003), it could still be viewed as a participant with maximally general reference. This would nicely suit an analysis of worth as consistently transitive but would of course complicate the analysis of other extraposition structures with intransitive adjectives. I believe this analytical dilemma is one manifestation of the slippery slope on which transitive worth finds itself, as will be argued below.


(18), where worthwhile appears to take the same construction type as does worth in

(14) above, with the subject (that) simultaneously functioning as the missing object

of the gerund clause following the adjective. The close resemblance to the transitive

uses of worth will justify referring to it as the transitive use of worthwhile (despite

small semantic differences; see below). The most important thing to note is that a

similar transitive use is not allowed for any other intransitive adjectives (cp. *the

news hadn’t been easy breaking to Nicole).

(17) it is certainly worthwhile stopping off on the way. (CB)

(18) there is much that is worthwhile visiting (CB)

From a historical point of view, both synchronic discrepancies in the use of

worth and worthwhile are innovations, and it is important to establish with some

precision the sequence of events leading up to them. The emergence of ing-

complements with worth figures against the background of the more general diffu-

sion of ing-complements during the Modern period (see Chapter 9). Ing-

complements with worth appear in the second half of the sixteenth century. The first

instances have a definite article introducing the ing-form, as in example (19a), but

soon ing-forms appear without the article, as in (19b). As both (19a-b) indicate, the

early instances already show the control relationship between the main clause sub-

ject and the implicit patient of the ing-clause/phrase that is also characteristic of later

ing-clauses with transitive worth. Instances of intransitive worth turn up soon after

the introduction of ing-complements with worth. The earliest example is given in

(20a), drawn from the Diary of Samuel Pepys;22 the following example attested is

rendered in (20b).

22 The example from Pepys in (20a) may not be a genuine instance of intransitive worth: if it is not a dummy-it but refers cataphorically to the ill state a Minister of State is in (with an intonation break between considering and the ill state), (20a) still instantiates transitive worth. The only other example in Pepys, given in (i), is some-what problematic as well, since the silliness of the quarrell must be the understood subject of the following clause (is a kind of emblem etc.), whereby the possibility of subject ellipsis suggests that the silliness of the quarrell might be the subject of the preceding clause (it is worth remembering) as well. (i) Here Creed did tell us the story of the dwell [i.e. ‘duel’] last night, in

Conventgarden, between Sir H. Bellasses and Tom Porter. It is worth remembering the silliness of the quarrell, and is a kind of emblem of the general complexion of this whole kingdom at present. (1667, PEPYS)


(19) a. and yf their women haue any thing about them, as apparell or lynnen,

that is worth the selling, they laye the same to gage [i.e. ‘they pawn

it’], or sell it out right (1567-8, PPCEME)

b. they haue left nothing here but short riggle-tayle-Comfits [i.e.

‘wriggling sweetmeats’], not worth mouthing (1620, PPCEME)

(20) a. But it is worth considering the ill state a Minister of State is in, under

such a Prince as ours is (1667, PEPYS)

b. Surely, then, it is worth paying some regard to the principles of fitness

and consistency, in order to avoid the consequences necessarily result-

ing from every striking deviation from these rules (1839, CLMETEV)

One fact to be added to this is that the history of intransitive worth is marked by a

long period of sporadic occurrence (as is also apparent from the time-lag between

(20a) and (20b)), followed by a sudden surge in frequency halfway the twentieth

century, as is nicely reflected in the frequency curve depicted in Figure 4.3.23 As

Figure 4.3 clearly shows, the dramatic increase in the pattern’s use began sometime

during the first half of the twentieth century, between the period represented by the

CLMETEV and that represented by LOB.

The development of worthwhile runs in the opposite direction, from intransi-

tive to transitive uses. The history of worthwhile starts with the semi-fixed phrase

worth one’s while. The latter expression is current as early as the seventeenth cen-

tury, when it is typically used in combination with a to-infinitive, as in (21a). As is

illustrated in (21b), alongside the pattern with a possessive or genitive preceding

If none of the examples in Pepys’ Diary turn out to be genuine instances of intransi-tive worth, however, this only reinforces the view developed above that the emer-gence of intransitive worth is due to semantic analogy with worthwhile. 23 Figures are from CEMET, PPCEME, LC, CLMETEV, LOB and FLOB. The very low frequency of intransitive worth in the period 1850-1920 is further confirmed by data from CEN, evidencing some 11 examples, or 0.4 instances per million words, while figures from CB confirm the sharp rise of the pattern in the twentieth century, with 12.2 instances per million words. Whether or not the S-curve in Figure 2 has in fact levelled out in the second half of the twentieth century, as is suggested by the FLOB data, is difficult to ascertain – the frequency for intransitive worth in CB is higher than in LOB, but not significantly so.

0

2

4

6

8

10

1640-1710

1710-1780

1780-1850

1850-1920

1961 1990s

intransitive worth

Figure 4.3. Frequency of intransitive worth over time (frequencies per million words).


while, seventeenth-century English also has the pattern with while immediately fol-

lowing worth – though still, apparently, as a separate word.

(21) a. but yet I thinke it worth our while to seeke the ways and practise them

when found (1685, PPCEME)

b. Sir, I fancy ‘tis not worth while to trouble Sir John upon this imperti-

nent Fellow’s desire (1696, PPCEME)

The earliest attestation of worthwhile in an extraposition construction with an ing-

clause dates from the beginning of the nineteenth century and is given in (22a). This

new use is soon followed by instances of transitive worthwhile. The oldest instance

is given in (22b) and dates from the end of the nineteenth century.

(22) a. Sir William does not think it worth while making another application

(1813, CLMETEV)

b. They fear in many cases he takes the refuse in order to have the oppor-

tunity of finding something which may be worth while “picking up,”

(1890, CLMETEV)

The question now is: how did these changes come about? One element in the

explanation, no doubt, is syntactic reinterpretation, allowing the innovative uses to

slip in through ambiguous instances of existing patterns of use. Worth in particular

often figures in the type of context illustrated in (23), where dummy-it can either be

interpreted as the dummy of an extraposed that-clause (in line with the transitive

interpretation), or the dummy of an extraposed ing-clause (implying an intransitive

reading).

(23) The value of the studentship is slightly under £200 a year. It is worth

noticing that persons of both sexes are received as candidates. (1883,

CLMETEV)

Note, though, that reinterpretation cannot tell us how the ambiguity arose in the first

place. Nor can reinterpretation as such account for the sudden surge in the frequency

of intransitive worth after 1920. Moreover, for worthwhile the role of ambiguous

sequences is definitely problematic, as the examples that would allow an alternative

interpretation, such as (24), are very infrequent in the historical data. The whole

CLMETEV corpus contains only a single instance of worthwhile with ing-clause


that is structurally ambiguous between an intransitive and a transitive reading, while

no more than four such instances are found in CEN. Of these, the only example that

predates the earliest unambiguous instance of transitive worthwhile ((22b) above)

does so by only one year.

(24) It would be worth while remembering, when he had gone, that he had

been gentle with her (1900, CEN)

A second element in the historical development is a subtle semantic change in

the transitive construction that may have facilitated the subsequent shifts in transitiv-

ity. Specifically, worth with an ing-complement has, in some instances, come to

deviate semantically from other uses of worth, probably as a result of semanticisa-

tion of pragmatic implicatures. In some Late Modern and Present-Day English ex-

amples, worth no longer assigns a particular value to its subject; instead, it desig-

nates the activity or situation expressed by the gerund clause as profitable – note that

paraphrases with an intransitive construction are certainly not far-fetched. This is

illustrated in (25).

(25) a. It has been said of him that he never made a mistake, that the plunder

he took was always large. His victims, too, were always those who had

bad reputations; and, one thing more, Mistress Lanison, his victims

have always won largely at Aylingford Abbey. […] He knew when

they were worth robbing. (1910, CLMETEV)

b. If we do not consider that the costs are worth paying, then we must

frankly acknowledge the human implications that some children will

die to preserve the freedom of others. (CB)

As a result of this semantic shift, transitive worth already approaches the semantics

of an intransitive construction before it is actually used as an intransitive adjective.

Even so, semantic change alone still cannot explain the syntactic changes that fol-

lowed.

It is semantic analogy that can tie everything together. Because of the seman-

tic developments in transitive worth, evidenced in (25), the adjectives worth and

worthwhile came more closely to resemble one another semantically. We can there-

fore hypothesize that, on the basis of semantic similarity, transitive worth provided

an analogue for the innovative transitive use of worthwhile, while intransitive worth-


Figure 4.5. Schematic representation of the hypothesised development of worth and

worthwhile.

while suggested the model for the new intransitive use of worth. As will be argued,

there is good evidence supporting the first assumption; the second needs to be quali-

fied but retains a good degree of plausibility. For clarity, the hypothesised historical

cross-over of constructional behaviour is schematised in Figure 4.5.

The fact that the syntactically transitive use of worth approaches an essen-

tially intransitive meaning suggests that what looks like a transitive construction for

worthwhile need not be radically different semantically from worthwhile’s usual

intransitive pattern. In other words, at the time worthwhile began to combine with

extraposed ing-clauses – a use still in line with its status as an intransitive adjective

– the transitive worth-construction already allowed an interpretation that was not

incompatible with the semantics of worthwhile. The subsequent use of worthwhile in

the transitive pattern can therefore be seen as an instance of analogical extension of

the transitive pattern. Put differently, the transitive construction simply recruited a

new lexical item. That transitive worth played a crucial part in the emergence of

transitive worthwhile is plausible from the fact that ambiguous sequences with

worthwhile may have been too infrequent to trigger the change independently (see

above), and that other adjectives occurring with extraposed ing-clauses as yet did

not undergo a similar shift towards transitive uses, despite the availability of poten-

N is worthwhile Ving

N is worth Ving (1)

N is worth Ving (2)

It is worth Ving

It is worthwhile Ving

Sem

antic

ch

ange

R

eint

er-

pret

atio

n

Rei

nter

-pr

etat

ion

Semantic analogy


tially ambiguous sequences that could give rise to syntactic reinterpretation. Thus,

the potentially ambiguous sequences in (26) did not trigger new transitive uses, as is

shown by the ungrammaticality of the sentences in (27).

(26) a. it’s very interesting noting here that erm you were brought up in Bur-

ton-on-Trent (CB)

b. It was very difficult deciding whether these are a beanbag or a ball.

(CB)

c. Of course, it is not easy knowing how many mourners will appear to

toast your memory. (CB)

d. It is hopeless trying to get him to give any definite dates for anything.

(CB)

(27) a. *one thing is very interesting noting

b. *the question was very difficult deciding

c. *the number of mourners is not easy knowing

d. *that is hopeless trying

As for the intransitive uses of worth, it is clear that the existence of intransi-

tive worthwhile cannot have been the only cause of change, since intransitive worth

(attested halfway the seventeenth century; yet see footnote 22 above) already oc-

curred before the appearance of the gerundial extraposition structure with intransi-

tive worthwhile (appearing in the early nineteenth century). These early occurrences

of intransitive worth are likely to have been modelled on the extraposition construc-

tions with other adjectives already occurring at the time, as illustrated in (28). Pre-

sumably, it were these extraposition structures that made some uses of worth with an

ing-clause ambiguous in the first place (see example (23) above).

(28) a. in the mean time it is best consulting Merchants and Seamen of most

fame for honesty, ability, and publick-heartedness, who can give you

an account of the state of our several Trades abroad (1659, LC)

b. When it is come thus far, it is hard untwisting the Knot (1673,

PPCEME)

c. Then I thought that it is easier going out of the way, when we are in,

than going in when we are out. (1678-84, CEMET)

0

2

4

6

8

10

1640-1710

1710-1780

1780-1850

1850-1920

1961 1990s

intransitive worth

intransitiveworthwhile

Figure 4.6 Frequency of intransitive worth and intransitive worthwhile over time (frequencies

per million words).


Still, the emergence of intransitive worth and the possible role in it of intran-

sitive worthwhile deserve closer scrutiny. Although interference with worthwhile

could not have been responsible for the earliest sporadic instances of intransitive

worth, it may account for the sudden surge in frequency of intransitive worth after

1920. Some evidence suggests that this is in fact so. For one thing, the surge in fre-

quency of intransitive worth follows very closely – say within fifty years – on the

first significant rise in frequency of intransitive worthwhile. This is shown in Figure

4.6, which reveals a modest but significant increase in the frequency of intransitive

worthwhile between the periods 1780-1850 and 1850-1920 – just prior to the dra-

matic increase in frequency for intransitive worth.24 Note, moreover, that the figures

for 1961 indicate a continuation of the rise of intransitive worthwhile,25 suggesting

an S-curve parallel to that of intransitive worth, but slightly preceding the latter in

time. The subsequent drop in frequency for intransitive worthwhile in the last dec-

ades of the twentieth century might be due to competition with intransitive worth,

which, after all, conveys much the same meaning.26 What this suggests, of course, is

that the post-1920 increase in frequency of intransitive worth was triggered by in-

transitive worthwhile.

A further finding suggestive of analogy between intransitive worth and in-

transitive worthwhile comes from a comparison between British and American Eng-

lish. In particular, the geographical distribution of intransitive worth matches the

geographical distribution of intransitive worthwhile. Observe first that intransitive

worth is less frequent in American English than in British English. This is suggested

by a comparison between LOB and FLOB on the one hand and BROWN and

FROWN on the other. Together, LOB and FLOB, representing British English, con-

tain 15 instances of intransitive worth; BROWN and FROWN, representing Ameri-

can English, contain only 5. Interestingly, now, a similar geographic distribution can

be observed for intransitive worthwhile, which is relatively frequent in British Eng-

lish (with 7 instances in LOB and FLOB) but largely absent in American English

24 Significance has been calculated using a Fisher’s exact test (based on the raw fre-quencies). For the transition from the second to the third sub-period in CLMETEV, the difference between the respective frequencies of worthwhile (1 and 10 instances respectively) lies at p < 0.05. Somewhat more convincingly, perhaps, if the figures for the second sub-period of CLMETEV are compared to those for CEN (containing 52 instances of intransitive worth), the change is significant at p < 0.001. 25 The difference between the 52 instances of CEN and the 6 instances of LOB is significant at p < 0.01, that between the figures for LOB and the third sub-period of CLMETEV at p < 0.05. 26 The drop in frequency is significant (at p < 0.05) when comparing the 6 instances in LOB to the 10 instances in the ukbooks section of CB.


(no instances in BROWN or FROWN). As is demonstrated in Table 4.1, this state of

affairs is confirmed by the figures obtained by comparing the ukbooks and usbooks

sections of CB.27, 28 The Table shows that both intransitive worth and intransitive

worthwhile are British English constructions. Important is also the absence of intran-

sitive worthwhile in BROWN, which suggests that in earlier stages of the language,

too, worthhwhile was largely absent in American English. Thus, there seems to have

been no American developments paralleling the British English developments

shown in Figure 4.4 above, involving first a rise in frequency of intransitive worth-

while, closely followed by a rise in frequency of intransitive worth. If the surge in

frequency of intransitive worth occurred under the influence of intransitive worth-

while, it is to be expected that this happened primarily in the regional variety that

had intransitive worthwhile. The data bear out this expectation, adding support to a

connection between intransitive worth and intransitive worthwhile (see De Vogelaer,

Devos & Van Der Auwera 2006: 221 for a similar argument using lectal variation to

support an analogy-based account of historical change; see also Section 1.3 above).

This interpretation of the historical course of events throws up one question:

why did intransitive worthwhile not outcompete worth at the time when intransitive

worth was still extremely marginal, between 1850 and 1920 – or, more precisely,

how can one claim that intransitive worthwile first promoted the use of intransitive

worth, yet at the same time maintain that the increased use of intransitive worth did

not promote intransitive worthwhile? A possible answer is that, even though the fre-

quency of intransitive worth was extremely low, the overall frequency of worth has

always been much higher than that of worthwhile (the frequency of worth in the

transitive gerundial pattern alone lies at 36.8 instances per million words for the

third sub-period of CLMETEV, and at 33.0 for CEN). Added to that is the fact that

27 The figure for intransitive worth in ukbooks is not entirely accurate. Intransitive worth occurs 124 times in the ukbooks section of CB, which makes for 23.2 in-stances per million words. Of the 124 instances, however, 38 occur in the same text on gardening (D. Kitto, 1986, Planning your Organic Vegetable Garden). To com-pensate for this, I have omitted all examples from the gardening text, and recalcu-lated relative frequencies. Because there is no information on the size of the omitted text, relative frequencies have been calculated on the basis of the full sub-corpus’ size and are therefore slightly too low. The difference between American and British English is clear enough, however. 28 The same picture emerges when we compare the number of instances of intransi-tive worth to the number of instances of transitive worth: these are 124 and 105, respectively, in ukbooks, as opposed to 21 and 97 for usbooks. Using these figures, the difference between British and American English is significant at p < 0.0001.

Corpus Corpus size Time Regional variety Intransitive

worth

Intransitive

worthwhile

LOB 1 million words 1961 British English 8.0 6.0

FLOB 1 million words 1990s British English 7.0 1.0

CB; ukbooks 5.35 million words 1990s British English 16.1 1.9

BROWN 1 million words 1961 American English 3.0 0.0

FROWN 1 million words 1990s American English 2.0 0.0

CB; usbooks 5.63 million words 1990s American English 3.7 0.2

Table 4.1. Frequency of intransitive worth and intransitive worthwhile in present-day British and American English (frequen-

cies per million words)


worth is shorter and therefore more economical than worthwhile, while worthwhile,

having become morphologically opaque, has no obvious advantage over worth in

terms of transparency. Arguably, then, it was impossible for the dwarf to beat the

giant, even when given a head-start. This kind of interaction is comparable to what

Hopper (1991: 25-8) describes as ‘specialization’, which starts from a parallel and

roughly simultaneous development of functionally equivalent constructions but ends

with the dominance of one construction over the others.

In conclusion, semantic analogy provides a crucial explanatory element in the

historical developments of worth and worthwhile, and there is good circumstantial

evidence supporting its role. The extension of the transitive pattern of worth is lim-

ited to the semantically related adjective worthwhile. The extension of the intransi-

tive extraposition construction to worth, on the other hand, is facilitated by the ap-

pearance of worthwhile in the same environment, as appears both from the historical

sequence of events and the geographical patterning of the changes. What the case-

study also demonstrates, is that change is sensitive to changing conditions of use.

This is most evident in the development of intransitive worth, which visibly re-

sponds to changes in the analogical model that underlies its formation, the innova-

tion with worth becoming increasingly acceptable, as its analogical model, with the

emergence of extraposed ing-clauses with worthwhile, extends its sphere of influ-

ence in its direction. Similarly, the timing of changes suggests that the emergence of

transitive worthwhile may have been facilitated by the new alternation that had be-

come evident in the semantically related adjective worth between an intransitive

extraposition structure and a transitive control construction. The extension of alter-

nation patterns, however, leads us from the realm of semantic analogy into that of

paradigmatic analogy (see Section 2.2.2).

Finally, we can think of the developments of worth and worthwhile as two

minor diffusional changes. On the one hand, the gerundial extraposition construction

has spread to worth, while on the other a preliminary gerundial object-to-subject-

raising construction has spread to worthwhile (and may continue to spread to other

adjectives in the future). Taking this perspective, we see how the sanctioning princi-

ple has determined the course of diffusion: worth, being dissimilar from other adjec-

tives and historically even incompatible with intransitive uses, has never received

strong sanction to adopt the extraposition construction, until worthwhile increased

the pressure through semantic analogy. Worthwhile, on the other hand, has so far

been the only adjective to enter in the transitive construction characteristic of worth,

because in contrast to other adjectives its use in the transitive/object-to-subject-

raising construction is strongly sanctioned by semantic similarity to worth.


2.2.2. Paradigmatic analogy

Paradigmatic analogy is less familiar a mechanism of change than semantic analogy.

Paradigmatic analogy involves the extension of a construction from one environ-

ment to another on the basis of a link between the spreading construction and some

other paradigmatically related construction. A possible interpretation of paradig-

matic analogy is that language users construe a paradigmatic link between the

spreading construction and some other construction, extending the former’s range of

occurrence by copying the distribution of the latter. On this interpretation of the

process, language users work with a second-order generalisation not usually incor-

porated in constructional frameworks of language, as it implies knowledge that re-

sides not at the level of the construction but in the comparison between construc-

tions (see Cappelle 2005: 43-52, 2006).29 For example, in purely replacive changes

(e.g. the replacement of the third person singular suffix -th by -s) it is likely that lan-

guage users simply work on the logic that where one pattern occurs the other pattern

can also occur, thus extending the distribution of the latter pattern on the basis of the

former. The paradigmatic relation between the old and new pattern are then suffi-

cient for the new pattern to spread to new environments.

A number of important remarks should be added to this. First, there is a ter-

minological asymmetry between semantic and paradigmatic analogy. Semantic

analogy, as understood here, is based on a semantic relation of similarity between

two environments that are syntagmatically related to the spreading pattern (e.g.

worth and worthwhile are syntagmatically related to the extraposed gerundial ing-

clause). Paradigmatic analogy, by contrast, is based on a semantic, formal and/or

distributional relation of similarity between the spreading pattern and some other

pattern to which it is paradigmatically related (e.g. the -s-ending is a paradigmatic

variant of the -th-ending to mark the third person singular present; their similarity is

semantic, distributional, and to some extent formal). The reason for not distinguish-

ing sub-types of paradigmatic analogy is that in practice they are typically indistin-

guishable. The complex relation between semantic and paradigmatic analogy is

visualised in Table 4.2. Second, as is further apparent from Table 4.2, paradigmatic

analogy is in practice (though not in theory) indistinguishable from syntagmatic dis-

29 These second-order generalisations are of course the transformational alternations of generative syntax – or in systemic-functional terms, the nodes of an ‘agnation network’ (Gleason 1965).


RELATION TO MODEL

SYNTAGMATIC PARADIGMATIC

MEANING Semantic analogy

FORM Formal analogy

BA

SIS

OF

SIM

ILA

RIT

Y

DISTRIBUTION

Paradigmatic analogy

Table 4.2. Types of analogy.

tributional analogy. The latter mechanism applies when a pattern spreads to new

environments on the basis of a distributional similarity between the environments –

i.e. apart from the spreading pattern, the environments combine with the same con-

structions (e.g. both do and take combine with the same -th-ending, on which ground

-s, combining with take, can also spread to do ).30 Third, as is again indicated in Ta-

ble 4.2, yet another type of analogy is possible, though not discussed here, namely

formal syntagmatic analogy. This type of analogy, based on formal similarity be-

tween the source and target environments involved in analogical extension, does

occur but is relatively rare (a few possible examples of formal syntagmatic analogy

are pointed out in Chapter 9). Fourth, even the boundaries between semantic and

paradigmatic analogy are not always clear-cut because in many cases, the two con-

spire (see e.g. the diffusion of for...to-infinitival complements in Chapter 6) and also

because there is often a relation between an item’s semantics and the other patterns

it can combine with (see below). Even so, it is useful to keep semantic and paradig-

matic analogy apart, as this section is to argue.

Turning to actual examples of paradigmatic analogy in the system of com-

plementation, the most far-reaching claims have been put forward by Los (1999,

2005) in her discussion of the spread of to-infinitival complements in Old English.

She argues that to-infinitives go back to prepositional phrases with to and on that

30 Replacing paradigmatic analogy by distributional syntagmatic analogy would re-move the theoretical problem of situating paradigmatic relations in a construction grammar (see above). However, this reduction would miss the important point that the distributional similarity underlying many cases of analogical extension is not merely distributional, but also involves a relation of semantic and sometimes formal similarity between the spreading construction and its paradigmatic variants.


ground hypothesises that the first verbs that would combine with to-infinitival com-

plements must have been those that also licensed goal arguments in the form of a

prepositional phrase with to. By the time of the first Old English records, however,

to-infinitival complements already occurred beyond the confines predicted by the

distribution of prepositional phrases with to. This observation, suggestive of a pre-

historic process of diffusion, raises the question of what accounts for the extended

distribution of the to-infinitive. According to Los (2005: 99), “the most probable

explanation is the functional equivalence of the to-infinitive with the subjunctive

that-clause”, which allowed to-infinitives to spread to new verbs “by analogy with

the subjunctive that-clause”. In support of this view, Los demonstrates that there is a

good degree of semantic equivalence between the two complement-types, as well as

positional similarity and, most importantly, a remarkably close match between the

distributions of to-infinitives and subjunctive that-clauses in Old English, which

occur with practically the same set of verbs, and both of which are used as purpose

adjuncts as well. The close equivalence between both structures is illustrated in (29).

(29) a. ða [...] domas, ðe <ic> [...] bebeode ðæt ge don

‘the judgements that I order that you do’ (Los 2005: 123)

b. þis sindan ða domas þe se ælmihtega God [...] him bebead to

healdanne

‘these are the laws which the almighty God [...] ordered him to keep’

(Los 2005: 123)

In addition, Los (2005: 179-89) shows that between Old English and Middle Eng-

lish, increases in the use of the to-infinitives in specific environments invariably go

at the expense of subjunctive that-clauses. Comparing differently-dated manuscripts

of the Old English translation of Gregory’s Dialogues, she even finds that the newer

version of the text has in various places quite literally substituted to-infinitives for

subjunctive that-clauses.

Of course, as the spread of to-infinitival complements began before the Old

English period, the actual process of diffusion is largely unobserved, so part of Los’

(1999, 2005) account remains conjectural. It is particularly unfortunate that the Old

English data do not allow us to isolate the effect of paradigmatic analogy from that

of semantic analogy (many of the predicates with which to-infinitival complements

occur are semantically related). In other instances of change, however, the role of

paradigmatic analogy can be observed more directly. Consider, for example, the


Figure 4.7. Paradigmatic analogy in start after keep.

appearance of start in the causative construction. As (30a) shows, this use first ap-

peared with start at the end of the nineteenth century, about half a century after start

had begun to combine with subject-controlled ing-complements, as in (30b) (see

Chapter 9).

(30) a. I saw Munroe leaning against the railings in front of a house in Wharf-

dale Road – I started him walking to see if he could take care of him-

self, and he walked through into All Saints’ Road (1886, POB)

b. I had before this written to Rose how we had best start agitating (1833,

Visser 1963-73: 1896)

These innovations took place as start changed in meaning from ‘(cause to) make a

sudden motion’ to ‘begin’. The fact that the appearance of start in the causative con-

struction follows relatively closely on its appearance with subject-controlled ing-

complements suggests that the use of the causative construction may have been

based on a first paradigmatic analogue between start and verbs such as keep or get,

which displayed the same combinatorial behaviour. Based on the examples in (30)-

(31), the analogy would have looked as in Figure 4.7:

(31) a. Why not? – he kept asking himself. (1889, CEN)

b. this latent repulsion occupied my mind, and kept me wondering on

what grounds it rested (1887, CEN)

Matters are not quite so simple, however. Because both keep (or get) and start could

be used as aspectual verbs (as in (31a) above), we could also argue that the analogy

was not paradigmatic but semantic. Furthermore, regardless of whether analogy was

semantic or paradigmatic, it is unclear why verbs such as begin or continue, near-

keep Ving

keep NP Ving

start Ving

x = ?


synonyms of start and likewise used with controlled ing-complements, have not

come to be licensed in the causative construction. Compare the examples in (32):

(32) a. he burst rudely past me, dashed into the room, caught up a jar and a

small horn-spoon, and began ladling snuff into his nose in most exces-

sive quantities. (1886, CEN)

b. *I began him walking to see if he could take care of himself.

Finally, the alternation between subject-controlled ing-complements and causative

constructions is defective in the opposite direction as well. The verb set, for in-

stance, was and is used in the causative construction, but somewhat puzzlingly never

appeared with subject-controlled ing-complements, as shown in (33).

(33) a. I noticed a stench from the varnish factory when they were boiling

varnish, [...] it set me coughing (1857, POB)

b. *He set ladling snuff into his nose in most excessive quantities.

These problems and discrepancies can be explained by invoking two further para-

digmatic analogues. On the one hand, keep and start (as well as set) further resem-

bled one another in that they could combine with resultative prepositional phrases,

as in (33) – an option that was unavailable for begin or continue. That start, unlike

other aspectuals, could be used in this pattern has to do with its original meaning

‘cause to make a sudden movement’ and the associated motional semantics, which

of old allowed the verb to combine with prepositional phrases expressing sources,

goals and paths (see Chapter 9).

(33) a. I gave that man the cheque and started him into the city (1859, POB)

b. She was white to the lips, but kept her eyes on him steadily. (1888,

CEN)

c. *I gave the man the cheque and began him into the city.

On the other hand, set differed from keep, get, start, begin or continue in that the

verb’s basic argument structure showed no additional paradigmatic analogue to sup-

port the use of subject-controlled ing-complements, either in the form of a subject-

complement, as with keep or get (see (34a)), or in the form of an ordinary direct ob-

Figure 4.8. Network of paradigmatic analogies underlying the appearance of start in the causative construction.

keep Ving

keep NP Ving

start Ving

x = ?

begin Ving

keep NP PP start NP PP

set NP Ving

set NP PP

keep ADJ start NP begin NP


ject denoting an action by the matrix subject, as with start, begin or continue (see

(34b-c)) (these paradigmatic links are discussed at length in Chapter 9).

(34) a. The three or four who seemed likeliest to be obedient and to keep so-

ber he drew aside (1884, CEN)

b. Our friend here is a wonderful man for starting a chase. All he wants is

an old dog to help him to do the running down. (1892, CEN)

The network of analogies that underlies the appearance of start in the causative con-

struction, then, must have resembled that in Figure 4.8.

It may be noted that this discussion of the development of start brings back

two conclusions already arrived at earlier. First, not everything that happens in the

system of complementation is dictated by semantics: semantically related predicates

may still behave differently with respect to the selection of different complement-

types. The explanatory power, in this particular case, resides with paradigmatic links

between constructions. Second, multiple paradigmatic analogies may contribute to

shaping a sub-system, in this case the alternation between subject-controlled ing-

complements and causative constructions.

Paradigmatic analogy also plays a role in determining which environments

first see new constructions arise. It is again Los (2005) who provides us with an ex-

ample, in her discussion of the emergence of exceptional case marking constructions

in the Early Middle English period. An example of the innovation in question is

given in (35), where the inanimate noun phrase þy comaundement, though still case-

marked by the higher verb, cannot be an argument of that verb but must be inter-

preted as the subject of to ben.

(35) þou comaunded þy comaundement to ben greteliche kept

‘you commanded your commandment to be carried out to the letter’

(Los 2005: 242)

Los shows that exceptional case marking constructions as in (35) first appeared with

verbs of commanding and permitting and never spread to verbs of urging and per-

suading. This is no coincidence. In all likelihood, exceptional case marking con-

structions go back on three-place object-control constructions that had a recipient

argument of the matrix verb controlling the to-infinitival theme argument. An exam-

ple of this source construction is (36a), where the dative case of eow betrays its

status as a recipient – in parallel to (36b) where the dative constituent seocum man-


num and cildum takes up the recipient role and the accusative form nan fæsten fills

the role of theme argument.

(36) a. ða ðincg ðe ic bebeode eow to gehealdenne

‘those things that I order you to hold’ (Los 2005: 103)

b. Seocum mannum, and cildum, we ne bebeodað nan fæsten.

‘we do not order sick people and children to fast’ (Los 2005: 102)

The loss of distinctive dative case marking in combination with the occurrence of

structurally ambiguous instances allowed for recipient arguments to be reinterpreted

as the subject of the following to-infinitive. Crucially, however, this change was

only possible because the two-place construction that thus arose was sanctioned by

the verbs of commanding and permitting. For example, the verbs of commanding

and permitting already combined with the AcI construction illustrated in (37) in

what is undoubtedly a two-place pattern.

(37) Nu ic bebeode beacen ætywan, wundor geweorðan on wera gemange

‘now I order a portent to appear, a miracle to happen among men’ (Los

2005: 103)

Verbs of persuading and urging, by contrast, also combined with object-controlled

to-infinitives, as in (38a), but never developed an exceptional case-marking con-

struction – witness the ungrammaticality of (38b). The fundamental reason, accord-

ing to Los (2005: 252), is that there was and is no two-place construction with these

verbs that licenses the change. Instead, these verbs invariably combine with a theme

argument and a goal argument to form a three-place construction.

(38) a. On hwilcum godum thihst þu us to gelyfenne?

‘Which gods do you urge us to believe in?’ (Los 2005: 57)

b. *He persuaded the town hall to be demolished. (Los 2005: 251)

What we see then is that the reinterpretation of a three-place object-controlled con-

struction as a two-place exceptional case-marking construction was only licensed in

environments that already associated with two-place constructions.

To round of the discussion of paradigmatic analogy, it is worth considering

the relation between paradigmatic analogy and the other mechanisms of change dis-

cussed above, namely blocking and – once again – semantic analogy. Paradigmatic


analogy may occur in a replacive change, just like blocking, and in that respect

seems to be a contrary phenomenon: whereas in blocking the presence of an old

variant hinders the spread of the new variant, in cases of paradigmatic analogy the

old variant invites the new variant. The two mechanisms appear to be irreconcilable

but there are a number of reasons why this is not so. For one thing, there is no prob-

lem, in principle, with competing motivations in language use. Moreover, blocking

is strictly speaking grounded in a frequent surface combination, whereas paradig-

matic analogy is based on a recurrent alternation. It is only on a superficial interpre-

tation that the application of the two mechanisms is dependent on the same structure.

Finally, in practice it turns out that paradigmatic analogy is not strictly linked to

replacive changes. Mostly, the paradigmatic analogy is between structures that are

functionally equivalent but not identical so that there is no strong competition. For

example, in the spread of for...to-infinitives, discussed in Chapter 6, paradigmatic

analogy works on the alternation between for...to-infinitives and to-infinitives, two

patterns that are functionally similar yet not in direct competition. Even in the rela-

tionship between to-infinitives and subjunctive that-clauses, which may have been a

driving force in the diffusion of to-infinitives in pre-Old English times, there was

only competition if the subject of the subjunctive that-clause was co-referential with

the subject or object of the matrix clause, so that the subjunctive that-clause could

be replaced by a controlled non-finite clause. In brief, what we usually find in cases

of paradigmatic analogy is a paradigmatic link between two variables on the same

parameter – that is, two variants that fill different if closely related functional niches.

If in theory the distinction between semantic and paradigmatic analogy is

straightforward, in practice it is not always possible to isolate the effect of either

mechanism. This is to be expected. The distributions on which paradigmatic analogy

works are likely to be in part determined by semantic regularities – indeed, it is no

coincidence that distributional information can be used to determine semantic rela-

tions such as polysemy and synonymy (Sinclair 1992). As a result, it is often unclear

whether a given extension is based on paradigmatic ties or simply rekindles a se-

mantic link between two environments. For example, it is conceivable that semantic

similarity between to-infinitives and subjunctive that-clauses causes them to get

paradigmatically related, allowing the to-infinitive to spread by copying the distribu-

tion of the subjunctive that-clause. But the same semantic similarity predisposes to-

infinitives and subjunctive that-clauses to be reinterpreted as complement clauses

with the same set of complement-taking predicates and would allow them to diffuse

in exactly the same manner over the inventory of predicates through semantic anal-

ogy. Thus, if we start by thinking of the subjunctive that-clause, whose distribution


probably took shape without a paradigmatic analogue, the question that arises natu-

rally is why to-infinitives could not have gone through exactly the same develop-

ment, with the same mechanisms and without the need to posit a paradigmatic ana-

logue – after all, the semantic similarity of to-infinitives to subjunctive that-clauses

would be sufficient to ensure a similar outcome. Even so, in various cases it is clear

that change relies on paradigmatic ties in addition to semantic ties – as in the exten-

sion of the causative construction to start but not begin. In fact, as will become in-

creasingly evident in the chapters to come, paradigmatic analogy is a force not to be

ignored in the diachronic development and synchronic organisation of the system of

complementation.

2.3. Complicating factors

In the picture of diffusional change suggested so far, there is implicit – yet rather

misleadingly so – the notion that change spreads smoothly outwards from a single

point of entrance in the grammatical system, like the concentric waves that spread

from the point where a stone has hit the water. The environments that, in terms of

sanctioning, are furthest removed from a spreading pattern’s first use will be last

affected by diffusional change. An important correction to this picture has already

been pointed out. Because there may be more than one kind of regularity involved in

change, combined with the special role of frequency, as well as a number of addi-

tional extra-grammatical factors, diffusion may take unforeseen turns and proceed in

new unpredictable directions or may be hindered in unforeseen ways. For example,

as a result of blocking, there are little islands of resistance that the spreading waves

of diffusion do not affect.

There are some further complicating factors, however, that make the meta-

phor suggested here inadequate as a representation of diffusional change. Without

going into the possible complications in detail, I will just point out two. First, diffu-

sional changes may slip into the system of grammar at more than one point of en-

trance. For example, it will be shown in Chapters 9 and 10 that some ing-

complements derive from participles, others from gerunds and that some may even

result from confusion of the ing-ending with the old en-ending of the infinitive. Dif-

fusion may thus proceed from different sources, which complicates the picture of

diffusional change considerably, giving rise to different diffusional strands, which

may or may not intertwine. Second, if diffusion is lexically determined and if it is

sensitive to lexical semantics, there is not just the diffusing construction to reckon

with, but also the lexical items that host the diffusing construction. These too are in


continual flux and the point at which diffusion reaches a specific item is often not so

much determined by the dynamics of diffusion as by the history of the lexical item

itself (the extension of causative constructions to start discussed above being a case

in point). If it might be assumed that these lexical changes are usually independent

of ongoing diffusional processes, they still pose a challenge to the student of diffu-

sional change, as lexical developments are bound to cut across the (supposedly) neat

progression of diffusion.

3. Conclusions

The preceding discussion has pieced together the various elements that a model of

diffusional change in the system of complementation has to contain. On this basis,

we can formulate hypotheses about what we may expect to find in the diffusion of

for...to-infinitives and ing-complements. A new construction establishes a gravita-

tional pole that can attract new items (cf. Krug 2000). Presumably, the more fre-

quent the construction gets, the stronger its gravitational pull, and the more items it

can recruit, by which means it spreads to an ever growing number of environments.

The diffusion to new environments does not occur in an unprincipled way, however.

In general, the environments that first ‘fall victim’ to the spreading pattern are ex-

pected to be the environments for which the new pattern is most strongly sanctioned,

on the basis of existing constructions and their frequencies, as well as any other fac-

tors that also affect synchronic usage. In many cases, new uses are sanctioned on the

basis of their resemblance – in terms of semantic or paradigmatic analogy – to the

already established uses from which the pattern spreads. At the same time, other

(sanctioning) factors may work to hinder or confuse this process of analogical exten-

sion. Blocking, for instance, causes some environments to resist change because

they are strongly associated with a competing pattern. In other words, sanctioning is

not a uniform principle, but incorporates the complexities of synchronic usage. Fur-

ther complicating factors are the possibility of there being more than one source

from which a pattern may spread, and the lexical histories of the environments

themselves to which diffusional change occurs. In the detailed case-studies that fol-

low in Parts III and IV, this model of change in the system of complementation can

now be put to the test.

PART III For...to-infinitives

“language is a real-time activity,

whose regularities are always provisional

and are continually subject to negotiation,

renovation, and abandonment.”

(Paul Hopper)

CHAPTER 5: A brief history of for...to-infinitives

The first diffusing construction to be subjected to closer scrutiny is the for…to-

infinitive. Under for…to-infinitive is understood here the kind of structure illustrated

in (1), where a noun phrase introduced by for (henceforth a for-NP) functions or can

be interpreted to function as the subject of a following to-infinitive, with which it

forms a single clausal constituent. Though probably not unrelated historically (see

below), for…to-infinitives are not to be confused with forto-infinitives, a typically

Middle and Early Modern English construction, where for functions as an infinitival

reinforcer and usually immediately precedes the to-infinitive, as in (2).

(1) a. It was neither my intention or aim for this to happen (CB)

b. I would rather they got off and we knew the truth than for them to die

and us never to know. (CB)

(2) Then from that day foorth, they tooke counsell together for to put him

to death. (1611, PPCEME)

Starting from the examples in (1), it is immediately clear that in Present-Day English

the for…to-infinitive can fulfil a number of different syntactic functions – in (1a),

for instance, the for…to-infinitive is an extraposed subject, while in (1b) it can be

analysed as a prepositional complement to than. The purpose of this chapter is to

describe the various functions for…to-infinitives can have and to give a broad pic-

ture of the historical development of for…to-infinitives in their different functional

slots. This description serves as a background to the more detailed investigation of

the diachronic development of for…to-infinitival verbal complements – an issue

reserved for Chapter 6 – and at the same time provides a first application of the prin-

ciples of diffusion introduced in Chapter 4.

The rest of this chapter looks as follows. Section 1 briefly discusses the pro-

cedures for data gathering and the typical problems encountered in dealing with

for...to-infinitives. Section 2, by way of introduction, gives an overview of the kind

of syntactic functions for…to-infinitives can fulfil. Section 3, finally, looks into the

overall historical development of for…to-infinitives in Modern and Present-Day

English. As will be demonstrated, not only for…to-infinitival complements but

for…to-infinitives in general, are a diffusing construction, and the way in which dif-

fusion has proceeded can be seen to be at least in part determined by the way

for…to-infinitives arose and by the characteristics of the grammatical environments

132 – CHAPTER 5: A brief history of for...to-infinitives

to which they have spread. It will thus be shown that patterns of diffusion are not

unmotivated – even if no single factor is alone responsible for the order of diffusion

that is revealed by the data.

1. Methodology

The discussion in this as well as the following chapter is largely based on data from

PPCME, IMEPC, PPCEME, CLMET and the two parallel corpora LOB and FLOB,

with additional examples collected from CB. Wherever quantitative information is

given, figures are based on searches for the form for followed by to, separated by

five words or less. This means that, even though for…to-infinitives with for-NPs of

more than five words have largely slipped through the mazes of the net, it can be

assumed with a good degree of confidence that the corpus searches have not been

biased by any preconceived ideas about the use and distribution of for…to-

infinitives. That is, there is no reason to believe that results are not representative of

the overall use of the for…to-infinitive.

To be sure, the investigator’s interpretation of the data still butts in early on in

the process of data gathering, as the immediate results of the corpus searches contain

a good deal of corpus ‘noise’ and call for careful further sifting of examples. The

most distressing problem in this is to make sure that a given sequence of a for-NP

and a to-infinitive is in fact a for…to-infinitive. Whether or not this is so is often

unclear because for-NPs and to-infinitives can both function independently as con-

stituents, and unfortunately the traditional tests to tell for…to-infinitives apart from

non-for…to-infinitival for…to-sequences (see esp. Jørgensen 1975) can only con-

firm that a given sequence is ambiguous. The for…to-sequences in (3), for example,

could be counted as for…to-infinitives, but it might also be argued that in (3a) the

for-NP is a benefactive to the preceding adjective good and the to-infinitive alone

functions as extraposed subject and as referent of anticipatory it, or that the for-NP

in (3b) marks an addressee and the to-infinitive alone functions as appositional com-

plement to call.

(3) a. I don’t think it’d be good for them to know that I am around you know

(CB)

b. A call for the West to come now to Gorbachev’s aid was heard today

across the region. (CB)

CHAPTER 5: A brief history of for...to-infinitives – 133

One possible solution is to accept only those examples that cannot but be genuine

for…to-infinitives. However, this would inevitably entail rejecting a great number of

examples on uncertain grounds (cf. Wagner 2000), thereby considerably impover-

ishing the data. Moreover, one soon is to face the complication that not all ambigui-

ties are equally compelling – on purely intuitive grounds, it may be found for in-

stance that in (3a) above the non-for…to-infinitival reading is less farfetched than in

(3b). The alternative procedure to follow is to include examples liberally, accepting

ambiguous cases as for…to-infinitives next to the more clear-cut examples, simply

on the ground that a for…to-infinitive reading is possible. Clearly, less interpretative

intervention is needed this way, which gives this approach some attractiveness.

Frustratingly, however, ambiguity is not equally rife in all syntactic environments

(as will appear further below). Because of this it is not only difficult to compare de-

velopments across syntactic contexts but even to get a firm grip on developments

within environments. Generally, in trying to work out how for…to-infinitives have

evolved in different syntactic contexts, I have therefore first collected all examples

that could be for...to-infinitives, but at the same time it has occasionally proven use-

ful to draw on only part of the available material to tease ambiguous instances apart

from non-ambiguous cases wherever this is possible. To avoid terminological confu-

sion, the term for…to-infinitive is used here to refer to both the ambiguous examples

where for may be a subject-marker but also a preposition, and the non-ambiguous

examples where for cannot but be a subject marker. When reference is made to ei-

ther of these sub-sets, the former are labelled ambiguous for…to-infinitives (or

for…to-sequences), the latter unambiguous or genuine for…to-infinitives.

If a given for…to-sequence is decided to be a for…to-infinitive, the problem

that arises next is to figure out the function of that for…to-infinitive, and here too

decisions may be hard to make. Fortunately, at this point there is only one sensible

procedure to follow, which makes the choices to be made less ambivalent. For the

purpose of corpus counts, for…to-infinitives that are ambiguous between two or

more functions have simply been listed as separate categories, so that the major

functional categories only include examples whose functional status is straightfor-

ward (even if their internal structural status as a genuine for…to-infinitive might be

debatable). For example, for anyone to answer in (4a) might be analysed as modify-

ing the adjective simple (the interpretation of (4a) is then equivalent to ‘here is a

question that is simple for anyone to answer’) or it may be considered a relative

postmodifier to question (in which case the example should be read as ‘here is a

simple question that anyone may answer’). Similarly, in (4b) for apartheid to be


given a strong moral basis could be treated as a complement to the noun pressure or

as a purpose adjunct to the entire preceding clause.

(4) a. Here’s a simple question for anyone to answer (LOB)

b. But it can be reasonably hoped that responsible leaders of those two

powerful branches of the Dutch Reformed Church will bring

increasing pressure to bear on the Government for apartheid to be

given a strong moral basis. (LOB)

Listing functionally ambiguous cases separately increases the inventory of ‘func-

tions’ to be kept track of, but otherwise it is the only way to get a reasonably reliable

picture of how for…to-infinitives have evolved across environments.

2. Functions of the for…to-infinitive

With the exceptions of Poutsma (1926) and Visser (1963-73), little attention has

been paid to the considerable range of functions in which for…to-infinitives can

occur – presumably because the functions of the for…to-infinitive are almost en-

tirely encompassed within the functional range of the to-infinitive. It is nevertheless

helpful to give an overview of those various functions, in order to get the terminol-

ogy straight for the discussion to come, as well as to get a first taste of the construc-

tion’s syntactic flexibility, which foreshadows the intricacy of the historical devel-

opments it underwent. Based on a collection of examples from LOB and FLOB as

well as CLMET, the following functions can be distinguished – important classifica-

tory difficulties will be pointed out along the way:

SUBJECT: For…to-infinitives can function as clausal subjects, either in a copular

clause, as in (5a), or in a non-copular clause, as in (5b). Importantly, while in the

examples in (5) the for…to-infinitive also fills the clausal subject slot preceding the

matrix verb, the great majority of for…to-infinitives functioning as subject are ex-

traposed, in which case the subject slot is filled by anticipatory it. Clauses of this

kind are usually copular, as in (6a-b), but may also be non-copular, as in (6c).

(5) a. Under the circumstances, for Laura Herbert to encourage his courtship

was an act of most uncharacteristic rebelliousness. (FLOB)

b. For the principle to apply requires that four conditions be fulfilled.

(FLOB)


(6) a. In chapter 3 of Book V Aristotle defines what it is for one thing to be

continuous with another. (FLOB)

b. But it’s rare for two people to talk about murder on a personal level.

(FLOB)

c. Because facility with serious art requires skill and knowledge not ac-

quired incidentally, it makes sense for the school to offer a programme

of art education. (LOB)

SUBJECT-COMPLEMENT: Rather less common but not unattested is the use of for…to-

infinitives as subject-complement. Two examples are given in (7a-b), where the

meaning of the construction is identifying. In the extremely infrequent use in (7c),

by contrast, a for…to-infinitive is used to attribute a function to the subject – i.e. to

say what the subject is for.

(7) a. when he was a boy in Bromsham the custom was for the young men to

gather thorn branches the night before May Day, and these they

planted in front of the door of all the unmarried women of the village.

(LOB)

b. A practical solution would be for the Commonwealth to draft a set of

principles excluding race discrimination. (LOB)

c. Then he undid the chain, and showed her how to fasten it round her

neck, and to hide it away under the edge of her frock. “It’s for you to

keep, you know,” he said, “not for other people to see.” (1889,

CLMETEV)

OBJECT: For…to-infinitives may serve as direct objects to verbs, as in (8), where

they complement the verbs call and wait.

(8) a. the script called for him to seduce Deneuve (CB)

b. He gave the driver first aid and checked the condition of the other

casualties while waiting for rescuers to arrive. (FLOB)

In addition to these straightforward verbal complements there are a number of other

object constructions that deserve mentioning. With a number of causative and cogni-

tive verbs (make, render, think, consider, etc.), for…to-infinitival objects are obliga-

torily extraposed, with anticipatory it filling the object position, as in (9). Rather

ambiguous in status are the for…to-infinitives following verbal expressions such as


allow time or make room, as in (10), as they might be analysed as complements to

the preceding verbal idiom, as postmodifiers to the nominal element of the verbal

idiom, or sometimes even simply as adjuncts. Finally, for…to-infinitives may be the

focus of a cleft sentence, as in (11), in which case their function is ambiguous be-

tween those of verbal complement and subject-complement.

(9) a. Unfortunately, he was mother-dominated to an extent which made it

impossible for him to manage his life or get the full value from his tal-

ents. (LOB)

b. we consider it appropriate for me, as temporary Family head, to take

over the control of things in general (CB)

(10) a. All I ask is that you allow time for the required improvements to be

implemented and for them to show through (CB)

b. So something had to be done; either new markets had to be found for

the gut rot wines, or the outmoded wine-making traditions had to be

overturned to make room for ‘new style’ wines to develop – wines ca-

pable of holding their own in the international market. (FLOB)

(11) What we want is for the Government to force the brewers to either

allow us to buy the pubs in a competitive market, or to rent them at a

commercial rent (FLOB)

ADJUNCT: The for…to-infinitives functioning as adjuncts fall into two major groups.

First, there are adjuncts of purpose, as in (12a-b), and – closely related but much less

common – adjuncts of result, as in (12c), as well as conditional adjuncts in modal

clauses, as in (12d).

(12) a. For a marriage to be valid under French law one of the partners has to

have lived for a month in the area where the ceremony takes place.

(FLOB)

b. Rules may originate in conventions but they must then be elevated

above conventions in order for the game to proceed at all. (FLOB)

c. Inevitably, it was Fondriest who cracked and picked up the pace again,

only for Museeuw to sprint past him with 300m to the line to gain a

comfortable win. (CB)

d. She had written the note and left it in the summerhouse. Which meant

they must have found it after she’d gone. But for Daniel to bring it


with him to England they must already have appreciated its ambiguity.

(FLOB)

Second, there are adjuncts of comparison. Where these are used, the possibility or

impossibility of the situation in the for…to-infinitive functions as a point of refer-

ence against which some qualification of degree or quantity can be measured. The

qualification in question is expressed by the grading items too, enough and suffi-

cient(ly), with which the for…to-infinitive works in tandem. Some examples are

given in (13). Notice that when used in combination with sufficient there may be

ambiguity with adjectival complementation or nominal postmodification, as in

(13d).

(13) a. The war surplus aircraft Mr Murphy is worried about were thrown

from ships in about 300m of water in 1946, too deep for ordinary di-

vers to reach. (CB)

b. He […] knew that if she cried she would make it loud enough for him

to hear. (LOB)

c. It is to be hoped that sufficient mutual respect exists between them for

such an exercise to be entirely counter-productive. (FLOB)

d. if someone has been stabbed and the police have a description of the

suspect, information as to where they have fled and an idea of the

weapon used, these are sufficient grounds for an arrest to take place.

(CB)

NOMINAL POSTMODIFIER: For…to-infinitives that are immediately dependent on a

noun may function either as a relative clause to that noun or as a nominal comple-

ment. In the former case, illustrated in (14a), the noun usually controls a secondary

participant in the for…to-infinitive clause. Nominal complement uses, on the other

hand, further fall apart into two categories, as the for…to-infinitive may be used

appositively, spelling out the symbolic content of the communicative act denoted by

the head noun, as in (14b), or may simply complete the semantics of the head noun,

as in (14c-d).31

31 Delimiting noun complementation or adjective complementation (see below) is more difficult than delimiting verbal complementation because noun and adjective complements, unlike verbal complements, are not strictly obligatory. As a conse-quence, in these domains, complementation becomes something of a fuzzy category (see also Chapter 10). I follow Keizer (2004), however, in assuming that comple-ments still distinguish themselves in that the semantic structure of the head on which


(14) a. At the end of this chapter are examples of radio and television scripts

for you to read as an exercise. (FLOB)

b. This was, of course, untrue, the “negotiation” was a flat demand for a

Soviet officer to be brought to whom they would talk and with whom

they would, no doubt, establish their identity. (LOB)

c. The war presented the opportunity for the whole area to be requisi-

tioned and to begin the development of a civil airport under the pretext

that it was needed as a base for the RAF. (FLOB)

d. Although wildland fires in America are increasingly fought with water-

carrying helicopters and planes spraying fire retardant, there is still a

need for people to fell trees and dig dirt. (CB)

Notice that for…to-infinitives following nouns such as time or place are often am-

biguous between relative postmodifiers and noun complements. In (15a), for in-

stance, if time is interpreted as ‘occasion, opportunity’, the for…to-infinitive might

be analysed as a complement filling in what it is time for, but it is also conceivable

that an entire prepositional phrase containing the noun time has been gapped in the

for…to-infinitive so that the structure really involves relative postmodification (‘the

time at which the black community must show its support’). More problematic still

are constructions introduced by it is time, as in (15b), for in such examples the

nominal status of time is questionable as well.

(15) a. Now is the time for the black community to show its support. (CB)

b. It is time for us to get off the fence – to speak up, and to vote. (CB)

ADJECTIVAL MODIFIER: Though by no means the most frequently attested, the class

of adjectival modifier constructions is certainly the most complex. A first group of

they depend, be it a verb, noun, or adjective, raises the expectation of some extra participant whose semantic role they more or less pre-specify. For precision’s sake, it should be added to this that semantic expectations can be reversed. As shown by Goldberg (1995), the presence of an argument can impose a certain interpretation on the main predicate rather than vice versa. I hypothesise that this reversal of the se-mantic relations in a construction can be thought of as an instance of form-function reanalysis (Croft 2000), akin to the processes of meaning relocation discussed in Chapter 3. At the same time, lacking obligatoriness, patterns of noun and adjective complementation do not appear to reach the appropriate levels of semantic predict-ability for such a reversal to become possible.


uses distinguish themselves by the fact that the adjective can be interpreted as stand-

ing in a predicative relation to the for…to-infinitive. Thus, although predication as

such is usually not the main point of the message (rather, these for…to-infinitives

tend to specify in what respect the quality of the adjective holds), uses of this type

can be recognised by the possibility of paraphrasing the construction by an extrapo-

sition construction in which the for…to-infinitive serves as extraposed subject and

the adjective as subject complement, predicating a quality of the situation in the

for…to-infinitive. Consider, for instance, the pair of examples in (16):

(16) a. These little things, trifling as they may seem, are yet very necessary for

a politician to know (1748, CLMETEV)

b. it is very necessary for a politician to know these things

Uses of this type have in common that the for…to-infinitive contains a gap con-

trolled by an element of the matrix clause (e.g. these little things in (16a)). Important

to see, further, is that the adjective may occur in predicative position, as in (16a), but

may also function as a postmodifier or premodifier to a noun, as in (17a) and (17b-c)

respectively. At this point, however, some serious classificatory difficulties come in,

for when the adjective precedes the noun it modifies, the construction as a whole is

highly prone to ambiguity with the relative postnominal modification construction

(see also (4a) above). For example, on a relative modification reading (17b) can be

interpreted as meaning ‘among houses for a drunken man to put up at this is a bad

one’, while on an adjectival modification reading it can be taken to mean ‘it would

be bad for a drunken man to put up at this house’. As I have come across no fully

convincing examples where the second reading is the only possibility, it is assumed

that constructions of this type are invariably ambiguous, presenting a merger of sorts

between two more basic construction types, predication on the one hand and rela-

tivisation on the other.

(17) a. I pay hard-working, conscientious ladies to teach this child things right

and proper for her to know. (1909, CLMETEV)

b. It would also be a bad house for a drunken man to put up at. (1889,

CLMETEV)

c. Could the strike have been broken? An unusual question for a marxist

sic! to ask perhaps, but the strike is very much an instance where the

capitalist state adopted the celebrated pose of ‘the dog that did not bark

in the night’. (FLOB)


A second group of uses allows no predicative paraphrase of the relation be-

tween the adjective and the for…to-infinitive that depends on it. They are found with

three kinds of adjectives. The first set of adjectives express a quality related to grad-

able categories such as size or duration, as in (18a-b). The second set of adjectives

occur as subject complements to animate subjects, and typically describe a state of

mind, as in (18c). The third contains adjectives that also occur as predicates in ex-

traposition constructions yet exceptionally invite no extraposition paraphrase, as in

(18d).

(18) a. Oakly paid his attorney ten golden guineas, remarked that it was a

great sum for him to pay, and that nothing but the love of justice could

make him persevere in this lawsuit about a bit of ground, “which, after

all,” said he, “is not worth twopence. (1796-1801, CLMETEV)

b. Besides, with only one leg exposed, there is but a very small object for

a snake to strike at. (1894, J.J. Astor, A journey in other worlds)

c. Meanwhile, Labour are anxious for Mr Major to set an election date at

the Tory conference in Blackpool next month. (FLOB)

d. […] it is nevertheless an expression of the traditional teaching of the

Church, based on St. Thomas who 600 years earlier had taught that a

degree of material well-being is necessary for a person to live a virtu-

ous life. (FLOB)

The latter two uses, illustrated in (18c-d), can be classed together as adjectival com-

plement uses, since the for…to-infinitive can be said to complete the meaning of the

adjective (see footnote 31). Notice that the for…to-infinitives in these uses are not

gapped. By contrast, for…to-infinitives with adjectives expressing size or duration,

as in (18a-b), are gapped. Their use partly resembles the use of adjuncts of compari-

son but also resembles that of for…to-infinitives with premodifying adjectives dis-

cussed earlier, showing the same ambiguity between adjectival modification and

relativisation (albeit without the possibility of a predicative paraphrase).32 For that

reason, this last use too may be thought of as a constructional blend.

32 Only very occasionally do for...to-infinitives of this kind not qualify an adjective in premodifying position. A possible example is given in (i).


As if things were not complicated enough, there is a final adjectival modifica-

tion construction that is transitional between those uses implying predication and

those not implying predication. The for…to-infinitives in (19) are used as adjectival

complements and invite no predicative paraphrase, yet they are gapped and (with the

exception of ready in (19c)) belong to the class of adjectives that in other contexts

can be predicated of a for…to-infinitive.

(19) a. In the 1870s R. F. Horton’s father, himself a Congregational minister,

refused to let his son enter a Congregational college because, he wrote,

“there is not one of them fit for you to enter. […]” (FLOB)

b. the unoccupied apartments, which I don’t pay for, may serve as lum-

ber-rooms, if I have anything to put in them; and they are very useful

for my little boy to run about in on rainy days when he can’t go out

(1848, CLMETEV)

c. and here has this spot remained, perhaps for centuries, all ready for

man to live in, and to enjoy whenever he should come to it. (1841,

CLMETEV)

Under the heading of adjectival modification, then, fall at least five different func-

tions: for…to-infinitives can function as gapped and non-gapped adjective comple-

ments; they may function as modifiers of respect, in which case they are gapped and

allow a predicative reading; and they can occur in one of two constructional blends

(both involving gapping), one between modifiers of respect and relativisation, the

other between relativisation and adjuncts of comparison.

OTHER USES: For the sake of completeness, two other uses should be drawn attention

to. First, for…to-infinitives may depend on prepositions (albeit hardly prototypical

ones). There are only three prepositions found with for…to-infinitives in the corpus

material, viz. than, as and but, the latter occurring only once. Some examples are

given in (20) (for an example with than see (1b) above).

(20) a. we should perceive it to be as impossible for him not to exist, as for

twice two not to be four. (1751, CLMETEV)

(i) What it means in practical terms is that our exports to Germany will

now be a little cheaper for Germans to buy, while the goods which Germany exports will be made a little dearer. (LOB)


b. there is nothing for them but for the father to hang his head and the

mother to cry herself to sleep (1890, CLMETEV)

c. to reduce the rotation to a spiral, but to a spiral with so little deviation

from perfect cycularity as for each revolution to appear practically a

cycle (1880, CLMETEV)

In general, the prepositional complement uses have in common that the interpreta-

tion of the for…to-infinitive is linked to the interpretation of some other clausal con-

stituent. In (20a), it is construed in parallel to the earlier for…to-infinitive for him

not to exist; in (20b) it relates back to nothing. The example in (20c) is an exception

in this respect, in that the use of the for…to-infinitive resembles that of a clause of

comparison, as discussed earlier.

Second, a construction that resembles the for…to-infinitive is found in exam-

ples as those in (21). Mind that constructions of this kind are not for…to-infinitives

proper. Although neither the for-NP or the to-infinitive can be shifted to other

clausal positions, it is not easy to conceive of for…to-sequences of this kind as sin-

gle constituents either, as this violates the semantics of the pattern (roughly ‘it is

NP’s duty to VP’). Notice further that the for-NP is obligatorily animate. The best

conclusion that can be drawn is that these sequences present a strongly idiomaticised

pattern, the relation of which to genuine for…to-infinitives is dubious.

(21) a. It remains, however, for us to look in more detail at theories of syllabic

quantity (FLOB)

b. It is not for us to say how and when change will come. It is for us to

speak our minds and at the same time reason with those who govern

China. (FLOB)

3. Historical developments

Allowing that the various functions discussed in the preceding overview are all

grammatical in Present-Day English, the natural question to ask is how the present-

day state of affairs came about. The current section looks into two aspects of this

question, on the one hand examining how for…to-infinitives first arose (Section

3.1), on the other taking a bird’s-eye view of their subsequent development in dif-

ferent functional slots (Section 3.2).


3.1. The origins

Starting where we left off, consider again the range of functions in which for…to-

infinitives are found to occur. On closer inspection, these functions can be subdi-

vided into predicand and non-predicand uses. Predicand uses are those uses where

something is predicated of a for…to-infinitive, and encompass the for…to-infinitives

functioning as (extraposed) subject, the for…to-infinitives functioning as extraposed

object, and the adjectival modifier uses where predication is implied. Non-predicand

uses represent the somewhat less coherent residue of functions, including the func-

tions of non-extraposed object, adjunct, noun postmodifier and adjectival comple-

ment. The distinction between predicand and non-predicand uses is of some impor-

tance to the history of for…to-infinitives. It is in predicand uses that for…to-

infinitives are commonly claimed to have emerged, and although I believe this rep-

resentation of the emergence of for…to-infinitives to be somewhat inaccurate, it is

beneficial to maintain the general distinction in describing the construction’s devel-

opment.

Received wisdom has it that the genesis of the for…to-infinitive is a prime

example of reanalysis and subsequent actualisation, whereby – to adopt Visser’s

(1963-73) terminology – for developed from an ‘organic’ preposition to an ‘inor-

ganic’ subject marker. Importantly, this change is generally assumed to have taken

place in predicand environments, more specifically in constructions containing a

benefactive for-NP controlling an extraposed to-infinitive. The view that this envi-

ronment was the locus of reanalysis was first articulated by Jespersen (1940: 302-6)

and Visser (1963-73: §914) and has been repeated without any significant alteration

by a number of authors (including Harris & Campbell 1995: 62; Haspelmath 1998:

324-5; Newmeyer 1998: 241).

For clarity, the various steps in the assumed development are here recon-

structed from present-day usage, taking the examples in (22) as representative of the

different stages. Example (22a) illustrates the original situation or source structure,

where an organic for-NP and a to-infinitive occur in immediate consecution but each

behaves as a separate constituent. That the for-NP is not part of the to-infinitive is

evident from the possibility to omit or replace either the for-NP or the to-infinitive

without producing an ungrammatical sentence. Semantically, it is clear that the for-

NP specifies the participant to whom the preceding adjective applies (the thief’s

breaking in is difficult for the thief). Example (22b) is more ambiguous: the extra-

posed subject of the sentence could be either the simple to-infinitive or the whole

sequence of for-NP and to-infinitive. In the former interpretation the for-NP is or-


ganic and marks the benefactive of good; in the latter interpretation, it is the subject

of the to-infinitive and a new (implicit) benefactive could be inserted to complement

good, e.g. good for society. Example (22c) very plausibly contains a genuine

for…to-infinitive, because despite the fact that the adjective impossible can also take

organic for-NPs, it is unlikely that the inanimate NP the poison should be used in

this way. Example (22d), finally, illustrates the extension of the construction to new

environments – wise being an adjective that does not normally take for-NPs. Here

for must be inorganic.

(22) a. Window locks can make it extremely difficult for the thief to break in

without making a lot of attention-drawing noise. (CB)

b. It is generally recognised to be good for people to own their own

houses. (LOB)

c. I can assure you that short of a conspiracy among these three it is abso-

lutely impossible for the poison to have been administered in Wynter’s

breakfast. (LOB)

d. In these cases it is wise for patients to be taken to casualty first. (CB)

In brief, (22a) illustrates the source construction, (22b) the locus of reanalysis, and

(22c-d) illustrate different steps in the actualisation of the new pattern.

The view of the change implied by the foregoing description is, however,

somewhat imprecise and certainly incomplete. As far as predicand for…to-

infinitives are concerned, the first unambiguous instances can – with Jespersen

(1940: 309) and Fischer (1988) – be dated to the first half of the sixteenth century,

although ambiguous examples occurred earlier. By way of illustration, (23) presents

some early for…to-sequences in the relevant environment; (24) the first unambigu-

ous predicand for…to-infinitives.

(23) a. Þus endeþ þe secunde degre of Contemplacion, in Holi writ; wherof

and [i.e. ‘if’] þou take good hede, hit schal ben liht for þe to holden

eueri sarmoun. (c1390, PPCME2)

b. hit is a foule þing for a kyng to iangle moche at þe feste and nouȝt fiȝte

in batayle. (1387, PPCME2)

(24) a. by the whiche I do perceyve that the Gentilwoman beyng accompaned

with your said doughter unto your howse, hath informed you that it

was my mynde for hir to certyfye you that the Controwler of the


Pryncesse howsolde dothe bere hys synguler favour to your said

doughter. (1538, PPCEME)

b. it is not possible for all things to be well (1551, Jespersen 1940: 309)

c. But forasmuch as it is parte of thy medicine, for the to know these

thynges, although I haue lytle leysure to do it, yet I wyl endeuor my

selfe to declare somwhat thereof. (1556, PPCEME)

Looking beyond predicand environments, however, it transpires that unambiguous

for…to-infinitives also occurred in other uses in the sixteenth century. These other

for…to-infinitives would either function as postmodifier to a noun, adjunct, adjec-

tive complement, or verb complement. Example (25a) very plausibly illustrates a

for…to-infinitive postmodifying a noun (maners or wayes); (25b) illustrates a

for…to-infinitive either postmodifying a noun (hertes), complementing an adjective

(apte),33 or functioning as purpose adjunct; and (25c) illustrates a probable for…to-

infinitive either functioning as verb complement (to had wrytten) or as purpose ad-

junct.

(25) a. goddes wyll and plesure, beyng set fast in the towre, or profounde

altitude of hys simplicitie or puritie, hath appoynted many maners or

wayes, for thynges to be done: (1556, PPCEME)

b. Nowe clottinge them, by breakynge their stonie hertes, and by making

them supple herted, and makyng them to haue hertes of fleshe, that is

soft hertes, and apte for doctrine to enter in. (1549, PPCEME)

c. And when he came for his ox, he answered him and sayd; Sir John

Roclife had wrytten for certayne tenaunts to be so taryed by him, and

spirred [i.e. ‘asked’] him, whose tenaunt he was (1502-4, PPCEME)

What is more, unlike the for…to-infinitives functioning as extraposed subjects, the

patterns illustrated in (25) are in fact older than the sixteenth century. As shown in

(26), in Late Middle English, too, one finds early for…to-infinitives functioning as

adjunct or noun postmodifier.

33 Example (25b) marginally allows an organic reading of for, since apt could take an independent for-NP. However, if the for-NP in (25b) is analysed as a separate constituent, it is unclear what function should be attributed to the following to-infinitive, which makes the for…to-infinitival reading more plausible. The same can be said about (26b).


(26) a. Also it ys a certayn techinge for hele [i.e. ‘health’] to be keped, þat a

man vse metys þat accordyn to his complexioun and nature (1400-49,

IMEPC)

b. and whan tyme was, the cordes were cutt / and the Trumpetis blew vp,

for euery man to do his deuoir (1450-99, IMEPC)

c. the Bysshop of Norwych makyth but delayes in my resonable desyre

for an eende to be had in the xxv. marc of Hykelyng (1400-49,

IMEPC)

The explanation given so far for the emergence of for…to-infinitives has evidently

not done full justice to the historical facts, in that the environment commonly

claimed to have given rise to the new structure was not the environment where the

new structure actually first appeared.

On more closely examining the distribution of for…to-infinitives in different

syntactic environments in the early sixteenth century, this finding is confirmed, and

SYNTACTIC FUNCTION ± inorganic for + inorganic for

Predicand uses:

Extraposed subject 50 8.6 2 0.3

Adjective postmodifier 12 2.1 0 0.0

Extraposed object 2 0.3 0 0.0

Non-predicand uses:

Noun postmodifier (relativisation) 10 1.7 0 0.0

Noun postmodifier (complementation) 8 1.4 1 0.2

Adjunct of comparison 3 0.5 0 0.0

Purpose adjunct 2 0.3 3 0.5

Verb complement 2 0.3 0 0.0

Purpose adjunct / noun postmodifier 2 0.3 1 0.2

Purpose adjunct / verb complement 1 0.2 2 0.3

Purpose adjunct / adjective complement

/ noun postmodifier

0 0.0 1 0.2

TOTAL 94 16.2 10 1.7

Table 5.1. Absolute and relative frequencies (per 100.000 words) of for…to-

sequences with organic/inorganic for and unambiguously inorganic for in PPCEME

(1500-1570).


it is further discovered that there is something quirky about the role attributed to

ambiguity in the traditional account of the emergence of the for…to-infinitive. Based

on the first sub-period in PPCEME, Table 5.1 gives the number of for…to-infinitives

found in different environments, distinguishing between predicand and non-

predicand uses and separating instances where for is unambiguously inorganic from

instances where it may still be organic. As indicated above, the predicand uses also

include (potential) for…to-infinitives functioning as extraposed direct objects, or as

adjectival modifier.

From Table 5.1, it is clear at once that most instances with unambiguously

inorganic for are found among the non-predicand uses. At the same time, it also ap-

pears that there is no correlation between the incidence of ambiguous examples and

the environments where unambiguously inorganic for first turns up. Predicand uses

do in fact account for the great majority of ambiguous instances, but only for a mi-

nority of the unambiguous cases. Indirectly, what this suggests is that it might not be

ambiguity between organic and inorganic for that is in itself responsible for the ini-

tial development of the for…to-infinitive. Reanalysis, in other words, is not the most

promising avenue of explanation.

How did for…to-infinitives arise then? Looking at the earliest for…to-

infinitives, one recurrent feature strikes the eye: the majority of early for…to-

infinitives is passive. This is apparent from the examples already given above – see

(25a, c) and (26a, c) – and is illustrated anew in (27):34

(27) a. Moyses at all tymes had recourse to þe tabernacle for doutes &

questions to be assoiled, & fled to þe helpe of praier for releving of

perels & of myschaunces of men. (15th c., IMEPC)

b. And for vengaunce to be taken of the same / Reynawd sendeth you

worde by vs, that he shall hange tomorowe rycharde of normandy vpon

the gret gate of his towne and thus shall be doon of all your men that

he shall take. (1450-99, IMEPC)

Needless to say, an account in terms of reanalysis through ambiguous for…to-

sequences would be hard put to explain the prevalence of passive clauses. But the

same prevalence does hint at an alternative explanation.

34 In late Middle English usage, passive instances account for some 73% of all ex-amples of unambiguous for…to-infinitives (i.e. 11 out of 15 instances, including 3 instances – 2 passive and 1 active – where an organic reading is marginally possi-ble). This figure is based on a 33% sample from IMEPC.


To-infinitives underwent a number of changes in the course of the Middle

English period. One was the introduction of (periphrastic) passive infinitives, as

illustrated in (28).

(28) leue me to ben I-wreken On him þis shome me haueþ speken. (?a1300,

HC)

‘let me be avenged on him that delivered this shame on me’

Another change is the introduction of for (lost again in standard varieties in the

Early Modern period) as an infinitive marker reinforcing infinitival to, as in (29)

(Fischer 2000).

(29) Ne cam ic noht te ȝiuen ȝew for-bisne of mire aȝene wille to donne, ac

i cam for to donne mines fader wille. (a1225, HC)

‘I did not come to do my own will in setting you an example, but I

came to do my father’s will’

Finally, as a result of overall changes in word-order, Middle English also saw the

gradual loss of infinitives with fronted objects. Structures as in (30), where an object

is fronted and ends up in between the infinitive markers for and to, thus disappeared.

(30) a. men must suffre for better to haue (1450-99, IMEPC)

b. and he besohte at gode þat naht ne scolde reinin [i.e. ‘rain’], for ðe

folke to kastin [i.e. ‘chastise’]. (1200-49, IMEPC)

c. for none envy ne yvel have I drawe this mater togider; but only for

goodnesse to maintayn, and errours in falsetees to distroy. (1400-49,

IMEPC)

We can now see what may have happened. When infinitives with fronted ob-

ject as in (30) became less acceptable two solutions were available: either to move

the object to post-verbal position or to turn it into a subject by passivising the verb.

Both solutions bring the ‘forto-infinitives’ with fronted objects in line with the new

rigid SVO order of English. The last solution of course gives rise to passive for…to-

infinitives as those found in (27) and other examples above – or in (31) below,

where the old structure (the people is the object of susteyne) stands side by side with

the new pattern (alle soules is the subject of be redemed). In this view, the first

for…to-infinitives are a syntactic blend between forto-infinitives with fronted object


and passive infinitives, motivated as an analogical realignment under the pressure of

a new word order pattern.

(31) That god that created the firmamente, and made alle thynges of

noughte, for the people to susteyne / And in the crosse suffred deth and

passyon for alle soules to be redemed out of the peynes of helle, kepe

and saue the / kynge charlemayne, emperoure of Almayne and kynge

of Fraunce, and all his noble lynee / (1450-99, IMEPC)

This account fits Fischer’s (1991) more general proposal on the motivating factors

underlying the emergence of passive infinitives. She states that “the strong pressure

exerted by the grammar on NPs before infinitives to take the role of subjects ulti-

mately led to the introduction of passive infinitives on a large scale” (1991: 174).

Further corroborating evidence comes from the distribution of the new for…to-

infinitives. The fact that the new for…to-infinitives primarily function as adjuncts

echoes Warner’s (1982: 123) observation that the older forto-infinitives are particu-

larly current in adjunct position – which is confirmed by the fact that nearly all

forto-infinitives with fronted object attested in the corpus data clearly function as

adjuncts (compare the examples under (30) above; see also Table 5.1).

No object-fronted forto-infinitives are attested in predicand environments in

the Middle English or early sixteenth-century data, nor do the same data show a pre-

dominance of passive for...to-sequences, so the appearance of for…to-infinitives

here must be accounted for in some other way. It cannot be logically excluded that

for…to-infinitives, once they had come into existence, extended from non-predicand

to predicand contexts, or that predicand for…to-infinitives have been modelled on

NP-to-infinitives, as in (32), which developed earlier in the same environment

(Fischer 1988).

(32) hit is nat commendable one knyght to be on horsebak and the other on

foote. (1470-85, Visser 1963-73: 964)

This ignores, however, the great number of ambiguous for…to-sequences found in

the latter environment and the role they may have played in the emergence of predi-

cand for…to-infinitives. There are two relevant points to be made in this respect:

first, independent of their status as for…to-infinitives proper, the for…to-sequences

in predicand contexts appear to function as a construction in their own right; second,

the relation between such for…to-sequences and genuine for…to-infinitives can be


described in terms of constructional polysemy, with semantic ties running across the

organic/inorganic divide implicated by constituent structure.

The constructional behaviour of for…to-sequences in predicand environments

consists primarily in a strong collocational tie between the for-NP and the following

to-infinitive. For Present-Day English, it has been shown by Wagner (2000) that the

presence of a for-NP with some nouns or adjectives strongly predicts a following to-

infinitive. A very similar argument can be set up for Middle English, at least for

some predicates. For example, in the Middle English data from IMEPC the propor-

tion of for…to-sequences following (im)possible to autonomous for-NPs is 13 to 2,

or about 87%. This means that even if a for-NP is organic and syntactically inde-

pendent of the following to-infinitive, for-NP and to-infinitive are already tied up

collocationally. In that light, it is plausible that the sequence of for-NP and to-

infinitive is stored and exploited by speakers as a single chunk of language, a semi-

fixed routine. Further note that the incidence of ambiguous for…to-sequences is

strongly tied to predicand environments (cf. Table 5.1 above). Distributional skew-

ing of this kind is unlikely to be a coincidence. Again, it makes sense to assume that

for…to-sequences in this context already were a productive construction before the

emergence of predicand for…to-infinitives.

More interestingly still, there have been cases where the occurrence of a for-

NP was fully dependent on a following to-infinitive (Garret 2004) A very clear ex-

ample is the adjective lawful, which combined with for…to-sequences early on in

the Middle English data, invariably licensed an organic reading for the for-NP, yet

never combined with an autonomous for-NP, the latter appearing only in the six-

teenth century. The explanation here may lie in the phenomenon of horror æqui, for

when in Middle English an independent benefactive did appear with lawful it took

the form of a to-NP, suggesting that the replacement of a to-NP by a for-NP before a

to-infinitive conveniently avoided the repetition of to. Still, the very possibility of

replacing to by for may have been sanctioned by the availability of for…to-

sequences as a conventionalised grammatical unit and would in turn definitely have

contributed to the pattern’s further conventionalisation.

In light of all this, there is certainly some justification in treating the for…to-

sequences in predicand contexts as a conventionalised construction. Once we take

this step, it is also possible to see how this construction, through semantic extension

and its own internal dynamics, could drift into manifesting itself as an unambiguous

for…to-infinitive. That is, the for…to-sequences and for…to-infinitives in predicand

environments are tied together by gradients of constructional polysemy. Specifi-


cally, there are two pathways of gradual semantic extension that may account for the

shift from organic to inorganic for in predicand contexts.

The first pathway pertains to for…to-sequences with adjectives that mark

relations in the modal domains of potentiality and necessity, most commonly impos-

sible, possible and necessary, as in (33)

(33) a. For Rasis sayth, Whose face is not seemely, it is vnpossible for him for

to haue good manners. (1548, PPCEME)

b. And if a yonge husbande shulde bye all these thynges, it wolde be

costely for hym: wherfore it is necessarye for hym to lerne to make his

yokes, oxe-bowes, stooles, and all maner of plough-geare. (1534,

PPCEME)

It is generally known that items in the modal domains of potentiality and necessity

tend to undergo specific semantic developments from agent-internal potentiality or

necessity to agent-external potentiality or necessity (Goossens 1987, 2000; Van Der

Auwera & Plungian 1998). The modal meanings involved are known for their inher-

ent fuzziness (Coates 1983), and any attempt at establishing a cut-off point between

one sense and another is likely not only to be defeated by the data but also to actu-

ally misrepresent speakers’ linguistic knowledge.35 In this light, it is no surprise that

constructions with a modal adjective and a for…to-sequence as in (33) should un-

dergo a semantic development whereby a modal force (marked by the adjectival

predicate) inherent in an agent (marked by the for-NP) turns into a modal force af-

fecting that agent from an external source. For example, the possibility and necessity

in (33a) and (33b) respectively may either inhere in the referent of the for-NP, or

affect that referent from without. This semantic development yields the same result

as the transition presupposed by the reanalysis of for from organic to inorganic, but

– as the fuzziness of the modal domains implies – involves no particularly abrupt

transition.

The second pathway consists in a gradual opening up of constructions with

predicand for…to-sequences to accommodate a greater range of predicative nouns

and adjectives – that is, the transition from organic to inorganic for is a reflex of a

diffusional change driven by step-wise extensions through semantic analogy (cf.

35 The reason is as philosophical as it is linguistic: a person is responsible for and identified with the situation he or she is part of. Just as readily as a situation can thus be seen to be a natural extension of its primary participant, a participant-internal modal force can be interpreted as a situational element.


Israel 1996; see also Chapter 4). Eventually, the adjectives and nouns combining

with predicand for…to-sequences have ended up forming a highly complex family

resemblance category, as depicted in Figure 5.1. For example, lawful shares seman-

tic traits with proper (both lawful and proper denote social desirabilty), possible

(lawfulness implies potentiality) and necessary (lawfulness implies necessity), but

not with safe or useful. Similarly, lawful relates to customary (lawfulness is deter-

mined by and creates custom), which in turn relates to common (both customary and

common denote a form of predictability or habit). At the same time, lawful and

proper are semantically linked to right (all three imply moral judgement), which in

turn relates to adjectives such as rational or wise (similarly implying judgement,

though more in attitudinal than in moral terms). The crucial point is that these se-

mantic relations bridge the divide between organic and inorganic uses of for. There

Figure 5.1. Semantic map of predicates selecting predicand for…to-sequences (over-

lapping predicates share semantic features).

POSSIBLE

CONVENIENT

EASY

ADVANTAGEOUS

USEFUL

LAWFUL

PROPER

NECESSARY

NATURAL

CUSTOMARY

GOOD

COMMON

RIGHT

RATIONAL

SAFE


are only two sub-classes of predicates that more or less resist combination with or-

ganic for – predicates expressing external judgement, e.g. rational, and predicates

expressing commonality, e.g. common – yet both of these are semantically linked to

other predicate types that do readily allow for…to-sequences with an organic read-

ing. The inorganic interpretation of for, then, is a corollary of semantic extension in

the fillers of the predicate slot in predicand for…to-sequence constructions, and

arises as this construction extends.

An interesting reflex of the latter mechanism of change is that the data occa-

sionally suggest vagueness between the two possible readings of a for…to-sequence

rather than simple binary ambiguity. This is specifically the case when the main

predicate expresses some form of judgement. In (34), for example, it is on the one

hand difficult to separate the for-NP from the following to-infinitive (e.g. by front-

ing) or even to think of the noun foly as capable of taking an autonomous benefac-

tive (either in the form of a for-NP or a to-NP), yet on the other the referent of the

for-NP will still be interpreted as being held accountable for the gret foly which

characterises his actions. Because neither the organic nor the inorganic reading are

thus fully satisfactory, (34) is better treated as vague than as ambiguous.

(34) By thys ye may lerne yt ys gret foly for a master to put a seruant to that

besynes wherof he can nothing skyll and wherin he hath nat be vsyd.

(1526, PPCEME)

The uncertainty concerning the status of for in (34) can be taken to reflect an inter-

mediate stage in the bleaching of the prepositional semantics of for, occasioned by

the pressure of an expanding range of main predicates combining with for...to-

sequences.

In summary, the genesis of for…to-infinitives is a fairly complex process.

Contrary to the commonly held view, for…to-infinitives first appeared in non-

predicand environments, especially in adjunct function, probably as a result of the

realignment of object-fronted forto-infinitives in concordance with the new SVO

word order of English. More or less simultaneously, however, Middle English saw

the appearance of an organic for…to-sequences in predicand environments that

formed a productive construction in its own right. This construction too, through its

own internal dynamics, could give rise to for…to-infinitives, which dutifully ap-

peared about halfway the sixteenth century. This accounts for the two environments

where for…to-infinitives are found at the beginning of the Early Modern period.


3.2. Developments in Modern and Present-Day English

Obviously, the history of the for...to-infinitive does not stop with its introduction as

a clause-type around 1500. For...to-infinitives grew more frequent in the environ-

ments where they appeared and at the same time spread to new environments (see

also Cuyckens & De Smet 2007). To get some feel of the developments in the use of

for...to-infinitives in Modern and Present-Day English, this section examines the

distribution of for...to-infinitives over different functional slots over time. This is

done mostly by comparing three historical periods: the period 1500-1570 (based on

PPCEME), the period 1710-1780 (based on CLMET) and the period 1961-1990s

(based on LOB/FLOB). Developments in the use of for...to-infinitives are at times

difficult to interpret because there is sometimes no fully satisfactory way of dealing

with ambiguous for...to-sequences (cf. Section 1). Still, some clear tendencies can be

discerned, showing that for...to-infinitives are a diffusing pattern and also that diffu-

sion is syntactically constrained, the new pattern spreading faster in some syntactic

environments than in others. In addition, the grammatical conditioning of diffusion

seems to be non-arbitrary. There are indications, first, that the diffusion of for...to-

infinitives proceeded faster in environments that resembled the environment from

which they spread and, second, that the potential presence of ambiguous sequences,

though not directly relevant to the first appearance of for...to-infinitives in Middle

English, played a part in furthering the pattern’s subsequent diffusion in Modern

English.

As a first major division, for...to-infinitives were, on the whole, faster to dif-

fuse in predicand than in non-predicand environments. This may look surprising,

seeing that the very first instances of genuine for...to-infinitives showed up in non-

predicand environments. However, it is to be recalled that in predicand environ-

ments the appearance of for...to-infinitives is fuelled by a productive construction,

while in non-predicand environments for...to-infinitives appeared as a result of ana-

logical realignment in a construction that was on the verge of disappearing (that is,

the Middle English forto-infinitive). It is probably this difference that explains why,

once they had made their entry on the scene, for...to-infinitives caught on faster in

predicand than in non-predicand environments. In what follows I will therefore start

by examining the diffusion of for...to-infinitives in predicand environments (Section

3.2.1) and then turn to non-predicand environments (Section 3.2.2).


3.2.1. Diffusion in predicand environments

Recall that the for...to-infinitives in predicand environments include for...to-

infinitives used as extraposed subjects, non-extraposed subjects, extraposed objects

and adjective modifiers.36 Table 5.2 gives the frequency of each of these con struc-

tion types for the three sub-periods, allowing us to make out some important quanti-

tative developments. These are, first, the increase and subsequent decrease of

for...to-infinitives in extraposed subject position; second, the decrease of for...to-

infinitives functioning as adjective modifier; third, the very sharp increase in for...to-

infinitives in extraposed object position between the first and second period, fol-

lowed by stability between the second and third; and fourth, the similarly sharp rise

of for...to-infinitives in non-extraposed subject position between the second and the

third period.37

Judging by the figures in Table 5.2, different grammatical environments have

yielded to the spread of the for...to-infinitive at different points in time. Specifically,

taking stability or a decrease in frequency as a sign that diffusion has run its course

for a given environment, the order of diffusion appears to be as follows:

Extraposed subject

Adjective modifier > Extraposed object > Non-extraposed subject

To be sure, we cannot take the figures in Table 5.2 entirely at face value, as they

represent both unambiguous for...to-infinitives and ambiguous for...to-sequences,

SYNTACTIC FUNCTION 1500-1570

(PPCEME)

1710-1780

(CLMET)

1961/1990s

(LOB/FLOB)

Extraposed subject 52 89.7 368 108.6 153 76.5

Non-extraposed subject 0 0.0 2 0.6 12 6.0

Extraposed object 2 3.4 50 14.7 35 17.5

Adjective modifier 12 20.7 13 3.8 3 1.5

Table 5.2. Absolute and relative frequencies (per 1 million words) for for...to-

infinitives in predicand environments in three historical periods.

36 Not included here among adjective modifier uses are those uses where the for...to-infinitive is ambiguous between a predicand adjective modifier and a relative clause or a clause of comparison (cf. Section 2). 37 These changes correspond to the differences between periods that are statistically significant using a Fisher’s exact test.


and for that reason potentially include an unknown number of misclassified exam-

ples. Indeed, this uncertainty affects also our interpretation of underlying factors in

the order of diffusion. Observe, for a start, that for...to-infinitives in non-extraposed

subject position are invariably unambiguous, because fronting the entire for...to-

sequence to subject position requires that the for...to-sequence be a single constitu-

ent.38 In all other predicand positions, by contrast, for...to-sequences can be ambigu-

ous. It is possible, therefore, that the order of diffusion suggested by Table 5.2 is a

red herring, an artefact of the data. This would be the case if developments in genu-

ine for…to-infinitives are obscured by independent quantitative developments in the

ambiguous for…to-sequences. On the other hand, if the order of diffusion implied by

Table 5.2 turns out to be correct, this would indicate that the environments sanction-

ing organic for (i.e. extraposed subject, extraposed object, adjective modifier) are

also ahead in sanctioning inorganic for.

To the extent that potential distortions in the data can be filtered out, the order

of diffusion suggested by the figures in Table 5.2 appears to be accurate. That is,

within the environments allowing organic for, fluctuations in the number of ambigu-

ous instances do not seem to cover up any steady rise on the part of the unambigu-

ous instances after 1780. This is first of all apparent for the class of for...to-

infinitives in extraposed subject position. From these a number of subgroups can be

singled out where the status of for as organic or inorganic is relatively certain.

Among these are the for...to-infinitives combining with adjectives expressing cus-

tom or habit (e.g. common, rare, unusual, etc.), where for is typically inorganic;39

the for...to-infinitives with adjectives of attitudinal judgement (e.g. absurd, rational,

wise, etc.), where for is vague or inorganic;40 for...to-infinitives with adjectives ex-

38 Jørgensen (1975: 134) gives a possible counterexample with organic for in a fronted for...to-sequence, rendered here as (ii). Examples of this kind are extremely rare though, and unattested in my own data. (ii) For the citizens to appeal to the police for protection became more

dangerous than to suffer assault and robbery in silence (1955, Jørgen-sen 1975: 134).

39 The full group of adjectives of custom or habit (as found in the corpus data) con-sists of common, commonplace, customary, exceptional, normal, rare, traditional, uncommon, unheard of, unusual, usual. 40 The full group of adjectives of attitudinal judgement (as found in the corpus data) consists of absurd, acceptable, arrogant, competent, contrary to reason, criminal, decent, discourteous, immoral, inappropriate, inconsistent, in order, irrational, jus-


pressing difficulty (difficult, easy, hard, simple), where for almost invariably allows

an organic interpretation;41 and for...to-infinitives with adjectives of possibility (pos-

sible, impossible), where for is usually thoroughly ambiguous. Taking these groups

as test-cases, we find no strong evidence showing that genuine for...to-infinitives are

still on the increase in extraposed subject position after the beginning of the Late

Modern period.

To demonstrate this, Table 5.3 gives the frequencies of for...to-sequences

with the four groups of adjectives for CLMET (1710-1780) and for LOB and FLOB.

While Table 5.3 shows a significant decrease in the ambiguous for...to-sequences

following possible and impossible, and a significant increase in the vague or inor-

ganic for...to-sequences following adjectives of attitudinal judgement (i.e. absurd

etc.), for...to-infinitives appear to be more or less stable with the adjectives of diffi-

culty and the adjectives of habit (i.e. common etc.). This tells us a number of things.

First, there is no strong correlation between inorganic for and quantitative increase

during the Late Modern period (true, the absurd-group increases, but the common-

group appears to have been firmly established already in eighteenth-century Eng-

lish). Second, there is no straightforward indication that organic for is unstable (wit-

ness the stability of the difficult-group). Third, changes in the frequency of for...to-

infinitives in extraposed subject position appear to be due to fluctuations within

lexically specified sub-groups (compare the rise of the absurd-group to the fall of

the possible-group). Fourth, if we take the increase in the absurd-group as an indica-

tion that inorganic for...to-infinitives were still a spreading pattern in subject extra-

position contexts at the beginning of the Late Modern period, they are nonetheless

ahead of the non-extraposed subject for...to-infinitives in Table 5.2 above. In gen-

eral, it appears that the various combinatorial possibilities of for...to-infinitives had

been more or less established by the beginning of the Late Modern period and that

later shifts in the frequency of for...to-sequences in this environment are mostly due

to shifts in lexical preferences, without any strong correlation with lexical items’

associations with organic or inorganic for.

tifiable, self-contradictory, rational, refreshing, ridiculous, unpardonable, unrea-sonable, vain, wise, wrong. 41 Regarding these, Jørgensen (1975: 134) states: “In the present inquiry 81 exam-ples have been found with ‘easy’, ‘difficult’, and ‘hard’ as predicative complements, and in none of them does the latter [i.e. inorganic] interpretation seem possible or likely.” Even so, occasional counterexamples occur, with inorganic for following easy or difficult, witness (iii). (iii) it’s fairly easy for tourist visas to be issued for major destinations or at

least places worth visiting. (CB)


1710-1780

(CLMET)

1961/1990s

(LOB/FLOB)

Possible etc. 151 44.7 32 16.0

Difficult etc. 39 11.5 29 14.5

Common etc. 31 9.2 15 7.5

Absurd etc. 11 3.3 17 8.5


infinitives in extraposed subject position with four groups of adjectives.

If not identical, the situation with extraposed objects is fairly similar. Impor-

tantly, in this environment there is a very persistent tendency for for...to-sequences

to allow an organic interpretation when the matrix verb is causative (make, render).

This is so even in Present-Day English, as is illustrated in (35a).42 Among the extra-

posed for...to-infinitives with cognitive verbs such as believe, consider or think,

however, organic for is more readily licensed, already in the Late Modern period, as

is illustrated in (35b-c).

(35) a. his diffidence and his conservative temperament made it virtually im-

possible for him to adapt himself to the very different life of the Brit-

ish capital. (LOB)

b. A person, who has hunted a hare to the last degree of weariness, would

look upon it as an injustice for another to rush in before him, and seize

his prey. (1751, CLMETEV)

c. If you had known the friendship which hath always subsisted between

the colonel and my husband, you would not imagine it possible for any

description to exceed it. (1751, CLMETEV)

With respect to quantitative developments in the course of Late Modern and Present-

Day English, however, for...to-infinitives functioning as extraposed objects with

42 In LOB and FLOB, only 1 out of 33 instances must be analysed as having unam-biguously inorganic for. The example is given in (iv): (iv) The congestion made it impossible for the market to operate efficiently

and after much argument it was decided to abandon the idea of re-building Old Covent Garden and to look for another site. (FLOB)


1710-1780

(CLMET)

1961/1990s

(LOB/FLOB)

Causative verbs 45 13.3 33 16.5

Cognitive verbs 15 4.4 2 1.0


infinitives in extraposed object position with causative and cognitive verbs.

causative verbs remained stable, whereas they decreased significantly with cognitive

verbs, as is shown in Table 5.4. This indicates again that in as far as genuine, non-

ambiguous for...to-infinitives spread to extraposed object position, diffusion had

more or less run its course by the beginning of the Late Modern period. What we

can conclude at this point is that for...to-infinitives – whether ambiguous or unambi-

guous – did appear earlier in extraposed subject and extraposed object positions than

in non-extraposed subject positions. In other words, for...to-infinitives did not in-

crease at the same rate and over the same period in all predicand environments, and

it appears that those contexts that sanctioned organic for were also more welcoming

to inorganic for.

There is, however, another way of splitting out the data that reveals some

further instructive trends. Notice, for a start, that among the extraposed subjects is a

subgroup of uses that invariably call for an inorganic interpretation of for, namely

uses where the matrix clause is non-copular, as in (36) (see also (6c) above). Simi-

larly, non-extraposed subject for...to-infinitives can be divided into subjects of copu-

lar and non-copular clauses, as in (36b-c) (see also (5) above).

(36) a. It has taken a long time for academic concern to reach the policy arena

but current international interest in the ‘polluter pays principle’ pro-

vides an example. (FLOB)

b. For would-be tenants to claim some further rights over a landowner’s

freehold property is impertinent and illogical. (FLOB)

c. For him to become involved in this blackmailing business would only

serve to continue the relationship between then them, even if in a non-

sexual fashion. (FLOB)

It is worth considering the evolving frequencies of these three constructions

throughout the Late Modern and Present-Day periods. The relative frequencies of


0

1

2

3

4

5

6

7

8

9

1710-1780 1780-1850 1850-1920 1961/1990s

non-extraposed,copular

extraposed, non-copular

non-extraposed,non-copular

Figure 5.2. Relative frequencies (per 1 million words) of for...to-infinitives in three

types of subject position in Late Modern and Present-Day English (based on

CLMET and LOB/FLOB)

the three patterns are charted in Figure 5.2. It is immediately clear, first, that unlike

the more familiar extraposed subjects in copular constructions (see Tables 5.2 and

5.3), for...to-infinitives in extraposed non-copular subject position only began to

appear at the beginning of the Late Modern period, which means that they were be-

hind on other extraposed subject for...to-infinitives yet in step with for...to-

infinitives functioning as non-extraposed subject. Second, among non-extraposed

subject for...to-infinitives, the for...to-infinitives in non-copular clauses are again

behind on those in copular clauses, not appearing in the data until after 1850.

The pattern of diffusion witnessed in Figure 5.2 offers an excellent illustra-

tion of the sanctioning principle discussed in Chapter 4. If we assume that for...to-

infinitives spread from extraposed subject positions in copular clauses, the environ-

ment that shares fewest features with the source environment is that of non-

extraposed subject position in non-copular clauses, whereas non-extraposed copular

subject positions and extraposed non-copular subject positions take an intermediate

position. The degree to which environments differ from the source environment cor-

responds to the degree to which they accept the spreading construction.

For predicand environments, the picture is now beginning to clear up. Two

general tendencies can largely account for the timing and pace of quantitative

changes in the use of for...to-infinitives in different syntactic contexts. On the one


hand, for...to-infinitives appeared earlier in environments that more closely resem-

bled the environment in which they first arose – that is, extraposed subject of a

copular clause. On the other, for...to-infinitives appeared earlier in environments that

licensed organic for. The first tendency explains why for...to-infinitives spread faster

to extraposed objects, extraposed non-copular subjects and non-extraposed copular

subjects than to non-extraposed non-copular subjects. The second tendency explains

why for...to-infinitives spread faster to extraposed object positions than to non-

copular and/or non-extraposed subject positions, all of which inevitably impose an

inorganic reading on for. Alternatively, we could say that extraposed object posi-

tions sanctioned the diffusion of for...to-sequences regardless of the distinction be-

tween organic and inorganic for, whereas non-copular and/or non-extraposed subject

positions sanctioned the appearance of only the inorganic subset of for...to-

sequences.

3.2.2. Diffusion in non-predicand environments

In addressing the issue of diffusion in non-predicand environments, it is important to

point out from the start that the database considered here is somewhat impoverished

as instances of functional ambiguity have been withheld (cf. Section 1). Most rele-

vantly, among adjective complements are reckoned only complements proper;

among verb complements are reckoned only complements to simple verbs (i.e. not

verbal idioms) and the complements in cleft clauses have been similarly excluded;

for...to-infinitives that are ambiguous between purpose adjunct and noun modifier

(either noun complement or relative clause) are likewise excluded, and so are the

noun modifiers following nouns such as time when they show ambiguity between

relativisation and complementation; finally, excluded also are examples where

for...to-infinitives show ambiguity between adjective complementation and compari-

son following sufficient. (For examples of these ambiguous patterns, see Section 2

above.)

This being said, Table 5.5 details the diffusion of for...to-infinitives in non-

predicand environments, giving the frequency of occurrence of for...to-infinitives

functioning as relative clause to a nominal head, as noun complement, as purpose

adjunct, as adjunct of comparison, as verb complement, adjective complement, and

as subject complement. As Table 5.5 shows, even though most non-predicand pat-

terns are attested as early as the sixteenth century, dramatic increases in frequency

took place in the course of the nineteenth and twentieth centuries. In fact, the only


SYNTACTIC FUNCTION 1500-1570

(PPCEME)

1710-1780

(CLMET)

1961/1990s

(LOB/FLOB)

Noun complement 4 6.9 20 5.9 75 37.5

Relative clause 10 17.2 18 5.3 25 12.5

Adjunct of comparison 3 5.2 31 9.1 46 23.0

Adjunct of purpose 5 8.6 5 1.5 36 18.0

Verb complement 2 3.4 6 1.8 95 47.5

Adjective complement 0 0.0 2 0.6 13 6.5

Subject complement 0 0.0 1 0.3 15 7.5

Table 5.5. Absolute and relative frequencies (per 1 million words) of for...to-

infinitives in non-predicand environments in three historical periods.

significant difference between the first and second period in Table 5.5 is a decrease

in the frequency of for...to-infinitives functioning as relative clauses and as purpose

adjuncts. By contrast, a comparison between the second and third sub-period shows

significant increases in all environments. The figures also suggest that these fre-

quency increases in the Late Modern period did not occur at the same pace in all

environments. The fact that some of the increases are more dramatic than others

indicates that in some environments change had already advanced further than in

others by the beginning of the Late Modern period. Following this kind of reason-

ing, the data suggest that diffusion in non-predicand environments proceeded

roughly as follows:

Noun complement

Relative clause

Adjunct of comparison

>

Adjunct of purpose

Verb complement

Adjective complement

Subject complement

Again, however, before we can begin to make sense of these developments,

we have to confront the problem of ambiguous for...to-sequences. Filtering out am-

biguous instances from the data reveals that the quantitative head start of some

for...to-infinitive constructions is exaggerated by a higher incidence of ambiguous

examples in some environments. At the same time, even with ambiguous examples

separated out, the order of diffusion suggested by Table 5.5 remains almost – if not

entirely – unaltered. To show this, Table 5.6 divides the examples from CLMET and

LOB/FLOB represented in Table 5.5 into ambiguous and non-ambiguous instances,


1710-1780

(CLMET)

1961/1990s

(LOB/FLOB)

SYNTACTIC FUNC-

TION

± inorganic + inorganic ± inorganic + inorganic

Noun complement 14 4.1 6 1.8 31 15.5 44 22.0

Relative clause 11 3.2 7 2.1 16 8.0 9 4.5

Adjunct of comparison 13 3.8 18 5.3 17 8.5 29 14.5

Adjunct of purpose 1 0.3 4 1.2 5 2.5 31 15.5

Verb complement 5 1.5 1 0.3 5 2.5 90 45.0

Adjective complement 0 0.0 0 0.0 3 1.5 10 5.0

Subject complement 0 0.0 1 0.3 0 0.0 15 7.5

Table 5.6. Absolute and relative frequencies (per 1 million words) of ambiguous and

unambiguous for...to-sequences in non-predicand environments in CLMET (1710-

1780) and LOB/FLOB.

depending on whether or not an organic reading of for is allowed and contextually

plausible.

Confining ourselves to the unambiguous examples in Table 5.6, the patterns

that are best-represented at the beginning of the Late Modern period and that show

the least subsequent growth are for...to-infinitives functioning as relative clause and

as clause of comparison. The position of for...to-infinitives used as noun comple-

ment and as purpose adjunct is somewhat more difficult to interpret, as these con-

structions appear to be slightly ahead of the for...to-infinitives used as verb comple-

ment, adjective complement and subject complement yet still undergo a very sub-

stantial increase in the course of the Late Modern period. For...to-infinitives in verb

complement, adjective complement and subject complement positions, finally, are

clearly the slowest patterns to emerge. The order of diffusion in non-predicand envi-

ronments can thus be (tentatively) refined as follows:

Relative clause

Adjunct of comparison >

Noun complement

Adjunct of purpose >

Verb complement

Adjective complement

Subject complement

How to account for these differences between grammatical environments?

Much as in predicand environments, it appears that to some extent the potential

presence of ambiguous examples has actually furthered the use of unambiguous

for...to-infinitives. Taking back into consideration the ambiguous for...to-sequences

in Table 5.6, we find that the environments most readily sanctioning organic read-


ings of attested for...to-sequences are adjuncts of comparison, relative clauses and

noun complements. The ambiguities typically encountered are illustrated for each of

these environments in (37).

(37) a. There is nothing now too difficult for me to undertake. (1751,

CLMETEV)

b. there was left a little kind of an esplanade for him and the corporal to

confer and hold councils of war upon. (1759-67, CLMETEV)

c. There would otherwise be no motive for any one of them to enter into

that scheme of conduct. (1751, CLMETEV)

The four other non-predicand environments, by contrast, show no such readiness to

accommodate ambiguous for...to-sequences – neither in early Late Modern English

nor in Present-Day English – generally because they do not depend on lexical items

sanctioning benefactive for-NPs. In other words, the potential for organic for

roughly correlates with the ease with which inorganic for has spread in different

contexts.

The potential for ambiguous for...to-sequences cannot give the whole answer,

however, because it does not tell us why in the diffusion of for...to-infinitives ad-

juncts of purpose are slightly ahead of verb complements, adjective complements

and subject complements; or why noun complements are slightly behind on relative

clauses and clauses of comparison. I see no easy explanation to these subtler effects,

but would suggest that two further factors might be involved. First, as was suggested

earlier (Section 3.1), the first non-predicand for...to-infinitives primarily appeared in

purpose adjunct positions because this was the environment where forto-infinitives

occurred most commonly. In this light, the slight head start of for...to-infinitives in

purpose adjunct positions apparent in Table 5.6 might be interpreted simply as per-

sistence of the initial distribution of for...to-infinitives. Second, a similarity effect

might be invoked with respect to the advantage of for...to-infinitives functioning as

relative clause and as clause of comparison over those in noun complement position.

Specifically, the first two grammatical environments show some connections to the

extraposed subject and adjective modifier positions in which for...to-infinitives in-

creased most quickly. For...to-infinitives used as relative clauses merge with predi-

cand for...to-infinitives in adjective modifier constructions where the adjective

modified is used attributively, as in (38a) (cf. Section 2). For...to-infinitives used as

clause of comparison, on the other hand, occasionally show a close affinity to adjec-

tive modifiers following predicatively used adjectives, as is illustrated in (38b).


(38) a. Next to the two books that I have already mentioned, I do not know a

better for you to read, and seriously reflect upon, than ‘Avis d’une

Mere d’un Fils, par la Marquise de Lambert’. (1751, CLMETEV)

b. the sovereigns and the great lords came universally to consider the

administration of justice as an office both too laborious and too ignoble

for them to execute in their own persons. (1766, CLMETEV)

4. Conclusions

This chapter has discussed the emergence of for...to-infinitives in English, covering

both their first appearance in the language and their subsequent diffusion to new

grammatical environments. For...to-infinitives appeared, probably more or less in-

dependently, first in non-predicand and then in predicand environments. In non-

predicand environments, however, the new construction was relatively marginal and

appears not to have served as a major model for further innovations. In predicand

environments, by contrast, the new construction was grounded in a relatively fre-

quent and productive pattern manifesting itself both in organic and inorganic for...to-

sequences. Given this firm basis, it could also spread to other environments.

By considering the timing and pace of major increases in different grammati-

cal environments, the following rough order of diffusion can be established:

Ext

rap

ose

d su

bjec

t

Adj

ectiv

e m

odi

fier

>

Ext

rap

ose

d o

bje

ct

>

Re

lativ

e c

lau

se

Adj

unct

of c

ompa

rison

>

No

un c

ompl

eme

nt

Adj

unct

of p

urp

ose

>

No

n-e

xtra

pose

d su

bje

ct

Ve

rb c

ompl

em

ent

Adj

ect

ive

com

ple

me

nt

Su

bje

ct c

ompl

em

ent

Closer consideration of the data suggests that this order of diffusion is the outcome

of a number of interacting factors. First, taking the for...to-sequences in extraposed

subject position as the centre from which for...to-infinitives spread, the order of dif-

fusion has been influenced by similarity to the source environment. This similarity

may explain why for...to-infinitives in non-extraposed copular subject position ap-

pear earlier than those in non-extraposed non-copular position, or (more tentatively)

why for...to-infinitives functioning as relative clauses or clauses of comparison de-


velop faster than those functioning as noun complements. Second, it also appears

that for...to-infinitives spread faster to environments that allowed ambiguous for...to-

sequences as well as unambiguous for...to-infinitives. This explains why for...to-

infinitives were quicker to emerge as extraposed object, relative clause, clause of

comparison or noun complement than as non-extraposed subject, verb complement,

adjective complement, or subject complement. Third, a persistence effect may ac-

count for the fact that in the process of diffusion for...to-infinitival purpose adjuncts

are ahead on other environments that neither sanction organic for nor closely resem-

ble the predicand source construction.

CHAPTER 6: For...to-infinitives as verb complements43

As shown in the previous chapter, the diffusion of for...to-infinitives during the

Modern period also led to the appearance of for...to-infinitives in verbal complement

positions. As the present chapter will demonstrate, the process whereby for...to-

infinitives are increasingly used as verbal complements constitutes a diffusional

change in its own right, in close interaction with the constraints characteristic of the

system of complementation.

From what we know of present-day usage, the distribution of for...to-

infinitival complements is lexically determined (Visser 1963-73; Rudanko 1984;

Egan 2003), presenting us with the familiar problem of matching a complement-type

to the right complement-taking predicates and vice versa. The examples in (1) show

the kind of situation Present-Day English confronts us with.

(1) a. we all hoped for the miracle to happen. (Google, 16 November 2007)

b. we all waited for the miracle to happen.

c. *we all believed for the miracle to happen.

d. *we all saw for the miracle to happen.

One question to address in this context is what principles – if any – underlie the dis-

tribution of for...to-infinitives as verbal complements. That is, we have to ask if and

how synchronic organisation is motivated. Yet from the discussion in Chapter 5, we

also know that the distribution of for...to-infinitival complements has been unstable

over time, for...to-infinitives having gained currency as verbal complements only in

the course of the Late Modern period. Therefore, resolving the first question inevita-

bly leads us to a second issue: if a distribution is synchronically motivated, how did

this distribution emerge and how does the diachronic process of emergence relate to

the synchronic principles of grammatical organisation? These two questions are the focus of this chapter. It is shown that the distribu-

tion of for...to-infinitival complements presents a straightforward example of how

simple motivational principles interact to shape the synchronic system of comple-

mentation. The present-day distribution of for...to-infinitives, though not entirely

without its irregularities, is fairly predictable and, on that ground, definitely non-

arbitrary. When next we turn to the construction’s historical development, it can be

shown that the historical trajectory of the pattern’s extension consists in a bearing

43 This chapter has been adapted from De Smet (2007b).

168 – CHAPTER 6: For...to-infinitives as verb complements

out of the synchronic regularities of its combinatorial behaviour. With few excep-

tions, what is innovated is what is maximally sanctioned by the system, with each

innovation setting new horizons to what the synchronic system can generate. The

rest of this chapter looks as follows: Section 1 briefly addresses some methodologi-

cal issues. Section 2 discusses the synchronic distribution of for...to-infinitives used

as verbal complements and analyses the kinds of principles by which present-day

usage is motivated. Section 3 turns to the process of diffusion, describing how diffu-

sion proceeded through Late Modern and Present-Day English and proposing prin-

ciples that account for why diffusion took the course it did. Section 4 combines the

results from the previous sections and shows how they confirm and further substan-

tiate the view of complementation and diffusional change developed earlier (cf.

Chapters 3 and 4).

1. Methodology

Because the diffusion of for...to-infinitival complements did not take off until the

Late Modern period (see Chapter 5), the discussion to follow is primarily based on

CLMET and LOB/FLOB. The search procedures are more or less identical to those

detailed in the previous chapter: the corpora were queried for all instances of for

followed by to within a five-word span. From the resulting hits, all instances where a

for...to-infinitive complements a verb have been selected. No pre-existing lists of

complement-taking predicates have been relied upon, so the results can be taken to

be unbiased and highly representative of actual usage. For Present-Day English,

however, the material has been further enhanced with data from CB and BNC.

Given their size, these corpora could not be searched systematically, so here use has

been made of pre-existing verb lists (Visser 1963-73; Rudanko 1984; Egan 2003)

combined with simple intuition to further complete the inventory of verbal predi-

cates taking for...to-infinitival complements in present-day usage.

In deciding whether specific examples should be accepted or rejected, the

procedures followed are similar to those in the previous chapter. With respect to

ambiguity on the level of constituency, examples that allow a two-constituent read-

ing for the for...to-sequence have been included along with genuine for...to-

infinitives. Given that ambiguity of this kind is not so widespread in the area of ver-

bal complementation (cf. Table 5.6 in Chapter 5) this can hardly have disturbed the

results. Still, some verbs exclusively combine with ambiguous for...to-sequences

(most notably send), in which case their membership of the class of for...to-

infinitive-taking predicates has been considered doubtful. As to functional ambigu-

CHAPTER 6: For...to-infinitives as verb complements – 169

ity, only the for...to-infinitives with simple verbs have been taken into account, and

only if they are found in a genuine object position. This means that for...to-

infinitives following verbal idioms, as in (2a), have been excluded, as well as

for...to-infinitives functioning as (extraposed) subject to a passivised verb, as in (2b),

or for...to-infinitives used in the focus of a pseudo cleft, as in (2c).

(2) a. Some homes are mere dormitories, with only the week-end allowing

time for the parents to meet as a family. (LOB)

b. It would be totally impracticable and nor is it proposed for a Labour

Government to set up hit squads of state funded marksmen with rifles.

(CB)

c. What Wendy really wants is for her and her partner to read each

other’s minds. (CB)

2. For...to-infinitives as verb complements in Present-Day English

The list in (i) gives all the verb types taking for…to-infinitives found in LOB and

FLOB and a good number added from CB. Since for…to-infinitives are productive

as a complement-type, the list is necessarily open-ended, but it should at least pre-

sent a more or less unbiased selection, representative of the for…to-infinitive’s com-

binatorial potential. The verbs for which attestations are suspicious are marked with

a question mark; these are verbs that are never found to combine with unambiguous

for...to-infinitives.

(i) Ache, afford, agree, aim, allow, appeal, apply, argue, arrange, ask,

bear, beg, call, care, choose, consent, crave, cry, cry out, desire, elect,

expect, gesture, hate, hope, indicate, intend, like, lobby, long, look,

love, mean (‘intend’), mention, motion, move (‘propose’) (?), pay,

phone, plan, plead, plot, pray, prefer, prepare, press, provide, push,

ring, say, scream, seek, send (?), shout, signal, thirst, vote, wait, watch

(‘anticipate, wait’), wave, wish, write (‘ask in writing’), yearn, yell

From (i) we can see that the distribution of for…to-infinitives as verbal complements

in Present-Day English is principled or non-arbitrary in at least two ways.

A first principle underlying the distribution involves a form of semantic regu-

larity. Typically, the predicate-complement construction expresses an activity or a

state of mind that is directed to or conducive to the realisation of the event referred


to by the for…to-infinitive (for a semantic characterisation along similar lines see

Egan 2003). Most of the verbs combining with for...to-infinitival complements lend

themselves to this in that the verbs themselves already express some goal-directed

activity or state of mind. Thus, semantic regularity is reflected, first of all, in the fact

that many verbs taking for…to-infinitives express anticipation, volition, or goal-

oriented activities, as shown in (3).

(3) a. Talk to your bank rather than waiting for the bank manager to send for

you. (CB)

b. He has to talk to the country in terms which will make the voters yearn

for him to replace John Major before another year is out. (CB)

c. Perhaps the most deeply disturbing aspect of Dr Tate’s intervention

was his mention on Radio 4’s Today programme that he would seek

for his “ten commandments” to be part of the national curriculum

when it comes up for review. (CB)

Another set of verbs expresses communication. These systematically manifest a di-

rective sense when combined with a for…to-infinitive, and as such conform to the

same principle of semantic regularity as found in other for…to-infinitive-taking

verbs. Compare, in that respect, the use of the for…to-infinitive and that-clause with

the verb indicate, as illustrated in (4).

(4) a. “What now?” she asked looking at a second steel door at the end of the

corridor. “We take over the control room,” he replied then took up a

position at the side of the door and indicated for her to do the same on

the other side. (CB)

b. UN sources, however, indicate that the Secretary-General is unlikely to

be willing to send a mission which did not have the co-operation of the

Israeli government […] (CB)

Equally significant is the virtual absence of for…to-infinitives with the whole class

of believe-type verbs, such as believe, know, think, claim, suppose, and so on. In-

stead, these verbs typically take an infinitive clause with a raised NP, equivalent to a

that-clause. Very occasionally, however, believe-type verbs do combine with a

for…to-infinitive, which then predictably gives rise to the semantic contrast illus-

trated for argue in (5). As the examples show, the contrast again conforms to the


principle of semantic regularity formulated above, and, as such, provides further

evidence for its validity.

(5) a. In any case, I argue this to be true. (CCB)

b. Germany and Denmark argued for it to be made plain to Belgrade that

any breach of the ceasefire by the federal army would prompt the

community to accept the declarations. (CCB)

Likewise, verbs expressing negative volition, such as decline, refuse or forbid

strongly resist combination with a for...to-infinitive. None of these verbs are attested

with for...to-infinitive, either in CB or in BNC. The absence of for...to-infinitives

with these verbs again underscores the goal-oriented semantics of for...to-infinitival

complementation.

It may be pointed out that semantic regularity is most easily described on the

level of the predicate-complement construction. Specifically, some verbs have goal-

oriented semantics outside their use with the for...to-infinitive, whereas others only

seem to acquire a goal-oriented meaning when combined with the for…to-infinitive.

As a result, it is (outside a constructional framework) unclear where exactly the

goal-oriented semantics reside: in the main verb, in the complement clause, or in the

construction as a whole. If we assume that goal-oriented semantics reside in the

for…to-infinitive, we somehow have to deal with for…to-infinitives in other con-

structions that clearly do not have goal-oriented semantics, as in (6).

(6) a. For would-be tenants to claim some further rights over a landowner’s

freehold property is impertinent and illogical. (FLOB)

b. […] the custom was for the young men to gather thorn branches the

night before May Day, and these they planted in front of the door of all

the unmarried women of the village. (LOB)

If we have the goal-oriented meaning come from the verb, we need to posit a

slightly awkward kind of polysemy in verbs such as say or indicate, with one goal-

oriented sense taking for…to-infinitives and one neutral sense not taking for…to-

infinitives. The easiest solution, therefore, is to locate goal-oriented semantics at the

constructional level, with some verbs fitting the selectional constraints imposed by

the matrix predicate slot, and others being made to fit by semantic coercion.

The second principle underlying the distribution of for…to-infinitives in-

volves distributional associations with other constructions. More specifically, the


distribution of for…to-infinitives as verb complements is confined by two other dis-

tributions: that of the to-infinitive as verbal complement, and that of the preposition

for as a marker of prepositional objects. Thus, the verbs selecting for…to-infinitives

form a distributionally homogeneous class, in that nearly all of them also select to-

infinitives, and the great majority of them take prepositional objects with the prepo-

sition for. In fact, in Present-Day English, this tendency is strong enough to predict

that if a verb takes to-infinitives and can form a phrasal verb with for, then it will

also sanction for…to-infinitives. The remarkable parallelism between different dis-

tributions will here be referred to as the principle of paradigmatic regularity. The

principle is illustrated in (7) for the verb ask, whose behaviour is typical of most

verbs taking for…to-infinitives.

(7) a. If the problem persists, ask your doctor to do a swab test to identify the

organism responsible. Ask for the swab to be tested for everything – if

the laboratory is instructed to check only for one or two conditions,

something obscure may be missed. (CB)

b. When you first become unemployed call in at your Unemployment

Benefit Office or Jobcentre and ask to see a New Client Adviser. (CB)

c. Please ask for an explanation if you do not understand what they are

doing and if you wish to know something about your illness or your

treatment. (CB)

Because both the to-infinitive and prepositional for are themselves associated with

directional or goal-oriented semantics,44 the restraining effect of their distributions

on that of the for…to-infinitive can – at least in part – be held responsible for the

association with goal-oriented semantics found for the for…to-infinitive. In this

sense, the two principles underlying the distribution of for…to-infinitives are, on the

whole, mutually reinforcing.

This is not to say that semantic and paradigmatic regularity are always in

agreement, or that they fully determine the actual distribution of for…to-infinitives

over the inventory of complement-taking verbs. One problem is that some verbs

seem to flaw both semantic and paradigmatic regularity. A good example is found in

(8a): semantically, could hardly bear is clearly not directed to (let alone conducive

to) the realisation of the event referred to by the complement clause, while distribu-

44 For the semantics of the to-infinitive, see the discussion in Chapter 3; for the se-mantics of for, see Lindstromberg (1998: 224-5) and Tyler & Evans (2004: 146-54).


tionally, bear does not combine with prepositional objects introduced by for (*bear

for NP). Likewise, the verb afford in (8b) resists combination with a prepositional

object (*afford for NP) and does not readily fit in with the goal-oriented semantic

characterisation of for...to-infinitival complementation.

(8) a. […] the forefront of her attention had been much more pleasantly oc-

cupied. So pleasantly, in fact, that now she could hardly bear for it to

be over. (BNC)

b. Can you afford for your computer not to be fully operational? (CB)

The main problem, however, is of a more subtle nature. Notice, for a start,

that many verbs would seem to fall under the scope of the semantic regularity prin-

ciple, but nevertheless do not take for…to-infinitives. To illustrate this, (ii) lists

some of the verbs that have not been found patterning with for...to-infinitives, nei-

ther in LOB and FLOB, nor in CB.45

(ii) Advise, decide, demand, force, guarantee, offer, order, permit, prom-

ise, propose, recommend, request, suggest, summon, urge, want

Most of the verbs in (ii) would appear to fit well into the semantically homogeneous

class of for…to-infinitive-taking verbs. The reason why the verbs in (ii) do not

(readily) take for…to-infinitives can only be sought in the fact that they would then

violate the principle of paradigmatic regularity. While the verbs in (ii) allow simple

to-infinitives, they are not used in combination with prepositional objects with for

(e.g. *advise for NP, *decide for NP, *force for NP, etc.), and are, in that respect,

different from most other verbs taking the for…to-infinitive. However, if invoking

paradigmatic regularity gets us some way, it also leaves us with a set of verbs that

do not combine with for-objects, but take for…to-infinitives nonetheless. The list in

(iii) is based on (i) above and presents the verbs found violating paradigmatic regu-

larity in Present-Day English. Comparing (iii) to (ii), it becomes clear that in some

cases semantic regularity overrules paradigmatic regularity, while in other cases the

opposite relation holds.

45 To be sure, some of these verbs do occur with for...to-infinitives in Google, which means that the ban is not absolute. At the same time, CB and BNC are sufficiently large for the absence of a syntactic pattern in a given environment to be significant.


(iii) Agree, afford, bear, consent, desire, expect, gesticulate, hate, indicate,

intend, like, love, mean, mention, prefer, say

From this two things can be gathered. First, positing two forms of regularity

rather than one helps to reduce the problem posed by the verbs in (ii) and (iii). This

underscores the necessity of seeing complementation as a dynamic system allowing

for multiple regularities influencing the matching between complement-types and

predicates and sanctioning different complement-predicate matches to different de-

grees. Second, the two forms of regularity still do not explain away the difference

between the verbs in (ii) and (iii). Together these verbs form a fuzzy area where the

application of the different principles underlying the distribution of for…to-

infinitives becomes more or less arbitrary.46 The only possible conclusion is that

while the distribution of for…to-infinitives looks not unmotivated and is even pre-

dictable to some extent, it is by no means deterministic. As I will argue next, this

area of indeterminacy is the space within which language change takes place.

3. The diffusion of for...to-infinitives as verb complements

We already know that the token frequency of for…to-infinitives functioning as ver-

bal complements has increased dramatically in the course of the Late Modern period

(cf. Chapter 5). This is only confirmed by Figure 6.1, which shows the relative fre-

quencies for the target construction in the three sub-periods of the CLMET, and in

LOB and FLOB, showing a steady increase from the beginning of the eighteenth

century until the end of the twentieth. Simultaneously to this rise in token frequency,

for…to-infinitival complements have diffused over the inventory of complement-

46 The absence of for...to-infinitives in direct consecution to the verb want, espe-cially in British English, has been the object of a specific study by Erdmann (1993). Erdmann invokes Givón’s binding scale (see Chapter 3) to explain why want, unlike the semantically related verb like, seems to resist complementation by for...to-infinitives, arguing that want is higher on the binding scale and therefore requires stronger clausal integration than like. This account is unconvincing, however, be-cause want does in fact combine with subjunctive that-clauses, as in (i), which would seem to offer less syntactic integration in the matrix clause than the for...to-infinitive. For an alternative solution, see Section 4. (i) a. I want you should follow your heart. (CB) b. You want I should be like him? (CB)


0

10

20

30

40

50

1710-1780 1780-1850 1850-1920 1961/1990s

Figure 6.1. For…to-infinitives as verbal complements (frequencies per 1 million

words) in CLMET and LOB/FLOB.

taking verbs. To show this, (iv), (v), and (vi) list the verbs taking for…to-infinitives

for each of the three sub-periods of the CLMET. (For the for…to-infinitive-taking

verbs in Present-Day English see (i) above.) Verbs for which attestation is system-

atically dubious are again marked with a question mark.

(iv) 1710-1780: call, long, move (‘propose’), wait

(v) 1780-1850: call, cry, cry out, halloo (?), like, long, pray, proffer, ring,

scream, send (?), wait, watch, wish

(vi) 1850-1920: arrange, ask, care, contrive, cry, fret, hang (‘desire, wait’),

hope, like, long, manage, pray, prepare, ring, scream, seek, send (?),

shout, sign (‘signal’), wait, whistle (‘signal by whistling’) (?), wish,

write (‘ask in writing’), yell

As the three sub-periods of the CLMET are more or less equal in size, it is safe to

conclude from (iv)-(vi) that along with token frequency the type frequency of

for…to-infinitive-taking verbs has undergone a marked increase during the Late

Modern English period.

Of course, it is the process of lexical diffusion apparent from (iv)-(vi) (and (i)

for Present-Day English) that is of particular interest here. On closer examination of

the data, for…to-infinitives appear to have diffused in a non-arbitrary way, reflecting

the principles outlined above for the synchronic distribution of the construction.

That is, the principles of semantic and paradigmatic regularity have been maintained


throughout the development of the construction. On the one hand, diffusion has fol-

lowed a path of minimal semantic variation, with for…to-complements clustering at

all times around sets of semantically related verbs, and expanding to new uses by

minimal steps, so that we can think of the spread of for…to-infinitives as a gradual

process driven by semantic analogy. On the other hand, the mechanism of semantic

analogy has worked more or less within the confines set by paradigmatic regularity

– or its diachronic counterpart, paradigmatic analogy (see Chapter 4). In this way

paradigmatic regularity has, for the greater part, constrained the spread of the new

complement type (although in some specific cases it may have played a more active

role in the diffusional process).

In what follows, the development of for…to-infinitives as verbal comple-

ments will be discussed in more detail, paying attention both to the specific course

of events, and to possible motivating factors. I will first deal with the period before

1850 (Section 3.1), and then with the period from 1850 up to the present day (Sec-

tion 3.2).

3.1. For…to-infinitives as verb complements before 1850

As can be seen from (iv) and (v) above, before 1850 for…to-infinitives concentrated

around verbs of volition and anticipation (like, long, wait, watch, wish) and verbs of

communication (call, cry, cry out, halloo, move, pray, proffer, ring, scream, send).

Notice, first, that in semantic terms these verbs form two fairly homogeneous

groups, that are, moreover, not entirely unrelated to each other: all verbs denote

some kind of expectation on the part of the subject, either felt or expressed. Second,

the principle of paradigmatic regularity is already firmly in place: all verbs in (iv)

and (v) can take to-infinitives, and with the exception of like, halloo, and proffer, all

verbs combine with prepositional objects with for.

Significantly, like, halloo, and proffer all three appear with for…to-infinitives

only at the very end of the period, in Elizabeth Gaskell’s Mary Barton (1848) and

Emily Brontë’s Wuthering Heights (1847). As they violate the principle of paradig-

matic regularity, these extensions are most likely to be semantically-based. By that

reasoning, the time when these extensions occurred is a good indicator of when the

group of for...to-infinitive-taking verbs was perceived by language users as a seman-

tically homogeneous class. Prior to this point, the distribution of for...to-infinitives

may, semantically speaking, have been accidental. Indeed, with the source and target

domains of extension both lying clearly within the area of verbal complementation,


these extensions are also an unambiguous sign that for...to-infinitival complements

began to extend by virtue of their own construction-internal dynamics.

However, retracting back in time a little, the most pertinent question to be

addressed is how the initial distribution of for...to-infinitives arose, prior to the point

when for...to-infinitival complementation turned into a productive construction. That

is, why should it have been exactly the verbs in (iv) above that first took for…to-

infinitives and served as a point of departure for the subsequent process of diffu-

sion? Presumably, the factors responsible for the introduction of the first for...to-

infinitival complements were incidental to the system of complementation proper.

The solution I propose therefore relates back to the more general discussion of how

for...to-infinitives have spread in the course of the Modern period (cf. Chapter 5).

Specifically, I suggest that for...to-infinitives first appeared with those verbs where

they were most strongly sanctioned by the overall use of for...to-infinitives at the

time.

Importantly, there are various ways in which the first for...to-infinitival com-

plements were sanctioned by existing language use. A first sanctioning factor con-

sists in the principle of paradigmatic regularity, which was already inherent in the

use of for...to-infinitives before their actual appearance as verbal complements. In

the domain of noun complementation, for example, the distributional parallels char-

acteristic of verb complementation emerged early on. Nouns such as manner, means,

need, occasion, order, room, time, and way are all found in Early Modern English

with for-NPs, to-infinitives, and for…to-infinitives – as is illustrated for way in the

following set of examples.

(9) a. Yf it were thus, God hadde lefte none ordinarye waye for his ghospell

and fayth to be taught. (1529, OED)

b. there were ij. waies for his deliuery (1526, PPCEME)

c. The readiest way to worke this conclusion, is to tourn that rightlined

figure into triangles (1551, PPCEME)

Indeed, the associations with to-infinitives and for-NPs that make up para-

digmatic regularity came into being as a natural consequence of the appearance and

spread of for...to-infinitives. Concerning the distributional parallelism between

for...to-infinitives and to-infinitives, the mechanisms accounting for the appearance

of for...to-infinitives ensured that the new construction would come to be perceived

as a functional variant of the to-infinitive. In non-predicand environments for...to-

infinitives arose as a word order variant on forto-infinitives (which in turn func-


tioned as a variant of the to-infinitive). In predicand environments, the benefactive

for-NP preceding the to-infinitive was drawn into the to-infinitive clause and lost its

benefactive function, while the infinitive expanded, essentially retaining the gram-

matical function it had had before (e.g. extraposed subject). The result is that from

the very beginning for…to-infinitives could be seen as a special kind of to-infinitive,

and fell entirely within the confines of the distribution of the simple to-infinitive – a

situation that still very much holds true in Present-Day English.

With respect to the distributional parallelism between for...to-infinitives and

for-NPs, it may be assumed that the formal and semantic similarity between for…to-

infinitives and (especially purposive) for-NPs naturally resulted in some degree of

distributional overlap. That for-NPs and for...to-infinitives did in fact get distribu-

tionally and semantically associated outside the domain of verb complementation is

illustrated by occasional examples where a for-NP is coordinated with a for…to-

infinitive and construed as functionally parallel, as in (10).

(10) a. Robsart advised, and Esclairmonde concurred in the counsel, that Lord

Glenuskie should set forth for Vincennes immediately, before there

should be time for any more cabals, or for Queen Isabeau to have

made her daughter repent of having delivered up the signet-ring. (1870,

CLMETEV)

b. “We shall still get plenty of horses. The horses that come to Warwick

do not come for the prize money, but for the public to bet on,” he

added. (LOB)

Another – more compelling – indication of the association between for…to-

infinitives and for-NPs comes from minor extensions beyond the distribution of the

to-infinitive, where for...to-infinitives invariably copy the function and distribution

of for-NPs. An Early Modern English example involves the use of for…to-infinitives

as a topicalising construction, as in (11a), in parallel with prepositional phrases with

for meaning ‘concerning NP’ as in (11b).

(11) a. And for any longer stay to have brought a more quantity, which I hear

hath been often objected, whosoever had seen or proved the fury of

that river after it began to arise, and had been a month and odd days, as

we were, from hearing aught from our ships, leaving them meanly

manned 400 miles off, would perchance have turned somewhat sooner


than we did, if all the mountains had been gold, or rich stones. (1596,

CEMET)

b. And for Your Boast, That you Exposed your Lives and Fortunes in the

Kings Ser-vice, it would have been better for Him if you had not been

so forward; you say, You took great Pains in Perswading to Loyalty. I

do not know what you did for the Father, but I am sure you did it not

for the Son: And for You to Examine the Behaviour of the Church, Is it

not known to the World that they both Perished Together? (1687, LC)

In a similar vein, for Late Modern English, Poutsma (1926) notes the use of the con-

struction illustrated in (12), where the for...to-infinitive copies the meaning of

prepositional phrases with for meaning ‘in favour of NP’.

(12) The father was for his son to go into the army. (Poutsma 1926: 787)

For Present-Day English, the example from FLOB in (13a) is worth pointing out. It

resembles the one in (11a) above but, given the long time-lag between attestations,

probably arose independently. Here the for...to-infinitive is found in the slot nor-

mally reserved for for-NPs in the semi-productive and-now-for-construction, illus-

trated in (13b) and best-known from the Monty Python catch-phrase and film title

And now for something completely different.

(13) a. And now for the performance to begin. (FLOB)

b. And now for our financial report. (CB)

Each of these minor extensions indicates that for...to-infinitives associate with for-

NPs outside the domain of verbal complementation as well as within. Moreover, the

earlier examples indicate that the association was in place before diffusion to com-

plement positions began.

In this light, the extension of for...to-infinitives to verbal complement posi-

tions can be seen as a straightforward instance of paradigmatic analogy. Thus, the

appearance of for...to-infinitives with verbs such as wait, call, cry, and so on was

sanctioned by the linguistic system, simply on the grounds that these verbs could

combine with to-infinitives and had some kind of association with for-NPs in the

form of prepositional objects. Still, paradigmatic analogy underspecifies the range of

verbs for...to-infinitives actually appeared with. In particular, it remains unclear why


the diffusion of for...to-infinitival complements had to begin with verbs such as call

and wait, and none of the other verbs taking both for-NPs and to-infinitives.

As a second sanctioning factor, therefore, I suggest there were a number of

more specific triggers that conspired – together with paradigmatic analogy – to fa-

vour verbs such as call and wait as the most likely candidates for for...to-infinitival

complementation around the beginning of the eighteenth century. First, while inter-

nally ambiguous for...to-sequences were relatively marginal in the area of verbal

complementation (cf. Chapter 5), some ambiguities did occur. Some early examples

of verbs followed by a for…to-sequence show ambiguity between a construction

involving a genuine for…to-infinitive functioning as verbal complement and a con-

struction involving a to-infinitive postmodifying the prepositional object of the ma-

trix verb.

(14) a. While he was going on, waiting for some certain evidence to accom-

pany his discovery, he perceived they were jealous of him, and so he

durst not trust himself among them any more. (1703, PPCEME)

b. How are you! I long for your next letter to answer me. (1742,

CLMETEV)

c. Then he took him by the hand, and led him into a very large parlour

that was full of dust, because never swept; the which after he had re-

viewed a little while, the Interpreter called for a man to sweep. (1678-

84, CEMET)

d. you must know, some time ago, before the change, they had moved

for a committee to examine, and state the public accounts (1742,

CLMETEV)

In other cases, the ambiguity is between a two-place construction and a three-place

construction, the distinction between which is disambiguated under passivisation, as

in (15a), but often imperceptible in active clauses, as in (15b-c).47

47 Interestingly, drawing on examples from POB, it appears that especially in early instances of call and wait with a for...to-sequence the for-NP – even if a three-place reading is contextually less plausible – is typically animate and could have served as an independent prepositional object to the matrix verb if it were not for the follow-ing to-infinitive, as in (ii). This tendency might be attributed to the lingering influ-ence of the three-place construction, and is probably another manifestation of the sanctioning principle (innovations are introduced where they impose a minimal breach on existing usage; see Chapter 4). As (iii) shows, the tendency is less outspo-ken in later usage.


(15) a. the Surgeon depos’d, That he being called for to dress the Deceased,

he found him held up by two Men, and a Bottle of Harts-Horn held at

his Nose (1724, POB)

b. When we had drank together a while, I call’d for the Landlady to give

me Change for the Liquor (1738, POB)

c. At last I got away, and went to Tom Smith, and told him how I had

been abused, and had called for him to help me. (1735, POB)

Second, the use of for...to-infinitival complements may also have been fur-

thered by ambiguities in the grammatical function fulfilled by the for...to-infinitive.

One area of ambiguity is that between verbal complements, noun complements and

purpose adjuncts following verbal idioms, as illustrated in (16).

(16) a. Therefore some of their Masters, or other Traders must either die,

break, or being grown too rich give over their Trades to make room for

some of them to have places to Trade in, besides those that are fur-

nish’d with places by the new Houses. (1685, LC)

b. Wherefore did Phillip the Second of Spain, obtain License of Queen

Mary for his Subjects to Fish upon the North Coasts of Ireland for the

Term of Twenty one Years, for which, was yearly paid One thousand

Pounds into the Exchequer of Ireland as by the Records appear? (1700,

LC)

More interesting perhaps are the ambiguities in for…to-infinitives with verbs of

communication, which may allow an adjunct or a complement reading. In the former

case the verb is intransitive and the for…to-infinitive refers to the purpose of the

action; in the latter the verb is transitive and the for…to-infinitive refers to the sym-

bolic content of a message.

(ii) a. while the Coach was driving along, he never called for the Coachman

to stop, but forced the Door open, and jump’d in. (1732, POB) c. I order’d him to be carry’d to the Round-House, where I waited for the

Deceased to come and confirm his Charge. (1738, POB) (iii) a. THE Federal Government and the ACTU yesterday called for workers

to be allowed to combine sick and bereavement leave for family needs. (CB)

b. Wait for pans to cool before putting them in cold water. (CB)


(17) a. Then I heard some people cry out for me to take care of my self for he

had his hand upon a knife (1744, POB)

b. My landlord hallooed for me to stop ere I reached the bottom of the

garden, and offered to accompany me across the moor. (1847,

CLMETEV)

Combined with the analogical pressure exerted by paradigmatic regularity,

these various ambiguities may have provided an ideal breeding ground for for...to-

infinitival complements to emerge. Most importantly, the combination of factors

may explain why for…to-infinitives, when they appeared as verbal complements,

first appeared with verbs of volition, anticipation, and communication, because the

ambiguities that could lead to for…to-infinitival complements are largely restricted

to this fairly specific set of verbs. Thus, ambiguity between a for…to-infinitive-

complement and an adverbial clause can only occur with verbs that have both transi-

tive and intransitive uses, such as the verbs of communication call and cry, and pos-

sibly the verb of anticipation wait. Likewise, ambiguity between three-place and

two-place constructions is especially plausible with verbs of communication, which

imply an addressee and a message but do not require that both be expressed at the

same time. Ambiguity between a for…to-infinitival complement and a prepositional

object with a relative to-infinitive-clause, on the other hand, can only occur with

verbs such as move, wait, long, or call taking prepositional objects.

Admittedly, other verbs combining with for-NPs existed at the beginning of

the Late Modern English period, and could in theory give rise to the same kind of

ambiguities – or at least some of them. Verbs such as look, seek or prepare, for in-

stance, could in principle be followed by ambiguous for…to-sequences. Notice,

however, that on a scale of agentivity these verbs take highly agentive subjects,

while, at the same time, they typically combine with inanimate objects. One conse-

quence of this is that following to-infinitives are more readily interpreted as taking

the main verb subject as their controller, precluding reinterpretation as a for…to-

infinitive, as shown in (18):

(18) In seeking for phenomena to prove this proposition, I find only those

of two kinds; but in each kind the phenomena are obvious, numerous,

and conclusive. (1751, CLMETEV)


Even some of the verbs that are semantically related to the first for...to-infinitive-

taking verbs do not commonly show the same patterns of ambiguity. By way of il-

lustration, the examples in (19) give the first instances in POB of wish and pray with

for...to-infinitive, showing clearly how both tend to resist alternative interpretations

in which for-NP and to-infinitive function as separate constituents. The delayed ap-

pearance of such verbs corroborates the role of ambiguities in the appearance of the

first for...to-infinitival complements.

(19) a. I wish for one witness to be called and that is Ann Thomas. (1784,

POB)

b. he went down upon his knees, and begged and prayed for me to let him

go (1800, POB)

Thus, the factors that may have led to the use of for…to-infinitives as verbal com-

plements are consistent with the actual verbs first attested with the for…to-infinitive,

implying that the initial distribution of for...to-infinitival complements was not acci-

dental.

In conclusion, it is possible to construct a complex but essentially coherent

account of how the distribution of for...to-infinitives before 1850 came about. Vari-

ous forms of ambiguity conspired with paradigmatic regularity to determine the first

environments where for...to-infinitival complements showed up. This means that the

role of ambiguous for...to-sequences as a furthering factor in the diffusion of for...to-

infinitives is consistent with their role in the overall spread of for...to-infinitives in

the Modern period (cf. Chapter 5), as it appears once more that ambiguous se-

quences prepared the ground for the arrival of the for...to-infinitive in a new lexico-

grammatical environment. A certain degree of semantic regularity may initially have

been an incidental side-effect of this development, but the extensions to verbs such

as like and proffer at the end of the period form an indication that about halfway the

nineteenth century a semantically homogeneous class of for…to-infinitive-taking

verbs was taking form and being recognised by at least some speakers. As the fol-

lowing section will show, the recognition of a principle of semantic regularity will

come to play a more important role in the further history of for…to-infinitival com-

plementation.


3.2. For…to-infinitives as verb complements after 1850

After 1850, for…to-infinitives have begun to appear with a number of new classes of

verbs. Although it is often difficult to prove that these extensions resulted from se-

mantic and/or paradigmatic analogy, all the changes observed seem to be consistent

with at least one of these mechanisms. Let us therefore consider the extensions that

occurred and the possible role played by semantic and paradigmatic analogy.

The first important development after 1850, taking place still before the end

of the nineteenth century, is the extension of for...to-infinitives to verbs of (non-

communicative) activity (arrange, contrive, manage, prepare, seek). This first ex-

tension is in accordance with the principle of semantic regularity. Firstly, there is a

natural relation between the verbs of activity and those of communication, since the

verbs in both groups designate an act (either a directive speech act or a more general

kind of act) that is directed to the realisation of some other event. The affinity is

especially apparent in the verb arrange, which tends to be ambiguous between com-

municative and non-communicative action. To illustrate this, (20a) gives an example

of the verb arrange where it most probably means ‘make an agreement’; in (20b),

by contrast, arrange is ambiguous between ‘make an agreement’ and ‘make prepara-

tions’.

(20) a. On the fifth day the physician came again and gave us a little hope. He

said the tenth day from the first appearance of the typhus would proba-

bly decide the result of the illness, and he arranged for his third visit to

take place on that date. (1859-60, CLMETEV)

b. As soon as ever it’s possible, we’ll arrange for you to live with some-

one who will preserve appearances. (1893, CLMETEV)

Secondly, in a very similar fashion, the category of activity verbs runs over into that

of verbs of volition and anticipation. Two bridging verbs are wait and watch, which

might be considered either verbs of anticipation, or (sometimes) verbs of activity.

Thus, (21a) illustrates wait in the kind of context where waiting can be seen as an

action that makes it possible for another event to take place; (21b) illustrates a simi-

lar use of the for…to-infinitive with the verb watch.

(21) a. She did not wait for Alice to curtsey this time, but walked on quickly

to the next peg, where she turned for a moment to say ‘good-bye,’ and

then hurried on to the last. (1871, CLMETEV)


b. A spider will remain for months watching for the fly to enter its web;

but this quality is to be observed in every animal in the pursuit of its

prey. (1841, CLMETEV)

These ambiguities show that, in semantic terms, the innovative verbs of activity are

only a small step removed from both verbs of communication and verbs of volition

or anticipation.

Noteworthy is the occurrence of the verbs contrive and manage with for…to-

infinitives. This use is only attested in the last sub-period of Late Modern English,

and is entirely absent from Present-Day English. On the face of it, the use of manage

and contrive as for…to-infinitive-taking verbs might suggest that in nineteenth-

century English for…to-infinitives had briefly spread to factive verbs. However,

when the instances in question are considered more closely, they display a strong

(non-factive) ‘arrange’-sense, deriving from the modification by modal auxiliaries.

(22) a. ‘You haven’t been ill since I saw you?’ he inquired. ‘Oh no!’ ‘But you

look as if you might have been. I say, we must manage for you to have

a fortnight off, you know, this month.’ (1891, CLMETEV)

b. “Allow me to present my son,” said the Vice-warden; adding, in a

whisper, “one of the best and cleverest boys that ever lived! I’ll con-

trive for you to see some of his cleverness. […]” (1889, CLMETEV)

In this view, the exceptional use of a for…to-infinitive with manage and contrive

constitutes a fine illustration of semantic analogy: the extension of the construction

only pertains to those uses of the verbs manage and contrive that still fall under the

principle of semantic regularity.

Still, semantic analogy may not have been the only triggering factor in the

emergence of for…to-infinitives with verbs of activity: the principle of paradigmatic

regularity may have played a covert but equally important role. It is presumably no

coincidence that the verb arrange starts combining with for…to-infinitives at about

the same time it has been first attested with prepositional objects introduced by for.

The latter use is unattested in the first sub-period of the CLMET (1710-1780), and

has only a few questionable instances in the second sub-period (1780-1850). It is not

until the third sub-period (1850-1920) that, with some twelve instances, arrange for

NP gains currency. The earliest clear example from CLMET is given in (23).


(23) They had descended one sultry evening on the little inn at Kyle Rhea

ferry; and while Tom and another of the party put their tackle together

and began exploring the stream for a sea trout for supper, the third

strolled into the house to arrange for their entertainment. (1857,

CLMETEV)

Plausibly, the emergence of for…to-infinitives with arrange at the end of the nine-

teenth century was aided by the occurrence of prepositional objects with for – just

as, conversely, the absence of for…to-infinitives before that period may be linked to

the non-occurrence of the same pattern.

Other important developments in the post-1850 period are a number of more

recent innovations. Judging from the verbs in (i) in Section 2 above, the extension of

for…to-infinitives to new complement-taking verbs has continued into Present-Day

English. It must be pointed out, however, that to some extent the type frequencies in

the corpus data have simply increased along with the size of the corpus. The ‘inno-

vations’ exposed by the Present-Day English material may therefore merely be mar-

ginal uses that remained unnoticed in the earlier material.

Still, given their frequency, it should be safe to consider at least the verbs

allow (attested three times in LOB/FLOB) and provide (four times in LOB/FLOB)

to be newcomers to the class of for…to-infinitive-taking verbs. These can be classed

as verbs of condition as they denote a situation that supplies the necessary condi-

tions for the event in the complement clause to take place. Interestingly, these verbs

take inanimate subjects, and in this sense break with the semantic consistency found

so far, as all other verbs taking for…to-infinitives quite strictly designated a mental

state or an activity directed to or conducive to the realisation of the event in the

complement clause. At the same time, semantic analogy cannot be wholly excluded,

because allow and provide still designate a state of affairs that enables the realisation

of the event in the complement clause and are, as such, still semantically related to

other for…to-infinitive-taking verbs – albeit more distantly.

(24) a. The marked slope of the top tube allows for the seat quick release to be

mounted on the front of the seat tube, well away from flying mud and

water – a nice touch. (FLOB)

b. Unlike the present access order, which normally provides for a parent

to have access to the child, the new contact order provides for the child

to visit or stay with the person named in the order. (FLOB)


Again, a second factor motivating the appearance of for…to-infinitives may well be

the systematic application of paradigmatic regularity. For provide in particular, the

role of the corresponding prepositional object construction is hard to ignore, as the

construction with the for…to-infinitive inherits or copies the meaning of provide for

NP (‘create a legal possibility’):

(25) The government’s latest views are contained in a short report on rec-

ommendations made by the House of Commons Foreign Affairs

Committee. It points out that amendments to Hong Kong’s Basic Law

provide for intervention by Peking only if turmoil is beyond the control

of the Hong Kong authorities and endangers China’s national security.

(CB)

This semantic link underscores the idea that the for…to-infinitive has been intro-

duced with provide as a clausal equivalent to the prepositional objects found with

the same verb.

A more marginal Present-Day English innovation is the use of for...to-

infinitives with the verb afford, as illustrated in (26).48 The combination of afford

with a for...to-infinitive is used in negative constructions to indicate that the condi-

tions are not met that would make a certain event’s taking place allowable or toler-

able for the matrix subject. On this ground, afford can be classified along with the

other verbs of condition allow and provide, even though its subject is typically ani-

mate. At the same time, if afford is used in a more literal sense (i.e. ‘have the finan-

cial means to’), as it is in (26b), its use with the for...to-infinitive is reminiscent of

the use of pay in the same kind of construction. Consequently, the verb could some-

times be classified as a somewhat unprototypical action verb as well.

(26) a. It is essential that we make this new parliament work. There are forces

inside and outside of Scotland who are desperate to make it fail. They

cannot afford for it to succeed. (CB)

b. as a young couple with a mortgage they could not afford for him to go

on competing at the top level of sport. (CB)

48 The fact that the combination of afford with for...to-infinitive is not only unat-tested in CLMET but also in CEN and POB suggests that the pattern is indeed a Pre-sent-Day English innovation – not a marginal use that remained between the thresh-old of attestation in the historical data. In the TIME corpus, which covers the larger part of the twentieth century as well as the beginning of the twenty-first, afford does not occur with a for...to-infinitive until after the year 2000.


Another innovation in Present-Day English is the use of for…to-infinitives

with emotive verbs (bear, hate, love).49 These verbs flaw the semantic regularity

found with for…to-infinitive-taking verbs as well as the principle of paradigmatic

regularity (cf. bear in (8a) above). At the same time, it is clear how for…to-

infinitives could spread to these verbs, as volition and emotion are often overlapping

categories (De Smet 2004; De Smet & Cuyckens 2005). Even the verb bear can be-

tray traits of volition when combined with the to-infinitive, as (27) illustrates. In

addition, with the verb like – a verb typically hovering between emotive and voli-

tional meanings – the emotive verbs already had a half-representative among the

for…to-infinitive-taking verbs in earlier periods, as (28) shows.

(27) I think this is quite realistic except during that period when you are ‘in

love’, which I think is a hysterical state where you can’t even bear for

your lover to go to the loo without you going too. (BNC)

(28) My Dear Colonel Keith, – I am just come out of court, and I am to wait

at the inn, for Aunt Ailie does not like for me to hear the trial, but she

says I may write to you to pass away the time. (1865, CLMETEV)

Finally, in addition to these extensions in semantic space, the Present-Day

English period also sees a number of new verbs taking for…to-infinitives that fall

more or less within the already established categories: verbs denoting activities (ap-

ply, elect, lobby, pay, vote), verbs of communication (gesticulate, indicate, mention,

phone, say), verbs of volition (ache, desire, expect, thirst), and verbs that are am-

biguous either between communication and activity (agree, appeal, argue, consent,

plead, push, press), or between volition/anticipation and activity (choose, intend,

look, mean, plan, plot). All this is within the lines of what might be expected on the

basis of the verbs already taking for…to-infinitives: the majority of the new verbs

49 Considering the lower frequency of this pattern (cf. Wagner 2000: 205), its at-testation in the comparatively large-sized synchronic corpora may not reflect real historical change in Present-Day English. As with afford, however, the construction is not only unattested in CLMET but also in CEN and POB, with the exception of one instance of bear with for...to-infinitive, given in (iv). This again indicates that the construction is probably new. (iv) But I couldn’t bear for him to know how I have disappointed you.

(1889 CEN)


conform both to semantic and distributional regularity and want little further com-

ment. The result is the distribution that we know today.

4. Synchrony and diachrony reconciled

The history of for...to-infinitival complements is primarily interesting for what it

tells us about the system of complementation and how it can change over time. With

respect to the synchronic organisation of a complement-type’s use, the most impor-

tant observations are, first, that motivational principles exist, and second, that they

are nondeterministic and may be in conflict. The use of the English for...to-infinitive

is clearly subject to semantic and paradigmatic regularities that even provide for

some degree of predictability. For example, in Present-Day English all verbs that are

fully qualified as for...to-infinitive-taking predicates by virtue of their meaning and

combinatorial behaviour effectively take for...to-infinitival complements. At the

same time, there is no full overlap between the ranges of verbs predicted to take

for...to-infinitives by different synchronic regularities, which means that in some

environments the principles of use may diverge. Where the sanctioning force exerted

by the linguistic system is low – that is, in environments falling under one regularity

but not the other – predictability decreases, with the result that arbitrariness slips in

or that other determining factors come to the fore. For instance, it is conceivable that

the absence of for...to-infinitives with the verb want in Present-Day English is due to

the very high degree of entrenchment of the exceptional case-marking construction

with which for...to-infinitives have to compete.

Turning to diachrony, subsequent synchronic stages each show the same rela-

tionship between motivation and indeterminacy. In the development of for...to-

infinitival complementation, the motivational factors – semantic and paradigmatic

regularity – appear to have emerged early on, but are provisional at all stages of dif-

fusion. The diachronic changes that occur, then, typically consist in bearing out ex-

isting regularities. In other words, the principles of synchronic organisation and

those of diachronic change are basically two sides of the same coin.

Thus, from a diachronic point of view, paradigmatic regularity corresponds to

paradigmatic analogy. That the principle determines the progression of change as

well as synchronic organisation is evident in changes where the appearance of a

for...to-infinitive with some new verb follows on the appearance of a prepositional

object with for with the same verb, as has happened in the case of arrange. Or, it

may manifest itself in the order in which semantically based extensions occur. For


VOLITION/

ACTIVITY

EM

OT

ION

VO

LIT

ION

/

EM

OT

ION

VOLITION/

ANTICIPATION ANTICIPATION/

ACTIVITY

COMMUNICATION

COMMUNICATION /

ACTIVITY

AC

TIV

ITY

CO

ND

ITIO

N

Figure 6.2. Semantic distances between verbs taking for…to-infinitives

example, although both wish and like are volitional verbs, for...to-infinitives ap-

peared with wish before they appeared with like, presumably because their use with

wish is paradigmatically sanctioned in addition to being semantically sanctioned.

Semantic regularity, on the other hand, corresponds to semantic analogy. This

correspondence implies that semantic links are not only reflected synchronically in

the semantic homogeneity between for…to-infinitive-taking verbs, but also dia-

chronically in the process of diffusion that gave rise to the present-day distribution.

The order in which for…to-infinitives have appeared with new verbs serves as a

mirror of the semantic distances between verbs. Figure 6.2 thus renders schemati-

cally the semantic distances between classes of verbs in semantic space, as reflected

in the diachronic development of for…to-infinitives. As described in the previous

section, for…to-infinitives first appeared in the shaded areas at the centre of the dia-

gram and then spread through transitional areas to the extremes of the diagram.

Two further points should be made in this respect. First, Figure 6.2 also

shows that semantic domains such as that of emotion and that of activity or condi-

tion are only distantly related to each other. In a way, the occurrence of for…to-

infinitives with verbs classifying under both these domains, could be seen as a his-

torical accident. Indeed, the use of for…to-infinitives with emotive verbs such as

bear, hate or love is motivated by the close semantic relation between volition and

emotion, and has little to do with the use of for…to-infinitives with verbs of activity

such as arrange, prepare, or seek, or verbs of condition such as allow and provide.

In fact, as pointed out earlier, the use of for…to-infinitives with emotive verbs can

be said to deviate from the overall regularity of the distribution of for…to-

infinitives, as the mental state expressed by an emotive verb is not directed to the

potential realisation of the event referred to by the complement clause. Thus, while

semantic analogy can often be seen as a levelling force that extends the regularities

in form-meaning mapping, it can also work as a disruptive force, when it only cre-


ates local regularities that do not necessarily concord with the overall regularity of a

distribution. The extension of for…to-infinitives to emotive verbs is justified locally,

not by the system of for...to-infinitival complementation as a whole. In the end, this

may actually come to undermine the homogeneity in the global distribution of

for…to-infinitives. Thus, the same principle that creates regularity can give rise to

irregularity.

Second, this in turn ties in with the notion that some uses are more strongly

sanctioned than others. Some uses will be sanctioned both locally and globally, oth-

ers only locally. Likewise, some uses will be sanctioned both paradigmatically and

semantically, others only semantically or paradigmatically. Once again, this point

pertains to synchronic organisation, but also to diachronic change. In the early ap-

pearance of for...to-infinitival complements, for example, it was the degree to which

different environments sanctioned the introduction of the for...to-infinitive that de-

termined where for...to-infinitives actually appeared. Similarly, where semantic

sanctioning applies equally, paradigmatic sanctioning can determine where for...to-

infinitives turn up first.

5. Conclusions

In this chapter it has been argued that the distribution of a complement-type, the

for...to-infinitive, is synchronically motivated, and it has been shown how syn-

chronic motivation can be reconciled with diachronic change. The crucial ingredi-

ents in reconciling synchrony and diachrony are, first, the finding that synchronic

motivation is nondeterministic; second, the finding that some uses are more strongly

motivated than others; and third, the notions that diachronic change is imminent in

synchronic organisation and that the principles of synchronic organisation readily

translate as mechanisms of historical change.

PART IV Ing-complements

“I find myself inherently suspicious

of anything that is inherently good”

(William Labov)

CHAPTER 7: Ing-clauses and their problems

On the face of it, ing-clauses and ing-complements are easy enough to define. Ing-

clauses are non-finite clauses whose verbal head is marked by an inflectional ing-

ending, as in (1). Ing-complements are a subset of ing-clauses, defined by their func-

tioning as complements to a matrix predicate, as in (2).

(1) a. This is as near as one will get to dining in a rainforest. (CB)

b. Speaking of which, how would you like to do the smallest of favors for

me? (CB)

(2) a. I like watching it but I don’t fancy playing it. (CB)

b. The dog took satisfaction in the fire, stretching out close enough for

warmth and far enough away to escape being singed. (CB)

c. I was at least able to start thinking partially straight again. (BNC)

d. In recent years this department has been busier repaying savings to the

public than attracting them. (BNC)

However, what the above definitions mark off as categories are not necessarily the

categories that underlie language users’ linguistic behaviour. The central question

addressed in this chapter, therefore, is how ing-clauses and ing-complements are

represented in speakers’ grammars. Specifically, do speakers work with a single

general category of ing-clauses from which they can derive all individual concrete

uses of the construction, as attested in actual corpus data, or do they (also) rely on

more specific representations, and if so, are those representations still related to one

another in some way?

This general issue is dealt with here in two of its manifestations. At the level

of ing-complements, we are confronted with another instance of the matching-

problem. Any resolution of the matching-problem follows from the constructional

characteristics attributed to ing-complementation. If language users generalise a

category of ing-complementation, this would imply that ing-complementation is a

single homogeneous construction – that is, that some common meaning (or a set of

closely-related meanings) is conveyed by all combinations of a predicate and an ing-

complement, and that there is some overall regularity in the distribution of ing-

complements over the inventory of complement-taking predicates. By contrast, if the

matching-problem proves irresolvable by a single constructional characterisation,

we may be forced to make further distinctions within the category of ing-

196 – CHAPTER 7: Ing-clauses and their problems

complementation or even to abandon such a highly abstract characterisation alto-

gether.

At the level of ing-clauses, there is the distinction between gerunds and pre-

sent participles to be dealt with. Following traditional grammar, gerunds are nomi-

nalisations, which means that they behave as noun phrases in terms of their external

syntax (i.e. they occur in clausal positions normally occupied by ordinary noun

phrases), even if their internal syntax is clausal, as in (3).

(3) a. Garner was also against “going to the cops”. (CB)

b. ending the bloodshed was his key task. (CB)

Participles similarly behave like clauses internally, yet they bear an external resem-

blance to adjectives and adverbs (e.g. occurring as modifiers to nouns or as adver-

bial clauses), as illustrated in (4). Note that, as (4) also shows, English participles

come in two subtypes, namely present and past participles.

(4) a. when I looked into her eyes I had this freezing feeling that the person

looking at me was someone completely different. (CB)

b. It’s a policy which is great for those in the NHS with new company

cars and office perks but is no good to the person left lying on a trolley

because of a lack of resources. (CB)

Because in English both gerunds and present participles are marked by an ing-suffix

on the verbal head, the two categories coincide, at least on the surface. What is less

clear is whether this superficial similarity has any deeper relevance. A category of

ing-clauses, if one is postulated, would evidently generalise over gerunds and parti-

ciples and sever the ties between nouns and gerunds or present and past participles.

So once again, is actual usage most accurately described by a single homogeneous

category of ing-clauses, or do we need to invoke more specific representations? In

the latter case, those more specific representations might correspond to the tradi-

tional distinction between gerunds and participles, but of course it is also conceiv-

able that the traditional categories are unrealistically general as well.

The position I ultimately take with respect to these issues is to remain cau-

tious about the psycholinguistic reality of high-level abstractions such as gerunds,

participles, ing-complements or ing-clauses. It is plausible that language users do

arrive at a highly abstract characterisation of ing-complements or ing-clauses in gen-

eral. In addition to their formal similarity, all ing-forms have in common that the

CHAPTER 7: Ing-clauses and their problems – 197

ing-suffix as such cannot independently ground the process or situation denoted by

the verbal stem to which it attaches (i.e. ing-clauses do not independently establish a

deictic link to the speaker-hearer situation). It is generally true, therefore, that -ing is

an ‘atemporalizing’ suffix (cf. Langacker 1991; Heyvaert 2003).50 It seems, how-

ever, that any further major generalisations have only some bearing on actual usage

and are certainly insufficient to account for all specificities of usage. Thus, most

generalisations, although partly relevant, are problematic in that they exaggerate the

consistency of actual linguistic behaviour. As in the case of for...to-infinitives, the

use of ing-clauses and ing-complements is organised around both global and more

local regularities, which once again emphasises the need for smaller-grained classi-

fications to arrive at accurate descriptions.

The rest of this chapter is organised as follows: The homogeneity or hetero-

geneity of ing-complementation is addressed in Section 1. The relevance or irrele-

vance of the distinction between gerunds and participles is dealt with in Section 2.

As a cautionary note, in both sections of this chapter attention primarily goes to ing-

complements that are controlled by the matrix subject, which are also the topic of

the historical analyses in the following chapters. Other types of ing-clause (non-

complements, non-controlled ing-clauses) are introduced in the discussion where

they are relevant (see especially Section 2) but mostly remain in the background.

This puts certain limits on the validity of what is to follow. However, where highly

specific constructional representations are argued for, the restriction is of course by

no means detrimental to the argument, and where claims are formulated with respect

to the more schematic constructions, I have attempted to take into account the whole

range of ing-clauses, even if they are not discussed explicitly.

1. Ing-complementation and the matching-problem

To understand the constructional status of ing-complementation is to resolve the

matching-problem (cf. Chapter 3). If ing-complementation functions as a single,

uniform predicate-complement construction, we may expect that, ideally, it consis-

tently maps a particular meaning (or a set of closely-related meanings) onto a given

form and that its constructional slots are subject to coherent, non-arbitrary selec-

tional restrictions. The consistent semantics of the predicate-complement construc-

50 Strictly speaking, even that may not be entirely true. Spoken language allows the omission of the auxiliary be in the progressive, especially in interrogatives (e.g. say your cousin she going to stay the night (COLT); Why you bugging her? (COLT)), which means that present tense reference is only marked by the ing-suffix. Conse-quently, -ing could here be analysed as a grounding element.


tion may translate as a particular way of construing the event denoted by the com-

plement clause or a particular relation between the event in the complement clause

and that expressed by the matrix predicate. The selectional restrictions imposed by

the construction are most likely to manifest themselves in non-arbitrary restrictions

on the set of matrix predicates ing-complements can combine with. It is to be ex-

pected that constructional semantics and selectional restrictions are not in contradic-

tion – indeed, one might even help to predict the other.

It turns out, however, that in this respect ing-complements are a tough nut to

crack. In most of the literature on ing-complements, the goal has indeed been to find

a consistent semantic characterisation of ing-complements as such, in the assump-

tion that the pattern’s distribution would fall out from this. These attempts have

largely failed, however. Establishing a consistent semantic characterisation has

proven difficult, and where such a characterisation has been provided, it typically

remains unclear how it should be translated to ing-complements’ actual distribution.

In other words, the only realistic semantic description of ing-complementation does

not solve the matching problem. A more promising approach to explaining the be-

haviour of ing-complements pursues the gerundial status of the majority of ing-

complements and their semantic and distributional links to noun phrases functioning

as direct objects to transitive verbs. This approach does not fully solve the matching-

problem either. It says nothing about participial ing-complements and even with

respect to gerundial ing-complements it has to be complemented with additional

functional principles and factors of language use as well as more specific construc-

tional representations in order to be compatible with actual usage. However, as I will

argue in this section, it offers the most satisfying generalisations about usage. (No-

tice, in addition, that the discussion here also gives us a first vista on the more gen-

eral issue of the relationship between gerunds and participles to be addressed in Sec-

tion 2 below.)

Turning to attempts to deal with ing-complementation, certainly problematic

– and even contradictory, as pointed out by Ney (1981) and Wierzbicka (1988) – are

proposals to describe ing-complements as referring to reified (i.e. actualised/realis)

situations (Bolinger 1968), or to situations in general, as opposed to specific situa-

tions (Wood 1956). The counterexamples in (5) dismiss the general applicability of

these labels: the ing-complement in (5a) refers to an event that is neither reified nor

conceived of ‘in general’, the one in (5b) refers to a situation that is reified but spe-

cific, and the one in (5c) must be thought of as referring to a situation in general

whose status as being reified or not remains unspecified (i.e. the ing-complement,


being generic, does not signal whether or not the situation in the complement clause

holds at the time of reference of the matrix clause).

(5) a. Am I going to stand there and ask somebody er Are y do you intend

stabbing somebody with this knife young man ‘cos obviously they’re

not going to say so (CB)

b. I will never forget hearing the Leningrad Philharmonic for the first

time (BNC)

c. He’s thin-skinned to criticism and likes being liked. (CB)

At the same time, Bolinger’s ‘reification’ and Wood’s ‘generality’ do not come out

of the blue. It is true that ing-complements can be used to mark reification in envi-

ronments where the to-infinitive marks potentiality, as in (6); and there are environ-

ments requiring complements with generic reference where ing-complements can be

used felicitously while to-infinitives cannot (cf. Fanego 2007), as in (7). As linguis-

tic facts these observations should not be ignored, even if they cannot be generalised

to the whole set of ing-complements.

(6) a. I don’t remember the explosion. I remember being plucked out of the

water. (CB)

b. I wonder if the padre will be there to greet me. I must remember to

watch my language. (CB)

(7) a. The tins were hidden behind a shelf of books because his landlady for-

bade cooking in the room (BNC)

b. *The tins were hidden behind a shelf of books because his landlady

forbade to cook in the room

The main alternative to characterising ing-complements in terms of reification

or generality is the view formulated by some proponents of cognitive grammar, who

argue that -ing has the effect of construing a verbal process as a homogeneous “se-

ries of internal states” (Langacker 1991: 443). Or, put differently, an ing-

complement “evokes a progressive sense whereby the verbal action is viewed as on-

going, in process, or viewed internally to the process” and it “focuses attention on

some interior part of a verbal process rather than the process as a whole” (Smith &

Escobedo 2001; see also Egan 2003). From this it follows that ing-complements can

be used to mark temporal overlap between the situation designated by the matrix

clause and that in the complement clause (Langacker 1991; Smith & Escobedo


2001). Wierzbicka (1988: 60) also seems to take this line of reasoning when she

argues that ing-complements tend to have a “present (current) orientation”. I take all

this to mean that ing-complements construe the event in the complement clause as

imperfective (i.e. they foreground only the middle section of an event, not its begin-

ning and ending, in contrast to perfective construal which highlights an event in its

entirety) and as simultaneous to the event in the matrix clause. Ing-complements

answering this description are regularly found with verbs such as enjoy, with percep-

tion verbs, or with adjectives such as busy, as in (8). In each example, the ing-

complement denotes a situation that is ongoing at some point in time simultaneous

to the situation denoted by the matrix clause.

(8) a. Mother, he was old and tough; you would not have enjoyed eating him.

(BNC)

b. she’d seen the witch’s cat riding through the air on a broomstick

(BNC)

c. She was so busy thinking about the people upstairs that she did not see

the figure standing in the shadows of the kitchen doorway. (BNC)

While the cognitive grammar account is at first sight harder to refute, it faces

serious problems as well. To begin with, the imperfectivising character of the -ing-

suffix in ing-complements is not as compelling as the cognitive literature suggests.

The semantic characterisation of -ing has no strong distributional reflex on the kinds

of verbs that occur within the ing-complements, as inherently imperfective verbs

(be, believe, have, know, like) are allowed to receive redundant imperfectivity mark-

ing:

(9) a. It was she who liked being in France. (BNC)

b. Heroin was found in their luggage but they’ve denied knowing it was

there. (CB)

In this, ing-complements differ from the ing-forms used to form the progressive,

which strongly disfavour imperfective verbs (cf. *she is being in France). The dif-

ference is not unmotivated, since the progressive has a paradigmatic non-progressive

alternative (she is in France), whereas ing-complements usually do not. That is,

without the -ing-suffix imperfective verbs could simply not be used as heads of the


complement clauses to a variety of predicates selecting only ing-complements.51

This granted, however, the suffix’s semantic salience is no doubt reduced by the fact

that the imperfectivising effect of -ing is not reflected distributionally, as well as by

the semantic redundancy that arises when -ing is used on verbs that are already im-

perfective.

But problems get worse. Imperfective meaning is sometimes absent alto-

gether, even where we would expect it (cf. Duffley 1999, 2003).52 Perfective verbs

can in fact be used perfectively in ing-complements. For example, what the subject

in (10a) will enjoy is not merely being engaged in the ongoing process of moving to

a new place, but crucially includes the end state of the process, namely being at the

new place. Likewise, the potential regrets of the subject in (10b) do not relate to her

having at one point been in the process of leaving Belfast, but to the entire process

of leaving, including its end state of being gone from Belfast.

(10) a. But he’ll enjoy coming up here. He misses his homeland. (BNC)

b. She did not know, now, if she regretted leaving Belfast, or was glad.

(BNC)

51 This is an extension of the argument originally produced by Langacker to account for the fact that -ing can be used on imperfective verbs to form participles modifying a noun. He writes: “when the participle modifies a noun rather than combining with be to derive a progressive clausal head, it can be formed on an imperfective verb as well as a perfective one: anyone knowing his whereabouts; students having difficulty with their homework; people still believing that the earth is flat. This difference be-tween the progressive and modifying uses is easily explained. As noted earlier, the progressive construction derives an imperfective verb, so using it with a verb that is already imperfective would be superfluous; hence -ing is limited to perfectives in the context of that construction. However, -ing by itself derives an atemporal relation, which is what the modification of a noun requires. Neither a perfective nor an im-perfective verb can modify a noun directly, so a verb of either sort needs the atem-poralisation effected by -ing to be usable for that purpose.” (1991: 210) 52 Duffley (1999, 2003) rightly points out the difficulties pertaining to the claim that -ing is always imperfective, but his argumentation is sometimes flawed. First, he wrongly assumes that imperfectivity and anteriority/posteriority are mutually exclu-sive (in fact, it is perfectly possible to profile only the ‘inner’ phase of an anterior or posterior situation, e.g. after he had been talking to her for some time, he asked...). Second, in his discussion, he also conflates verbs that are inherently imperfective with verbs that are inherently perfective. As indicated above, however, only the lat-ter can be used as strong evidence against the imperfective nature of ing-complements (see also footnote 50). As a result, Duffley’s criticism against the ca-nonical treatment of the -ing-suffix in cognitive grammar, though justified, is proba-bly too severe.


Similarly, the claim that imperfectivising -ing imposes a repetitive reading on punc-

tual verbs (blink, hit, nod, slap, tap), which holds true for progressives,53 does not

always extend to ing-complements (cf. Declerck 1982: 3-4). Thus, hitting with the

verb stop in (11a) can only be interpreted as repetitive (reasonably so, because oth-

erwise the hitting could not but stop), but with avoid in (11b) it designates a single

punctual event.54

(11) a. Stop hitting that man, Patel. (BNC)

b. she entered by the low door, bending to avoid hitting her head on the

lintel. (BNC)

Even the ing-complements that allow an imperfective reading often do not strongly

enforce it. For example, the ing-complement with enjoy in (12) can just as well be

rendered with a perfective paraphrase as with an imperfective one (‘I was reading

this book and that I enjoyed’ versus ‘I read this book and that I enjoyed’).

(12) I enjoyed reading this book and recommend it to anyone with an inter-

est in the way chemistry really develops (BNC)

In brief, the evidence indicates that the imperfectivity supposedly associated with -

ing can be neutralised.

A further argument against describing ing-complements as imperfectivising

comes from the distribution of ing-complements over the inventory of complement-

taking predicates. The actual distribution of ing-complements is hardly ever predict-

able on the basis of the semantic characterisation offered and in a number of cases

even seems downright incongruous with it. Simple unpredictability is exemplified

by the use of ing-complements with verbs such as remember or imagine. As a com-

plement-taking verb, remember is certainly compatible with an imperfectively con-

strued situation in the complement clause. Such construal rather expressively sug-

53 For example, Langacker states that “With punctual verbs like blink and flash, the repetitive construal is the only one for which an internal perspective is conceptually plausible” (1991: 209). 54 Incidentally, that the repetitive reading in (11a) probably does not derive from the -ing-suffix itself is suggested by examples such as the following, which shows that a punctual verb in the to-infinitive – which clause-type is typically described as per-fective (Langacker 1991) – also receives a repetitive interpretation when comple-menting an aspectual verb: Hurricane force winds continue to hit Shetlands (BNC) or then, slowly, he began to nod, a faint smile forming on his lips (BNC).


gests that the subject relives the situation as s/he brings it back to mind: the higher

clause subject is dropped back in the middle of the action, as it were, like a character

in the opening lines of a novel:

(13) I remember going to head the ball. The next thing I remember was ly-

ing on the ground surrounded by players and officials. (BNC)

At the same time, remember also allows a complement clause that construes its

situation perfectively:

(14) I remember I said to myself: “If this ship could turn round now, I could

return to Jaffa.” (BNC)

Therefore, the choice for a complement with imperfective construal can score a cer-

tain semantic effect, but the perfective or imperfective meaning of any given clause

type remains largely irrelevant to its being grammatically eligible as a complement

to the verb remember. One could not predict, for example, that remember sanctions

ing-complements but not bare infinitives, simply from the fact that ing-complements

are imperfective and bare infinitives perfective, because remember is compatible

with either aspectual construal types.

If this situation would hold across the inventory of complement-taking predi-

cates, there would be no strong distributional evidence for the imperfective meaning

supposedly associated with ing-complements. But the situation is worse than that.

With other complement-taking verbs, the imperfectivising character of -ing is actu-

ally incompatible with the semantics of the matrix verb. For example, verbs such as

avoid, (cannot) help, delay and reject signal that the subject takes action in order for

a situation not to occur. The avoided situations are obligatorily expressed as ing-

complements, yet in these constructions, there is no temporal overlap between the

action in the matrix clause and the avoided situation in the complement clause.55

55 Smith & Escobedo (2001) invoke the notion of subjectification to try to account for the lack of temporal overlap. They claim that with verbs like avoid overlap has been relegated from the objective axis to the subjective axis, and therefore obtains no longer between matrix clause and complement. Instead, the constructions in question allow “a subjective construal on the speaker’s part that conceptual overlap between the matrix subject and the complement process is obligated in some way, even if it is not objectively realised. [...] Thus, if I say that Mary dreads or avoids doing something, I imply that, although there is no overt objective overlap between the matrix subject Mary and the process she dreads or avoids doing, there ought to be such overlap”. This appears to be an entirely ad hoc explanation, and not a very


Moreover, it would seem that avoidance bears on a situation as a whole, including

the undesirable consequences that follow from completion, so there is no obvious

reason for imposing an imperfective construal on the avoided situation. Indeed, the

examples in (15) show that verbs of avoiding provide some of the prime contexts for

aspectual neutralisation of -ing. Each of the verbs in the ing-complements in (15) is

inherently perfective (become, fall, etc.); in none is the imperfective reading that

ought to follow from the use of the ing-suffix plausible.

(15) a. It may easily be objected that if he had wished to avoid becoming

archbishop he had simply to refuse. (BNC)

b. The boy who can’t help falling asleep. (BNC)

c. He had delayed publishing the proposals four times in the hopes of a

rapprochement. (BNC)

d. On the following day he condemned it as “illegitimate and invalid” and

rejected opening formal negotiations (BNC)

Similar problems attach to some other verbs with prospective orientation, such as

risk, intend or consider. Again, with such verbs there is no temporal overlap be-

tween the situations in the main clause and the complement clause, and it is often the

whole complement situation that is relevant to the main clause subject. If anything,

one would therefore expect perfective construal to be preferred over imperfective

construal. Accordingly, as shown in (16), these verbs readily allow aspectual neu-

tralisation of -ing:

(16) a. Serving officers who attempt constructive criticism of the police, risk

being labelled traitors and put their promotion prospects in jeopardy.

(BNC)

b. You will see that we intend sending a leaflet to each member with the

summer edition of Rural Wales. (BNC)

c. You might consider buying a screech alarm. (BNC)

Some retrospective verbs, too, cause difficulties. Imperfective construal appears,

once more, to be incongruous with verbs such as admit and regret. The reason is that

compared to the middle section of a situation, a situation in its entirety, including the

good one at that. Whatever is meant by “overlap between the matrix subject and the complement process”, there is as far as I can see no indication that ing-complements with verbs like avoid somehow subjectively mark speaker opinion.


resultant end state, has just as much or indeed more bearing on any admission or

regret on the part of its agent. For example, what the main clause subject in (17a)

admits is not that she has at one point been in the process of shooting Staudinger,

but that she has actually shot Staudinger. Similarly, the regrets of the subject in

(17b) do not specifically centre on the short period between her beginning and end-

ing her question, but on the entire act of asking the question.

(17) a. But as the trial opened, she changed her plea and admitted shooting

Staudinger last December, at his Manhattan apartment. (BNC)

b. She regretted asking the question as soon as the words were out.

(BNC)

In all, then, the whole idea that ing-complements carry an imperfective meaning is

ill-conceived. There is no distributional evidence supporting the imperfective mean-

ing of -ing (neither in the verbs within the ing-complement nor in the predicates

combining with ing-complements) and there is a considerable amount of counterevi-

dence.

In search of a semantic characterisation of ing-complements, we are eventu-

ally left with Duffley’s solution, which abandons the imperfective meaning of -ing

and states instead that in ing-complements “the interiority of the -ing’s event is sim-

ply taken as a whole, as the totality of positions contained between its beginning and

its end-points” (1999: 10), or “the -ing evokes the interiority of its event holistically

as that which is “[verb]ed” in the event expressed by the matrix” (2000: 227-8).56, 57

56 Given that Duffley (1999, 2000) refutes the cognitive grammar position that sees -ing as an imperfectivising suffix, I take it that by “the interiority” of an event he does not mean the period excluding beginning and end-point of that event, but the entire event – though taken as such the wording of the semantic definitions might lead one to think otherwise. 57 In fact, Duffley (1999, 2000) limits this characterisation to gerundial ing-complements, simultaneously proposing that participial -ing does not neutralise the imperfective meaning (see Section 2 below on the syntactic distinction between ge-rundial and participial ing-complements). A strict semantic contrast between gerun-dial ing-complements and participial -ing is not easy to maintain, however, as some participial ing-complements also neutralise the perfectivity/imperfectivity distinction (e.g. Kevin Blinco has no trouble remembering the events of October 10, 1993 (CB); see further the discussion of integrated participle clauses in Chapter 10; see also Declerck 1982). I do not exclude the possibility that in some constructions the char-acterisation of -ing as imperfectivising suffix holds true (as it does in the progressive construction). In the domain of complementation, the most likely candidate is the complements to perception verbs, where the aspectual distinction might be gram-matically encoded in the contrast between ing-clauses and bare infinitives. Even


I interpret this as meaning that ing-complements are in fact neutral with respect to

aspect (imperfective/perfective), tense (posterior/anterior/simultaneous) or modality

(reified/potential). Ing-complements do not add anything to the semantics of the

predicate-complement construction in terms of modality or temporality other than

what is already conveyed by the more schematic transitive construction and the

complement-taking predicate, nor do they, in the majority of uses, impose any other

internal temporal make-up on the situation in the complement clause than what is

contextually implied. Thus, catching 23 salmon before breakfast in (18) is reified

and anterior to the matrix verb because recall takes reified and anterior events as its

direct object (as opposed to he envisaged catching 23 salmon before breakfast) (cf.

Duffley 1999), and the complement clause is, if anything, pragmatically perfective

because perfectivity is implicated by the finite number of salmon caught (as opposed

to he recalled catching one salmon after another).

(18) He was later to recall catching 23 salmon before breakfast (CB)

This view has an important advantage in that it brings to the fore the close

affinity between ing-complements and the ordinary direct objects of transitive verbs

both on the level of interpretation and on the level of distribution. Visser already

noted this generalisation when he wrote:

Broadly speaking it might perhaps be said that the form in -ing seems some-

times to be preferred when the finite verb is distinctly meant to be appre-

hended as transitive for which an object in the form of -ing is more suitable

than one in the form of an infinitive, since the former is nearer to a real noun

than the latter (1963-73: 1861)

Concerning interpretation, the meaning of the whole predicate-complement con-

struction is typically analogous to the meaning of an ordinary transitive construction

here, however, the effect sometimes seems to be pragmatic – witness the apparent counterexamples in (i). (i) a. Alison still keens and whispers as if she’s just seen the sun being

snuffed out. (CB) b. Dressed in a robe, he half-listened to Lisa talk to him from the bed,

leaned against the window jamb and watched Autumn's supple body, a misty shadow in the moonlight, as it moved through the water. (CB)


with the same complement-taking predicate. This allows us to handle interpretative

differences between ing-complements and to-infinitives in the same environment, as

well as differences in the interpretation of ing-complements in different environ-

ments (e.g. the anteriority expressed by ing-complements with remember, as op-

posed to the posteriority expressed by to-infinitives; see (6) above). It also points the

way to how we are to deal with the generic ing-complements found with verbs such

as forbid or mean (see (7) above): these too can now be seen as clausal equivalents

to ordinary direct objects, as is illustrated in (19) (cf. Duffley 1999; Heyvaert 2006):

(19) a. to be loyal to my feelings means betraying the Party. (CB)

b. Packaging can mean waste. (CB)

The semantic neutrality of ing-complementation indicates that the distributional

problem is unlikely to be solved through semantics alone. Yet here too the relation

between ing-complements and direct objects offers a helpful perspective. It is a dis-

tributional fact that ing-complements tend to occur in nominal positions. Generative

studies, for instance, have unhesitatingly pointed to the relevance of ing-

complements’ nominalness to issues of distribution (Ross 1973; Emonds 1976: 125;

Schachter 1976: 223; Pullum 1991: 764; see also Huddleston & Pullum 2002: 1188,

1209), and although these generative approaches keep away from dealing explicitly

with the lexical restrictions characteristic of complementation, it is easy enough to

see that there is a good deal of sense in the line of reasoning they take. For example,

it can be assumed that the verb aim takes to-infinitives but no ing-complements at

least in part because aim is simply not a transitive verb:

(20) a. We aim to please.

b. *We aim pleasing.

All this can be summarised by saying that the interpretation and distribution of ing-

complements is determined by a form of paradigmatic regularity, with ordinary di-

rect objects providing the interpretative and distributional model whose behaviour

ing-complements copy.

The role of paradigmatic regularity needs to be qualified, however. It would

be an oversimplification to say that ing-complements add nothing at all to the inter-

pretation of the predicate-complement construction. Specifically, with a good num-

ber of verbs, ing-complements that do not express their own subject are obligatorily

controlled by the matrix subject. In this respect, ing-complements obviously differ


from ordinary direct objects which need not be controlled, as shown in (21). Further,

ing-complements typically convey active voice. This is apparent from the fact that

the passive must be marked explicitly by the auxiliary being, as in (22a). In this re-

spect, ing-complements again differ from semantic restriction not imposed on noun

phrases in the same lexico-grammatical environments, witness (22b).

(21) a. I regret saying what I did to those TV guys (CB)

b. Few would regret the introduction of electricity to a village, for exam-

ple, although opinion is likely to diverge with the consequent introduc-

tion of a juke box and numerous television sets. (CB)

(22) a. I don’t remember it being postponed. (BNC)

b. How many people now remember his humiliation as chancellor of the

Exchequer (BNC)

In a constructional network, what this implies is that at least some ing-complements,

including the ones in (21a) and (22a), must resort under a construction that is more

specific than the transitive construction – even though they may still instantiate the

more schematic transitive construction as well. Going one step further, notice that

special (passival) control relations, along with internal nominal syntax, also allow us

to single out the ing-complements with verbs such as want or need, illustrated in

(23), as instantiating a distinct construction; and the same goes for the ing-

complements with worth (discussed at length in Chapter 4).

(23) There’s no guesswork needed when changing these cooker hood filters

– the writing on them changes colour when they need replacing. (CB)

These smaller-grained sub-divisions are not exactly a problem to the notion of para-

digmatic regularity. It simply means that paradigmatic regularity is relevant only to

some aspects of usage and that the field of ing-complementation is more diverse

than the unifying principle implies.

A (simplified) constructional hierarchy capturing both the regularities and

specificities across constructions is given in Figure 7.1. For the sake of simplicity,

only instantiating relations are rendered (not integrating relations). Notice that the

maximally abstract category of ing-clause, as well as its instantiating relations to

other constructions, is marked in dotted lines to indicate the lack of evidence sup-

Figure 7.1. Ing-complementation in a (possible) constructional hierarchy.

...they need re-placing...

...I regret saying what I did...

...few would re-gret the introdu...

NP + need/want + Ving; [+ passival con-trol]; [+ nominal syn-tax in complement]

NP + V + Ving; [+ active voice in com-plement]; [+ subject-control]

Transitive clause NP + V + NP; [= ‘agent + action + pa-tient’]

Ing-complementation

Ing-clause

Regret Need


porting its existence as independent constructions. It is possible that language users

generalise up to this most abstract level of constructional representation, but apart

from the recurrence of the ing-marker, there is so far no strong reason to assume that

they do. The relation of paradigmatic regularity that is characteristic of ing-

complementation, both at the distributional and at the interpretative level, is marked

by the instantiating relation between ing-complementation (and its instantiations)

and the transitive construction. The fact that the predicate complement constructions

with ing-complement cannot be reduced to the transitive construction is marked by a

layer of less schematic constructions, specifying additional information regularly

conveyed by subtypes of ing-complementation.

Figure 7.1 is misleading, however, in as far as paradigmatic regularity leaves

one serious problem unresolved. Distributionally, there may be a tendency for ing-

complements to associate with nominal positions, but it is evident that not all ing-

complements fill nominal slots and that not all nominal slots are filled by ing-

complements. For example, busy oneself takes ing-complements, but is not a transi-

tive verb, as shown in (24a-b). Likewise, the object-controlled ing-complements to

causative verbs such as set, get, have, or keep do not correspond to noun phrases, as

shown in (24c-d). Conversely, several verbs do not take ing-complements despite

being transitive, such as promise in (25):

(24) a. She nodded and strode briskly to the sideboard, where she busied her-

self pouring coffee. (BNC)

b. *She nodded and strode briskly to the sideboard, where she busied

herself coffee.

c. Human Nature had him musing on the internal battle of animal and

civilized instincts. (2004, TIME)

d. *Human Nature had him some musings on the internal battle of animal

and civilized instincts.

(25) a. I’ll promise silence in exchange. (CB)

b. *I’ll promise keeping silent in exchange.

In other words, paradigmatic analogy at once over- and underspecifies the actual

distribution of ing-complements.

Solutions to this problem lead in two directions. On the one hand, the ing-

complements in non-nominal slots might have to be treated as participles and be

thereby relegated to a different category altogether. If we divide ing-

complementation up into gerundial and participial complementation, the violation of


paradigmatic regularity in one of these domains is a non-issue. Until I have ad-

dressed the issue of the gerund/participle distinction in Section 2, I take this to suf-

fice as a solution (although as Section 2 will show the participial status of non-

nominal ing-complements is not always equally clear either and it may be necessary

to invoke local patterns of indeterminate ing-clauses that do not stand any further

from typical gerunds than they stand from typical participles). In the case of transi-

tive verbs resisting ing-complementation, on the other hand, there is no such escape

hatch. To handle this second group of violations, the only thing we can do is to find

additional constraints on ing-complementation that might account for the exceptions

against paradigmatic regularity.

Pursuing the latter line of inquiry, there are some other principles that appear

to have a say in deciding which predicates combine with ing-complements and

which do not. First, there is at work a scattered version of semantic regularity. It is

clear that ing-complements appear with clusters of semantically related verbs.

Among these at least three major groups can be discerned: ing-complements associ-

ate with negative implicative verbs (avoid, defer, omit, etc.), emotive verbs (enjoy,

hate, like, etc.), and aspectual verbs (begin, continue, stop, etc.). In addition, some

smaller semantically constrained patterns can be made out: for instance, the use of

ing-complements with retrospective verbs (forget, recall, recollect, remember), the

use of ing-complements with proposal verbs (propose, recommend, suggest), the use

of passival ing-complements with necessity verbs (need, require, want), and their

use with valuing adjectives (worth, worthwhile). The semantic groupings of ing-

complement-taking verbs have been worked out in detail (if not comprehensively)

by Rudanko (1989: 44-6), and their possible relevance is further recognised at least

implicitly in several other complementation studies that start by dividing the inven-

tory of complement-taking predicates into sets of semantically related verbs (e.g.

Visser 1963-73; Ney 1981; Wierzbicka 1988; Fanego 1996a; Huddleston & Pullum

1240-1; Smith & Escobedo 2001).

The relevance of semantic groupings to the construction hierarchy is particu-

larly evident when a predicate-complement construction with a subset of semanti-

cally related verbs specifies some further restriction. For example, verbs of endur-

ance (bear, endure, stand) can combine with ing-complements, but only when pre-

ceded by a negated modal of ability (typically can’t), as in (26).

(26) I can’t bear picnicking on a slope. (BNC)


Still, while this underscores the importance of semantic groupings, note that seman-

tic regularity cannot work as a strong factor ruling out the use of ing-complements

with some transitive verbs, because in fact the distribution of ing-complements ex-

tends beyond the limits imposed by semantic groupings. For example, ing-

complements occur with verbs not clearly belonging to any of the above semantic

sub-classes, such as fancy, intend, mind, risk, or try. Even so, semantic regularity

may at least explain part of the puzzle, as it is at least true that the transitive verbs

disfavouring ing-complementation fall outside the major semantic groupings.

Second, the absence of ing-clauses with many transitive verbs can be ac-

counted for by the fact that an ing-complement does not give all the information that

is necessary for grounding the event it denotes. This is especially significant when

the matrix predicate makes no predictions about the temporal and modal interpreta-

tion of the situation in its complement clause. Tellingly, complement-taking predi-

cates of this kind usually require a finite complement (Noonan 1985; see the discus-

sion of dependent and independent time reference in Chapter 3). For instance,

among cognitive verbs, imagine can take controlled ing-complements, while think

cannot, as shown in (27a-b). Presumably, the reason is that imagine, unlike think,

implies that the event in the complement clause is necessarily situated in a space of

unreality, which reduces the need for modal and temporal grounding. Interestingly,

(27c) seemingly violates this principle, as it shows that think can in fact combine

with a non-finite complement-type, namely the to-infinitive. The exception proves

the rule, however, since the subject-controlled to-infinitive following think consis-

tently marks posteriority and potentiality (in line with many other to-infinitival

complements), thereby conveying enough information for situating the event in the

complement clause with respect to that in the matrix clause.

(27) a. He imagined opening the drawing-room door. (BNC)

b. *He thought opening the drawing-room door.

c. He thought to open the drawing-room door.

Third, to some extent the distribution of the to-infinitive negatively deter-

mines the distribution of ing-complements. That is, with some transitive verbs the

possibility of using a to-infinitival complement seems to pre-empt the use of ing-

complements, especially when the to-infinitive is highly common. This is the case

with verbs such as expect (or promise in (25) above).


(28) a. if you expect to make more than one call per day, choose the Standard

tariff. (CB)

b. *if you expect making more than one call per day, choose the Standard

tariff.

Notice that expect need not belong to the set of verbs whose semantics predict too

little about the temporal and modal status of the complement clause for a non-finite

complement to be permissible, for the potentiality expressed by the to-infinitive is in

fact redundant with the meaning of the matrix verb. That is, there is no semantic

difference between expect and (say) envisage (which is attested with ing-

complements) that is significant to the selection of ing-complements. Probably, the

difference that is significant is the highly frequent use of the to-infinitive with ex-

pect. That is, ing-complements are blocked by the to-infinitive (cf. Chapter 4).

Where does all this leave us? It is evident that even if ing-complementation is

a construction, it is not or not only at the most abstract level of representation that

usage is being orchestrated. To account for actual usage, we need to posit a number

of less schematic constructions. On the interpretative dimension, a fairly consistent –

if somewhat unrevealing – characterisation of ing-complements is possible in terms

of paradigmatic regularity, according to which ing-complements typically follow the

interpretation inherited from the transitive construction and add no temporal, modal

or aspectual information of their own. To the extent that some ing-complementation

patterns are subject to further interpretative constraints, however, separate construc-

tional levels must be postulated. For example, as pointed out above, there is a large

group of verbs where ing-complements without explicit subject are obligatorily con-

trolled (regret, remember, etc.); there is another smaller set of verbs where ing-

complements are controlled and passival (need, want); and so on.

On the distributional level, paradigmatic regularity again provides a powerful

generalisation but the distributions of the transitive construction and of ing-

complementation do not match fully, so that other factors have to be invoked to ac-

count for the distributional gaps. First, further selectionally constrained sub-

constructions can be postulated organised around sets of semantically related verbs.

Second, there are negative factors such as functional disfavouring (with matrix verbs

that do not specify how the complement event is grounded) and frequency-induced

blocking (with matrix verbs that frequently combine with to-infinitives) that keep

the distribution of ing-complements from being paradigmatically fully regular.

Third, participial ing-complements may have to be lifted out of the equation alto-

gether and treated as a separate phenomenon. The discussion in the following sec-


tion reinforces the idea that the use of ing-clauses is organised around a complex

network of constructions of different degrees of specificity and interrelated by cate-

gorial ties of varying strength.

2. Gerunds vs. participles

At least one reference grammar of English, Huddleston & Pullum (2002), treats ger-

unds and (present) participles as belonging to a single category: ‘gerund-participles’.

In doing so, Huddleston & Pullum react against the “general practice in traditional

grammar of describing subordinate clause constructions in terms of functional

analogies with the parts of speech” (2002: 1221). Indeed, the traditional view holds

that gerunds are somehow akin to nouns, while participles relate to adjectives and

adverbs, which relationships form the primary basis for distinguishing the two

clause-types (see e.g. Declerck 1991). It is of course correct to state that non-clausal

ing-forms can productively function as pure nouns or as pure adjectives, as the ex-

amples in (29) show. The sensibleness of distinguishing these purely nominal and

purely adjectival uses is not at issue here.

(29) a. The New Zealand experience shows that a complete recasting of tax is

entirely possible (CB)

b. But I must admit that what they are offering is very enticing. (CB)

The question is what to do with the clausal constructions, where phrase-internal syn-

tax is neither nominal nor adjectival. Here we have to resolve which categorial ties

Figure 7.2. Possible categorial relationships for gerunds and participles.

Gerunds Present participles

Nouns

Past participles/ Adjectives/

Adverbs


are strongest: the tie between gerunds and participles or the respective ties between

gerunds and nouns and between present participles and past participles, adjectives or

adverbs. The problem is schematised in Figure 7.2.

The position I wish to take here is a subtle one. First, there are some argu-

ments against the view expressed by Huddleston & Pullum (2002) that make it diffi-

cult to wholeheartedly embrace their position, especially in its strongest form. Sec-

ond, a straightforward functional equation between gerunds and nouns or between

participles and adjectives/adverbs is similarly problematic, even if it cannot be de-

nied that the behaviour of nouns and adjectives/adverbs is to some extent still rele-

vant to that of gerunds and participles. Third, as should already be clear from the

argument in the previous section, the categories of gerunds and participles – or, for

that matter, gerund-participles – can only be highly schematic super-categories un-

der which more specific constructions resort. In this light, it is not inconceivable that

at some lower level of abstractness a subset of the traditional class of gerunds and a

subset of the traditional class of participles can be collapsed – especially in uses

where ing-clauses maximally depart from the behaviour of nouns or adjectives and

adverbs. In what follows I will develop these points one by one.

Huddleston & Pullum’s (2002) argument hinges on the claim that there are no

inflectional or semantic differences and no differences in internal syntax that still

warrant the distinction between gerunds and participles. This is not entirely true,

however. For a start, coming from a somewhat unexpected corner, there is a formal

argument against seeing ing-clauses as an entirely undivided clause-type. Huddle-

ston & Pullum (2002: 82-3) argue that inflectionally there is no distinction between

gerunds and participles, the inflectional ing-suffix added to verbal stems being the

same for either grammatical category for any verb. However, as reported by Labov

(1989), where the ing-suffix is phonologically reduced to /in/, an interesting pattern

of grammatical conditioning comes to light, in that the reduced variant is “favored

most in progressives and participles, less in adjectives, even less in gerunds and least

of all in nouns like ceiling or morning” (1989: 87). Labov insists that “every study

of (ING) for almost every speaker has shown the same result” (1989: 87). It is unfor-

tunate that Labov does not factor out the role of frequency, which might have given

an alternative explanation to part of the grammatically determined variation.58 It is

58 In fact, based on patterns of geographical distribution, Labov (1989) argues that what today looks like a weakening tendency in final consonants is in fact a pattern of historical persistence, with /in/ ultimately not deriving through reduction of the ing-suffix but being inherited from the historical participial ending -inde. However, present-day speakers may have identified /in/ as a reduced variant of ing, and may have come to manage their use of the variants in accordance with the ordinary


also unfortunate that the classification of ing-forms is not more fine-grained, as there

might be further differences between lower-level constructions. In any case, how-

ever, formal differences, while not being categorical, provide evidence that speakers

somehow manage to keep track of grammatical distinctions between ing-clauses.

Respecting internal syntax, too, Huddleston & Pullum (2002) ignore certain

differences between ing-clauses. For one thing, traditional gerunds still allow a

range of determiners that remind of their nominal origins (Heyvaert 2006).

(30) a. If the four blade edges are sharpened with a file, it becomes a very

manageable hoe, cutting weeks in all directions. Its flat blade is useful

for grooving seed drills on deep beds and the consolidating the soil af-

terwards. (CB)

b. Well, I hope that sentence will deter you from any more writing on

walls (BNC)

c. So the, with this it’s the only thing I’ve got. with this changing the

rules, about eighteen months ago (BNC)

If this syntactic possibility can be dismissed as a marginal remnant of earlier stages

of the language, Huddleston & Pullum’s (2002) argument still stumbles over the

difference in the form of the subject in gerunds and participles. As they themselves

recognise, both clause types can take oblique subjects, but traditional gerunds can

take genitive subjects as well whereas traditional participles can also take nomina-

tive subjects. This is illustrated by the following pair of examples from Huddleston

& Pullum (2002):

(31) a. She resented his/him/*he being invited to open the debate. (Huddleston

& Pullum 2002: 1220)

b. We appointed Max, he/him/*his being much the best qualified of the

candidates. (Huddleston & Pullum 2002: 1220)

mechanisms organising reduction. In this light it is perhaps significant that the grammatical environment apparently most strongly favouring reduction, namely progressives, is also the most frequent environment, which would tie in with the notion that frequency favours reductive changes (see Chapter 4). It is less clear though whether the same frequency factor could account for the differences between participles, adjectives and gerunds, as frequency differences among these categories are less outspoken.


Huddleston & Pullum (2002: 1221) dismiss this kind of evidence for the ger-

und/participle distinction by claiming that genitive case is restricted to ing-clauses in

complement position, whereas nominative case is restricted to ing-clauses in non-

complement position, implying that the difference really lies between syntactic envi-

ronments and not between gerunds and participles.59 True though this may be, there

is a subtler difference, calling for structural divisions between ing-clauses independ-

ent of the complement/non-complement distinction. Specifically, some ing-clauses

in complement positions allow both genitive and oblique subjects, while others only

allow oblique subjects.

Most relevant in this context are two types of ing-clauses, both of which obli-

gatorily take an oblique subject. One type, illustrated in (32a), functions as comple-

ment to verbs of perception.60 Another, illustrated in (32b), is part of an adverbial

clause introduced by the preposition with. The absence of genitival subjects in these

ing-clauses contrasts with the (potential) use of genitival subjects in other ing-

clauses in complement-positions, as shown in (33).

59 This generalisation holds because the ing-clauses following prepositions can be regarded as complements (viz. to the preposition). Incidentally, because this means that gerunds only ever occur as complements, half of Huddleston & Pullum’s (2002) argument is not falsifiable. 60 A quick survey of corpus data reveals some (very few) possible counterexamples, such as that given in (ii). But even if examples as in (ii) are genuine counterexam-ples, there is an overwhelming tendency against using genitival subjects that would still set perception verb complements apart from other ing-complements. (ii) Man: I heard your playing last night. – Woman: My playing? – Man:

Shakuhachi. – Woman: Oh. – Man: You played very softly. I had to strain to hear it. (CB)

This being said, I am not convinced that examples as in (ii) are in fact genuine coun-terexamples. Intuitively, it seems that the possessive in (ii) inherits something of its role as a determiner in encoding a means of referential accessibility as well as con-veying an existential presupposition (consider the presuppositional difference be-tween did you hear him playing the piano? and did you hear his playing the piano?). I do not claim that this interpretative difference extends to all contexts of variation between possessive and oblique subjects, but it does appear to play up when the pos-sessive is highly marked, as in (ii) above. On a tentative note, though, the possibility that the possessive is more strongly reminiscent of a determiner marking accessibil-ity (or more precisely, the road to mental access; cf. Willemse 2005) might explain why ing-clauses functioning as subjects – a role associated with accessible informa-tion – strongly prefer possessive subjects, at least in formal registers (Heyvaert, Rogiers & Vermeylen 2005).


(32) a. this time we’d definitely seen him carrying what looked like wind-

screen wipers (CB)

b. With him playing like that, my job was to hang on to second place.

(CB)

(33) a. a strongish north wind prevented our seeing clearly the moment a

squadron moved (CB)

b. Despite his being a card-carrying Communist since 1961 he was one of

the first Russians to take an active part in the transition to capitalism.

(CB)

The contrast between ing-clauses as in (32) and ing-clauses as in (33) suggests that

the expression of the subject-argument in ing-clauses is in fact a moot point in Hud-

dleston and Pullum’s unified category of gerund-participles, in that certain ing-

clauses can be set apart on the basis of their internal syntax.

Note that with respect to the ing-complements following perception verbs, it

can be argued that the subject of the ing-complements is not really a subject but an

object to the matrix verb controlling a subjectless ing-clause. On this analysis, him

in (32a) above is the object of seen rather than the subject of carrying. The main

argument for this view is that him can be fronted to subject position under passivisa-

tion of the matrix verb, as in (34).

(34) he was seen carrying what looked like windscreen wipers

However, the object-controlled analysis is only one possible analysis – inevitable in

some uses, but very unlikely in others. In at least some examples the ing-clause and

the preceding noun phrase show signs of behaving as a single constituent, in which

case the conclusion is hard to escape that the oblique noun phrase between the verb

of perception and the ing-form is the subject of the latter rather than the object of the

former (cf. Declerck 1982).61 First, there are instances where, semantically, what is

61 The syntactic arguments invoked here are also invoked by Declerck (1982), but the conclusion is slightly different. Declerck distinguishes between one two-constituent analysis (the object-controlled construction) and two one-constituent analyses. His positing two one-constituent analyses is motivated by the fact that some of the tests revealing one-constituent status show the ing-complement to be equivalent to the bare infinitive construction, whereas other tests show ing-complements and bare infinitives to be non-equivalent. Most importantly, coordina-tion (as in (36) above) reveals equivalence, whereas the possibility of use in the fo-cus position of pseudo-clefts (as in (37) above) shows non-equivalence, since bare


perceived cannot be the referent of the oblique noun phrase as such, but must be the

whole situation in which it participates. Consider the examples in (35).

(35) a. Fortunately, because I wanted no part of any event featuring gospel

singers, I was not present to hear an incredulous crowd being exhorted

to “stamp on Satan”. (CB)

b. It is in the popular culture that our children hear black women being

referred to as ‘bitchez’ and ‘hoez’. (CB)

Second, on the object-controlled analysis, examples as in (36) force us to attribute a

double status to the oblique noun phrase, which must at once be the subject of the

bare infinitive and the object of the perception verb controlling the ing-clause –

unless of course the bare infinitive is analysed as another object-controlled secon-

dary predicate, but that would be in conflict with the impossibility of uses such as

*he was heard get up. While there is perhaps nothing in principle against syntactic

blends, it is in this case a lot easier to simply treat the oblique noun phrase as the

subject of both the bare infinitive and the conjoined ing-clause.

(36) a. I note also that he complains that goods come into this country from

third world countries where wages are far lower, but I’ve heard him get

infinitives, according to Declerck, cannot occur in this position. However, corpus data show that bare infinitives, too, can occur in the focus of pseudo-clefts, as in (iii), casting doubt on the validity of this argument. (iii) What I like to see is an actor really inhabit his or her body and then

inhabit the space around them so it reaches out. (CB) Declerck’s other arguments are based on constructions that I could not find to actu-ally occur in corpus data. First, he argues that ing-complements can be moved to subject position under passivisation of the matrix verb (e.g. the moon rising over the mountain was seen by many people last night (1982: 2), whereas bare infinitives cannot. Second, he uses the ing-complements fronted to subject position to show that the first NP can trigger agreement on the verb (e.g. your teachers quarreling with each other last night has/have been overheard by some of the students (1982: 6)) and must therefore be the head, unlike in ing-complements equivalent to bare infinitives. I do not exclude the possibility that these constructions exist, but am strongly inclined to question grammaticality judgements in such contrived areas of grammar. It is telling, for instance, that Gee (1977: 465) comes to the entirely oppo-site conclusion that the subject of a fronted ing-complement would remain in the oblique case. Pending empirical evidence, I take it therefore that one one-constituent analysis for the ing-complements of perception verbs suffices.


up several times before and saying third world countries need help

(ICE-GB)

b. Obviously, the cigarette commercials – I mean, we haven’t seen those

in 25 years. [...] Something almost a little painful about it, too, seeing

Lucy and Ricky light up and encouraging us to. (CB)

Third, the entire sequence of oblique noun phrase and ing-clause can be made the

focus of a pseudo-cleft, as in (37), again implying that it can function as a single

constituent.

(37) a. What we are seeing is history repeating itself Mr Court said. (CB)

b. There was one morning she had to literally drag him all the way round

here and all you could hear was him going No. No. No. (CB)

Fourth, existential there exclusively occurs in subject positions, which means that

the slot preceding the ing-marked verb in the examples in (38) must be a subject

slot.

(38) a. We couldn’t see there being any problem. (CB)

b. From a personal viewpoint I could not see there being sufficient Satur-

days for five successive sevens tournaments to be held in April if the

Five Nations were to be moved. (BNC)

Regarding the ing-clauses following with, similar arguments can be put for-

ward to prove that the slot between with and the ing-marked verb is a clausal subject

slot. Evidence to this effect is, for instance, the possibility of using existential there

in this position, as in (39).

(39) Well with there being so few people erm did you all do everything

[…]?

All this indicates that the ing-complements with perception verbs and in adverbial

with-constructions can be analysed as clauses expressing their own subject. The fact

that these subjects are obligatorily in the oblique case casts doubt on Huddleston &

Pullum’s (2002) argument, as there is now a type of ing-complement that behaves

differently from other ing-complements in terms of its internal syntax.


Still, this is not the time to jump to conclusions. First, as there is no telling

where generalisation stops, it is perfectly conceivable that language users do gener-

alise beyond syntactic and (non-categorical) formal differences to infer a category of

ing-clauses or gerund-participles. Indeed, the reduced /in/ variant of the ing-suffix

has been extended from participles to gerunds, even though the fact that the varia-

tion never levels out completely across the two (historical) categories indicates that

the generalisation underlying its extension must be relatively weak. Similarly, his-

torical evidence occasionally shows nominal means of marking the direct object (i.e.

with an of-phrase) in historically participial forms, as in (40). The only source of this

use can be the gerund (see Chapter 8), which means again that language users have

generalised across gerunds and participles, yet the fact that the alternative expression

of the object never caught on in participles indicates again that the generalisation is a

weak one.

(40) a. They .. came into his gardein .. and found him weding of his ground.

(1553, OED)

b. [Ulysses says] But I, not forcing of their giftes, did loue my wedlock

best. (1567, OED)

In other words, the evidence does not entirely exclude a broad category of gerund-

participles. All we can conclude for certain is that such a category is not enough. If

the category of gerund-participles exists, it has to generalise over more specific rep-

resentations that must be stored independently (for example, specific representations

telling language users which subject marker they can use).

Second, positing these lower-level representations, however, still does not

necessitate a recurrence to the traditional division between gerunds and participles.

It is true that both the complements to perception verbs and the ing-clauses intro-

duced by with historically derive from participles, but this need not mean that they

are still participial clauses in the eyes of present-day language users. The fact that

these ing-clauses take only oblique subjects is as much an argument against putting

them on a par with other ing-complements allowing genitive subjects as it is an ar-

gument against equating them with other historically participial constructions that

sanction nominative subjects. Moreover, looking beyond the constructions with ex-

plicit subject, there are, as Huddleston & Pullum point out, “numerous constructions

where no subject is permitted” and where “there is no difference at all in the internal

form of the constructions” (2002: 1221).


Let us therefore consider again the merits as well as the weaknesses of the

traditional distinction between gerunds and participles. The distinction is primarily

authorised by the persistent link between gerund clauses and noun phrases on the

one hand and between present participles and past participles, adjectives or adverbs

on the other. With respect to gerunds, this link is captured by the principle of para-

digmatic regularity operating in the use of gerundial ing-complements, the interpre-

tative and distributional relevance of which is hard to deny (see the discussion of

paradigmatic regularity in Section 1). As to a similar link between present partici-

ples and past participles, adjectives and adverbs, the evidence is less compelling, but

a relationship is visible nonetheless, again in the form of distributional and interpre-

tative parallels. A clear example comes from the use of participles with subordina-

tors, as illustrated in (41). Distributionally, it is impossible to ignore the fact that the

set of subordinators combining with present participles matches almost exactly the

set of subordinators taking past participles (cf. Quirk et al. 1985: 1004-6).62

(41) a. she is often depicted with a cat’s head while carrying a shield and rat-

tle, to frighten away thunder-storms (BNC)

b. It was beyond her mother’s skill to put the car in the garage while dis-

tressed. (BNC)

Likewise, both present and past participles can occur as relative postmodifiers to

nouns, as disjuncts, or as complements with causative verbs or perception verbs. The

62 Quirk et al. (1985: 1004-6) explicitly compare which subordinators go with past and present participles respectively. The only discrepancies they find are where, wherever, as (expressing manner) and as soon as, which combine only with past participles, and after, before and since, which only take ing-clauses. The latter ex-ceptions, however, are prepositions and the ing-clauses they combine with can be regarded as gerundial (Quirk et al. 1985: 1005-6). As to the former, the only real exception, as far as I can see, is as expressing manner (e.g. do as told vs. *do as being told). By contrast, where, wherever and as soon as are in fact attested with present participles, as shown in (iv). (iv) a. Make sure to articulate these clearly, alternating your plucking fingers

except where raking from the 1st to 2nd strings (notes E to A and D to A). (BNC)

b. Report to the ISSC (405 Main Building) as soon as arriving on cam-pus. (Google, 28 November 2007)

c. for the words “United Kingdom”, wherever occurring except in the expression “the United Kingdom’s obligations”, substitute “Isle of Man” (Google, 28 November 2007)


examples in (42) illustrate the latter parallel (cf. (35) above) (cf. Visser 1963-73:

ch.16).

(42) a. If you hear President Bush using the word “brutalize” you will hear the

word misused (CB)

b. And then quite suddenly ... he heard himself addressed. (1945, Visser

1963-73: 2381)

Given also their historical relevance (see Chapters 8 and 9), it is difficult to believe

that these paradigmatic relations would go unnoticed by language users.

At the same time, it must be recognised that the behaviour of gerunds and

participles diverges from that of their paradigmatic sister constructions (cf. Ross

1973). As already pointed out, paradigmatic regularity does not give a perfect pre-

diction of the distribution of gerundial ing-complements, which continues to show

important gaps (see Section 1), and the distribution of participles is similarly defec-

tive if compared to that of adjectives (Huddleston & Pullum 2002: 1221). Moreover,

apart from showing gaps, the distribution of ing-clauses also extends beyond the

confines set by paradigmatic regularities. The most obvious example is the use of

ing-clauses in extraposed subject position, as in (43a-b). The form of the subject

suggests that these ing-clauses are gerundial, but it is clear that the syntactic position

in which they occur is not a nominal one. Similarly, some ing-complements follow-

ing prepositions, such as those illustrated in (43c-d), have no nominal alternative (cf. ?stop her from it; *you have no way of it), and the same is true of the paradigmati-

cally deviant constructions involving worth and worthwhile in (43e-f) (see Chapter

4).

(43) a. it was freaky, it was scary watching it (CB)

b. It’s no use your keeping on going accidentally on purpose by Warren-

ers farm. (BNC)

c. Their instructions were simple: to stop her from getting the case. (CB)

d. You have no way of knowing what yours will be like, so be prepared.

(CB)

e. it’s worth listening to all the warnings from social scientists about our

Hurried Children (2003, TIME)

f. But as Tigan, 32, thought more about the idea, he decided it was

worthwhile pursuing seriously. (1979, TIME)


On the semantic side, ing-clauses and their paradigmatic variants may also

have grown apart in meaning, violating the interpretative dimension of paradigmatic

regularity. For example, the preposition in, used with ing-clauses, as in (44a), can

express semantic relations (e.g. causality) that it does not normally express with or-

dinary noun phrases. The same applies to ing-clauses introduced by the idiomatic

there is no, which can be used to express deontic modality, as in (44b), a meaning it

cannot convey when combined with a noun phrase. A particularly striking example

is that in (45a), where keep is used with an ing-complement as though it is synony-

mous to prevent in the same construction, in violation of the far more common and

paradigmatically supported interpretation of the same surface sequence illustrated in

(45b), where keep means ‘cause to continue’.

(44) a. Worse, in doing so, he held the indigenous population up to ridicule.

(CB)

b. there is no disguising the political charge packed by yesterday’s attack

(CB)

(45) a. This includes criticising a female partner’s family or, swearing at a

partner, preventing her from having money for her own use or “trying

to keep her doing something she wants to do”. (CB)

b. A competent laundress kept many irons warming in reserve (CB)

All this shows that ing-clauses can break out of their paradigmatic mould, histori-

cally extending their range of application on the basis of different alternative analo-

gies and through construction-internal semantic developments. It is through such

developments that the categories of gerunds and participles can begin to crumble

down.

The problem-constructions in examples (43)-(45a) depart from the canonical

use of gerunds and participles, but (with the possible exception of (45a)) still lean

closer to either one category in resembling other, more prototypical members of the

category. There are also environments, however, where the distinction between ger-

unds and participles gets blurred more thoroughly. One such shady area is formed by

the complements of negative implicative verbs, most notably prevent (cf. Jespersen

1940: 142-50). Historically, prevent started out with complements that were unam-

biguously gerundial, as shown by the genitival form of the subject in (46a). How-

ever, with genitival subjects becoming less current, the canonical form of the ing-

complements following prevent became that in (46b), with an oblique subject that


allows for an alternative structural interpretation, analogous to the causative con-

struction illustrated in (47a-b), with an object-controlled participle.

(46) a. it behoves them that I be convicted, to prevent their being indicted for

Felony. (1685, PPCEME)

b. And now for Lippius’s clock! said I, with the air of a man, who had got

thro’ all his difficulties – nothing can prevent us seeing that (1759-67,

CLMETEV)

(47) a. Keep him not waiting for his love too long. (1678, CEMET)

b. people were coming and going constantly, and the doorkeeper was

kept opening and closing the door. (1902, CLMETEV)

That this reinterpretation effectively took place at one point in the history of English

is illustrated by the examples in (48), where what used to be the subject of the ger-

und is fronted as the subject of passivised prevent. As the examples show, this pas-

sivised construction was particularly current in Late Modern English (see (48a-d)),

although it is still marginally attested with prevent and stop in present-day usage

(see (48e-f)).63

(48) a. Mrs. Bennett was prevented replying by the entrance of the footman

with a note for Miss Bennett (1813, CLMETEV)

63 I have no ready interpretation for the fact that present-day usage shows less ex-plicit evidence of the participial analysis in the form of the passivised construction. On the one hand, the absence of evidence may be an artefact of the cognitive com-plexity principle (Rohdenburg 1996; see Chapter 3). Among the complementation patterns with prevent, bare ing-complements compete with from-complements (as in (49a-b) above). Given that the variant with from is structurally more explicit it is to be expected that it is the more favoured one in cognitively complex environments. That this is in fact so is demonstrated by Rohdenburg (2003) for the active uses of prevent, but evidently, the tendency is the more likely to manifest itself in passive uses of prevent, which are cognitively complex by default. On the other hand, the historical tendency has been for bare ing-complements to replace the variant with from, at least from the nineteenth century onward (Mair 2002; Rohdenburg 2003), so the disappearance of the from-less variant with passive prevent goes against the his-torical trend. An alternative explanation, therefore, is that one of the constructions on which the participial interpretation is paradigmatically modelled has fallen out of use, namely the use of prevent with an animate object in a monotransitive clause (as in (49c) above). On this interpretation, the participial reading may in fact have lost some of its salience for present-day language users and it is this loss of salience that is responsible for the demise of bare controlled ing-complements with passive pre-vent.


b. I thought you would be prevented coming. (1900, CEN)

c. they got into the pantry, [...] but were stopped going further by the pan-

try door being fastened outside (1882, POB)

d. It may seem surprising that the balloon, which could not be prevented

falling in the water, is yet enabled to ascend from the grip of the waves

by the mere discharge of ballast. (1902, CLMETEV)

e. Nearly two tons of pure cocaine has been stopped getting on the mar-

ket. (CB)

f. There was the lady who wished to wear her miniscule cross around her

neck at work but was prevented doing so by the rules of the employers.

(Google, 29 November 2007)

This historical cross-over between categories shows that gerunds and partici-

ples may in fact be impossible to distinguish. At the same time, the reason why the

cross-over was possible at all is that the argument-structures of prevent themselves

caused the principles of paradigmatic regularity to get confused. First, prevent, like

other verbs in the participial causative construction (e.g. keep), also occurred in

other three-place constructions, particularly with an object-controlled ing-clause

introduced by from, as in (49a-b). Second, prevent could take animate objects, as in

(49c), which provided a model for the objects in the reinterpreted participial con-

struction.

(49) a. Many other reasons prevented them from daring to look their true

situation in the face. (1770, CLMETEV)

b. but I was prevented from putting my design in execution by a fit of

illness (1751, CLMETEV)

c. but the Baronet prevented him by bidding him not aggravate his mean-

ness (1766, CLMETEV)

In sum, then, since it is paradigmatic regularities themselves that cause all the con-

fusion, the development of prevent actually indirectly demonstrates the linguistic

relevance of these regularities. Yet at the same time the development shows that

where the generalisations of paradigmatic regularity happen to overlap, there is no

ground left to tell gerunds and participles apart.

There are at least two other shady areas of grammar where, through historical

developments, the distinction between gerunds and participles has become blurred.

In these cases distinctive opacity is not dependent on an accidental coincidence of


two patterns of paradigmatic regularity, as in the case of prevent. Rather, the use of

ing-clauses departs from any paradigmatic regularity, resulting not in ambiguity

between the traditional categories of gerunds and participles but in indeterminacy.

The first example of indeterminacy involves the ing-complements with aspec-

tual verbs (cf. Huddleston & Pullum 2002: 1221). Historically, these derive either

from gerunds or from participles (see Chapter 9), but grammatical changes have

dissociated them from their paradigmatic sister constructions. For example, in the

case of begin, incipient grammaticalisation has broadened the selectional restrictions

beyond what is normally allowed by the verb’s transitive use. This is illustrated by

the examples in (50), showing that when begin combines with ing-complements, the

subject need no longer be construed as an instigator, in contrast to the ordinary tran-

sitive construction.64 With go on, on the other hand, the syntactic function of the ing-

clause has changed from adjunct to complement, and on the new reading no para-

digmatic association is left – compare the examples in (51), with (51b) and (51c)

illustrating adjunct and complement uses respectively.

(50) a. police began a special operation (CB)

b. Police yesterday began investigating the security breach (CB)

c. Twenty-six hundred state employees in Connecticut begin receiving

their layoff notices today. (CB)

(51) a. Back home in surburbia [sic] life goes on as usual for Jane’s family

and friends. (CB)

b. I kept calling the hour down Little Wild-street; I came round to an ob-

scure place, then I saw a man moving; when I came up to him I saw

three men, one of which proved to be the prisoner, stand rapping at the

64 The grammaticalisation of aspectual verbs with ing-complements, as argued for by Brinton (1988), is certainly not the most straightforward example of auxiliation. There is no sign of formal reduction, for instance, and the first evidence of selec-tional restrictions getting loosened is surprisingly early to appear. One of Mair’ (2003: 331) oldest instances of begin with ing-complement already reveals the use of weather-it in subject position, well before the combination of begin and ing-complement started to gain any frequency. The only explanation I can give is that the use of ing-complements with begin has been modelled after that of to-infinitives and that (incipient) grammaticalisation is not isolated to a single highly specific con-struction type. The change resembles the grammaticalisation-through-analogy changes pointed to in recent grammaticalisation literature (e.g. Noël 2006), though in this case it is not a lexical item in a grammaticalising construction that attracts other lexical items in the same construction, but a lexical item in combination with one grammatical variant that attracts another grammatical variant.


back door [...]: I went on crying the hour into Great Wild-street, in or-

der to see for some assistance (1749, POB)

c. I want my buds to swell slowly, and go on swelling, so that the leaves

on my tree are full and fat (CB)

Arguably, these individual developments each present only relatively slight, seem-

ingly isolated breaches of the principles of paradigmatic regularity. However, the

(assumedly) independent developments have also converged in a more schematic

aspectual construction. Presently, ing-clauses are found with all aspectual verbs in

English (e.g. begin, cease, continue, finish, go on, keep, start, stop, etc.), so we can

safely postulate a construction specifying that aspectuals combine with ing-clauses

(further evidence of the uniform behaviour of aspectuals in English is given in

Chapter 9). Because the ing-clauses in this construction have developed from differ-

ent sources their status as gerunds or participles is highly uncertain. That is, para-

digmatic dissociation is double: at the level of individual verbs, ing-complements

have begun to deviate from paradigmatic regularity, while at the level of the more

schematic aspectual construction, there is no systematic paradigmatic analogue left

at all. In this light, it is highly unlikely that language users continue to primarily

associate the various ing-clauses following aspectual verbs either with gerunds or

with participles. Rather, the ing-clauses in the aspectual construction are indetermi-

nate.

The double-levelled constructional dissociation that characterises the use of

aspectual verbs with ing-complement is given a (simplified) schematic rendering in

Figure 7.3 (which again only marks instantiating relations between constructions,

not integrating relations). The full lines in Figure 7.3 represent relations of instantia-

tion; the grey dotted lines represent ties of instantiation that are weakening or sev-

ered. Constructional dissociation is found both at the level of specific verbs (aspec-

tual begin and go on) and at the level of the more abstract construction ‘Aspectual V

+ Ving’, which captures the regularity that aspectual verbs take some kind of ing-

clause but whose ing-clause no longer falls under either the category of gerunds or

participles.

The second example of indeterminacy involves the rather more marginal con-

structions I will discuss in Chapter 10 under the heading of integrated participle

clauses. These include a cluster of at best distantly related constructions, all display

ing subject-controlled ing-complements in non-nominal slots. One use of integrated

participle clauses is with emotive adjectives, as illustrated in (52).

Figure 7.3. The aspectual construction in a (possible) constructional network.

…began a special operation…

…began investi-gating the…

…began receiving their…

Lexical verb begin

Aspectual begin

Transitive con-struction

Gerundial ing-complementation

…go on swell-ing…

…went on cry-ing…

…goes on as usual…

Aspectual go on

Lexical verb go on

Adverbial modi-fication

Adverbial partici-ple

Aspectual V + Ving

Gerund Participle


(52) a. One English boy admitted that he was fed up walloping English kids.

He wished he could bash a couple of Yanks but he didn’t want to get

his father into trouble. (1952, TIME)

b. lease has been on a year-to-year basis but we are hopeful negotiating a

20-year deal with the Council. (CB)

It will be noted that the ing-clauses in (52) paradigmatically correspond neither to

nouns nor to past participles, adjectives or adverbs. As is argued in Chapter 10,

some such uses derive from adverbial participles, reinterpreted as complements

(along the lines of the reinterpretation giving rise to aspectual go on; see (51)

above). In some environments the distinction between adverbial clause and com-

plement is vague, as with comfortable in (53a), suggesting that in such environments

the break with the participial source construction is not a radical one. But the con-

struction has gained some degree of independent productivity so that the use of inte-

grated participle clauses has been extending beyond the contexts of reinterpretation,

as when combining with fed up and hopeful in (52) above.

(53) Yet Reagan is far more comfortable addressing human issues than ab-

stract interests (1988, TIME)

This means that, as in the case of aspectuals, paradigmatic dissociation is situated

both at the level of individual predicates and at the level of a more schematic con-

struction. Again, the most plausible conclusion is that at this point the distinction

between gerunds and participles has become irrelevant and that uses as in (52)-(53)

are simply indeterminate.

Let us recapitulate. The category of gerund-participles, postulated by Huddle-

ston & Pullum (2002), is an overgeneralisation. This does not mean that it cannot

have cognitive reality – for all we know, language users may overgeneralise (if only

a little) – but neither the form nor the internal syntax of ing-clauses can be described

accurately without specifications that pertain to less schematic constructions. The

best alternative, however, is not necessarily the traditional distinction between ger-

unds and participles. In as far as this distinction is based on paradigmatic correspon-

dences, it has a degree of linguistic relevance, but again, it generalises over more

specific constructions and at the same time it is not comprehensive of the variety of

constructions that exist. Between the folds of gerunds and participles new uses arise

that do not clearly belong to either category. These, in their turn, might rightly be

thought of as gerund-participles, though not in the sense of a global super-category


but rather as newly emerging local regularities that deviate from set regularities in

the use of ing-clauses.

3. Conclusions

If the use of English ing-clauses can be compared to a labyrinth, the purpose of this

chapter has been to walk along its various corridors and to show that none quite

leads to a centre or not to a single centre, unless perhaps if one is willing to cheat a

little and climb some of the obstacles. That is to say, no single generalisation can

account for the behaviour of ing-clauses in its totality. Indeed, even when we try to

carve up the domain into smaller areas (gerunds, participles, ing-complements),

generalisations are hard to arrive at. At the same time, it is impossible to believe that

language use is not organised around certain regularities. In the area of ing-clauses

and ing-complements, the most powerful of these are still the paradigmatic regulari-

ties that underlie the distinction between gerunds and participles. Yet at the same

time, it must be recognised that these generalisations are not fully deterministic,

leaving gaps that have to be accounted for by other means, and that part of usage

simply escapes their grasp.

Looking ahead, the labyrinth that exists today is the outcome of a complex set

of historical developments, and to explore some of these is the purpose of the fol-

lowing chapters. Apart from raising this central issue, the discussion in this chapter

also has some practical consequences. If the distinction between gerunds and parti-

ciples still makes some sense today, it makes more sense historically. This motivates

the division between gerunds and participles (or their historical offshoots) as the

respective topics of Chapters 8-9 and Chapter 10. At the same time, some uses sim-

ply defy classification, which I hope will pardon the inclusion of ing-complements

with aspectual verbs under the heading of gerundial complementation in Chapter 9.

CHAPTER 8: The rise of verbal gerunds65

The distributional resemblance that obtains between nouns and gerund clauses in

Present-Day English (see Chapter 7) is the current stage of a drawn-out historical

development. The syntactic peculiarities gerund clauses still show (their occasional

use with determiners, the variation in the form of their subject) likewise constitute

the layered reflex of previous changes. That is, part of the reason why gerund

clauses distributionally and syntactically behave the way they do is that they histori-

cally derive from a nominal construction.

Even this nominal source construction still has its descendents populating the

grammar of English, in the form of nominal gerund constructions. Thus, next to ger-

und clauses or (henceforth) ‘verbal gerunds’, which have a verbal head and whose

internal syntax is that of a clause, English has ‘nominal gerunds’ whose head is a

noun and whose internal syntax is entirely that of a noun phrase. To illustrate the

distinction, the internal syntax of the constituent headed by visiting in (1a) shows no

trace of nominal origins, containing a subject expressed by an oblique noun phrase

(them), a verbal negator (not), a direct object (you) and an adverbial (anymore). By

contrast, mocking in (1b) heads a purely nominal construction, expressing its (no-

tional) subject by means of a possessive determiner (his), taking a modifying adjec-

tive (casual), and using a periphrastic construction to express its (notional) object (of

the old left).

(1) a. You’re not so bothered about them not visiting you anymore. (CB)

b. Both his old comrades were irritated by his casual mocking of the old

left (CB)

Ignoring the categorial difficulties caused by the formal similarity between

gerund clauses and participle clauses (see Chapter 7), the variation in (1) betrays a

gradient that is familiar from language typology and generally characteristic of

nominalisations, manifesting itself in the cross-linguistically attested continuum

between action nominals and nominalised clauses (Comrie 1976; Comrie & Thomp-

son 1985; Koptjevskaja-Tamm 1993). An action nominal, such as refusal in (2a),

has as its head a “lexically derived noun”. In contrast, a nominalised clause, such as

the that-clause in (2b), contains “no evidence in favour of viewing its head as a lexi-

65 This chapter has been adapted from De Smet (forthc. b).

234 – CHAPTER 8: The rise of verbal gerunds

cal noun” – in other words, in nominalised clauses the head is a verbal form and it is

the clause as a whole that is nominalised (Comrie & Thompson 1985: 391-2).

(2) a. Sir Edwin’s refusal to communicate directly with his wife continued.

(CB)

b. the King of Thebes discovered that he had lived his life under a curse.

(CB)

As for the English gerund, nominal gerunds can be situated closest to the action

nominal end of the continuum, whereas verbal gerunds take a more intermediate

position (cf. Lees 1966; Ross 1973; Langacker 1991; Heyvaert 2003, 2004).

The purpose of this chapter is to examine the historical relation between

nominal and verbal gerunds, in order to get to a better understanding of how verbal

gerunds functioned in different historical periods and how they grew into the highly

successful clause-type they are today. The answers to both these questions are rele-

vant to the history of gerundial complementation, discussed in Chapter 9, yet they

are also of interest in their own right, in part because the type of change clearly re-

curs cross-linguistically (Disterheft 1981; Haspelmath 1989) but also because, as

will be argued here, the emergence of verbal gerunds provides an illustration of the

way in which functional pressures operate on language change and are mediated by

the functioning of synchronic systems of choices.

The following discussion gradually circles in on the general goal of unearth-

ing the language-internal factors driving the emergence of verbal gerunds. Section 1

provides a first rough picture of the development of the gerund, summarising the

existing literature and introducing the interlinked issues of historical causation and

the ‘functionality’ of language change, which have still been insufficiently ad-

dressed. This leads to a reassessment of the historical developments that affected

both nominal and verbal gerunds. Section 2 sets out to describe in more detail what

historical changes have in fact taken place, paying attention not only to verbal ger-

unds but also to their nominal counterparts. Section 3 then digs more deeply into the

precise relation between the different types of gerund constructions, examining

which factors determined synchronic choices between gerund constructions in actual

usage. Section 4, finally, summarises the findings and reconsiders the issue of

whether – and if so, in what sense – developments in the gerund were steered by

functional motivations. It is proposed that developments have primarily been shaped

by relations of functional similarity and dissimilarity between gerund constructions

as well as by the combined functional pressures of iconicity and economy, with lan-

CHAPTER 8: The rise of verbal gerunds – 235

guage contact and the relation of gerunds to infinitives as additional determining

factors.

Once again, for the purpose of the discussion, I have chosen to focus primar-

ily on a limited set of constructions (see Section 2) as well as on a restricted histori-

cal period. The period examined most closely starts in Middle English and stretches

into the Early Modern period. The restriction is justified by the fact that it is this

period that saw the major changes in the relationship between verbal and nominal

gerunds and by the finding that by the end of it usage already closely resembled pre-

sent-day usage.

1. The verbalisation of the gerund

Essentially, the history of the gerund consists in a gradual reconfiguration of the

structure of a noun phrase into that of a non-finite clause, coupled to a growing

quantitative predominance of the verbal construction over the nominal one. This

process of verbalisation and the changes concomitant to it have already received due

attention in the diachronic literature (Jespersen 1940: 108-50; Einenkel 1914; Dal

1952; Mustanoja 1960: 566-78; Visser 1963-73: 1165-1217; Emonds 1973; Tajima

1985, 1996, 1999; Donner 1986; Jack 1988; Houston 1989; Van der Wurff 1993;

Fanego 1996b, 1998, 2004a, 2004b; Miller 2002; Kranich 2006). The extant litera-

ture agrees on the following course of events:

Old English morphology possessed a derivational mechanism that added -ing

or -ung to verbal stems to create deverbal nominalisations, as illustrated in (3a). The

behaviour of these Old English gerunds was noun-like in all respects, much like that

of the derivations with the cognate suffixes -ing and -ung in Present-Day Dutch and

German. From Late Middle English onwards, however, gerunds with verbal features

began slipping into the language. As (3b) shows, these early verbal gerunds mostly

took the form of constructions with a non-genitive subject or object, and were espe-

cially used following prepositions. Still, it was not until the Modern English period

that the process of verbalisation came into full swing and verbal gerunds became a

truly indispensable part of English grammar. During the Modern period the fre-

quency of verbal gerunds rose dramatically, and, as shown in (3c), verbal gerunds

spread beyond their predominantly prepositional use. In addition, as illustrated in

(3d-e), the same period saw some further innovations in the internal syntax of the

gerund, such as the introduction of passive and perfective gerunds, and the increased

use of ‘hybrid’ gerunds, which combined nominal with verbal features. The latter

have again disappeared in the course of the 19th century – with the notable excep-


tion of verbal gerunds with genitive subjects, as in (3f). The result of these various

developments is the situation as it stands today.

(3) a. ðurh ðæra sacerda blawunge toburston ða weallas. (c1000, Visser

1963-73: 1165)

‘through the blowing of the priests, the walls broke down’

b. Sain Jon was … bisi In ordaining of priestes, and clerkes, And in cast-

ing kirc werkes (c1300, Tajima 1985: 76)

c. They come so to purpose, that hee can not refraine telling them. (1561,

Fanego 1996b: 38)

d. ... whose pillow she kissed a thousand times for having borne the print

of that beloved head. (1580-1, Tajima 1999: 269)

e. which … may very much contribute to … the driving Barbarity out of

our Streets. (1711, Visser 1963-73: 1214)

f. But there was no question of our taking the exam in it. (CB)

The picture gets more complicated when we turn from the superficial descrip-

tion of diachronic developments to the underlying causes. If there is anything on

which the existing literature agrees, it is that the verbalisation of the gerund is best

seen as the result of a complex interaction between different factors. At least the

following elements appear to have been involved: First, highly relevant to the ver-

balisation of the gerund is the occurrence of ambiguous constructions. Because such

constructions sanction both nominal and verbal readings they could invite structural

reinterpretation (Visser 1963-73: 1191; Jack 1988; Fanego 2004a). The ambiguities

arise when a gerund form has no modification, as in (4a) or has no modification that

is exclusive to noun phrases, as in (4b).

(4) a. And ȝyue to almes þat yche þyng Þat ȝe haue wune wyþ okeryng.

(a1400 (c1303), HC)

‘and give to alms everything that you have won by means of usury’

b. And þis poynt þorow þe ȝiftes of God he may at sum tyme haue

experience of, þat after sykyngge and mornynge for wrecched seruitute

of þis wordle, is so ileft vp in-to heiȝ clennesse of conscience and holy

contemplacioun þat he haþ in a maner forȝyten al þis wordle. (c1400,

PPCME)

‘and this may at some time through the charity of God be experienced

by him who after sighing and mourning for the wretched servitude of


this world is to such a degree lifted into the high purity of conscience

and holy contemplation that in a way he forgets all this world’

Second, there is the possible influence of the present participle, whose formal

identity to the gerund may have given rise to confusion between different construc-

tion types. Jack finds that the dialects that first saw the appearance of verbal gerunds

did at that time not have participles in -ing but in -nde so that “coalescence of the

verbal noun [i.e. nominal gerund] with the present participle cannot have been the

source from which the [verbal] gerund arose” (1988: 27). Still, he does not exclude

the possibility that subtler forms of interference have played a role in the further

development of the gerund. In this spirit, Fanego (1998) sees confusion with the

present participle as one reason for the increased use of non-genitive subjects with

the gerund in Early Modern English. Houston (1989), on the other hand, suggests a

link between participial adjuncts and the fact that verbal gerunds first emerged in

contexts where they followed a preposition – a construction she considers function-

ally similar to the participial adjunct construction. Even if both these claims are –

despite their intuitive appeal – hard to verify, a certain degree of confusion between

gerunds and participles is apparent from the historical data (cf. Chapter 7; see, in this

respect, also Denison 1993: 404-8 on the possible, albeit marginal, influence of the

gerund on the development of the progressive).

Third, a case has been made for French influence (Donner 1986; Jack 1988;

but see Jespersen 1940: 90; Miller 2002: 337-8; and especially Kranich 2006 for a

critical view). Jack comes to the conclusion that association with the French géron-

dif could have promoted the emergence of verbal gerunds. He finds that the French

gérondif, like the first verbal gerunds in English, was found with prepositions, and

could take non-genitive subjects, as in devant le soleil levant (‘before the sun ris-

ing’) (1988: 59). Notice that Jack finds his evidence in Old French and highlights

constructions that are specific to that period. The only gérondif construction that

survives into Modern French has no overt subject and is exclusively found with the

preposition en ‘in’. The other constructions treated by Jack as relevant seem to have

disappeared from the language early on, as is suggested by the fact that all instances

quoted in the historical grammars by Nyrop (1930: 245-7), Anglade (1966: 196),

and Brunot (1966: 246) date from the thirteenth century or earlier (see also Martin &

Wilmet 1980: 116-7). In light of this, the influence of these gérondif constructions

on the first use of verbal gerunds may have been fairly marginal (cf. Kranich 2006),

though, as will appear in Section 4, the controlled gérondif construction with en ap-


pears to have played a more influential part later on and the same may be true of

certain other French constructions.

Fourth, once verbal gerunds started to spread and (to some extent) replace

nominal gerunds, the verbalisation of the gerund began to be affected by the pres-

ence or absence of a determiner preceding the gerund. Fanego (1996a; 1998; 2004a)

has demonstrated that verbal features first appeared in gerunds without an overt de-

terminer, later spreading to gerunds with a possessive determiner, and finally to ger-

unds with an article (see also Chapter 4). This order of diffusion is still reflected in

the fact that the only type of nominal gerund that has more or less disappeared from

English usage is that illustrated in (5a), while the verbal features in gerunds with

definite articles, as in (5b), have later on even receded again.66

(5) a. By eating of Garlike, a man may the safelier goe into a suspected aire,

and by stinking places. (1563, OED)

b. I went to this Newcastle in Staffordshire to see the makeing the fine

tea-potts cups and saucers of the fine red earth. (1702, OED)

Fanego interprets this order of change in terms of Timberlake’s (1977) markedness

principle, arguing that verbalisation first affected those environments that were clos-

est to clauses anyway in showing the least explicit nominal features. In other words,

a ‘hierarchy of relative nominality’ affected the way verbalisation proceeded (2004a:

38).

Still, none of this tells us why verbal gerunds were as successful as they were

and managed to develop from a quirk of grammar into a major clause type. For the

success of verbal gerunds, the literature provides only one explanation. Einenkel

(1914) and Jack (1988) suggest that the reason for success lies in the failure of the

English infinitive to combine with prepositions other than to, implying that the ver-

bal gerund fills a syntactic gap. Jack writes:

66 Neither have fully disappeared. Though decidedly rare, the gerund construction with of-phrase but without overt determiner is not entirely ungrammatical in Pre-sent-Day English, as demonstrated by Heyvaert (2004). Some of the examples she reports are given in (i): (i) a. What shouldn’t be happening in schools is breaking of bones. (CB) b. tthe Ministry of Agriculture will allow burning of herbage seeds, reeds,

lavender. (CB).


[I]n its absence from use in prepositional phrases the infinitive lacked a syn-

tactic property of nouns, and accordingly there was a gap in the range of syn-

tactic patterns in which the infinitive functioned as a nominal form of the

verb. The development of the gerund filled this gap, by providing a nominal

form of the verb that could freely be used following prepositions; and, as

Einenkel […] argued, it seems likely that the development of the gerund was

either initiated or promoted by pres-sure to remove this syntactic gap. (1988:

61-2)

Although this view is certainly in line with the early association of verbal

gerunds with prepositional uses, it is also problematic in at least two respects. A first

shortcoming to Einenkel and Jack’s account is pointed out by Fanego (1996a; 2004).

Fanego recognises that the inability of the infinitive to combine with prepositions

created some kind of gap, but in her view this gap could be filled by nominal as well

as verbal gerunds. Thus she states that:

[I]t seems likely that the great expansion in the use of prepositions in the

course of the Middle English period […] gave rise to a situation in which a

form of the verb capable of being used prepositionally was often called for:

the gerund may have served to fill this gap. But this process, as I see it, could

have been independent of whether the gerund was syntactically nominal or

verbal, for in sequences like Wiþouten doying of any harme […] or in casting

kirc werkes […] the advantage of the gerund over the infinitive does not de-

pend on the nature of its object (i.e. of-phrase or noun phrase), but rather on

its ability to occur in a syntactic position from which the infinitive was ex-

cluded. (2004: 16–7)

A second problem is what precisely is to be understood by the idea of ‘pres-

sure to remove a syntactic gap’? The notion of a gap leads straight into the thorny

issue of the extent to which language change is driven by functional considerations.

The careful position to take in the matter is to be sceptical (see, for example, Harris

& Campbell 1995, Croft 2000, Heine & Kuteva 2005, and more extremely, Lass

1998), but then again the possibility of functionally driven language change is not to

be abandoned lightly (García 1999; Haspelmath 1999; Hawkins 2004). In this re-

spect, the history of the gerund provides a relatively straightforward test-case for

functional claims outside the realm of theoretical debate. In all likelihood, the pres-

sures referred to by Einenkel and Jack did not by themselves cause the verbalisation


of the gerund, yet they may, on a functional view of language change, have provided

the incentive to seize the opportunities that arose by fully exploiting verbal gerunds

when they came along as the more or less accidental result of other factors. How-

ever, as long as the evidence is restricted to the finding that verbal gerunds most

commonly occurred with prepositions, the notion of gap-filling will remain difficult

to assess.

The main questions to be addressed in this chapter relate back to the problems

emerging from the preceding discussion: Is there any indication that the gerund was

used to fill a syntactic gap or, more generally, that the rise of verbal gerunds was

somehow functionally motivated? And, if so, how did verbal gerunds relate to

nominal gerunds in this? As I will show, the second question can be answered con-

clusively, and gives us a number of important clues to answering the first. In order

to arrive at this point, however, we first need to get a more detailed picture of the

actual quantitative developments in verbal and nominal gerunds.

2. Quantitative developments in nominal and verbal gerunds

To come to grips with the rise of verbal gerunds, as well as with the relation of ver-

bal to nominal gerund constructions, three main types of gerund construction can be

distinguished, depending on the expression of the patient argument role and the de-

terminer preceding the gerund. The three types are represented in (6):

(6) a. I shall teach him a lesson, for filching hens or cocks (1552-63, HC)

b. That we fall not into disordering of ourselves by anger. (1559, OED)

c. He defouleth the whole faith of his testimony, by the falsifying of one

part. (a1555, OED)

The gerund construction in (6a) is verbal: it has no overt determiner and has a genu-

ine argument expressed by means of an oblique noun phrase. The constructions in

(6b-c), by contrast, are nominal: both have a notional argument expressed by means

of an of-phrase. The two constructions still differ, however, in that (6b) has no overt

determiner, whereas (6c) is marked by a definite article. These two types will be

referred to here as bare and definite nominal gerunds respectively (De Smet forthc.

a).67

67 The focus on definite, bare and verbal gerunds as defined above implies a certain degree of idealisation of the data, as a number of constructions have been left out of consideration. Most importantly, I do not discuss gerunds introduced by a genitive


In historical data, the three gerund constructions in (6) above are also the ger-

und constructions most abundantly attested and it is for that reason possible to use

them as reliable ‘indicators’ of the major developments in gerundial usage. To

achieve this, I have tracked the development of verbal gerunds, bare nominal ger-

unds and definite nominal gerunds through the course of much of the Middle Eng-

lish period and into the Early Modern period. The primary corpus used for this is

HC, which has been searched exhaustively for all forms ending in -ing, -yng, -inge,

and -ynge. The periods covered are 1250-1350, 1350-1420 and 1420-1500 for Mid-

dle English, and 1500-1570 and 1570-1640 for Early Modern English. For the pe-

riod 1500-1570, the data has been further extended with the supplements to the Hel-

sinki Corpus (Penn1 and Penn2) found in PPCEME.68

Throughout the period examined, the three gerund constructions have been

undergoing change in the form of shifting frequencies of use. Figure 8.1 charts the

use of different gerund constructions over the five sub-periods of the corpus data,

distinguishing uses with preceding preposition from uses without (i.e. oblique and

non-oblique respectively) (see also Appendix A1 at the end of this chapter). To start

from the familiar, as can be seen from Figure 1, verbal gerunds first appeared in the

Middle English period and came to be associated with prepositional environments

(cf. Section 1 above). The oldest period of the corpus data examined, 1250-1350,

contains two unambiguously verbal instances, given here in (7a-b), and one rea-

sonably convincing example, given in (7c).69 Note that the only example in which

the verbal gerund follows a preposition is (7b), which is a direct translation of a

NP or possessive pronoun (or, for that matter, a subject expressed by means of a bare NP). Other gerund constructions omitted from the discussion are those without at least one (notional) argument, the majority of which are in any case ambiguous between nominal and verbal readings. Except in Section 4.3, hybrid gerunds have not been considered either. The latter form an inherently interesting category for the purposes of this paper, but are often too infrequent in the corpus data to allow for significant conclusions. 68 Note that where qualitative aspects of change are involved, I have also drawn on further corpus data, especially from PPCME. 69 Example (7c) is translated from Latin. The Middle English text as well as the Latin original (Jerusalem, quæ ædificatur ut civitas, cujus participatio ejus in idip-sum, i.e. ‘Jerusalem, which is built as a city, of which his participation is in the same’) are somewhat obscure, and the Middle English text does not appear to trans-late the Latin accurately. Still, assuming as is only reasonable that the Middle Eng-lish form takyng part translates the Latin action nominal participatio, takyng part is a nominal in subject position with unambiguously verbal force.

0

20

40

60

80

100

1250-1350 1350-1420 1420-1500 1500-1570 1570-1640

DNG (obl)

BNG (obl)

VG (obl)

DNG (n-obl)

BNG (n-obl)

VG (n-obl)

Figure 8.1. Verbal gerunds (VG), bare nominal gerunds (BNG) and definite nominal gerunds (DNG) in

oblique (obl) and non-oblique (n-obl) environments (frequencies per 100,000 words).


Latin gerund form in the Vulgate Bible (in custodiendo sermones tuos, i.e. ‘in

guarding your words’ ). This indicates that the association with prepositional uses

need not go back to the earliest stages of the verbalisation of the gerund.

(7) a. Ac þer ne is non zuo propre ase þis word / ‘þet art.’ þet zuo

propreliche. zuo ssortliche. Zuo cleuiyndelyche. zuo sotilliche / þe

names nemneþ / ine zuo moche / ase onderstondinge him may

strechche. (1340, HC)

‘But there is no name so fitting as this expression, ‘that is’, which so

fittingly, so concisely, so convincingly, so truly gives the name in as

far as understanding him (i.e. God) may stretch.’

b. In what þynge amendeþ þe ȝenge man his waie in keping þy wordes?

(c1350, HC)

‘In which way does the young man correct his way of life? – In

keeping your words.’

c. Þe heuen þat is edefied as cite, of wham takyng part is in þat ich þing.

(c1350, HC)

‘Heaven, which is built as a city, participation of which is in that same

thing.’

From the earliest sub-period onwards, verbal gerunds become steadily more fre-

quent, and whatever environment first saw the use of verbal gerunds, it is clearly the

prepositional environments that are responsible for the pattern’s rise in frequency in

subsequent periods. Thus, while throughout the data non-prepositional uses are at-

tested with no more than one or two instances per sub-period, the frequency of ver-

bal gerunds following prepositions increases exponentially.

Less familiar are the developments in nominal gerunds. Tajima (1985: 60-73)

gives valuable figures but, not knowing the precise size of his printed corpus, cannot

give relative frequencies, nor does he indicate differences between syntactic envi-

ronments. As shown by Figure 8.1, bare nominal gerunds occur marginally before

1350; then, around the end of the fourteenth century their frequency increases spec-

tacularly, but the change breaks off abruptly and from the fifteenth century onwards

the pattern again recedes (to disappear almost completely later on, as shown by

Fanego 2004a). Significantly, the changes affecting bare nominal gerunds – in par-

ticular the surge in frequency after 1350 – are again specific to those uses where the

gerund follows a preposition, other uses remaining fairly stable throughout. Two

early examples are given in (8), illustrating both the oblique and non-oblique uses.


(8) a. For longing of þe niȝttegale, Þis foweles murie singeþ. (?a1300, HC)

‘For longing for the nightingale, these birds sing happily.’

b. Hwan he was king, þer mouthe men se Þe moste ioie þat mouhte be

Buttinge with sharpe speres, Skirming with taleuaces þat men beres,

Wrastling with laddes, putting of ston, Harping and piping ful god

won, Leyk of mine, of hasard ok, Romanz-reding on þe bok. (c1300,

HC)

‘When he was king, you could there see the greatest merriment

possible: thrusting with sharp spears, fencing with shields carried by

men, wrestling with young men, putting the stone, playing the harp and

the pipe a good deal, games of dice and chance also, and reading

romances from the book.’

Definite gerunds, too, become more frequent in the course of the period examined.

The increase here is less abrupt than in the case of bare nominal gerunds, continues

throughout the Middle and Early Modern periods, and more or less parallels the rise

of verbal gerunds. Note that even though a simultaneous increase can be noted in

non-oblique environments after 1500, the main environment of change is once more

formed by the uses following prepositions. Two early examples are given in (9),

again illustrating use in oblique and non-oblique environments.

(9) a. Ac huanne he heþ longe ymyned / and he heþ / alle his uelþes ykest

out: þanne uint he pays / and reste / and solas / and blisse / zuo þet him

þingþ þet al þe wordle by an helle to þe lokinge of þe ilke clyernesse /

and of þe ilke pays: þet he uint ine his herte. (1340, HC)

‘And when he has long been digging and has cast out all his impurities,

then he finds peace and rest and solace and bliss, so that he believes all

the world to be a hell compared to the beholding of that same

brightness and that same peace that he finds in his heart.’

b. Þe shewynge of þin wordes aliȝteþ me, & ȝeueþ vnderstondyng to litel.

(c1350, HC)

‘The revelation of your words enlightens me and gives understanding

to the young.’

Taken together, the developments in the gerund show convincingly that

prepositional environments were involved in all of the major changes affecting the


gerund in the course of the Late Middle and Early Modern periods. While this is in

itself an interesting conclusion, it is clear that, since the frequencies of all three ger-

und types rose spectacularly in prepositional environments, the original idea that

prepositional environments promoted verbal gerunds (Einenkel 1914; Jack 1988) is

either too specific or as yet insufficiently supported by the data. Indeed, if any gap-

filling occurred, Figure 8.1 would seem to confirm Fanego’s (1996a, 2004a) as-

sumption that both nominal and verbal gerunds would qualify to substitute for the

infinitive (see Section 1). That said, it is also clear that the findings raise more ques-

tions than they answer. There is the issue of what happened to bare nominal gerunds,

which represent the first gerund construction to receive a serious boost in frequency

but began dropping out of use again soon after. A related problem is why no less

than three types of gerund construction undergo such significant changes. Neither of

these issues can be resolved on the basis of the description offered so far, which

glosses over the variety of concrete uses to which the combination of a preposition

and a gerund could in actual fact be put, and further ignores any potential differences

between gerund constructions that might be significant to understanding the changes

they underwent.

3. Functional differentiation

Having highlighted the broad similarity between the three gerund types – that is,

their historical association with prepositional environments – we may now turn to

what distinguishes the three constructions. The focus is on the gerunds in oblique

environments, since these are the constructions that undergo significant changes in

the course of the period examined. Section 3.1 addresses in general terms the dis-

course-functional properties of the three gerund constructions, specifically focusing

on the functioning of gerund constructions as NPs in discourse. Against this back-

ground, Section 3.2 then returns to the use of the gerund constructions with preposi-

tions. It is first shown that different gerund constructions have collocational prefer-

ences in terms of the combinations they form with specific prepositions. These com-

binatorial patterns are then examined more closely, investigating how each preposi-

tional pattern interacts with the discourse-functional properties of the distinct gerund

constructions. Section 3.3, finally, examines the three gerund constructions from a

different perspective, highlighting the differences brought about by the nominal or

verbal syntax of gerund constructions. As the discussion will show, while there is a

degree of overlap in the range of application of different gerund constructions, the

constructions also differ in important ways. A proper understanding of these differ-


ences is crucial to any evaluation of the major changes affecting the gerund in Mid-

dle and Early Modern English.

3.1. Discourse-functional differences

One important difference between gerund constructions relates to their functioning

as NPs in discourse, with the presence or absence of a definite article serving as the

primary distinctive feature. As argued by Heyvaert (2004, 2006) for Present-Day

English, determiners with the gerund preserve their ordinary functioning. It would

be surprising if anything else were the case in older stages of the language. It may

further be assumed for Late Middle and Early Modern English that the function of

the definite article itself does not differ radically from its present-day function (as

described, for instance, by Hawkins 1978 or Langacker 1991: 98). While this as-

sumption clearly awaits further investigation, the historical differences that have so

far been noted in reference grammars pertain to very specific lexico-grammatical

environments and do not suggest any fundamental change in the function of the

definite article (Raumolin-Brunberg 1991: 173-84; Fischer 1992: 219; Rissanen

1999: 191-5; see also the quantitative evidence in De Smet forthc. a). In what fol-

lows, I will therefore examine in more detail how the use of the definite article inter-

acts with the referential properties of different gerund constructions.

Indeed, examples of gerund constructions with other determiners than the or

ZERO suggest that the determiner slot is essentially intact and functions with ger-

unds much as it does with other uncount nouns. For instance, the gerund construc-

tion a teching of anoþer þan it-self with the indefinite article a in (10a) marks an

indefinite, non-specific instance of a type of situation in the same way as a myrour

marks an indefinite, non-specific instance of a type of thing. Along similar lines, the

gerund construction this feyghtyng with that knyght with proximal this in (10b)

marks a specific situation that is construed by the speaker (Sir Lancelot) as immedi-

ately accessible to the addressee (the three cowardly knights) in the situation of dis-

course.

(10) a. And riȝt as þou seest þat ȝif a foule spot be in þi bodily visage, þe iȝe

of þe same visage may not see þat spotte, ne wite wher it is, wiþ-outyn

a myrour or a teching of anoþer þan it-self: riȝt so it is goostly.

(?a1400, HC)


‘And just as you see that if a physical face has a stain, the eyes of that

face cannot see that stain, nor know where it is without a mirror or a

remark by another; just so it is with your spiritual face (i.e. the soul).’

b. Whan sir Launcelot herde this he arose up and loked oute at the

wyndowe, and sygh by the moonelyght three knyghtes com rydyng

aftir that one man, and all three laysshynge on hym at onys with

swerdys; and that one knyght turned on hem knyghtly agayne and

defended hym. ‘Truly,’ seyde sir Launcelot, ‘yondir one knyght shall I

helpe […].’ And therewith he toke his harneys and wente oute at a

wyndowe by a shete downe to the four knyghtes. And than sir

Launcelot seyde on hyght, ‘Turne you, knyghtis, unto me, and leve this

feyghtyng with that knyght!’ (a1470, PPCME)

The working of the definite article is particularly easy to see in definitely marked

gerund constructions without further internal modification, as in the examples in

(11). As a marker of retrievability (cf. Ariel 1990; Langacker 1991; Givón 1992;

Epstein 2001; Gundel, Hedberg & Zacharski 2001), the definite article here signals

that the addressee has mental access to the referent of a NP on the basis of previous

knowledge and/or the information given in the NP itself. The referent may be acces-

sible through explicit mentioning in the previous discourse, as is the case for the

reading in (11b), or through an inferential relation and world knowledge, as is the

case for þe falling in (11a) (pushing a rider of his horse naturally involves the rider’s

falling), or the prickynge in (11c) (needles characteristically prick).

(11) a. Anoþer launce in honde he hent – Aȝein þe prince of Tyre he went. He

smoot hym þorouȝ þe breest þare, And out of sadel ouere croupe hym

bare, And J sigge, for soþe þing, He braak his nek in þe fallyng.

(?a1300, HC)

‘He seized another lance in his hande and went against the prince of

Tyre. He pierced him through the breast there and pushed him out of

the sadel and over the horse’s back, And I sigh, for in truth, he broke

his neck in the fall.’

b. And when he came to the ende, he began it afresh and read it ouer

again. And in the reading he made no maner hast (1555, PPCEME)

c. When all thinges were tombled and cleane out of fassion Whether it

were by fortune, or some other constellacion Sodenlye the neele [i.e.


‘needle’] Hodge found by the prickynge And drew it out of his

bottocke where he felt it stickynge (c1553-63, PPCEME)

Although their behaviour is more intricate and varied than that of the gerund

types in (10) and (11) above, definite nominal, bare nominal and verbal gerunds too

can be seen to fulfil specific functions in discourse. Definite nominal gerunds, like

the definitely marked gerunds in (11), can be used to refer to a specific event that is

accessible to the addressee on the basis of previous discourse, as in (12a-b), or the

situation of discourse, as in (12c).

(12) a. Bowȝes of dyuers trees hij kytten, And to her horses tales knytten. To

Darrie ward alle hij fareþ; Þe bowȝes þe dust heiȝe arereþ. Of þe

draweyng of bowȝes and stykke Þe eyre bicom trouble and þicke, Þat

to Darries folk it ferde, Als on hem com þe myddelerde. (?a1300, HC)

‘They cut off twigs of many trees, and tied them to their horses’ tales.

They all rode against Darius, and the twigs made the dust fly up high.

Because of the drawing forth of the twigs and branches, the air became

so troubled and thick with dust, that Darius’ men got frightened as

though the whole world were advancing against them.’

b. & þerfor, as honeste wolde, sche went to þe cherch þer þe lady herd hir

seruyse, wher þis creatur sey a fayr ymage of owr Lady clepyd a pyte

[i.e. ‘a Pietà’]. And thorw þe beholdyng of þat pete hir mende was al

holy ocupyed in þe Passyon of owr Lord Ihesu Crist & in þe

compassyon of owr Lady, Seynt Mary, be whech sche was compellyd

to cryyn ful lowde & wepyn ful sor (c1450, PPCME)

c. My lorde the cause of the sendyng of this man in so great haste unto

you, is, be cause that and [i.e. ‘both’] the Kyngs majestie and you shall

thynke it convenient to have hym to be brought to the Towre there to

be more streyghtlye examyned and to be put to torture (1538,

PPCEME)

At the same time, definite nominal gerunds can also involve other ways of identifi-

cation than the anaphoric or exophoric ties illustrated in (11) and (12). In particular,

definite nominal gerunds instantiate so-called reference point constructions (Lan-

gacker 2001: 35-6; Willemse 2005: 264-6), which provide mental access to one en-

tity through another entity. In definite nominal gerunds, the entity identified or ac-

cessed is the event designated by the gerund, and the entity leading to identification


or access – the reference point – is the event participant marked by the of-phrase.

The relation between a reference point and the target of identification is one of ‘for-

ward bridging’ (Willemse 2005; 2006). For example, because commissions can be

granted, the availability in discourse of a referential set of commissions, as in (13a),

gives easy mental access to an event of granting. Similarly, because a dinner, as in

(13b), needs to be furnished, the presence of a referent dinner in discourse gives

access to an event of furnishing.

(13) a. AN ACT FOR THE GRAUNTING FORTH OF CO~MYSSIONS FOR

SEWERS (1488-91, HC)

b. The ij day of October was bered [i.e. ‘buried’] the nobull duke of

Norffok at a plasse callyd Fremyngham chyrche; and ther was a goodly

hersse of wax as I have sene in thes days, with a dosen of banerrolles

of ys progene, and xij dosen penselles, xij dosen of kochyons, and with

standard, and iij cotes of armes, and a baner of damaske, and iiij banars

of emages, and mony mornars, and a gret dolle, and after gret dener.

For the furnishing of which dinner were killed forty great oxen and a

hundred sheep, and sixty calves, besides venison, swans, and cranes,

capons, rabbits, pigeons, pikes, and other provisions both flesh and

fish. (1553-9, PPCEME)

While the definite nominal gerunds in (12) and (13) have specific referents

that can be located in space and time, definite nominal gerunds can also be used

generically, referring not to a specific instance of an event but to a kind of event.

Normally, generic NPs are treated as proper names naming a commonly known

kind, and as such they require no definite article in English (Carlson 1977; Davidse

1999; Heyvaert 2006). However, in a ‘generic text’ – i.e. a text whose participants

interact on a generic rather than an episodic level – the ordinary mechanisms of ref-

erential tracking can be imposed on generic referents (Behrens 2005: 317-23). The

example in (14) illustrates this point: the drinking of wine is used generically but can

nevertheless receive definite marking because it is simultaneously an accessible par-

ticipant in a (generic) text about wine and its properties. Note, incidentally, that in a

similar vein the drinking of wine is also tracked anaphorically further down the dis-

course by means of the pronoun it.

(14) Plato in the .ij. booke de legibus, forbiddeth all children wine yt are

vnder .xij. houres old, for that inte~t yt they should not be driuen


therwith into madnesse, he suffreth them that are full growen in age, to

vse it, bicause it is a remedie against ye grieuousenesse of age, and

driueth away sorowes, & swageth the hardnesse of maners, the age of

springoldes or of growing children, is hote and full of muche bloude,

contrarywise the olde age is colde, and wanteth bloud, therefore the

drinking of wine is profitable for olde men, but to them that are in

growing, it is exceeding hurtfull (1568, PPCEME)

This special character of generic texts interacts with the properties of definite nomi-

nal gerunds as reference point constructions. For instance, þe takyng of þe nede in

(15a) marks an event of taking that is generic yet also accessible through its argu-

ment þe nede, which in turn functions as a participant in a generic text about sensu-

ality. Arguably, the reference point construction may even function as a minimal

generic text on its own:70 in (15b), the generic NP the bredyng, chesinge, and

kepyng, of them does not figure in a larger generic text, but definite marking is justi-

fied by the relation of forward bridging between horses and the associated kinds of

action known as breeding, choosing and keeping.71

(15) a. And it [i.e. sensuality] haþ two partyes: one þorow þe whiche it

beholdeþ to þe needfulnes of oure body, anoþer þorou þe whiche it

serueþ to þe lustis of þe bodely wittys. For þis same miȝt is it þat

gruchiþ when þe body lackyþ þe needful þinges vnto it, and þat in þe

70 This is not a wild hypothesis: both generic texts and reference point constructions are linguistic manifestations of ‘scripts’ or ‘idealised cognitive models’, reflecting how the world is structured in the eyes of the language user (cf. Behrens 2005; Wil-lemse 2006) 71 Another special type of generic definite nominal gerund derives from the use of the adjectives true and right to modify the gerund phrase. Here the motivation of the definite article does not lie in the organisation of generic text, but in the fact that true or right highlight the referent’s status as the single appropriate type among a wider set of types. In much the same way as the use of a superlative can serve to singularise an instance among instances and thereby make it automatically accessi-ble (e.g. the fastest swimmer), this singularity of a type among types appears to fa-vour (redundant) accessibility marking on a generic referent. Consider, as an exam-ple, the arte of the righte measuringe of likeiammes in (ii) (as opposed to, e.g., the arte of measuringe of likeiammes). (ii) By this Theoreme may you know the arte of the righte measuringe of

likeiammes [i.e. ‘bodies’], as in my booke of measuring I wil more plainly declare. (1551, PPCEME)


takyng of þe nede stereþ us to take more þan nediþ in fedyng and

forþeryng of oure lustys. (?a1400, HC)

b. In his [i.e. Virgils] Georgikes lorde what pleasaunt varietie there is: the

diuers graynes, herbes, and flowres that be there described, that, reding

therin, hit semeth to a man to be in a delectable gardeine or paradise.

What ploughe man knoweth so moche of husbandry as there is

expressed? who, delitynge in good horsis, shall nat be therto more

enflamed, reding there of the bredyng, chesinge, and kepyng, of them?

In the declaration whereof Virgile leaueth farre behynde hym all

breders, hakneymen, and skosers. (1531, PPCEME)

In contrast to definite nominal gerunds, bare nominal gerunds can be pre-

dicted to behave as uncount nouns without determiner. The expectation, then, is for

bare nominal gerunds to be either generic or indefinite. Generic uses are illustrated

by dredyng of our Lord in (16a), redynge of deuoute thinge in (16b) and wryting of

letteres in (16c).

(16) a. His name is holy & dredeful, þe biginnyng of wisdome is dredyng of

our Lord. (c1350)

b. for ther is no thinge that ouercometh so sone the fende as doth redynge

of deuoute thinge and prayer and meditacyon of Cristys passyon.

(a1450, PPCME)

c. Here regneth Artharxerses, þe vi kyng in Perse, whech is clepid Nothus

– xl ȝere he regned þere; vndir whom Esdras repayred þe lawe þat was

brent be hem of Chalde, whech Esdras broute in new maner of wryting

of letteres, þat were more esy for to write and more esi for to

pronounce, and þerfor was he called a swift writer. (a1464, PPCME)

Purely indefinite readings appear to be more exceptional, yet in some contexts they

provide a plausible interpretation of a given gerund construction. The gerund con-

structions grete hyndryng and anyntysement of the same cite in (17a) and great

shooting of the guns in (17b) each mark a specific but newly introduced situation.

Similarly, in (17c) the formal contrast between the definite nominal gerund the

Lodginge of wayfaring people and the bare nominal gerund harbouringe of lewde

and idle people (though both also allowing a generic reading) is suggestive of a

rhetorically exploited opposition in discourse between a preconceived plan that is


accessible to the addressee and the actual though unintended state of affairs that is

only first brought to the addressee’s attention by the speaker.

(17) a. More over the sayde Mayer and Citeseyns seyn that by

commaundement of the sayde Bysshop and speciall rule of the sayde

Master Harry, assessyng and levy was made amonge the sayde

Bysshoppis tenantis of ij dymes as a cite or a burgh by hym self, […]

and the money so assessed and made levy kepte and yet kepeth to the

use of the said Bishop to grete hyndryng and anyntysement [i.e.

‘destruction’] of the same cite (c1428, HC)

b. The ij day of January the king of Spain’s ambassadors landed at Tower

wharf. During whose landing there was great shooting of the guns.

(1553-9, PPCEME)

c. Whereas the ancient true and principall use of Innes Alehouses and

Victuallinge Houses, was for the Receipte Reliefe and Lodginge of

wayfaring people travellinge from place to place, and for such Supplie

of the wants of such people as are not able by greater Quantities to

make theire p~vision of Victuals, and not meant for entertainment and

harbouringe of lewde and idle people to spende and consume theire

money and theire tyme in lewde and drunken manner (1588-1604, HC)

More common, however – and especially typical of oblique uses – are bare nominal

gerunds that are ambiguous or even indeterminate between a generic and an indefi-

nite reading. Such instances conform to Behrens’s (2005: 293) hypothesis that the

distinction between a generic and a non-generic interpretation may in some envi-

ronments have very low salience. For example, heryng of holy bokys and heryng of

holy sermownys in (18a) can be conceived of as kinds of activities, generically con-

strued means to the achievement of contemplation, but they may also refer to an

indefinite homogeneous and unbounded series of occasions of hearing holy books

and sermons. Similarly, resceiuynge of yistes in (18b) may have to be read as refer-

ring simply to a kind of activity deemed dangerous by the speaker, or to any indefi-

nite non-specific instantiation of that kind of activity.

(18) a. Thus, thorw heryng of holy bokys & thorw heryng of holy sermownys,

sche euyr encresyd in contemplacyon & holy meditacyon. (c1450,

PPCME)


b. Firthermore I wolde thou were war of resceiuynge of yistes [i.e.

‘guests’] by way of hospitalite anempst wommen of deuocyon,

religious or other, for ofte-sithes amonge gode ben medled badde

(a1450, PPCME)

Next to generic and indefinite uses, whose occurrence is predictable from the

nominal nature of the construction, bare nominal gerunds also engage in uses that

are neither generic nor indefinite. Thus, berynge of the cros in (19a) refers to a spe-

cific – and, it follows, non-generic – event that needs no introducing in the dis-

course, as it has been mentioned previously and, in any case, constitutes common

knowledge to a fourteenth-century English audience. Along similar lines, smytyng of

Malcus here in (19b) and delyverynge off hym unto me in (19c) are unlikely to have

generic reference or to introduce a new specific text participant. The fact that no new

text participants are introduced shows that the gerunds in (19) cannot be indefinite.

The question, though, is whether they are definite.

(19) a. And on þeise greces [i.e. ‘stairs’] wente oure lord whan he bare the

cros on his schulder And vnder this grees is a chapell & in þat chapell

syngen prestes yndyenes [i.e. ‘Indian’] þat is to seye prestes of ynde

noght after oure lawe but after here [...] And þere nere is the place

where þat oure lord rested him whan he was wery for berynge of the

cros. (c1400, PPCME)

b. And here þese blynde heretykes wanton wyt as ydiotes, whan þei seyn

þat Petur synnede not in smytyng of Malcus here (c1400, PPCME)

‘And here these blind heretics lack sense like idiots, when they say that

Peter did not sin in cutting off Malcus’ ear’

c. These two off the garde declared unto mr. shreffe thatt they weare

commaunded by the councelle to delyver me unto hym, and he to

sende me unto Newgate, saynge, “Syr, if it please yow we wyll carye

hym thether.” […] “Masters, sayde he, you maye departe; I wyll sende

my offycers with this jentyllmane anone, when they be come in.” “We

wylle se hym caryed, syr, – sayde they, – for ower discharge.” Then

the shreffe sayde sharpely unto them, “Whatt! do you thynke that I

wyll nott do the councelles commaundementt? Yow are discharged by

delyveryng off hym unto me.” (a1561, PPCEME)


In fact, what we see in (19) is a shift from nominal means of grounding to clausal

means of grounding. Prototypical nominal grounding strategies rely on the time-

stability of nominal referents (Langacker 1987, 1991) and the possibility of identify-

ing a referent on the basis of earlier contact with it (because nominal referents are

time-stable, they can recur). Events being non-time stable, clausal grounding, by

contrast, has to rely on tense and modality, situating events in temporal and/or mo-

dal space, and on the time-stable referents that are involved in events as participants.

Examples as in (19), despite having specific reference, are not strictly speaking defi-

nite, in that they do not encode accessibility through recoverability. Referential ac-

cess to the event in the gerund is in this case realised by the control relationship with

the matrix clause (providing one of the crucial event participants as a reference

point) (cf. Heyvaert 2006) and by the inferentially recoverable modal and temporal

relations with the matrix clause. Uses of this type are analogous to non-finite clauses

with specific reference. In (20), for instance, pulling off a 13-7 victory is clearly not

definite, but is nevertheless accessible as a specific event through its modal and

temporal relation to the grounded event in the matrix clause (the Carolina Panthers

managed it) and by the identifiability of its subject participant through subject-

control.

(20) Never before had a defending Super Bowl champion been beaten by an

expansion team, but the Carolina Panthers managed it by pulling off a

13-7 victory – in San Francisco, no less. (CB)

Bare nominal gerunds as in (19) are in a sense functional hybrids, in that they have a

nominal form, but exploit a clausal grounding strategy.

The hallmark of this use is that it is obligatorily controlled (cf. Heyvaert

2006), in contrast to definite gerunds with specific reference but in line with the

canonical strategies of clausal grounding. In each example in (19) the implied agent

of the gerund form is identical to a participant in the matrix clause – the subject he

in (19a), the subject Petur in (19b) or the subject Yow in (19c).72 This restriction

72 The restriction is not absolute; for instance, (ii) presents an evident counterexam-ple, unless the duke of Orleans, who was not present at the time of the incident de-scribed, is nevertheless construed as controller of fyryng. Note though that instances of this type are extremely uncommon. (ii) But were it of rechelesness or of some euyll dysposyd persone, fyre

was put to the vesturis of the disguysers, the which anon was vpon suche a flame that no man there coulde quenche it; wherefore the


does not apply to definite nominal gerunds – witness the examples in (12) above or

(21) below.

(21) Nowe vppon the reporte made by the Lord Chauncelour and the other

Lordes to the kinge of all their whole discourse had with Sir Thomas

Moore, The kinge was so higlye offended with him, that he plainely

told them he was fully determined that thaforesaid parliament bill

should vndoubtedly proceede forth against him. To whom the Lord

Chauncelour and the rest of the Lords saide that they perceaved the

lordes of the vpper hous so precisely bente to heare him, in his owne

defence, make awneswere himself, that if he were not put oute of the

bill, it wold without faile be vtterlye an overthrow of all. But, for all

this, needes wold the kinge haue his owne will therein; or els he said

that at the passinge thereof, he wold be personally present himself.

(1555, PPCEME)

The observed difference in control behaviour has to be related to the function of the

definite article (De Smet forthc. a). Because the definite article signals that the refer-

ent of a NP can be construed on the basis of previous knowledge and/or the informa-

tion given in the NP itself, the interpretation of the NP draws less heavily on the

immediate linguistic context in which it is employed. If an interpreter is capable of

construing a sufficiently accurate mental image of the event referred to simply by

(re)activating information already available to him/her there is no need to recover

additional information from the immediate linguistic environment through a control

relationship. Thus in (21) above the reader homes in on the referent of the passinge

thereof by activating the commonsensical relation between a bill of parliament and

its being passed, and by recalling whatever information is already available on the

bill in question. This is sufficient for the reader to figure out that the implied agent

sayde dysguysers, beynge by reason of the pytche and oylys greuously turmentyd, ranne into pyttes and waters whiche they myght sonest attayne vnto, [...] In the which passetyme he sent into dyuerse placys of Fraunce sundry accusasyons of the duke before slayen, that he ente~dyd to depose the kyng, and to take vpon hym the rule and gouernaunce of the realme, and to haue poysoned the sayd kynge, as by dyuerse tokyns by hym affermyd for perfourmaunce of the same. And also yt the sayde duke of Orleau~ce was cause of fyryng of the dysgusers garmentis before shewyd, to the ende to brynge the kynge in more daunger of sykenesse (1516, PPCEME)


of passinge is the parliament, without searching the gerund’s matrix clause for a

controller, which indeed there is not.

Interestingly, the dependence on control that is witnessed in the referential

identification of the events in bare nominal gerund constructions as in (19) above

can be thought of as a syntacticised version of the way bare nominal gerunds (and

bare abstract nouns) are normally interpreted in discourse. Specifically, examples as

in (22) (and (18) above), which allow both a generic and an indefinite reading,73

tend to imply control on the indefinite reading. For example, kepyng of þi body fro

outrageous colde or hete in (22a) can be interpreted generically as one of many

kinds of doynges, and on this interpretation the gerund is non-controlled (cf. Hey-

vaert 2006); but if attributed an indefinite non-specific interpretation (meaning ‘any

future occasion of keeping your body from outrageous cold or heat’), the main

clause subject þou takes up the function of controller. The same interpretative pat-

tern is illustrated with vndewe makyng, medlyng, or contrefetyng of Romeney in

(22b), while gruching in (22c) shows that the observations extend to bare gerunds

without of-argument.74 One explanation for this interpretative effect is that an in-

definitely marked referent is construed as not previously accessible to the addressee,

so that an interpreter, in assembling a maximally enriched representation of that ref-

erent, cannot but draw on the immediate syntactic relations in which the referential

NP is found – a strategy that may be reinforced by a parallel speaker strategy of in-

troducing a referent in the context of the other referents it is most saliently related to.

73 The analysis of genericity by Declerck (1991b) may suggest that in case of ambi-guity between a generic and a non-generic reading, the generic reading will by Grice’s maxim of quantity get the upper-hand. Note, however, that Declerck’s analysis primarily applies to NPs in subject position, whose interpretation as generic or non-generic can indeed affect the interpretation of the whole sentence as a state-ment with generic or merely episodic bearing (e.g. the snake is dangerous). Fur-thermore, in episodic sentences, non-subject NPs might even be preferentially inter-preted as non-generic by the maxim of relevance (e.g. the prince was involved in financial fraud). It does not seem to be the case then that the indefinite reading of bare nominal gerunds is automatically the secondary option if a generic reading is also available. 74 Note that the association between indefiniteness and implied control is only a ten-dency. The clearest counter-examples are those where an of-phrase does not denote the patient but the agent of the gerund, as in (iii): (iii) Scabbe [i.e. a skin disease] wol brede in þe necke. & on þe dok of þe

taile riȝt as þe mangew [i.e. another skin disease]. [...] & of þe same cause þat þe mangew wol gendre so wol þis. of bi-stonding bi a schabbed hors or bi gnappynge of an hors. (a1450, HC)


(22) a. For in alle þin oþer doynges þou schalt haue discrecion, as in etyng &

in drynkyng, & in slepyng, & in kepyng of þi body fro outrageous

colde or hete, & in longe preiing or redyng, or in comounyng in speche

wiþ þin euen Crysten. In alle þees schalt þou kepe discrecion, þat þei

be nouþer to mochel ne to lityl. (?a1400, HC)

b. For as mych as by þe Mair, Aldermen, and Comunes of þis Citee hit

was late ordeynid & proclamyd, for the good & eese of al þe peple, þat

no coupere, fre ne foreyn, ne none oþer man shold wiþin þis Citee rase

ne gumme, or in eny other wyse medle with vndewe makyng, medlyng,

or contrefetyng of Romeney [i.e. a sweet red wine], vp peyne of Juesse

of þe pillory (1419, HC)

c. þat es godis wille, þat ye foliz yure ordir, and luue til yure prome, and

do with-uten gruching þat yu es cumandid. (a1425, HC)

‘It is God’s will that you follow your vows, and bear love to your

neighbour, and do without complaining what you are ordered to do.’

The categories used to describe the functioning of bare and definite nominal

gerund constructions, can be extended to the description of verbal gerunds. Doing

so, the most relevant property of verbal gerunds is that they broadly resemble bare

nominal gerunds. Like bare nominal gerunds, verbal gerunds can be used generi-

cally. For instance, þis psalme seying (with the object fronted) in (23a) describes the

kind of action that can be used to measure a particular length of time, while making

good common lawes for the hole Realme and obseruing priuate discipline in (23b)

are portrayed as the kinds of activities that can avert disaster.

(23) a. And ley þi gumme in watir to stepe al a nyght. And on the morwen

take þi gumme-water and þi pouuder of gallys [i.e. ‘oak-galls’], and

put hem togeder, and sette hem ouer the fyer, and lete hem sethen þe

space of þis psalme seying, “Miserere mei, Deus.” And than cast þi

powder of coporose [i.e. ‘a metallic sulphate’] þerin, and steret well

togeder, and þan take it don. (1470-1500, PPCME)

b. The remedie of this, doth not stand onelie, in making good common

lawes for the hole Realme, but also, – and perchance cheiflie – in

obseruing priuate discipline euerie man carefullie in his own house

(1563-8, PPCEME)


While purely indefinite examples are hard to come by, verbal gerunds do participate

in the ambiguity patterns also found for bare nominal gerunds. Thus, swech oþer

synnes vsyng (with fronted object) in (24a) can be interpreted generically as one of

various ways of breaking God’s commandments, but can also be read as indefinite

and non-specific, referring to an unspecified set of occasions of sinning that inevita-

bly meant the addressee’s breach of God’s commandments. Similarly, using the

same in (24b) can be taken to generically describe the kind of crime for which

twenty marks is to be forfeited, or it can refer to any as yet unspecified instance of

that crime.

(22) a. Þu brekyst þe comawndmentys of God thorw sweryng, lying,

detraccyon, & bakbytyng, & swech oþer synnes vsyng. (c1450,

PPCME)

b. And that noman undre the degre of an Erle were in his apparell any

Furr of Sables or black Jenett~ upon payne to forfeit the same apparell

And for using the same to forfeit for ev~y tyme so offending xx

mark~. (1514-5, PPCEME)

Finally, verbal gerunds are also found in uses with specific reference, with the

subject of the gerund obligatorily controlled by an element of the matrix clause.

Indeed, this use is decidedly more prominent in verbal gerunds than in bare nominal

gerunds. Examples are given in (25).

(25) a. And al be it so that it was greet peril to myne enemys to do me a

vileynye in takynge vengeance upon me, yet tooken they noon heede

of the peril, but fulfilleden hir wikked wyl and hir corage. (c1390,

PPCME)

b. Plumton is now owner of it, a man of fair land: and lately augmentid

by wedding the doughter and heir generale of the Babthorpes. (1535-

43, PPCEME)

c. Madame, I recommend me unto you, doyng you to undrestonde that I

have receyved your Lettres by your servante concernyng the maryage

of your doughter, by the whiche I do perceyve that the Gentilwoman

beyng accompaned with your said doughter unto your howse, hath

informed you that it was my mynde for hir to certyfye you that the

Controwler of the Pryncesse howsolde dothe bere hys synguler favour


to your said doughter. Trewly she mysusyd hir selff in gevyng you any

suche knowlege on my behalff (1538, PPCEME)

In conclusion, the three gerund constructions each have their own profile in

terms of referential properties and the way they function in discourse. The major

difference between gerund types stems from the presence or absence of a definite

article, which sets apart definite nominal gerunds on the one hand from bare nominal

and verbal gerunds on the other. The difference primarily manifests itself in definite

gerunds with specific reference, which freely occur without an additional control

relationship between the gerund and some participant in the matrix clause. In such

uses, definite nominal gerunds draw on accessible knowledge on the part of the

speaker for the identification of the event they refer to. Bare nominal and verbal

gerunds have seemingly similar uses, but identification of their referent typically

depends on clausal means of grounding, including a control relationship to the ma-

trix clause.

These differences apart, there is also a degree of overlap between the different

gerund constructions, so that in a number of environments the three types could be

used more or less interchangeably. This is particularly clear for bare nominal and

verbal gerunds, whose usage potential appears nearly identical, even if in actual us-

age they diverge somewhat in the functions they preferentially fulfil. Further, all

three gerund types can be used to express generic reference – albeit that in the case

of definite nominal gerunds, this referential possibility is typically dependent on the

generic character of the surrounding discourse. The corpus data testify to this func-

tional overlap by showing occasional instances of variation between the gerund con-

structions used in parallel environments. Thus, (26a-b) illustrate variation between

bare nominal and verbal gerunds; (26c) between bare nominal and definite nominal

gerunds. In (26a), both gerunds belong to the controlled type; in (26b-c) the different

gerunds are all generic.

(26) a. Afterward, in getynge of youre richesses and in usynge hem, ye shul

alwey have thre thynges in youre herte (that is to seyn, oure Lord God,

conscience, and good name). (c1390, PPCME)

b. Now been ther generale signes of gentillesse, as eschewynge of vice

and ribaudye and servage of synne, in word, in werk, and contenaunce,

and usynge vertu, curteisye, and clennesse, and to be liberal – that is to

seyn, large by mesure (c1390, PPCME)


c. But to knowyng of God þat is soþnesse, ne maiȝt þou not comen but

þorw knowynge of þi-self [...]. To þe knowyng of þy-self maiȝt þou

comen wiþ ofte þenkynge; to þe knowyng of God: wiþ clen

contemplacion. (c1390, PPCME)

3.2. Distributional differences

A second dimension on which the three gerund types must be compared is their dis-

tribution over the set of available prepositions. In this spirit, Figures 8.2, 8.3 and 8.4

show the usage frequencies from 1350 to 1640 of the three gerund types in oblique

environments, but this time differentiate between the six most common prepositions

combining with gerunds: in, for, of, by, to, and without (see also Appendix A1). This

further breaking down of the data is a reminder that the major shifts in frequency of

the three gerund types, already discussed in Section 2, are in actual fact produced by

the emergence and/or disappearance of a variety of more specific patterns, each with

their own function. To come to a better understanding of the general changes affect-

ing gerundial usage, it is therefore necessary to investigate the histories of the more

specific combinations in which the different gerund constructions occur.

In interpreting Figures 8.2, 8.3 and 8.4, it is important to be aware of the fact

that figures are sometimes distorted by the idiosyncrasies of a single text in the cor-

pus, often through repetitious use of a single formulaic expression. This is most

clearly the case for the frequency of definite nominal gerunds with of in the sub-

period 1420-1500, where a single text (The cyrurgie of Guy de Chauliac) provides

more than half of all instances for the period (typically in headings of the type of þe

wasshinge of butter). To a lesser extent, the same text may also somewhat exagger-

ate the figures for verbal gerunds with in for the same period. Another possible case

of misrepresentation involves the frequency of verbal and especially bare nominal

gerunds with for in the sub-period 1500-1570. Here two texts provide a dispropor-

tionately high number of instances, again due to formulaic or semi-formulaic uses

(The diary of Henry Machyn and the Act agaynst wearing of costly apparrell, which

has been included in the PPCEME in two slightly different versions). With these

qualifications in mind, it is possible to observe a number of tendencies in the data.

To begin with, it is evident from Figures 8.2 through 8.4 that the three gerund

types have somewhat different collocational preferences and, from a diachronic

point of view, have each followed a distinct course of development (once again


0

10

20

30

40

50

60

70

80

90

1350-1420 1420-1500 1500-1570 1570-1640

other + VB

to + VB

without + VB

by + VB

of + VB

for + VB

in + VB

Figure 8.2. Prepositions with verbal gerunds (VG) (frequencies per 100,000 words).

0

10

20

30

40

50

60

70

80

90

1350-1420 1420-1500 1500-1570 1570-1640

other + BNG

to + BNG

without + BNG

by + BNG

of + BNG

for + BNG

in + BNG

Figure 8.3. Prepositions with bare nominal gerunds (BNG) (frequencies per 100,000

words).


0

10

20

30

40

50

60

70

80

90

1350-1420 1420-1500 1500-1570 1570-1640

other + DNG

to + DNG

without + DNG

by + DNG

of + DNG

for + DNG

in + DNG

Figure 8.4. Prepositions with definite nominal gerunds (DNG) (frequencies per

100,000 words).

change is diffusional!). Thus, verbal gerunds (Figure 8.2) strongly collocate with the

preposition in, so much so that the initial increase of verbal gerunds is in fact largely

due to the extension of this single pattern. Not until the Early Modern period do ver-

bal gerunds spread to use with other prepositions and does the bias towards use with

in disappear. The new prepositions verbal gerunds spread to are first of, then for,

then by and without. With to, verbal gerunds remain almost entirely absent.

Bare nominal gerunds (Figure 8.3) show the same strong association with in

as verbal gerunds, but other combinations too are relatively current from the start,

most notably the combinations with for and of. The decline of bare nominal gerunds

sets in after 1350 with the preposition in. With other prepositions – of, by, for and to

– bare nominal gerunds are at this point still on the increase or at least stable. In

these last environments, decay sets in first with of and to (after 1500), then also with

by, and for (after 1570). The combination of a bare nominal gerund with without

remains marginal and is, if anything, not on the increase after 1500.

Finally, the rise of definite nominal gerunds (Figure 8.4) is primarily due to

use in combination with for and a number of prepositions outside the group of the

most frequent six (most prominently, these include after, before, and at – see Ap-

pendix A2). A slight but steady increase with the preposition in further contributes

to the rise of definite nominal gerunds. It is also worth noting that from 1420 on-


ward, definite nominal gerunds are the gerund type most strongly associated with

the preposition to. Usage with by is marginal but relatively stable throughout, and

the same goes for of. Apart from a single instance in the sub-period 1500-1570,

definite nominal gerunds do not combine at all with without.

While it is difficult to interpret these findings at face-value, they already

throw some light on the interaction between the three gerund types, as they are sug-

gestive of areas of competition and of functional differentiation. In what follows, the

interaction between gerund types is closely examined, focussing first on the rivalry

between bare nominal and verbal gerunds (Section 3.2.1) and then turning to the

status of definite nominal gerunds as a construction relatively isolated from the on-

going competition between the other two types (Section 3.2.2).

3.2.1. Bare nominal and verbal gerunds: the rivals

The data in Figures 8.2 and 8.3 are suggestive of competition between verbal and

bare nominal gerunds, with the former ousting the latter. Most strikingly, the fact

that the decline of bare nominal gerunds begins in the patterns with in puts obvious

suspicion on the verbal gerund as the responsible competitor, the latter having seen

its initial rise mainly in uses with in. Less spectacular but equally telling is the com-

petition in uses with of and by, where verbal gerunds may again carry responsibility

for the decline of bare nominal gerunds, first with of after 1500, then with by after

1570. These findings are not entirely surprising, since a replacement of bare nominal

by verbal gerunds is well in line with the overall similarity between bare nominal

and verbal gerunds in their discourse-functional properties (see Section 3.1 above).

However, the replacement hypothesis receives its strongest support from the fact

that nearly all of the various uses of verbal gerunds – with in but also with by, of, for

and without – have equivalents with bare nominal gerunds. This implies that the

notion of a replacement of bare nominal gerunds by verbal gerunds can be extended

to encompass the rise of verbal gerunds in general. In order to demonstrate this, it is

essential to look more closely into each of the recurrent combinations of a preposi-

tion and a bare nominal or verbal gerund, so as to verify to what extent they fulfil

similar functions.

Combinations of bare nominal or verbal gerunds with the preposition in can

be seen to function similarly, despite certain usage preferences. Use of the pattern

clusters between two extremes. On the one hand, gerunds with in can be used as

adverbial adjuncts expressing a relation of manner, means or accompanying circum-


stance.75 Used in this function, the gerund is typically controlled, and it may refer to

a specific event. The meaning of the preposition in here is relatively far removed

from its original spatial sense, and replacement of the gerund by an ordinary NP is

difficult (cf. Chapter 7). The examples in (27) show that the use is found both with

bare nominal (27a, c-d) and verbal gerunds (27b-c).

(27) a. Neuerþeles, ȝif I schal soþlier sey, a soule is more bleendid in felyng

of it for habundaunce of goostly liȝt, þen for any derknes or wantyng

of bodely liȝtte. (?a1400, HC)

b. & the heremyte asked him what he was. And the monstre answerde

him & seyde he was a dedly creature such as god hadde formed &

duelled in þo desertes in purchacynge his sustynance (c1400, PPCME)

c. whiche glory and reno~mee of name he [i.e. Joshua] obteyned in

lyftyng his hondes & scute [i.e. ‘bolts’] in shotynge of sharpe & fyry

arowes ayenst ye cyte of hay [i.e. Ai] (?1495, HC)

d. All your counsell must goe to this ende in chosing of wine fit for old

men, that it may be very thin or subtil, in color redish yellow, or

yellow, or pale yellow , which is of a middle color betweene bright

yellow and white. (1568, PPCEME)

On the other hand, gerunds with in also occur as obligatory adjuncts or prepositional

complements with a number of predicates. In this function, the gerund is unambigu-

ously generic, and can usually be replaced by an ordinary NP without affecting the

meaning of in or upsetting the syntax of the clause as a whole. Examples are given

in (28), with (28a-b) illustrating the pattern with a bare nominal gerund, and (28c)

with a verbal gerund.

(28) a. But in translating of wordis equiuok, that is, that hath manie

significacions vndur oo lettre, mai liȝtli be pereil (a1397, HC)

75 The preposition in occasionally appears to have a weakly purposive meaning (or, arguably, allows this meaning to be pragmatically inferred), as is indicated by the following example: (iv) The mooste noble and valiant princis of Grece often tymes, to recreate

their spirites, and in augmenting their courage, enbraced instrumentes musicall […]. (1531, HC).


b. For þe souerayne and þe Escencyalle Ioy es in þe lufe of Godd by

hym-selfe and for hym-selfe, and þe secundarye es in comonynge [i.e.

‘meeting, conferring with’] and byhaldynge of Aungells and gastely

creaturs. (c1440, HC)

c. Onely they that be called beneficiall, and do use the vertue of

beneficence, whiche consisteth in counsaylinge and helpinge other

with any assistence in tyme of nede, shall alway finde coadiutours and

supportours of their gentyll courage. (1531, PPCEME)

Between these two extremes, a substantial group of instances falls into an intermedi-

ary category. These can mostly be described as adjuncts of respect, specifying in

which respect the situation of the matrix clause holds true or takes place (cp. Quirk

et al. 1985: 483). Typically they are ambiguous between a controlled and a generic

reading, and may alternatively lean closer to the group of adverbial clauses or to the

group of prepositional complements. The pattern is illustrated in (29a-b) for bare

nominal gerunds, and in (29c-d) for verbal gerunds.

(29) a. Here take hede how worþy þe court of Rome hylde hym [...]. And also

how worþy we schal holde hym by manere of his leuynge and of his

techynge. He myȝte not be vicious þat spende so his witte and þouȝt in

expouninge of holy writte. (a1387, PPCME)

b. it be ordeyned by the seid auctorite that yf defaulte be in the

Constables of eny Cite Borough or Hundred in makyng of due serche

after open p~clamacion made in this behalve, [...] that they and ev~y of

theym being in suche defaulte forfeite to the King xxtis. (1488-91, HC)

c. Of this appereth morally. that one vertuous prelate dooth moche good

in crystis chyrche / as well for his owne vertue & zelose obedyence

vnto the lawe of god: as in ledynge other men to do the same by his

holy doctryne and vertuous example. (?1495, PPCME)

d. P~vyded alwey that this Acte be not p~iudiciall nor hurtefull to any

sp~uall and temporall man in wering any ornament~ of the Churche in

executing dyvyne s~vice nor to any m~chaunt straung~s. (1514-5,

PPCEME)

From the examples in (27)-(29) it is clear that considerable functional overlap ex-

isted between bare nominal and verbal gerunds in combination with the preposition

in. This supports the idea that verbal gerunds ousted bare nominal gerunds in this


environment. It should be pointed out, however, that replacement did not proceed

evenly in the different uses found with in. As the above examples indicate, verbal

gerunds with in first occurred in adverbials of means, manner and accompanying

circumstance, only spreading to adjuncts of respect toward the end of the Middle

English period, and to clear prepositional complement uses after 1500.

The combinations of by, without and of with bare nominal or verbal gerunds

present a similar albeit less complicated picture. The preposition by begins to com-

bine with bare nominal gerunds after 1350 to express an adverbial relation of means.

The bare nominal gerund at this point typically allows an indefinite reading with an

implication of control, but also a generic reading, as shown in (30a-b).76 Verbal ger-

unds begin to appear with by around 1500 and then quickly become more frequent.

Like bare nominal gerunds, the verbal gerunds with by allow a generic reading

alongside a controlled one, as in (30c), but especially after 1570, a controlled spe-

cific reading can be the only interpretative option left, as in (30d). In any case, see-

ing that in combination with by bare nominal gerunds had no functions that verbal

gerunds could not fulfil, replacement of one by the other is again plausible.

(30) a. & at þat selue tyme þe foure wardeyns schul chese oþere foure newe

for þe ȝeer comynge, be settyng of foure gerlaundes [i.e. ‘wreaths’]

(1389, HC)

b. By this bodely wirkynge that I spake of, may þou vndirstonde all

maner of goode werke that thi soule doth [...] to thine even cristen, by

fulfillynge of the dedis of mercy bodili or gostely, or vnto God, by

suffrynge of all maner bodely mischeves for the loue of rightwisnes.

(c1440, HC)

c. He hath then a busie worke I say, to bringe his flocke to a ryght fayth

and then to confirme them in the same fayeth, Nowe castynge them

downe with the lawe, and with threateninges of God for synne. [...]

Nowe weedinge them, by, tellinge them their faultes, and makynge

them forsake synne. (1549, PPCEME)

76 Similar behaviour is found in bare nominal gerunds with through, which express a semantically similar relation of means or causation. An example is given in (vi). The pattern, never current in the first place, disappeared after the Middle English period. (vi) Thys shuldest thou euere considre and beholde in al thy werkes,

thorugh weilynge [i.e. ‘penitential bewailing’] of thy wrecchednesse, that thou art but a straunger and a pilgrime in this wrecched worlde; (a1450, PPCME)


d. Sirra, quoth the Seruingman, you must come to my maister, you haue

broken his Trunke all to peeces, by letting it fall. (1597, HC)

Used with without, bare nominal and verbal gerunds again show little or no

differences. Both are typically used as controlled adverbial adjuncts expressing a

relation of non-co-occurrence between the state of affairs in the main clause and the

situation expressed by the gerund. Frequencies being low, the data do not directly

reveal replacement of bare nominal by verbal gerunds (indeed, the oldest attested

example with a verbal gerund precedes the earliest instance with a bare nominal

gerund), but given the high currency in Middle English of without with gerunds

having no modification, as in (31a), or having nominal modification, as in (31b), a

nominal origin of the pattern is nevertheless plausible. The semantics of the pattern

are constant, as the gerund constructions used are typically ambiguous or even

indeterminate between an indefinite reading, with an implication of control, and a

generic reading. This is illustrated for a bare nominal gerund with without in (31c),

and for a verbal gerund in (31d). Note that the same semantic characterisation can be

applied to (31a), and even to (31b), where the gerund, though certainly not generic,

is clearly indefinite and similarly suggestive of a control relationship between its

implied agent and a participant of the matrix clause. Once more, the functional

similarity between bare nominal and verbal gerunds with without is evident.77

(31) a. Yef sho haue whreþid ani of hir sistirs, o what maner sam it be,

hastelike with-vten dwellyng sal sho falle to her fete and aske hir for-

gifnes (a1425, PPCME)

b. After þe deth of Hesider, regnede xxxiij Kynges, eche after oþere, in

pees; and wiþout eny longe tariyng y shal tel ham alle, and how longe

euery of ham regnede, as þe story telleþ. (c1400, PPCME)

c. to spened itt [i.e. worldly good] with-oute loue or vaynlikynge of itt, as

reson askith, in worship of God, and helpe of thyne evyn cristyn.

(c1440, HC)

77 Exceptionally, a bare nominal gerund introduced by without is not controlled by a participant in the matrix clause. An example is given in (vii). This use appears to have disappeared along with the bare nominal gerund. (vii) As ȝif a lond wolde bere good corn wiþowte tylyng and donghyng

þerof, it were but ydel to traueyle þerfore, whonne it encresuþ not þe fruyt. (c1400, HC)


d. þe Erl of Cornwaile & Igerne his wif setten alþer next þe kyng. and þe

kyng saw þe fairenesse of þat lady þat she hade, and was rauisshede

for here beaute; and after, he made towarde here nyce semblant in

lokyng and leiȝhyng. so at þe laste þe Erl perseuede þe priue lokyng

and Laughing, and þe loue bituene ham, and arose vp fram þe table al

in wraþ, and tok his wif, and callede to him his knyȝtes, and went þens

al in wraþ, wiþouten takyng Leue of te Kyng. (c1400, PPCME)

Combinations with of, too, show functional similarity across the variation

between bare nominal and verbal gerunds. As illustrated in (32), bare nominal ger-

unds with of can be used as postmodifier to a noun (as in (32a)), complement to a

complex preposition (as in (32c, e)), or prepositional complement to a verb or verbal

idiom (as in (32b, d)). Referentially, bare nominal gerunds with of are predomi-

nantly generic (see (32a-b)), but occasionally they allow an indefinite reading, with

or without an implication of control by an element of the matrix clause (see (32c-d)

and (32e) respectively).

(32) a. [He] cursed dredfulliche al þat doþ aȝenst þat statute of worschippynge

of holy ymages. (a1387, PPCME)

b. Therfor hise disciplis seiden, Lord, if he slepith, he schal be saaf. But

Jhesus hadde seid [i.e. ‘spoken’] of his deth; but thei gessiden, that he

seide of slepyng of sleep. (c1395, PPCME)

c. No man take þe valew of IIId. but if he pay, up peyn of lesing of a

fynger, ne þe valew of VI d., up peyn of lesing of his hand, ne þe

valew of XII d., up peyn of his hed. (a1464, PPCME)

d. Thei were also accused of clipping of mony (a1464, PPCME)

e. And on þe same maner, wher anoþer man wolde bid þee gader þi

miȝtes & þi wittes holiche wiþ-inne þi-self, & worschip God þere – þof

al he sey ful wel & ful trewly, ȝe! [...] – ȝit for feerde of disseite &

bodely conceyuyng [i.e. ‘interpreting’] of his wordes, me list not byd

þee do so. (?a1400, HC)

As they emerge, verbal gerunds with of map onto the uses of bare nominal gerunds,

as illustrated in (33). In Middle English the first few verbal gerunds attested with of

are generic and function as postmodifier to a noun or complement to a complex

preposition. Later they also appear in the other environments and with non-generic

reference. As the examples clearly show, the lexico-grammatical environments in


which verbal gerunds with of appear are often identical to the environments in which

bare nominal gerunds were used earlier.

(33) a. þat no maner persone, fre ne forein, be so hardy betuene þis and

mighelmasse þat next cometh to make, medle, or consent in any wyse,

priue or appert, with any maner of brocage with-inne þis Cite, vp

peyne of inprisonement of ther bodyes tuelfmoneth and a day, and

makyng fyn to þe Chaumber after discrecion of the Mair and Aldermen

(1418-9?, HC)

b. as sche knelyd on hyr knes þe tyme of tweyn Messys heryng, owyr

Lord Ihesu Crist seyd on-to hir, “Þu comyst not hedyr, dowtyr, for no

nede but for meryte & for mede [...].” (c1450, PPCME)

c. Speake not of winnyng me: for it shall neuer be so. (1552-3, PPCEME)

d. Thus lay the poore Draper a long time in prison, in which space his

Wife which before for dayntinesse would foule her fingers, nor turne

her head aside, for feare of hurting the set of her neckenger, was glad

to goe about and wash buckes at the Thames side, and to bee a chare-

Woman in rich mens houses (1619, HC)

It should be pointed out that verbal gerunds did not replicate the entire functional

range of bare nominal gerunds following of. Specifically, in one use bare nominal

gerunds with of are strikingly at odds with verbal gerunds or even with bare nominal

gerunds following other prepositions. As shown in (34), some bare nominal gerunds

with of clearly cannot be generic nor indefinite, but neither can they receive a con-

trolled interpretation. Instances of this kind could be interpreted as double reference

point constructions where the definite marking of the intermediate reference point

(the gerund form) can be omitted because its role as text participant is less promi-

nent than that of the primary reference point and the eventual target of identification

(but see example (iii) in footnote 74 above, where this explanation cannot hold). In

this use bare nominal gerunds have presumably been replaced by definite nominal

gerunds rather than by verbal gerunds.

(34) a. EXAMINED vpon þe first article he seide at þe tyme of makyng of þe

certificat þe which was sent vnto þe kyng his felawe and he were in

difference and discorde (1424-6, HC)

b. This Robert began the priorie of blake chanons at Oseney by Oxford

emong the isles that Isis ryver ther makith. Sum write that this was the


occasion of making of it. [An account follows of what moved Robert to

build a monastery at the spot in question.]

With for, finally, verbal gerunds once more share their various uses with bare

nominal gerunds. Both bare nominal and verbal gerunds can be used with for to

form adjuncts of purpose, as in (35). Referentially, bare nominal and verbal gerunds

with purposive for often allow a generic reading next to a controlled reading (though

in (35c) the generic reading has become unlikely). Akin to these purpose adjuncts

are noun complements as in (36), which carry the same purposive flavour, typically

with generic reference.

(35) a. Preyer in it-self propirly is not elles bot a deuoute entent directe vnto

God, for getyng of goodes & remowyng of yuelles. (?a1400, HC)

b. when a pricke standeth in the middell of a circle – as no circle can be

made by co~passe without it – then is it called a centre. And thereof

doe masons, and other worke menne call that patron [i.e. ‘model’], a

centre, whereby they drawe the lines, for iust hewyng of stones for

arches, vaultes, and chimneies (1551, PPCEME)

c. and certain of them were for this time thought meet to passe, and to be

redd; other, for avoyding tediousnes, to be omitted (1550-2, PPCEME)

(36) a. The godlie counsels of Salomon and Iesus the sonne of Sirach, for

sharpe kepinge in, and bridleinge of youth, are ment rather, for

fatherlie correction, then masterlie beating, rather for maners, than for

learninge: for other places, than for scholes. (1563-8, PPCEME)

b. Sir Jhon Williams, who was committed to the Flete for disobeying a

commaundment gevin to him for not paying any pensions without

making my counsel prevy, upon his submission was deliverid out of

prison. (1550-2, PPCEME)

Both bare nominal and verbal gerunds can also be used with for to form adjuncts of

reason, specifying the causal motivation of the situation or event in the main clause,

as shown in (37). The causal use is somewhat later to develop than the purposive

use, appearing only marginally in Middle English, and at that point, only with bare

nominal gerunds, but it is certainly current in the Early Modern period, when the

pattern is found both with bare nominal and verbal gerunds. Mostly, gerunds in this

use allow both a generic and an indefinite controlled reading (see (37a, c)), but a

controlled reading without indefinite reference is possible as well (see (37b, d)).


(37) a. Þus fer inwardes comyn many; bot for greetnes of pyne þat þei fele and

for lackyng of counforte þei go bak in beholdyng of bodely þinges,

sekyng fleschly counfortes wiþ-outen, for lackyng of goostly þat þei

haue not ȝit deseruyd, as þei schuld ȝif þei had abeden. (?a1400, HC)

b. Than suffyrd sche schamys & repreuys for weryng of hir white clothys

& for [i.e. ‘because’] sche cryed so lowde whan owr Lord ȝaf hyr

mende of hys Passyon. (c1450, PPCME)

c. The xiij day of January ther was a man drane from the Towre thrugh

London a-pone a sled unto Tyborne, and ther hangyd, dran, and

quartered, for conterffeytyng the quen’s senett [i.e. ‘signet’]. (1553-9,

PPCEME)

d. Thus when they are cursed of God, they shall curse one another, curse

the Lord for condemning them, curse their sinnes for accusing them,

curse their parents for begetting them, and curse themselues, because

they cannot helpe themselues. (1591, HC)

Finally, there are some more marginal uses with for. Among others, these include

adjuncts expressing an apprehensive relation, where for can be paraphrased by ‘lest’

or ‘for fear of’, and topic-shifting disjuncts, where for can be paraphrased by ‘as to’

(cf. Chapter 6). These too, however, appear to combine with bare nominal and ver-

bal gerunds, as shown in (38) and (39) respectively.

(38) a. For worchipyng of þe crosse he seide þat body þat hing on þe crosse

schuld be worchipid, and noþing but he. (a1464, PPCME)

b. So ferforþ, þat neiþir of hem wole aȝeinseie oþere, for displesynge of

eþer oþer in ony þing þat neþer of hem wole do, þouȝ it were fully

deedli synne. (a1450, HC)

‘To such an extent that neither of them is willing to contradict the

other, for fear of displeasing each other with respect to anything that

either one wants to do, even if it were plain deadly sin.’

(39) a. Nay, as for charming me, come hither if thou dare. (1552-3, PPCEME)

b. I dare make none other comparison betwene them for offendinge the

frendes of them both (1531, PPCEME)

The examples in (35)-(39) above all point to a high degree of functional similarity

between bare nominal and verbal gerunds. Replacement of bare nominal gerunds by


verbal gerunds is not unequivocal, however. Figure 8.5 shows the frequency of the

two major patterns with for – purpose and reason adjuncts – for bare nominal and

verbal gerunds over the period 1350-1640.78 Judging by their shifting frequencies,

verbal gerunds are very likely to have replaced bare nominal gerunds with for as

reason adjuncts, since the increase of one goes hand in hand with the decrease of the

other. In the function of purpose adjuncts, however, the rise of verbal gerunds with

for seems to have halted after 1570 and bare nominal gerunds remain relatively sta-

ble throughout the period examined. The hesitancy in the rise of verbal gerunds with

purposive for may be due to competition with definite nominal gerunds in this envi-

ronment or with the to-infinitive (see Section 3.2.2 below).

As the foregoing discussion has shown, the great majority of uses of the ver-

bal gerund closely match uses of bare nominal gerunds. Thus, diachronically, the

decrease of bare nominal gerunds in a number of environments matches an increase

0

1

2

3

4

5

6

7

1350-1420 1420-1500 1500-1570 1570-1640

purp. for+ BNGpurp. for+ VGreas. for+ BNGreas. for+ VG

Figure 8.5. Bare nominal (BNG) and verbal gerunds (VG) with for as purpose and

reason adjuncts (frequencies per 100,000 words).

in the use of functionally equivalent verbal gerunds. This is most obvious for the

different uses associated with the preposition in, but also applies to uses with by and

of. Uses with without do not show an equally straightforward pattern of replacement,

but it is nevertheless clear that without increasingly combined with verbal gerunds,

while the functionally similar combination with a bare nominal gerund dwindled and

78 Figure 8.5 is based on the HC and the PPCEME with the exclusion of the two texts showing disproportionate use of reason adjuncts with for (see the introduction to Section 3.2 above).


disappeared. Combinations with for also show functional similarity across bare

nominal and verbal gerunds, and even though purpose adjuncts with for continue to

favour bare nominal gerunds into the first half of the seventeenth century, reason

adjuncts with for once more show a clear pattern of replacement.

In conclusion, the data again do not straightforwardly support the claim that

verbal gerunds were promoted by use in prepositional environments. Firstly, Middle

English usage of the verbal gerund is marked by the emergence of a single preposi-

tional pattern – the combination with in – with any extensions to other environments

occurring only sporadically. The claim that verbal gerunds preferred prepositional

environments, then, is certainly an overgeneralisation, at least as far as the initial

development of the pattern is concerned. Secondly, the data indicate that verbal ger-

unds did not really fill any functional gap – or, if they did, it was not exclusively due

to their ability to combine with prepositions. If anything, verbal gerunds were pri-

marily used to replace other gerund constructions – bare nominal gerunds, specifi-

cally – rather than to fill in for a construction that did not previously exist.

3.2.2. Definite nominal gerunds: watching from the side?

As the previous section focused on the interaction between bare nominal and verbal

gerunds, an important question has been ignored: what has been the part played by

definite nominal gerunds? Given the higher degree of functional differentiation be-

tween definite nominal gerunds on the one hand and bare nominal and verbal ger-

unds on the other (see Section 3.1 above), we might expect the use of definite nomi-

nal gerunds to be more indifferent to developments in the other two types. Indeed, it

appears that definite nominal gerunds have developed a number of uses of their own

that are dependent on the special semantics of the definite article and, as a result,

more or less unique when set against the use of bare nominal and verbal gerunds.

This suggests that the relation between definite nominal gerunds on the one hand

and bare nominal and verbal gerunds on the other is for the most part one of func-

tional differentiation rather than competition.

The use that is most evidently specific to definite nominal gerunds is that of

temporal adverbials situating an event or situation in the matrix clause with respect

to a temporal landmark. As pointed out above, more than the other two gerund

types, definite nominal gerunds tend to co-occur with prepositions other than the

most frequent six. Most prominently, these include prepositions such as after, at,

before, since or upon, each of which express a temporal relation between two events.

Accordingly, definite nominal gerunds combine with these prepositions to express a


temporal relation between the state of affairs in the main clause and some specific

event that then serves as a temporal reference point vis-à-vis the state of affairs in

the main clause. Definite nominal gerunds lend themselves well to this use due to

the nature of the definite article, which marks an event as specific and accessible to

the reader – and therefore ideally suited to serve as a temporal landmark in dis-

course. Consider the examples given in (40).

(40) a. Also telle hem how I am oute of preson, and what adventure befelle me

at the getynge of this swerde. (a1470, HC)

b. And that other book is named polycronycon / in whiche book ben

comprised briefly [...] the historical Actes & wonderful dedes syth the

fyrst makyng of heuen & erth vnto the begynnyng of the regne of kyng

edward the fourth (1477-84, HC)

c. So fell it oute, within a moneth or thereaboutes after the makinge of the

statute for the oathe of the supremacye and matrimonye, that all the

preistes of London and Westminster, [...] were sente for to appeare att

Lambethe before the Byshoppe of Canterbury (1555, PPCEME)

d. if it shall appeare upon the measuring therof, unto the Maire of the

saide Towne of Bristoll and to the King~ Customers [...], that there is

more Corne or graine loded to be transported into the parties beyonde

the Sea [...], than it shalbe lefull unto the saide maire and customers

there [...] to cause the owner or conveyer therof unto the saide towne

of Bristoll, to make sale of suche corne and graine as shalbe brought

over and above the content~ of the saide cocket or lycence (1542-3,

PPCEME)

By contrast, being typically generic or indefinite, bare nominal gerunds – and verbal

gerunds in their wake – do not favour use as temporal landmarks, which explains

why bare nominal and verbal gerunds rarely co-occur with temporal prepositions in

Middle and Early Modern English. Notice that, as repeatedly pointed out above,

bare nominal and especially verbal gerunds, can occasionally refer to a specific

event without being necessarily indefinite, provided that the implied agent of the

gerund is controlled by an element in the matrix clause. In such uses, the two gerund

types do occasionally combine with members of the set of temporal prepositions, as

shown in (41). As the use of controlled gerunds of this type became more current,

they may have turned into a competitor for definite nominal gerunds. This, however,

happened after the period under investigation (cf. Fanego 1998).


(41) a. Upon viewing of wich thinges there was devised tow fortes to be made

upon the entry of the haven (1550-2, PPCEME)

b. then, soone after, I tooke my Cotch and went to Linton, wher, I aftor

salutinge my mother, praied, and so went to supper (1599-1601, HC)

A certain degree of functional differentiation is also apparent in combinations

of definite nominal gerunds with the prepositions of, by and in. It is true that by and

large these prepositions express the same semantic relations when used with definite

nominal gerunds as when used with bare nominal or verbal gerunds. Indeed, occa-

sional examples show definite gerunds that closely correspond to uses with a bare

nominal or verbal gerund. For example, in (42) the definite nominal gerund the

turnynge of thi will is in fact coordinated with the functionally equivalent bare

nominal gerund with-drawynge of thyne herte fro luste, both referring to a specific

event whose implied agent is controlled by the subject of the matrix clause.

(42) Grace and the goodenes of oure lorde Ihesu Criste that he hath shewed

to the, – in with-drawynge of thyne herte fro luste and from likynges of

worldely vanite, and vse of flesshly synnes and in the turnynge of thi

will enterely to his seruyce and his plesaunce, – bryngith into my herte

much mater to loue hym in his mercy. (c1440, HC)

However, in the majority of cases, the choice of a definite nominal gerund with of,

by or in is, if not necessarily the only option available, at least clearly motivated by

the functional properties of definite marking. This is most evident when the gerund

refers to a specific non-controlled event, as in the examples in (43), where a bare

nominal or verbal gerund is virtually ruled out. In the extreme case, a definite nomi-

nal gerund may even have the effect of blocking a controlled interpretation. This is

illustrated in (44a), where reading is not controlled by the matrix subject (rather, the

matrix subject is being read to) but by a participant previously mentioned in the dis-

course, and in (44b), where inbreathing would be readily – but mistakenly – inter-

preted as being controlled by the highly agentive subject of the matrix clause, if it

were not for the definite article, which marks the referent of the gerund as accessible


through the alternative strategy of drawing on shared knowledge (the common ‘in-

breather’ of ill breaths is man and not the air).79

(43) a. In thys yere also, Charlys herynge of the subduynge of kynge

Rycharde, sent into Engla~de two of his houshold knyghtys,

requyrynge kynge Henrye the. iiii., tha~ newly made kynge, to send

home his doughter Isabell latelye maryed vnto kynge Rycharde (1516,

PPCEME)

b. through most Counties of this Realme Horstealinge is growen so

co~mon, as neither in Pastures or Closes nor hardlie in Stables the

same are to be in safety from stealinge, whiche ensueth by the redye

buyinge of the same, by Horscorsers and others in some open Fayres or

Markett~ farr distant from the Owner, and withe suche speede as the

Owner cannot by pursuyte possiblie helpe the same (1588-9, HC)

(44) a. All which Lettres and copies I have distinctely redde unto his Grace,

who hath in the reding therof substancially considered as well the

Quene his sisters Lettre with the Lettres agaynward devised and sent

by my lord Admirall to her, and his Lettres of advertisement to your

Grace, as your moost politique devises and answeres un to all the same

(1523, HC)

b. A man cannot finde a better remedie than this triacle [i.e. ‘ointment’]

against the pestilence, which being also as it were a wild beast, bred of

ye corruption of the ayre, leaping vpon men by the inbreathing of that

ill breaths which destroyeth, wasteth, and maketh hauock not only of

one man, but of whole tounes and Cities. (1568, PPCEME)

Sometimes the matrix clause provides a potential controller, but the event referred to

by the gerund is also an accessible discourse participant that is highly prominent in

the discourse. For example, in (45a) the declaring of the causes refers to an action

whose realisation is being negotiated, while in (45b) the churning thereof is used in

a text precisely about churning butter. It is a natural choice therefore to represent

these actions as retrievable through the definite strategy rather than the controlled

strategy.

79 It is further worth pointing out that, as appears from (43b) and (44), the meaning of the preposition by can change from a relation of means to a relation of cause when combining with a definite nominal gerund.


(45) a. But my Lordes of the Counsaile before whom I refused it, might well

perceiue [...] that all sturdy stubbernesse whereof obstinacy groweth,

was very farre fro my mynde. For the clearer profe wherof, sith they

semed to take for one argument of obstinacy in me, that refusing of the

othe, I wolde not declare the causes why, I offred with a full heauy

heart, that albeit I rather wolde endure all the payne and peryll of the

statute than by the declaring of the causes, geue any occasion of

exaspiracion vnto my most dradde Souerain Lorde and Prince, yet [...]

I wolde vpon such his gracious licence and cammaundement as

shoulde discharge me of his displeasure and peryll of any statute,

declare those poyntes that letted my poore conscience to receyue that

othe (1534, PPCEME)

b. if you churne your butter in the heate of Sommer it shall not be amisse,

if during the time of your churning you place your Churn in a paile of

cold water as deep as your Creame riseth in the Churne; and in the

churning thereof let your stroakes goe slow, and be sure that your

churne be cold when you put in your creame (1615, HC)

In yet other cases, the use of a definite gerund is simply triggered by coordination or

structural parallelism with another definite NP. The motivation for definite marking,

then, is the same for the gerund as for the coordinated or parallel NP. Thus, in (46a),

the redyng of historyes is construed parallel to the experyence of aduerse fortune,

both being sources of insight in wysedome and polycye. In (46b) the using and

mynistering of them is coordinated with the nature kind and operac~on of certeyne

herbes.

(46) a. He is / and euer hath ben reputed the wysest / whiche by the

experyence of aduerse fortune hath byholden and seen the noble

Cytees / maners / and variaunt condycions of the people of many

dyuerse Regyons / For in hym is presupposed the lore of wysedome

and polycye [...] / yet he is more fortunat / and may be reputed as wyse

[...] that may by the redyng of historyes conteynyng dyuerse customes

Condycyons / lawes & / Actes of sondry nacions come vnto the

knowleche of and vnderstandynge of the same wysedom and polycye

(1477-84, HC)

b. divers honest p~sones aswell men as woomen, whome God hathe

endued with the knowledge of the nature kinde and operac~on of


certeyne herbes rotes and waters, and the using and mynistering of

them to suche as been pained with customable diseases (1542-3,

PPCEME)

In all, it appears that the choice of definite nominal gerunds with in, by and of is

mostly motivated. This may explain why the use of these combinations – though

marginal – remains stable throughout the period examined (see Figure 8.4), despite

the rise of verbal gerunds with the same prepositions.

A further group of definite nominal gerunds – the most substantial group, in

fact – is found with the prepositions for. Regarding these, it is striking to see that

while for can express a variety of semantic relations, it nearly always designates a

purpose or goal when used with definite nominal gerunds. Accordingly, the dia-

chronic rise of definite nominal gerunds with for is entirely due to use with pur-

posive for. This strong bias is demonstrated in Figure 8.6, which compares the use

of definite nominal gerunds with for as purpose adjuncts to their use as reason ad-

juncts.

The preference for purposive uses distinguishes definite nominal gerunds

with for from bare nominal and verbal gerunds. In particular, as shown earlier, bare

nominal and verbal gerunds are much more commonly used with for to form reason

0

24

68

1012

1416

18

1350-1420 1420-1500 1500-1570 1570-1640

purp. for+ DNG

reas. for+ DNG

Figure 8.6. Definite nominal gerunds (DNG) with for as purpose and reason adjuncts

(frequencies per 100,000 words).

adjuncts (see Figure 8.5 and the discussion in Section 3.2.1 above). Predictably,

when a definite gerund is exceptionally used with for to form a reason adjunct, the


gerund refers to a specific and accessible event, contrasting and overlapping with

bare nominal and verbal gerunds along the by now familiar lines. An example is

given in (47).

(47) after the sharpe and vehement contention, betwene him [i.e. Achilles]

and Agamemnon, for the taking away of his concubine (1531,

PPCEME)

More interesting are the purposive uses of for, where the choice between definite

nominal gerunds on the one hand and bare nominal or verbal gerunds on the other

appears to be organised differently. This is plausibly the one area where definite

nominal gerunds on the one hand and bare nominal and verbal gerunds on the other

have come closest to competing and definite gerunds may in fact have partly re-

placed the other two types. Nevertheless, definite nominal gerunds are not entirely

equivalent to the other two types. Start by observing that purpose adjuncts typically

denote events that are still to be realised and that, being more or less hypothetical,

are not easily conceived of as having precise spatio-temporal coordinates. In purpose

adjuncts, then, the definite article does not mark specificity and accessibility in the

strictest possible sense. That is, accessibility lies not so much in the interpreter’s

ability to home in on a specific occurrence of a situation with fixed spatio-temporal

coordinates, as in the construal of a situation as a ‘common goal’. Basically, it is

because a goal can be known and recognised as a goal among a larger group of peo-

ple, that it can also be accessible as shared knowledge in the communicative setting

of discourse.80

In its most distinctive manifestation, a common goal has the following char-

acteristics: first, it benefits others than those who strive to achieve it, and, as a con-

sequence, it carries an aura of general importance; second, its existence as a goal is

at once prior to and continues after any actions taken to achieve it; and third, just as

it is not directly linked to the intentions or desires of a single agent, a single agent is

80 The function of the definite article does not always boil down to marking a com-mon goal. Occasionally, the choice for the definite article appears to be prompted by other considerations. For example, the definite marking in the perchinge and burlinge of Clothe in (viii) is motivated by the reference point relation between cloth and the activities that constitute its processing. (viii) in manye part~ of this Realme ys newly and lately devised erected

buylded and used, certen Milles called Gigge Milles, for the perchinge and burlinge of Clothe, by reason whereof the true Draperie of this Realme ys wonderfullye empayred (1551-2, PPCEME)


not capable of realising it but merely contributes towards its achievement. The defi-

nite nominal gerunds in (48) are used with for to express common goals that fully

answer the description given. For example, the edifiynge of theyr bretherne in (48a)

is a goal whose validity is generally recognised, that benefits the bretherne more

than the ploughmen, and that is presumably not achieved by the ploughmen’s efforts

alone. Likewise, the keepinge of our peace in (48b) benefits an entire shire, is an

ever-continuing task, and constitutes a responsibility that is shared by a variety of

people and which can even be transferred from one authority to another, as indeed

happens in the example. The extreme case is presented by (48c): as a goal, the dis-

closing of Nature potentially entails the enlightenment of all mankind, and certainly

represents an ongoing process that feeds on all kinds of inuentions and experiments

without ever being wholly completed. In fact, the preposition for here can almost be

said to mark a benefactive as much as it marks a goal.

(48) a. But they that wil be true ploughmen [i.e. preachers] muste worke

faythfullye for Goddes sake, for the edifiynge of theyr bretherne.

(1549, PPCEME)

b. And where as we at this time have written aswell to the sheriffe of that

our shire [...] as also to the justices of our peace within our said shire

[...], comaundinge and straitly chargeinge that as well the said sheriffe

as the said justices endeavor them for the keepinge of our peace and

entertainment of our subjectes in good quiet and restfullnes dureinge

the time of our jorney into the partes beyond the sea to the which wee

intend to dispose us aboute the latter end of this present month of May.

(1520, PPCEME)

c. so assuredly the search and stirre to make gold hath brought to light a

great number of good and fruitfull inuentions and experiments, as well

for the disclosing of Nature; as for the vse of mans life. (1605, HC)

In other examples, definite nominal gerunds combine with for to express goals that

approach the prototype of the common goal but do not possess all of its features. For

example, a man’s keeping of his health in (49a) is not a goal to which a great num-

ber of people contribute or from which anyone benefits more than the person whose

health is concerned. Still, good health will be commonly recognised as a valid goal

for any individual person, and is certainly a goal whose achievement cannot depend

solely on the choice between white wines and red and Clared wine, but which re-

quires continuous attention and care. A different set of features of a common goal is


highlighted in (49b): as a goal, the achieving thereof clearly originates from the per-

sonal agenda of one man, this Cardinall, but its achievement requires scheming and

intrigue, inevitably involving others, and the action taken towards its achievement

(casting doubt on the acceptability of another marital candidate) can contribute to

success only indirectly.

(49) a. Smal white wines scoure and driue out the vncleannesse of the bodie as

much as it is possible to be done by them, and red and Clared wine

stoppe and hold backe, and fill the bodie full of ill humors, now which

are most profitable to be taken most commonly of a man for the

keeping of his health (1568, PPCEME)

b. This Cardinall therefore [...] devised to allure the kinge [...] to cast

fantasy to [i.e. ‘get enamoured of’] one of the Frenche kings Sisters:

which thing, because of the Enmity and warre that was at that tyme

betweene the French king and the Emperour (whom, for the cause

afore remembred, he mortally maligned) he was very desirouse to

procure; And for the better atcheving thereof, requested Langland,

Bishoppe of Lincolne, and ghostly father to the kinge, to put a scruple

into his graces head, that itt was not lawfull for him to marry his

brothers wife (1555, PPCEME)

The examples in (48)-(49) can be contrasted with bare nominal and verbal gerunds

with purposive for, as in (50). Here, the goal designated by the gerund is immedi-

ately and exclusively linked to the desires and intentions of the controller in the ma-

trix clause, and can be realised directly by the action described in the matrix clause

(see also the examples in (35) above).

(50) a. I am in a greet angonye howe js best fore me to sywe to hym fore

rehauyng off my place. Þat goode lorde woot full lytell how moche

harme he doothe me, and how lytell goode ore worshyp it dothe hym.

(1472, HC)

b. It shall be therfore, and also for refreshing the witte, a conuenient

lesson to beholde the olde tables of Ptholomee, where in all the worlde

is paynted (1531, PPCEME)

c. For besydes money, he looketh for meate and drinke for doinge his

dame pleasure. (1567-8, PPCEME)


The contrast established between common goals and direct or personal goals

underscores the semantic impact of the definite article and implies that the choice

between definite nominal gerunds and bare nominal or verbal gerunds is not arbi-

trary in the context of purposive for. Still, it is clear that conceptually the distinction

may not be discrete (between the two we can imagine a semantically fuzzy transition

area), and that the choice of one variant over another may in many cases primarily

be a matter of construal (i.e. most goals might be presented as common goals if

rhetoric purposes so allow). Especially in the Middle English period, bare nominal

gerunds could combine with for to express goals that look suspiciously like common

goals. Thus, putting a-wey of þis opyn & sclaunderouse deseyt in (51a) can easily be

conceived of as a common goal (and is in fact coordinated with a definite nominal

gerund). Conversely, in the Early Modern data definite nominal gerunds occasion-

ally occur with for to designate goals showing very few of the features of common

goals. For instance, the avoiding of singularity in (51b) may be thought of as a per-

sonal goal that contributes little to general well-being (even if it perhaps represents

an example of modesty that others would recognise as a virtuous goal).

(51) a. And noght only for exclusion & puttyng a-wey of þis opyn &

sclaunderouse deseyt, here with-yn þis Citee late practisid and

bygonne, but also for þe redy remevyng of grete multitude of such

wynes, deceyuablych contrefetyd and medlid on þe other syde of þe

see, and broght hydir to selle, The peticion of þe Comons hath oftyn

here-to-for requirid a couenable remedie (1419, HC)

b. This Lord Chauncelour, albeit he was to god and the world well

knowen of notable vertue [...] yeat, for the avoiding of singularity,

wold he appeare none otherwise then other men in his apparell and

other behaviour. (1555, PPCEME)

This hints at a certain degree of competition between definite nominal gerunds on

the one hand and bare nominal and verbal gerunds on the other. It must be pointed

out, however, that this view is only weekly corroborated by the quantitative data.

There is no decrease in bare nominal and verbal gerunds concomitant to the marked

increase of definite nominal gerunds with purposive for (compare Figures 5 and 6

above). At the same time, compared to other environments, where the frequency of

verbal gerunds continually increases for several centuries, it is remarkable that the

rise of verbal gerunds with purposive for seems to peter out very soon after its in-

ception. A blocking effect due to the use of definite nominal gerunds in the same


environment is therefore not entirely inconceivable. Alternatively, the blocking fac-

tor may also be the to-infinitive, which is also used to form purpose adjuncts, and

more closely resembles bare nominal and verbal gerunds with for in being typically

controlled and in expressing direct goals rather than common goals.

A few final words are in place concerning combinations of definite nominal

gerunds with to. Of the three gerund types, it is definite nominal gerunds that most

commonly combine with to, bare nominal gerunds being rare and verbal gerunds

hardly occurring at all. In part, definite nominal gerunds are used with to just as with

for to express a common goal. Thus, the definite nominal gerunds in (52) represent

preconceived or pre-established, often socially relevant goals to which the action of

the main clause more or less indirectly contributes. Note that, as (52a) shows, com-

mon goals with to historically predate those with for and may therefore be thought

of as a predecessor to the more frequent pattern.

(50) a. Also y bequeth to þe Mendyng of þe heye way be-twen Hillindon and

Akton, xl.s. (1402, HC)

b. For if ye bryng it to passe, that the yomanry be not able to put their

sonnes to schole – as in dede vniuersities do wonderously decaye al

redy – and that they be not able to mary their daughters to the

auoidyng of whoredome, I say ye plucke saluation from the people and

vtterly distroy the realme. (1549, PPCEME)

In other cases the combination of to with a definite nominal gerund functions more

as a complement than as an adjunct, yet the sense of a commonly recognised goal is

still eminent. A clear example is given in (53), where the understanding of autors

and the most swete and pleasant redinge of olde autours are construed as the ulti-

mate gratification and actual goal of the strenuous process of learning grammar, and

must therefore be thought of as the preset end-stage of grammar teaching.81

(53) Grammer beinge but an introduction to the understanding of autors, if

it be made to longe or exquisite to the lerner, hit in a maner mortifieth

his corage: And by that time he cometh to the most swete and pleasant

redinge of olde autours, the sparkes of feruent desire of lernynge is

extincte with the burdone of grammer [...]. (1531, PPCEME)

81 Note, incidentally, that the explanation here applies to definite nominal gerunds used within a generic text (see Section 3.1 above).


More marginally, definite nominal gerunds can also be used to specify a (typically

undesirable) result, which is then presented as an accessible fact. An example is

given in (54).

(54) so great haboundaunce of cappes and hattes redy wrought and made

have been and daily be brought from the p~ties of beyond the See into

this Realme, and here have been and daylly bee uttred and sold [...] to

the great Idelnesse enpov~ysshyng and utter undoing of great multitude

of the Kyngis naturall subjectis borne wythin this said Realme (1511-2,

PPCEME)

It is difficult to assess the degree to which definite nominal gerunds with to

competed with bare nominal gerunds in the same environment. In resultative uses as

in (54) above, definite nominal gerunds are likely to have replaced bare nominal

gerunds (despite a slight semantic difference – compare (17a) above). Otherwise,

replacement is doubtful. In the entire data set, only two bare nominal gerunds with

to function as purpose adjuncts, which is hardly enough to speak of replacement. At

best, competition may have blocked any growth of the pattern – though again the

main blocking factor may in actual fact have been the to-infinitive rather than the

definite nominal gerund. The most important remaining use of bare nominal gerunds

with to is that of prepositional complement, especially in idiomatic combinations

where the purposive sense of to is largely absent, as in the examples in (55a-b). Dis-

regarding the occasional intrusion of a definite nominal gerund, bare nominal ger-

unds appear to have survived in this use into Early Modern English – a fine illustra-

tion of which is the continued association of the idiom fall to with bare nominal ger-

unds up until the eighteenth century, as shown in (55c).

(55) a. Schynyng eyn [...] qwan thei be of amenabyl gretnes, moyste and

schynyng, thei sygnyffye gret nobylnes and gret ymagynacionnys off

wysdam, dysposycion also to wrathe and to drynkyng off wyne, gret

dysyre off wurchyp; also thei sygnyfye manffulnes and hardynes

(c1450, HC)

b. And out they goe, in a fustian fume, into the backe syde, where was a

great Axiltrye [i.e. ‘axle’], and there fell to pitching of the barre [i.e. a

game involving a heavy piece of wood that the contenders throw as far

as possible] (1567-8, PPCEME)


c. words could not vent half their rage; and they fell to pulling of caps,

tearing of hair, and dragging the clothes off one another’s backs (1749,

CLMETEV)

To summarise, definite nominal gerunds do not engage in competition to the

same extent as do bare nominal and verbal gerunds. This is evident from their distri-

bution over the set of prepositions. Definite nominal gerunds have specialised into

two major uses – that of temporal landmarks with prepositions such as after, at and

so forth, and that of purpose adjuncts designating a ‘common goal’ with the preposi-

tions for and to. In the former use, competition with bare nominal or verbal gerunds

is as good as non-existent prior to 1640, whereas in the latter use a certain degree of

competition is possible but is only weekly apparent (if apparent at all) in the data.

With other prepositions, definite nominal gerunds remain relatively marginal

throughout the period examined. This is especially true for their use with the prepo-

sitions without, by, for (in its causal sense) and to (in non-purposive prepositional

complements) and, to a lesser extent, of and in. When definite gerunds are occasion-

ally used with these prepositions, their use tends to be clearly motivated by the

properties of the definite article. While it is not incorrect to state that definite nomi-

nal gerunds potentially overlap in use with other gerund types, they never become

very current in such overlapping functions and at no point is there very straightfor-

ward quantitative evidence that definite nominal gerunds replaced or were replaced

by another gerund type in any specific environment. All in all, then, definite nominal

gerunds can be said to go their own way, more or less independently of the other

two gerund types, specialising into their own distinct set of uses.

3.3. Syntactic differences

The preceding discussion has emphasised the similarity between bare nominal and

verbal gerunds, which competed over the same environments, as opposed to definite

nominal gerunds, which specialised into their own niche of language use. This im-

plies that verbal gerunds did not fill a syntactic gap in any strong sense. The appear-

ance of verbal gerunds did not bring about a dramatic innovation involving an en-

tirely new set of expressive options suddenly available to users of English. Instead,

the uses to which verbal gerunds were put were to a large extent established on or at

least prefigured by the various uses of bare nominal gerunds. The question therefore

remains what drove the spectacular increase of verbal gerunds that began in the

Middle English period and continued into Modern English.


An answer to this problem is suggested by a final functional difference be-

tween the three gerund types. While from a discourse-functional and distributional

point of view verbal gerunds align themselves with bare nominal gerunds, they dif-

fer from the latter in terms of their internal syntax. This difference reflects on the

way the three gerund types are used, in that verbal gerunds display greater syntactic

flexibility. That this is so is most obvious in later stages of the development of the

gerund. For example, in Present-Day English, there is no nominal alternative to the

auxiliary being in (56a), the object complement responsible in (56b), or the subject

complement a minimalist in (56c).

(56) a. She showed no sign of being overwhelmed by the atmosphere of

Schloss Langenbach. (CB)

b. Only by returning it to the peasantry, he argued, only by making farm-

ers responsible for marketing their own produce, would it be possible

to ensure regular food supplies. (CB)

c. I will have to get back to being a minimalist to retain such sanity as I

have. (CB)

The syntactic configurations illustrated in (56) are attested only sparingly in the his-

torical data. Nevertheless, the data show that from the very start the internal syntax

of verbal gerunds was more flexible and more strongly favoured complex patterning

0%

20%

40%

60%

80%

100%

VG BNG DNG

complex

simple

Figure 8.7. Proportion of verbal (VG), bare nominal (BNG) and definite nominal

gerunds (DNG) with simple and complex internal syntax.


than that of nominal gerunds – even when the syntactic potential of either construc-

tion type was in theory roughly identical.

This is demonstrated in Figures 8.7 and 8.8. Figure 8.7 collapses the different

sub-periods of the corpus and categorises the three whole sets of verbal, bare nomi-

nal and definite gerunds into two distinct categories. The first category is that of

‘simple’ gerund constructions, which contain only a patient argument, either ex-

pressed as a bare NP (in the case of verbal gerunds) or as an of-phrase (in the case of

nominal gerunds). The second category is that of ‘complex’ gerund constructions,

which in addition to a patient argument contain some other modifier to the gerundial

verb (e.g. an adjective, a prepositional phrase, a second non-subject argument and so

on). The examples in (57) illustrate the distinction between simple and complex ger-

und constructions.

(57) a. they would not obey hym, but refused and defendid the~ by takynge

sanctuarye (1556, PPCEME)

b. I praye you, my Lordes, be good vnto vs, and lette vs not bee molested

for dischargyng our Consciences truelye (1554, HC)

The same sub-categorisation is repeated in Figure 8.8, with the only difference that

the set of complex gerund constructions has been reduced by only including those

examples whose additional modifier is a prepositional phrase. The rationale for this

reduction is that of all possible extra modifiers prepositional phrases are most likely

to be acceptable in both nominal and verbal gerund constructions, as is illustrated in

(58).82 In this way, Figure 8.8 indicates more reliably how simple and complex syn-

tactic configurations correlate with the choice between nominal and verbal construc-

tions.

(58) a. The Lord kepe me from doyng this thing vnto my maister yat is the

lordes anointed. (1549, PPCEME)

82 Adverbials too can have a nominal alternative in the form of an adjective preceding the gerund. However, a good number of adverbials have no adjectival equivalent, and some adjectives used in a nominal gerund do not fulfil the function of a corresponding adverbial in a verbal gerund. The correspondence is not as complete therefore as that between prepositional phrases in nominal and verbal gerunds.


0%

20%

40%

60%

80%

100%

VG BNG DNG

add. PP

simple

Figure 8.8. Proportion of verbal (VG), bare nominal (BNG) and definite nominal

gerunds (DNG) with simple syntax or with additional PP.

b. Humfrey Boucher, base sunne to the late Lorde Berners, did much

coste in translating of the priorie into a maner place (1535-43,

PPCEME)

c. what with the shotte & gorynge [i.e. ‘stabbing’] of their horses with the

sharpe stakes, they stumbelyd one vpon another [...], so yt in a short

whyle a great multytude of horse & men were layde vpon the grounde.

(1516, PPCEME)

Figures 8.7 and 8.8 reveal a clear preference for verbal means of expression when

the gerund construction’s internal syntax is more complex.83 This tendency is not

merely the result of the greater syntactic potential of verbal gerunds. As Figure 8.8

shows, verbal means of expression are preferred in complex gerunds even when the

nominal alternative is perfectly viable.

Interestingly, the observations in Figures 8.7 and 8.8 extend to another area of

gerundial usage. The same skewed distribution between simple and complex internal

patterning is found in the relation between definite nominal gerunds and definite

hybrid gerunds. As defined here, definite hybrid gerunds combine a definite article

with a verbally expressed object argument, as illustrated in (59).84

83 Subjecting the raw figures to a χ2-test, the difference between verbal gerunds and the two nominal gerund types is significant at p < 0.001. 84 This is a very narrow definition. In general, the mixture between nominal and ver-bal features that characterises definite hybrid gerunds can also take other forms. In some cases the definite article is missing, in others it is not the object that is ex-pressed by verbal means but some other modifier. The only possible comparison in


0%

20%

40%

60%

80%

100%

DNG DHG

complex

simple

Figure 8.9. Proportion of definite nominal (DNG) and definite hybrid gerunds

(DHG) with simple and complex internal syntax in 1570-1640.

(59) a. Now they proceed to the Reading the Proofs. (1600, HC)

b. First the childe is to be taught, how to call every letter, pronouncing

each of them plainely, fully and distinctly; [...] More specially to bee

carefull, for the right pronouncing the five vowels, in the first place, as

a, e, i, o, u. (1627, HC)

Important for our purposes is that, as the examples in (59) indicate, definite hybrid

gerunds show the same variation between simple and complex internal syntax as

other gerund types. Figure 8.9, then, gives the proportion of definite nominal and

definite hybrid gerunds in simple and complex environments for the period 1570-

1640 (see Appendix B). Only one sub-period is represented, because definite hybrid

gerunds only began to gain some currency in the Early Modern period. The result is

evident: the use of definite hybrid gerunds conforms to the tendency for complex

syntax to associate with verbal means of expression.85

Taken together, Figures 8.7 through 8.9 hint at the possible functional moti-

vations behind the rise of verbal gerunds (and, by extension, definite hybrid ger-

unds). Knowing that verbal gerunds grew proportionately more frequent over time,

terms of simple vs. complex constructions is between definite nominal gerunds and definite hybrid gerunds with a verbally expressed object. Bare hybrid gerunds are complex by definition, as they contain at least one verbal and one nominal modifier, and the same applies to definite hybrid gerunds whose patient argument is expressed by means of an of-phrase. 85 The difference is significant at p < 0.01 (applying a χ2-test to the raw figures).


0

10

20

30

40

50

60

1350-1420 1420-1500 1500-1570 1570-1640

VG (simple)

VG (complex)

Figure 8.10. Simple and complex verbal gerunds from 1350 to 1640 (frequencies per

100,000 words).

it can be deduced that the use of verbal gerunds has brought about an increase in the

use of gerund constructions with more complex internal patterning (see also Appen-

dix B). If we interpret this overall trend as the manifestation of an underlying func-

tional motivation, the emergence of verbal gerunds can be seen as at least in part

driven by the need for syntactically more versatile means of expression. Some inter-

esting evidence in favour of this view comes from a diachronic shift in the propor-

tion of simple to complex constructions in verbal gerunds. Figure 8.10 tracks the

frequency of verbal gerunds with simple and complex internal syntax, revealing that

whereas the increase in verbal gerunds between 1500 and 1640 is felt in roughly

equal measure in complex and in simple gerund constructions, the initial rise in the

frequency of verbal gerunds between 1350 and 1500 seems to be mostly due to ver-

bal gerunds with complex internal syntax.86 This suggests that the original impetus

for the emergence of verbal gerunds may have lain in the drive for greater syntactic

flexibility.

This is not all to it, however. Recall that the preference for verbal gerunds

over nominal gerunds is not confined to complex syntactic configurations that can

only be expressed verbally. It is clear that in all complex environments speakers

86 The proportion of simple to complex verbal gerunds between 1420 and 1500 dif-fers significantly from that between 1500 and 1570 (p < 0.01, using a χ2-test). Due to the paucity of data, the difference between the sub-periods 1350-1420 and 1420-1500 does not reach the 0.05 threshold of statistical significance, but the probability that the difference is due to chance is still fairly low (p < 0.1, using a χ2-test).


prefer to code the patient of a gerund by means of a bare noun phrase rather than an

of-phrase, regardless of whether a nominal alternative is available or not. As such,

the tendencies observed in gerundial usage are reminiscent of Rohdenburg’s (1995,

1996) ‘cognitive complexity principle’, which states that “In the case of more or less

explicit grammatical options the more explicit one(s) will tend to be favored in cog-

nitively more complex environments” (1996: 151) (see Chapter 3). Indeed, although

Rohdenburg primarily associates explicitness with formal weight,87 a number of

arguments suggest that despite the greater formal weight of the of-phrase bare noun

phrases represent the more explicit option. First, given that in any environment apart

from the gerund patient arguments are expressed by means of a bare noun phrase,

the bare noun phrase is by far the more canonical way of coding the patient. Second,

being a prepositional phrase, an of-phrase is less explicit in marking the difference in

syntactic and semantic status between an argument and a non-argument – a differ-

ence that becomes increasingly relevant exactly in expressions that contain both

arguments and non-arguments, and in which the difference between a prepositional

phrase and a bare noun phrase can be seen as an iconic reflex of degrees of semantic

‘closeness’ to the action denoted by the verb. Third, of-phrases, unlike bare noun

phrases, can be ambiguous between a patient interpretation and an agent interpreta-

tion (cf. Jespersen 1940: 102), as is illustrated in (60).

(60) it shoulde bee harde for him to brynge his purpose to passe, without the

gathering and great assemble of people and in maner of open warre

(1514-8, PPCEME)

If the assumption is correct that bare noun phrases are more explicit than of-phrases

in marking the patient of a verb, this has the important consequence that, compared

to their nominal counterparts, verbal gerunds appear at once to offer a more

economic (i.e. shorter) and a more iconic (i.e. more explicit) way of saying the same

thing (cp. Haiman 1983). In other words, economic and iconic motivations converge

in the verbal gerund, giving it a plausible functional advantage over its nominal

competitors.

87 As Rohdenburg states, however, “differences in grammatical explicitness may be achieved in many other ways” (1996: 151).


4. Functional and other motivations in the development of the gerund

Pulling the various strands of the preceding discussion together, we get a compli-

cated but largely consistent picture of the various developments in gerundial usage

and their underlying motivations. To work out this broader picture is the purpose of

this section. In general, functional motivation can be conceived of as arising through

motivated decisions made by language users on the basis of what synchronic gram-

mar has to offer (García 1999; Haspelmath 1999; Hawkins 2004). In this respect, the

history of the gerund indicates that functionally driven language change is in fact

possible, provided that it is implemented through repeated motivated choices in syn-

chronic usage. The factors determining synchronic choices are a complex bunch,

however. Most prominently, they consist in functional similarity and dissimilarity,

coupled to iconicity and economy, though other factors too demand their share of

the pie, such as blocking. At the same time, these language-internal factors might

not be able to fully explain the development of the gerund, so that it may be neces-

sary to invoke a language-internal factor as well, in the form of French influence.

Clearly, the history of the English gerund cannot be understood without a

close understanding of the functioning of the system of choices in which gerund

constructions participate. Use of the three major gerund types examined in this chap-

ter is organised along two major divisions. On the one hand, verbal gerunds – and

definite hybrid gerunds – contrast with bare and definite nominal gerunds in allow-

ing greater syntactic flexibility (Section 3.3). On the other hand, bare nominal and

verbal gerunds resemble each other in the way they function in discourse, as op-

posed to definite nominal gerunds (Section 3.1). These discourse-functional proper-

ties of gerund types interact with the concrete uses to which they are put, as is re-

vealed by the distribution of the gerund types over the set of available prepositions.

Definite nominal gerunds have specialised into a set of uses that are uniquely moti-

vated by the semantic import of the definite article (Section 3.2.2). Bare nominal and

verbal gerunds compete with one another over the same set of environments (Sec-

tion 3.2.1).

Together, functional differentiation and overlap between the different gerund

constructions can explain some of the historical developments witnessed in the use

of the gerund. The functional specialisation of definite nominal gerunds can account

for the rise of definite nominal gerunds alongside verbal gerunds and the ultimate

survival of definite nominal gerunds into Present-Day English. In contrast, the simi-

larity between bare nominal and verbal gerunds, along with the greater syntactic

flexibility of verbal gerunds, can account for the loss of bare nominal gerunds and


the concomitant rise of verbal gerunds, since we may assume that one got systemati-

cally preferred over the other in selecting a variant to fill the same functional slots.

Because the unique syntactic flexibility of the verbal gerund is likely to have played

its part in furthering the construction, verbal gerunds did fill a kind of gap, in that

they provided a more flexible means of expression, at once iconic and economic, in

an area of grammar where it was hitherto unavailable. The gap, then, lay in internal

rather than external syntax – that is, in internal syntactic flexibility rather than in

external combinatorial potential.

Regarding the question of whether the gerund also filled a syntactic gap in

terms of external combinatorial potential, a nuanced answer must be formulated. It is

not correct to say that verbal gerunds in particular substituted for the infinitive in

prepositional environments because they presented the only clausal structure that

could combine with prepositions. The use of verbal gerunds is clearly anticipated by

the use of bare nominal gerunds, and the rise of verbal gerunds consists more in a

large-scale replacement of bare nominal gerunds than in the filling of a grammatical

void. Still, there is the dramatic rise of bare nominal gerunds halfway through the

Middle English period, which, in speed and scope, is unparalleled by any later de-

velopment in the gerund (Section 2). This may reflect the fact that at that time the

assumed prepositional gap was indeed still a gap.

How exactly this other gap is to be conceived of is a different issue and one

about which I have no ready answer available. It may be thought of as an attraction

pole for a new construction, but we might also take the opposite and decidedly safer

perspective by arguing that the mechanism of blocking (Chapter 4) is the real moti-

vating factor and that the syntactic gap and, in particular, its getting filled are

epiphenomenal. In general, what we can say is that gerund constructions only gained

frequency in those areas where they somehow came in useful, resulting in a division

of labour between the different constructions. There is no evidence in the data to

support the strictly teleological point of view (although there is no strong counter-

evidence either). The only thing the data show is that there are gaps in grammar

which leave space for new constructions to develop, without necessarily being the

reason of the development.

The point is best illustrated by definite nominal gerunds, which remained

marginal in those uses where they could not sufficiently distinguish themselves from

bare nominal or verbal gerunds, but prospered in uses where the semantic import of

the definite article was most fully and uniquely felt (Section 3.2.2). A similar com-

plementary relationship may be identified between the three gerund constructions

and the infinitive. It is certainly reasonable to assume that the spectacular rise in


gerundial usage in prepositional environments was possible only because of the ab-

sence of the infinitive in those environments. On a smaller scale, we may note the

absence of purpose adjuncts formed by to with a bare nominal or verbal gerund and

the somewhat half-hearted use of bare nominal and verbal gerunds with purposive

for. The reluctance of bare nominal and verbal gerunds to combine with purposive to

and for may result from a blocking effect due to the use of definite nominal gerunds

with the same prepositions, but it is more plausible that they have been blocked by

the to-infinitive, whose purposive semantics and control behaviour more closely

resemble that of bare nominal and verbal gerunds than do the characteristics of pur-

posive definite nominal gerunds (Section 3.2.2).

Of course, if we assume that gap-filling is epiphenomenal to blocking, we are

still in want of a driving force behind the surge in frequency of bare nominal ger-

unds halfway the Middle English period. As the functional forces discussed so far

operate on choices, they require variation, which means that they can account for

most of what happened after (say) 1400, but cannot tell us what first set the machin-

ery of interacting factors going. A possibility hitherto unexplored is that the devel-

opment of the gerund has in part been fuelled by language contact, with French pro-

viding the model that English gerundial usage has aimed to replicate. So far, the

influence of French, in as far as it is considered relevant to the development of the

gerund, has been sought in the first appearance of verbal features in gerundial con-

structions in Early Middle English (see Section 1), but it is conceivable that French

has also had some impact on the subsequent spread of gerund constructions, regard-

less of whether they were nominal or verbal.

There is, for example, no denying the role of French as far as adverbial con-

structions with in and gerund are concerned, as these clearly replicate the French

pattern with gérondif introduced by en (explaining at once why English in in this

construction deviates from its usual semantics; see Section 3.2.1).88 Compare in this

light the examples in (61) and (62), the latter of which shows a particularly obvious

rendering of a French idiom.

88 In fact, with its ending in -ant, the French gérondif could easily occasion confu-sion with the French present participle, as indeed happened in the history of French (Martin & Wilmet 1980: 216-7). If, likewise, users of English perceived of the French gérondif as a present participle, this may have led to confusion between the gerund introduced by in and the present participle in English as well, thus promoting the use of verbal characteristics in this specific environment. This may at once ex-plain the remarkable preponderance of verbal gerunds with in in this use evidenced in Middle English (see Section 3.2).


(61) a. Now how this myghty prynce executyd the hyghe co~maundement of

god in helpynge the people & sauyng ye same by the grete power

geuen vnto hym of god ye sayd story of Josue paynly doth declare

(1495?, HC)

b. l’ancienne dame lui en tendra plait en le ramentevant et devisant les

bons moz qu’elle lui aura ouÿ dire de l’amour qu’il a d’elle, (1405,

Halmøy 2003: 27)

‘The old lady will plead [his] cause with her in reminding [her] of him

and talking about the fine words that she has heard him say about the

love he feels for her.’

(62) a. “And I it yow graunte,” quod sir Gawein, all in laughynge, “for in yow

it is right well employde.” (1450-1460, IMEPC)

b. Et quant ele oï le respons Yselt, ele li dist tot en riant (13th c., Halmøy

1988: 72)

‘And when she heard Yselt’s response, she said to her all in laughing.’

French influence may have extended beyond this one construction, however. Al-

though (pending further research) the evidence remains anecdotal, it does not seem

improbable that the use of gerunds in English was promoted by various French con-

structions that were otherwise hard to render. Next to the gérondif with en, these

included derivational nominalisations, nominalised infinitives, and infinitives with a

preposition. Among the latter, for instance, are infinitives introduced by de (‘of,

from’), pour (‘for’), sans (‘without’) and other prepositions, most of which combi-

nations have a gerund with preposition as their closest English equivalent. The

French pattern is illustrated by the examples in (63) from the Anglo-Norman Dic-

tionary:

(63) a. Il le trespasserent sanz pardun demander (c1235, AND)

‘They wronged him without asking for pardon’

b. Le tierz si est a la feste Seint Johan, pur garder la cite de arsoun pur la

graunt secheresce. (1419, AND)

‘The third (general assembly) is at the feast of Saint John, for guarding

the city against fire because of the great draught.’

c. le roy luy comaunda lever, et sur peine de perdre vie et membre qe

hastivement luy feist venir [...] le seignour de Wake (c1343, AND)

‘the king commanded him to rise, and upon pain of losing life and

limb, had him promptly bring Lord Wake over to him’


As another illustration, it is worth considering the relation between diverse French

constructions and English gerunds in Chaucer’s Tale of Melibee, a direct and very

close translation from Renaud de Louens’ French rendering of the Liber consola-

tionis et consilii by Albertano of Brescia (Severs 1958). In the following examples,

Chaucer’s translation (integral part of PPCME) is set off against the French original

(after the edition in Severs 1958).

(64) a. his flatereres maden semblant of wepyng (c1390, PPCME)

b. les losengiers firent semblant de plorer (a1336, Severs 1958: 571)

(65) a. musik in wepynge is a noyous thyng (c1390, PPCME)

b. musique em plour est ennuyeuse narracion (a1336, Severs 1958: 572)

(66) a. by moneye and by havynge grete possessions been alle the thynges of

this world governed (c1390, PPCME)

b. par argent et par avoir se gouvernent les choses (a1336, Severs 1958:

600)

(67) a. we conseille yow aboven alle thyng that right anon thou do thy

diligence in kepynge of thy propre persone (c1390, PPCME)

b. nous te conseillons, Mellibee, que tu sur toutes choses ayes diligence

de garder ta personne (a1336, Severs 1958: 571)

Closer analyses of Old and Middle French usage would be necessary to prove the

point, but it is clear from the above examples that gerunds could be exploited to ren-

der a variety of French constructions in English. Consequently, the influx of all

kinds of French idioms in the course of the Middle English period (Dekeyser 1986)

may bear responsibility for the increased reliance on nominal and by extension ver-

bal gerunds.

5. Conclusions

The goal of this chapter has been to get a firmer grip on the rise of verbal gerunds

and their relation to other gerund constructions. Examining verbal, bare nominal and

definite nominal gerund constructions, it has been proposed that developments in

gerundial usage have been largely determined by language-internal functional fac-

tors. Roughly, if a new form did not differ substantially from a more common older

form, the older form has been maintained, but if a new form had a unique functional

advantage, it has either come to replace or to complement the older form. This ac-


counts for the replacement of bare nominal by verbal gerunds and for the selection

of a specialised functional niche by definite nominal gerunds. It might not fully ac-

count for the rise of bare nominal gerunds in Middle English. With respect to the

latter development, we might have to recur to the old teleologically formulated

claim that gerunds filled a syntactic gap, but it is also possible that their use began to

increase as an indirect consequence of French influence.

These findings link up to the discussion in both the previous and the follow-

ing chapter. With respect to the relation between gerunds and nouns, discussed in

Chapter 7, we can see that with the development of verbal internal syntax gerunds

did not straightaway abandon other aspects of their nominal behaviour. In terms of

grounding, clausal grounding strategies came to be exploited next to nominal strate-

gies, but in terms of distribution, verbal gerunds have stuck closely to the distribu-

tion of their nominal counterparts, usurping bare nominal gerunds rather than ex-

tending to new environments. In essence, this confirms the relevance of the principle

of paradigmatic regularity proposed in Chapter 7 to account for the distribution of

gerund clauses. With respect to the discussion of gerundial ing-complementation in

Chapter 9, it may be noted that the discussion in the present chapter has shown that

verbal gerunds are a spreading pattern. In the period considered so far, the use of

gerund clauses as verb complements is underway, only beginning to establish itself.

Importantly, however, the interaction of gerund clauses with nominal constructions,

apparent in their development in prepositional environments, will continue to play

an important part in the extension of gerunds to complement positions.

APPENDIX A1

1250–1350 1350–1420 1420–1500 1500–1570 1570–1640

AF RF AF RF AF RF AF RF AF RF N-OBL 2 2.1 1 0.5 1 0.5 3 0.5 6 3.2

in 1 1.0 9 4.9 31 14.5 111 19.5 40 21.1 purp. 1 0.5 21 3.7 6 3.2 reas. 35 6.1 11 5.8 other 9 1.6 6 3.2

for

TOTAL 1 0.5 65 11.4 23 12.1 of 2 1.1 6 2.8 19 3.3 23 12.1 by 2 1.1 1 0.5 35 6.1 44 23.2 to 1 0.2 2 1.1 without 1 0.5 1 0.5 14 2.5 6 3.2 other 1 0.5 3 1.4 17 3.0 14 7.4

VG OBL

TOTAL 1 1.0 15 8.1 43 20.1 262 46.0 152 80.1 N-OBL 6 6.2 10 5.4 19 8.9 24 4.2 6 3.2

in 1 1.0 55 29.9 43 20.1 50 8.8 5 2.6 purp. 1 1.0 11 6.0 10 4.7 32 5.6 11 5.8 reas. 1 1.0 4 2.2 5 2.3 44 7.7 1 0.5 other 4 2.2 6 2.8 14 2.5 1 0.5

for

TOTAL 2 2.1 19 10.3 21 9.8 90 15.8 13 6.8 of 1 1.0 13 7.1 17 7.9 35 6.1 3 1.6 by 4 2.2 9 4.2 30 5.3 6 3.2 to 2 1.1 8 3.7 5 0.9 2 1.1 without 2 1.1 3 1.4 5 0.9 1 0.5 other 8 4.3 5 2.3 17 3.0 3 1.6

BNG OBL

TOTAL 4 4.1 103 55.9 106 49.6 232 40.7 33 17.4

N-OBL 5 5.1 14 7.6 12 5.6 66 11.6 32 16.9 in 2 1.1 5 2.3 28 4.9 13 6.8

purp. 2 1.1 4 1.9 52 9.1 29 15.3 reas. 1 0.5 2 0.9 6 1.1 1 0.5 other 10 1.8 3 1.6

for

TOTAL 3 1.6 6 2.8 68 11.9 33 17.4 of 2 2.1 3 1.6 22 10.3 16 2.8 4 2.1 by 5 2.7 3 1.4 12 2.1 3 1.6 to 1 1.0 3 1.6 8 3.7 25 4.4 6 3.2 without 1 0.2 other 6 3.3 11 5.1 42 7.4 25 13.2

DNG OBL

TOTAL 3 3.1 22 11.9 55 25.7 192 33.7 84 44.3 Verbal (VG), bare nominal (BNG) and definite nominal gerunds (DNG) in oblique (OBL) and non-oblique (N-OBL) environments and their distri-bution over different prepositions, with absolute (AF) and relative frequencies per 100,000 words (RF).

APPENDIX A2

1350–1420 1420–1500 1500–1570 1570–1640 VG BNG DNG VG BNG DNG VG BNG DNG VG BNG DNG

about 4 5 1 afore 1 after 2 1 2 9 1 6 against 1 1 1 1 as 1 1 2 1 3 at 1 3 1 1 6 1 5 5 before 1 1 6 besides 2 contra 1 for/concerning 2 from 1 4 7 3 4 1 1 into 2 1 lest 1 of/for 2 1 1 on 1 since 1 1 1 through 2 to/for 1 toward 1

towards 1 unto 1 1 1 upon 2 3 with 4 1 1 2 6 1 4 2 1 4 TOTAL 1 8 6 3 5 11 17 17 42 14 3 25 Verbal (VG), bare nominal (BNG) and definite nominal gerunds (DNG) with other prepositions than the most fre-quent six (absolute frequencies).

APPENDIX B

1350–1420 1420–1500 1500–1570 1570–1640 TOTAL

AF % AF % AF % AF % AF % simple 10 66.7 18 41.9 154 58.8 93 61.2 275 58.3 complex 5 33.3 25 58.1 108 41.2 59 38.8 197 41.7

(i)

TOTAL 15 100.0 43 100.0 262 100.0 152 100.0 472 100.0 simple 10 71.4 18 56.3 154 70.0 93 72.7 275 69.8 add. PP 4 28.6 14 43.8 66 30.0 35 27.3 119 30.2

VG

(ii)

TOTAL 14 100.0 32 100.0 220 100.0 128 100.0 394 100.0 simple 85 82.5 91 85.8 196 84.5 26 78.8 398 84.0 complex 18 17.5 15 14.2 36 15.5 7 21.2 76 16.0

(i)


BNG

(ii)

TOTAL 94 100.0 99 100.0 211 100.0 31 100.0 435 100.0 simple 19 86.4 47 85.5 154 80.2 69 82.1 289 81.9 complex 3 13.6 8 14.5 38 19.8 15 17.9 64 18.1

(i)


DNG

(ii)

TOTAL 20 100.0 49 100.0 169 100.0 73 100.0 311 100.0

simple 5 45.5 complex 6 54.5

(i)

TOTAL 11 100.0 simple 5 62.5 add. PP 3 37.5

DHG

(ii)

TOTAL 8 100.0 Verbal (VG), bare nominal (BNG), definite nominal (DNG) and definite hybrid gerunds (DHG) with (i) simple vs. complex syntax and (ii) simple syntax vs. with additional PP, in absolute frequencies (AF) and percentages (%).

CHAPTER 9: The diffusion of ing-complements I: Ger-unds as verb complements

The purpose of this chapter is to work out a first extensive case-study on the devel-

opment of ing-complementation, with a specific focus on the use of subject-

controlled ing-complements of gerundial origin and some similar-looking patterns.

The construction primarily at issue is that illustrated in (1), displaying the use of an

ing-clause as complement to a transitive verb.

(1) a. Mr Comben could not help mentioning the poky and cluttered envi-

ronment that is his upstairs shared city office (CB)

b. Would you mind putting Bessie’s exercise book back exactly where

you found it? (CB)

c. Lenny eased the door closed quietly but did not escape being noticed.

(CB)

The literature available (Visser 1963-73; Fanego 1996a, 2007) tells us that verb

complementation, and gerundial ing-complementation in particular, has been a ma-

jor area of change throughout the Modern period, with gerundial ing-complements

spreading to an ever increasing number of verbs. The rise of ing-complements, as

part of the rise of verbal gerunds (Chapter 8), reveals the characteristic pattern of

diffusional change and its study can therefore illuminate our understanding of this

type of change. At the same time, the history of gerundial complementation also

leads us back to the discussion on the status of ing-complementation as a construc-

tion (Chapter 7), throwing a fresh light on the synchronic data.

As to diffusional change, the historical development of ing-complements

manifests a tension between emergent regularity and a scattering of apparently unre-

lated innovations. Like that of for...to-infinitival complements (Chapter 6), the diffu-

sion of gerundial ing-complements can be said, on the whole, to have followed a

path of least resistance, with ing-complements occurring first where they were most

straightforwardly sanctioned by the synchronic system. This is reflected in a gradual

increase in productivity of ing-complementation as a syntactic construction, and in

an overall shift in the kind of mechanisms that produce the innovations, with the

requirements for new verbs to be recruited as ing-complement-taking predicates

continually loosening up. At the same time, it is found that across time there is a

veritable hotchpotch of sources and factors that have determined the appearance and

306 – CHAPTER 9: The diffusion of ing-complements I

further development of ing-complements. It is also found that innovations may be

triggered or blocked locally rather than by the ‘system’ of ing-complementation as a

whole.

This tug of war between the ‘centralising’ force of emerging regularity and a

‘decentralising’ force introducing (or restraining) new ing-complements more or less

independently of the emerging system is highly relevant to our understanding of

complementation from a synchronic point of view. While over time regularity in the

use of gerundial ing-complements appears to have increased, there are historical

stages where the distribution of ing-complements, if considered from the point of

view of the regularity of ing-complementation as a whole, is fairly unsystematic. As

such, the diffusion of gerundial ing-complements in English lays bare the danger

inherent in thinking of the grammar of complementation as a tightly-organised sys-

tem of functionally motivated choices (a conclusion that is confirmed by the analysis

of integrated participle clauses in Chapter 10). This is not to say that use is not moti-

vated, only that it is not motivated by a simple and highly coherent underlying sys-

tem.

However, before we can begin to see how diffusion works and how its work-

ing bears on our understanding of complementation, we need to get the historical

facts straight. This in itself proves to be something of a challenge, as the landscape

of ing-complementation is marked not only by constant innovations over a period of

more than five centuries but also by ever shifting frequencies of use. Neither Visser

(1963-73) nor Fanego (1996a), though both commendable for their descriptive rig-

our, provide the detailed information actually needed. Visser (1963-73) gives a

wealth of data (rivalling the accuracy even the largest of today’s historical corpora

can offer), both for the period approximately corresponding to Present-Day English

and for earlier stages of the language, but the little quantitative information he gives

remains hard to interpret. Fanego (1996a), on the other hand, produces a detailed

overview of gerundial ing-complements between 1400 and 1760 and gives very pre-

cise information on frequencies, but her corpus is relatively small, especially so for

the period prior to 1570. Moreover, it is unfortunate that the focus of her study is

restricted to verbs where ing-complements compete with the to-infinitive (though

excluding the aspectuals) and that the period she describes only covers the beginning

of the diffusion of ing-complements. For these reasons, much of this chapter is in

fact devoted to the further description of the historical changes at issue.

The organisation of this chapter is such that it moves from description to in-

terpretation. Section 1 briefly details the specific methodology I have followed to

get at a sufficiently accurate picture of gerundial ing-complementation over time.

CHAPTER 9: The diffusion of ing-complements I – 307

Section 2 puts the methodology in practice and sets out to describe the historical

data, while Section 3 further refines the description and at the same time addresses

the question of causation. Through analyses of the history of ing-complementation

with different verbs, it is shown how diffusion proceeded and how this bears on is-

sues of complementation.

As usual, a preliminary note is in place on the precise demarcation of the ob-

ject of inquiry. As indicated above, the focus is on subject-controlled gerundial ing-

complements following simple verbal predicates. Gerundial ing-complements are

defined by the criterion used also in Chapter 7: ing-clauses that fill the slot of a di-

rect object to a transitive verb have been regarded as gerundial ing-complements.

The simple verbal predicates taking such gerundial ing-complements that are thus

considered here include a variety of regular transitive verbs, the aspectual verbs and

some phrasal verbs. Excluded is the very substantial group of ing-complements fol-

lowing predicates with a fixed preposition (e.g. the doctor objected to removing the

patient’s head) (on which see Rudanko 1998, 2000). Also excluded from the discus-

sion, except where relevant, are non-controlled ing-complements, which include

both generic ing-complements (on which, see Fanego 2007) and ing-complements

with their own explicit subject. The somewhat marginal class of ‘integrated partici-

ple clauses’ is also ignored for now but these are taken up in Chapter 10.

At the same time, some ing-complements have been included that, strictly

speaking, do not belong under the heading of gerundial ing-complements. In particu-

lar, I have opted to deal with the class of aspectual verbs (begin, cease, continue, go

on, keep, start, stop etc.) exhaustively here. The ing-complements found with the

majority of these verbs are clearly of gerundial origin (see especially leave, finish,

give up, stop). Others clearly derive from participial adjuncts and have been reinter-

preted as ing-complements in the context of specific verbs (see especially go on,

keep, keep on, remain); the eventual status of these ing-complements as either ger-

unds or participles is unclear, since they show none of the paradigmatic parallels

characteristic of either class of ing-clauses (see Chapter 7). Finally, with other as-

pectual verbs even the origin of ing-complements is uncertain: with cease and con-

tinue, in particular, ing-complements seem to translate a Latin participial construc-

tion, but they do fill a nominal slot in English, which means that they may in fact be

gerunds (then again, as cease and continue can be intransitive as well as transitive

this latter conclusion is not compelling). In all, however, it makes more sense to

treat all ing-complements with aspectual verbs in a single chapter than to treat them

separately, since the pattern seems to have developed into a productive construction

in its own right.


1. Methodology

Ideally, to complete the picture of change already available from Visser (1963-73)

and Fanego (1996a), we would have to have access to two kinds of information

relevant to diffusion. First, we would want to know exactly when ing-complements

appeared with which predicates for all predicates that ever took ing-complements.

Second, we would want to know whether and, if so, at which pace ing-complements

became established as regular complement-type with each predicate – that is,

whether some new combination remained merely exceptional, or if it developed into

a stable conventionalised chunk of grammar, how long this development took. This

ideal state of empirical knowledge is out of reach, but to approximate it, I have here

adopted the following two-step procedure.

As a first step, two ‘slices’ of historical usage were examined, covering the

periods from 1640 to 1710 and from 1850 to 1920. Instances of ing-

complementation were gathered from PPCEME for the period 1640-1710, while

data for the period 1850-1920 was collected from a sample from CLMETEV.89 Be-

cause both corpora were searched for all occurrences of word-final -ing, the various

predicates combining with ing-complements could be listed exhaustively for the data

examined. Once a list of predicates taking ing-complements had been established for

the two periods, the data were supplemented with additional material, using the

specified set of complement-taking verbs to search for other instances of ing-

complements in data from LC, CEMET (for 1640-1710) and the remaining material

from CLMETEV (for 1850-1920). This has yielded a fairly detailed picture of ing-

complementation in the two periods examined that is also to some extent representa-

tive of the progression of change. (The results of this examination are summarised in

Tables 9.1 and 9.2 in Section 2 below.)

As a second step, the predicates most frequently combining with ing-

complements were selected and taken as a starting point for further examination. A

predicate was selected if it occurred more than once per million words in either one

of the two periods examined. By applying this criterion, those predicates were se-

lected that can certainly be considered ‘good’ (i.e. non-aberrant and non-accidental)

examples of ing-complementation and that are sufficiently current in the data to

89 The relevant section of the PPCEME contains about 0.57 million words; the CLMETEV, whose relevant section contains about 6.25 million words, has been sampled at 10%. As a result, the two sets of historical data are approximately equal in size, allowing some degree of comparison and avoiding a bias towards one his-torical period.


make further investigation possible (in other respects, the criterion is purely arbi-

trary). The individual developments of these predicates were then traced throughout

the relevant sections of all the corpus material available (i.e. IMEPC, PPCME, HC,

PPCEME, LC, CEMET, POB, CLMETEV, CEN, OED, MED and various other

sources), with an eye to establishing as accurately as possible the first appearance of

ing-complements for any given predicate. The results of this second round of analy-

sis can be rendered in the form of a time-line of change, as given in Table 9.3 in

Section 3 below. At the same time, the close analyses of individual predicates also

provides an opportunity to move beyond the direct empirical facts of first occurrence

and unearth for every combination of predicate and ing-complement the most plau-

sible mechanisms and factors at work in its emergence and further development.

This brings to light the variety of sources, mechanisms and factors that interacted in

the diffusion of ing-complements, but also reveals the broad associations between

different mechanisms and different stages of the change.

2. Two synchronic slices: 1640-1710 and 1850-1920

As explained in Section 1, first two preliminary lists of complement-taking predi-

cates were compiled on the basis of an exhaustive search for ing-forms in PPCEME

(1640-1710) and a sample from CLMETEV (1850-1920). The predicates found

combining with controlled ing-complements are given in (i) and (ii). A question

mark following a predicate indicates that instantiation is doubtful (e.g. because the

only ing-complements attested are clearly nominal gerunds, or have an a-prefix, or

do not immediately depend on the complement-taking predicate, etc.; see further

below).

(i) 1640-1710: Avoid, bear, begin (?), cannot help, carry on (?), continue,

defer, delay, desist, escape, fall (?), fear, finish, forbear, forget, give

over (‘give up’), hate, intend, leave, leave off, love, miss, neglect, omit,

own (?), refrain, try (?), wait

(ii) 1850-1920: Admit, anticipate, avoid, bear, begin, cannot help, cease,

commence, contemplate, continue, decline, defer, delay, dislike, enjoy,

escape, fall, fancy, fear (?), finish, forbear (?), give up, go on, hate, in-

tend, keep, keep on, leave off, like, love, mind, need, prefer, prevent,

propose, purpose, put off, recollect, regret, remain, remember, require,

resume, risk, save, shun, start, stop, suggest, take up, try, want


Subsequently, the predicates in (i) and (ii) were examined in the full data sets for the

periods 1640-1710 and 1850-1920. The results of this are rendered in Tables 9.1 and

9.2 below, which also give the frequency of ing-complements with each predicate.

Note that the classification into subtypes in Tables 9.1 and 9.2 diverges

somewhat from that adopted in Chapter 8, as counts are no longer restricted to those

gerund constructions that encode their own (notional) object, and the primary inter-

est is no longer in the verbalisation of the gerund as such. The ing-complements that

are of primary interest here are of the kind illustrated in (2): they are clausal or can

be interpreted as such, and allow a reading whereby the complement clause is con-

trolled by the matrix clause subject. Two sub-types are distinguished: ‘verbal or am-

biguous’ ing-complements, as in (2a-b), show internal modification that is or can be

interpreted as clausal; ‘indeterminate’ ing-complements, as in (2c), show no modifi-

cation at all and their clausal status is therefore undecided.90 All of these ing-

complements are ‘bare’ (i.e. there is no determiner introducing the ing-clause).

(2) a. He was no sooner come to himself, than they continued pressing him

to know the occasion of his sighs and cries (1688, CEMET)

b. Up early, and to Petersfield, and there dined well; and thence got a

countryman to guide us by Havant, to avoid going through the Forest

(1662, CEMET)

c. Does not your Lordship love reading then? (1696, PPCEME)

In addition to the two types in (2), a number of other constructions are thrown up by

the data that are not the immediate target of investigation but that nonetheless carry

potential interest because of their obvious relatedness to the ing-complements under

consideration. One subset of these is a group of constructions whose internal syntax

contains unambiguous features of a noun phrase but still allow (pragmatically if not

90 Thus, clearly clausal ing-complements are counted along with ing-complements whose modifiers could be clausal as well as nominal. The reason for not splitting up these categories is that doing so adds little to the general picture, as the use of am-biguous ing-complements is parallel to that of unambiguously clausal ing-complements (the distinction between clausal/ambiguous and indeterminate, by con-trast, does have some further bearing – see esp. the discussion of love in Section 3.1). Note additionally that also included in this category are ing-complements con-taining a verb of saying and directly rendered speech as their direct object, ing-complements with a wh-extracted direct object, and some examples that could be either ambiguous or indeterminate (usually with adverbials that might belong with the matrix verb or with the verb in -ing: for example, Mr. Franklin and I remained waiting to see what might happen (1868, CLMETEV)).


syntactically) a controlled interpretation. Of these, ‘definite nominal’ ing-

complements are preceded by a definite article and their internal modifiers are ex-

pressed by nominal means (e.g. an of-phrase to express a patient argument), as in

(3a). ‘Hybrid’ ing-complements also have a definite article although otherwise their

ing-form behaves as a verb in taking verbal modifiers, as shown in (3b). ‘Bare

nominal’ ing-complements again show explicitly nominal behaviour but lack the

definite article, as in (3c).

(3) a. when he had finished the reading of his first prayers he could see that

he had lived down some of the derision due to his adventure with the

old woman. (1897, CLMETEV)

b. his wrath revenged him, and he feared the being stripped of it (1895,

CLMETEV)

c. Go on wasting of our blood and treasure (1653, LC)

A second subset consists of constructions with an ing-constituent that expresses its

own (notional) subject. Of these, ing-complements with a ‘repeated subject’ are in-

troduced by a possessive that is co-referential with the subject of the main verb, as in

(4a); their internal syntax appears to be nearly always nominal. Ing-complements

with a ‘different subject’ have their own possessive or oblique subject that is not co-

referential with the subject of the main verb, as in (4b); their internal syntax is most

commonly clausal.

(4) a. with no thought of anything serious having happened to the ship, I

continued my reading (1912, CLMETEV)

b. Wholly regardless even of heaven and man, he made a public

confession of his passion, denied her being married to Brilliard, and

weeps as he protests her innocence (1684, CEMET)

The last category of ing-complements is illustrated in (5) and is characterised by an

a-prefix introducing the ing-form.

(5) A jackdaw isn’t to be called a religious bird because it keeps a-cawing

on the steeple (1897, CLMETEV)

In addition to the various distinctions between types of ing-complement in

Tables 9.1 and 9.2, some classificatory difficulties and finer contrasts still need


pointing out. To begin with, a few words are in place concerning the notion of sub-

ject-control. Subject-control has been understood here quite loosely as a pragmatic

relation of identity or near-identity (inclusion)91 between an explicit matrix clause

subject and an implicit participant of the situation denoted by the ing-complement.

This has a number of implications, one of which is that purely nominal construc-

tions, too, can be controlled.92 To illustrate this, controlled definite nominal ing-

complements are exemplified in (6a-b) and contrasted with non-controlled examples

in (6c-d) (see also the controlled examples in (3) above).

(6) a. Notwithstanding the seeming Rationality of Achitophel’s Advice,

Absalom declines the taking of it (1696, LC)

b. and for the Instructions themselves, my Lord Ranelagh doth owne ye

drawing of them, (1675-6, PPCEME)

c. Why Ermine! you could not bear the jarring of that crutch (1865,

CLMETEV)

d. She foresaw an immense family fuss, endless tomfoolery, the upsetting

of her existence, the destruction of her calm. (1908, CLMETEV)

It is also assumed that subject-control may provide the active subject of the com-

plement clause – in the canonical case – or its object/passive subject. The latter op-

tion, illustrated in (7a-b), is found most commonly with the verbs need, require and

want, but it is attested occasionally with the verbs bear, decline, escape, fear, and

prevent as well. I will refer to ing-complements of this type as ‘passival’ (rather than

‘passive’ as they do not syntactically encode passive voice). An analogous phe-

nomenon shows up in repeated subject constructions, where the possessive introduc-

ing the ing-complement is at once co-referential with the subject of the matrix clause

and provides the (notional) object of the complement clause, as illustrated in (7c-d).

91 Near-identity between matrix subject and implied subject of the complement clause arises when the subject of the complement clause includes but does not coin-cide with the subject of the matrix clause (see e.g. propose and suggest). 92 It is assumed here that control is not necessarily a syntactic relationship, involving PRO or a deleted constituent in the complement clause. Control also occurs in a pragmatic form, where it is simply an implied relationship of identity between the matrix clause subject and a ‘participant’ present (if necessarily backgrounded) in the semantic structure of a noun (e.g. he never came to actually enjoy detention). No doubt this is also part of the reason why some nominal constructions can neverthe-less strongly require a controlled interpretation (see the discussion of bare nominal gerunds in Chapter 8).


(7) a. Such a character and such conditions as his required nicer handling

than hers. (1885, CLMETEV)

b. when the poor Debitor feeling the Rope about his neck, expresses an

unwillingnesse to be made thus accessory to lying and his owne ruine,

it is told him, the Chancery will forgive and pardon him; whereupon to

prevent strangling at that instant, he sets his hand and seale, and gets a

Reprieve for six moneths longer, or some such breathing time (1653,

LC)

c. the maides that could scant stand for laughing, at last askt how hee

liked his washing? (1619, PPCEME)

d. Well, those who mean to escape their catching must get ready. (1897,

CLMETEV)

Further, the data contain many instances allowing both a controlled and a

non-controlled reading. Because the two readings are typically not mutually exclu-

sive, and because control is at least implicated, instances of this kind have been re-

garded as controlled. By way of example, consider the complements to the verb

avoid in (8):

(8) a. it was impossible for him to avoid any longer giving them battle (1684,

CEMET)

b. they must return home without delay if he was to avoid getting frozen

in for the winter. (1902, CLMETEV)

The ing-complements giving them battle and getting frozen in for the winter may be

construed generically as a kind of situation that the matrix subject wishes to avoid,

or specifically, as a concrete (if as yet unactualised) instance of that general situa-

tion. On the generic reading, the ing-complement is strictly-speaking non-controlled,

since what is referred to is a type of event that need not specify its own subject

(Heyvaert 2006). At the same time, the verb avoid projects another world of dis-

course in which the avoided situation occurs with respect to the matrix subject. At

this level of interpretation, the ing-complement is non-generic and can be regarded

as controlled. It is only where this projection is missing, as in the examples in (9),

that ing-complements can be regarded as purely generic and non-controlled.

(9) a. And for preventing clandestinely carrying out of the said Ware-houses

any of the said Goods hereby prohibited and by this Act intended for


Exportation as aforesaid Be it further enacted by the Authority afore-

said [...] (1698-9, PPCEME)

b. No! my backsliding friend, there is but one way back to peace and joy

and usefulness, and that is your Lord’s own way – repentance, which

always implies forsaking sin (1879, CLMETEV)

Notice, finally, that in some cases an additional interpretative dimension is

involved, on top of the ambiguity between a generic and a non-generic reading. Es-

pecially verbs such as propose or suggest can occur with controlled ing-

complements in which the matrix subject may be either identical to or merely in-

cluded in the agent implied by the complement clause, as in (10a-b) respectively. In

yet other cases (not counted in Tables 9.1 and 9.2) these verbs project an action car-

ried out by a third party not including the matrix subject, as in (10c). For all these

examples, notice that a generic reading is not unacceptable (especially so in (10c)).

(10) a. he accordingly proposed shifting his inquiries to the jugglers in the

prison at Frizinghall. (1868, CLMETEV)

b. So I propose confining our attention to the elementary rules. (1909,

CLMETEV)

c. [...] Wise entertained no doubt as to the practicability of such a ma-

chine. For its inflation he suggests inserting a muslin balloon filled

with air within the copper globe, and then passing hydrogen gas be-

tween the muslin and copper surfaces, which would exclude the inner

balloon as the copper one filled up. (1902, CLMETEV)

Next to these distinctions relating to subject-control, Tables 9.1 and 9.2 hide

some classificatory shortcuts that should be pointed out. First, the passival ing-

complements following the verbs need, require and want can take the form of inde-

terminate ing-complements, as in (11a), or of bare ing-complements, as in (11b).

However, these three verbs also productively combine with nominal ing-

complements introduced by an indefinite determiner or quantifier such as a, a little,

any, some or no, as in (11c). For need, as for require and want, examples of this type

have been classified as bare nominal ing-complements, even though they are not in

actual fact ‘bare’ (i.e. determinerless).

(11) a. My fire would not need replenishing for an hour or so. (1888,

CLMETEV)

ING-COMPLEMENT Subject-controlled

Definite Bare Own subject

VERB

Definite nominal

Hybrid Bare nomi-

nal Indetermi-

nate Clausal /

ambiguous A-prefix

Repeated subject

Different subject

Admit 1 (0,3) Anticipate Avoid 2 (0,6) 2 (0,6) 3 (0,9) 1 (0,3) 35 (10,4) 1 (0,3) Bear 1 (0,3) 1 (0,3) 1 (0,3) Begin 3 (0,9) 2 (0,6) 1 (0,3) 1 (0,3) Carry on 1 (0,3) Cease 4 (1,2) Commence Contemplate 1 (0,3) Continue 1 (0,3) 16 (4,8) 14 (4,2) 4 (1,2) Decline 1 (0,3) 3 (0,9) 1 (0,3) 2 (0,6) Defer 2 (0,6) 1 (0,3) 2 (0,6) 3 (0,9) 2 (0,6) 2 (0,6) Delay 1 (0,3) 1 (0,3) 1 (0,3) 2 (0,6) 1 (0,3) Deny 4 (1,2) 2 (0,6) 1 (0,3) 1 (0,3) Desist 1 (0,3) Dislike 1 (0,3) Enjoy Escape 2 (0,6) 4 (1,2) Fall 1 (0,3) 1 (0,3) 20 (6,0) Fancy

Fear 1 (0,3) 1 (0,3) 4 (1,2) Finish 2 (0,6) 1 (0,3) Forbear 1 (0,3) 1 (0,3) 16 (4,8) 34 (10,1) Forget 1 (0,3) Give off 2 (0,6) Give over 2 (0,6) 5 (1,5) Give up 1 (0,3) Go on 1 (0,3) 2 (0,6) 3 (0,9) Hate 1 (0,3) 2 (0,6) Help 1 (0,3) 10 (3,0) 1 (0,3) Hold 1 (0,3) 2 (0,6) Intend 1 (0,3) 1 (0,3) Keep 2 (0,6) 1 (0,3) 16 (4,8) Keep on Leave 1 (0,3) 9 (2,7) 6 (1,8) 2 (0,6) Leave off 1 (0,3) 3 (0,9) 4 (1,2) 2 (0,6) Like 1 (0,3) 1 (0,3) 1 (0,3) Love 15 (4,5) 2 (0,6) Mind 1 (0,3) 1 (0,3) Miss 2 (0,6) Need 1 (0,3) 1 (0,3) Neglect 1 (0,3) Omit 2 (0,6) 1 (0,3) 8 (2,4) Own 1 (0,3) 1 (0,3) 1 (0,3) Prefer 1 (0,3) 1 (0,3) 1 (0,3) 1 (0,3)

Prevent 1 (0,3) 4 (1,2) 48 (14,3) Propose 2 (0,6) 1 (0,3) Purpose Put off 1 (0,3) 1 (0,3) Recollect Refrain 1 (0,3) 2 (0,6) Regret 1 (0,3) Remain 2 (0,6) 1 (0,3) Remember 1 (0,3) 1 (0,3) Require 1 (0,3) Resume 1 (0,3) Risk Save 1 (0,3) 1 (0,3) Shun Start Stop 1 (0,3) 4 (1,2) Suggest Take up Try 1 (0,3) 1 (0,3) Wait / a-wait 1 (0,3) 8 (2,4) Want 2 (0,6) Table 9.1. Ing-complementation in 1640-1710 in PPCEME, CEMET and LC (frequencies per million words given in brackets).

ING-COMPLEMENT

Subject-controlled Definite Bare

Own subject

VERB

Definite nominal

Hybrid Bare nomi-

nal Indetermi-

nate Clausal /

ambiguous A-prefix

Repeated subject

Different subject

Admit 1 (0,2) 2 (0,3) Anticipate 1 (0,2) 1 (0,2) Avoid 1 (0,2) 6 (1,0) 71 (11,4) Bear 10 (1,6) 2 (0,3) Begin 3 (0,5) 43 (6,9) 246 (39,4) 1 (0,2) 5 (0,8) Carry on Cease 15 (2,4) 20 (3,2) 4 (0,6) Commence 8 (1,3) 21 (3,4) Contemplate 10 (1,6) 2 (0,3) Continue 13 (2,1) 21 (3,4) 3 (0,5) Decline 9 (1,4) Defer 4 (0,6) Delay 1 (0,2) 5 (0,8) 1 (0,2) Deny 2 (0,3) 1 (0,2) Desist Dislike 1 (0,2) 7 (1,1) 1 (0,2) Enjoy 1 (0,2) 1 (0,2) 26 (4,2) 1 (0,2) Escape 2 (0,3) 13 (2,1) 1 (0,2) Fall 1 (0,2) 1 (0,2)

Fancy 1 (0,2) 24 (3,8) 16 (2,6) Fear 1 (0,2) 3 (0,5) 2 (0,3) Finish 1 (0,2) 22 (3,5) 19 (3,0) 7 (1,1) Forbear 1 (0,2) 1 (0,2) 6 (1,0) Forget 2 (0,3) 2 (0,3) Give off Give over 2 (0,3) Give up 13 (2,1) 24 (3,8) 1 (0,2) Go on 1 (0,2) 37 (5,9) 117 (18,7) Hate 1 (0,2) 6 (1,0) 21 (3,4) 1 (0,2) Help 15 (2,4) 314 (50,2) 5 (0,8) Hold 1 (0,2) Intend 12 (1,9) Keep 7 (1,2) 185 (29,6) 1 (0,2) 93 (14,8 Keep on 8 (1,3) 34 (5,4) Leave Leave off 24 (3,8) 38 (6,1) Like 11 (1,8) 79 (12,6) 9 (1,4) Love 2 (0,3) 12 (1,9) Mind 7 (1,1) 83 (13,3) 27 (4,3) Miss 15 (2,4) Need 20 (3,2) 13 (2,1) Neglect Omit Own Prefer 1 (0,2) 2 (0,3) 29 (4,6) 2 (0,3)

Prevent 1 (0,2) 1 (0,2) 131 (21,0) Propose 13 (2,1) Purpose 4 (0,6) Put off 1 (0,2) Recollect 6 (1,0) 4 (0,6) Refrain Regret 1 (0,2) 16 (2,6) 2 (0,3) Remain 13 (2,1) 20 (3,2) Remember 118 (18,9) 34 (5,4) Require 1 (0,2) 6 (1,0) 10 (1,6) 1 (0,2) Resume 2 (0,3) 1 (0,2) 5 (0,8) Risk 1 (0,2) 1 (0,2) 16 (2,6) 1 (0,2) Save 1 (0,2) Shun 1 (0,2) Start 2 (0,3) 10 (1,6) 18 (2,9) 2 (0,3) Stop 2 (0,3) 50 (8,0) 20 (3,2) 4 (0,6) 13 (2,1) Suggest 1 (0,2) 7 (1,2) 8 (1,3) Take up 1 (0,2) 5 (0,8) 2 (0,3) Try 4 (0,6) 13 (2,1) Wait / a-wait 8 (1,3) Want 10 (1,6) 34 (5,4) 1 (0,2) 2 (0,3) Table 9.2. Ing-complementation in 1850-1920 in CLMETEV (frequencies per million words given in brackets).


b. “The story will need good telling,” admitted Sapt. (1898, CLMETEV)

c. borne along by our passionate desire, we needed no convincing at all.

(1898, CLMETEV)

Second, the different-subject complements following keep and, more marginally,

continue and start do not strictly speaking have their ‘own’ subject. These verbs

occur in causative constructions with a participial complement, where what might

look like the subject of the ing-complement is actually the object of the matrix verb,

as appears from the fact that it can be fronted to subject position under passivisation,

as shown for keep in (12a-b). A similar causative participial construction is respon-

sible for the different-subject constructions found with want. The different-subject

constructions following prevent, like and stop, finally, might also occasionally resort

under this pattern (cf. Jespersen 1940: 142-50): although their different-subject com-

plements are more canonically analysed as one-constituent gerundial ing-

complements, at least for prevent there is clear evidence that the alternative analysis

as a causative construction is also valid – consider (12c) (see Chapter 7).

(12) a. Keep him not waiting for his love too long. (1678, CEMET)

b. people were coming and going constantly, and the doorkeeper was

kept opening and closing the door. (1897, CLMETEV)

c. It may seem surprising that the balloon, which could not be prevented

falling in the water, is yet enabled to ascend from the grip of the waves

by the mere discharge of ballast. (1902, CLMETEV)

If anything, Tables 9.1 and 9.2 clearly confirm the diffusional nature of the

emergence of ing-complements. Some predicates were already commonly or fre-

quently used with ing-complements in the period 1640-1710 and have remained so

until the period 1850-1920, or have even declined somewhat again. Among these are

avoid, continue, defer, escape, fear, forbear, give over, leave, love, omit and prevent.

Other predicates are attested with ing-complements in 1640-1710 but were probably

still on the increase at this time, as is suggested by the fact that the frequency of ing-

complements with these predicates is higher in 1850-1920. These predicates include

begin, cease, decline, finish, give up, go on, hate, help, intend, keep, leave off, like,

miss, need, prefer, remain, stop, and want. Other predicates still do not combine

with ing-complements at all in the period 1640-1710, yet ing-complements have

appeared by the period 1850-1920. Among them are bear, commence, contemplate,

dislike, enjoy, fancy, keep on, mind, propose, regret, remember, require, risk, start,


suggest, and try. While this is only an approximation of the actual course of diffu-

sion, there can be no doubt as to the fact that ing-complements did not appear simul-

taneously with all predicates.

Some further remarks can be made regarding the data in Tables 9.1 and 9.2.

First, the data show a decrease in the number of explicitly nominal complement con-

structions. Thus, excluding the group of passival ing-complements following need,

require and want, bare nominal ing-complements shrink back from approximately 3

instances per million words in 1640-1710 to 0.3 in 1850-1920; the joint frequency of

definite nominal and hybrid ing-complements reduces from a total of 9.5 instances

per million words for 1640-1710 to 3.4 in 1850-1920. The decrease in ing-

complements with explicitly nominal syntax is of course in line with the loss of bare

nominal gerunds (see Chapter 8), but in as far as it also affects definite nominal ger-

unds, it suggests a reversal of the trend for definite nominal gerunds to grow more

frequent prior to 1640 (see again Chapter 8). Still, although this is not immediately

apparent from the data in Tables 9.1 and 9.2, the behaviour of definite gerunds is

otherwise in line with the functional characterisation of definite gerund construc-

tions, in that the majority of instances are found to be non-controlled, as in (5c-d)

above, as opposed to verbal gerunds (cf. De Smet forthc. a). In this light, it is plausi-

ble that definite gerunds held their ground where they maximally differed from ver-

bal gerunds, while the definite gerunds with (implied) control were more vulnerable

to competition. Indeed, the discussion in Section 3 below indicates that from the

second half of the seventeenth century onwards verbal gerunds also began to replace

definite nominal gerunds.

Second, a number of lexico-grammatical patterns can be discerned in the data

with relative ease. Most clearly, the three verbs strongly associated with passival

control (need, want, require) cling on to complements with nominal internal syntax.

Further, the repeated-subject construction is primarily associated with aspectual

verbs, especially in the data from 1850-1920. Likewise, the use of ing-complements

with a-prefix is (if anything) characteristic of aspectuals. Finally, the group of nega-

tive implicative verbs stands out early on as a semantically coherent group of verbs

taking ing-complements (cf. Fanego 1996a, 2004a, 2007). Each of these observa-

tions gives evidence of the relevance of semantic patterning. In the discussion in

Section 3 below, regularities of this kind are frequently invoked to corroborate

claims about semantic analogy and semantic relatedness between predicates.


3. Innovations and their causes

We may now turn to individual predicates and investigate their history more closely,

in order to break down diffusion into a chain of innovations. As indicated above

(Section 1), the predicates selected for this purpose are those that are firmly attested

with ing-complements (i.e. occurring more than once per million words) in one or

both of the two periods represented in Tables 9.1 and 9.2 above. The predicates an-

swering this criterion are given in (iii).

(iii) Avoid, bear, begin, busy, cease, commence, contemplate, continue,

decline, defer, dislike, done, enjoy, escape, fancy, fear, finish, forbear,

give over, give up, go on, hate, help, intend, keep, keep on, leave, leave

off, like, love, mind, miss, need, omit, prefer, prevent, propose, regret,

remain, remember, require, risk, start, stop, suggest, try, want

In what follows, the predicates in (iii) will be discussed one by one, in the order in

which they began to combine with ing-complements. The analyses are carried out

with an eye to obtaining two pieces of information, one quite empirical, the other

more speculative. In the venerable tradition of first-quotation-hunting, the first piece

of information looked for is the earliest appearance of a predicate with an ing-

complement. The second is the possible causal factors leading to the appearance of

ing-complements in a new environment. With respect to the latter, a ‘most reason-

able guess’ is in many cases the best any individual analysis can arrive at, but as will

appear, with the accumulation of best guesses reliability increases and the resultant

picture of diffusion turns out a relatively coherent one. The goal, eventually, is to

show that the mechanisms of diffusion subtly shift in the course of diffusion.

The major mechanisms involved are familiar from preceding chapters, con-

sisting in semantic and paradigmatic analogy and blocking (see especially Chapter

4). Nevertheless, a few words are in place on their specific application here. With

respect to semantic analogy, Fanego (1996a; forthc.) has appealed to this mechanism

to account for the early association of ing-complements with negative implicative

verbs (such as forbear, avoid, decline etc.), and also to explain the appearance of

ing-complements with the members of other groups of semantically related verbs

(emotive verbs, retrospective verbs). Obviously, as a mechanism of change, seman-

tic analogy is the most likely diachronic correlate of the clusters of semantically

related verbs attested in the present-day distribution of ing-complements (see Chap-

ter 7).


Paradigmatic analogy, on the other hand, links back to the historical sources

of gerundial ing-complements and their distributional associations with noun

phrases. From a diachronic point of view, paradigmatic analogy can be interpreted

weakly as a tendency for gerundial ing-complements to extend to transitive verbs

(see Chapter 7). However, paradigmatic analogy also allows for a stronger interpre-

tation. To see how this works, we need to consider the paradigmatic relationship

between noun phrases and gerundial ing-complements somewhat more closely. The

similarity between gerundial ing-complements and noun phrases varies according to

different parameters, with ing-complements resembling different noun types to dif-

ferent degrees. For one thing, the fact that, historically, gerundial action nominals

were noncount nouns still translates in a closer similarity between gerundial ing-

complements and mass nouns or bare plural nouns. For another, the fact that gerun-

dial ing-complements necessarily refer to an event, action or situation makes them

lean closer to nouns either denoting or implying an event, action or situation than to

nouns denoting beings or things, especially so if the action referred to by a noun is

also (pragmatically) controlled by the subject of the matrix clause (see footnote 92

above). As a result, a cline of similarity arises, as is illustrated in (13), with (13a)

and (13e) as outer poles:

(13) a. he likes playing practical jokes (CB)

b. he likes outdoor activities

c. he likes television

d. he likes dogs

e. he likes his dog

Although it is difficult to operationalise the contrasts between different noun types

in corpus data (especially because many nouns weakly imply some action, as in

(13c), where television is suggestive of watching), it appears that the similarities

between gerundial ing-complements and other noun types played an important role

in the development of ing-complementation. Because verbs collocate with different

noun types, they paradigmatically attract gerundial ing-complements to different

degrees.

For that reason, the discussion below distinguishes between extension

through ‘narrow’ paradigmatic analogy (based on the occurrence of bare noncount

nouns denoting actions and their association with bare gerundial ing-complements)

and extension through ‘broad’ paradigmatic analogy (based on the association be-

tween gerundial ing-complements and other noun types). Further, a special subtype


of broad paradigmatic analogy is distinguished, termed here ‘indirect’ paradigmatic

analogy. As Tables 9.1 and 9.2 show, there are certain alternation patterns that are

fairly systematic. For example, where different-subject complements occur we also

find controlled ing-complements. Consequently, these alternations too could give

rise to innovations through paradigmatic analogy – for instance, environments with

different-subject complements could begin to select controlled complements (which

is what seems to have happened in the histories of mind and remember; see below).

The term ‘indirect’ paradigmatic analogy will here be used to refer to such exten-

sions. Its motivation lies in the fact that the first member of the alternating pair may

itself have appeared as a result of paradigmatic analogy, so that the appearance of

bare controlled ing-complements is based on a chain of paradigmatic analogies,

rather than being arrived at directly through a single mechanism.

Finally, note that paradigmatic analogy has also affected the small group of

aspectual verbs whose ing-complements are not of gerundial but of participial ori-

gin. In these cases, paradigmatic analogy is based on the distributional tie between

participles and adjectives/adverbials (see Chapter 7). Thus, paradigmatic analogy

can introduce participial ing-complements to verbs that are used copulatively and

sanction adjectival subject-complements (see e.g. remain, keep). Alternatively, para-

digmatic analogy can introduce adverbial participles with intransitive verbs that of-

ten take adverbials (such as go on or keep on), which adverbial participles can sub-

sequently be reinterpreted as complement clauses (cf. the emergence of integrated

participle clauses discussed in Chapter 10).

In addition to semantic analogy, paradigmatic analogy and blocking, the dif-

fusion of ing-complements is also governed by other mechanisms and factors, such

as borrowing from French or Latin or the reinterpretation of participial adjuncts as

complements with some aspectuals. A mechanism that is entirely specific to ing-

complementation involves the loss of the a-prefix in constructions of the type illus-

trated in (14), either as a result of phonetic reduction or of a combination of para-

digmatic analogy and competition.93

93 Since the majority of ing-forms with a-prefix occur in environments that also li-cense a gerund clause preceded by a preposition (mostly in or on), it is plausible that the use of the a-prefix results from phonetic reduction of the preposition and that its consequent loss is simply a further step in this development. However, the historical data also throw up a number of environments where ing-complements with a-prefix occurred side by side with ing-complements without a-prefix (for example, with causative set). It is therefore possible that a paradigmatic relationship between the two variants was recognised and extended through paradigmatic analogy with regu-lar ing-complements ultimately outcompeting the ing-forms with a-prefix.


(14) A married Gentleman coming through Canterbury, his Horse threw

him, which a young Gentlewoman seeing, fell a laughing (1684-7,

PPCEME)

A final but crucial factor in the diffusion of ing-complements is presented by lexical

changes. Changes in the lexical semantics of a predicate may cause that predicate to

engage in new syntactic constructions, including ing-complementation. Such lexical

changes are not directly related to the diffusion of ing-complements, but they play

an important part in explaining the course of diffusion of ing-complements. Most

importantly, they can account for the timing of certain innovations. For example, the

meaning of the verb avoid changed from ‘empty’ to ‘shun, escape’. Only in the latter

sense is avoid compatible with clausal complementation, so the timing of this se-

mantic development is crucial also to the timing of the appearance of ing-

complements with avoid and thereby to the diffusion of ing-complements in general

(see the discussion of avoid below). Because lexical changes have the potential of

suddenly dropping new lexical items into the range of the new and spreading con-

struction, they add a measure of unpredictability to diffusional change. Moreover,

note that lexical items are not only subject to semantic change but also to consider-

able (and often unaccountable) fluctuations in terms of discourse frequency, which

further complicates the history of the syntactic constructions with which they com-

bine, including that of ing-complements.

As a final practical remark, historical analyses of individual verbs are pre-

sented here in some detail. Because the early attestations (either of new senses of

verbs or new collocations) sometimes prove hard to find and are often not recorded

by the standard works of reference (most prominently Visser 1963-73 or the OED),

they have some documentary value and I have for that reason not omitted quoting

abundantly. Moreover, the individual histories of verbs are the key to understanding

the relevance of very local developments to the macro-change that is diffusion. As a

disadvantage, amidst the discussions of lexicological intricacies it sometimes be-

comes easy to lose sight of the bigger patterns of change. The time-line and sum-

mary of the separate analyses that is given in Table 9.3 at the end of this section is

meant to remedy this and can at once serve as an index to the individual analyses.

Likewise, the division of the process of diffusion into stages is in part rhetorical, as a

device to break up the flow of analyses into manageable segments. That is not to say

that there are no differences between the stages, but as such differences are (I be-


lieve) continuous rather than discrete, segmentation is not intended to reflect any

fundamental break in the process of diffusion.

3.1. Stage I: narrow paradigmatic analogy

Stage I in the diffusion of ing-complements coincides with the Middle English pe-

riod. During this period, ing-complements are attested (in order of historical appear-

ance) with the verbs love, begin, continue, cease, hate, need, leave, forbear and es-

cape. The ing-complements that come in use in this period are probably of a nomi-

nal nature; more precisely, they are typically syntactically indeterminate, but when

they show internal modification it tends to be of the nominal type. However, since

bare nominal gerunds are functionally very similar to verbal gerunds (cf. Chapter 8),

I see no strong grounds for treating them as anything other than ing-complements,

especially as they are usually indistinguishable from the present-day patterns of ing-

complementation with the same verbs.

The mechanism characteristic of this stage of diffusion is narrow paradig-

matic analogy. Many of the verbs ing-complements appear with in Middle English

naturally collocate94 with bare abstract nouns denoting or implying an action, event

or situation controlled by the matrix subject. Consequently, bare ing-complements

can be said to behave and distribute just like other similar noun phrases. Sanctioning

is thus extremely strong; indeed, given certain verbs’ collocational preferences there

seems to be nothing remarkable about their use with ing-complements. This applies

very clearly to the verbs love, hate, need, leave, forbear and escape, and the same

probably holds for begin, though the evidence here is less clear. The verb that most

clearly resists an account in terms of paradigmatic analogy is cease, though with

cease there is evidence of borrowing from Latin, along with the possibility of inter-

ference with the prepositional verb cease of and, possibly, indirect paradigmatic

analogy. The use of ing-complements with continue, finally, appears to be the result

of borrowing as well.

94 The notion of collocation is used here in a loose sense. Ing-complements are said to fit the collocational behaviour of a given predicate if they are not at odds with the other objects selected by that predicate. Quantifying this kind of observation is ex-tremely difficult, however, since there are different points of resemblance (i.e. for-mal and semantic) between ing-complements and other object noun phrases and some (especially semantic) resemblances are themselves gradational (see also the introduction to Section 3). Furthermore, polysemy in the verbal predicates would seriously complicate any count.


In all, there are no straightforward signs that ing-complementation is begin-

ning to behave as a construction in Middle English. Nothing sets ing-complements

clearly apart from other action nominals occurring in direct object position; nor does

anything set specific combinations of a predicate and ing-complement clearly apart

from the transitive construction. From this point of view, we cannot say that ing-

complementation is already diffusing as a construction (which is not to say that no

changes are taking place). At the same time, it is conceivable that specific predicate-

complement combinations came to be stored and employed as very concrete gram-

matical patterns. What is certainly clear is that the patterns first showing up in Mid-

dle English remained in existence and as such determined the starting point of vari-

ous analogical extensions when ing-complementation began to gain some independ-

ent productivity in the Early Modern period, as will appear below. Note further that

the different sources for ing-complements in this period bring in from the start a

measure of arbitrariness – not in the sense that particular predicate-complement

combinations are unmotivated, but with respect to ing-complementation as a whole,

the distribution of ing-complements being determined by very different factors from

the very start.

LOVE: The verb love is very early in combining with controlled ing-complements,

albeit that all are of the syntactically indeterminate type, as illustrated in (15). The

earliest ing-complements tend to denote a sin or virtue, or by extension, an activity

that the matrix subject typically indulges in. As such, the use of ing-complements is

firmly embedded in a more general collocational pattern consisting of the verb love

in combination with all kinds of abstract and action nouns denoting the sins and vir-

tues characteristic of the matrix clause subject – see (16). Incidentally, as (15a)

shows, this collocational connection is supported by occasional instances of coordi-

nation between an ing-complement and an ordinary noun. In other words, the ap-

pearance of gerundial ing-complements with love is a straightforward case of narrow

paradigmatic analogy.

(15) a. Þe luueden tening & stale, hordom & drunken. (a1225 (?c1175),

MED)

‘(you) that loved doing harm and theft and whoring and drunkenness.’

b. And he loued waryynge (c1350, PPCME)

c. and halde þe in chastite, and iuil langingis do away; luue fasting;

(a1425, PPCME)

(16) a. þa þe luueden unriht and ufel lif leden. (a1225 (?c1175), MED)


b. Hue loued so lecherie & lustes of synne, Þat her chylder hue chase

unchastly to haue (1340-70, OED).

c. He loued malloc [i.e. ‘foul language, cursing’] dai and nighte. (a1400,

MED)

Despite the early appearance of indeterminate ing-complements, it took a relatively

long time for love, compared to other predicates, to select unmistakably clausal ing-

complements, as already pointed out by Visser (1963-73: 1866) and Fanego (1996a).

The pun in (17a), hinging on the ambiguity of out of measure as a manner or degree

adverbial, is on one of its readings the earliest such instance in the data; yet the fol-

lowing example attested, given in (17b), is of much later date, and probably better

signals the point at which love abandons its reluctance to combine with clausal ing-

complements. As a reason for the late appearance of clausal ing-complements,

Fanego (1996a) suggests the blocking effect of the to-infinitive, which is indeed

plausible, seeing that to-infinitives are fairly frequent with love throughout the Mid-

dle and Early Modern English periods (and came semantically very close to the con-

struction with ing-complement – see De Smet & Cuyckens 2005).

(17) a. Yea for I loue singyng out of measure, It comforteth my spirites and

doth me great pleasure. (1566 (?1552), PPCEME)

b. Dost thou love picking meat? (1678, CEMET)

BEGIN: The history of ing-complements following begin is marked by a long period

of marginal occurrence. Visser (1963-73: 1890) adduces a few very early and uncon-

troversial-looking Middle English examples (see e.g. (18a)) but an attempt at repli-

cating his findings yields only a handful of additional instances (see (18b-d)) indi-

cating that the pattern occurred in Middle English but that it was certainly not cur-

rent, especially in view of the high frequency of the verb begin as such. This state of

affairs continues into the Modern Period. The sixteenth century yields no examples,

except the two quoted by Visser (1963-73), one of which is somewhat suspect, as its

ing-complement is clearly nominal and precedes the matrix verb. For the seven-

teenth century, Mair (2002) reports one fairly reliable-looking example dated 1614,

found in the OED quotation data base. To this may be added the two examples from

CEMET in (19a-b), the first of which, however, quite unaccountably shows passival

control. It is not until the second half of the eighteenth century that the period of

irregular occurrence at last comes to an end and ing-complements start to be used

with begin on a fairly regular basis, witness the mid-eighteenth-century examples


from POB in (19c-e) (still, Visser’s 1963-73: 1890 first instance postdating the six-

teenth century is from 1813). In striking contrast to its hesitant start, however, the

construction has become very common in Present-Day English, with over 40 in-

stances per million words in CB, and its frequency still on the increase, especially in

American English (Mair 2002).

(18) a. he sone bi-com i to Brutlonde & anan bigon ræuinge (c1275 (?a1200),

Visser 1963-73: 1890)

‘he arrived in Britain without delay and immediately began

plundering’

b. Vmben ane stunde; heo bigunnen striuinge. al-se hit wes auer laȝe;

imong childrene plæȝe. (c1275 (?a1200), HC)

‘After a while they began quarrelling, just as it has always been in

children’s play.’

c. In euel tyme bigan she tomblyng To make his heed of be brouȝt.

(a1400, MED)

‘Unluckily, se began dancing acrobatically, to have his head cut off

and brought to her’

d. that day ys gode to begynne edyfying of placys, and to begynne alle

odyr werkys (c1450, HC)

(19) a. Dined alone at home, and was glad my house is begun tiling (1662,

CEMET)

b. Faxiondono growing more arrogant by this moderation of his prince,

began raising his note by extolling the profession of a Bonza (1688,

CEMET)

c. I believe, if these unhappy Men had been told, when they first began

Smuggling, that the Time would come when they would cooly bathe

their Hands in the Blood of two innocent Men, (bad as they now are,)

they would then have been shocked, and startled at the very

Imagination of it (1748, POB)

d. accordingly taking him out of the Chair, they began searching his

Pockets, where they found two half Crowns, four Shillings, and about

three Pennworth of Halfpence, but no Watch (1750, POB)

e. they came up with me, and began pulling me about, saying, I should go

along with this and the other. (1750, POB)


This course of development makes begin the verb that can lay claim to the longest

interval between the initial stage of sporadic occurrence with ing-complements (dat-

ing back to the Early Middle English period) and the rise of the pattern to full pro-

ductivity (starting certainly not earlier than halfway the eighteenth century). It is

plausible that the to-infinitive, which has been frequent with begin from the start,

again has to be held responsible for this hampered development.

The most plausible mechanism that could have led to the first ing-

complements following begin is narrow paradigmatic analogy. On this account, the

use of ing-complements with begin is the result of collocational extension from bare

abstract nouns implying an action controlled by the matrix subject. Example (39b)

above may thus resort under the Middle English idiom begin battle / fight illustrated

in (20a). Further examples supporting the possibility of collocational extension are

given in (20b-e). It is to be noted, however, that begin more commonly combined

with objects with a definite determiner, which may account for the relatively fre-

quent use of definitely marked ing-complements when these became more current in

the Early Modern period, as witnessed in Table 9.1 above.

(20) a. Ich bigon strong fæht. (c1275 (?a1200), MED)

b. For the lachesse [i.e. ‘negligence’] Of half a Minut .. Fro ferst that he

began laboure, He loste all that he hadde do. (a1393, MED)

c. Þou maist wel se þat .. Seynt Poul tauȝte ȝou to bigynne first charyte at

ȝousilf. (?a1425, MED)

d. for whoso wolde bygynne good lyf and kepe him fro synne he schulde

loue God brennyngliche wiþ all his herte and all his myght aboue all

oþer þing (c1425, IMEPC)

e. here is an ensaumple that no woman shulde take striff nor wordes with

suche men, for there is mani women that beginnithe langage [i.e. ‘a

discussion’] with a man that canne not ende it well (1484, IMEPC)

There is another possible mechanism of change, however. Ing-complements may

have arisen through confusion with bare infinitives, as a result of formal similarity,

particularly when the infinitive still had its en-ending (evidence of such confusion is

reported by Miller 2002). Indeed, the bare infinitives with en-ending that are neces-

sary for such confusion to arise are attested in Layamon’s Brut, the same text that

provides the earliest instances of ing-complements ((18a-b) above), as shown in

(21).


(21) a. bigunnen feollen (c1275 (?a1200), Visser 1963-73: 1374)

‘began to slay’

b. Þa he to Euerwic com ... heo bigunnen fehten (c1275 (?a1200), Visser

1963-73: 1374)

‘When he came to Euerwic ... they began to fight’

CONTINUE: Example (22a), from Visser (1963-73: 1897), shows that the verb con-

tinue occurred with ing-complements in Middle English. It appears, however, that

all early instances are translations from Latin: (22a), for instance, translates per-

severarent interrogantes eum (from the Vulgate Bible).95 The earliest instance not

linked to a Latin source is (22b), which, however, comes from The three kings’s

sons, a romance based on a French original. As (22c) is again inspired by Latin

(from Petrus autem perseverabat pulsans (again from the Vulgate Bible)), the first

unproblematically native examples are those in (22d-e) (assuming, for (22e), that

alle the belles is the subject of contenuyd). The first unambiguously clausal instance

appears in the early seventeenth century and is given in (22f).

(22) a. thei continueden axinge with greete voices (c1380, Visser 1963-73:

1897)

b. all his lif he contynued wepyng & teeres (c1500, IMEPC)

c. But Peter continued knocking (1534, PPCEME)

d. Al this month the Frenchmen continued spoiling of th’emperour’s

frountiers (1550-2, PPCEME)

95 Visser (1963-73: 1897) quotes three more Middle English examples. One comes from the Wycliffite Bible (just like (22a) above), and similarly goes back on Latin perseverare with a present participle. Another comes from Trevisa’s translation of Bartholomaeus’ De proprietatibus rerum and is thus presumably another Latin-based instance, though I could not check the Latin original. The last example comes from an original English text and is given in abridged form: “yif þou continue ... knokking, at þe last ... I must nedis graunte þin axing”. On checking the original, however, the full quotation turns out to be: “of þis oure lord seyth þus.Si perseueraueris pulsans propter improbitatem surget et dabit, yif þou continue in prayeris.bisily knokkyng: at þe last he seith I must nedis graunte þin axing.” (A deuout treatyse called the tree & xii. frutes of the holy goost, edited by J.J. Vaissier). In other words, the example is again Latin-inspired, and the complement-status of the ing-clause is moreover somewhat questionable. As to Early Modern English, Visser’s 1534 example from Thomas More’s writings actually comes from the 1557 translation of More’s Latin original, while his examples from Tyndale (like (22c) above) are translations from the Vulgate Bible, implying that these examples, too, plausibly result from Latin influence.


e. and at durge and masse contenuyd ryngyng alle the belles till vij at

nyght. (1555, PPCEME)

f. The vice-admirall .. began with her chace to salute her with three or

foure peeces of artillery, and so continued chasing her and gunning at

her. (1622, OED)

For Middle English, borrowing is thus likely to be the sole mechanism responsible

for the use of ing-complements with continue. Whether Latin-based uses also gave

rise to the native use of continue with ing-complement in Early Modern English is

hard to determine but as a hypothesis it faces no obvious counterevidence. There is

continuity between the period of borrowing and the first native use (see also the evi-

dence in footnote 95), so it is certainly possible that native use found its source in

the clearly borrowed instances. The original Latin structure is participial, which

means that the English ing-forms in direct translations are probably also meant as

participles. However, the fact that the nominal syntax of (22d) suggests a gerundial

interpretation for at least some instances is not a problem to the borrowing account,

for once the construction had attained native status, there is no reason why speakers

should have kept in mind the participial status of the original Latin structure, and it

is easy to confound gerunds and participles. Whether, in the end, we are to regard

the ing-complements with continue as participial or gerundial remains unclear

(paradigmatic clues point in both directions).

Still, this need not mean that native developments or language-internal factors

can be excluded from the picture. First, the use of controlled ing-complements,

whether native or non-native, received increasing support from paradigmatic anal-

ogy as the verb continue, which appeared in English only in the Late Middle English

period, gained currency in the course of the Early Modern period. More specifically,

ing-complements were paradigmatically supported especially on the participial read-

ing, on the one hand by the use of continue as a copular verb, as in (23a-b), and on

the other by the use of continue with semi-obligatory adverbials, as in (23c). Pre-

sumably, it is also this collocational behaviour that could make the Latin borrowings

acceptable in the first place. Second, the emergence of (native) ing-complements

with continue in Early Modern English coincides with the rise of to-infinitives (cf.

Visser 1963-73: 1383). These parallel developments are certainly remarkable and

suggest that continue was somehow reclassified and entered a new sphere of ana-

logical influence, in all likelihood as a new member of the class of aspectual verbs

(most of which indeed took both ing-complements and to-infinitives). This in turn

implies that semantic analogy may have also played its part in the use of continue


with ing-complements. At the same time, the late appearance of to-infinitives with

continue means that a possible check on the rise of ing-complements was absent.

This may explain why, judging by the dates of first attestation and by the quantita-

tive information in Tables 9.1 and 9.2 above, controlled ing-complements rose to

high frequency with continue in a short time-span as compared to other aspectual

verbs such as begin or cease. Third, the appearance of repeated-subject complements

with continue at the end of the sixteenth century, as shown in (24a-b), is certainly

not the outcome of borrowing, and is probably best explained as being due to para-

digmatic analogy, based on uses as in (24c), as well as to semantic analogy on the

model of other aspectual verbs selecting repeated-subject complements (see e.g.

cease or leave; see also Tables 9.1 and 9.2 above).

(23) a. And so [he] ran aboute the stretys in london nakyd & mad cryenge

alway I shall be stole. I shall be stolen. And so contynuyd mad durynge

his lyfe (1526, PPCEME)

b. so longe as thou continuest loving & kinde to me, I shall never let slipp

the least occasion dulye to observe thee. (?1635-9, PPCEME)

c. that he or they so continuing in service or other lawfull woorck or

occupac~on , shall [...] be dischardged of all Paines and Forfeitures

conteined in this Acte. (1554-5, PPCEME)

(24) a. Continue .. your skimming so long as any skim doth arise. (1594, OED)

b. Of Saint Peran, wee haue spoken before, which too well brooketh his

surname, in Sabulo: for the light sand, carried vp by the North wind,

from the sea shore, daily continueth his couering, and marring the land

adioynant (1602, CEMET)

c. I require you and every of you duely to observe and precisely to kepe

according to your Oath and duties, as you will retaine my favour and

would have me to continue my careful government over you (1586-8,

PPCEME)

d. Wyriott continuing his Malice, or rather his Madnes in this, made the

lyke Complaynt unto the Queene agaynst hir Privey Counsell (1592-

1603, PPCEME)

CEASE: As also reported by Visser (1963-73: 1901), the verb cease occurs with ing-

complements already in Middle English. Some of the early instances in the data are

given in (25). There are a number of different mechanisms that may account for the

early appearance of the combination and it is possible that each of these contributed


to the use of ing-complements with cease – certainly, none excludes the others. To

begin with, some of the early instances are clearly translations from Latin sources,

where the ing-complement is used to render the Latin present participles following

cessare: thus, (25a) translates Latin non cesso gratias agens (from the Vulgate Bi-

ble), while (25b) translates non cessabant docentes & evangelizantes (from the same

source). By contrast, the examples in (25c-d), coming from original English sources,

cannot be directly based on Latin, which indicates that the combination of cease

with ing-complement had attained native status in Middle English. While it is hard

to prove that the native uses actually descend from earlier borrowings, a connection

between native and borrowed uses might be suggested by the recurrence of negation

of cease in both Latin-based and non-Latin-based examples as a possible shared

constructional feature – then again, this evidence is not so strong, since cease gener-

ally collocated with negation in Middle English.

(25) a. I .. ceesse not doynge thankyngis for ȝou. (c1384, MED)

b. þei ceessid not teching and preching Crist Iesu (?a1425 (a1415),

IMEPC)

c. For we see wel þat þei seese neuer criing on þis litil worde OUTE, or

þis lityl worde FIIR, er þe tyme be þat þei haue in greet party getyn

help of þeire angre. (a1425 (?a1400), IMEPC)

d. She .. shalle nevyr cese cryyng .. to Christe for your good estate ..

duryng thys present lyff, and afterward for a perpetuall and a joyfull

blysse with owte ende. (c1461, MED)

An alternative scenario, noted by Visser (1963-73 : 1901), is that the use of cease

with ing-complements descends from the more common combination of cease of

with ing-complement as a result of phonetic reduction and eventual loss of the

preposition of (see (26a-b)). The problem with this account is that the supposed in-

termediary stage – cease with ing-complement with reduced of or a-prefix – is

hardly attested: the data yield only one instance, given in (26c). On the other hand,

the fact that cease of has indeed disappeared by the Early Modern period might be

taken to support the reduction scenario. Moreover, a possible reduction would have

been facilitated by the availability of transitive cease without preposition. None of

this constitutes strong evidence, though.


(26) a. þan Jone praiede hir to stynte of sech sorouful wordes, & to cese of

wepyng, & confortede hir in þe beste manere he kuoþe. (c1430

(a1410), IMEPC)

b. As sone therfor as the brethren, cese of syngynge, the ebdomadary

schall begynne: (c1500, IMEPC)

c. Þe movand heve[n]s .. Sal þan ceese o turnyng obout, And na mare

obout in course wende. (a1425 (a1400), MED)

Finally, ing-complements may also have appeared with cease through an indirect

form of paradigmatic analogy. As a transitive verb taking ordinary direct objects,

Middle English cease is typically used with respect to a situation not instigated by

the matrix clause subject, but originating from an outside source, as illustrated in

(27a-b). Accordingly, when the subject of cease ends his/her own action, this has to

be explicitly signalled by a possessive determiner in the direct object noun phrase, as

in (27c-d). As such, the collocational characteristics of cease did not immediately

favour controlled ing-complements (which rules out narrow paradigmatic analogy

with bare action nominals as a possible mechanism of change). However, given the

use with possessive determiner, it is likely that at least the repeated-subject ing-

complements in (28) arose as a result collocational extension. The next step –

though this again remains pure speculation – may have been for the repeated-subject

complements to trigger controlled complements as well, in analogy to leave (see

below) (i.e. indirect paradigmatic analogy).

(27) a. Ȝif eny sclaundre were i-rise, I cessed hit what I myȝte. (a1387, MED)

b. An oinement þat sesses ache. (a1400, MED)

c. We wold .. cese our sute and lete hym sytt yn pees. (1432, MED)

d. [I] beseche you [...] that ye wyll ceasse your sorowe, and take ayen

vnto you thastate of Ioyfull lyuyng (c1489, IMEPC)

(28) a. Þe houndes .. sesed al here sechyng. (a1375, MED)

b. Ces now youre blaberyng in the develis name. (a1400, OED)

In sum, the early use of ing-complements with cease is very likely to have been in-

fluenced by Latin usage, while it may have received further support from preposi-

tion reduction in cease of and, at least indirectly, from paradigmatic analogy. As

with continue, the eventual status of the ing-complements as either participial or

gerundial is uncertain: the Latin source suggests a participial interpretation, but the


fact that ing-complements with cease occupy a nominal slot (or can be interpreted as

such) allows a gerundial reading as well.

HATE: Contrary to Visser (1963-73: 1865), whose first instance is dated 1816, hate

is first attested with ing-complements as early as the Middle English period, as

shown by (29a-c), although the combination is admittedly uncommon. Example

(29c) shows that, though still rare, the combination of hate with ing-complement

continues in use in the Early Modern period (see also Table 9.1). Example (29d)

gives the first unambiguously clausal ing-complement with hate, which is found in

the first half of the eighteenth century. The increase in the pattern’s frequency that is

evident from Table 9.2 occurs in the course of the nineteenth century, probably un-

der the influence of like and love (for more detailed figures, see De Smet &

Cuyckens 2007: 199).

(29) a. He [Pope John XXII] hated fongyng [i.e. ‘receiving’] of ȝiftes. (a1387,

MED)

b. He sholde bene sothefaste in worde and dedd, and lowe throuth abowe

al thynge, and hate lesynge [i.e. ‘lying’]. (1422, MED)

c. But I know as many, or mo, and some, sometyme my deare frendes,

for whose sake I hate going into that countrey the more, who, partyng

out of England feruent in the loue of Christes doctrine, and well

furnished with the feare of God, returned out of Italie worse

transformed, than euer was any in Circes Court. (1570, CEMET)

d. One hates writing descriptions that are to be found in every book of

travels; but we have seen something to-day that I am sure you never

read of, and perhaps never heard of. (1740, CLMETEV)

As the examples in (30) show, the early ing-complements found with hate are a

natural extension of the verb’s collocational behaviour, which was marked by co-

occurrence with bare abstract nouns and plurals implying an action characteristic of

the subject, typically in the domain of moral behaviour or misbehaviour. In this re-

spect, the verb hate resembles its antonym love (see above), differing from the latter

mainly in being less frequent. This similarity is further underscored by the coordina-

tion of love and hate in (29b) and (30c-d). The fact that hate, like love, is late in ap-

pearing with unambiguously clausal ing-complements further indicates that these

verbs have followed the same historical trajectories (see the discussion of like be-

low).


(30) a. Kende hit is, wimman te be Schamfaste and ful of corteisie & hate

dedes of fileinie. (c1330 (?c1300), MED)

b. An emperour moot nede Be vertuous and hate tirannye. (c1375, MED)

c. to hate synne and loue vertues. (c1390, MED)

d. Þow louest soþ [i.e. ‘truth’] and hatest lees. (c1390, MED)

NEED: Bare ing-complements first appear with need in the Late Middle English pe-

riod. From the start, they show the passival control interpretation characteristic also

of later use – see the examples in (31). The appearance of ing-complements appears

to involve paradigmatic analogy. It is certainly no coincidence that the data reveal

(non-gerundial) action nominals with need that display exactly the same passival

control behaviour as their gerundial counterparts – see (32). However, as the exam-

ples in (31)-(32) show, ing-complements and ordinary passival action nominals ap-

pear in English precisely at the same time, at the end of the fourteenth century. In

view of this, we cannot speak of collocational extension in a strong sense – that is,

with the extended use diachronically following its paradigmatic model. Neverthe-

less, the parallel emergence bears witness to a strong connection between both ob-

ject types, essentially showing that ing-complements were firmly embedded in the

group of bare abstract nouns. What happened, therefore, is that a lexical change in

the collocational behaviour of need triggered the appearance of ing-complements as

part of a larger class of predominantly bare action nouns implying passival subject-

control.96

(31) a. Euerich ȝere þe spraie [of a vyne] nedeþ kuttinge and paringe. (a1398,

MED)

b. Alle þat nedeþ mete and fedyng. (a1398, MED)

c. I saye nought that me nedes na mare techynge (c1450 (c1400),

PPCEME)

d. We haue leyd þe siege afore the cite of Roan .. atte which siege vs

nedeþ gretly refresshing for vs and for our hoost. (1418, MED)

(32) a. So that it nedeth no demande, To axe of me if I be scars [i.e. ‘stingy’]

To love, for as to tho pars I wole ansuere and seie no. (a1393, HC)

96 The data suggest that the new construction is favoured in translations from Latin: (31a-b) and (32b) come from Trevisa’s translation of Bartholomaeus’ De proprieta-tibus rerum, while (32d) comes from the Middle English translation of the Chirurgia Magna by Guy de Chauliac. Still, (31c-d) and (32a, c) are native examples.


b. Also vines .. nedeþ pampynacion, þat is to menynge pullinge awey of

superfluite of leues. (a1398, MED)

c. Hit nedeth exposicyon. (a1425, MED)

d. Þe whiche .. hauynge ferse brennynge, nedeþ larger infrigidacioun.

(?c1425, MED)

As the construction in (31) remains in use throughout the further history of English,

it is remarkable that need, unlike other verbs (see e.g. escape below), never aban-

doned the passival subject-control pattern in favour of more canonical active sub-

ject-controlled ing-complements, especially when ing-clauses with the passive auxil-

iary being became available at the end of the sixteenth century. It may be surmised

that the extremely high frequency of to-infinitives with need worked as a powerful

check against this otherwise natural development. As (33a) shows, only the avoid-

ance of two consecutive to-infinitives can occasionally weaken this resistance (cf.

the discussions of horror æqui in Chapter 3).97 Finally, the availability of the to-

infinitive may also explain why ing-complements following need rarely show

clausal features – examples such as (33b) crop up only very sporadically (which

contrast significantly with worth, discussed in Chapter 4). As another interesting

oddity, notice the construction in (33c-d), suggesting that passival ing-complements

are considered analogous to passive to-infinitives with a wh-extracted subject – the

result, though, is another awkward breach of the ordinary principles of paradigmatic

regularity (cf. Chapter 7) and another example of a very local sub-regularity.98

(33) a. that this faculty tends to restrain men from doing mischief to each

other, and leads them to do good, is too manifest to need being insisted

upon. (1726, CLMETEV)

97 Remarkably, the author of (33a), in avoiding two consecutive to-infinitives, could just as well have taken recourse to the more common pattern with passival ing-complement. As such, the use of an active clausal ing-complement is not entirely enforced by repetition avoidance. 98 Specifically, the closest analogue to the alternation any question that needs an-swering / any question you need answering is the to-infinitival alternative any ques-tion that needs to be answered / any question you need to be answered. The use il-lustrated in (33c-d) certainly deviates from normal gerundial usage. Wh-extraction of a subject is only marginally acceptable among normal, non-passival gerundial ing-complements (e.g. ?Is he the one you remember carrying the weapon?) and defi-nitely shows no alternation pattern where the extracted complement subject func-tions as the subject of the matrix clause.


b. You can say all the good about her you like, I know it’ll be true. It’s a

cursed shame to treat her in this way, I don’t need telling that. (1886,

CEN)

c. We should be glad to answer any questions you need answering before

you make an application. (CB)

d. If you need any questions answering I’ll do my best to help! (Google,

10 January 2008)

LEAVE: Already in Middle English leave combined very frequently with indetermi-

nate and repeated subject ing-complements, as shown in (34a-d). Although the Mid-

dle English data also yield the first instance of an ing-complement with a potentially

clausal modifier, given in (35a), the first unambiguously clausal ing-complement

with leave appears at the end of the sixteenth century – see (35b). As the examples

in (36a-d) indicate, bare ing-complements were embedded very naturally in the col-

locational behaviour of leave – a point underscored by recurrent instances of coordi-

nation between ing-complements and bare nouns implying actions of the matrix sub-

ject in (36b-d). The examples (36e-f), on the other hand, show that the use of re-

peated-subject complements is motivated by another collocational pattern of leave,

whereby a noun is preceded by a possessive determiner co-referential with the ma-

trix subject.

(34) a. In þy etynge þow shalt reule þy hond, þat ys to say, to leue etynge

whenne þy wyl and desir lastys [i.e. ‘lusts’] ȝyt to etynge (c1425,

Visser 1963-73: 1904)

b. if thefes here berkyng of houndis, thei castith hem brede or flessh, &

so thei leve berkyng (a1500 (?a1450), IMEPC)

c. The heraudes lefte hir prikyng vp and doun. (c1385, MED)

d. then must ye promise first to leave your stealing and roving. (1481,

IMEPC)

(35) a. To restrene hur wepyng sho lefte thynkyng of þe manhede of Criste &

toke hur to vmbethynkyng of His godded. (c1440, MED)

b. Leaue colouring thy tresses (c1593, Visser 1963-73: 1904)

(36) a. Knyȝtes of Rome after sixty ȝere lefte dedes of armes. (a1387, MED)

b. Leve cuttyng and Iaggyng of clothes, ffelawship of women and

tauernes alsoo (c1450, Visser 1963-73: 1904)


c. and on that other side was the kynge Ban so a-raied that he left pley

and laughinge at the table, and cowde not wite how it was to hym

come (a1500 (?c1450), IMEPC)

d. and [he] sayth he shal dye and he can not leave syghing and many

foolysh and intemprat words why God shold thus abase hym (1582,

PPCEME)

e. Clement abatede hys herte ylle And lefte hys cheste [i.e. ‘strife,

fighting’]. (a1500 (a1375), MED)

f. Sumtyme þei þat sat þer wold speke largely of hem þat wer absent, &

þann wold Augustyn say on-to hem þat, but [i.e. ‘unless’] þei wold

leue her [i.e. ‘their’] detraccion, he must rase oute þe vers [i.e. a line of

verse written on the table against slander]. (1451, IMEPC)

FORBEAR: The use of ing-complements with forbear dates back to the Late Middle

English period, the first instances appearing around 1450 (cf. Visser 1963-73: 1871),

as shown in (37a-c). The Middle English examples all involve nominal or indeter-

minate ing-complements. Instances where the ing-complement has a potentially

clausal modifier appear in the course of the sixteenth century (the earliest example

found is (37d)), while the first unambiguously clausal instance, given in (37e), dates

from around 1600. The examples in (38) demonstrate that already in Middle English

forbear readily combined with bare abstract nouns or bare plurals implying an action

of the matrix clause subject. This shows that the first ing-complements were essen-

tially a natural extension of the collocational behaviour of forbear – another exam-

ple of narrow paradigmatic analogy.

(37) a. If freris of Frauncessis religioun forbering handling and bering of

money .. whi forberen not thei telling of money with a stikkis eende?

(c1449, MED)

b. Þar handis war so sore at þai might not forbere crying, & þai durst not

cry in þe cetie for purseyvyng. (c1450, MED)

c. We muste forbere blasphemyng, chiding .. mowyng, scornyng, and

suche oþire toward god. (c1475 (c1445), MED)

d. A great part .. forbear coming to church, and participating of the

Sacraments. (1561, OED)

e. I can hardly forbeare hurling things at him (1623 (1600-1), CEMET)


(38) a. Fastynge stant in thre thynges: In forberynge of bodily mete and

drynke and in forberynge of worldly iolitee and in forberynge of

deedly synne. (c1390, MED)

b. Quen þaim biheld þat kinges here, was nan þat lahuter [i.e. ‘laughter’]

miht forbere. (a1400, MED)

c. Among the ten comandementz y rede How that manslaghtre schulde

be forbore. (c1400, MED)

ESCAPE: The verb escape first appears with an ing-complement at the end of the

Middle English period, as shown in (39a). Early usage of the construction is marked

by a certain semantic restriction on the kind of complement selected: the comple-

ment is typically hanging or drowning or, by extension, denotes some form of pun-

ishment or unpleasant death – see (39a-c). In later examples these collocational re-

strictions are relaxed, as is apparent from (39d-e). Notice that if the verb in the ing-

complement is transitive the interpretation is passival (e.g. in (39a, c)). In this light,

it is not surprising that of the first unambiguously clausal ing-complements with

escape, as in (39d-e), the majority are passive, as in (39e).99 The characteristics of

ing-complements match neatly the wider collocational behaviour of escape: ordinary

objects too can imply passival control, and often reflect the association with pun-

ishment or death and drowning (e.g. shipwreck, suffocation, etc.). It is therefore very

likely that narrow paradigmatic analogy was the mechanism that gave rise to the

first ing-complements following escape. Some relevant examples are given in (40) –

with passival control implied in (40c-e).

(39) a. rycharde of normandy sholde not scape hangyng, what somever it

sholde hap therof (c1489, IMEPC)

b. [He] escaped drowning verye narrowely. (1560, OED)

c. those Spaniards that were fled from Trinidad, and escaped killing

(1596, CEMET)

99 That the first-attested clausal example in (39d) is not passive but nevertheless con-tains being as a copula is probably no coincidence. A very similar situation is found for bear (see below), where the transition from passival uses with nominal or inde-terminate syntax to clausal uses is similarly marked by a stage in which all clausal uses are either passive or contain the copula being. Presumably the formal similarity between the passive and the copular construction motivated this otherwise odd re-striction – another illustration of the markedness principle in diffusional change (cf. Chapter 4).


d. Gon. Had I plantation of this Isle my Lord — Ant. Hee’d sow’t with

Nettle-seed [...] — Gon. And were the King on’t, what would I do? —

Seb. Scape being drunke, for want of Wine (1623 (?1611), CEMET)

e. the rooms being all rarely furnished, and escaped hardly being set on

fire yesterday. (1662, CEMET)

(40) a. He þat passeþe þe duwe maner of Aristotle .. in moevinges or in restis,

in .. leting .. bloode, in dissolucion or withholding of þy wombe .. may

not escape seeknesse. (a1456, MED)

b. By god, ye shall not this daye escape dethe. (c1489, IMEPC)

c. he wist well be present tribulacion he schuld escape euerlastyng

dampnacion (a1500, IMEPC)

d. The headers of that truth that God techyth cannot escape just judgment.

(1537, OED)

e. which if you should, the fault Would not scape censure (1603-6,

CEMET)

3.2. Stage II: semantic analogy

Stage II, covering the period 1500-1666, is marked by the appearance of ing-

complements with leave off, remain, give over, require, want, fear, avoid, defer,

miss, omit, like, prevent, decline, prefer and bear. Stage II differs from Stage I in

two respects. The first is that ing-complements increasingly show evidence of being

proper clauses. The second and most interesting difference is that while the innova-

tions in ing-complementation in Stage I appeared to occur more or less independ-

ently of one another, Stage II sees the emergence of a number of productive patterns,

showing that ing-complementation is beginning to operate as an independent predi-

cate-complement construction. This is apparent from the fact that ing-complements

turn up in environments where they are not collocationally sanctioned – that is,

where there is no evidence of narrow paradigmatic analogy. Invariably, however,

these environments are semantically related to other environments where ing-

complements are in use already. In other words, if innovations at this stage do not

arise through narrow paradigmatic analogy, they are consistently supported by se-

mantic analogy. The consistency of this conditional relation provides independent

support to the role of semantic analogy. In turn, with semantic analogy ‘kicking into

action’, there is evidence of schematisation beyond combinations of specific verbs

and ing-complements, implying that schematic constructions have come into being.


At the same time, the domain of ing-complementation looks scattered and

regularities appear to be local rather than global. Specifically, different groups of

semantically related ing-complement-taking verbs can be distinguished, forming

constructional clusters. These clusters are internally coherent but there is no evi-

dence of schematisation beyond the separate clusters. This is evident from other

local regularities that apply within but not beyond constructional clusters. For exam-

ple, the use of ing-complements with negative implicative verbs is characterised by

the frequent occurrence of interrogative or negative matrix clauses (cf. Fanego

1996a; Rudanko 2000). This negative/interrogative pattern can be thought of as a

construction in its own right, typically used to signal a strong disposition of the ma-

trix subject towards the realisation of the action denoted by the complement clause,

which is portrayed as something to which the subject is inevitably driven, either by

their own predispositions or by external circumstances. Early on, this use is fore-

shadowed by forbear, as shown in (41):

(41) a. Who is it, that reading Beuis of Hampton, can forbeare laughing, if he

marke what scambling shyft he makes to ende his verses a like. (1589,

OED)

b. Our soildiers could not forbeare dauncing in the holie quire, whereat

the Baalists were sore displeased. (1642, OED)

Other groups of semantically related verbs selecting ing-complements show similar

group-specific collocational behaviour, supporting at once their status as a coherent

group and their independence from other groups of verbs.

Another indication of the absence (or low salience) of an overarching gener-

alisation is that ing-complementation becomes productive for the different groups of

verbs at different times – an observation that brings us back to the issue of diffusion.

The first group to combine with ing-complements productively are the aspectual

verbs (see especially leave off, give over), followed by necessity verbs taking pas-

sival ing-complements (need, require), then negative implicative verbs (see espe-

cially omit, decline), and finally endurance verbs (bear) and emotive verbs (like,

possibly prefer).

To explain the order in which verbs in different groups are affected by seman-

tic analogy, both the strength of the sanctioning model and the degree to which the

new environment resembles its model have to be taken into account. For example, I

take it to be no coincidence that leave off and give over are probably the first verbs

selecting ing-complements through semantic analogy. First, their sanctioning model


– the aspectual verbs – is, as a result of the developments in Middle English, the

group of verbs already best-represented among ing-complement-taking predicates

(see the appearance of ing-complements with begin, continue, cease, leave in Stage

I). Second, both leave off and give over are nearly synonymous to leave (and leave

off is formally similar on top), unlike for instance finish or stop (both appearing with

ing-complements in Stage III), whose original lexical semantics are further removed

from that of the aspectuals (i.e. finish implies completion of an action rather than

abandonment, whereas stop primarily applies to the actions of other agents than the

matrix subject). In the same vein, the relatively late appearance of ing-complements

with emotive verbs can be linked to the fact that Middle English provided only one

emotive verb taking ing-complements with some frequency, namely love (hate being

very infrequent), and that the development of ing-complements with this verb was

furthermore stalled by the blocking effect of the to-infinitive.

Analogical sanctioning apart, independent factors and developments continue

to interfere with the progression of diffusion. It is a lexical change, for instance, that

creates the necessary conditions for ing-complements to spread to avoid, which in

turn marks the point at which ing-complementation becomes productive for negative

implicative verbs. Similarly, paradigmatic analogy is still lurking in the background.

In the appearance of ing-complements with want and require, for example, narrow

paradigmatic analogy converges with semantic analogy (which probably explains

why ing-complementation with the group of necessity verbs is comparatively early

to expand). In general, all changes that occur are sanctioned by broad paradigmatic

analogy, even if ing-complementation is still insufficiently productive for broad

paradigmatic analogy alone to operate as a driving force of change.

LEAVE OFF: The first ing-complement attested with leave off is a repeated-subject

complement found in the Middle English data – see (42a). A probable first example

with controlled ing-complement appears at the very end of the Middle English pe-

riod and is given in (42b) – surprisingly, this early example already shows clausal

features. The following attestation of leave off with controlled ing-complement is

given in (42c) and is dated 1530 – a date very close to Visser’s (1963-73: 1904) old-

est example, which is dated 1534. It therefore seems safe to assume that the pattern

with controlled ing-complement was more or less absent in Middle English. This

absence in the Middle English data can be explained by the fairly low frequency of

leave off in Middle English and by the fact that the verb’s collocational behaviour

did not particularly welcome bare ing-complements: IMEPC and PPCEME yield

only two instances of leave off with a noun phrase having the same characteristics as


an ing-complement – see (43a-b). Somewhat better-represented, but by no means

frequent, are noun phrases with a possessive determiner referring to the matrix sub-

ject, suggesting that at least the repeated-subject complement attested with leave off

might have arisen through narrow paradigmatic analogy. As to the controlled ing-

complements, the most plausible explanation is semantic analogy to other aspectual

verbs and to leave in particular. An indication to this effect comes from the fact that

leave off also begins to pattern with the to-infinitive in the sixteenth century, copy-

ing the behaviour of leave and other aspectuals – some examples are given in (44)

(cf. Visser 1963-73: 1387-8).

(42) a. it is speedful sumtyme to leue of þi corious worching in þi wittes &

leere þee to taast sumwhat, in þi felyng goostly, of þe loue of þi God.

(a1425 (?a1400), IMEPC)

b. Themperour coude not wele leue of embracyng and kissing the kynge

of ffraunce. (c1500, IMEPC)

c. And God left off talkyng with him, and departed vp from Abraham.

(1530, PPCEME)

(43) a. Trewly I trowe, bot ȝif þei haue grace to leue of soche pipyng

ypocrisie, þat bitwix þat priue pride in þeire hertes wiþ-inne & soche

meek wordes wiþ-outyn, þe sely soule may ful sone sinke into sorow.

(a1425 (?a1400), IMEPC)

b. it was tyme nowe to leue of alle sorowe & lamentacion for any fortune

that was befalle / & to put alle thynge yn foryetyng (c1500, IMEPC)

c. Thus was this wise king excused, And thei lefte of here evele speche.

(a1393, MED)

(44) a. For when the words of fortune do leue of to sound in myne eares, the

former greife commeth againe, and ouermuch greuyth my mynde.

(1556, PPCEME)

b. Now my Lord Hastings, and Sir William Stanley Leaue off to wonder

why I drew you hither (1595, CEMET)

REMAIN : As the examples in (45) show, remain began to occur with ing-

complements in the beginning of the sixteenth century. Remain being an intransitive

verb, its ing-complements are clearly not nominalisations and are therefore likely to

be participial in nature. The mechanism responsible is most plausibly a mixture of

semantic and paradigmatic analogy. At the end of the Middle English period, remain

developed a copulative use that sanctioned adjectives and adjectival participles. As a


result, it acquired an argument slot for participial ing-complements, while at the

same time it came to resemble more closely the verb continue – which developed a

copulative use around the same time (see continue above) – and so became more

prone to analogical influence. The examples in (46a-b) show a pattern that is par-

ticularly frequent among the first copulative uses of remain, consisting of remain

with an adjectival subject-complement formed from a past participle with the prefix

un-, while (46c) gives an example with unprefixed past participles, foreshadowing

the present participles that will appear after 1500. The examples in (47), on the other

hand, show some of the similarities that arose between remain and continue at the

beginning of the Early Modern period, providing a basis for semantic analogy and

the spread of ing-complements to remain: (47a-b) demonstrate the use of continue

and remain with predicative noun phrases, (47c-d) with prepositional phrases. A

final witness to the close relation between remain and continue is their occasional

co-occurrence in discourse, as in (48) (note, incidentally, the shared ing-

complement).

(45) a. still remayne they vnto the planke cleuing (1515, Visser 1963-73:

1899)

b. If it shall happen the Cattel or Sheep of the one Realm to be staff-

herded, or to remain depasturing upon the ground of the opposite

Realm. (1563, OED)

c. And seeing I cannot be proved guyltie of any Crime, and that I

remayne here repayring my selfe [i.e. ‘abiding’] to serve your Highnes

[...]; I trust your Majestie will cause it to be examyned, wherfore and

by whom such unusuall Letters were sent out, with Intent to doe me

Harme. (1592-1603, PPCEME)

d. The Lyon .. remaines feeding .. whilest his small seruant [sc. the

Jackal] stands barking, and yalping by. (1623, OED)

(46) a. The violence .. yet remayneth unpunysshed, to theire grete hurte,

hyndryng, and damagis. (1447-8, MED)

b. Youre dettes remaynyng ungadered. (1461, MED)

c. þerfore þei remaynen implied & encombred in hemself, þat þei mowe

not be lifte up in spirit aboue hemself. (a1500, IMEPC)

(47) a. Sythen whiche tyme your seid Subgiect hathe contynued .. your

feythfull and true liegeman. (1503-4, OED)


b. Yf any wood other than .. his owne he aforestid [i.e. ‘convert into

cultivated forest’] .. it shalbe disforestid, and yf he afforested his owne

propur wood remayne it forest. (1502, OED)

c. Nor certes you that be set in the increase or waye of vertue haue not

come to abounde in pleasurs and to continue in lustes of the fleshe.

(1556, PPCEME)

d. But all maner of fortune is euell to them that remayne in wickednes

(1556, PPCEME)

(48) to remaine and contynue drinkinge or tiplinge in the saide Inne Victu-

allinge House Tiplinge house or Alehouse (1603-4, PPCEME)

GIVE OVER: The examples in (49) show the earliest attested ing-complements fol-

lowing give over. The verb first appears with a repeated-subject complement in the

first half of the sixteenth century, but examples with controlled ing-complements

follow very soon, and have become relatively current by the first half of the seven-

teenth century. As to mechanisms of change, evidence for narrow paradigmatic

analogy is virtually absent: only one example was found that might provide a collo-

cational match for bare controlled ing-complements and that actually predates the

latter’s appearance (see (50a)). The first example that might motivate the appearance

of repeated-subject ing-complements post-dates the emergence of the latter by al-

most fifty years (see (50b)). Semantic analogy, on the other hand, provides a plausi-

ble explanation for the appearance of ing-complements, given the use of both con-

trolled and repeated-subject ing-complements following other aspectual verbs, and

therefore remains as the most likely mechanism of change. Support for this point of

view comes from the fact that the appearance of ing-complements follows very

quickly on a change in the lexical semantics of give over from ‘pass on, transmit’ to

‘finish, stop’ ((50a) is the oldest example with the new meaning), and furthermore

coincides with the first appearance of (very sporadic) to-infinitives (see (51)), both

which observations are in line with the view that give over was recruited as a new

member of the class of aspectuals.

(49) a. Except these geese go from theyr olde flock and giue ouer all theyr

olde gagelynge [etc.]. (1532, OED)

b. For if ye plough men yat now be, were made lordes they woulde cleane

gyue ouer ploughinge (1549, PPCEME)

c. Let the lasses giue over laaking in the greene. (1599, OED)


d. Seeke to ouertoile him, and make him glad to giue ouer striuing to get

the leading. (1607, OED)

(50) a. It ryseth on my owne mynd to give over grett tuggs [i.e. ‘efforts’] of

husbandry which I had, and take me to lesse charge. (1504, OED)

b. Clitus .. would not giue ouer his impudencie and malapertenesse.

(1579-80, OED)

(51) a. Certain persones ... saiyng that Demades had now geuen ouer to bee

suche an haine, as he had been in tyme past (1542, Visser 1963-73:

1387)

b. Awhile he stood in this astonishment, Yet would he not for all his

great dismay Give over to effect his first intent. (1590, OED)

REQUIRE: An early example of require with ing-complement is found in Middle

English, and given here as (52a). The example is highly dubious, however, since the

ing-form is probably a lexicalised gerund meaning ‘reward, repayment’ (in fact, the

example is classified under this sense in MED s.v. guerdoning). It is safer therefore

to assign the appearance of ing-complements with require to the middle of the six-

teenth century, in accordance with the examples in (52b-d). Incidentally, the figures

in Table 9.1 above suggest that the pattern was absent in the period 1640-1710, but

the material from the OED easily corrects this false impression, as shown in (53).

Already in Middle English, require frequently combined with bare mass nouns, as in

(54a-b), but only occasionally did these nouns also imply a passival control relation,

as in (54b). Examples of the latter type became more frequent, however, in the be-

ginning of the Early Modern period, as shown by (54c-d), and the appearance of

bare ing-complements (with passival interpretation) can be seen as a natural exten-

sion of this use (i.e. narrow paradigmatic analogy).

(52) a. In love, free yeven thing Requyrith a gret guerdoning. (a1425

(?a1400), MED)

b. For, great shippes, require costlie tackling, and also afterward

dangerous gouernment (1563-8, PPCEME)

c. And yet neuerthelesse, in all this good proprietie of wordes, and

purenesse of phrases which be in Terence, ye must not follow him

alwayes in placing of them, bicause for the meter sake, some wordes in

him, somtyme, be driuen awrie, which require a straighter placing in

plaine prose (1570, CEMET)


d. After the ground is sowne, it requireth weeding, sarcling, or raking.

(1601, OED)

(53) a. It had also a rawishnesse in it, as if the fat required boiling. (1662,

OED)

b. with this [i.e. a saddler’s strainer] the Girth web is fastned and drawn

streat upon the Sadle trees; or in such places where the Girth requires

straining. (1688, OED)

(54) a. The mater requyreth haste, for such that ye woll not ymagen. (1463,

MED)

b. In somych forsoþ as an vlcere is an vlcere, it requireth desiccacioun

after..Ipocras. (?a1425, MED)

c. The mater is spirytuall, and requyreth moche declaracion in englysshe.

(1526, OED)

d. for I wold nut onely commynycat thyngs vnto yow wherin for my

comfort and relief I wold haue your good, sad, dyscret aduyse and

counsell, but also opon the same commytt sertyng thyngs requyryng

expedition [i.e. ‘executing’] to yow (c1529, PPCEME)

WANT: The verb want, originally meaning ‘lack, not have’, already combined with

ing-forms in Middle English, as shown in (55). Typically, however, the ing-forms

found with want belong to a group of lexicalised nouns such as being, understand-

ing, moving, or teaching. These lexicalised nouns are characterised by their slightly

specialised meanings, merely implicating the verbal process expressed by their mor-

phological stem as a non-focal element of their semantic structure. Thus, moving in

(55a) denotes the ability to move; teaching in (55b) denotes the result of teaching,

i.e. a received education.

(55) a. for whiche I wondre gretly that men mervailen on swiche thinges [i.e.

precious stones]. For-why, what thing is it, that yif it wanteth moeving

and Ioynture of sowle and body, that by right mighte semen a fair

creature to him that hath a sowle of resoun? (?a1425 (c1380), IMEPC)

b. þus þey ben goostly disseyued, boþe for hem wantiþ teching to wende

to heuene bi cristis weye, & for þey ben led to helle bi errour of þe

fendis weye. (a1500 (?c1378), IMEPC)

The passival control construction familiar from Present-Day English appears only in

the Early Modern period. The first examples are given in (56). As the examples


show, when used with passival ing-complement, want usually means ‘need, require’.

The examples are representative of following periods, with the exception of some

very occasional departures from strictly nominal syntax in the ing-complement in

later usage, as illustrated in (56c-d). Still, two additional (though admittedly minor)

innovations occurred: as (57a) shows, ing-complements with want can also be used

in analogy to a passive to-infinitive with wh-extracted subject (cf. need above) –

indeed, the OED (s.v. want) already reports the equivalent with ing-form of a to-

infinitive with explicit subject (or else of the two-place object-complement construc-

tion with past participle), as in (57b). Example (57c), on the other hand, shows that

the passival construction also occasionally combines with want in the sense in which

it is presently most common, namely ‘desire, wish’. Notice finally that want like

need never shows convincing evidence of selecting fully clausal ing-complements

with normal active control. As with need, this is presumably due to blocking by the

to-infinitive.

(56) a. It was but rough hewen by one of the prentises, and wanted sum

polishing by the forman. (1574, OED)

b. But hereunto we want presently indifferent using [i.e. ‘impartial

treatment’]. (1563-83, Visser 1963-73: 1887)

c. ‘Then you do delight to oppress her.’ ‘I don’t, I tell you! – only when

I’m in a bad humour [...] or when she looks flat and wants shaking up a

bit.’ (1848, CLMETEV)

d. Lupin proceeded to put on a bright-blue coat that seemed miles too

large for him. Carrie said it wanted taking in considerably at the back

(1894, CLMETEV)

(57) a. Anyway, Reg has questions he wants answering. (CB)

b. I want that doing (OED)

c. She did not call for help because she did not want saving. (CB)

Several mechanisms may have conspired to introduce passival ing-

complements with want. For a start, ing-complements have a conspicuous colloca-

tional analogue in ordinary direct objects following want. As shown in (58), want

combined commonly with such bare abstract nouns already in Middle English, in-

cluding ing-forms (see (58b)). In Middle English, however, the implication of a pas-

sival control relation hardly ever manifested itself, showing up at best exceptionally

(see (58c-d)). By contrast, bare abstract nouns implying passival control became

more common in the course of the sixteenth century, as shown in (59).


(58) a. Me wantede verray contrycyone [i.e. ‘remorse’], wythowtten þe

whilke, all othere thynges avayles noghte. (c1440 (a1349), PPCME)

b. And also yn þe selue manere to hym þat hauys vsyd to ete but oon

meel, and he begynne to ete twyes; ffor he shal wante defyinge [i.e.

‘digesting’] of stomak, and so his mete dwellys nought defyed. (c1450,

IMEPC)

c. We rede in ‘Libro de Dono Timoris’ of a womman þat was devowte; &

on a tyme when sho wantyd comfurth at sho was wunte vnto, & was

ferd at it suld tary lang or it come, [...] sho spakk vnto hur awn vertues

at was with-in hur & sayd [...] (c1450, IMEPC)

d. And whan this wyse man saugh that him wanted audience, al

shamefast he sette him doun agayn. (c1390, IMEPC)

(59) a. Many times woulde they come into the Citie, neither wanted that

thinge great suspicion. (1560, OED)

b. Buck. Is all things ready for the Royall time? — Darb. It is, and wants

but nomination. (1623 (1591), CEMET)

c. If their Sonnes be ingenuous, they shall want no instruction (1623

(1595-6), CEMET)

It is more than plausible that the increase in passival noun objects with want is

linked to the appearance of passival ing-complements. However, as both develop-

ments take place more or less simultaneously, the link may not be directly causal. It

is more likely that the joint appearance of the two patterns reflects the interaction

between two factors: on the one hand, the ground for passival bare nouns and ing-

complements had been prepared by the collocation of want with bare abstract nouns

in general; on the other, the shift to bare nouns with passival control may have been

triggered by the analogical influence of need and require (see above). In this respect,

notice that want shifted meaning from ‘lack, not have’ to ‘need, require’ around the

same time as it appeared with passival ing-complements. The new sense is often

implicated in early sixteenth century examples, as in (60a-b), and is undeniably se-

manticised by the end of the century, as shown in (60c-d).

(60) a. I Thrust [i.e. ‘thirst’], I want drinke. (1530, OED)

b. I pray to god to parfourme [i.e. ‘supply’] that she wantethe. (a1533,

OED)

c. Oh welcome Oxford, for we want thy helpe. (1593, OED)


d. Though we have not beef and mutton, &c. yet (God be praised) we

want them not; our Indian corn answers for all. (1630, OED)

FEAR: The earliest example of fear with ing-complement is to be found in Visser

(1963-73: 1869) and is given here as (61a). Notice that some examples show pas-

sival control, for instance (61b). The first clausal examples appear at the end of the

seventeenth century, as shown in (61c-d).

(61) a. those who feare drowning come neere no wells, nor they that dread

burning neere to fire (1580, Visser 1963-73: 1869).

b. But on another bark while they take hold, They now full fraught, and

fearing overtrimming [i.e. ‘being overbalanced’], With cruell sword ..

Cut of their hands. (1591, OED)

c. And I, of all your sex, Most fear displeasing you (1678, CEMET)

d. haste to me then, my lovely maid, and fear not being discovered (1684,

CEMET)

As to possible causes, one mechanism of change that may have triggered the appear-

ance of ing-complements is narrow paradigmatic analogy: fear combines with bare

nouns implying actions and situations involving the matrix subject, either with im-

plicit subject-control, as in (62a, d), or with implicit passival control, as in (62b-c,

e). Notice that the transitive use of fear (in the sense ‘be afraid of’) is relatively re-

cent, emerging only at the end of the Middle English period (cf. OED s.v. fear),

which explains why ing-complements did not appear with fear earlier. Next to para-

digmatic analogy, however, it is not inconceivable that the verb fear also modelled

its behaviour after its nominalised counterpart, which became very common with

gerundial ing-complements with of (including passival examples) in the course of

the sixteenth century, as illustrated in (63a-b). As suggested by (63c), the latter con-

struction may have arisen through French influence.

(62) a. Some feareth synne & payne bothe, hauynge an eye and respecte to

bothe in maner indifferently. (1526, OED)

a. fear ye imprisonment? (1552-3, PPCEME)

b. But my Lord’s purpose to have Men planted at the Court, was in regard

he feared hindrance by private Enemies (1600, PPCEME)

c. Why he that cuts off twenty yeares of life, Cuts off so many yeares of

fearing death (1623 (1599), CEMET)


d. Feare not Slander, Censure rash (1623 (c1609), CEMET)

(63) a. Put not to many shepe in a penne at one tyme .. for feare of murtheryng

or ouer pressyng of their felowes. (1523, OED)

b. For feare of decaying the common wele, men are driuen to put

malefactors to pain. (1529, OED)

c. que se ce ne fut de peur de gaster mon lit, je vous eusse gectés ung

chauldron d’eaue (1515, FRANTEXT)

‘that if it were not for fear of spoiling my bed, I would have thrown

you a jar of water’

AVOID: The verb avoid first appears with bare controlled ing-complements at the

very end of the sixteenth century, as shown in (64a), at which point its complements

still appear to be nominal (see (64b)). The first potentially clausal ing-complement

with avoid, given in (64c), dates from 1635, while the first unambiguously clausal

instance, in (64d), appears in 1646. This fairly rapid succession of examples an-

nounces the quick rise in frequency of avoid with ing-complement that took place in

the second half of the seventeenth century (cf. Table 9.1 above). Finally, as (64e)

illustrates, avoid also quickly entered in the construction with negated can that is

characteristic of the period and that accounts for about one third of all instances of

avoid with ing-complement between 1640 and 1710.

(64) a. Avoyde filchinge and robbinge. (1597, OED)

b. I thinke they would not haue denied vs wine, .. yet to auoide troubling

of them, my selfe and my brother carried some flaggons of rich wine.

(1617, OED)

c. If .. I could have avoided meddling with him, I should not desirously

have begun with a Gentleman .. of so .. turbulent a Disposition. (1635,

OED)

d. How you avoid being goared [i.e. ‘pierced’] by the three hornes of my

Syllogisme. (1646, OED)

e. go which way you will, it should seem that without a representative of

the people, your commonwealth, consisting of a whole nation, can

never avoid falling either into oligarchy or confusion. (1656, CEMET)

As for mechanisms of change, paradigmatic analogy is probably at least partly re-

sponsible for the introduction of ing-complements. As the examples in (65) show,

ing-complements nicely fit the wider collocational behaviour of avoid, which natu-


rally favours bare nouns implying an action or situation involving the matrix subject.

However, paradigmatic analogy may have converged with the analogical pull ex-

erted by semantically related verbs already taking ing-complements, such as forbear

and escape (see above), or the somewhat less frequent eschew or shun.100

(65) a. they be false witnesses saith S. Jerome, which do ad, alter, wrest,

double, or do speake for hope to auoid Death, or for malice to procure

another Man’s Death (1554, PPCEME)

b. surelie than rash ignorant heads [...] either durst not, for lacke of such

learnyng: or els would not, in auoyding such labor, be so busie, as

euerie where they be (1570, CEMET)

c. The name of Christians, which verie title enioyneth a serch to avoyd

contention, euen by submission of the wronged. (1582, OED)

What is still somewhat puzzling is the relatively late emergence of ing-

complements with avoid as compared to other verbs such as forbear or escape. To

understand the reasons for this we must consider the lexical semantics of avoid.

Originally, avoid denoted the purely physical action of emptying a container of its

contents (e.g. (66a)). By extension, the meaning of avoid could also involve either

movement of the subject away from an object (‘abandon, leave from’; e.g. (66b)) or

expulsion of an object by the subject (‘send away, dispel’; e.g. (66c)). The semantic

change that eventually produced the present-day meaning of avoid ‘shun, escape’

100 The earliest examples of ing-complements with eschew and shun predate the ear-liest instances with avoid, as shown in (i). Eschew especially occurs with ing-complement with some frequency in Middle English, judging by the corpus data from IMEPC and PPCEME. Note that the use of ing-complements with these verbs can be explained as the result of narrow paradigmatic analogy (cf. the examples in (ii)), in line with the general development of ing-complementation. (i) a. To eschewe Engrutynge [i.e. ‘overeating’]. Kepe þe wel þat þou ete

noght anoþer tyme, vnto þou vnderstonde certanly þy stomak voyde, þat ys to wete, þat it be clensyd of þe ferste etynge, and þat shalt þow knowe by appetyt of etynge (c1450, IMEPC).

b. O Atalanta, thou at all of husband hast no need, Shun husbanding. (1565, OED)

(ii) a. For to eschewe leccherye and vileynye. (c1390, MED) b. Shune lustliche wil. for þat it dereð swiðe and beð afterboht [i.e.

‘payed for afterwards’] mid bitere sor (a1225, PPCME)


may have drawn on all three previous meanings, with figurative uses bringing the

necessary opportunities for confusion and, thereby, semantic extension.101

(66) a. Yee shall drawe waters .. Which have vertue to curen .. Hertes

avoydoying off alle theire hevynesse. (?1435 (1432), MED)

b. And thenne syre Gawayn and syr Borce dyd their message / and

commaunded Lucius in Arthurs name to auoyde his lond / or shortly to

adresse hym to bataylle (1485 (a1470), IMEPC)

c. Diadose is a stone which is pale, & he is found in water, & he is gode

to avoid deuelis (a1500, IMEPC)

Importantly, even though the present-day sense of avoid was a Middle English inno-

vation, it is at best marginally attested in the Middle English period itself. No clear

examples are found in HC prior to 1500, nor in PPCME; while IMEPC yields a

number of examples, many of which, however, also allow one of the readings illus-

101 For example, avoid could be used with reference to arguments or legal docu-ments in the sense of ‘hollowing out, invalidating’ (a blend, on a figurative plane, of the original senses ‘empty’ and ‘dispel’), as in (iii.a). The present-day sense of ‘es-cape’ arises when invalidation also implies the cancellation of future negative con-sequences, or applies to a foreseen counterargument – as in (iii.b). In another meta-phorical extension, avoid could be used to refer to the expulsion of mental contents, as in (iv.a). The step needed to get to the present-day meaning of avoid then is a stronger interpretation, according to which the undesirable mental contents are not allowed to enter in the first place – an interpretation that is triggered, for instance, by prescriptive contexts such as (iv.b). The inferential reasoning applied here is that prevention (‘shunning’) is better than cure (‘dispelling’), by which reasoning, a re-quirement for cure automatically translates into a requirement for prevention. Simi-lar inferences arise in other (literal or metaphorical) domains where avoid denotes expulsion of an undesirable element from some (figurative) container, for example with respect to unwelcome characters in a population, immoral qualities in a person, or corruption in the system of justice. (iii) a. I will avoiden and dispreuen Conceites whiche þe to þis errour meven.

(c1450 (1410), MED) b. But for þe more cleere declaring of þis mater. and avoiding of

obiecciouns þat mai be putt forþe; we schullen vndirstonde þat þer ben þre chirchis (?a1425 (a1415), IMEPC)

(iv) a. þese veyn þoutes þat come sodeynly on us, avoyding þe swete deuocion þat we wold haue (c1450, IMEPC)

b. þei schuld no euele speke, but be occupied with orisones and meditaciones to avoyde euel þoutes. (c1450, IMEPC)


trated in (66) (the most common ambiguities are discussed in footnote 101). One of

the first clear instances of avoid in its new sense is given in (67).102 It is reasonable

to assume, then, that the late appearance of controlled ing-complements with avoid

(meaning ‘shun, escape’) is best explained by the late emergence of the relevant

sense of the verb.

(67) whan I haue on my self compassion / Thenne am I enioyed of thyn

ease / & take grete playsir / in this, that thou auoydest the myserries

that I suffre euery day (1484, IMEPC)

DEFER: The first example of defer with bare controlled ing-complement turns up in

1630, as shown in (68a). The combination soon becomes fairly common, witness the

examples in (68b-c) (cf. also Table 9.1). The appearance of ing-complements with

defer cannot be the result of narrow paradigmatic analogy, the objects of defer being

nearly always definite (which makes sense, for as the object of defer is typically a

preconceived plan, it can be presented as specific and discursively given; consider

(69a-c)). Still, the strong collocational preference might account for the definite and

repeated-subject ing-complements found with defer, as in (69d-e) (see also Table

9.1), whose presence may in turn have furthered the use of bare ing-complements

(i.e. indirect paradigmatic analogy). At the same time, it is plausible that the use of

ing-complements with defer was supported by semantic analogy, the pattern being

modelled on the use of ing-complements with other negative implicative verbs. One

indication to this effect is the firm association of the pattern with negated can (see

(68b-c), which can be compared to the analogous examples with forbear in (41)

above), underscoring its membership of the class of negative implicative verbs typi-

cally found in the negated can construction.

(68) a. These deferre giving, till they cannot give, making their executors their

almoners. (1630, OED)

b. If I come not home Saturday nigh, I cannot possibly defer coming

longer then Wedensday in Whitson weeke (1647-8, PPCEME)

c. The lad so commanded, could no longer defer telling him Sylvia was

gone (1684, CEMET)

102 See further the examples in MED s.v. avoiden (5), some of which are in fact also ambiguous, or the one Middle English example in OED s.v. avoid (8), which, how-ever, is not entirely convincing either.


(69) a. I deferred the search of the country on Guiana side till my return down

the river. (1596, CEMET)

b. I doe beseech your Grace to pardon me, Who earnest in the seruice of

my God, Deferr’d the visitation of my friends. (1623 (?1591),

CEMET)

c. he deferred his journey until August, because I was delivered on the

30th of July of a daughter. (1676, CEMET)

d. A letter .. beeing delivered him .. at supper, he deferred the opening of

it, pronouncing this by word, To morrow is a new day. (1603, OED)

e. Madame I am verie sorye that I am necessitated upon my brother

Gerard’s letter to me, so freely letting me know the unseasonableness

of my intention to have visited yow this weeke, to deferr my waiteing

upon yow somm few dayes longer. (1629-30, PPCEME)

M ISS: As a verb taking ing-complements, miss appears in two different senses. First,

miss may mean ‘fail’ and in this sense, which is also the most common, it first ap-

pears with ing-complements in the first half of the seventeenth century, starting

1630, as shown in (70a-c). Second, miss may mean ‘avoid, escape’, in which sense it

first combines with an ing-complement, albeit quite infrequently, at the end of the

seventeenth century, as shown in (70d).

(70) a. I haue knowne [...] a Marchants wife a quicke Gamester at Irish [i.e. a

game resembling Backgammon] [...] that she would seldome misse

entring. (1630, PPCEME)

b. I will not undertake [i.e. ‘guarantee’] that though you’l hardly miss

changing the Colour of our shining Tinglass, yet you will the first or

perhaps the second time hit Right upon the way of making the

Glistring Sublimate I have been mentioning. (1664, PPCEME)

c. Pardon me for haveing mist writing to you so many posts (1664, Visser

1963-73: 1872)

d. They very narrowly missed being taken by the Christians. (1687, OED)

How ing-complements appeared with miss is something of a puzzle. The wider col-

locational profile of miss does not at all welcome ing-complements, which makes

paradigmatic analogy very unlikely. To illustrate this, some examples of transitive

miss are given in (71), showing the verb’s preference for concrete definite objects

without any strong implication of an action by the matrix subject. By contrast, there


is some evidence that the pattern may have arisen through semantic analogy. It may

be observed that most examples show miss combining with a negative polarity item

(see (70a-b)), suggesting the influence of the negative constructions found with

other negative implicative verbs (see esp. the examples with forbear in (41) or defer

in (68b-c)). Yet it is clear that the negative pattern with miss lacks modal can or

could and so does not match exactly the construction familiar from other negative

implicative verbs.103 Further note that, unlike other negative implicative verbs, the

frequency of miss with ing-complements does not peak in the period 1640-1710 but

continues to rise in the eighteenth century (compare Tables 9.1 and 9.2).

(71) a. You finde not the apostraphas, and so misse the accent. (1623 (1595-

6), CEMET)

b. Two Beggers told me, I could not misse my way. (1623 (?1609),

CEMET)

Foreign influence may have played its part as well, with the French verb manquer

(‘miss’) providing a model for the use of English miss. It is certainly striking that in

the FRANTEXT corpus manquer first appears with an infinitival complement in

1610, about twenty years before ing-complements appear with miss in English –

there is a good chronological match, in other words (see (72) for one of the earliest

French examples). Even so, if speakers of English replicated the French pattern, that

still does not explain why in doing so they used an ing-complement rather than an

infinitive. In sum, then, we had perhaps best account for the appearance of ing-

complements with miss as the result of a mixture of French influence and semantic

analogy. It ought to be clear, though, that the evidence for neither mechanism is en-

tirely satisfactory.

(72) je fis semblant de dormir, et toutesfois je tenois les yeux entr’ouverts

pour voir ce qu’il deviendroit, et certes il ne manqua point de faire ce

que j’avois pensé (1612, FRANTEXT)

‘I pretended to be asleep and all the time I kept my eyes half open to

see what he would do, and for sure he did not miss doing what I

thought he would do’

103 Then again, the absence of modal can or could is not inexplicable. Semantically, miss denotes an unintentional action, of which it consequently makes less sense to say that the matrix subject is ‘able’ to achieve it.


OMIT : The appearance of bare controlled ing-complements with omit can be dated to

1647-8 (see (73a)). The ing-complement in the first example is nominal, but in the

following examples, which appear after 1660 (see (73b-c)), they clearly allow

clausal syntax. The first example may have been triggered by horror æqui, the direct

consecution of two to-infinitives being stylistically awkward, but it is unlikely that

this could explain the later instances of the pattern. For these semantic analogy pro-

vides the most plausible mechanism of change. Supporting semantic analogy is the

fact that, as (73c) demonstrates, omit behaves like other negative implicative verbs

in entering the negated can construction. There is no straightforward evidence for

other mechanisms of change.

(73) a. and therfore am forced to omit buying of some things (1647-8,

PPCEME)

b. and [I] have not omitted writing to thee by the Fryday and Tuesday

Post since I came to London (1662-3, PPCEME)

c. And here I cannot omit relating the ensuing story, confirmed by Sir

Thomas Barton (1676, CEMET)

LIKE: Starting in Middle English times, the verb like has had a complicated history

which also reflects on the development of ing-complements. Ing-complements with

like make their entry in the data with the example given in (74a), dated a1649. The

start is clearly hesitant though, as appears from the following examples attested,

given in (74b-e).

(74) a. Of all pastimes and exercises I like sailing worst. (a1649, OED)

b. Seizing their factories I like well enough (1673, Visser 1963-73: 1866)

c. They did not like working, so that the Cudgel was forc’d to walk now

and then to quick’n their Laziness. (1686, OED)

d. How say you, mistress, would you like going to sea? (1695, CEMET)

e. a Man of Spirit should contemn the Praise of the Ignorant and like

being applauded for nothing but what he knows in his own Heart he

deserves. (1711, Visser 1963-73: 1866)

Turning to mechanisms of change, paradigmatic analogy can hardly account for the

appearance of ing-complements, for, as illustrated in (75), the object of like is nearly

always either a person or a concrete thing, never an action carried out by the matrix

subject.


(75) a. I like the gentleman exceeding well by site and discourse. (1629-30,

PPCEME)

b. I am glad you liked my litle pulletts I sent you (1629-30, PPCEME)

On the other hand, examples (74a, c) above recall the examples found of ing-

complements with the verbs love and hate. Much like love and hate, like marks the

subject’s attitude towards some general kind of activity expressed by an indetermi-

nate ing-complement. Additionally, the pattern can be used to convey further impli-

cations about the subject’s characteristic behaviour (this is at least true for (74c);

(74a), by contrast, more strongly foregrounds the attitudinal relation, as a result of

the comparative worse). The semantic similarity between patterns of ing-

complementation suggests that the introduction of ing-complements with like may

have been due to semantic analogy.

There is some additional evidence in favour of semantic analogy. First, like

very regularly occurred side-by-side with the verb love, indicating quite clearly that

language users perceived a degree of semantic similarity between these verbs and

considered them as belonging to the same lexical field (see the examples in (76)).

(76) a. For it is well knowne, that I both like and loue, and haue alwaies, and

do yet still vse, all exercises and pastimes, that be fitte for my nature

and habilitie. (1563-8, PPCEME)

b. We had rather follow the perfections of them whom we like not, than

in defects resemble them whom we love. (1597, OED)

Second, the assumed semantic analogy fits comfortably in a broader diachronic con-

text. As is well-known, like used to occur preferentially in a ‘cause-subject’ con-

struction, where the cause or trigger of liking fills the subject role and the sentient

experiencer is coded as object, as in (77) (Fischer & Van der Leek 1983). The pre-

sent-day ‘experiencer-subject’ construction appeared with like in Late Middle Eng-

lish and entirely ousted the cause-subject construction in the Early Modern period,

the latter having more or less disappeared after 1570. If we think of this develop-

ment as a reclassification of like as an emotive verb akin to love and hate, the ap-

pearance of ing-complements follows naturally as a further step in the same direc-

tion.


(77) Myn herte weore holly his, Þat no þing lykede me but he. (c1390,

MED)

‘My heart was so entirely his, that I liked nothing but him.’

Third, other developments too testify to growing similarity among the emotive

verbs. The appearance of ing-complements coincides with the re-appearance of to-

infinitives with like, after the latter had virtually disappeared in the beginning of the

Early Modern English period.104 Just like the introduction of ing-complements, the

(re-)introduction of to-infinitives causes like to get syntactically more closely

aligned with love and hate (for figures see De Smet & Cuyckens 2007). Further-

more, later changes in the to-infinitival complement construction throw up striking

parallels between like and love or hate, both semantically (De Smet & Cuyckens

2004) and collocationally (De Smet & Cuyckens 2007) (see also the extension of

for...to-infinitives from like to love and hate discussed in Chapter 6). At the same

time, other constructions in which like engaged but which were not shared by love or

hate have tended to disappear or become archaic. This holds true for the use of

prepositional complements introduced by of, as in (78a), and for the use of that-

clauses, as in (78b).

(78) a. I have sent unto you Goodman Cooper, one hoome I thinke you will

well like of. (1629-31, PPCEME)

104 The (near-)disappearance of to-infinitives with like after 1570 is the more per-plexing for its empirical robustness. An explanation has been attempted in De Smet (2003) along the following lines: The disappearance of to-infinitive complements with like coincides with the disappearance of the cause-subject construction, so it seems plausible that the latter has triggered the former. The causal relation between both historical events, then, is probably to be sought in the fact that by the Late Middle English period use of the to-infinitive with like had largely become restricted to formulaic expressions marking negative politeness (of the type may it like your Highness to etc.). This formulaic use did not join the transition to experiencer-subject constructions, because the cause-subject construction was instrumental in downplaying the speaker’s impingement on the addressee: quite literally, the wish formulated by may it like you to etc. is not for the addressee to do anything but for some state of mind to occur to the addressee, thus implying minimal imposition on the addressee’s negative face wants. When the cause-subject pattern was lost for like, rather than switching to the experiencer-subject pattern, like was replaced by the verb please, thereby retaining the politer cause-subject syntax of the formula. The unavoidable side-effect, however, was the loss of the dominant pattern of to-infinitival complementation with like.


b. he liked not that any man shold be troubled for Religion (1600,

PPCEME)

All in all then, we may safely assume that like, love and hate form a group of simi-

larly behaving verbs. It also seems reasonable to assume that like entered this group

only in the Early Modern period, and to interpret the emergence of ing-complements

with like in this light.

One intriguing aspect to the change is still to be pointed out, however. The

appearance of ing-complements with like in the second half of the seventeenth cen-

tury coincides with an increase in the productivity of ing-complements following

other emotive verbs. Emotive verbs initially patterned only with indeterminate ing-

complements, and unambiguously clausal instances began to appear later than with

other verbs: with love (as well as with fear) this happened at the end of the seven-

teenth century; with hate it happened in the first half of the eighteenth (see above). It

is tempting to look for some causal link between the changes, yet this is difficult,

since the facts can be interpreted in different ways. The increased productivity of

ing-complementation with emotives may have set off the emergence of ing-

complements with like, but, conversely, the appearance of ing-complements with

like may also have contributed to the increase in productivity with other emotives.

Either way, though, the fact that these small changes occur in close succession adds

credibility to the status of the emotives as a cluster of related verbs.

PREVENT: The use of prevent with controlled ing-complements is now obsolete but

was current in the seventeenth century and continued in use into the Late Modern

period (cf. Tables 9.1 and 9.2). The first example of the pattern appears in 1653, as

shown in (79a), and is soon followed by other examples, witness (79b-c).

(79) a. when the poor Debitor feeling the Rope about his neck, expresses an

unwillingnesse to be made thus accessory to lying and his owne ruine,

it is told him, the Chancery will forgive and pardon him; whereupon to

prevent strangling at that instant, he sets his hand and seale (1653, LC)

b. then tie him about it with thread, but no harder than of necessity, to

prevent hurting the fish (1656, CEMET)

c. and so home in great pain in my back by the uneasiness of Sir W.

Batten’s coach driving hard this afternoon over the stones to prevent

coming too late. (1667, CEMET)


The appearance of ing-complements is likely to have been furthered by semantic

analogy, given the negative implicative semantics of prevent. Still, the pattern fails

to engage in the familiar construction with negated can, and the passival character of

its first attestation ((79a) above) is uncommon with negative implicative verbs (but

see escape). As this points to one or more other mechanisms of change being in-

volved, it is worth observing that prevent also co-occurred with bare abstract nouns,

as the examples in (80) show. While these abstract nouns did not imply active con-

trol by the matrix subject, they occasionally did allow a passival reading (see (80d)),

and provided a plausible collocational basis for the construction illustrated in (81),

in which prevent combines with a non-controlled bare ing-complement that typically

denotes theft and, again, occasionally allows a passival interpretation (see (81c)).

Together, these constructions constitute a plausible collocational model for the un-

ambiguously passival ing-complement with prevent in (79a) above. A final player of

some significance in the history of prevent is the different-subject construction illus-

trated in (82), which appeared around 1600 (though it gained currency only after

1640), and which, through indirect paradigmatic analogy, could give further support

to the use of controlled ing-complements.

(80) a. To be more upon their keeping, to prevent treachery. (1571, OED)

b. The best meanes of preuenting leakage, is to let three or foure shouels

full of earth fall softly downe, by the inner side of the flood-gate

(1603, CEMET)

c. The first multiplyeth endeuour, the second preuenteth error (1605,

PPCEME)

d. The murtherers to preuent pursuit, strewed galthrops behinde them.

(1611, OED)

(81) a. I doe nowe like some that haue an Orcharde ill neighbored, that gather

their fruit before it is ripe, to preuent stealing. (1598, OED)

b. What boxing them vp to preuent embezeling [i.e. ‘stealing’]! (1607,

OED)

c. they did but stand on their own defence, to prevent Plundering (1645,

OED)

(82) a. if wisdome and princely authority be not by you used to prevent

perilles appearing (1592, Visser 1963-73: 2352)

b. He built a house without his Camp for all strangers .. , whereby he

prevented their sneaking into his Camp. (1656, OED)


DECLINE: The verb decline begins to combine with ing-complements in the second

half of the seventeenth century, as shown in (83). Somewhat remarkable is the more

or less simultaneous appearance of controlled definite ing-complements, which turn

up with striking regularity in the following decades, as shown in (84) (cf. Table 9.1).

As the collocational behaviour of decline offers no evidence of narrow paradigmatic

analogy, the best explanation for the emergence of ing-complements (bare as well as

definite) is semantic analogy to the other negative implicative verbs. The best sup-

port for this view comes from the use of decline in the negated can construction, as

in (83c).

(83) a. Unsuccessful Sallyes, whose retail I decline telling you. (1654-66,

OED)

b. but though I took a good occasion of going to the Trumpet she

declined coming, which vexed me. (1664, CEMET)

c. And next in Conscience, I could not decline giving Judgment in this

Case (1688, LC)

(84) a. A wise man is not pragmatical; for he declines the doing of any thing

that is beyond his office. (1656, OED)

b. Sir Henry, though heir to his Brother Richard after his death; yet

perceiving himself over-titled or rather under-stated, for so high an

honour, .. declined the assuming thereof. (a1661, OED)

c. There was but one Reason given herein for declining the granting

Money, and that is the Unpresidentedness. (1678, OED)

PREFER: There are some difficulties dating the first appearance of ing-complements

with prefer. Late Middle English material throws up one fairly clear-cut instance of

prefer with a controlled definite ing-complement, given in (85a), and one example

with a potential bare ing-complement in (85b). However, the bare gerund phrase

solitary lyuynge in the latter example may in fact involve a lexicalised gerund,

lyuynge, that does not refer to the process of living but means ‘way of life’. The first

Early Modern instances appear only in the seventeenth century, as shown by the

examples with bare ing-complements in (86), and with definite and own-subject ing-

complements in (87). Note here too that the first instance with bare controlled ing-

complement in (86a) is dubious, as the form preferring might not instantiate the verb

prefer but be a lexicalised noun meaning ‘preference’. I therefore, assume that the

use of bare controlled ing-complements starts with (86b), dated 1664, though it is to


be regarded as plausible that bare controlled ing-complements were marginally ac-

ceptable before that time.

(85) a. Mevyng hem to preferre so grete goodes and theschuyng of the grete

and innumerable mischiffes and inconvenientes [i.e. ‘offenses’] befor

the rigour and hardinesse [i.e. ‘obduracy and arrogance’] of her propre

ententes and desirs. (1439, MED)

b. First take heed why thou shalt preferre solitary lyuynge aboue the

conuersacion of the worlde. (a1450, PPCME)

(86) a. The Apostle here wisely directs his Corinthians to a preferring of

prophecying, i. e. an abilitie of understanding the mysteries of the

kingdome, of explaining Scripture, of praying and praising God,

foretelling things to come, and discovering the secret and hidden things

of men: all which this discourse concludes under Prophecying in its

latitude (1653, LC)

b. Turning up of the Earth .. is to be preferr’d to Hand weeding, and more

expeditious. (1664, OED)

c. preferring, which seems perhaps best to express the act of volition,

does it not precisely. For though a man would prefer flying to walking,

yet who can say he ever wills it? (1689, CEMET)

d. They should prefer doing anthing [sic] to the ennui of their own

conversation. (1732, OED)

(87) a. and there [he] did confess to me of his own accord his having

heretofore discovered as a complaint against Sir W. Batten, Sir W. Pen

and me that we did prefer the paying of some men to man “The Flying

Greyhound” to others, by order under our hands. (1667-8, PEPYS)

b. Prefer not the righting of your selves before the winning of him by

Love. (1685, OED)

c. Preferringe your plesinge before myne own desire. (1596, OED)

d. For though he has volition, though he prefers his not falling to falling;

yet the forbearance of that motion not being in his power, the stop or

cessation of that motion follows not upon his volition; and therefore

therein he is not free. (1689, CEMET)

Interestingly, prefer not only selects for an ing-complement as direct object but also

for a prepositional object introduced by to, before or some other preposition to mark

the disfavoured second option. With the example in (88) there is some evidence that


this position, too, could be filled by ing-forms in Middle English, albeit very spo-

radically.105 The sixteenth century again fails to yield continuous attestations, how-

ever, so that we have to wait until 1590 for the first Early Modern examples, given

in (89a-b), both of which contain a definite ing-phrase. The first Early Modern in-

stances with bare ing-construction are those in (86b-c) and (87d) above. The devel-

opment here thus roughly parallels that in the first complement position of prefer,

with ing-complements being marginally acceptable before the seventeenth century,

but only catching on around 1650.

(88) a. And in þe bygynnyng, why þu schalt preferre solitarye lyf beforn

lyvynge in felaschepe of men, bysilyche tak hede. (a1450, PPCME)

b. Yf part of eny olde wall or yate sodenly fall, hit of reason owe first to

be made and to be preferred a-fore oþer wallying. (1480, OED)

(89) a. Philanax [...] returnes to his Lorde, and like one that preferred truth

before the maintaining of an opinion, hidde nothing from him (1590,

SIDNEY)

b. Captaine Robert Stewart .. was preferred before the In-taking of

Virtzberg (1637, OED)

As to causation, one cannot fail to observe that the timing of the development

of bare ing-complements with prefer resembles that found for the emotive verbs (see

like, love, hate). Given that prefer is sometimes classed along with the emotive verbs

(e.g. Bladon 1968), we might take this as an indication that it developed along with

the emotives, through semantic analogy. The appearance of to-infinitival comple-

ments with prefer at the end of the seventeenth century, as shown in (90a-b), is cer-

tainly an indication to this effect, as is the use of prefer along with like, love and

hate in the would-like-to-construction later on, as illustrated in (90c).

(90) a. He would prefer to fight with any mean person, if cried up by the volge

[i.e. ‘crowd’] for a tall man. (1639, OED)

105 The OED explicitly lists the Middle English example in (v) under the sense ‘building walls’ (OED s.v. walling), implying that it could be regarded as an ing-complement with prefer. The reading ‘defensive structure’ for wallying seems more plausible, however, in which case it is simply a lexicalised gerund. (v) Yf part of eny olde wall or yate sodenly fall, hit of reason owe first to

be made and to be preferred a-fore oþer wallying. (1480, OED)


b. There is no reproof however pointed, no punishment however severe,

that a woman of spirit will not prefer to neglect (1792, CLMETEV)

c. She would also like to be raised in a pollution-free environment and as

a postscript, would prefer to have an attractive body. (CB)

At the same time, the use of bare ing-complements (as well as other types of ing-

complement) is weakly motivated on independent grounds through (broad/narrow)

paradigmatic analogy, as shown in (91). Examples are by no means abundant, but

this can be thought to be in line with the marginal acceptability of ing-complements

hypothesised for prefer prior to 1650.

(91) a. Crist jugid .. werkis of þe iiije table to be preferid bifore werkis of

worschiping. (c1443, MED)

b. Thei .. vsed not to renne to wrastillingis, ber-baytingis, and swech oþir

onthrifty occupaciones, whech summe men now on dayes preferr be-

for dyuyne seruyse. (1451, MED)

c. yet the Lord President preferred the Love to his Countrie, and the

Service of his Prince, before any particular Love to any Person (1592-

1603, PPCEME)

d. An indiscreet builder, who preferreth the care of his frontispice before

the maine foundation. (1630, OED)

e. The Neapolitane Gentry, who stand so on the puntoes of their honour,

that they preferre robbery before industry. (1642, OED)

BEAR: Pepys’ Diary yields the first ing-complement with bear, as shown in (92a),

but it is not until some thirty years later, around 1700, that ing-complements start to

turn up regularly in the data, witness the examples in (92b-e). At this point, passival

ing-complements, as in (92b-c) are clearly in the majority, representing the primary

use of the pattern in eighteenth-century English. As (92a, d-e) show, however,

clausal ing-complements with normal subject-control are also found from the begin-

ning and are typically passive, with the exception of (92e) which contains the only

non-passive ing-complement among the early examples (though still with the form

being, reminiscent of the passive; see footnote 99 above).

(92) a. away goes Alce, our cooke-mayde, a good servant, whom we loved

and did well by her, and she an excellent servant, but would not bear


being told of any faulte in the fewest and kindest words (1666,

CEMET)

b. I would summ up the Particulars of this Second Head, if the

Examiner’s Performance could bear recapitulating. (1699, OED)

c. They used a sweet fluent kind of Rhetorick .. which .. serv’d only to

put a present good Face upon an Argument, but would not bear

Scanning [i.e. ‘close examination’]. (1704, OED)

d. I can bear being told that I am in the wrong, but tell it me gently.

(1711, OED)

e. a Fellow that is capable of shewing an impudent Front before a whole

Congregation, and can bear being a publick Spectacle, is not so easily

rebuked as to amend by Admonitions. (1711, Visser 1963-73: 1877)

The ing-complements with bear most probably result from a combination of para-

digmatic and semantic analogy. With respect to narrow paradigmatic analogy, the

seventeenth-century data show bare abstract nouns allowing the same passival inter-

pretation as the majority of early ing-complements (see (93)). The examples are rela-

tively few, however, especially considering the fact that bear is a highly frequent

verb. Even so, judging by the instances attested, the construction appeared slightly

earlier in the language than did the ing-complements and may therefore have served

as a model.

(93) a. Thy great imployment Will not beare question. (1605, OED)

b. The remainder can hardly beare such deminution, as all Armies are

subiect vnto. (1617, OED)

c. The demons are too proud to bear contempt without revenge, when

God permits them. (1688, CEMET)

d. No Modern Latin can bear criticism. (1697, OED)

As to semantic analogy, the pattern of bear with ing-complement is foreshadowed

by the use of ing-complements with the verbs abide and endure earlier in the seven-

teenth century.106 Some representative examples are given in (94), which immedi-

106 For both abide and endure, the most plausible source for ing-complements is narrow paradigmatic analogy, though the evidence is not abundant: (vi) a. The Greeke Article is so placed, as it can abide no such patcherie.

(1579, OED)


ately reveal important similarities to the pattern with bear, sharing with it the collo-

cationally biased use of negators and modal auxiliaries with the main verb, as well

as the passival interpretation of the ing-complement.

(94) a. This Eagles feathers will not abide blending with others. (1622, Visser

1963-73: 1877)

b. they will endure handling, which before would not abide looking on

(1623 (?1599), CEMET)

c. To enquire .. what Metals endure Subliming; and what Body the Sub-

limate makes. (a1626, OED)

d. You may have occasion sometimes to Braze .. a piece of work; but it is

used by Smiths only, when their work is so thin, or small, that it will

not endure Welding. (1677, OED)

Still, neither mechanism can easily account for the surprising fact that the first in-

stance of bear with ing-complement, (92a) above, is not of the indeterminate pas-

sival type but is unambiguously clausal. In fact, both paradigmatic and semantic

analogy would seem to predict otherwise, since both would primarily trigger inde-

terminate passival ing-complements. As the indeterminate passival type is quantita-

tively better represented in eighteenth-century English, and the first clausal instances

are all attested in rather informal prose that may be considered ahead of its time if

judged by the standards of formal writing (Pepys’ Diary for (92a), Lady Montague’s

Letters in (92d), and Addison and Steele’s Spectator in (92e)), the precedence of a

passive clausal instance over the expected passival indeterminate examples might be

discounted as no more than a peculiarity of the corpus data. On the other hand, we

may also see the apparently immediate transition to clausal ing-complements as a

sign of the increasing productivity of ing-complementation that is going to be char-

acteristic of the Late Modern period (see below).

3.3. Stage III: indirect paradigmatic analogy

Stage III in the diffusion of ing-complements runs from 1666 to 1736 and sees the

appearance of ing-complements with finish, propose, give up, keep, help, go on, in-

tend, stop, keep on, remember and mind. Compared to Stage II, ing-

b. That he might bee hable with a paciente and contentefull mynde, to

endure banyshemente. (1542, OED)


complementation in Stage III gains further productivity, which manifests itself in a

productive regularity that holds over and above specific clusters of ing-

complementation constructions. In particular, Stage III witnesses a number of inno-

vations that cannot be accounted for on the basis of narrow paradigmatic analogy or

semantic analogy. Instead, they manifest a first phase of broad paradigmatic anal-

ogy, namely indirect paradigmatic analogy.

The most relevant innovations in this respect are the introduction of ing-

complements with remember and mind at the end of Stage III. These verbs do not

straightforwardly fall in the semantic groups established earlier. However, at the

time of their appearance with ing-complements, they collocated with noun phrases

implying some action, event or situation either controlled by the matrix subject or

some other agent (e.g. don’t mind my fatigue). Although these collocate noun

phrases did not take the form of bare action nouns or abstract nouns, they could indi-

rectly trigger the use of bare subject-controlled ing-complements. That is, ing-

complements do not appear all at once: the appearance of bare ing-complements is

preceded by the use of another type of ing-complement – in the case of mind and

remember, different-subject ing-complements. I see two (mutually compatible) rea-

sons for the temporal precedence of non-bare ing-complements over bare ing-

complements. First, the former more closely resemble the paradigmatic sanctioning

model – that is, a noun phrase with explicit determiner. In the case of mind and re-

member, different-subject complements are triggered by their resemblance to other

direct objects used with the same verbs (a first instance of paradigmatic analogy),

and in turn they can then trigger bare controlled ing-complements which already

alternate with different-subject complements in the context of other verbs (a second

instance of paradigmatic analogy). Second, the order of appearance may be another

manifestation of Fanego’s (2004a) hierarchy of relative nominality (see Chapters 4

and 8), with the more nominal-looking complement-types being sanctioned more

easily in environments that otherwise only allow noun phrases.107

The introduction of ing-complements with remember and mind present pure

cases of indirect paradigmatic analogy. Leading up to this, there are a number of

other extensions of ing-complementation in which indirect paradigmatic analogy

probably played a central (if not the only) role. These innovations show the same

pattern of non-bare ing-complements being selected before bare ing-complements.

107 Fanego’s (2004a) hierarchy is here of course used not to explain the emergence of verbal features within gerunds, but to account for the external behaviour of differ-ent gerund constructions. Otherwise, however, the argument is similar, with the ranking from more to less (explicitly) nominal playing its role in a change affecting canonically nominal environments.


Thus, with finish and propose, definite ing-complements precede bare ones, whereas

with stop the appearance of bare ing-complements is foreshadowed by different-

subject ing-complements. With propose, however, the use of ing-complement may

in part have been furthered by French influence. With finish and stop, indirect para-

digmatic analogy may have been assisted by semantic analogy to the aspectual

verbs, even though semantic analogy was probably relatively weak, seeing that the

original lexical semantics of finish and stop are less closely related to aspectual

meanings (moreover, the use of different-subject complements with stop is entirely

inexplicable on the basis of semantic analogy with the aspectuals).

Not all is indirect paradigmatic analogy, however. Stage III shows further

evidence of the importance of independent lexical developments. Give up, keep and

help, for one, underwent lexical changes that brought them semantically closer to

some sanctioning model (the aspectuals in the case of give up and keep; the negative

implicative verbs in the case of help). The subsequent appearance of ing-

complements with these verbs is probably due to semantic analogy, while the fact

that semantic analogy is later to affect these verbs than other semantically-related

verbs is explained by the independent lexical change they had to undergo in order to

become subjected to semantic analogy in the first place. A different scenario is evi-

denced in go on and keep on. Go on first came to be used with participial adjuncts of

manner, which were subsequently reinterpreted as complements, possibly on the

model of the ing-complements found with continue (and other aspectuals). Keep on

first underwent independent lexical change that caused it to resemble go on; ing-

complements then appeared with the intermediate step of a reinterpretation akin to

that witnessed in go on. The appearance of ing-complements with intend, finally,

might be a first case of broad paradigmatic analogy (without the mediation of indi-

rect paradigmatic analogy), announcing the mechanism of change characteristic of

Stage IV, though the case is not a pure one, given the possible influence on com-

plement selection of the nominal counterpart of intend, the noun intention.

FINISH: The first example of finish with bare ing-complement again appears in

Pepys’ Diary and is given in (95a). The next attested examples are (95b-c), which

are at once the first instances whose ing-complement shows clearly clausal syntax.

Strikingly, in the seventeenth-century data definite controlled ing-complements, as

illustrated in (96), are more common than bare ing-complements, especially in

Pepys’ Diary, even though by the eighteenth century, this preference for definite

ing-complements has again disappeared (cf. Tables 9.1 and 9.2).


(95) a. and then to Knipp’s again, and there staid reading of Waller’s verses,

while she finished dressing, her husband being by. (1666, CEMET)

b. when you have finished placing the Earth about them, you may – if the

Land be stiff, cold, wet, or barren – cover the Earth with Dung, round

about the Tree (1696, PPCEME)

c. I had but just finished taking a copy of this, and laid the letter where I

had it, [...] when she came up in a great fright, for fear I should have

seen it (1740, CLMETEV)

(96) a. and so away home to my business at the office, and then home to sup-

per and to bed, after having finished the putting of little papers upon

my books to be numbered hereafter. (1666, CEMET)

b. to supper with my wife, and then to finish the writing fair of my ac-

counts, and so to bed. (1666, CEMET)

c. Finish the gathering and drying of your Hops; cleanse the Poles of the

Hawm. (1669, OED)

It is plausible that bare ing-complements arose under the influence of semantic anal-

ogy from the group of aspectuals. In support of this is the fact that alongside bare

and definite ing-complements we also find the repeated-subject complements typical

of aspectuals (cf. continue, leave, give over etc.), again in Pepys’ Diary, as shown in

(97) (but note that in (97a) trimming might also be lexicalised and simply mean

‘costume’). In addition, the use of bare ing-complements may have been furthered

through indirect paradigmatic analogy from the presence of definite ing-

complements in the same environment, though there is no further evidence for this.

(97) a. After that to finish my trimming, and while I was doing of it in comes

Mr. North very sea-sick from shore, and to bed he goes. (1660,

CEMET)

b. and so by night home, where to my chamber and finished my pricking

out [i.e. ‘notating’] of my song for Mr. Harris (“It is decreed”) (1668,

CEMET)

The early preference for definite ing-complements is of course also in need of an

explanation. As illustrated in (98a-b), finish can combine with definite direct objects

implying an action by the matrix subject. The first definite ing-complements may

have been modelled on this use through narrow paradigmatic analogy, although,

admittedly, the definite objects in question are not very frequent. Some interesting


supportive evidence comes from Pepys’ usage, where a certain collocational paral-

lelism can be observed between definite direct objects and definite ing-

complements, in that both tend to be used in the context of paperwork – compare

(98c-d) to (96a-b) above.

(98) a. He beganne the foundation of a new colledge at Winchster, and in sixe

yeares finished the same (1580, PPCEME)

b. as God is described by Plato to have done when he had finished the

creation of the world, and saw his own orbs move below him (1656,

CEMET)

c. Up betimes and to my office, and anon we met upon finishing the

Treasurer’s accounts. (1663, CEMET)

d. So home by water, and there hard till 12 at night at work finishing the

great letter to the Duke of Yorke against to-morrow morning (1666,

CEMET)

PROPOSE: The history of propose presents another complicated picture. Let us start

by observing that, as pointed out by Fanego (2007: 182-4), propose occurs in two

relevant senses, the almost obsolete sense ‘plan, intend’ and the now normal sense

‘put forward (to others) as a plan’. In the intentional sense propose naturally com-

bines with a controlled complement, but in the communicative sense the control

relationship may take various forms: the action proposed may be construed as car-

ried out by the subject (control proper), by the subject together with the implied ad-

dressee or a third party (involvement), or by a third party to the exclusion of the sub-

ject (no control) (see Section 2 above). These distinctions must be borne in mind

when considering the emergence of ing-complements with propose.

The first-attested bare ing-complement with propose occurs in Pepys’ Diary

and is given in (99a): the sense of the verb here is the communicative one ‘put for-

ward as a plan’ and the matrix subject is involved in the action expressed by the ing-

complement. The next attested examples with bare ing-complement are given in

(99b-c): in both examples propose is again used meaning ‘put forward as a plan’. As

to control, whereas in (99c) the matrix subject is involved in the action denoted by

the ing-complement, (99b) allows any reading (that is, control, involvement, or no

control). As shown by (99d-e), the first bare ing-complements with propose in its

intentional sense appear in the first half of the eighteenth century, in another source

of relatively informal register, Walpole’s Letters.


(99) a. it being a fine clear day, I did, ‘en gayete de coeur’, propose going to

Bow for ayre sake, and dine there, and so W. Batten and I (setting W.

Pen down at Mark Lane end) straight to Bow (1666-7, CEMET)

b. also sevll of ym made speeches yt it was a shame yt Strangrs should

Come into ye bowells of ye City to Insure ye houses from fire & pro-

posd insuring ‘em at ye same rates as ye others & ye chambr of Lond

to be ye fund (1681, Fanego 2007: 184)

c. In a few Days my old Lady [...] thought of turning me out of Doors,

because I put her in mind of her Son. Sir Stephen proposed putting me

to Prentice; but my Lady being an excellent Manager, would not let her

Husband throw away his Money in Acts of Charity. (1711, SPECTA-

TOR)

d. I am so sick of it all, that if we are victorious or not, I propose leaving

England in the spring (1741, CLMETEV)

e. Ceretesi tells me that Madame Galli is dead [...]. I am quite sorry for

Madame Galli, because I proposed seeing her again, on my return to

Florence (1743, CLMETEV)

Things get more complicated if we consider also other types of ing-complement.

Definite ing-complements seem to emerge earlier than bare ing-complements, mak-

ing their appearance in the data with the example in (100a), dated 1606, followed

fifty years later by the one in (100b). In Pepys’ usage, illustrated in (100c-e), defi-

nite ing-complements are decidedly more frequent than bare ing-complements,

which can probably be taken to reflect the fact that the former pattern is the older

one. As illustrated in (101), Pepys’ Diary also provides us with a good number of

ing-complements with an explicit subject, which can either refer back to the pro-

poser, or refer to the addressee or some third party. As for interpretation, the mean-

ing of propose in (100a) remains unclear without further context, but the other ex-

amples in (100) show propose in both its intentional sense and its communicative

sense (see (100b, d) and (100c, e) respectively) and either with straightforward con-

trol (see (100b, d)), involvement (see (100e)), or ambiguity between involvement

and no control (see (100c)). The examples with explicit subject in (101) exclusively

associate with the communicative sense of propose and give overt expression to the

various possible control interpretations of the subjectless constructions.

(100) a. They have there [Rome] newly proposed .. the processing of the Duke

by way of Inquisition. (1606, OED)


b. as I propose not the gaining of credit by this undertaking, so I would

not willingly lose any part of that to which I had a just title before I be-

gan it (1656, CEMET)

c. Among others arguing with the Commissioner about his proposing the

laying out so much money upon Sheerenesse (1665, CEMET)

d. While W. Batten and I were alone, we had much friendly discourse,

though I will never trust him far; but we do propose the getting “The

Flying Greyhound,” our privateer, to us and W. Pen at the end of the

year when we call her home, by begging her of the King, and I do not

think we shall be denied her. (1666-7, CEMET)108

e. I to the office, whither Creed come by my desire, and he and I to my

wife, to whom I now propose the going to Chetham (1667, CEMET)

(101) a. Mr. Warren proposed my getting of £100 to get him a protection for a

ship to go out, which I think I shall do. (1665, CEMET)

b. he proposes his and my looking out into Scotland about timber (1666,

CEMET)

Concerning mechanisms of change, it is to be noted that French proposer was

relatively common with non-finite complement and by the seventeenth century oc-

cured in the two relevant senses ‘intend, plan’ and ‘put forward as a plan’, as shown

in (102).

(102) a. Le desir d’obeyr à sa maistresse eut tant de pouvoir sur Renaud, qu’à la

fin il se resolut à s’enfermer dedans la prison de Floran, proposant de

se descouvrir dans peu de jours au concierge. (1615, FRANTEXT)

‘The desire to obey his mistress had such power over Renaud that in

the end he resolved to have himself locked up in the prison of Floran,

intending to make himself known to the guard after a few days.’

b. Cent fois ma mere m’a proposé de quitter le sejour de nostre ville pour

aller à Torrillos (1625, FRANTEXT)

‘A hundred times my mother has proposed to me to leave the abode of

our city to go to Torrillos’

108 The definite article in this example is only found in the 1974 edition of Pepys’ Diary by Latham and Matthews.


Consequently, the use of a non-finite clause with English propose may have been

modelled on French. In this, the choice for the to-infinitive to translate the French

construction may have been hindered by the vague control relationship characteristic

of propose in its sense ‘put forward as a plan’ (Fanego 2007).109

Even so, the appearance of ing-complements remains surprising, because it is

sanctioned neither by semantic nor by narrow paradigmatic analogy. Of the ing-

complement-taking verbs in the 1660s, decline and deny (in the sense ‘refuse’) are

the only ones semantically related to propose, at least in as far as their semantic

structure presupposes some proposal, yet the relation is somewhat distant and per-

haps far-fetched. In any case, there is no additional evidence that these verbs might

have analogically influenced propose. Collocationally, propose is not found combin-

ing with bare abstract nouns on which the use of ing-complements might have been

modelled, so narrow paradigmatic analogy is similarly unlikely. As seventeenth-

century usage shows a clear preponderance of definite ing-complements with pro-

pose, it is conceivable that bare ing-complements simply supplanted the definite ing-

complements. In the end, this last possibility offers the most plausible account, since

the two constructions are in fact functionally equivalent in the environment of pro-

pose (comparing the examples above, (99a, c) closely match (100e), (99b) matches

(100c), and (99d-e) match (100b, d)).

In turn, this leads us to the question of what gave rise to the use of propose

with definite ing-complements. Semantic analogy, being equally problematic for

definite as for bare ing-complements, does not offer any ready solutions. At the

same time, it is clear that propose combines naturally with definite patient argu-

ments referring to an action by the subject, the addressee, a third party, or any com-

bination of these participants. The patient is often pronominal and most commonly

precedes the verb propose, either as a relativised object, or as the subject in a pas-

sive clause, as in (103a-b), though exceptionally, the patient can be non-

pronominalised and does not precede the matrix verb, as in (103c-d). Such objects

testify to the semantic compatibility between propose and an object noun phrase

referring to an action.

109 Specifically, Fanego (2007) suggests that to-infinitival complements in monotransitive constructions strongly tend to imply subject-control, which easily clashes especially with the new meaning of propose ‘put forward to others as a plan’. I would add that to-infinitives are furthermore disfavoured when used in a construction with explicit addressee argument for propose because of horror æqui, since the typical surface sequence in such an environment is propose to NP to VP, with repetition of to.


(103) a. whatsoever, upon debate of the senate, is proposed to the people, and

resolved by them, is enacted by the authority of the fathers (1656,

CEMET)

b. the expedition, which was proposed to him, was the most glorious that

the Portuguese could undertake (1688, CEMET)

c. And in like manner, he proposeth the Consideration as well of the

Earths Aphelium and Perihelium, as of the Æquinoctial and Solsticial

Points, in order to the finding a Reason of the Annual Vicissitudes

(1666, LC)

d. he was the man that did propose the removal of the Chancellor (1667,

CEMET)

In the absence of semantic or narrow paradigmatic analogy as explanatory mecha-

nisms of change, it appears that the appearance of definite ing-complements with

propose (though still helped forward somewhat by foreign influence) marks a point

of increasing productivity, that is, an increasing awareness on the part of language

users that ing-complements can readily serve as clausal complements with transitive

verbs that can select an action as their patient argument. That is, the mechanism un-

derlying paradigmatic regularity in the distribution of ing-complements is beginning

to shift from narrow to broad paradigmatic analogy.

GIVE UP: As (104) shows, the rise of ing-complements with give up took off at the

end of the seventeenth century and gathered momentum in the latter half of the

eighteenth. The emergence of ing-complements must be seen in light of the history

of the lexical item give up and its relationship to the near-synonymous give over.

Semantically, give up developed a variety of meanings in close parallel to give over.

The development most immediately relevant is that from the original sense ‘hand

over’, as in (105a-b), to ‘abandon’ and ‘cease from’, as in (105c-d) – a change com-

parable to the one witnessed earlier in give over (see above) – but give up and give

over also came to share a number of other uses, such as the use in (106), where both

verbs combine with reflexive pronouns and mean ‘devote/abandon oneself to’.

Given this high degree of overlap in meaning and syntactic behaviour, the appear-

ance of ing-complements with give up can be seen as a further instance of conver-

gence through semantic analogy. It is worthwhile at last to point to the subsequent

history: as give up with ing-complement rose in frequency, the pattern with give


over disappeared, strongly suggesting replacement of one by the other, and thus

bearing further witness to the high degree of overlap between the two verbs.110

(104) a. But after I was arrived at the fifteenth year of my age, or thereabouts, I

began to shoot up, and gave not up growing till I had attained the mid-

dle size and stature of men. (1683, CEMET)

b. I told the person who told me this, that, on the contrary, you seemed,

by your letters to me, to be extremely pleased with Lord Albemarle’s

behavior to you: but that you were obliged to give up dining abroad

during your course of experimental philosophy. (1751, CLMETEV)

c. The price of silver in the European market might, perhaps, have fallen

still lower, and it might have become necessary either to reduce the tax

upon it [...] or to give up working the greater part of the American

mines which are now wrought. (1751, CLMETEV)

d. after giving up searching, Mr. Henry Davis saw his fingers on his hat,

and took his hat from him, and searched it, and found the chief of the

gold now in question (1786, POB)

(105) a. Mr. Wotton gave up his secretaryship, and mr. Cicil toke it. (1550-2,

PPCEME)

b. For which I doe discharge you of your office, Giue vp your keyes

(1623 (1603-4), CEMET)

c. but that she had already given her honour up, it would have been

something more surprising: but she was used to questions of that na-

ture, and therefore received this with so much the less concern (1684,

CEMET)

d. and so home, and there did resolve to give up my endeavours for ac-

cess to the leads, and to shut up my doors lest the being open might

give them occasion of longing for my chamber, which I am in most

fear about. (1662, CEMET)

(106) a. Why take they such paines to abandon and put out from their harts all

sense, all tast, all feeling of religion? but only to this end and purpose,

110 The example in (vii) shows that give up could very occasionally combine with ing-complements in its sense ‘abandon, forbear’. In this use it may be taken to fall under the negative implicative verbs. (vii) You will have had a letter from me by this time, to give up sending the

Dominichin by a man-of-war, and to propose its coming in a Dutch ship. I believe that will be safe. (1742, CLMETEV)


that they may without inward remorse and grudging of conscience giue

over themselues to all vncleanenes. (1614, PPCEME)

b. At noon he dined with me, and we sat all the afternoon together, dis-

coursing of ways to get money, which I am now giving myself wholly

up to (1662, CEMET)

KEEP: In the first two examples with keep followed by an ing-complement, given in

(107a-b), the ing-form in question is waking and might be best interpreted as purely

adjectival and lexicalised, meaning ‘awake, attentive’ – witness the use of waking in

(107c). The ing-complements in (108), by contrast, are likely to be clausal (or can be

interpreted as such), so that it seems safe to date the appearance of ing-complements

with keep to the end of the seventeenth century.111

(107) a. it will concern him then to keep waking, to stand in watch, to set good

guards and sentinels about his received opinions (1644, CEMET)

b. thy unwearied diligence Has still kept waking (1682, Visser 1963-73:

1898)

c. He was thought negligent, and therefore they sent a wakinger spirit.

(1601, OED)

(108) a. we kept walking to keep us warm (1683, CEMET)

b. You keep soaking in Taverns, and come and make such Complaints to

me. (1687, OED)

c. when he walk’d, he kept Cringing on his Larboard Quarter, not pre-

suming to go Cheek by Jowl with one of the Representatives of the Na-

tion (1700, LC)

Various mechanisms may have contributed to the appearance of ing-complements

with keep. First, crucial to the timing of change is the lexical development of keep

and, in particular, the emergence of copulative uses of keep around 1600, as illus-

trated in (109). This copulative use sanctioned predicative adjectives, as in (109a),

and, as is evident from the examples in (107a-b) above, adjectival participles. From

these it was probably only a small step to sanctioning participle clauses proper,

through the paradigmatic tie between participles and adjectives. Second, the copula-

111 Visser (1963-73: 1898) also gives the following Middle English example, dated 1391: Kep bydding ay, and lyf clenly. This example, however, differs from present-day usage, in that keep most probably means ‘respect, abide by’ and bidding is a gerund form.


tive use attracted a variety of constructions to fill the subject complement slot in

close parallel to the object-complement slot of the older complex-transitive (causa-

tive) construction with keep. For instance, constructions filling both subject-

complement and object-complement slots with keep included gerund clauses with

from (compare (110a-b)) or certain specific adjectives such as close (compare (110c-

d)). Given the occurrence of participle clauses in the complex transitive construction

(see (111)), it is conceivable that the same parallelism between the copulative and

complex transitive constructions inspired the use of participle clauses with copula-

tive keep. Collocational similarities between participle clauses in either use testify to

this, most prominently the frequent use of going, as illustrated in (112). A further

feature found across the two constructions is the use of ing-forms with an a-prefix,

as in (113), again underscoring constructional parallelism.112

(109) a. This seruitude makes you to keepe vnwed. (1590, OED)

b. after diner I went about the house, and kept with my Maides till all

most night (1599-1601, PPCEME)

(110) a. all Cœlestial Bodies whatsoever, have an attraction or gravitating

power towards their own Centers, whereby they attract [...] their own

parts, and keep them from flying from them (1674, LC)

b. However, I could not keep from peeping at them, and there I saw him

again (1722, DEFOE)

c. the lippes [...] should be to the mouth as a doore to a house , and to

keepe the mouth close tyl the meate were kindly chewed (1548,

PPCEME)

d. Again Captain Boanerges sendeth his trumpeter to Ear-gate, to sound

as before for a hearing; but they again kept close, came not out, nor

would they give him an answer (1682, CEMET)

(111) a. Tye vp the Libertine in a field of Feasts, Keepe his Braine fuming.

(1623 (1603-7), CEMET)

b. and as soone as your horse hath past ouer his course, and is in a hie

sweat, you shall instantly haue him home, and there lay more cloaths

112 Visser (1963-73: 1899) proposes that the a-prefix is the reduced form of on in constructions with keep on and ing-complement (see below keep on). This is unlikely, however, seeing that all examples of keep on with ing-complement post-date the use with a-prefix. It seems more plausible that the a-prefixes extended from the ing-forms used with complex-transitive keep, which in turn may have been mo-tivated by the use of a-prefixes in other object-controlled participle clauses, in par-ticular those with the verb set.


vpon him, and keepe him stirring till hee haue sweat so in the stable an

houre or more (1615, PPCEME)

(112) a. the Factor will let them have no more money then what will suffice to

keep their Trade going (1681, LC)

b. that we might lose no time, we kept going (1719, Visser 1963-73:

1898)

(113) a. With lesse paines to keepe agoing that which he had moved, and set a

going. (1583, OED)

b. he keeps a puffing and a blowing (1606, Visser 1963-73: 1899)

Third, copulative keep not only underwent the analogical influence of complex-

transitive keep, it also shared characteristics with continue – another verb with copu-

lative uses (see continue above) – and may have been reshaped after its model. A

connection between the two verbs is evident from the occasional influence exerted

by keep on continue. For example, the object-controlled participial construction with

continue in (114) is almost certainly inspired by the analogous pattern found with

keep illustrated in (111) above. In this light, it is plausible that influence also ran in

the opposite direction and that the use of subject-controlled ing-complements with

keep may have been modelled after the ing-complements that were already in com-

mon use with continue (and whose status as participial or gerundial may have been

sufficiently uncertain to analogically sanction the participial ing-complements with

keep; see continue).

(114) This Lady Sands continues her Clack going ever since (1650, LC)

HELP: The examples in (115a-c) represent the first attestations of the verb help with

an ing-complement. The pattern’s appearance can thus be dated to the last two dec-

ades of the seventeenth century. As the examples immediately make clear, the use of

help with ing-complements is very strongly tied to the negated can construction, so

much so, in fact, that the pattern is not attested outside this construction – (115d) is

the closest we get to a counter-example. Interestingly, the first instances of help with

ing-complement succeed by one or two decades the emergence of a similar construc-

tion, illustrated in (116), in which help combines with can or could in a negative or

interrogative clause and takes as its object the pronoun it. It seems very plausible

that the recurrent string cannot help, on top of the negative implicative semantics of

the construction, provided a more than easy target for semantic analogy to introduce


ing-complements, with the negated can construction found with other negative im-

plicative verbs serving as the model.

(115) a. and though she could not help flattering both, while by; yet she ever

loved the absent best (1684, CEMET)

b. but when he talked freely, he could not help letting himself out against

the liberty that under the Reformation all men took of inquiring into

matters (1683-1713, PPCEME)

c. I cannot help defending an Opinion in which now I am more con-

firm’d, that probable conjectures may be made of the ingenious Dispo-

sition of the Mind, from the fancy and choice of Apparel. (1692,

CEMET)

d. But the worst of it is [...] that I don’t see how I am to help disliking him

doubly henceforward (1894, CLMETEV)

(116) a. I would not throw myself away upon this fool, if I could help it. (1662,

CEMET)

b. Suppose we are as bad as you bespeak us; how can we help it? (1674,

LC)

c. But if this will not do, we cannot help it. (1680, PPCEME)

GO ON: The first ing-clause found following go on, given in (117), is doubly am-

biguous. On the one hand, go on might be interpreted as an intransitive phrasal verb,

meaning ‘move on, continue’, as in (118), with the ing-clause functioning as a man-

ner adverbial. On the other, the example might instantiate the prepositional pattern

go on NP meaning ‘undertake, start’, as illustrated in (119).

(117) He him selfe went on wayfarynge frome place to place. (1548, OED)

(118) a. Pray, said I, Madam, be so free and ingenous with me, and so kind to

yourself, as if there be any other Person that is concealed in any part of

your House, [...] as to deliver him up, and you shall come to no further

trouble. She denied it, and said, I know nothing of them: But we went

on, and searched, and at last discovered the other man Nelthorp, hid in

a Hole by the Chimney. (1685, PPCEME)

b. I go on then; Whatever doth desire to subsist and endure, doth also

desire Unity (1695, PPCEME)

(119) a. Some tyme thei wylle goo on pylgremage, somtyme they wyll be

recluse. (c1491, OED)


b. This Action I now goe on, Is for my better grace. (1611, OED)

c. And on a tyme goyng on huntyng, when he had lost his people, he was

destroyed of Wolues. (1568, OED)

These ambiguities make it difficult to date the appearance of ing-complements with

aspectual go on, meaning ‘continue, keep’. The examples in (120), for instance, all

allow an adjunct reading for the ing-clause, while those in (121) exclude the adjunct

reading but might instantiate the prepositional pattern go on NP.

(120) a. Marke how well shee singeth .. And goeth on alwayes continuing her

songe. (1583, OED)

b. Then he vouched one Appleyard’s Case, a Traitor in Norfolk, who

said, a Man must have two Accusers. [...] but Mr. Justice Cotlin said,

that that Statue was not in force at that Day. [...] Then he went on

speaking of Accusers, and made this difference: [...] (1600, PPCEME)

c. Thus, therefore, he now went on bewailing his sinful sleep, saying, O

wretched man that I am that I should sleep in the day-time! (1678-84,

CEMET)

(121) a. they began according to their Predecessors footsteps, to pretend they

might by the Prerogative thereof give us stones instead of bread, and

Serpents instead of fish, Mat.7. Go on wasting of our blood and treas-

ure, without securing us in the liberty either of our own consciences,

persons, or Estates (1653, LC)

b. The Wolfishness of those which .. ought to have been the Christian

Pastors, but went on scattering their Flocks, if not devouring. (1676,

OED)

c. We shall goe on buildeing to, as soon as spring begins. (1677, OED)

The examples in (122), however, look like unambiguous instances of aspectual go

on with ing-complement, indistinguishable from the pattern that is in common use

today. The ing-clauses in these examples cannot be interpreted as denoting a manner

of continuation (they denote that which is continued), nor can the semantics of go on

NP ‘start, undertake’ be read into the examples. I therefore take it that the pattern of

aspectual/continuative go on with ing-complement emerged at the beginning of the

eighteenth century at the latest (implying that five of the six examples listed in Table

9.1 above are questionable).


(122) a. As soon as I observ’d the Political posture of their Affairs, [...] and

remembring what is said for our Instruction, That a Kingdom divided

against its self cannot stand; I ask’d the Old Gentleman if he had any

Estate in that Country? He told me, no great matter; but ask’d me why

I put that Question to him? Because, said I, if this People go on fight-

ing and snarling at all the World, and one among another in this man-

ner, they will certainly be Ruin’d and Undone, either subdu’d by some

more powerful Neighbour; [...] or else they will destroy and devour

one another. (1705, CEMET)

b. Absorbed in looking at the lady, Sir Martin foolishly goes on opening

and shutting his mouth and fumbling on the lute after the man’s song, a

version of Voiture’s ‘L’Amour sous sa Loi’, is done. (1711, SPECTA-

TOR)

c. This well-managed Officer of yours has, to my Knowledge, been the

Ruin of above five young Gentlemen besides my self, and still goes on

laying waste wheresoever she comes, whereby the whole Village is in

great danger. (1711, SPECTATOR)

The mechanisms of change that gave rise to the use of go on with ing-

complement have already been hinted at in the preceding discussion. The participial

adjuncts found with intransitive go on may have invited a reinterpretation by which

they came to be perceived as complements to an aspectual verb. What happened,

then, is that the manner of continuation was semantically reinterpreted as that which

is continued (cf. the development of integrated participle clauses; Chapter 10). A

reinterpretation of this kind is the more likely because it is analogically supported by

at least two other constructions. The first is a construction involving prepositional

complements with go on in the form of a gerund introduced by in, as in (123), simi-

larly denoting that which is continued (paradigmatic analogy). The second is the use

of ing-complements with continue, which as a potentially intransitive verb to some

extent resembled go on, not only semantically but also distributionally – for exam-

ple, in being used to mark a speaker’s progression in discourse (compare (118b)

above to (124a)) and in being used with prepositional complements with in, (com-

pare (124b-c)) (semantic/distributional analogy). Note that the role of analogical

influence is supported by the fact that other developments in go on are similarly

marked by convergence with continue – consider the use of aspectual go on with to-

infinitive in (125a), or the use of go on to mark duration in (125b-c) (in parallel to

continue in (125d)).


(123) a. Go on in drinking, whoring, .. and dicing, hating and malicing, fretting

and chafing. (a1640, OED)

b. They went on in ingrossing the militia. (1650, OED)

(124) a. “What could urge him,” continued he, “to so prompt a resolution? [...]”

(1688, CEMET)

b. he would go on, he was resolved to go on in his wickedness. (1680,

CEMET)

c. their continuing in their rebellion did but chafe and heat the spirit of

the captains (1682, CEMET)

(125) a. If you go on to learn at this rate, you will soon puzzle me, in Greek

especially. (1739, OED)

b. This permixture going on for some few yeares. (1604, OED)

c. The souring process had been going on for two days. (1756, OED)

d. We had not been a day at sea before we had a storm begun, that con-

tinued two days and two nights in a most violent manner (1676,

CEMET)

INTEND: As the first apparent example of intend with ing-complement given in

(126a) is doubtful (writing and cyphering are probably best regarded as lexicalised

ing-forms denoting scholarly acquirements rather than the verbal processes of writ-

ing and ciphering as such), the appearance of ing-complements with intend must be

dated to the beginning of the Late Modern period, starting with the examples given

in (126b-d).

(126) a. the Scholars which intend writing or cyphering, or the like, may go to

the Writing-schoole, as they yet use to doe about London. (1660,

PPCEME)

b. After we have got the castle, I hear the fleet will go for the Islands of

Minorca and Majorca, and, after that, I hope home. If the St George

should not do, intend asking Sr Jno Leake leave for my self. (1706,

PPCEME)

c. I intend being at London .. to attend the Mortmain Bill. (1736, OED)

d. To prevent his design being pirated, he intends petitioning the Parlia-

ment. (1754, OED)


At the time of their appearance, there is no close semantic analogue on which the

ing-complements with intend could have been modelled, which means that straight-

forward semantic analogy is unlikely as the mechanism responsible for the change.

On the other hand, looking at the more general collocational behaviour of intend, it

is found that while intend readily combines with objects denoting or implying some

action by the matrix subject, those objects tend not to take the form of bare abstract

nouns, but are typically marked with either a definite or indefinite determiner and

refer to specific actions. This collocational behaviour is illustrated in (127)-(128):

two rare instances of intend with bare abstract noun are given in (127), whereas

(128) is representative of the more common object type with determiner and with

specific reference. On quantitative grounds alone it seems unlikely that the use of

ing-complements with intend would have been modelled on the few bare abstract

nouns found with intend, but note further that the ing-complements in question do

not even resemble such bare abstract nouns, since all ing-complements refer to spe-

cific actions (see (126) above). To the extent, then, that paradigmatic analogy has

played a role in the appearance of ing-complements with intend, the analogical in-

fluence must have come from the direct objects with specific reference. The occa-

sional appearance of a definite controlled ing-complement, as in (129), can be taken

as evidence to this effect.

(127) a. The Sting of a Bee [...] is as great an Instance, that Nature did realy

intend revenge as any (1665, PPCEME)

b. by their Instinct they are urged to intend and pursue bodily Delight.

(1695, PPCEME)

(128) a. I understand that the Erle of Essex, with a great Rout, intendeth the

Conquest of the North (1592-1602, PPCEME)

b. Fathers will speak loudest .. and look angerliest, that intend not the

severest correction. (a1631, OED)

c. We hear the French are breaking ground, as if they intended a formal

Siege. (1678, OED)

d. I intended only a visit hither, not a continuance (1683, CEMET)

(129) a. these greater Schooles rather intend the forwarding of such children as

are already grounded, then busie themselves about meere Rudiments

(1660, PPCEME)

b. It is made very doubtful whether the King do not intend the making of

the Duke of Monmouth legitimate (1663, CEMET)


The effect of broad paradigmatic analogy may have been reinforced by a develop-

ment that precedes the appearance of ing-complements with intend by no more than

a few decades, namely the use of ing-complements introduced by of with the mor-

phologically related noun intention. The latter construction appears in the course of

the seventeenth century, as shown in (130), and is gaining currency at the time of the

appearance of ing-complements with intend.

(130) by the Entring of Satan (that is the Enemy) into him, is meant, the hos-

tile and traiterous intention of selling his Lord and Master. (1651,

CEMET)

STOP: The first examples of stop with ing-complement appear in the first half of the

eighteenth century, as is clear from (131). Although the pattern has ended up very

frequent by the end of the Late Modern period (cf. Table 9.2), it remained very rare

at first, no examples being attested in CLMETEV prior to 1800, and only a handful

before 1850.

(131) a. My Lord, I shall make a Pause here, and stop going on farther in my

Discourse, ‘till I see farther, if his Grace, my Lord Commissioner re-

ceive any humble Proposals for removing Misunderstandings among

us (1706, PPCEME)

b. Women who stop’d childing [i.e. ‘bearing children’] early. (1748,

OED)

c. immediately I ordered my people to stop working, and take up a rest

(1761, POB)

d. I pursued him into the Castle and Faulcon yard: he stopped running,

and was opening the paper to look at the ring: I got up to him, and laid

hold of him, and said, my friend, you shall not drop the ring (1764,

POB)

e. We have stop’d Paving as it was when you was at Haydock (1771,

CLECP)

The collocational behaviour of stop presents little or no occasion for narrow

paradigmatic analogy. In seventeenth and eighteenth-century usage, stop typically

involves an opposition, marking the matrix subject’s throwing up of a (figurative)

barrier to an external opposing force. Consequently to these oppositional semantics,

the patient argument of stop is almost never an action controlled by the subject. Rep-


resentative examples are provided in (132). When contrary to normal expectations

the patient is an action controlled by the matrix subject, this has to be marked explic-

itly by the use of a possessive determiner, as in (133). Note though that examples of

the latter kind are highly uncommon.

(132) a. [He] began to have some tender for his own Safety, and would will-

ingly have groped his Way back again; when he heard a Voice, as from

a Person whose Breath had been stopp’d by some forcible Oppression,

and just then, by a violent Effort, was broke through the Restraint.

(1692, CEMET)

b. It appear’d that the Cart was going very soberly along, and that the

Child fell down just under the Wheel, so that the Prisoner could not

possibly stop it (1715, POB)

c. Mrs. Bennet having fastened the door, and both the ladies having taken

their places, she once or twice offered to speak, when passion stopt her

utterance; and, after a minute’s silence, she burst into a flood of tears.

(1751, CLMETEV)

d. But I considered that the intent of my father’s endeavouring to stop me

was to hinder me from obeying the call of my heavenly Father, and to

stop me from going to worship Him in the assembly of His people

(1683, CEMET)

(133) a. Come, come, my scholar, you see the river stops our morning walk:

and I will also here stop my discourse (1656, CEMET)

b. Stay, thou dear innocence, retard thy flight, O stop thy journey to the

realms of light (1753, CLMETEV)

c. You ask me, what is the cause of this cause? I know not; I care not;

that concerns not me. I have found a Deity; and here I stop my inquiry.

(1779, CLMETEV)

There are two mechanisms left that may have brought about the use of ing-

complements with stop. The first is semantic analogy with the group of aspectual

verbs. However, although a classification of stop as an aspectual might seem self-

evident from a present-day perspective, seventeenth and eighteenth-century usage

shows stop to have little enough in common with the aspectuals. Indeed, the aspec-

tuals semantically and collocationally contrasted with stop in that they did take ac-

tions carried out by the matrix subject as their patient, while stop differed from the

aspectuals for instance in allowing concrete objects and persons as patients (see


(132b, d) above), in combining with different-subject ing-complements (see (136)

below), and in entering object-controlled from-constructions (see (132d) above).

Still, one point of similarity between stop and some of the aspectuals lay in the in-

transitive uses of stop, illustrated in (134), which can be interpreted as implying the

cessation of an action by the matrix subject. The fact that intransitive uses of this

kind increased in frequency in the first half of the eighteenth century, when ing-

complements first appeared as well, suggests that the intransitive pattern may have

been the point of entry for further analogy-induced modifications in the direction of

aspectualhood. At the same time, apart from this temporal coincidence, there is little

further evidence in favour of semantic analogy. For example, that stop also came to

combine with repeated-subject complements (see (135)) indicates that stop was in-

deed at some point recruited as a new member of the aspectual class, but the re-

peated-subject complements only appeared around 1850 – quite late, in comparison

to the appearance of bare controlled ing-complements.113

(134) a. While I was thus employed in thought, Monsieur – pulling me (eager

of joys to come,) and I holding back, he stopped and cried, ‘Sure,

Melinda, you came not hither to bring me a denial.’ (1684, CEMET)

b. He ascended the staircase without stopping till he arrived at the gallery,

at the door of which he met Hippolita and her chaplain. (1764,

CLMETEV)

(135) a. She looked up. And her heart stopped its beating to listen. (1848,

CLMETEV)

b. Hearing this, the young swaggerer stopped his step-dancing and cried:

“What cheer, myte? [...]” (1897, CLMETEV)

The second mechanism that may have set off the use of ing-complements with stop

is indirect paradigmatic analogy. Given the collocational behaviour of stop dis-

113 Stop also differs from other aspectuals in that it has never accepted to-infinitives as complements (two examples in Visser 1963-73: 1389 excepted). It is true that aspectuals of cessation are in general somewhat less hospitable to the to-infinitive, but the total absence of to-infinitives with stop remains remarkable. Arguably, com-plement to-infinitives have been blocked by the homonymous purpose adjuncts, which in the eighteenth century grew sufficiently common with stop to become a firmly-entrenched collocate. An example is given in (viii): (viii) My father could not stop to answer, for fear of breaking the thread of

his discourse (1759-67, CLMETEV)


cussed above it comes as no surprise that the verb could easily combine with non-

controlled complements, taking the form of definite as well as different-subject con-

structions, as in (136). Different-subject ing-complements may have inspired bare

controlled ing-complements – as appears also to have happened with the verbs re-

member and mind (see below).

(136) a. I recommend to you a Note of Platerus in his Observations [...] how an

Hangman of Basil cut off mens Hands, and stopped the Bleeding of the

Veines and Arteries by slitting the Arse of an Hen, and thrusting the

Hand into it. (1666, LC)

b. upon our coming we sent away an express to Sir W. Batten to stop his

coming (1662, CEMET)

c. By such a progression as this, setting out from the place where it is, or

any other place, it can proceed and pass beyond all those lengths, and

find nothing to stop its going on, either in or without body. (1689,

CEMET)

KEEP ON: The first ing-clauses following keep on show up in the first half of the

eighteenth century, and are given in (137). As with go on (see above), however,

most of these can be interpreted as the combination of a motion verb (roughly mean-

ing ‘keep moving on a set course’) and an adverbial participle expressing manner or

an accompanying activity. The first examples for which such an adverbial interpreta-

tion is impossible are given in (138): keep on does not denote motion but continua-

tion of some activity, while the ing-clause functions as complement, marking that

which is continued.

(137) a. We kept on crouding [i.e. ‘speeding’] till Night. (1699, OED)

b. I kept on steering directly for the island (1719, Visser 1963-73: 1899)

c. I took the occasion .. to keep still on southing. (1725, OED)

d. I was afraid of his Knife, or I could have closed him then; but I kept on

pursuing and crying Stop Thief till he was taken. (1733, POB)

(138) a. This stagnation of our manufacturing trade in the country would have

put the people there to much greater difficulties, but that the master-

workmen, clothiers and others, to the uttermost of their stocks and

strength, kept on making their goods to keep the poor at work (1722,

DEFOE)


b. The mason took the mortar out of the bucket; and if any was spared, he

still kept on beating. (1793, OED)

The primary mechanism giving rise to the use of keep on with ing-complement is in

all likelihood reinterpretation, in combination with semantic analogy to the verbs

continue and go on (which began to take ing-complements slightly earlier; see

above), as well as formal resemblance to keep. As an intransitive motion verb, keep

on first appeared around 1600 (see (139)) – about a century ahead of the use with

ing-complement – but it subsequently remains unattested in the data throughout the

seventeenth century, only to crop up again around 1700 (see (140)).114

(139) a. In this manner doth the Greeke dactilus begin slowly and keepe on

swifter till th’ end. (1589, OED)

b. Like to the Ponticke Sea, Whose Icie Current, and compulsiue course,

Neu’r keepes retyring ebbe, but keepes due on To the Proponticke, and

the Hellespont (1623 (1621), CEMET)

(140) a. He that ploughs must keep on, and make no balks. (a1703, OED)

b. In ev’ry Thing the Course of Nature still Keeps duly on, concernless in

its Road. (1721, OED)

c. We kept on all night. (1724, OED)

d. we resolved to stand away to the southward; and [...] to keep on to the

southward, and see if we could not leave not only the Moluccas, or

Spice Islands, behind us, but even Nova Guinea and Nova Hollandia

also (1720, DEFOE)

The intransitive pattern’s reappearance around 1700 coincides with the emergence

of adverbial ing-clauses with keep on (see (137) above) and is followed relatively

soon by the emergence of the first ing-complements proper. The latter development

can be thought of as the result of reinterpretation, as witnessed in go on (see above).

114 The early-eighteenth-century example in (ix) shows that the semantic change in keep on from ‘keep moving on a set course’ to ‘continue’ took place in part inde-pendently of the construction with participial ing-clause (note though in (ix) that there is still a figurative trajectory implied by in my Way). (ix) I desire [...] that you would advance me in my Way, and let me keep

on in what I humbly presume I am a Master, to wit, in representing human and still Life together. (1711, SPECTATOR)


Again, there are a number of analogues that may have invited the complement-

reading – most obviously go on of course, but also continue and keep.

REMEMBER: Whereas Visser (1963-73: 1876) comes up with an example of remem-

ber followed by a bare ing-form that is dated c1535 (see (141a)), all following ex-

amples are attested almost two centuries later (see (141b-e)), leaving an awkwardly

long gap in the evidence. On closer inspection, however, it is clear that the construc-

tion in Visser’s example in (141a) is not exactly equivalent to that in the later exam-

ples in (141b-e). Specifically, the action referred to by nombrynge in (141a) is

probably not anterior to that referred to by remembereth. An alternative interpreta-

tion of (141a) is that nombrynge is in fact an adjunct (meaning ‘by enumeration’)

and that the outwarde partes of it is the real direct object of remembereth (meaning

‘bring to mind’, a verb of communication in parallel to the earlier spake of). I take it

therefore that the examples in (141b-e), and not (141a), are representative of the first

use of bare ing-complements with remember, which must then go back to the begin-

ning of the eighteenth century. Finally, as the examples in (142) show, remember

also selects different-subject complements, which came in use about half a century

before remember appeared with bare ing-complements – an order of events that is

further reflected in the quantitative predominance of the different-subject construc-

tion until about 1800.

(141) a. He spake before of the inwarde partes of mysery, now he remembereth

nombrynge the outwarde partes of it. (c1535, Visser 1963-73: 1876)

b. about Midnight, when he came Home I heard him swear, he’d make

her remember leaving the House. (1724, POB)

c. I perfectly remember carrying back the Manuscript you mention and

delivering it to Lord Oxford (1740, Fanego 1996b: 76)

d. Besides, in my last letter I remember telling you that the Archduchess

was dead; she did not die till a fortnight afterwards. (1744,

CLMETEV)

e. I don’t remember ever reading of any [goats] in the country about the

lakes. (1759, OED)

(142) a. The child [...] only remembered his falling into the sea, not being able

to give any account how he returned into the ship. (1688, CEMET)

b. I remember his going away with Sir John Southcoat, but I cannot tell

the Time. (1685, PPCEME)


c. I remember, added she, your asking Mr. Williams, If there were any

gentry in the neighbourhood? (1740, CLMETEV)

No straightforward semantic analogues being available, it is hard to think of the

emergence of ing-complements with remember as being caused by semantic anal-

ogy. Similarly, there is no indication that the use of ing-complements could result

from narrow paradigmatic analogy. Given the sequence of events apparent from

(141)-(142) above, the use of bare controlled ing-complements may have been mod-

elled on the slightly earlier use of different-subject ing-complements. At the same

time, the appearance of the different-subject construction is itself hard to account

for. The only circumstance that might have promoted its use is the common colloca-

tion of remember with noun phrases referring to events, actions or situations, as il-

lustrated in (143) (i.e. broad paradigmatic analogy).

(143) a. This matter I only communicate to yourself and desire yr Lp will keepe

it private, only you may please to finde some means to have my Lord

of Ormond askt whether hee remembers not such a discourse between

himself and mee (1675-7, PPCEME)

b. Besides, said he, he remembers the affronts that Mansoul has given

him, and he is resolved to be revenged of them. (1682, CEMET)

c. I expect it from all you Gentlemen of the King’s Counsel, and others

that are concerned, that you take notice and remember what has passed

here (1685, PPCEME)

d. L. C. J. Had one of the two Men a Cloak on? — J. Sikes. I cannot re-

member that. (1692, POB)

M IND: The history of ing-complementation with mind is complicated by semantic

changes in the verb. Used in its original sense, mind means ‘direct one’s attention to,

take notice of’, as illustrated in (144a-b). This original sense branches off into a va-

riety of different meanings, including ‘remember’ and ‘intend’. For our purposes,

however, it is especially important that the original sense often implies that the sub-

ject is concerned or worried about something or cares about it, as in (144c-d). This

new meaning in turn invites an inference of annoyance, objection or unwillingness

on the part of the subject, which has likewise become semanticised, witness (144e-

g). Because of the ambiguities between different senses, it is difficult to date the

semantic changes with any great precision, but it appears that the sense ‘worry


about’ dates from the end of the seventeenth century, whereas the sense ‘object to’

goes back to the second half of the eighteenth.

(144) a. Here comes a Spirit of his, and to torment me [...]: I’le fall flat, Per-

chance he will not minde me (1623 (1610-1611), CEMET)

b. The lords did not much mind Gunning’s arguments, but passed the bill

(1683-1713, PPCEME)

c. And in a word, by all such as mind more the Benefice, then the office

[...]: All such do falsly wrest, and wickedly pervert, this double charge

of the great Visitor in the Text: Take heed to your selves and to your

Flock (1669, LC)

d. And this [i.e. measures to the opening of trade] he recommended to his

Prudence to persuade the Rajah thereto; who being a Soldier from his

Infancy, its possible minded not such concerns. (1672-81, PPCEME)

e. it would only furnish them with matter for silly jests; which, though I

would not have you mind, I would not have you invite. (1749,

CLMETEV)

f. Why, yes, you may venture, Sir Harry: it is not minded in London.

(1776, OED)

g. “[...] if he comes this way, I shall certainly make a renounce, and re-

tire.” “Why so?” said Sir Robert, “what the d---l do you mind him

for?” “O he is the greatest bore in nature!” cried the Captain (1782,

CLMETEV)

Against this background, we can return to the development of ing-

complements. Bare controlled ing-complements first turn up with mind in two iso-

lated instances. The first is given in (145a) and shows mind in the sense ‘intend,

plan’, a meaning that derives independently from the original sense of the verb ‘di-

rect one’s attention to’. Used in this sense, mind regularly combined with a to-

infinitive (see (145b-c)), so the ing-form is probably an incidental case of horror

æqui, here used in order to avoid two consecutive to-infinitives.

(145) a. to devise the death of the lordis was felony; to mind resisting his

attachment was felony; to raise London was treason (1550-2,

PPCEME)

b. after the inditements rede, 6 in number, the lerned counceill laid to my

lord of Somerset Paulmer’s confession. To wich he answerid that he


never minded to rayse the North, and declared al ill he could devise of

Paulmer (1550-2, PPCEME)

c. but also he did mynde shortlye after to associate himselfe with those

Traytours; for hee minded to haue departed with the Earle of

Deuonshire Westwardes. (1554, PPCEME)

The second occurrence of mind with bare ing-complement is found in Pepys’ Diary

and is given in (146). The example instantiates a use that is very frequent in Pepys’

language, as is illustrated in (147). Mind here is intermediate between its original

sense ‘direct one’s attention to’ and its later sense ‘worry about’: the verb means

‘occupy oneself with, pay full attention to’ and is employed to characterise the ma-

trix subject as being preoccupied with one particular activity, often to the exclusion

of anything else. As the collection of examples shows, the object of mind was most

commonly a noun phrase introduced by a possessive determiner referring back to the

matrix subject (see (147a-b), but could also be a repeated-subject ing-complement

(see (147c-d)) or a bare abstract noun, typically business (see (147e)). In this light, it

is plausible that the bare ing-complement in (146) results from some form of para-

digmatic analogy (either narrow, broad or indirect).

(146) my Lord Treasurer, he minds his ease, and lets things go how they will

[...]. My Lord Chancellor he minds getting of money and nothing else

(1665, PEPYS)

(147) a. he and the Duke of York mind their pleasure, as they do and nothing

else (1666, PEPYS)

b. he [...] did then discover his thoughts, that Sir J. Minnes was too old,

and so was Colonel Middleton, and that my Lord Brouncker did mind

his mathematics too much. (1668-9, PEPYS)

c. I walked with him to Paul’s, he telling me how my Lord is little at

home, minds his carding and little else, takes little notice of any body

(1664, PEPYS)

d. At noon home, and contrary to my expectation find my little girle Su

worse than she was, which troubled me, and the more to see my wife

minding her paynting and not thinking of her house business (1666,

PEPYS)

e. I was wiling to carry something fresh that I may look as a man minding

business (1666, PEPYS)


Bare controlled ing-complements with mind as we know them today first ap-

pear in the eighteenth century. It is assumed here that this event is a separate innova-

tion, despite the two earlier examples in (145a) and (146) above. The reason is that

neither the pattern in (145a) nor that in (146) has survived in later usage, that each is

restricted to a single occurrence, and that, semantically, there is no direct continuity

between the two early examples and later usage. The oldest examples representative

of the present-day pattern, then, are given in (148). In these examples, the verb mind

is invariably negated, which in combination with the semantics of mind (here typi-

cally intermediate between ‘worry about’ and the later sense ‘object to’) gives rise to

the interpretations ‘not care about’ or, later, ‘be prepared to’. As (149) shows, from

the end of the eighteenth century, the use of bare controlled ing-complements also

extends to the idiomatic sequence never mind, which is used in conversation to dis-

charge the addressee from an obligation or intention.

(148) a. I shall not mind (says the Prosecutor) giving her the Trifle of Money,

but the Watch I value, and would not part with it for 20 l. (1736, POB)

b. I heard him say he had robbed a man with a bundle under his arm of 1

s. 6 d. and said, he hoped they would not hang him; but he did not

mind being transported. (1755, POB)

c. I hope you have been long recovered from your worry on Thursday

morng, & that you do not much mind not going to the Newbury Races.

I am withstanding those of Canterbury. Let that strengthen you. (1808,

CLMETEV)

d. They hope she will not mind sleeping in the attic – indeed they are sure

she will not, she is such a dear, good creature (1839, CLMETEV)

(149) a. O says he, never mind going back with the half guinea, let us [illegi-

ble] it. (1792, POB)

b. Come, never mind looking at it now; I want to find out the green chaise

that has John Nelson’s name upon it. (1796, CLMETEV)

c. ‘O never mind shewing me the way’ .. and [he] sallied into the apart-

ment. (1796, OED)

Trying to account for the appearance of bare ing-complements with mind, it is

again found that bare ing-complements are diachronically preceded by different-

subject constructions, as is shown in (150). The examples in (150) all show a

negated form of mind with the same meanings as in (148)-(149) above (‘not care

about, not object to’). It is therefore possible that the use of bare controlled ing-


complements was modelled after these different-subject complements. As to the

mechanism that gave rise to the different-subject complements, there is again no

strong trigger other than the occasional use of mind with noun phrases denoting a

situation or an action, as is illustrated in (151), which at least demonstrate the

compatibility of mind with patient arguments denoting some action or state of

affairs.

(150) a. Men do not minde their [i.e. travelling merchants’] going away; for if

one be gone, be sure another will quickly come. (1681, LC)

b. I did not mind his being a little out of humour, but comforted him, by

giving him several instances of men of our acquaintance, who had no

one quality in any eminence, that were much more esteemed than he

was with very many: but the thing is, if your character is to give pleas-

ure, men will consider you only in that light, and not in those acts

which turn to esteem and veneration. (1710, TATLER)

c. He told me, if it had not been for the Bitch his Wife, it had never hap-

pen’d, for he had never minded the Prisoner whetting his Knife, if she

had not put him in a Passion (1732, POB)

d. He has preached twice at Somerset Chapel with the greatest applause. I

do not mind his pleasing the generality, for you know they ran as much

after Whitfield as they could after Tillotson; and I do not doubt but St.

Jude converted as many Honourable women as St. Paul. But I am sure

you would approve his compositions, and admire them still more when

you heard him deliver them. (1742, CLMETEV)

(151) a. Oh, now the danger’s past, your general comes! He joins not in your

joys, nor minds your triumphs; But, with contracted brows, looks

frowning on, As envying your success. (1678, CEMET)

b. and so where her body laid this spring burst forth and remaines till

now, a very rapid current, which runs off from this Well under a barre

by which there are stone stepps for the persons to descend which will

bathe themselves in the Well; [...] but there is nothing to shelter them

but are exposed to all the Company that are walking about the Well

and to the little houses and part of the streete which runs along by it;

but the Religeuse are not to mind that (1698, PPCEME)

c. Why, says Mr. How, I have got a Writ against you for 9 l. which you

owe your Father-in-Law Thorowton Bocklingson for Board. O, says


the Prisoner, if that’s all, I don’t mind that, and so he sat down again,

and sung us a Song or two (1732, POB)

d. I am resolved to return to our master; and as he is not so well as were

to be wished, the more haste you make the better: and don’t mind my

fatigue, but consider only yourselves, and the horses. (1740,

CLMETEV)

3.4. Stage IV: broad paradigmatic analogy

Stage IV, starting in 1740 and thus more or less coinciding with the Late Modern

period, is the final stage in the diffusion of ing-complements here described. During

Stage IV, ing-complements spread to the verbs regret, enjoy, commence, dislike,

risk, fancy, contemplate, suggest, try and start. Characteristic of Stage IV is that ing-

complements now spread on the basis of simple broad paradigmatic analogy, with-

out the mediating step that is typical of indirect paradigmatic analogy.

This is particularly clear for the introduction of ing-complements with enjoy,

fancy, contemplate and try, and may apply also to regret and risk (though for the

latter two verbs, French influence plausibly assisted in the furtherance of ing-

complements). Furthermore, note that enjoy collocationally patterned with nouns

denoting actions by the matrix subject, which are still relatively close to ing-

complements, but this is not true of fancy, contemplate, and try, where the introduc-

tion of ing-complements appears to be sanctioned simply by the fact that these verbs

are transitive. Interestingly, the historical data even suggest that with fancy and con-

template, the appearance of ing-complements triggered slight semantic changes that

subsequently affected the verbs’ collocational behaviour, which came to select noun

phrases more closely resembling ing-complements. This is a reversal of the usual

order of change and as such bears further testimony to the increase in the productiv-

ity of ing-complementation.

Next to broad paradigmatic analogy, Stage IV also sees the familiar changes

due to lexical developments. Particularly clear examples are the histories of the as-

pectual verb start and of fancy used in the sense ‘imagine’. In addition, Stage IV

also provides a counterexample to the general trend in that ing-complements appear

with dislike, probably as a result of semantic analogy. Since semantic analogy af-

fected most members of the class of emotive verbs at the end of the seventeenth cen-

tury (see above), dislike is remarkably late taking ing-complements, especially con-

sidering the fact that there seems to be nothing that could delay the appearance of

ing-complements with dislike (e.g. blocking or the necessity of a prior semantic


change in the verb). It is not inconceivable, however, that the counterexample is

simply a quirk of the data.

REGRET: The combination of regret with ing-complements begins to occur halfway

the eighteenth century with the examples in (151), thus concurring with a consider-

able increase in the frequency of regret itself, which before 1710 is barely attested in

the corpus data. As the examples show, the ing-complements with regret are pecu-

liar in the eighteenth century for two features not found in the ing-complements with

other verbs at the time: the use of a negator within the complement clause and the

use of the perfective auxiliary have.

(151) a. I have heart enough, sir, to do a deed that would make you regret using

me thus (1740, CLMETEV)

b. How much do I regret not having had more opportunities of showing

you my esteem and love, before this new attention, to Mr. Mann.

(1742, CLMETEV)

c. “I suppose,” he said, “you will not permit an old friend to visit you in

town, lest the sight of him should prove a disagreeable memorial of the

time you will soon regret having wasted in the country?” (1782,

CLMETEV)

d. I regretted not having visited them sooner (1783, CLMETEV)

As to mechanisms of change, the collocational behaviour of regret offered little op-

portunity for narrow paradigmatic analogy. Semantic analogy on the model of the

emotive verbs cannot be entirely excluded but there is certainly no further evidence

pointing in this direction (such as further distributional parallels). On the contrary,

what evidence there is points to borrowing as the main mechanism of change. The

use of regret with ing-complement closely resembles a French construction with the

verb regretter (which is obviously also the lexical source of English regret) and, in

particular, seems to mirror the use of negation and perfective avoir ‘have’ that is

prevalent also in the French pattern, witness the examples in (152). Thus borrowing

can account for the somewhat idiosyncratic lexico-grammatical behaviour of ing-

complements with regret.

(152) a. Vous êtes aimable, sans doute, il n’est pas difficile de le voir, et j’ai

regretté cent fois de n’y avoir pas fait assez d’attention (1733,

FRANTEXT)


‘You are lovely, without a doubt, it is not hard to see that, and I have

regret a hundred times not having paid enough attention to it’

b. Permettez-moi, monsieur, de remarquer que, si l’on est sourd aux

expressions de ma douleur, jusqu’à refuser d’entendre ce que j’ai à

dire, et de lire ce que j’écris, on pourra regretter bientôt de m’avoir

traitée si durement. (1751, FRANTEXT)

‘Allow me, sir, to observe that, if one is deaf to my expressions of

grief, to the point of refusing to listen to what I have to say and to read

what I write, one could soon regret having treated me so severely.’

What borrowing cannot explain is the specific choice for ing-complements. In this

respect, the appearance of ing-complements with regret again bears witness to the

increasing productivity of ing-complementation. At the same time, this does not

signify that the choice for ing-complements is thereby arbitrary. The choice is still

weakly constrained, in that regret, like other ing-complement-taking verbs, is a tran-

sitive verb that readily combines with patient arguments denoting an action by the

matrix subject (see (154)) (i.e. broad paradigmatic analogy). Moreover, as a rela-

tively new verb, regret did not yet strongly collocate with some other complement

type, so that language users could quite unrestrainedly resort to ing-complements

(while simultaneously exploring the possibilities of the to-infinitive; see (155)).

(154) a. He for a few hours regretted his regret, and from that time bent his

whole mind upon the means of escaping from the Valley of Happiness.

(1559, CLMETEV)

b. time, perhaps, if our declarations are ineffectual, will convince your

highness we are so, and you will then regret the injustice you have

done us. (1744, CLMETEV)

(155) He was pleased to take notice of my dress; and spanning my waist with

his hands, said, What a sweet shape is here! It would make one regret

to lose it; and yet, my beloved Pamela, I shall think nothing but that

loss wanting, to complete my happiness. (1740, CLMETEV)

ENJOY: Before tackling the appearance of ing-complements with enjoy, it is fruitful

first to look at the verb’s semantic development. What is presently the most common

sense of enjoy, ‘take pleasure from, delight in’, is in fact a relatively recent innova-

tion. While the transitive uses of enjoy have always had a positive semantic prosody,

combining with objects denoting something useful or pleasurable, the verb origi-


nally did not foreground the experience of pleasure that such objects excite in the

subject but rather emphasised the subject’s having the object at his/her disposal –

often as a specially granted privilege – as well as his/her drawing on or partaking of

the exclusive benefits it offers, as shown by the examples in (156). Two factors,

however, contributed to the emergence of the new sense ‘delight in’. One was the

intransitive and especially the reflexive uses of the verb, which (for independent

historical reasons) did semantically foreground the experience of pleasure, as illus-

trated in (157). The other was the recurrent implicature in transitive uses that the

subject not only receives, possesses or exploits something valuable or pleasant but

also derives pleasure and satisfaction from it, as shown in (158). Although the new

sense of enjoy that arose as a result probably may have appeared in the seventeenth

century, it certainly remained very infrequent at the time. In eighteenth-century us-

age, by contrast, the new sense is commonly attested, as is illustrated in (159).

(156) a. your license and liberty shall be to take, hold, enjoy, and make your

own all that is pleasant from the east to the west. (1682, CEMET)

b. I governed myself in a free yet respectful carriage towards her, that I

thereby both preserved a fair reputation with my friends and enjoyed as

much of her favour and kindness in a virtuous and firm friendship as

was fit for her to show or for me to seek. (1683, CEMET)

c. this, and many other like blessings, we enjoy daily. And for the most of

them, because they be so common, most men forget to pay their praises

(1656, CEMET)

d. [the Court] is made up of fraud and titles, and flattery, and many other

such empty, imaginary, painted pleasures; pleasures, that are so empty,

as not to satisfy when they are enjoyed. (1675, CEMET)

(157) a. Yet he neuer enioied after, but in conclusyon pitifully wasted his pain-

ful lyfe. (1549, OED)

b. we’ll sit still, and enjoy ourselves a little longer under this honeysuckle

hedge. (1656, CEMET)

(158) a. we can’t say that we have spent our Fortunes, but that we have enjoy’d

‘em. (1707, PPCEME)

b. [he] went into Ireland with a person powerful there in those times, by

whose means he was quickly preferred to a place of trust and profit,

but lived not long to enjoy it. (1683, CEMET)

c. I confess I had the happiness to be particularly known to him for about

the space of twenty years; and, in Oxon, to enjoy his conversation, and


his learned and pious instructions while he was Regius Professor of

Divinity there. (1673, PPCEME)

(159) a. My Lovers are at the Feet of my Rivals, my Rivals are every Day be-

wailing me, and I cannot enjoy what I am, by reason of the distracting

Reflection upon what I was. (1711, SPECTATOR)

b. I am writing to you in one of the charming rooms towards the park: it

is a delightful evening, and I am willing to enjoy this sweet corner

while I may, for we are soon to quit it. (1742, CLMETEV)

c. “You may at least recreate yourself,” said Imlac, “with the recollection

of an honourable and useful life, and enjoy the praise which all agree

to give you.” (1759, CLMETEV)

With this in mind, let us return to the development of ing-complements. The

verb enjoy is first attested with an ing-complement in an example from 1751, given

here as (160). This first example stands apart from later usage, however, because the

meaning of enjoy is not exactly ‘delight in’, foregrounding the experience of pleas-

ure, but ‘have the opportunity to engage in’, a semantic variant on the original sense

of the verb that still conveys the idea of being granted something pleasant. By con-

trast, the examples under (161) are representative of present-day usage, with enjoy

simply marking the subject’s pleasure at an activity in which s/he engages. As these

examples show, this use first appears in the data in 1777, yet as the long interval to

the following attested examples suggests, its frequency must have remained very

low until the latter half of the nineteenth century.

(160) We were visited by Mr. H. B –, a relation of my lord, and one Mr. R –,

of the Guards, who, with the little Scotchman and my lover, made an

agreeable set, among whom I enjoyed hunting, and all manner of coun-

try diversions. (1751, CLMETEV)

(161) a. The girls, Betsy and Beckey, were upon the high gig all the time, for

they enjoyed seeing me thus whisked about. (1777, OED)

b. During my stay here I thoroughly enjoyed scrambling about these huge

mountains. (1839, CLMETEV)

c. I never remember wanting to throw a rotten egg at any of my fellow-

creatures before – but I feel certain that I should enjoy having a shy at

Mr. Jubber! (1854, OED)


d. I followed her, and tried to make her talk. All in vain. It was my mis-

fortune to be a man – and Limping Lucy enjoyed disappointing me.

(1868, CLMETEV)

From the examples in (160)-(161) it can be argued that whereas the original senses

of transitive enjoy – with the exception of (160) – resisted complementation by ing-

complements, the new meaning sanctioned ing-complementation and thus, on its

emergence, provided the occasion for the spread of ing-complements to enjoy.

This conclusion fits the attested collocational behaviour of enjoy. To begin

with, the fact that early enjoy never combined with ing-complements could be re-

lated to the receptive semantics of the verb, by which the verb’s object was con-

strued as a favour or privilege granted to the subject, disfavouring as objects such

actions as are carried out by the subject him/herself and are under the subject’s con-

trol. The collocational behaviour of enjoy in seventeenth-century usage thus gives

little occasion for paradigmatic analogy – compare the examples in (156) above.115

The isolated example with ing-complement in (160), then, is best explained as an

instance of a somewhat specialised use, illustrated in (162), in which enjoy combines

with a particular pastime the subject has the chance to engage in. The use is espe-

cially characteristic of Smollett’s language, which is the source both of (160) and of

the examples in (162). It needs no pointing out that hunting in (160) above is to be

conceived of as a kind of pastime, as is clear from the coordinated NP all manner of

country diversions. Finally, the use of ing-complements with enjoy ‘delight in’, as in

(161) above, is supported by broad paradigmatic analogy, with examples as in (163)

providing the necessary model.116

115 Some interesting indirect evidence of the semantic incompatibility between re-ceptive enjoy and controlled ing-complements comes from the fact that enjoy did regularly combine with subject-controlled ing-clauses provided the latter were em-bedded in a noun phrase marking the subject’s action as a granted privilege, as shown in (x). (x) ever since the compassionate Edella had procured them to be removed

from the dungeon, they had enjoyed the privilege of walking on the leads [i.e. lead-covered roofs] (1744, CLMETEV)

116 Notice that although semantically enjoy is superficially akin to the emotive verbs, it does not seem to undergo their analogical influence. Dissimilarities between enjoy and the emotives are the absence of to-infinitives with enjoy and certain colloca-tional differences, for instance, the low likelihood of enjoy (in the sense ‘delight in’) combining with object noun phrases denoting persons. The reason, presumably, is that, semantically, enjoy misses the evaluative component characteristic of verbs


(162) a. being disabled by the gout from enjoying any amusement abroad, he

entertains himself within doors, by keeping open house for all corners,

and playing upon the oddities and humours of his company (1771,

CLMETEV)

b. We have had princely sport in hunting the stag on these mountains.

These are the lonely hills of Morven, where Fingal and his heroes en-

joyed the same pastime (1771, CLMETEV)

c. For my part, I love swimming as an exercise, and can enjoy it at all

times of the tide, without the formality of an apparatus [i.e. a device on

the beach that hides female bathers from view but only at high tide]

(1771, CLMETEV)

(163) a. See but how he simpers, and enjoys, as one may say, the relations of

his own rakish actions, when he tells a bad story! (1742, Samuel

Richardson, Pamela (Vol. 3))

b. I never, myself, so much enjoyed the sight of happiness in another, as

in that woman when I first saw her after the death of her husband.

(1782, BURNEY)

c. I have enjoyed retirement as much as any man in England; but then I

have been always surrounded with my family and friends (1820-2,

CLMETEV)

d. I so much enjoyed my rambles among the rocks and mountains of St.

Helena, that I felt almost sorry on the morning of the 14th to descend

to the town. (1839, CLMETEV)

A final remark is in place on the subsequent development of the verb enjoy,

which underwent further semantic changes in the course of the Late Modern period.

Thus, in the examples in (164), the meaning of enjoy is not ‘delight in’ but is a

weakened version of the earlier receptive senses that can be paraphrased as ‘engage

such as like, love or hate. This evaluative character makes the relations expressed by verbs such as like more time-stable: liking extends beyond the time-interval of actual interaction, whereas enjoying is restricted to the moment of contact between enjoyer and enjoyed (compare I like Jack Nicholson, which can be said without actually watching Jack Nicholson at the time of utterance, to I enjoy Jack Nicholson, which must be uttered while seeing Jack Nicholson in action – the alternative I always en-joy Jack Nicholson is felicitous as a general statement because always adds the time-stability that enjoy lacks). The same semantic contrast probably underlies the differ-ence in acceptability between ?I am liking this and I am enjoying this (Cuyckens p.c.).


in’ or simply ‘have’. As the examples show, this use of enjoy readily patterns with

actions by the matrix subject, so that the fact that ing-complements never spread to

this new use poses an apparent counterexample to the principle of paradigmatic

analogy.

(164) a. May you enjoy a better night’s repose than I am likely to have. (1787,

CLMETEV)

b. Following one of the arms of the bay, we enjoyed a pleasant row, and

passed through pretty scenery, until we came to a village, beyond

which the boat could not pass. (1839, CLMETEV)

c. “Whereas the King of Ruritania –” “Would enjoy a long and prosper-

ous reign, God willing, sir.” (1898, CLMETEV)

What explains this incongruence is the status of enjoy in (164) as a light verb. The

semantics of enjoy have generalised to such a degree that now the primary verbal

process denoted by the clause is in fact that coded by the object of enjoy. The func-

tionality of this kind of construction lies in the nominal nature of the complement

and the greater (and different) range of modification the nominalised object allows,

such as quantification, comparison or adjectival modification (Huddleston & Pullum

2002: 291). It is to be noted, for instance, that the light verb uses of enjoy nearly

always take an indefinitely marked object (adding quantificational information) with

an adjectival modifier, and tend to become ungrammatical or change meaning under

omission of either of these (cf. *the King of Ruritania enjoyed a reign; or we en-

joyed the pleasant row, where enjoy must mean ‘take pleasure from’). Since the

light verb construction is employed exactly to bring the added semantic possibilities

of nominal modification into play, it is understandable that ing-complements are

never found in this use, since their clausal nature would obliterate the functional

advantages gained by the light use of enjoy.

COMMENCE: Ing-complements first appeared with commence at the end of the eight-

eenth century, as shown in (165). The construction was in frequent use in the early

nineteenth century, but began to decline again after 1850 and has become restricted

to very formal usage in Present-Day English.117

117 The following passage from George Eliot’s Middlemarch (published in 1871), in which the narrator comments on the pompous style of the character Borthrop Trum-bull, indicates that commence was regarded as highly formal also in the Late Modern period:


(165) a. were I to commence making apologies ... , I fear my letter would be too

long. (1778, Visser 1963-73: 1891)

b. As I commence farming at Whitsunday, you will easily guess I must be

pretty busy (1788, CLMETEV)

c. He commenced being a severe and ardent student. (1797, OED)

d. when the pot was on the fire, Ready, Mr. Seagrave, and William set off

with the cross-cut saw and hatchets, to commence felling the cocoa-nut

trees for the building of the outhouse (1841, CLMETEV)

The history of ing-complements with commence is tied to the lexical history of

commence itself. Another obvious French loan, the verb commence had made spo-

radic appearances in English since the Middle English period (OED s.v. commence),

but it began to gain currency only in the Late Modern period and especially around

the end of the eighteenth century (e.g. in CLMETEV, the frequency of commence

almost quadruples between the first sub-period 1710-1780 and the second 1780-

1850). This increase in frequency goes hand in hand with a general loosening of the

collocational restrictions on the kind of direct objects with which commence could

combine. Prior to 1750, the typical object of commence referred to legal – or, by

extension, antagonistic – action taken by the subject, as in (166a), but in the course

of the eighteenth and nineteenth centuries its selection of objects broadened, as illus-

trated in (166b-e).

(166) a. This Richard Muckleston .. commenced a suite against the Towne of

Shrewsbury for exacting an imposition upon him which they call ten-

sorship. (1700, OED)

b. she had an invention so fruitful she was not at a loss to commence new

demands (1757, POB)

“You have an interesting work there, I see, Miss Garth,” he observed, when

Mary re-entered. “It is by the author of ‘Waverley’: that is Sir Walter Scott. I have bought one of his works myself – a very nice thing, a very superior pub-lication, entitled ‘Ivanhoe.’ You will not get any writer to beat him in a hurry, I think – he will not, in my opinion, be speedily surpassed. I have just been reading a portion at the commencement of ‘Anne of Jeersteen.’ It commences well.” (Things never began with Mr. Borthrop Trumbull: they always com-menced, both in private life and on his handbills.)


c. This amiable maiden has actually commenced a flirting correspon-

dence with an Irish baronet of sixty-five. (1771, OED)

d. I know nothing of these goods, he brought them to my house when we

commenced partnership (1785, POB)

e. Mr. Hill commenced a practice .. of going .. into the kitchen .. to take a

slice from the roast or the boiled before it went up to table. (1804,

OED)

It is clear that the appearance of ing-complements coincides with other changes in

the verb, and it only makes sense to look for some causal link. Presumably, as its

collocational behaviour changed, commence became increasingly eligible for ing-

complementation. On the one hand, the verb could more readily combine with noun

phrases denoting actions by the matrix subject, which provided a paradigmatic

model for ing-complementation (i.e. broad paradigmatic analogy). On the other, the

verb came to resemble more closely other aspectual verbs and entered their sphere of

influence – as is also reflected in the appearance of to-infinitives, as in (167), and

the occasional use of repeated-subject ing-complements typical of aspectuals, as in

(168).

(167) a. The barbarians have commenced ... to furbish their professions and

vocations with rather whimsical skirts and linings (1824, Visser 1963-

73: 1376)

b. It was about that time when I [...] commenced to live upon an incon-

siderable annuity and my past reputation. (1841, Charles Dickens, Ba-

rneby Rudge)

(168) I [...] had commenced my sewing, when I was called out by Juno, and

Caroline went with me, and Tommy was left in the house. (1841,

CLMETEV)

Notice that the appearance of ing-complements (as well as to-infinitives) with com-

mence might also be regarded as a continuation of French influence, since French

commencer is extremely common with a non-finite (infinitival) complement. In this,

it is virtually impossible to tell a language-internal development apart from contact-

induced change, but given that commence comes to behave as other English aspec-

tual verbs (as illustrated in (167)-(168)), the language-internal component must be

an essential part of the story.


DISLIKE: Bare controlled ing-complements emerged with dislike during the last dec-

ades of the eighteenth century, as shown in (169). Although this is more than a cen-

tury after the first attestation of a bare ing-complement with like (see above), the

innovation is best accounted for as based on semantic analogy to the class of emo-

tive verbs. An important indication to this effect is that to-infinitives – though disal-

lowed in Present-Day English – appeared at about the same time as ing-

complements, witness the examples in (170), thus nicely copying the behaviour of

other emotives such as like, love and hate. In addition, dislike often patterns with

other emotive verbs, especially like, as shown in (171), indicating that the verbs

were classified by language users as belonging to the same lexical field. Finally, a

very clear example of dislike paralleling the behaviour of like is found in (172),

where dislike echoes the fixed phrase whether you like to or not.

(169) a. Cecilia much disliked thus taking possession of the house in the night-

time, though Delvile, solicitous to relieve her, desired she would not

waste a thought upon the subject (1782, CLMETEV)

b. At present, I must confess, I should not dislike submitting to his em-

pire, for a few months or years, just as it might happen, whilst Europe

is distracted by demons of revenge and war (1783, CLMETEV)

c. He disliked proceeding parliamentarily in this business. (1797, OED)

d. I was prevailed upon to become the payer, they promising that I should

always have sufficient assets to answer any demand that should be

made upon me. I disliked becoming the payer, principally from its be-

ing out of my usual business (1807, POB)

(170) a. I am angry at your thinking that I can dislike to receive two or three of

your letters at once. (1743, CLMETEV)

b. “Miss Beverley,” she said, “has your little rattling tormentor ac-

quainted you who is coming?” “Lord Derford, do you mean, ma’am?”

“Yes, with his father; shall you dislike to see them?” (1782,

CLMETEV)

c. I always think how much I should dislike to be obliged to do so my-

self, and therefore very sincerely pity those who must. (1783,

CLMETEV)

d. I desired he might be examined, suspecting he had some tea: the pris-

oner disliked to be examined: I sent for a constable: (1790, POB)

(171) a. I will tell you every thing I like or dislike, as it occurs to me: and I

would have you do the same, on your part; that nothing may be upon


either of our minds that may occasion the least reservedness. (1740,

CLMETEV)

b. Great merit, or great failings, will make you be respected or despised;

but trifles, little attentions, mere nothings, either done, or neglected,

will make you either liked or disliked, in the general run of the world.

(1749, CLMETEV)

(172) Therefore, whether you dislike to or not, you must tell me where to

find your grandfather. (1916, CEN)

The account in terms of semantic analogy faces some problems but these are

not insurmountable. First, the time-lag in the use of ing-complements with dislike, as

compared to other emotive verbs, may be accounted for by the fact that the ing-

complements with emotive verbs were relatively slow to increase in general, and by

the fact that the verb dislike itself, coined at the end of the sixteenth century (OED

s.v. dislike), has always remained a low-frequency verb. Chances are, therefore, that

the picture of change is somewhat disturbed by lack of data – a suspicion that is at

least weakly confirmed by the attestation of a different-subject ing-complement in

seventeenth-century English (cf. Table 9.1). Second, the subsequent loss of to-

infinitives after 1920, also reported in Visser (1963-73: 1864), is probably in part

related to the general decrease in the frequency of dislike in the twentieth century

(compared to the last sub-period in CLMETEV, 1850-1920, dislike has decreased in

frequency by more than half) but may also manifest a more general tendency for to-

infinitives to be replaced by ing-complements with the emotive verbs during the

same period (De Smet & Cuyckens 2007). Indeed, the fact that dislike, unlike like or

love (De Smet & Cuyckens 2005), never developed specialised uses or strongly

routinised half-idioms with to-infinitives, may have made to-infinitives with dislike

more vulnerable to competition with ing-complements than with like or love.

RISK: As is shown by the examples in (173), the verb risk is first found combining

with an ing-complement at the end of the eighteenth century, even though the pat-

tern gets to be commonly used only in the second half of the nineteenth.

(173) a. The worst .. will not risk losing their only abettors and palliators in this

kingdom. (1792, OED)

b. I wouldn’t risk losing sight of the charge committed to me a minute, so

we climbed the slope of heath together. (1847, CLMETEV)


c. and I was just worked up to risk coming to the rescue, when of a sud-

den his fingers relaxed (1847, CLMETEV)

d. Risk dying in an unknown landless sea. (1868, OED)

As with commence, the development of ing-complements with risk is linked to the

lexical development of the verb. Risk, from French risquer, entered English at the

very end of the seventeenth century (OED s.v. risk) and became progressively more

frequent in the course of the eighteenth and nineteenth centuries. The verb is used in

a number of different senses, which correlate with different types of direct objects.

Risk may mean ‘run the chance of losing, put in danger’ and take as its object that

which may be lost or damaged, as in (174a). This is the sense that is by far the most

common in eighteenth-century usage, accounting for almost 95% of the instances in

the period 1710-1780 in CLMETEV. However, the object of risk may also denote a

certain danger, as in (174b-c), or it may denote an action with potentially dangerous

consequences, as in (174d-e). In such cases, risk means ‘be in danger of’ or ‘take the

hazardous action of’. It is the latter two senses, which became more current in the

nineteenth century, that are also relevant to the development of ing-complements,

witness the examples in (173) above.

(174) a. Lately the King risqu’d both his kingdoms for offering to imprison

Philander. (a1687, OED)

b. Risquing the loss of Heaven. (1705, OED)

c. Darnford could not put off his journey, without risking the loss of his

property (1798, CLMETEV)

d. I now go so far as to risk a proposal of my own. (1775, CLMETEV)

e. Her next difficulty was in what manner to have this note conveyed; to

send her own servant was inevitably betraying herself, to employ any

other was risking a confidence that might be still more dangerous, and

she could not trust to the penny-post, as her proposal required an an-

swer. (1782, CLMETEV)

As with other verbs in this section, there is no obvious semantic analogue to risk that

could explain the appearance of ing-complements. It is likely, however, that the use

of ing-complements replicated French usage, as illustrated in (175). At the same

time, the examples in (174b-e) above indicate that risk, when used in the relevant

senses, was highly compatible with noun phrases denoting some action or situation


involving the matrix subject. Together, foreign influence and broad paradigmatic

analogy appear to have sufficed to trigger the use of ing-complements.

(175) en prenant la route contraire, je risquois de tourmenter son coeur

(1771, FRANTEXT)

‘in taking the opposite direction, I risked tormenting his heart’

FANCY: The history of ing-complements with the verb fancy is complicated by the

polysemy of the verb. The examples in (176) present some of the first attested in-

stances of fancy with ing-complement where fancy is used meaning ‘want’. The ex-

amples in (177), on the other hand, show the first instances where fancy means

‘imagine, picture’. As the examples show, both uses have a somewhat idiomatic

quality: in the former fancy is typically negated; in the latter it typically appears in

the form of an imperative and is used to draw attention to something extraordinary

or surprising (hence it soon develops into a marker of surprise as in (177d)). Al-

though it appears in the data slightly later, it is especially the second construction

type that is responsible for the high frequency of fancy with ing-complements in

nineteenth-century English (cf. Table 9.2).

(176) a. it’s a sin to kill a PET LAMB, I’m thinking – any way, it’s what I’m

not used to, and don’t fancy doing, and I’ll go and say as much to At-

torney Case (1796-1801, CLMETEV)

b. He did not quite fancy making one of that crowd of irregular-horse

who appear on a Wednesday at Crick or Misterton (1861, OED)

c. Neither he nor I seemed to fancy dealing with this part of the inquiry.

We looked at each other, and then we looked at the tide, oozing in

smoothly, higher and higher, over the Shivering Sand. (1868,

CLMETEV)

d. “Have you told mother?” “No, miss,” Maggie whimpered, absently

wiping her wrinkled cheeks with ineffectual muslin. “I couldn’t seem

to fancy telling your mother. [...]” (1908, CLMETEV)

(177) a. Carriages and opera-boxes, thought he; fancy being seen in them by

the side of such a mahogany charmer as that! (1847-8, CLMETEV)

b. Please, Ma’am, is this New Zealand or Australia?’ (and she tried to

curtsey as she spoke – fancy CURTSEYING as you’re falling through

the air! [...]) (1865, CLMETEV)


c. just fancy calling everything you met “Alice,” till one of them an-

swered! Only they wouldn’t answer at all, if they were wise.’ (1871,

CLMETEV)

d. Can I do anything to help? Liddy is afraid to come out. Fancy finding

you here at such an hour! Surely I can do something? (1874,

CLMETEV)

Turning to the mechanisms involved in the development of ing-complements

with fancy, it is convenient to start with the second construction type. Observe first

that I assume there to be no direct historical link between the two uses of fancy with

ing-complement. It is not impossible that the historically slightly older construction

in (176) exerted some influence on that in (177), but no real case can be made to this

effect. Instead, the use of fancy, meaning ‘imagine, picture’, with ing-complements

probably goes back to a reinterpretation of the complex-transitive construction found

in (178). As this complex-transitive construction allowed participles as object-

complement, including present participles (see (178c-d)), it is possible that the com-

bination of an object and a participial object-complement was reinterpreted as a ger-

und clause (and thus as a single clausal constituent). The construction such a reinter-

pretation is expected to give rise to – an unambiguous gerund clause, its subject in

the possessive form – is indeed attested for the relevant period, as shown in (179). It

is conceivable, then, that the use of bare controlled gerund clauses with fancy was

the result of reinterpretation, giving rise to a different-subject construction, which in

its turn triggered the use of controlled ing-complements through indirect paradig-

matic analogy (i.e. on the basis of the alternation between different-subject and sub-

ject-controlled ing-complements in other environments).

(178) a. One, who is inflamed with lust, feels at least a momentary kindness

towards the object of it, and at the same time fancies her more beauti-

ful than ordinary (1739-40, CLMETEV)

b. Here I lay, in peace and sunshine, a few happy moments; contemplat-

ing the blue sky, and fancying myself restored to the valley at F.,

where I have passed so many happy hours, shut out from the world,

and concealed in the bosom of harvests. (1783, CLMETEV)

c. You can fancy me ascending Monte Cavallo, leaning against the ped-

estal which supports Bucephalus; then, spite of time and distance, hur-

rying to St. Peter’s in performance of my vow. (1783, CLMETEV)


d. the beauties of holiness seem to radiate about her, and the by-standers

are almost induced to fancy her already worshipping amongst her kin-

dred angels! (1792, CLMETEV)

(179) a. No right to interfere! Venetia, my little fellow-labourer, no right to

interfere! Why all is yours! Fancy your having no right to interfere at

Cadurcis! (1837, CLMETEV)

b. She fancied their falling into the hands of some speculator, who, if he

did not break the mother’s heart by putting up a gasometer, would cer-

tainly wring it by building hideous cottages, or desirable marine resi-

dences. (1865, CLMETEV)

Apart from the fact that no other mechanism seems capable of accounting for the

appearance of bare controlled ing-complements (e.g. semantic analogy, narrow

paradigmatic analogy), there are two arguments giving some further support to a

connection between the construction with bare controlled ing-complement and the

complex-transitive construction. Firstly, it is in the latter construction that the verb

fancy underwent a subtle semantic development from its original sense ‘think, be-

lieve (wrongly)’ to the sense ‘picture, conjure up before the mind’s eye’ (compare

(178a) and (178b-d) respectively). As the pattern with ing-complement is only found

with the second meaning, it is not implausible that it inherited its semantics from the

complex-transitive construction with which the new meaning was primarily associ-

ated. Secondly, the use of the imperative form fancy to draw attention to something

unexpected or surprising also seems from its emergence to favour the complex-

transitive construction, as shown in (180).118 This too supports a link to the bare ing-

complement construction, which, as pointed out earlier, is exclusively attested with

imperative fancy, as well as to the different-subject gerund construction, which, too,

is first attested with imperative fancy (see (179a) above).

(180) a. Fancy me boxed up in the narrow vehicle. (1834, OED)

118 Counterexamples do occur, as in (xi) with the simple transitive fancy me. Still, even here the complex-transitive pattern is implied by the phrases preceding fancy me, of which at least the first two stand in a predicative relation to the direct object of fancy. (xi) Very snug, in my own room, lovely morng [sic], excellent fire, fancy

me. (1813, OED)


b. But fancy then the banners all placed on the steps of the Rock; high-

mass chaunted; and the civic oath of fifty thousand (1837, CLMETEV)

b. Fancy him bowing his little frizzle head. (1840, OED)

The slightly older construction, in which ing-complements combine with

fancy meaning ‘want’, presents a different puzzle. The problem here is that the

meaning of fancy, as well as the apparent association with negation, seems to be

exclusive to this particular construction at the time of its appearance. That is to say,

when transitive fancy is used with an ordinary direct object, the best paraphrase for

the verb is ‘like’ and the element of volition only shows up as a pragmatic side-

effect. Thus, in (181) all instances of fancy can be interpreted as meaning ‘like’ and

only some – viz. (181d) and perhaps (181c) – allow the additional sense of ‘wish,

want’. In the examples with ing-complement, however, the negated form of fancy

can only be paraphrased as ‘have no desire to’, while the evaluative sense of ‘like’ is

positively ruled out (see (176) above). The new sense of fancy did extend to other

uses of the verb, as shown in (182), but this only happened after the appearance of

fancy ‘want’ with ing-complements. It seems then that the behaviour of ing-

complements with fancy at first violated the principle that the addition of an ing-

complement always leaves verbal semantics intact.119

(181) a. The Prisoner being a Parish Child of St. Andrews Holborn, was nurs’d

next door to the Prosecutor’s House, who fancying the Boy, took him

upon liking to be a Servant (1714, POB)

b. I reached down a Bundle, but they did not fancy the Colour; they must

be lighter, and I must shew them some of 2 s. 3 d. (1740, POB)

c. They had a pot of hot, I believe brandy and rum together. I drank but

little to my share. I do not fancy strong liquor in a morning. (1779,

POB)

d. Those fourteen rings you would have sold, if any body had fancied

them? (1798, POB)

119 The problem may be partly due to the inaptness of paraphrases. For instance, fancy in the nineteenth-century examples in (176) and (181) can be rather consis-tently paraphrased with ‘find attractive’. But even this paraphrase has as a disadvan-tage that it leaves the later examples in (182b-c) uncovered and disguises their se-mantic relatedness to the examples with ing-complement in (176) above. In all, the different periphrastic options indicate that the semantic shift is subtle and that the use with ing-complements in (176) takes an intermediate position.


(182) a. To tell you the truth, M. le Baron, [...] the sooner [w]e are beyond

Blois, the better I shall be pleased. I think we run some risk there, and,

besides, I do not fancy a shambles. (1893, CEN)

b. Steve had rarely been lucky with bets when he fancied a flutter before.

(CB)

c. Antonia and Clive met at a dinner party 19 months ago. Clive asked if

she fancied a curry, she said yes and they have been together ever

since. (CB)

What we see then is a semantic shift that is concomitant to a shift in collocational

behaviour; what is new and surprising is that the shift is heralded by ing-

complements. It is the appearance of ing-complements that first broadens the collo-

cational behaviour of fancy and that primarily contributes to the semanticisation of

the volitional implicatures in the verb’s interpretation., which then begins to com-

bine with noun phrase objects denoting actions or events. As for the broader picture

of the diffusion of ing-complements, fancy shows clearly that diffusion has reached

the point where the appearance of new ing-complements is licensed simply by the

productivity of ing-complementation and the transitivity of the complement-taking

predicate.

CONTEMPLATE: With the examples in (183), the use of bare controlled ing-

complements with contemplate can be dated to the beginning of the nineteenth cen-

tury. Typical of the first examples with ing-complement is that contemplate means

‘intend, purpose’ (cf. Fanego 2007: 177). Later examples, however, also show con-

template meaning ‘imagine, think of’ (see (184)).

(183) a. To avoid starving, I again contemplated keeping a school. In that coun-

try, knowledge was viewed as a handicraft trade. (1797, Fanego 2007:

177)

b. I felt great relief in being the fellow-pupil with my friend, and found

not only instruction but consolation in the works of the orientalists. I

did not, like him, attempt a critical knowledge of their dialects, for I

did not contemplate making any other use of them than temporary

amusement. (1818, CLMETEV)

c. he never contemplated writing this Memoir, nor would he have made

the attempt, had it not been urged on him as a duty by friends (1838,

CLMETEV)


d. He had just established himself in lodgings in Alfred Place, Tottenham

Court Road; and he desired to see Mr. Luker immediately, on the sub-

ject of a purchase which he contemplated making. (1868, CLMETEV)

(184) a. For three years he had sung the praises of the Italians, but he had never

contemplated having one as a relative. (1905, CLMETEV)

b. Carson – or Bayle – had doubtless contemplated seeing a frightened

servant trying to prepare a stammering obvious lie. He confronted a

tall, thin man about whom – even if his clothes had been totally differ-

ent – there could be no mistake. (1922, CEN)

The fact that ing-complements first appeared with contemplate in the sense ‘intend’

is significant, since the use of contemplate meaning ‘intend’ is itself a nineteenth-

century innovation. In CLMETEV, for instance, the sense ‘intend’ is entirely absent

for contemplate in the period 1710-1780, but shows up regularly in the period 1780-

1850. Before the nineteenth century, contemplate was used meaning ‘look at’ or

‘view with the eye of the mind’ and would rarely take an action or situation as its

patient, as is shown in (185). The shift towards the sense ‘intend’, however, reorgan-

ises its collocational behaviour into allowing object noun phrases denoting actions

by the matrix subject, as shown in (186).

(185) a. The trifling and frivolous mind is always busied, but to little purpose

[...]. Knick-knacks; butterflies; shells, insects, etc., are the subjects of

their most serious researches. They contemplate the dress, not the

characters of the company they keep. (1748, CLMETEV)

b. If we contemplate a savage nation in any part of the globe, a supine

indolence and a carelessness of futurity will be found to constitute their

general character. (1776, CLMETEV)

(186) a. Their opinions, however, contemplate the actual employment of force.

(1807, OED)

b. had I contemplated murder, should I have fetched the sister to give

original evidence against me – it is contrary to nature, which dictates

self-preservation (1827, POB)

c. and if I had contemplated escape as the only alternative to save my

life, I might have avoided returning to London (1828, POB)

d. The admiration of Mr. De Quincey was so great that inquiring where

Coleridge was to be found, and learning that he was in Malta, he con-


templated an immediate visit to that island, but the fear of a French

prison reconciled him to remaining in England. (1838, CLMETEV)

With respect to contemplate meaning ‘intend’, note that both examples with ing-

complement and examples with action-denoting noun phrase still allow the original

reading ‘view with the eye of the mind’. This indicates that the transition between

the two senses is not a radical one.

Strikingly, however, the first instance of contemplate with ing-complement in

(183a) above precedes by about a decade the first example of contemplate ‘intend’

with ordinary action-denoting noun phrase in (186a). The temporal precedence is of

course narrow and might be an accident of the data, but it still remains hard to see

how the usual sequence of events (semantic change leads to collocational change,

which leads to the appearance of ing-complements) could apply here. Rather, what

we again witness here is that a verb that does not usually take actions or events as its

patient argument begins to select ing-complements, simultaneously undergoes a

slight semantic change and at about the same time or slightly later also starts com-

bining with ordinary action-denoting nouns. While this still points to the relevance

of paradigmatic regularity, it also indicates that paradigmatic analogy cannot have

been the mechanism behind the appearance of ing-complements with contemplate,

except in its most general form, as a principle that sanctions the use of ing-

complements with transitive verbs. The development is therefore analogous to that

witnessed in fancy (see above).120

SUGGEST: The verb suggest shows the same interpretative possibilities with respect

to control as the near-synonymous propose (see above). The first attested examples

testify to this. In (187a-b), both from the first decades of the nineteenth century, the

ing-complement may be exclusively controlled by the subject – at least, the dialogic

interaction in each of these examples indicates that the addressee of the act of sug-

120 Incidentally, though contemplate is of French (or Romance) origin, the develop-ment described causes its use to deviate from that of French contempler and is there-fore unlikely to be contact-induced. Likewise, I believe it cannot be argued that con-template has been attracted to the small group of intention verbs including intend and propose, because the semantic similarity between contemplate and these other verbs only arises as contemplate begins to take ing-complements – that is, prior to the appearance of ing-complements there is no ground for the analogical attraction that would have to explain the appearance of ing-complements (which is not to say that the group of verbs did not afterwards come to be conceived of as a coherent set).


gesting thinks of the suggester as the person who is also to carry out the action sug-

gested. In (187c-d), on the other hand, it is unclear whether or not the person utter-

ing the suggestion is also to be involved in its execution. As with propose, this am-

biguity can be resolved by making the subject of the ing-complement explicit by

means of the different-subject construction, which, as shown in (188), arose more or

less simultaneously with the controlled construction.

(187) a. When I suggested searching him, he said, “you find nothing but im-

plements of house-breaking.” (1817, POB)

b. I was called in to assist, and suggested sending for a constable [...]; the

prisoner fell on his knees, and begged I would not till Mr. West came

home. (1829, POB)

c. At length Tom’s friend, being of an ingenious turn of mind, suggested

sealing with ink; and the letter was accordingly stuck down with a blob

of ink, and duly handed by Tom, on his way to bed, to the housekeeper

to be posted. (1857, CLMETEV)

d. I next respectfully suggested writing to an agent at Torquay, but I was

met here by being reminded of the imprudence of taking lodgings

without first seeing them. (1859-60, CLMETEV)

(188) a. When I first knew him he wore his own black wool. I suggested his

shaving his head – he wore a wig after that, continually. (1820, POB)

b. the officer could not have searched thoroughly; it was myself who

suggested his going in again. (1825, POB)

c. Did she not suggest your sending for an officer? (1834, POB)

Once more the historical development of ing-complements is caught up in the

lexical history of the complement-taking verb. Specifically, suggest underwent a

semantic development from ‘inspire, cause one to be aware of’, as in (189), to ‘rec-

ommend, speak to endorse as a course of action’, as in (190). As (191) illustrates,

the first reading commonly implicates the second, especially when the mental con-

tent that is brought to one’s attention is or implies some possible action and the sub-

ject of suggest is a wilful human agent. While this implicational relation is in all

likelihood the cause of change, it also makes dating the change difficult. However, a

concomitant change in the behaviour of that-clauses following suggest can help in

this respect. As (192a-b) show, prior to 1800 that-clauses with suggest referred to

ideas or beliefs respecting factual states of affairs, whereas after 1800 that-clauses

could also be used to refer to possible courses of action, with suggest marking a di-


rective speech-act, as in (192c-d). This indicates that the semantic shift in suggest

probably took place around 1800, only one or two decades before the appearance of

the first ing-complements.121

(189) a. A jealous Woman believes every thing her Passion suggests. (1728,

CLMETEV)

b. I shall suggest an argument to this purpose, which, I believe, has not

been insisted on by any writer. (1779, CLMETEV)

(190) a. he then proposed to leave a deposit, and asked me what was usual; I

had not suggested it to him – I rather declined it (1832, POB)

b. I have heard that the king said to one who suggested another divorce to

him, ‘No, if the queen comes within the scope of the divorce, she also

comes within the pale of the scaffold.’ (1843, CLMETEV)

(191) a. Insensible to pity, and fearless of consequences, he was the ready in-

strument of every act of cruelty which the policy of that artful prince

might at once suggest and disclaim. (1776, CLMETEV)

b. This latent civil war already distracted the whole government, when a

scheme was suggested that seemed of mutual benefit to the hostile

brothers. (1776, CLMETEV)

(192) a. What reason have you to suggest that the prisoner is guilty of perjury?

(1779, POB)

b. It has been suggested, that the mortification was produced by the ap-

plication of fire. (1789, POB)

c. How came any body to be foolish enough to suggest, that a note should

be written to let the wife know where he was? (1799, POB)

d. what the prisoner said was not taken down, it was suggested that he

had better leave what he had to say until the time of his trial. (1811,

POB)

As to mechanisms of change, the examples in (190)-(191) above indicate that the

semantic change suggest underwent on the one hand made suggest semantically and

121 It is further worth pointing out that prior to 1800 suggest could combine with ing-clauses marking a recommended course of action, yet only on the condition of being embedded in a noun phrase that made them compatible with the selectional restric-tions of suggest at the time (cf. footnote 115 on enjoy). (xii) Did you suggest the propriety of his being received into St. Luke’s?

(1787, POB)


collocationally compatible with actions carried out by the suggester and/or the im-

plied addressee, and on the other allowed suggest to resonate to the analogical influ-

ence exerted by propose.

TRY: The first examples of try with ing-complement make their appearance in the

first decades of the nineteenth century, as shown in (193). These first examples con-

firm the view expounded in various reference grammars of English that try is used

with ing-complement only when it means ‘test’ as opposed to ‘strive for, attempt the

achievement of’, in which sense it typically combines with the to-infinitive, as in

(194) (cf. Quirk et al. 1985; Declerck 1991).

(193) a. Both hoped to get money by journalism, but opportunities failed; and

they tried lecturing (1819, Visser 1963-73: 1880)

b. but being long and thin, and of a delicate habit of body, he was obli-

gated to refrain from this recreation; so he betook himself to books,

and from reading he began to try writing (1821, CLMETEV)

c. Swift as the machine [i.e. the guillotine] is, it will not serve; the

Headsman and all his valets sink, worn down with work; declare that

the human muscles can no more. [...] Whereupon you must try fusillad-

ing; to which perhaps still frightfuller methods may succeed. (1837,

CLMETEV)

d. She tried keeping house with a female friend; then the double ménage

began to quarrel and get into debt. (1848, OED)

(194) a. as our little boat must come round this side of the island, it is at the

point on this side that I must try to find an entrance. (1841,

CLMETEV)

b. yet you deliberately insult him, and try to turn him out of my house.

(1908, CLMETEV)

The timing of the appearance of ing-complements with try is in part deter-

mined by a semantic development within the ‘test’ meaning of the verb, involving a

shift in semantic focus. Originally, try typically denoted a test for its own sake, in

which an object is subjected to critical examination to determine its inherent value,

strength, truthfulness etc. Increasingly, however, try could also be used to denote the

tentative use of an object, with an eye to achieving some goal to which that object is

a means. The distinction between these two sub-senses is illustrated in (195) and

(196) respectively. The second sub-sense is attested already in the Early Modern


period, but gained more currency and collocational variability in Late Modern Eng-

lish. As the examples with ing-complement clearly belong under this second sub-

sense, their appearance must to some extent have been conditioned by its increase.

(195) a. I judge of men by their actions, – a rule, I believe, it will not suit you to

be tried by. (1777, CLMETEV)

b. they have certainly got the credit of understanding more of love, and

making it better than any other nation upon earth; but, for my own part,

I think them arrant bunglers, and in truth the worst set of marksmen

that ever tried Cupid’s patience. (1768, CLMETEV)

c. At Noon try’d the Current, and found it set South 3/4 East, 1/3 of a

Mile per Hour (1768-71, CLMETEV)

(196) a. Her brother advised her to try platted packthread instead of hemp for

the soles (1796-1801, CLMETEV)

b. His Errand as you may suppose, is health. It had been often recom-

mended to him to try Bath, but his coming now seems to have been

chiefly in consequence of his sister Susan’s wish that he would put

himself under the care of Mr Bowen. (1805, CLMETEV)

c. They are ashamed in company, and so disconcerted, that they do not

know what they do, and try a thousand tricks to keep themselves in

countenance (1748, CLMETEV)

Still, the slight semantic shift in try is only a conditioning factor, not a mechanism

that could trigger the appearance of ing-complements. The only mechanisms proper

that can be invoked here are semantic analogy and broad paradigmatic analogy. Of

these, the former is somewhat problematic: the only semantic analogue available

taking ing-complements is attempt, as in (197), yet as the examples show, attempt

means ‘strive to achieve’, which is precisely the sense of try that does not come to

combine with ing-complements but takes to-infinitives (see (194) above).122, 123

122 Incidentally, attempt is quite frequent with ing-complement in the period 1780-1850, though it gradually drops out of use again after 1850, which is presumably why it was not attested in the sampled survey of ing-complementation described in Section 2 above. The appearance of ing-complements with attempt may be another instance of (broad) paradigmatic analogy – witness the examples in (xiii) – though it must be added that many of the early examples are in part also triggered by horror æqui (as in (197a) above).


(197) a. thus happy might he have continued, had he not been persuaded to

attempt becoming rich (1753, CLMETEV)

b. We bowled smoothly over the lawns I attempted describing in my last

letter, amongst myrtles in flower, that would have done honour to the

island of Juan Fernandes. (1783, CLMETEV)

As to paradigmatic analogy, note that the appearance of ing-complements again

triggers further paradigmatic innovation. In particular, the collocational behaviour of

try changes subsequent to the appearance of ing-complements, with try increasingly

combining with action nominals, as in (198), in the course of the nineteenth century.

(198) a. The rational woman, whose conversation on this occasion is to serve

her purpose more effectually than tears, knows better than to speak of

what her husband would probably consider a most unreasonable sub-

ject of complaint. [...] If conscious of beauty, she tries a little raillery,

and plays gently upon some of her husband’s not unpleasing peculiari-

ties, looking all the while as disengaged and unsuspecting as she can.

(1839, CLMETEV)

f. Catherine was near distraught; still she persisted that she must go

home, and tried entreaty in her turn, persuading him to subdue his self-

ish agony. (1847, CLMETEV)

g. When I wrote sternly and coldly, as I confess I frequently did at the

last, he blamed my harshness, and said it was enough to scare hung

from his home: when I tried mild persuasion, he was a little more gen-

tle in his replies, and promised to return; but I had learned, at last, to

disregard his promises. (1848, CLMETEV)

(xiii) a. if he attempted the least insult, he would be himself immediately the

executioner of vengeance on him. (1749, CLMETEV) b. as we are about to attempt a description hitherto unassayed either in

prose or verse, we think proper to invoke the assistance of certain ae-rial beings (1749, CLMETEV)

123 Another possible semantic analogue is risk, as hinted by Fanego (2007: 178). Note, however, that risk is at this point still extremely infrequent with ing-complements and that the majority of early examples do not combine with risk in its conative sense, ‘take the hazardous action of’, but in its sense ‘be in danger of’ (see risk).


A final word is in place on blocking. In its sense ‘attempt the achievement

of’, try also takes nominal objects, as in (199), so paradigmatic analogy would pre-

dict the appearance of ing-complements in this use as well. However, the to-

infinitive being extremely frequent with try, blocking is likely to have hindered ing-

complements from appearing in this (semantic) environment, though it may be noted

that occasional intrusions appear to occur, as illustrated in (200), suggesting a cer-

tain tension between blocking and paradigmatic analogy.

(199) a. He had her hand in his, and was bowing over it to kiss it, when Becky

started up with a faint scream as she caught sight of Rawdon’s white

face. At the next instant she tried a smile, a horrid smile, as if to wel-

come her husband; and Steyne rose up, grinding his teeth, pale, and

with fury in his looks. (1847-8, CLMETEV)

b. I would try a shot upon him were it not a waste of powder and lead.

(1893, CEN)

(200) a. What shall we do if she and Denton don’t get on? It will never answer

if she tries meddling in the kitchen – I must tell her. (1898, OED)

b. For years Dion’s brothers and sisters tried breaking into scene singing

songs by the Beatles, Creedence Clearwater Revival and the Doobie

Brothers with little success. (CB)

START: As the examples in (201) demonstrate, start began to combine with ing-

complements in the first half of the nineteenth century. Prior to the 1890s its fre-

quency remains low, with only two examples in CLMETEV. But with the start of

the twentieth century, the frequency of start with ing-complement soon rises (wit-

ness Table 9.2 above), to grow exceedingly frequent in Present-Day English.

(201) a. I had before this written to Rose how we had best start agitating (1833,

Visser 1963-73: 1896)

b. I in company with two of my fellow passengers started taking with us

some sea bread water &c. determined to camp out that night. (1834,

OED)

c. I’d no heart to sweeping an’ fettling, an’ washing pots; so I sat me

down i’ th’ muck – who should come in but Maister Weston! I started

siding stuff then, an’ sweeping an’ doing; and I expected he’d begin a

calling me for my idle ways as Maister Hatfield would a’ done (1848,

CLMETEV)


d. He started ‘prospecting’, struck gold, entered his claim. (1875, OED)

e. without giving us any further time to get scared, she started walking

along the spur, leaving us to follow her as best we might. (1887,

CLMETEV)

To understand how ing-complements with start entered the language, we

must look at the verb start itself. The now-common sense ‘begin’ was still rare for

start in the eighteenth century; instead the verb at this point typically meant ‘(cause

to) make a sudden movement, (cause to) take a sudden fright’, as is illustrated in

(202).

(202) a. it is your duty and office, whereto you are also qualified by the orders

of this commonwealth, to have the people as you have your hawks and

greyhounds, in leashes and slips, to range the fields and beat the bushes

for them, for they are of a nature that is never good at this sport, but

when you spring or start their proper quarry. (1656, CEMET)

b. It is alike troublesome to both the rider and his beast, – if the latter

goes pricking up his ears, and starting all the way at every object which

he never saw before. (1768, CLMETEV)

Start acquired the sense ‘begin’ around the beginning of the nineteenth century, just

prior to the emergence of ing-complements. This semantic development proceeded

along two paths. First, when used transitively, start could mean ‘introduce, put for-

ward’ with reference to conversational topics, ideas, problems, or objections, as

shown in (203). Presumably through bridging-examples such as (204a), this use also

began to accommodate noun phrases such as a dispute or an argument, as in (205b-

c), reflecting a new interpretation of start as ‘begin’. Second, when used intransi-

tively, start often denoted the beginning of movement and in this capacity typically

combined with a particle or prepositional phrase denoting a source, path or goal, as

in (205). From this it is a relatively small step to use start to mark the action of be-

ginning a journey or, by extension, a metaphorical trajectory, such as a story, a song,

life etc, as shown by the examples in (206).

(203) a. In the afternoon went and sat with Mr. Turner in his pew at St. Greg-

ory’s, [...] and heard Dr. Buck upon “Woe unto thee, Corazin,” &c.,

where he started a difficulty, which he left to another time to answer,


about why God should give means of grace to those people which he

knew would not receive them (1661, CEMET)

b. You alone, or almost alone, disturb this general harmony. You start

abstruse doubts, cavils, and objections (1779, CLMETEV)

(204) c. There is a social respect necessary: you may start your own subject of

conversation with modesty, taking great care, however, ‘de ne jamais

parler de cordes dans la maison d’un pendu’. (1751, CLMETEV)

d. at length he started a dispute upon the hackneyed comparison betwixt

blank verse and rhyme (1771, CLMETEV)

e. after a supper of the most homely fare, he tried to start an argument

with me, proving that everything for which I had interceded in my

prayer was irrelevant to man’s present state. (1824, CLMETEV)

(205) a. For I should be sorry that this Advisor [...] should like a Rabbit start

out of his Borough, and look about him, and then run in again, and

hide himself, and think no body observ’d him. (1674, LC)

b. our little company, hearing the sound of trumpets and kettle-drums,

[...] suddenly started from their seats, running directly to the terrace

(1749, CLMETEV)

(206) a. It was therefore highly unlucky for her, that she had gone to the very

same town and inn whence Jones had started, and still more unlucky

was she in having stumbled on the same guide (1749, CLMETEV)

b. Who Peter was, let that be told, And start from the beginning. (1798,

OED)

c. The high tone with which the tune started, died away in a quaver of

consternation. (1818, OED)

d. How many generous purposes, what bright and heart-thrilling visions

of beneficence and honour, does the young man, just starting in the

race of life, conceive! (1831, CLMETEV)

These semantic changes obviously brought start into the sphere of influence of the

aspectual verbs. A further contributory factor in this is the fact that, again from the

end of the nineteenth century, start also began to combine with to-infinitives, as

illustrated in (207). As is indicated by the fact that the infinitival verb was always a

verb of motion, these to-infinitives must initially have been purpose adjuncts rather

than complement clauses, and therefore comparable to the adjuncts found in (208).

However, a semantic and syntactic reinterpretation would have readily suggested

itself, especially given the aspectual meaning of start that was surfacing around the


same time in other uses. With such a reinterpretation, the similarity of start to other

aspectual verbs of course further increased, in that start now also patterned with one

of the complement types typical of aspectuals.

(207) a. he took to his heels and ran round the corner of Mr. Day’s house; and

as soon as Brown started to run away, the other started to run the other

way (1798, POB)

b. It was then about half after eleven; we staid there till half after twelve.

After this I started to go home, as I suppose. I cannot recollect any

thing more till half after five in the morning. (1810, POB)

(208) a. upon saying that, Gale started up to go away (1801, POB)

b. the other man made a start to run away, but Butler popped on him

(1805, POB)

With these various developments feeding the analogical pressure on start, it is not

surprising that start began to select ing-complements, as modelled after the aspec-

tual verbs. That start in fact underwent such influence is evident from a number of

later changes, all of which contribute to further convergence between start and the

aspectuals. First, there is the near-inevitable introduction of repeated-subject com-

plements, as shown in (209). Second, start copies the behaviour of keep by taking

present participles in a causative construction, as in (210). Third, start also occa-

sionally combines with ing-forms with a-prefix, as in (211) – a feature that also has

come to be associated with aspectual verbs (cf. Table 9.2).

(209) a. Yet what else can be expected, when the youngster starts his beer-

drinking with a “Fruhschoppen” at 10 a.m., and closes it with a

“Kneipe” at four in the morning? (1900, CEN)

b. on Susy’s curt refusal they started their rambling again, circuitously

threading the vague dark lanes and making for the Piazza and Florian’s

ices. (1922, CEN)

(210) it was the thought of it that started me laughing just now. (1892, CEN)

(211) and he starts a-tapping his foot too (1994, CB)


3.5. A bird’s-eye view of the diffusion of ing-complements

Table 9.3 summarises the individual analyses of the previous sections, thus giving a

bird’s-eye view of the major changes in subject-controlled ing-complementation

over a period of about 600 years. One general finding is that the diffusion of ing-

complements can be roughly broken down into four stages, which reflect gradual

shifts in the mechanisms driving diffusion. Interpreting this, we may say that ing-

complements spread with increasing ease, growing increasingly productive at the

same time. Broadly speaking, in Stage I ing-complements behaved much like other

bare abstract nouns and their distribution was determined by narrow paradigmatic

analogy. In Stage II, a number of semantic regularities emerged from the initial dis-

tribution of ing-complements and these gave rise to the first extensions, through the

locally operating mechanism of semantic analogy. In Stages III and IV, verbs no

longer needed to be semantically related to other ing-complement-taking verbs in

order to acquire an ing-complement themselves: their being transitive was a suffi-

cient condition. The mechanism driving diffusion, at that point, is broad paradig-

matic analogy, initially operating indirectly (Stage III), later directly (Stage IV).

Before addressing the relevance of these findings to our understanding of

diffusional change, complementation and the history of ing-complements, it is im-

portant to assess their validity. On the one hand, it is true that historical explanation

inevitably proceeds through the formulation of hypotheses that ultimately remain

difficult to test. The preceding discussions of individual verbs have amply demon-

strated that it is difficult enough to arrive even at plausible conclusions, let alone

certainties. The findings in Table 9.3 should, in that respect, be interpreted cau-

tiously. At the same time, however, if a series of hypotheses results in a coherent

picture, this adds credibility to those hypotheses. By studying a large body of

changes, it is therefore possible to circumvent the uncertainties that cling to individ-

ual developments, by fitting them into a bigger historical picture. One point of criti-

cism against historical explanations is that they are applied ad hoc (cf. Chapter 4).

Justified though this criticism is, it is not impossible to reduce the ad hoc character

of explanation: the more findings we can fit into our explanation, the more authority

it gets.

This being said, let me return to the shift in mechanisms observed in Table

9.3. Two points of comment should be added to this major generalisation, both relat-

ing to the nature of productivity and its relation to diffusion. First, it is remarkable in

Table 9.3 that there is no direct correlation between the degree of productivity of

DATE VERB MOST PLAUSIBLE MECHANISMS OF CHANGE EXAMPLE

STAGE I: narrow paradigmatic analogy

a1225 Love Narrow paradigmatic analogy (+ Blocking) (15a)

c1275 Begin Narrow paradigmatic analogy + Confusion bare infinitive (+ Blocking) (18a)

c1380 Continue Borrowing (+ Semantic / narrow paradigmatic analogy) (22a)

c1384 Cease Borrowing + Reduction of cease of + Indirect paradigmatic analogy (25a)

a1387 Hate Narrow paradigmatic analogy (29a)

a1398 Need Narrow paradigmatic analogy (+ Blocking) (31a)

c1425 Leave Narrow paradigmatic analogy (34a)

c1449 Forbear Narrow paradigmatic analogy (37a)

c1489 Escape Narrow paradigmatic analogy (39a)

STAGE II: semantic analogy

c1500 Leave off Semantic analogy (42b)

1515 Remain Paradigmatic / semantic analogy (45a)

1549 Give over Lexical change > Semantic analogy (49b)

1563 Require Narrow paradigmatic / semantic analogy (52b)

1574 Want Narrow paradigmatic / semantic analogy (+ Blocking) (56a)

1580 Fear Lexical change > Narrow paradigmatic analogy + Influence nominal fear (61a)

1597 Avoid Lexical change > Narrow paradigmatic / semantic analogy (64a)

1630 Defer Semantic / indirect paradigmatic analogy (68a)

1630 Miss Semantic analogy + Borrowing (70a)

1647-8 Omit Semantic analogy (73a)

a1649 Like Lexical change > Semantic analogy (74a)

1653 Prevent Semantic analogy + (Indirect) Paradigmatic analogy (79a)

1654-66 Decline Semantic analogy (83a)

1664 Prefer Semantic / paradigmatic analogy (86b)

1666 Bear Semantic / narrow paradigmatic analogy (92a)

STAGE III : indirect paradigmatic analogy

1666 Finish Semantic analogy + Indirect paradigmatic analogy (95a)

1666 Propose Borrowing + Indirect / broad paradigmatic analogy (99a)

1683 Give up Lexical change > Semantic analogy (104a)

1683 Keep Lexical change > Semantic / paradigmatic analogy (108a)

1684 Help Lexical change > Semantic analogy (115a)

1705 Go on Reinterpretation + Semantic analogy (122a)

1706 Intend Broad paradigmatic analogy + Influence nominal intention (126b)

1706 Stop Semantic / indirect paradigmatic analogy (131a)

1722 Keep on Lexical change > Reinterpretation + Semantic analogy (138a)

1724 Remember Indirect / broad paradigmatic analogy (141b)

1736 Mind Indirect / broad paradigmatic analogy (148a)

STAGE IV: broad paradigmatic analogy

1740 Regret Borrowing + Broad paradigmatic analogy (151a)

1777 Enjoy Lexical change > Broad paradigmatic analogy (161a)

1778 Commence Lexical change / borrowing > Semantic / broad paradigmatic analogy (165a)

1782 Dislike Semantic analogy (169a)

1792 Risk Borrowing + Broad paradigmatic analogy (173a)

1796-1801 Fancy (i) Broad paradigmatic analogy (176a)

1797 Contemplate Broad paradigmatic analogy (183a)

1817 Suggest Lexical change > Semantic / broad paradigmatic analogy (187a)

1819 Try Broad paradigmatic analogy (+ Blocking) (193a)

1833 Start Lexical change > Semantic analogy (201a)

1847-8 Fancy (ii) Reinterpretation > Indirect paradigmatic analogy (177a)

Table 9.3. Time-line of diffusion of bare controlled ing-complements across verbal predicates with dates of first-

attested examples and most plausible causes of change.


ing-complementation and the incidence of change. The period that produces the

most innovations is the time-span from 1630 to 1740 – a period of about one century

that accounts for almost half of all the innovations discussed. By contrast, the period

for which ing-complementation is claimed to be most productive is the period after

1740. This discrepancy is presumably due partly to methodology and partly to the

nature of diffusional change itself. On the methodological side, the period of greatest

productivity happens to follow directly on the period that is worst represented in the

corpus data (namely 1500-1640) (see Chapter 2). This means that some innovations

might in actuality predate 1630 but only turn up as innovations in the data after

1640. At the same time, there is a methodological bias against innovations after1740

as well. Because an innovative pattern needs time to gain frequency, it is very

unlikely that innovations after 1850 would have shown up as sufficiently frequent in

the synchronic slices (1640-1710 and, in particular, 1850-1920) to be selected for

further examination (see the description of methodology in Section 1). These two

methodological weaknesses may have led to a somewhat misleading concentration

of innovations in the period 1640-1800 (or thereabouts), a period that more or less

coincides with the period that actually shows the highest incidence of change in Ta-

ble 9.3.

Nonetheless, it may still be true that the pace of diffusion has been slackening

after 1740. A brief survey of present-day usage in LOB/FLOB on the basis of Ru-

danko’s (1989) list of ing-complement-taking verbs (itself an updated version of the

listings in Visser 1963-73) adds relatively few major innovations that must be dated

to the period after 1740. The most important additional verbs appearing with ing-

complements and not included in Table 9.3 are admit, consider, face, imagine, put

off, report, resent and resist. Even if we make the (quite unlikely) assumption that

all these verbs first appeared with ing-complements between 1740 and 1850, this

period sees fewer innovations than the period 1630-1740, which is of equal length.

This means that change has in fact been slowing down a little and that to some ex-

tent the paradoxical finding holds true that constructional productivity does not nec-

essarily correlate with a greater rate of diffusion.

Why should this be so? Provisionally, constructional productivity can be un-

derstood as the ease with which a construction spreads to new environments. Pro-

ductivity is higher if new environments have to meet fewer conditions to qualify as

potential environments for the spreading construction. This means that diffusion

proceeds by grace of increasing productivity. There is no theoretical need, however,

for an increase in productivity to correlate with quicker diffusion. For one thing, as

diffusion proceeds, the number of environments susceptible to the change decreases,


and as diffusion runs out of new environments to spread to, the pace of change must

necessarily slacken. On top of this, there are always the idiosyncrasies of a given

linguistic system. In the case of the English system of complementation, the diffu-

sion of ing-complements runs up against patterns that are already established, most

notably the to-infinitive. The evidence of this is apparent even from the environ-

ments where ing-complements do eventually appear. In the preceding discussion,

various instances of blocking have been pointed out (also marked in Table 9.3),

showing that blocking is consistently associated with environments that strongly

collocate with the to-infinitive. Indeed, the verbs at issue, love, begin, need, want

and try are or were all very common with the to-infinitive (see also the discussion of

the would-like-to-construction in Chapter 4). It is, in that light, conceivable that

many transitive verbs that of old combined with to-infinitival complements continue

to resist the diffusion of ing-complements, hindering the further progression of ing-

complementation, despite increased productivity.

Second, a word is in place on the precise interpretation of indirect paradig-

matic analogy, which as I will argue casts a slightly different light on the notion of

increasing productivity. It seems correct to say that a shift from semantic analogy to

broad paradigmatic analogy as driving mechanisms of diffusion implies an increase

of productivity, concomitant to a reduction of the conditions an environment has to

fulfil to be an acceptable host for ing-complements (in as far as semantic analogy

only affects transitive verbs, broad paradigmatic analogy also implicitly applies at

Stage II, so that semantic analogy is in essence an extra condition specific to Stage

II). It is entirely in accordance with expectations concerning diffusional change,

therefore, that the stage of semantic analogy precedes the stage of broad paradig-

matic analogy. Indirect paradigmatic analogy fits this picture in as far as it can be

seen as a preparatory step to broad paradigmatic analogy. For the innovations that

fall under this intermediary stage, it is in any case true that the end-result closely

resembles that of broad paradigmatic analogy: an ing-complement is used in an en-

vironment where it is neither sanctioned by semantic analogy nor by narrow para-

digmatic analogy.

In actual fact, however, indirect paradigmatic analogy involves two steps,

neither of which requires broad paradigmatic analogy. The first step is the appear-

ance of some non-bare ing-complement (either a definite or an own-subject com-

plement), typically on the basis of the collocational behaviour of a verb (that either

combines with definite noun phrases denoting subject-controlled actions, or noun

phrases with possessive determiner denoting actions by the referent of the deter-

miner). This is just another type of narrow paradigmatic analogy, for the paradig-


matic tie is between a type of noun phrase and a type of ing-complement that for-

mally and functionally closely resemble one another. The second step, then, is the

introduction of bare ing-complements on the basis of paradigmatic associations be-

tween bare and non-bare ing-complements in other environments (e.g. with other

complement-taking predicates). This, too, is not broad paradigmatic analogy, for

bare ing-complements do not spread exclusively on the basis of the transitivity of

the target environment but also on the basis of their paradigmatic link with other

functionally similar types of ing-complement. In this light, the stage associated with

indirect paradigmatic analogy is perhaps not so much marked by an increase in pro-

ductivity as by speakers’ recognition of new regularities, namely the connections

between bare and non-bare ing-complements.

Importantly, then, it is too simplistic to see diffusion as driven merely by in-

creasing productivity in terms of ever loosening selectional restrictions. Diffusion

does not necessarily proceed by grace of a dropping away of conditions that held

previously. It is also advanced by the discovery of new regularities that can generate

new innovations, without old regularities necessarily being abandoned. Increasing

productivity then may either involve the loss of conditions, or the introduction of

new conditions as alternatives to the old ones. Put differently, diffusion is not neces-

sarily a matter of existing ‘rules’ being generalised; it may also proceed as addi-

tional rules or regularities arise.124

Indeed, strictly speaking, only Stage IV in the diffusion of ing-complements

is marked by a loss of conditions; the other stages involve new alternative conditions

for change. This is shown in Table 9.4, which makes explicit the conditions verbs

have to meet at different stages of diffusion to qualify as potential ing-complement-

taking predicates. Notice that broad paradigmatic analogy is implicit at all stages, as

it is logically implied by narrow paradigmatic analogy and indirect paradigmatic

analogy, while semantic analogy too never occurred without its additional support.

The transition to Stage IV, then, is marked by broad paradigmatic analogy becoming

operative as an independent mechanism of change. To be sure, the point of Table 9.4

is not that there is a fundamental difference between the transition to Stage IV and

the transitions to Stages II and III. In each case, what happens is that on the basis of

existing usage speakers infer new generalisations which may or may not encapsulate

124 Yet another way of putting this is that by the addition of new alternative condi-tions old conditions are transformed from necessary to sufficient conditions.


Stage I (+BPA)

+NPA

Stage II (+BPA)

+NPA or

(+BPA)

+SA

Stage III (+BPA)

+NPA or

(+BPA)

+SA or

(+BPA)

+IPA

Stage IV +BPA

Table 9.4. Conditions for complementation by ing-complements (BPA = broad

paradigmatic analogy; NPA = narrow paradigmatic analogy; SA = semantic anal-

ogy; IPA = indirect paradigmatic analogy).

(and thereby replace) the old ones. It is, however, a refinement of our understanding

of the relation between productivity and diffusion.

Turning now to a different aspect of the diffusion of ing-complements, the

summary of results in Table 9.3 also points to the importance of local developments.

Indeed, the transitions from Stage I to Stage II and from Stage II to Stage III hinge

on local changes through local regularities – that is, regularities that apply to only

part of the domain of ing-complementation. Stage II is characterised by speakers

distilling different groups of semantically-related verbs and inferring that they com-

bine productively with ing-complements, which results in the emergence of con-

structional clusters organised around a number of different verb classes. Stage III

involves the recognition of a new regularity, that cuts across the verb classes of

Stage II but that still does not apply to all ing-complement-taking predicates.

Going a step further, it is not only the analogical mechanisms that result in

innovations, but also a number of ad hoc mechanisms, whose local character is even

more outspoken. Most prominently, these include of course borrowing and reinter-

pretation, both of which draw on linguistic phenomena that are alien to the lan-

guage-internal system of complementation (foreign complement constructions, ad-

verbial clauses). Interestingly, however, these mechanisms do not operate in full

independence of the major patterns of change. Reinterpretation, for instance, gives

rise to ing-complements with aspectual verbs (go on, keep on) at a time when ing-

complements are already common with aspectual verbs and are thereby partly sanc-

tioned by the existing system. Borrowing, likewise, interacts with the general pro-

gression of diffusion. For example, ing-complements appear with miss to replicate

French usage at a time when negative implicative verbs become productive with ing-

complements; they appear with propose at a time when indirect paradigmatic anal-


ogy (which co-sanctions the use of ing-complements with propose) is coming into

operation; and they appear with regret at a time when use of an ing-complement is

sufficiently sanctioned by broad paradigmatic analogy (which similarly co-sanctions

the use of ing-complements with regret). Borrowing thus does not seem to strongly

interfere with the existing system at a given point in time. Rather, it seems to be the

state of the system that determines which borrowings are possible.125 Another type

of local developments are the various lexical developments involving shifts in the

semantics and collocational patterning of individual lexical items. Obvious examples

are avoid, fear, enjoy and start. Of course, here too it is true that lexical change only

gives rise to ing-complementation if the new use of a lexical item invites ing-

complementation on the basis of existing patterns of usage.

Still, if borrowing, reinterpretation and lexical changes only seem to feed the

diffusion of ing-complements in as far as they fit the existing system, this does not

rob them of having potentially further-reaching impact on that system. Consider, for

instance, the case of aspectual begin. By hypothesis, we can think of the history of

ing-complementation with begin as an interplay between the pressure of semantic

analogy (to other aspectuals) and blocking (due to the to-infinitive that is in frequent

use with begin). Presumably, the sparse use of ing-complements with begin in the

Early Modern period and their rise after 1750 then reflects the shifting balance of

power between these two factors, with blocking yielding to analogical pressure. If

this view is correct, one of the reasons why the effect of blocking eventually gets

surpassed is that the sanctioning model underlying the use of ing-complements has

grown stronger, which in part is due to the fact that more and more aspectual verbs

have come to select ing-complements, itself a consequence of various local devel-

opments (e.g. keep, go on, keep on). In a similar vein, the appearance of ing-

complements with avoid is probably due to narrow paradigmatic analogy, but it adds

another negative implicative verb to the set of negative implicative verbs already

selecting ing-complements and may thus have contributed to such verbs being con-

ceived of as a semantic class sanctioning ing-complementation. Because of local

developments, then, diffusion does not proceed solely through the internal dynamics

of the spreading construction but is in part steered by changes that happen independ-

ently.

125 Only the appearance of ing-complements with cease and continue could be re-garded as an infringement on existing usage (ing-complements not being productive at the time), but here the ing-form is inspired by a participle in the Latin source con-struction.


What does all this mean for our understanding of the synchronic system of

complementation? It would make no difference, synchronically speaking, how the

use of ing-complements in a given environment arose, whether through local or

global developments, whether as a result of some local analogy or through a highly

abstract generalisation, just as long as the end result fully fits some global regularity.

However, this happens not to be the case. There are synchronic phases in the history

of ing-complementation where the productive use of ing-complements is sanctioned

by local regularities only – most clearly so in Stage II, when ing-complementation is

organised around constructional clusters with semantic constraints on their matrix

verbs (true, the global regularity of broad paradigmatic analogy holds throughout,

but it does not motivate innovations until Stage IV). Even in Present-Day English,

certain patterns do not fit any broad generalisation but are nevertheless motivated by

locally operating regularities. For example, as a violation of broad paradigmatic

analogy, the construction of the type which questions do you want answering (dis-

cussed under need and want above) is a case in point. Similarly, among the aspectual

verbs we have to confront the question which categorial ties are strongest: those

connecting certain ing-complements to their gerundial origins or those connecting

them to other ing-complements with aspectuals, including the ing-complements that

are not of gerundial origin (cf. Chapter 7)? As a final example, the effects of block-

ing continue to defy overall regularity in the distribution of ing-complements. Ing-

complementation is productive only where it is not barred by blocking – a factor that

cannot be captured in a description in terms of the regularities of ing-

complementation only.

Translating this to the organisation of the system of complementation as a

whole, it is clear that overly neat conceptions of this system have to be abandoned.

Most tellingly, the initial distribution of ing-complements has been largely shaped

by a factor that lies entirely outside the system of complementation, namely the af-

finity between gerunds and abstract nouns. It is this affinity that largely determined

the use of ing-complements in Stage I and that also gave rise to the first productive

classes of ing-complement-taking verbs in Stage II (it will be noted that almost all

productive classes in Stage II ultimately go back to uses motivated by narrow para-

digmatic analogy in Stage I). It cannot be maintained, therefore, that the system of

clausal complementation is, as a linguistic subsystem, functionally organised by

principles strictly internal to that system. The use of gerundial ing-complements has,

at least in part, been determined by the organisation of another subsystem, that of the

noun phrase. This is endemic to diffusional change in general: whatever pattern dif-

fuses in a given subsystem of the language is likely to originate from outside that


system and will continue to bear the traces of its historical origins. In other words,

grammar cannot be compartmentalised. As to complementation, it should be clear

from this that the organisation of the system of complementation cannot be under-

stood in isolation.

4. Conclusions

In this chapter, I have analysed the diffusion of subject-controlled ing-complements

of gerundial origin. The conclusions are in line with earlier findings. Diffusion pro-

ceeds as a result of new regularities being inferred from existing usage. Generally, a

given regularity is worked out in language use in a relatively short time, but no

sooner is a regularity applied to a group of environments, than it can trigger some

new regularity that again gives rise to new uses. It is also true that by the end of the

process of diffusion, environments have to meet fewer conditions to qualify as hosts

to the spreading construction. This means that diffusion becomes easier as it pro-

ceeds, though its progress does not necessarily become quicker.

From the historical developments, it is clear that the use of a complement-

type is not determined by a single principle, but is the outcome of a complex interac-

tion between various factors, including different kinds of analogy and blocking. Of

course, this finding has synchronic relevance as well. It implies that the synchronic

use of a complement-type result from more than one regularity operative at a given

time, and also that a synchronic distribution may reflect an amalgam of different

considerations that may have functioned at different times. In a sense, the very fact

that diffusion proceeds indicates that speakers continue to attempt making sense of

synchronic usage, but also that the regularities of synchrony are provisional and in-

complete.

CHAPTER 10: The diffusion of ing-complements II: Inte-grated participle clauses126

In the course of the Modern and Present-Day English period, participle clauses,

which are typically used as adverbial clauses, have in a variety of constructions

come to be more closely integrated in the syntax of the matrix clauses they combine

with, thus developing into another type of subject-controlled ing-complement. The

construction at issue is illustrated in (1):

(1) a. The receptionist is busy filling a fifth box. (CB)

b. I am tired hearing of the Duchess of Chiselhurst’s ball (1899, CEN)

c. Mr Jones said because he was not being properly paid he had trouble

getting a housing loan and feared he might lose his new home. (CB)

Being participial in origin, the ing-complements in (1) cannot be regarded as nomi-

nalisations. This appears from their inability to alternate with a noun phrase (con-

sider, for instance, *I am tired all this talk of the Duchess of Chiselhurst’s ball). In

this respect, they differ from other ing-complements, which occur in the argument

slots of transitive verbs (cf. Chapter 9) or with prepositions, as in (2).

(2) a. Along with the rest of his partners, he will have to weigh up whether to

go public now, or risk remaining private. (CB)

b. I think people are tired of hearing about it. (CB)

At the same time, the ing-clauses in (1) differ from adverbial participle clauses, be-

cause their relation to the matrix predicate is not one of adverbial modification but

of complementation – notice, for instance, that (1b) above is semantically very close

to (2b). In what follows, I will refer to the ing-complements in (1) as integrated par-

ticiple clauses, or IPCs for short – the term ‘integrated’ referring to the higher de-

gree of syntactic integration between matrix and subordinate clause that is character-

istic of complements. In light of the close tie between ing-complements and noun

phrases (Chapter 7, 9), IPCs are of course of special interest.

IPCs as in (1) are uncharted territory in English grammar. The type in (1a),

headed by the adjective busy, has been recognised as an adjective complement con-

126 This chapter is adapted from De Smet (forthc. c).

440 – CHAPTER 10: The diffusion of ing-complements II

struction by Quirk et al. (1985: 1230) and Huddleston & Pullum (2002: 1259). Vis-

ser (1963-73: 1127-8) recognises the busy-type in (1a), the tired-type in (1b), and

further spots instances with long and late, yet classes all of these together with vari-

ous unrelated constructions. Noteworthy is also the fact that Quirk et al. give an ad-

ditional example with the adjective fortunate (“we’re fortunate having Aunt Mary as

a baby-sitter”, 1985: 1231), which indicates that they consider the construction pro-

ductive, though ironically, the pattern with fortunate is unattested in any of the large

corpora of Present-Day English. No further description is given of the kinds of

predicate that occur in the IPC-construction, nor is there any discussion of what dis-

tinguishes the construction from other construction types such as gerund clauses and

adverbial participle clauses.

The purpose of this chapter is to explore the use and history of IPCs. Because

IPCs do not fill nominal argument positions, their status as complements is less

straightforward. For that reason, Section 1, which is synchronically oriented, dis-

cusses criteria for distinguishing IPCs from adverbial clauses. Applying these crite-

ria to different patterns reveals that some ing-clauses take an intermediate position,

suggesting a syntactic cline from adverbial to complement. With respect to the dis-

tribution of IPCs, the synchronic discussion further shows that use is extremely che-

quered, organised around a variety of highly specific constructions that have little

overall coherence. Section 2, then, turns to the history of IPCs. The origins of vari-

ous IPC-constructions are traced, showing that they emerged one by one in the

course of the Modern English period. It is argued that, initially, the mechanism of

change is reinterpretation of adverbial participle clauses. However, just as in the

development of nominalised ing-complements (discussed in Chapter 9), at a later

stage some regularities begin to assert themselves and patterns are extended through

various forms of analogy. The question what this means for the categorial status of

ing-complementation is re-addressed.

The method of data-gathering applied in this chapter has been less systematic

than in the previous chapters. The reason is that IPCs are very marginal compared to

other ing-clauses, which means that they are difficult to find by means of corpus

searches, except by searching for the predicates they might combine with. The

method employed, therefore, has been to keep track of instances encountered inci-

dentally, in combination with specific corpus searches on potential complement-

taking predicates.

CHAPTER 10: The diffusion of ing-complements II – 441

1. Synchronic characterisation

Looking at the contrast between IPCs and adverbial participle clauses, it should be

pointed out from the outset that IPCs do not differ from all adverbial participle

clauses to the same extent. For one thing, adverbial participle clauses can themselves

be divided into adjuncts and disjuncts. This distinction, based originally on

Greenbaum (1969: 15-25) and Quirk et al. (1985: 1070-3), is illustrated in (3), with

(3a) exemplifying an adjunct, and (3b) presenting a disjunct.

(3) a. See how many words of four or more letters you can find using the

letters above (CB)

b. Fishermen in Scotland have taken a tennis club to court, claiming that

its floodlights are driving away the fish in an angling river. (CB)

Particularly adjuncts, which, unlike disjuncts, form part of the propositional content

of the matrix clause (Verstraete 2004), share certain features with IPCs, as will be-

come clear in the following discussion. Let us start, therefore, by contrasting IPCs,

adjuncts, and disjuncts as three broad and somewhat idealised, maximally differenti-

ated categories (Section 1.1), only then to turn to the more subtle differences be-

tween the IPCs in more specific environments, as well as the more problematic cases

where the distinction between IPC and adverbial participle are obscured (Section

1.2).

1.1. Complements and adverbial clauses

One set of characteristics distinguishes IPCs from disjuncts but not from adjuncts.

Firstly, IPCs are not separated from the matrix clause by an intonational boundary

or, in writing, a comma. In this they differ from disjunctively used participle clauses

(cf. Kortmann 1991). Thus it is (intuitively) evident that adding an intonational

break changes the semantic relation between the participle clause and its matrix

clause. Compare example (4) below to (1b) above: in (4) the participle clause gives

the reason why (the speaker assumes) the subject of the matrix clause must be tired,

while in (1b) the participle clause specifies what the matrix subject is tired of.

(4) You must be tired, wandering about on the hills as you do! (1887,

CEN)


While the difference between IPCs and disjuncts is clear, the absence of an intona-

tional boundary does not distinguish IPCs from adjunctively used participle clauses,

which can similarly form a single intonation unit with the matrix clause, as in (5).

(5) Bold, 31, now in Glenochil jail, claimed he hurt his knee playing at

Perth prison. (CB)

Secondly, IPCs, like adjuncts but unlike disjuncts, can be the focus of nega-

tion and polarity questions in the matrix clause. Compare the possible interpretations

of the IPC in (6), the adjunct in (7) and the disjunct in (8):

(6) a. Yet, for most people, the impression gained of Mr Hanley is that he

has trouble deciding which shoe to put on first in the morning. (CB)

b. Mr Hanley doesn’t have trouble deciding… (‘Mr Hanley has trouble

doing some things, but he does not have trouble deciding…’)

c. Does Mr Hanley have trouble deciding… ? (‘Mr Hanley has trouble

doing some things, but does he have trouble deciding…?’)

(7) a. Ian went to check the cars and found that they were missing their wip-

ers and he went and told the hotel manager, who came out looking very

worried. (CB)

b. The manager did not come out looking very worried. (‘the manager

came out but he didn’t look very worried’)

c. Did the manager come out looking very worried? (‘the manager came

out but did he look very worried?’)

(8) a. Yesterday the group issued its strongest warning yet, telling foreigners

to leave the country. (CB)

b. The group didn’t issue its strongest warning yet, telling foreigners to

leave the country (*’The group issued its strongest warning yet but did

not tell foreigners to leave the country’)

c. Did the group issue its strongest warning yet, telling foreigners to leave

the country? (*’The group issued its strongest warning yet, but did it

tell foreigners to leave the country?’)

Thirdly, neither IPCs nor adjuncts can have explicit subjects – again in con-

trast to disjuncts, which do readily take their own subjects, as in (9a). Examples (9b-

c) illustrate the effect of adding a subject to an IPC and adjunct; with the former the


result is clearly ungrammatical, with the latter the result is acceptable only under a

disjunct reading.

(9) a. It was only to be expected, he being thirty-five years older than me, but

I can tell you expecting makes no difference. It’s still an outrage to be

left without a husband. (CB)

b. *Mr Hanley has trouble his wife deciding which shoe to put on first in

the morning.

c. *The manager came out his assistant looking very worried.

Moreover, when they do not express their own subjects disjuncts can be controlled

by the speaker/interlocutor, instead of a participant in the matrix clause, as in (10a),

or the controller can be the matrix clause as a whole, as in (10b) (Kortmann 1991).

Again, neither adjuncts nor IPCs have this possibility.127

(10) a. Speaking of which, why can’t veterans just forgive and forget Japan

over its treatment of allied prisoners of war so long ago? (CB)

b. In North Antrim at the last election, about 33,600 people voted for Un-

ionist parties and 8,400 for nationalists, indicating that Catholics form

about 20 per cent of its population. (CB)

A second, more important set of characteristics distinguishes IPCs from ad-

verbial participle clauses in general – i.e. disjuncts as well as adjuncts – establishing

them as a separate phenomenon. Firstly, omitting the IPC from the main clause will

affect the meaning of the main clause predicate. Specifically, the IPC fills or maps

onto a semantic role evoked by the predicate – in (11a), the ing-clause denotes what

the subject is fed up about, in (11b) what the subject needs help with. Its omission

thus brings about a shift from a (semantically) transitive interpretation to an intransi-

tive one, affecting the lexical semantics of the predicate as such. This is particularly

clear in (12): Without IPC, tired in (12a) would by default designate a state of

physical (and possibly mental) exhaustion, rather than a feeling of weariness with

127 Incidentally, adjuncts can occasionally be controlled by the subject of a previous clause, as in the following example: (i) Third party developers were reluctant to release applications for Win-

dows, complaining that it was slow. This was certainly true running on the 6MHz ATs of the day (ICE-GB).


relation to one specific activity. Happy in (12b) without IPC would designate a state

of positive psychological excitement, but with IPC designates an attitudinal relation

between a sensing subject and an object of emotional judgement. In contrast to all

this, omission of a participle clause has no effect on the semantics of the matrix

clause when the participle clause functions as an adjunct or disjunct, as can be tested

by leaving out the participle clauses in (13) and (14) respectively.

(11) a. I was fed up sitting at the station doing nothing. (CB)

b. Thousands like us need help finding someone special. (CB)

(12) a. The day I say I’m tired playing for my country is the day I hang up my

boots. (CB)

b. I wasn’t happy being described as cute, but seeing as I got the part I

didn’t care. (CB)

(13) a. As ever he stormed away refusing to speak, along with his sulking

team. (CB)

b. I shall make so much money exploring Africa I shan’t know what to do

with it. (1902, CLMETEV)

(14) a. In a classroom in Farmington […] about a dozen farmers are lined up

in desks, looking at charts of farm prices projected on a screen. (CB)

b. Cyril glanced at Amy, who averted her head, putting spoons into three

saucers. (1908, CLMETEV)

Judging by the examples in (11) and especially (12), the semantics of IPC-

constructions resemble those of complement constructions in general. The semantic

relation between the complement and the matrix clause is typically determined by

the semantics of the matrix clause predicate. More precisely, a relation is predicated

between the matrix clause subject and the situation described by the complement.

This relation – abstractly defined – consists in (real or potential) psychological or

physical energy being either exerted by the matrix clause subject on the situation of

the complement or triggered by the situation of the complement in the subject.

Secondly, another indication of the higher degree of integration of IPCs in the

matrix clause is the fact that IPCs allow wh-extraction (cf. Los 2005: 48-9, who,

following Chomsky 1980, uses this argument to assert the argument status of to-

infinitives in Old English). An example of wh-extraction is given in (15). As to ad-

verbial participle clauses, both adjuncts and disjuncts resist this operation (albeit to

different degrees; see below). This is illustrated in (16) and (17).


(15) It’s difficult to get the actual legislatures to act, and therefore one has

to activate the legislators to do something. And this is what we are

busy trying to do, and we have been preparing for this for the last 18

months. (CB)

(16) a. Police late yesterday were still looking for the youth, who escaped on

foot wearing a baseball cap and a false beard. (CB)

b. *The false beard the young man escaped on foot wearing.

(17) a. The operation was successful for the commandos who managed to

push the enemy (infantry and armoured units) onto the back foot and

keep them at bay for three days, taking over the town of Wyndham,

WA, in process. (CB)

b. *The town that the commandos kept the enemy at bay for three days,

taking over in process.

Thirdly, IPCs can never be introduced by subordinating conjunctions such as

when, while, if or though. Adverbial participle clauses allow this option, provided

that the semantics of the inserted conjunction do not disagree with the meaning of

the sentence. Compare the effect of inserting a conjunction before the IPC in (18)

and the adverbial participle clause in (19).

(18) a. The old man, 65-year-old Anatole Pierre, is busy digging up the roots

of a felled mahogany. (CB)

b. *The old man is busy while/when digging up the roots of a felled ma-

hogany.

(19) a. Stan and his remaining survivor a Chinese man, had radio contact with

the Allied forces who tried to send submarines in to rescue them but

lost men doing that. (CB)

b. The Allied forces lost men while/when doing that.

Fourthly, adjuncts – though not disjuncts – can often be questioned by a wh-

question using adverbial interrogative pronouns such as how, when or why, as in

(20). IPCs, by contrast, generally resist questioning – compare (21a-b). If IPCs are

questioned, the most natural interrogative pronoun to use is what – as is illustrated in

(21c) – but notice that questioning then typically requires the addition of a preposi-

tion to the matrix clause predicate, which suggests that what is questioned in such

examples is not the IPC but a semantically similar gerundial construction with

IPC Adjunct Disjunct

+ + - i. Ing- clause is not separated from the matrix clause by an intonational boundary

+ + - ii. Ing-clause falls under the scope of negation, polarity questions and epistemic modals in the matrix clause

+ + - iii. Ing-clause has no explicit subject

+ - - iv. Omission of ing-clause changes the semantics of the matrix clause predicate, broad-ening the scope of the predication

+ - - v. Ing-clause allows wh-extraction

+ - - vi. Ing-clause resists insertion of subordinating conjunctions

+ - n/a vii. Ing-clause cannot be questioned by how or other adverbial interrogative pronouns

Table 10.1. Characteristics of IPCs, adjuncts and disjuncts.


preposition (e.g. the new group has had no more success in prying men away from

…).128

(20) a. We started the season aiming for the top four and a place in Europe.

(CB)

b. How did we start the season?

(21) a. But, despite some influence among Protestant swing voters in the

Northeast, the new group had no more success prying men away from

their entrenched partisan loyalties than did the Greenbackers and the

other small, alternative parties of the day. (CB)

b. *How/when has the new group had no more success than the Green-

backers …?

c. What has the new group had no more success in than the Greenbackers

…?

In order to summarise the discussion so far, the various characteristics of

IPCs, adjuncts and disjuncts are repeated here in Table 10.1. At this point, it should

be clear that the differences setting apart IPCs from adverbial participle clauses are

exactly the characteristics of complement clauses in general. That is, under the vari-

ous tests listed in Table 10.1, IPCs and more familiar types of complement clauses

display roughly the same behaviour. This is shown in the following examples where

the tests distinguishing IPCs from adjuncts (iv-vii in Table 10.1) are applied to con-

structions with a gerundial and infinitival complement ((22) and (23) respectively).

As is clear from these examples, there is little difference between IPCs and gerun-

dial or infinitival complements in terms of the relation that obtains between the ma-

trix clause and the dependent clause.

(22) a. She’ll throw her toys around and will enjoy making a mess with her

dinner. (CB)

b. *She’ll throw her toys around and will enjoy. [Omission]

c. The mess she will enjoy making. [Wh-extraction]

d. *She’ll throw her toys around and will enjoy while making a mess with

her dinner. [Adverbial subordinator]

128 An alternative questioning strategy avoids the addition of a preposition and in-stead makes use of extraposition, keeping the IPC-slot filled with semantically ‘empty’ doing and using the question word what as its extraposed object, e.g. what did the new group have no more success doing than the Greenbackers …?


e. *How does she enjoy? [Adverbial interrogative]

(23) a. I guess I wanted to be a hero. (CB)

b. *I guess I wanted. [Omission]

c. The hero that I wanted to be. [Wh-extraction]

d. *I guess I wanted while/in order to be a hero. [Adverbial subordinator]

e. *How did I want? [Adverbial interrogative]

At the same time, it must be recognised that in at least some respects IPCs differ

from gerundial and infinitival complement constructions. One difference is that

other complement constructions often more strongly resist omission of the comple-

ment clause than IPC constructions, as the main clause predicate not just changes

meaning, but in fact becomes ungrammatical (as in (22b) and (23b) above). Further,

IPCs seem less inhospitable to the insertion of adverbial material between the com-

plement-taking predicate and the complement clause, an operation which at least

gerundial ing-complements tend to resist, as is illustrated in (24).129 Regular com-

plement clauses standardly allow questioning by what, unlike IPCs, which some-

times simply cannot be questioned by a wh-pronoun or at best require the addition of

a preposition – compare (21) above with (25). Other complement types can have

their own subject, as shown in (26). And, finally, because IPCs are not nominalised,

they resist some additional operations that are allowed by the more regular comple-

ment types, such as (pseudo-)clefting or pronominalisation, as in (27) and (28).

(24) a. Police departments don’t change easily, and Williams had trouble early

in his tenure identifying commanders who wanted to follow the com-

munity policing model. (CB)

b. In Bao Loc – a highlands hole-in-the-wall four hours north of Ho Chi

Minh City, dozens of families were busy last month attempting to turn

homes into mini-hotels. (CB)

c. *She enjoys every day making a mess with her dinner.

d. I wanted all my life to be a hero.

(25) a. What will she enjoy?

b. What did I want?

129 The fact that to-infinitival complements less strongly resist the insertion of ad-verbial material between main verb and complement may reflect the lower degree of ‘nouniness’ of to-infinitives as compared to gerund clauses (Ross 1973), or may reflect the fact that, like IPCs, to-infinitival complements historically derive from adverbial constructions (see Section 2).


(26) a. Suzie’s parents find it hard to enjoy her making a mess with her dinner.

b. I guess everybody wanted him to be a hero.

(27) a. What she’ll enjoy is making a mess with her dinner.

b. What I wanted was to be a hero.

c. *What Williams had trouble was identifying commanders who wanted

to follow the community policing model.

d. *What dozens of families were busy was attempting to turn homes into

mini-hotels.

(28) a. She’ll enjoy it.

b. (?) I guess I wanted it.

c. *Williams had trouble it.

d. *Dozens of families were busy it.

Constructions where a gerundial ing-clause is introduced by a preposition, as

in (29a), resemble IPCs even more closely, because it is often the same predicates

that take IPCs that also occur in the prepositional pattern, without much appreciable

difference in meaning. At the same time, the characteristics relating to the nominal

nature of the gerund also apply to the patterns with preposition, and serve to distin-

guish IPCs from their prepositional variants. Tests demonstrating the similarities

between IPCs and prepositional variants are applied in (29b-e); tests highlighting the

differences in (29f-i).

(29) a. I could never get tired of looking at a hog that big (CB)

b. (?) I could never get tired. [Omission]

c. The hog that I could never get tired of looking at. [Wh-extraction]

d. *I could never get tired of while/when looking at a hog that big. [Ad-

verbial subordinator]

e. *Why/how/when could I never get tired of? [Adverbial interrogative]

f. What could I never get tired of? [What-question]

g. My parents got tired of me looking at that hog. [Explicit subject]

h. What I could never get tired of was looking at that hog. [Pseudo-cleft]

i. I could never get tired of it. [Pronominalisation]

It may be concluded that the differences between IPCs and adverbials are suffi-

ciently convincing to treat IPCs as a distinct phenomenon, closely resembling other

means of clausal complementation. Differences that exist between IPCs and other

complement clauses mostly relate to the non-nominal status of IPCs.


1.2. Between complements and adverbial clauses

While the characteristics listed in Table 10.1 can be used to set IPCs apart from both

adjuncts and disjuncts, they do not exhaust the differences that exist between com-

plement and adverbial constructions and say as yet nothing about the possibility of

intermediate category membership. With a view to refining syntactic description, it

is therefore crucial also to examine the various clusters of related constructions that

may qualify – to various degrees – as IPCs. It will become clear from the discussion

that not all of the tests used to distinguish IPCs from adverbial participle clauses

yield clear-cut results, and that not all constructions are equally clear examples of

IPCs.

At least superficially, IPC-constructions can be classified by the syntactic

structure of their main clause predicate. Thus, as a starting point, two major groups

of (sometimes only seemingly) similar constructions can be distinguished. The first

group consists of predicatively used adjectives combining with an IPC that functions

as postmodifier or complement to the adjective. Amply illustrated above, the con-

struction type under consideration is exemplified once more in (30).

(30) New Man, that sociological phenomenon said to treat women as his

equal and who is happy sharing domestic chores, was pronounced

dead yesterday. (CB)

Semantically, the adjectives used in constructions of this kind fall into a number of

sub-categories. The first set of adjectives express an emotive relation between the

subject of the matrix clause and the situation designated by the ing-clause, specify-

ing how the former is emotionally affected by the latter. These adjectives include

bored, comfortable, fed up, happy (as in (30)), hopeful, tired, uncomfortable and

unhappy. A second group of adjectives does not denote a psychological state, but an

external judgement predicated of the subject in respect of the situation denoted by

the ing-clause. Such adjectives include right, as in (31a), better off, brave, and lucky.

A third group of adjectives express a relation of active occupation of the main clause

subject in the situation denoted by the ing-clause. These include most notably the

adjective busy, as in (18a) above, but also, more marginally, employed, engaged and

occupied. Potential members of this group are also the expressions be gone, be off

and be out, as illustrated in (31b). Semantically related to the adjectives of active

occupation is a fourth group denoting the manner or degree to which the matrix

clause subject is advancing or has advanced in realising the situation denoted by the


ing-clause. Adjectives of this kind are late, as illustrated in (31c), long, quick and

slow; and the group might be further expanded with the expressions be done and be

finished, as illustrated in (31d-e).

(31) a. The Supreme Court in a five to four decision declared that the officer

was right in arresting her; he was right in putting her in handcuffs; he

was right taking her into custody, taking her to jail; and it was right to

force her to post a bail of more than $300. (Google, August 2006)

b. [G]enerally I was out shoveling long before my ‘young lady’ had her

nightcap off. (1869, CLMETEV)

c. What happens if I’m late paying my VAT? (CB)

d. Karen came through the door, lugging the bulky file. “Schultz is done

burning copies,” she said as she strode to Winters’s desk and plopped

the bundle down in front of him. (CB)

e. They must be finished painting by now. (CB)

Some of the adjectival constructions under discussion respect all the defining

characteristics of IPC-constructions, but as is to be expected, this is by no means

always the case; indeed, the superficially similar structures exemplified in (30)-(31)

reveal various differences when considered more closely. The most straightforward

members of the category of IPCs are the constructions with emotive adjectives:

omission of the participle clause has a clear semantic effect on the matrix clause

predicate, wh-extraction is invariably allowed, insertion of an adverbial subordinator

is impossible, and so is questioning by means of an adverbial interrogative (while

questioning by what works with relative ease, provided a preposition is added to the

matrix clause predicate). Moreover, unlike most IPCs, these emotive adjectives re-

sist the insertion of adverbial material between the adjective and the IPCs – when

such material is inserted, as in (32), the participle clause automatically receives an

adverbial interpretation (compare this to the examples in (24) above). In light of this,

we can safely treat the IPCs with emotive adjectives as genuine adjectival comple-

ments.

(32) a. They were never tired telling me. (1913, CEN) (‘they were never tired

of telling me’)

b. They were tired yesterday telling me. (‘they were tired yesterday when

/ as a result of telling me’)


A number of the non-emotive adjectives, viz. busy, late, done and finished, also take

IPCs approaching genuine complements, even if the semantic effect of omitting the

participle clause is sometimes less dramatic (e.g. if a person is busy making their

bed, it is fair to say they are busy, while it is not the case that a person who is happy

sleeping on the coach is simply happy). Note further that the expression be done

shows a remarkable alternation between the copula be and the perfect auxiliary have

– a pattern obviously not shared by any of the other adjectives.130

More problematic are the adjectives of the right-type. With these adjectives

the semantic effect of omitting the participle clause is again less outspoken. Further,

the adjectives right, lucky and brave differ from the above adjectives in that they

occur in an alternative though semantically reasonably similar structural configura-

tion in which the situation otherwise expressed by the participle clause is now ex-

pressed in a that-clause or to-infinitive and takes the syntactic function of extraposed

subject – compare the examples in (33) (see also (31a) above, where the two uses

occur together in a single sentence).

(33) a. Listen, we were really lucky getting home at all. (BNC)

b. Listen, it was really lucky we got home at all.

As to the adjectival expression be better off, this pattern resembles right, brave and

lucky in being commentative in meaning, but better off does not occur in the con-

struction illustrated in (33b). What is more, it is ill-behaved in allowing how-

questions instead of what-questions and licensing the insertion of the subordinating

conjunctions when or if, suggesting that although they license wh-extraction, the

participle clauses following better off are really adverbial, as is further supported by

the possibility of replacing the ing-clause by a prepositional phrase or adverb with-

out a change in meaning (cp. I’d be better off living under the bridges of Paris and

I’d be better off there).131

Finally, the same possibility of replacing the ing-clause by a functionally

similar adverb seems to be supported by the expressions be gone, be out, and be off

(cp. He is out working and he is out in the garden), but unlike better off they resist

questioning by an adverbial wh-pronoun. In this light, these expressions might be

130 Note that if finished shows the same alternation, this remains invisible since the pattern with have is indistinguishable from the verb complement construction with gerund clause. 131 The same possibility of inserting when or if applies to the adjectives occupied, engaged and employed when preceded by the adverb better.


more suitably analysed as being akin to catenative constructions with lie, sit or stand

and a participle clause, as in (34). As the latter are characterised by the reduced se-

mantic prominence of the main verb (cf. Quirk et al. 1985: 506), a shared analysis

for the catenatives and be gone, be out and be off ties in with the intuition that with

all of these constructions omission of the ing-clause is not impossible but strongly

increases the prominence of the matrix clause predicate.132

(34) He stood breathing gusts of vapor into the snowflakes that flitted about

his face and clogged his eyelids. (CB)

The second major group of interrelated constructions involve an IPC attached

to a light verb (typically have but sometimes also find or experience) and its seman-

tically ‘heavy’ noun phrase. The IPC can be interpreted as a complement to the

whole light verb idiom or as a postmodifier or complement to only the noun phrase.

This construction type is illustrated in (35):

(35) a. I give advice to people who are having difficulties getting a job, […].

(CB)

b. […] to believe in the possibility of events that I have a hard time be-

lieving will come to pass. (My name is red, p.233)

c. The biggest imponderable is the extent of the damage likely to result

from the energy crisis, which is sure to bring something that econo-

mists have no experience charting: a slowdown caused not by lack of

demand but by shortage of supply. (1974, TIME)

Again, different semantic sub-classes can be recognised. The most prominent group

is constituted by light verb constructions with the noun phrases (no) difficulty(/ies),

problem(s) and trouble. Some semantically related but less frequent expressions also

take IPCs, such as have (no) success, have a tough job or have a hard time and IPCs

also appear with some formally similar expressions such as have no hesitation and

have no experience. Of all these, especially the constructions with difficulty(/ies),

problem(s), trouble and success conform to all characteristics of IPCs and can be

fairly naturally interpreted as clausal complement constructions, as they establish a

relation between the matrix clause subject and a participle clause, specifying to what

132 For be gone this statement may have to be qualified: John is gone fishing implies that John went out purposely in order to go fishing – an implication of intentionality that is missing with John is gone.


degree the subject is successful in realising the situation denoted by the participle

clause. It is to be noted, however, that the possibility of using the noun phrase along

with the IPC without a light verb, as in (36), supports an alternative analysis of the

IPC as a postmodifier or complement to the noun phrase.

(36) a. I know the point you made about the difficulties finding evidence but

you know the trail doesn’t go cold just because forty fifty years have

passed. (CB)

b. Bailey White is on summer vacation from her job teaching first grade

in south Georgia. (CB)

Other light verb constructions are built on the noun phrases business and

right, as exemplified in (37). Characteristic about these highly idiomatic construc-

tions is that they always contain a negative element, typically no.

(37) a. The state has no right telling the people what they can and can’t do

with their own body. (CB)

b. But they can only search the parts of the house that a person could be

hiding in. They have no business looking in a one foot square box for a

6ft. 20stone man. (CB)

Besides the obligatory negative element, these constructions differ somewhat from

those of the difficulty-type in that they strongly resist questioning by any question

word. Also, it seems impossible to use the sequence of noun phrase and IPC without

the light verb, so the IPC is unambiguously a complement to the light verb idiom

rather than a complement or postmodifier to the noun phrase. The meaning of the

construction certainly fits in comfortably with a complement reading, as a relation of

(external) permission is predicated of a subject in relation to the activity denoted by

the IPC. A final characteristic is that the idiomatic meaning of the construction is

lost without the IPC, and that under omission of the IPC the expressions feel incom-

plete (unless its content is contextually given).

These two major sets of constructions – IPCs with predicative adjectives and

IPCs with light verb idioms – do not cover all the constructions showing IPCs. Thus,

(38) lists and illustrates a number of less clearly related constructions that appear

nonetheless also to involve a complement-like participle clause, judging by the tests

defining IPCs. These include mostly IPCs with other light verbs or verbal idioms –

take one’s time (38a), take turns (38b), have one’s hands full (38c), need help (38d)


– but also with some verbs – hesitate (38e), assist (38f), succeed (38g), bother (38h)

and tire (38i).

(38) a. The conman turns up with a woman said to be his mother, drops the

names of fancy hotels, flashes his mobile phone and asks to test drive

expensive motor cars. The problem is, according to the local newsletter

Motor Industry News, he takes his time bringing them back. (CB)

b. We took turns opening the refrigerator door and hoping, but no matter

how many times we looked inside, the contents never changed. (CB)

c. Ms Brydges has her hands full putting on the finishing touches, […].

(CB)

d. He needed help rethinking his gendered, monolithic definition of fam-

ily provider. (CB)

e. I wouldn’t hesitate moving to Norway where he comes from. We’re

thinking of it and it’s an exciting thought. (CB)

f. The only times to avoid travelling in the Metro are peak hours when

you will see, and it is fascinating to watch, people performing the most

extraordinary manoeuvres to get in (and out) of an already-full car-

riage; indeed, the guard will sometimes give the final coup de grâce

with his boot to assist closing the doors. (CB)

g. A little tailor is sent on a quest with a glass key and succeeds rescuing

a sleeping beauty, slaying the sorcerer and living happily ever after in

the palace with his beautiful bride. (CB)

h. you know why bother arguing with her (CB)

i. Back Home women never tire asking that question. (1964, Visser

1963-73: 1868)

Further, there are a number of construction types whose status as IPC-

constructions is doubtful even among its most central members. The ing-clauses in

these constructions can be said to approach the adverbial end of the cline from IPC

to adjunct. By far the most important group of constructions of this kind is clustered

around the prototype illustrated in (39a), which has the verb spend in combination

with a noun designating a period of time, and an ing-clause designating the activity

that the matrix clause subject is taken up in during that period of time. The status of

the participle clauses in these constructions is dubious because the construction al-

lows questioning by how (39b), but also wh-extraction (39c) and, arguably, ques-

tioning by what (39d). Insertion of an adverbial subordinator is not allowed (39e),


but this may be due to the fact that English lacks a subordinator of manner. The con-

struction is also remarkable, however, because omission of the participle clause of-

ten makes the sentence sound oddly incomplete (39f).

(39) a. He in turn would spend his time boozing and nightclubbing with mates

[…]. (CB)

b. How did he spend his time?

c. The mates he spent his time boozing and nightclubbing with.

d. (?) What did he spend his time in?

e. *He in turn would spend his time while/when boozing and nightclub-

bing with mates.

f. *He in turn would spend his time.

Variants of the construction arise when the TIME-NP is replaced by a noun

phrase denoting material goods, or when the verb spend alternates with the verbs

employ (marginally), lose, pass and waste. Notice that with all of these variants

omission of the ing-clause is more acceptable. This indicates that the reason for the

obligatoriness of the ing-clause with spend TIME is pragmatic rather than syntactic:

whereas all things existing spend time by definition so that stating so is hardly in-

formative, wasting time and spending goods are optional activities, alternative to

using time well and keeping goods, which suffices to make them of interest by

themselves. Support for this view comes from the fact that spend TIME too can be

used without an additional participle clause if the TIME-NP receives extra modifica-

tion, as with the noun phrase evening in (40).

(40) After the long conversation between herself and Lord Lackington

which followed on the momentous confession of her identity, Julie

spent a restless and weary evening, which passed into a restless and

weary night. (1903, CEN)

Noteworthy is also that if the participle clause in a spend-construction is absent and

the TIME-NP receives no extra modification, the participle clause’s place will typi-

cally be taken by a PP or adverbial, as in (41). It is this paradigmatic relationship to

clearly adverbial elements that suggests that the participle clauses in this construc-

tions are (semi-)obligatory adverbials rather than IPCs proper.


(41) She arrived in Jamaica in April, intending to spend six months there.

(CB)

That it is a thin line separating adverbials from IPCs, however, appears from

the fact that some variants of the spend TIME construction do seem to take pure

IPCs. Compare in this respect examples (42a-b): example (42a) can still be inter-

preted as the negated answer to the question ‘how did x waste time?’; in (42b), by

contrast, the participle clause does not denote a manner of wasting time but an activ-

ity carried out by the subject without wasting time – that is, waste no time here func-

tions as a verbal idiom specifying a relation of immediate (unhesitating) and inten-

tional realisation between its subject and an action of which the subject is the agent.

Accordingly, the participle clause in (42b) cannot be the focus of a how-question,

and omission of the participle clause alters the meaning of the matrix predicate from

‘not hesitate’ to ‘not idle’.

(42) a. Manchester United wasted no time mourning the loss of their Premier-

ship crown. (CB)

b. Handball by a keeper outside his area in this competition warrants an

instant dismissal and the referee wasted no time waving a red card.

(CB)

Finally, apart from the group of constructions clustered around the expression

spend TIME, there are some other constructions whose participle clause allows wh-

extraction but otherwise shows no signs of being a genuine IPC or whose IPC-status

is at best questionable.133 Such constructions are therefore best classified as contain-

ing a participial adjunct. Characteristic of the main clauses in question is that they

contain only one referential participant (the subject) and therefore carry a relatively

low information load, which may be what makes the extraction possible. Examples

are go to bed / wake up (43a), die / get killed (43b), get ADJ (43c), VERB of motion

(43d), VERB of rest / motion + around (43e) and so on.

133 In general, wh-extraction is not the most reliable of tests to distinguish comple-mentation from adverbial modification. Consider the example in (ii): (ii) What Demoyte [the headmaster at a school] cared about was profi-

ciency in work. This his masters were engaged to produce and sacked for failing to produce (The sandcastle, p.20)

Enjo

y

Hav

e n

o

bu

sine

ss

Be

tire

d

Be

fin

ishe

d

Have

tro

u-

ble

Be

bu

sy

Be

luck

y

Be

out

Spe

nd

TIM

E

Wast

e

GO

OD

S

Be

be

tte

r off

Make

m

one

y

+ ? - - - - - - ? - - - i. Omission of ing-clause renders matrix clause ungrammatical

+ + + ? - - - - - - - - ii. No insertion of adverbial material between predicate and ing-clause

n/a ? + ? + ? ? ? - - - - iii. Omission of ing-clause changes the semantics of the matrix clause predicate

+ + + + + + + + - - ? - iv. Ing-clause cannot be questioned by adverbial interrogative pronouns

+ + + + + + + + + ? - - v. Ing-clause resists insertion of subordinating conjunctions

+ + + + + + + + + + + ? vi. Ing-clause allows wh-extraction

Table 10.2. IPCs as a gradient category.


(43) a. You can go to bed thinking about something and wake up thinking

about it next morning. (CB)

b. They’re pulling statues down in Moscow these days. And to replace

them there are men and women many Russians feel should be honored,

such as Andrei Sakharov or the young men who died resisting last Au-

gust’s coup […]. (CB)

c. You may not get rich backing them, but put your cash elsewhere and

you’re throwing it away. (CB)

d. The woman enters holding a box of stockings. (CB)

e. At night workers just sat around playing cards or sleeping. (CB)

In view of the preceding discussion, IPCs may be thought of as a gradient

category, with some constructions exhibiting all defining features, others exhibiting

some, and yet others none at all (cf. Quirk 2004 [1965]). How such a gradient might

have to be conceived of is illustrated in Table 10.2, where a number of constructions

are judged with respect to the defining characteristics of IPCs (versus adjuncts) and

a few further characteristics that have surfaced in the preceding discussion.134 Real

IPC-constructions are found at the left hand side of the table, whereas constructions

whose ing-clause shows more adjunct-like behaviour are found at the right hand

side.

Start by noting that there is a dangerous elegance to this kind of representa-

tion and the implicational hierarchies it might suggest. The gradient ordering in Ta-

ble 10.2 is in some respects imperfect and furthermore glosses over certain syntactic

differences between constructions whose occurrence it cannot predict. Thus there is

the semi-obligatory nature of participle clauses with spend TIME that cuts across the

IPC-gradient at an apparently arbitrary point. Then there is the occasional possibility

of replacing an ing-clause with a functionally equivalent prepositional phrase or

adverb such as there, again found with spend TIME, but also with waste GOODS

and be better off and, marginally, with have no business (though not have no right!).

Similarly, looking at the possibility of questioning the participle clause by what, this

operation seems fairly acceptable for spend TIME and waste GOODS, yet not for be

gone, be finished and have no business, which are nevertheless higher up the IPC-

134 Notice that in Table 2 characteristics a. (omission of ing-clause renders matrix clause ungrammatical) and c. (omission of ing-clause changes semantics of matrix clause predicate) are logically linked. If under omission of the ing-clause the mean-ing of a predicate changes it still has to be grammatical, while if a predicate becomes ungrammatical, presumably we can no longer say anything about its meaning.


cline in other respects – note though that because what-questioning requires the ad-

dition of a preposition to the matrix clause predicate, the outcome of the test de-

pends on the availability of a (suppletive) prepositional alternative to the non-

prepositional ing-clause-construction and may reveal little about the ing-clause as

such. Finally, the gradient in Table 10.2 ignores the possibility that expressions like

be out (including be off and be gone) approach yet another construction type – that

of aspectual auxiliary constructions with verbs such as lie or sit.

Still, one cannot but wonder what the gradient in Table 10.2, in spite of its

imperfections, tells us about the relation between complements and adverbial

clauses. Following Croft (2001: ch.9), language users have different ways of con-

ceiving of (or ‘construing’) the relation between two situations. This is grammati-

cally reflected in the difference between adverbials and complements in relation to

their matrix clauses: in a complement construction one situation is perceived as a

core participant in another situation (e-site elaboration), while in an adverbial con-

struction one situation functions as the background against which the other situation

unfolds (figure-ground configuration). A possible interpretation of the gradient in

Table 10.2, then, is to see it as the reflex of a conceptual cline running from typical

figure-ground configurations to e-site elaboration. Arguably, it is a matter of degree

whether a situation is seen as providing the background against which another situa-

tion is profiled, or as one element in an interaction with another participant.

If it is true that the conceptual boundaries between complementation and ad-

verbial modification are non-discrete, we still have to account (in a non-circular

way) for the specific positions that predicates take along the cline from one relation

type to the other. The factor most likely to matter is of course the semantics of the

matrix predicate involved. That is, somehow the semantics of different matrix predi-

cates cause those predicates to assume different positions on the gradient from ad-

verbial to complement when combining with a subordinate ing-clause. On a tenta-

tive note, I would suggest that what ultimately determines the position of a predicate

is the strength or impact of the interaction it expresses between two participants –

that is, the degree to which the state of the subject can be conceived to be dependent

on or responsible for the state of the situation expressed in the participle clause. The

more clearly the process denoted by a predicate follows from a specified state in one

participant and results in a specified state in the other participant, the higher the

probability that the predicate will indeed be perceived as designating an interaction

between the two participants, and the stronger and more intrinsic the cognitive link

that will activate (the expectation of) the second participant essential to a relation-


ship of e-site elaboration.135 For example, be busy and have trouble are ranked

higher than spend TIME because they imply greater activity and intentionality on the

part of the subject and thereby create a stronger link to a second participant in the

form of a particular goal. In turn be tired and other emotive predicates are ranked

slightly higher than be busy and have trouble because while the outcome of the sub-

ject’s goal-oriented activity is indeterminate in the case of be busy and have trouble,

there is an outspoken effect of one participant (the situation in the ing-clause) on the

other (the subject) in the case of be tired: the ing-clause designates a source that

triggers a certain emotion in the subject.136 By hypothesis, a scale of this kind from

weaker to stronger interaction may provide the required semantic underpinning of

the gradient attested in the data.137

In any case, however, the gradient in Table 10.2 licenses the conclusion that

the non-nominalised ing-clauses in certain environments show complement-like

behaviour and differ from adverbial clauses. Moreover, in addition to revealing cer-

tain classificatory difficulties pertaining to IPCs with respect to the distinction be-

tween adverbials and complements, the preceding discussion revives a familiar

theme with respect to the use of complement types. It is clear from the overview of

IPC-constructions in this section that the use of IPCs is organised around fairly spe-

135 The hierarchy suggested resembles Givón’s (1980) binding scale or Hopper and Thompson’s (1980) scale of transitivity, but only in part since it does not immedi-ately matter whether influence runs from the first participant to the second or vice versa. Primarily, what matters is whether and to what degree some influence or in-teraction is likely to be perceived by the language user. 136 As Langacker (1991: 326-7) points out, an experiencer is to some extant an active participant in event structure since experiencing requires the establishment of mental contact with the thing experienced as well as the activation of some cognitive repre-sentation (cf. Hollmann 2003: 59). 137 It is not to be assumed that the prototypical organisation of categories sufficiently excuses the disparate syntactic behaviour of all IPCs. Part of the behavioural varia-tion between different IPC-like constructions is explained by other factors. To name only those that surfaced in the preceding discussion, there is the role of pragmatics (e.g. spend TIME, unlike waste TIME, is uninformative without further modifica-tion), and there is the interference of other construction types than complement and adverbial constructions (in particular, the role of catenative patterns as in the case of the participle clauses following verbs of rest). Further, the results of syntactic tests may be obscured by accidental (or at least not directly relevant) characteristics of the language (e.g. not all adverbial relations have a corresponding adverbial interroga-tive) and by strong idiomaticity in the constructions tested (as might, for instance, be the case for the IPCs following have no business). Thus, while a semantic cline from adverbial to complement relations may be responsible for part of the attested syntac-tic variation, the situation is certainly complicated by a number of unrelated factors.


cific constructions. Even among the patterns whose complement-status is relatively

unproblematic, there is no obvious way in which the distribution of IPCs is princi-

pled. Among adjectival complements, for example, we find IPCs complementing

emotive adjectives as well as semantically unrelated adjectives such as late or busy.

In the domain of verb complementation, the distribution of IPCs is even less consis-

tent: synchronically, there is no principle that could motivate the use of a given

complement type with verbs as disparate as assist, bother, tire, hesitate and succeed.

Indeed, to understand this distribution, we have to turn to the history of IPCs.

2. Historical developments

IPCs are a relatively recent phenomenon in English and if they form a coherent

category at all, it is an emergent one. This section examines more closely when and

– to the extent possible – how IPCs came into being. As will become clear, IPCs did

not appear simultaneously, but made their appearance with one predicate type after

another over a period of about four centuries (Section 2.1), and the manner in which

IPCs arose similarly varies from predicate to predicate (Section 2.2). What we see,

in other words, is another diffusional change, with a shift in the mechanisms driving

diffusion (cf. Chapter 9). In this respect the development of IPCs resembles the his-

tory of other complement types, although in other respects it shows interesting dif-

ferences (Section 2.3).

2.1. Diffusion

Two of the constructions discussed or mentioned in the preceding section go back a

long time, but both are dubious as to whether they are in fact IPCs from the start.

Specifically, done is attested with ing-clauses as early as the sixteenth century, but

only in the pattern where done is preceded by the auxiliary have, and it is not incon-

ceivable that at this point the construction is in fact gerundial. A very early but

somewhat hard-to-interpret example is given in (44a), a clearer instance of slightly

later date is given in (44b). Evidence for the possible gerundial origin of the con-

struction is given in (44c-d), revealing nominal characteristics in the -ing-forms fol-

lowing have done. Ing-clauses following the expression be gone turn up around the

same time, but only in the fixed phrases be gone a-hunting / a-fishing / a-birding, as

in (45a-b), with a characteristic a-prefix that renders the syntactic status of these


forms dubious as well138 – moreover, as pointed out above, constructions of this kind

may have come to lean closer to catenative uses of verbs of motion and rest than to

IPCs proper, at least as far as Present-Day English is concerned.

(44) a. and when they had don plahyng, and then begane the sagbottes pla-

hyng, (1553-59, PPCEME)

b. ‘twill be this hour ere I haue done weeping (1591, Visser 1963-73:

2209)

c. soþ it is þat dymes ben due vn-to prestis in þe olde lawe, but þey weren

holdun to do aȝen sleyng of beestis & hard seruyss. (a1500 (?c1378),

IMEPC)

‘it is true that tenths were due to priests according to the old law, but

they were obliged to do their own slaughtering of animals and hard

work.’

d. than risith on of the wisist lordis and reportith to the peple gret re-

comendacioun and preysyng of the kyng, and of þe good governaunce,

and done gret thankyng vnto god þat hath sent so excellent a witt vnto

the kyng of Iewes to gouerne hem in suche wise (a1500, IMEPC)

(45) a. her husband is this morning gone a Birding: (1599, PPCEME)

b. sure I thinke she be gone a fishing for her. (1630, PPCEME)

The earliest straightforward IPCs attested date from the seventeenth century,

and occur in combination with the adjective busy (46a-b). The use of obligatory par-

ticiple clauses with spend TIME (the IPC-status of which is again questionable, see

above) shows up around the same time (46c), although it only becomes highly cur-

rent in the course of the nineteenth century. Another late seventeenth-century inno-

vation is the use of IPCs with long, meaning ‘late, slow’ (46d). Other types lag be-

hind considerably, however: There are the IPCs combining with light verb idioms

have no business (46e), have difficulty or have trouble (46f), as well as with take

time (46g), or IPCs with other adjectives than busy – including done (preceded by

be) and tired, and possibly also happy and late (46h-i) – all of which first show up in

the second half of the nineteenth century. The twentieth century sees further innova-

tions, among which the use of IPCs with the adjectives bored, comfortable, finished,

fed up, slow and quick and the light verb idioms have success and have no right

138 It is very likely that the pattern derives from the prepositional construction go on NP (see the discussion of go on in Chapter 9).


(46j). Another clearly twentieth-century innovation is the occasional use of IPCs as

complements or postmodifiers to nouns such as difficulty, job, problem, (one’s) time

or work (46k-l).

(46) a. [A]nd so home to supper -- my people busy making mince-pies (1666,

P participle clauseEME3)

b. Charles Smith and William Moon were both tried for stealing a Silver

Tankard from one John Morris, value 6 l. they came to drink at Mor-

ris’s House, and whilest the Man of the House was busy waiting on the

other Guests, the Tankard was gone, and the Men too, without paying

the Reckoning: […] (1693, Proceedings of the Old Bailey)

c. and she spent the whole day making herself clean. (1668, Diary of

Samuel Pepys)

d. Jo. I have almost broke my Brains with studying & contriving, but now

I think I have hit on’t. Ka. Tis long coming out. (1684-7, P participle

clauseEME3)

e. he said he told me the truth and nothing but the truth, and all that he

knew – I said he had no business telling lies and making statements he

could not prove (1862, POB)

f. Dear Sir, – For a long time past I have had considerable difficulty de-

ciding the important question, ‘Who is the master of my own house?

Myself, or YOUR SON Lupin?’ (1894, CLMETEV)

g. We were no sooner done eating than Cluny brought out an old,

thumbed, greasy pack of cards, such as you may find in a mean inn,

and his eyes brightened in his face as he proposed that we should fall

to playing. (1886, CEN)

h. “Lina” took a long time answering the question, but eventually spelt

out “ROSES, LILIES, AND COWS.” (1894, CLMETEV)

i. I am quite happy standing here alone in a crowd, knowing nobody!

(1894, CLMETEV)

j. “I was slow figuring it out,” Leaphorn said. “I smelled something

about Jackson. But I figured him to act like a Navajo and he was acting

like a white man.” (CB)

k. Work on the horses. Work around the yard. Work trying to get preg-

nant. And, nearly a full-time vocation in itself, work raising Marley.

(Marley and me, p.81)


l. Terry […] said I would only ever get a job collecting supermarket trol-

lies or cleaning donkey shit at an animal sanctuary (The curious inci-

dent with the dog in the night time, p.33)

The non-simultaneous appearance of IPCs with different types of heads pre-

sents us with another diffusional change, in which a construction becomes increas-

ingly frequent over time and is gradually matched to an increasing number of host-

ing constructions. The diffusional character of the development of IPCs is also ap-

parent in other respects. As IPCs begin to occur in more and more environments,

they also run into other (and older) constructions that are equivalent or roughly simi-

lar in meaning and with which they have to compete. The diffusion of IPCs is espe-

cially marked by competition with gerund clauses introduced by the preposition in,

which in many cases turn out to present a very adequate semantic and syntactic al-

ternative to the IPC. Compare in this light the examples in (46) above with the fol-

lowing examples in (47):

(47) a. It was proved that the three Prisoners coming into the house of Tem-

ple, and calling for Wine, whilst Wilson and Pain were busie in drink-

ing, Ellenor Davis makes use of the opportunity, taking the silver Salt-

seller, marchs off unknown to her Companions, whereupon the said

Temple missing his Salt seller, apprehends the said Wilson and Pain,

as Accessary in the Theft; (1686, Proceedings of the Old Bailey)

b. A friend of his had spent much time in composing a book, and went to

Sir Thomas to have his opinion of it; (1753, CLMETEV)

c. I had more and more difficulty in keeping the fat landlady at arm’s

length, and the nasty child was well beaten one day for lingering about

my door. (1888, CEN)

d. [O]ur cousin of France is happy in having a cavalier who is so fit to

uphold his cause either with tongue or with sword. (1891, CEN)

e. Several agencies have complained that Turkey has been slow in ap-

proving projects. (CB)

In order to give some idea of the competition between the two clause types,

Table 10.3 documents the percentage of IPCs to gerund clauses introduced by in

with the most common predicates taking either clause type. Figures are given for

two historical periods, the transition from the nineteenth to the twentieth century


(the period 1850-1922) and the end of the twentieth century (the period 1990-1995),

on the basis of material from CLMETEV and CEN, and from CB respectively.139

MATRIX PREDICATE 1850-1922 1990-1995

Be busy 93.9% 100.0%

Spend TIME 43.0% 98.9%

Be happy 25.0% 98.9%

Be late 24.2% 46.9%

Take turns 22.3% 85.7%

Have trouble 17.6% 94.3%

Have difficulty 1.5% 72.1%

Be engaged 1.3% 2.8%

Be slow 0.0% 18.2%

Be right 0.0% 3.2%

Be successful 0.0% 0.0%

Be justified 0.0% 0.0%

Table 10.3. Percentage of IPCs to gerund clauses with in.

139 The manner of obtaining figures differs slightly for the two periods represented in Table 10.3, due to the different concordancing programs used to access different corpora: CEN and CLMETEV were accessed using Wordsmith Tools 3.0; CB could only be accessed using the interface that comes with the corpus. Differences apply to sampling methods and to the search strings used. Thus, the percentages in Table 10.3 are based on samples when corpus searches yielded too many instances. Sam-ples were taken for be busy and spend TIME in the period 1850-1920, which were sampled at 1/3 hits and 1/12 hits respectively, and for spend TIME, be happy, be late, be slow, have difficulty and have trouble in the period 1990-1995, which were sampled at 200 hits for be happy and be slow and 150 hits for the other predicates (with the differences in sample sizes compensating for the amount of junk hits with some predicates). The search strings that were used were based on the main lexical word in the predicate (e.g. busy, happy, and so on) followed by a gap of zero to one words, followed by a word ending in -ing for the period 1850-1920, or a form tagged VBG (i.e. verbal form in -ing) for the period 1990-1995 – for the expression spend TIME, the gap between spend/spends/… and the form in -ing was zero to five words, so as to leave room for the TIME-NP. Note finally, that it is not always easy to distinguish between IPCs and participial adjuncts (see further Section 2.2) or be-tween the gerunds with in that function as complement and those that function as adverbial clause. For this reason, counts are based on instances that can be inter-preted as IPCs or complement uses of the gerund with in.


Of course not all IPC constructions compete or compete exclusively with gerund

clauses introduced by in (tired typically takes gerund clauses introduced by of;

happy also takes gerund clauses introduced by about, spend TIME has also come to

combine with gerunds introduced by on; and so forth). Moreover, the figures in Ta-

ble 10.3 should be interpreted cautiously, as what is counted is potential rather than

certain IPCs (i.e. constructions allowing an IPC-reading). This granted, Table 10.3

still provides a strong indication that at least with a number of predicates, the spread

of IPCs progressed at the cost of the older gerund clauses with in. Most relevantly

however, the table shows that the progression of IPCs did not proceed simultane-

ously and at the same pace with all predicates. Clearly confirmed by the data in Ta-

ble 10.3 is the finding that busy and – to a lesser degree – spend TIME started occur-

ring much earlier with IPCs than other predicates. Another striking observation is

that IPCs began to combine with some predicates such as be engaged already in the

nineteenth century, but have failed to catch on in the course of the twentieth. For the

twentieth century, too, it is intriguing to find that while some predicates have come

to combine with IPCs exclusively, others are still highly resistant to use with IPCs,

despite the fact that their use does not seem ungrammatical. For instance, construc-

tions like those in (48), though apparently well-formed, remain unattested in CB:

(48) a. Values are important in career exploration because people who believe

in the goals of their employers and whose values are in synch with

their coworkers’ values are more likely to be successful getting and

keeping a job than those whose values conflict with others in the work-

place. (Google, August 2006)

b. One is justified taking the elevator one floor when no stairs are avail-

able. (Google, August 2006)

c. This book was very helpful getting me up and running and doing what

I wanted to do – make my website. (Google, August 2006)

This type of variation is of course typical of diffusional change. IPCs have appeared

in a number of different lexico-grammatical environments at different times, replac-

ing competing constructions with varying degrees of success.


2.2. Mechanisms of change

The questions that these findings raise are straightforward but remain hard to an-

swer: First, what mechanisms give rise to the emergence of IPCs? And second, what

determines the course and pace of diffusion? Put differently, why are some predi-

cates affected by a change earlier and to a greater degree than others? While it is

probably impossible to resolve these issues in all detail, some relevant insights can

be gained by examining the mechanisms that could give rise to IPCs in particular

environments, which is the main purpose of this section. As will be shown, the early

development of IPCs is primarily steered by syntactic reinterpretation (Section

2.2.1), while later stages are marked by local analogies coming into operation (Sec-

tion 2.2.2).

2.2.1. Reinterpretation

In all likelihood, the major mechanism in the emergence of the first IPC-

constructions has been the reinterpretation of adverbial clauses. For the adjective

busy, potentially ambiguous sequences are illustrated in (49).140 If the presence or

absence of a comma in writing is ignored, many instances of busy followed by an

ing-clause allow both an adverbial disjunct interpretation and an IPC-reading. On

either reading the ing-clause eventually describes what the subject of the matrix

clause is busy doing, but on an IPC-reading, the participle clause restricts the seman-

tic scope of the predicate by narrowing down the matrix clause subject’s activity (as

evoked by busy) to the activity denoted by the participle clause, while on a disjunct

reading, the participle clause elaborates on the main clause and gives the additional

information that justifies the speaker in calling the subject busy – information that

typically consists in a description of the subject’s current activity.

(49) a. Up, and to the office betimes; and there all the morning very busy,

causing papers to be entered and sorted, to put the office in order

against the Parliament. (1666, PPCEME)

140 The construction with busy and IPC or disjunctive participle clause is particularly frequent in the diary of Samuel Pepys (49a-b), although as the examples show, it occurs elsewhere as well (49c-e). The same is true for the construction with spend TIME, which is also strikingly frequent in Samuel Pepys’ diary; here too, however, the ambiguous instances are found both in the writings of Samuel Pepys (50a) and elsewhere (50b-c).


b. Thence took coach and I all alone to Hyde Park […], and so all the

evening in the Park, being a little unwilling to be seen there, and at

night home, and thereto W. Pen’s and sat and talked there with his wife

and children a good while, he being busy in his closet, I believe pre-

paring his defence in Parliament, and so home to bed. (1668, PEPYS)

c. When nature was most busie, the first weeke / Swadling the new-borne

earth God seemd to like, / That she should sport her selfe sometimes,

and play, / To mingle, and vary colours euery day. (a1631, John

Donne, An anatomy of the world)

d. Mr. John Collins deposed, […] That he saw the Prisoner, who had a

Black Patch upon his Nose, in the House Five or Six Minutes, very

busy, breaking the Sashes and Frames of the Windows (1716, POB)

e. and by Course a new Consolidator being to be built, they were as busie

as ever. Bidding, Offering, Procuring, Buying, Selling, and Jobbing of

Feathers to who bid most; and notwithstanding several late wholesome

and strict Laws against all manner of Collusion, Bribery and clandes-

tine Methods, in the Countries procuring these Feathers; never was the

Moon in such an uproar about picking and culling the Feathers, such

Bribery, such Drunkenness, such Caballing […], as the like has never

been known. (1705, CEMET)

For spend TIME similar ambiguities are attested, originating from original

disjunct uses, as is illustrated in (50). In each of the examples in (50), the ing-clause

may be read as a participial disjunct, elaborating on the matrix clause (whose prag-

matic/syntactic requirement for some sort of adverbial is already met by another

element), but may also be read as being itself part of the (semi-)obligatory adverbial

modification of the spend TIME construction. For instance, talking again about

Creed’s folly in (50a) may be a disjunct elaborating the matrix clause, but may also

be interpreted as an obligatory modifier of spend TIME, syntactically on a par with

the prepositional phrase with her. Instances of this type could have given occasion to

reinterpretation as early as the seventeenth century – though of course, given that the

participle clauses found with spend TIME are perhaps not strictly speaking IPCs, we

may have to think of the proposed reinterpretation as running from disjunct to

obligatory adjunct, or we may even do without reinterpretation altogether and limit

ourselves to saying that adjuncts began to appear in this environment around the

seventeenth century (but see below as to why this last possibility is somewhat less

plausible).


(50) a. So to dinner to my Lady Sandwich’s, and there after dinner above in

the diningroom did spend an houre or two with her talking again about

Creed’s folly; (1665, Diary of Samuel Pepys)

b. Husbandry is another thing that doth occasion men to break; Some will

spend their time in Drinking and Gaming, neglecting their business,

until they are undone. (1681, LC)

c. but most of the little time I had with them was spent in a silent retired-

ness of spirit, waiting upon the Lord. (1683, CEMET)

For the adjectival predicates tired and happy a similar situation obtains,

though here the ambiguity seems to lie between IPCs and adjuncts. Thus, participle

clauses following tired (especially get tired), as in (51a-b), may be adjuncts specify-

ing how or why the subject is getting tired, or IPCs specifying what the subject is

getting tired of. Similarly, participle clauses following happy, as in (51c-d), may

function as adjuncts describing the circumstances under which the subject is happy,

or as IPCs denoting the source or object of the subject’s happiness. Note that

whereas tired is found in unambiguous IPC-constructions already in the nineteenth

century (see examples (1b) and (32a) above), the evidence for happy is more diffi-

cult to interpret as the sequences with participle clause found in the nineteenth-

century data are nearly all ambiguous between an adjunct and IPC reading, and con-

vincing instances of the IPC construction are only available for Present-Day English

(see (12b) and (30) above).

(51) a. But I think I’ll try for the mule-buyer. I’m getting tired looking at these

slab-sided cowmen. Now, just look at those mules – haven’t had a har-

ness on in a month. (1904, CEN)

b. I hope Hannah and John do not get tired doing my chores. (1903,

CEN)

c. Yes, I see; but oh, I was so happy being a garden flower with the sun-

shine on my head, and I can’t seem to care the least little bit for being a

banian-tree! (1893, CEN)

d. Dearest mother, should we not be very happy living together in Lon-

don? (1850, CLMETEV)

Very similar ambiguities are found in the Present-Day English data for adjectives

such as (un)comfortable, bored and fed up, and it is likely that these constructions


went through the same development. Thus, the evidence in (52) for (un)comfortable

is particularly suggestive: (52a-b) illustrate clear adjunct uses, (52c-d) represent the

majority of present-day instances in allowing both an adjunct and IPC reading, and

(52e-f) favour the IPC reading. The variable interpretations compare neatly to those

with the other emotive predicates: adjuncts specify the conditions or circumstances

accompanying the situation of feeling comfortable as depicted by the main clause

predication, while IPC constructions profile an attitudinal relation of willingness or

reluctance between the main clause subject and the action denoted by the participle

clause.

(52) a. [T]hough tall for a woman Jill was still short enough to be comfortable

sitting on the bed. (CB)

b. Many immigrants couldn’t use traditional banks because they required

a Social Security card, and they also felt more comfortable doing busi-

ness in their own language. (CB)

c. it was clear that Deborah didn’t feel comfortable exploring these issues

at this juncture. (CB)

d. I think a lot of men would feel very uncomfortable going to work

without a shirt and tie because in their particular field that’s what gives

them the confidence that they’re part of of of the business world (CB)

e. I’m not sure we’re looking yet at a real paradigm shift, but the accumu-

lating evidence does make it very difficult to evade the conclusion that

here, as in many other areas, there is a good deal more built in than

most developmental psychologists had supposed (or felt comfortable

assuming) a decade or two ago. (CB)

f. The psychiatrist impressed us as a sensitive and cautious man. After

four visits, during which he played with Ted or interviewed Sara and

me, he confessed that he was uncomfortable making a diagnosis. (CB)

For the adjective late, evidence is scantier, but it is possible that examples

like (53) could give rise to reintepretations from adjunct to IPC. On the adjunct read-

ing, the participle clause specifies an activity carried out at a late time – i.e. an activ-

ity concomitant to the subject’s being late – while on an IPC reading, the participle

clause specifies a goal that is associated with some (implicit) predefined time of

realisation, and that the subject fails to achieve at that time (for a clear IPC instance,

see (31c) above). Note, however, that the use of IPCs with late might also have been

analogically extended from the earlier use of IPCs with long (see below).


(53) a. Heavens, how sleepy I am! No wonder either! Late going to bed last

night and up so early this morning. (1913, CEN)

b. He was merely a young man who had been rather late visiting one of

the girls. (1893, CEN)

Finally, the IPCs with light verb idioms too appear to derive from adverbial

adjuncts. In late-nineteenth-century English, have difficulty and have trouble or

clauses with trouble or difficulty in general could combine with participial adjuncts,

as in (54a), but they typically show ambiguity between adjunct and IPC-readings. In

ambiguous examples, the ing-clause can either be taken to describe the activity con-

comitant to the subject’s having trouble (adjunct), or the activity the subject’s trou-

ble specifically pertains to (IPC), as is illustrated in (54b-d). In such cases, adverbial

participle clauses and IPCs are in practice indistinguishable.141 By way of contrast,

(54e-f) allow no adverbial interpretation.

(54) a. he wrote the two notes he asked leave to – he gave me no trouble going

to the station; he walked in front of me voluntarily and posted the two

notes. (1885, POB)

b. the defendant was struggling with difficulties trying to pay his credi-

tors. (1879, POB)

c. Indeed, the captain told me he met difficulty enough navigating the

shallow Main, and I think he prefers the deeper Rhine. (1910, CEN)

d. I will not tell the needless trouble I had breaking into that house – af-

terwards I found the front door was on the latch – nor how I ransacked

every room for food, until just on the verge of despair, in what seemed

to me to be a servant’s bedroom, I found a rat-gnawed crust and two

tins of pineapple. (1897, CLMETEV)

e. Carleton’s President Cowling, a bouncing Yaleman who in 28 years

has transformed a small Congregationalist school [...] into a prosper-

ous, top-ranking college, should have no trouble recruiting two fac-

ultymen of suitable calibre. (1934, TIME)

141 As (iii) conveniently illustrates, the adjunct reading of participle clauses is still available in Present-Day English: (iii) customers returning from Indonesia have experienced difficulties when

trying to cash American Express traveller’s cheques (CB)


f. she [...] proved herself a routine interpreter with a big pleasant voice

which she had trouble controlling. (1935, TIME)

As to IPCs with have no business, the pattern is announced by the phrase have no

business followed by a locative phrase, as in (55). The fact that the first ing-clauses

following have no business typically also imply a locative relation, as in (56), sug-

gests that at this point they are also still regarded as adjuncts.

(55) a. I placed myself in a cart-house, where I could not be seen, and saw the

prisoner go to my stable, bring out some loose hay, and put it into his

cart, which was already loaded – he had no business in that stable

(1851, POB)

b. you have no business here, and out you shall go (1852, POB)

(56) a. I did not say you were selling false tickets; I said, “You have no busi-

ness selling them tickets here” (1857, POB)

b. he told me to mind my own business, and I had no business looking

under his window (1879, POB)

As far as nineteenth and early-twentieth-century English is concerned, then, most

IPCs occur in environments where participial adjuncts occur and probably derive

from them. There are few exceptions to this. The most important exception is pre-

sented by be done, which acquires IPCs without any evidence of a preceding use of

participial adjuncts with the same predicate. There is an obvious alternative explana-

tion, however, since the already existent pattern have done with IPC (or gerund?)

was in all likelihood extended to be done through analogy. In the nineteenth century

have done occurs with and without IPC, as illustrated in (57a-b). On this basis, the

use of be done without IPC in the same meaning, as in (57c) (a use attested from the

end of the eighteenth century; Visser 1963-73: 2079), could be naturally extended to

the use with IPC as in (57d).

(57) a. Interrupt me again, and I have done. (1884, CLMETEV)

b. And when he has done eating, say I should like to have a few words

with him, if he doesn’t mind coming up here. (1873, CLMETEV)

c. One further favor and I am done. (1771, Visser 1963: 2079)

d. We were no sooner done eating than Cluny brought out an old,

thumbed, greasy pack of cards, such as you may find in a mean inn;


and his eyes brightened in his face as he proposed that we should fall

to playing. (1886, CEN)

Another plausible exception is the case of be long. In all likelihood, IPCs with

be long, which first appeared at the end of the seventeenth century (see (46d) above)

go back on a gerundial pattern with the same a-prefix as found in be gone a-fishing

(cf. (45) above), as is illustrated in (58a). There is no evidence to support any form

of reinterpretation from adverbial participle clauses, but adverbial uses of long, as in

(58b), may have contributed to the loss of the a-prefix. Note that, once occurring

with IPCs, be long may have provided an analogical model for IPCs with be late,

and is – along with spend TIME – particularly likely to have provided the model for

IPCs with the light verb take time. The latter fails to show evidence supporting re-

analysis, and the occurrence of the blend take long, as illustrated in (59), strongly

points in the direction of analogical extension.

(58) a. She shall be welcome Sir, I’le not be long A clapping you together.

(1630, PPCEME)

b. it soon became a stormy time. The clouds had been long gathering and

threatening a tempest. (1683, CEMET)

(59) I’ll warrant you wouldn’t take long getting things shipshape. (1890,

CEN)

Finally, a somewhat difficult case is presented by the development of IPCs

with bother and be bothered. Some of the first examples of bother with ing-clause in

fact seem to allow an adjunct reading: the examples in (60) are not negative implica-

tive as today’s use of not bother with IPC; instead, bother here denotes an emotive

state of worry or annoyance that arises concomitantly to the situation denoted by the

ing-clause. It is possible therefore that ing-clauses as in (60) were adjuncts. It is fur-

ther possible that they gave rise to the IPCs in (61), where bother denotes unwilling

engagement in an action. However, examples as in (60) are scarce and, in addition,

there is an alternative explanation: early IPCs typically follow the infinitive form to

bother (see (61a-b)) and may therefore have been triggered by horror æqui, as a

replacement for the more canonical to-infinitival complement with bother illustrated

in (62). Given that bother gave rather ample occasion to horror æqui, it is possible

that this was sufficient to trigger the use of IPCs.


(60) a. “Oh, dear Mrs. Charrington, it was not you I was afraid of, it was what

Dr. Bulling –” Again Iola hesitated. “Don’t bother telling me,” said

Mrs. Duff Charrington, observing her confusion. “No doubt Bulling

gave you to understand that he worked me to invite you. Confess

now.” (1906, CEN)

b. we never refused to give a man who had been in our employment a

reference on the ground that we had given him a lot of references al-

ready, and could not be bothered giving them constantly (1903, POB)

(61) a. The U. S. has 2,761 hospitals good enough for the American Medical

Association to bother inspecting. (1928, TIME)

b. We didn’t want to bother lugging him around the countryside. (1933,

TIME)

c. As you walk with a high German officer, the salutes of passing soldiers

come so aimlessly that you wonder why he bothers taking them (1942,

TIME)

(62) I have not bothered to find out who has taken the van. (1909, POB)

Apart from these exceptions, in all cases discussed, reinterpretation gives a

very natural account of the semantic and structural change at hand. Note in this con-

text that the reinterpretation from adverbial to complement follows a recurrent pat-

tern, reflecting the distinction between complementation and adverbial modification

in terms of e-site elaboration and figure-ground construal. On the one hand, the main

clause predicates whose participle clause is reintepreted all implicitly evoke an

elaboration site or e-site, i.e. a schematic participant that is activated along with a

predicate (Langacker 1987; Keizer 2004). Busy implies an activity one is busy with;

as emotive predicates, happy and tired imply a source that triggers the emotion they

denote and in relation to which the emotion holds; the light verbs have trouble and

have difficulty imply an intended goal on the part of their subject that circumstances

hinder the subject from reaching, and so on. On the other hand, adverbial participle

clauses with these predicates tend to map onto these e-sites by pragmatic implica-

ture. Disjuncts used with busy are used to support the speaker’s claim in the main

clause by describing what the subject is busy with. Adjuncts with emotive predicates

such as happy or tired strictly speaking only specify the circumstances under which

the subject is happy or tired, but these often and naturally coincide with the source

of happiness or tiredness. Adjuncts with have trouble or have difficulty describe an

activity unfolding concomitantly to the subject’s experiencing difficulty, and since

such an unfolding activity is also related to the subject’s intentions and goals – it is


the activity the subject is trying to develop successfully – it naturally maps onto the

intended goal implicitly evoked by the predicate.

Eventually, what happens under reinterpretation is that a pragmatic implica-

ture becomes semanticised (cf. Traugott & König 1991; Lopez-Couso forthc.). The

primary consequence of this is that the mapping of the activity in the participle

clause onto the e-site of the main clause becomes encoded as part of the construc-

tion’s meaning and therefore restrictive. That is, the (schematic) activity filling the

predicate’s e-site now gets narrowed down to the activity in the participle clause.

Roughly, the activity in the participle clause becomes the only activity that the sub-

ject is claimed to be busy with, happy or tired about, or experience trouble or diffi-

culty with.

The invocation of an implicit e-site suggests that reinterpretation is guided by

an underlying form of analogy (Hollmann 2003; Denison 2004; Fischer 2007: ch.3,

p.c.; De Smet 2007a). After all, the e-site attributed to adjectives or light verb con-

structions in most cases already receives formal expression by other means than the

IPC well before the reanalysis from adverbial to IPC takes place. For example, the

use of IPCs with be tired virtually mirrors the older use of gerunds introduced by of

with the same adjective. As pointed out above, nearly all IPC constructions have

semantic equivalents of this kind in the form of prepositionally marked gerund con-

structions, and these equivalents may be thought of as providing the model on which

the reinterpretation from adverbial to IPC takes place. That is, adverbial clauses

were reinterpreted as complements on the model of already existent complement

constructions in the same environments.

The mechanism proposed to give rise to IPCs receives credibility from the

fact that it accords well with general observations from the history of English. That

is, the emergence of IPCs instantiates a development towards closer integration in

the matrix clause that is typical of adverbial clauses in general and that in each case

appears to depend on the same interaction between implicit e-sites and the pragmatic

implicatures conventionally attaching to adverbials in specific environments. For

example, to-infinitives presumably derive from purposive adjuncts but have also

acquired the possibility of functioning as verb complements (Los 2005), probably

through reinterpretation in environments such as (63a). A somewhat similar mecha-

nism is likely to have given rise to the use of lest-clauses as complements (Lopez-

Couso forthc.), again with ambiguous environments serving as the trigger (63b).

For…to-infinitives first showed up as complements with predicates ambiguous be-

tween an intransitive and a transitive reading, as in (63c) (Chapter 6). Finally, the

typological literature on complement clauses seems to confirm this pattern of devel-


opment for other languages than English (Haspelmath 1989; Croft 2001: ch.9). Ob-

viously, the recurrent pattern of adverbials pragmatically mapping onto implicit e-

sites and eventually turning into complements supports the plausibility of this sce-

nario in the emergence of IPCs.

(63) a. & blodig regn & fyren fundiaþ þas eorþan to forswylgenne & to for-

bærnenne. <HomS 26 206> (Los 2005: 48)

‘and bloody rain and fire make haste / strive to devour and consume

the earth.’

b. Alyse me of Esaues handa, mines broðor, for þam ðe ic hyne swyðe

ondræde, þe læs ðe he cume & ofslea ðas modra mid heora cildum

(c.1000, Lopez-Couso forthc.).

‘Deliver me from the hands of Esau, my brother, because I fear him

very much, lest he come and kill the mothers with their children / be-

cause I fear him very much, (I fear) that he might come and kill the

mothers with their children.’

c. First missis’s children fell ill of the measles, just when th’ week I’d

asked for came, and I couldn’t leave them, for one and all cried for me

to nurse them. (1848, E. Gaskell, Mary Barton) (De Smet 2007)

What can the account in terms of reinterpretation tell us concerning the pat-

tern of diffusion attested for IPCs? It is clear that the earliest instances of IPC-

constructions mostly arose in environments that gave occasion for reinterpretation

from adverbials to complement or complement-like clauses (indeed, where reinter-

pretation is unlikely, some alternative mechanism provides a plausible explanation:

for example, phonetic reduction in gerunds with an a-prefix in be gone and be long

and analogical extension in take time or be done). It is therefore likely that reinter-

pretation largely determined the initial distribution of IPCs over the set of predicates

potentially available for IPC-complementation. By spelling out its consequences,

this conclusion can in turn serve to explain some facts about the order in which IPCs

emerged in different environments, at the same time further corroborating the rein-

terpretation hypothesis itself.

For example, the account in terms of reinterpretation presented above rules

out the initial occurrence of IPCs in at least one syntactic environment, well in ac-

cordance with the actually attested pattern of diffusion. If IPCs appeared in envi-

ronments where they were reanalysed from participial adjuncts and disjuncts, it

makes sense that they did not at first occur as noun complements or noun postmodi-


fiers (as in Present-Day English examples (46k-l) above), since participial adjuncts

or disjuncts do not attach to nominal heads. For the sake of illustration, consider the

use of complements to the noun difficulty as illustrated in (64).

(64) a. Doctors often may not recognize the symptoms of clinical depression,

which can be sadness, low energy, loss of interest in usual activities,

difficulties concentrating, changes in eating or sleeping habits, and

suicidal thoughts. (CB)

b. He accepted the defence’s argument that Miss Short would not receive

a fair trial because of the difficulties in tracing witnesses and evidence

that would back her denial of the allegations. (CB)

Taking once more the percentage of IPCs (64a) to gerund clauses introduced by in

(64b) as a measure of the success of IPC-constructions in a given lexico-

grammatical environment, the lag of IPCs as noun complements is particularly strik-

ing: Table 10.4 compares the percentage of IPCs to gerund clauses with in as com-

plement to the light verb idiom have / find difficulty(/ies) and as complement to the

freely occurring nominal head difficulty(/ies), showing clearly that IPCs are far less

advanced in the latter environment (see footnote 139 on counting practices). Dia-

chronically, this indicates that IPCs are cascading down from the environments in

which they first appeared through reinterpretation to new environments unaffected

by the original reinterpretation, with some form of analogical extension operating as

the underlying mechanism. Thus, the initial emergence of IPCs through reinterpreta-

tion ties in with and partly explains the order of appearance of IPCs in different en-

vironments.

Reinterpretation explains differences in timing in another respect as well. The

account presented above distinguishes between disjunct-based reinterpretation for

busy and spend TIME, and adjunct-based reinterpretation for have difficulty, have

trouble, be happy, be tired and so on. In other words, IPCs with busy and IPCs or

obligatory adjuncts with spend TIME are claimed to derive from disjuncts, while

Head 1850-1920 1990-1995

Have difficulty 1.5% 72.1%

Difficulty 0.0% 13.6%

Table 10.4. Percentage of IPCs to gerund clauses with in.


other IPCs are claimed to derive from adjuncts. This distinction is of some impor-

tance because it provides a possible explanation for the remarkable head start of

busy in taking IPCs as compared to other IPC-taking predicates (cf. Table 10.3 and

the examples in (46) above). Specifically, adjunctively used participle clauses are

themselves a spreading construction, having gradually grown more frequent over the

past four centuries. To show the progression of the construction, Table 10.5 gives

estimates of the relative frequency of participial adjuncts in three sub-periods.142, 143

The possibility of reinterpretation from adjuncts in a given environment is of course

dependent on the recurrent presence of adjuncts in that environment. Although it is

impossible and probably even nonsensical to formulate a critical frequency value

above which reinterpretation can start taking place, it may be reasonably assumed

that the increase in the use of participial adjuncts, by introducing participial adjuncts

142 The figures for the period 1640-1710 are based on a subpart of PPCEME, con-taining about 95,000 words of running text, on the basis of a corpus search on the orthographical sequence -ing. The figures for the period 1850-1920 were obtained in similar fashion, carrying out a similar search on -ing on a subpart of CLMETEV containing about 89,000 words. The figures for Present-Day English were obtained from the tagged and parsed ICE-GB corpus. The ICE-GB corpus was not searched for -ing (which the corpus interface does not allow) but for all verb phrases tagged as -ing-participles (ICE-GB was preferred to CB because the latter’s interface nei-ther allows searching on parts of words nor on tags unaccompanied by lexical mate-rial). To increase comparability with the historical data only subsections of the cor-pus were sampled, using different sampling rates to adjust the balance between text genres: the sections Non-academic writing, Reportage, Instructional writing and Persuasive writing were sampled at 25%, the section Creative writing was sampled at 50%. Together, the sampled sections are good for an estimated 69,000 words of text. 143 Figures should be interpreted cautiously, since it is sometimes difficult to distin-guish participial adjuncts from participial disjuncts. The presence of an explicit sub-ject has been regarded as a fully reliable criterion to separate disjuncts from ad-juncts. Less reliable is the presence or absence of a (written) comma; instead, in case of doubt the scopal tests described in section 1.1.1 were given priority in determin-ing the adjunct/disjuncts status of a given adverbial participle. The estimated fre-quencies given in Table 5 are conservative, in that adjuncts have only been recog-nised as such when an adjunct reading yields the most plausible interpretation in context. One specific difficulty is presented by the Present-Day English data, which contain the fairly frequent use of the participle using in a somewhat grammaticalised form, as a semi-preposition roughly meaning ‘with, by means of’ (e.g. do not open the can using a knife). Because it is practically impossible to distinguish preposi-tional from non-prepositional uses, instances of this kind have not been counted. When included, the estimated frequency of participial adjuncts would rise to about 29 instances per 100,000 words for the period.


1640-1710 1850-1920 1990-1993

Adjunct partici-

ple clauses 3.3 13.5 20.2

Table 10.5. Adjunctively used participle clauses (frequencies per 100,000 words).

into more environments, created more opportunities for reanalysis, and therefore

bears indirect responsibility for the fairly simultaneous appearance of a variety of

IPC-constructions around the end of the nineteenth century.

Turning from adjuncts to disjuncts, and comparing the figures in Table 10.5

to the situation for disjunctively used participle clauses, it is found that the latter also

saw an increase in use, but the main surge in frequency occurred much earlier in the

transition from the Middle to the Early Modern period. Moreover, in all periods dis-

junctively used participle clauses vastly outnumbered their adjunctive counterparts

(relative frequencies for disjuncts are found in Killie & Swan 2006; see also Swan

2003). There is considerably less ground, therefore, to suspect that disjunct-based

reinterpretation would have been delayed by frequency-dependence in the same way

as adjunct-based reinterpretation. In combination, then, the differently timed devel-

opments of participial disjuncts and adjuncts, and the different starting points for

reinterpretation to IPCs can account for the time-lag between the use of IPCs with

busy and other predicates. Furthermore, disjunct-based reinterpretation may also

account for the use of obligatory adjuncts with spend TIME at a time when adjuncts

were still highly infrequent in other environments.

Other details of the order in which various predicates began to select IPCs can

be explained in a similar fashion. For example, it has been pointed out above that

among the emotive predicates, be tired was probably the first to acquire IPCs, fol-

lowed by be happy and other adjectival predicates. This sequence of events ties in

with changes in the use of participial adjuncts, which not only grow in number in the

course of the Late Modern period (as shown in Table 10.5 above), but also spread to

new sub-uses. Already in the nineteenth century adjuncts quite readily combine with

change of state predicates, especially with the verb get, specifying the manner or

cause of the subject’s moving from one state to another, as shown in (65). The use of

adjuncts with (get) tired that probably gave rise to IPCs with the same adjective can

be seen to instantiate this schematic pattern of co-occurrence (as is illustrated in

(51a-b) above).


(65) a. You couldn’t walk to-morrow if you took all the free samples of solid

gold the boys would offer you. You’d get dizzy looking down prospect

holes. (1905, CEN)

b. then Snowdon and a son as he had both got drowned going over a river

at night. (1889, CEN)

c. I began to hope, although I’d got wrinkles crying about him. (1900,

CEN)

By contrast, in the nineteenth century, predicates expressing an emotive state (rather

than a change of state) only sporadically combine with adjuncts specifying the cir-

cumstances related to that state, as in (66a-b). This schematic pattern accounts for

the adjuncts getting reinterpreted as IPCs outside the change of state pattern, and the

fact that it only became current in the course of the twentieth century correlates with

the time of emergence of IPCs with emotive adjectives such as happy and comfort-

able.

(66) a. He could not go on with his honeymoon, so he would go up to London

and work – he felt too miserable hanging about. (1910, CLMETEV)

b. She remembered how she was reproved for peeping over her

neighbour’s shoulder, and how proud she felt sitting among all the

workwomen. (1885, CEN)

Again, then, the dependence of the first IPC-constructions on reinterpretation can be

brought to explain facts about the specific order of emergence of different uses.

2.2.2. Analogical extension

The early development of IPCs, which is mostly characterised by mutually inde-

pendent developments through reinterpretation, is followed by a stage of growing

coherence in the use of IPCs, as a result of analogical extensions introducing IPCs in

environments that do not invite reinterpretation, showing that IPC-complementation

is gaining productivity. Analogy surfaces in its familiar forms, involving either se-

mantic analogy (on the basis of semantic similarity between complement-taking

predicates) and paradigmatic analogy (with gerunds introduced by in serving as the

favourite paradigmatic analogue). Thus, despite the predominance of reinterpretation

in the development of IPCs, not all is reinterpretation, and IPCs also appeared in a

number of environments where their use had not been foreshadowed by adverbial


participle clauses. Assessing the influence of analogy, analogical extension seems to

be a weak force that mostly works locally, and that, in the case of IPCs, seems still

far from having run its full potential course. In that respect, the analogical exten-

sions in IPCs resemble the local analogies in the diffusion of for...to-infinitives

(Chapter 6), or the first semantic analogies in the development of gerundial ing-

complements (Chapter 9). Still, while its effects are limited, analogy can not be ig-

nored as an independent mechanism of change.

There are some relatively successful IPC-uses whose existence is most read-

ily explained by analogical extension. There is the example of be done discussed

earlier (see (57) above), whose use with IPCs from the end of the nineteenth century

onwards is probably not the result of reinterpretation but is based on the use of IPCs

(or some kind of -ing-clause) with have done, and which in turn is likely to have

formed the basis for the occasional use of IPCs with be finished in Present-Day Eng-

lish. Another nineteenth-century example discussed above is the possible extension

of IPCs from be long to take time (with the possible help of the spend TIME pattern

and the blended pattern take long). A more recent example is the use of IPCs with

have no right, which in all likelihood derives from the use of IPCs with have no

business.

In other cases, innovations arising through analogical extension are less suc-

cessful. A truly innovative but otherwise relatively unsuccessful case of analogical

extension is presented by the use of IPCs with simplex verbs, such as assist, hesi-

tate, succeed or tire, as in (67) (or (38) above), the occurrence of which has so far

remained restricted to a few scattered instances in the corpus material. Note further

that the analogical models on which the simplex verb formations are based are not

altogether straightforward, although there do exist semantic and formal relations

linking the simplex verbs to other, more common IPC-taking predicates. Semanti-

cally, assist can be linked to need help (see example (38d) above), hesitate to have

no hesitation which in turn relates naturally to have (no) difficulty, succeed to have

(no) success which in turn is again linked to have (no) difficulty. Paradigmatically,

assist and succeed (though not hesitate) can be construed with gerund clauses intro-

duced by the preposition in, just like many other IPC-taking predicates. The use of

IPCs with tire, on the other hand, can be linked easily to the IPCs with emotive ad-

jectives, especially tired and fed up.

(67) The Town Council, who are the Local Health /Authorities, have been

condemning many old properties recently, and are pressing proprietors

to put their houses in order; but they have been considerably hampered


by the great scarcity of houses, and naturally hesitate taking action

which would have people removed with no place to house them.

(Google, August 2006)

Another likely instance of analogical extension, motivated both semantically

and paradigmatically, is presented by the use of IPCs with engaged, occupied, and

employed as in (68), in all likelihood mirrored on the use of IPCs with busy. But as

the figures in Table 10.3 above indicate, here too analogical pressure seems to be

insufficient to forward these uses beyond occasional occurrence.

(68) a. Winter in a Flat racing yard is the most hated time of year for stable

lads. It’s the time they’re engaged breaking in the yearlings and, apart

from the odd all-weather card, there are no race meetings to break the

monotony. (CB)

b. […] in a densely peopled quarter of the city, such as in our own day we

should call a slum, where folk were employed making those articles

which ministered to the comfort of the luxury of the more fortunate

[…]. (1903, CEN)

c. He had hurled his lasso with the rest, and it was trailing. He jerked

about and fled for a mile or more, holding on with his legs while both

hands were occupied gathering in the rope and coiling it about the high

pommel of his saddle. (1898, CEN)

Similarly, analogical extension is likely to have been at play in the use of IPCs as

noun complement to, for instance, the noun difficulty. Again, however, the data dis-

cussed earlier (see in particular Table 10.4 above) reveals that the extension of IPCs

from one environment to another is hesitant. This lack of productivity is most con-

spicuous, finally, for emotive adjectives: while these seem to be well-represented

among the IPC-taking predicates and seem to form a natural basis for analogical

extension, there still exist a good number of common emotive adjectives that remain

unattested in the pattern in CB, including annoyed, delighted, excited, glad, sad,

sorry, surprised, upset and so on. Only occasionally do IPCs extend to new emotive

adjectives: (69) is a probable example.

(69) lease has been on a year-to-year basis but we are hopeful negotiating a

20-year deal with the Council. (CB)


In sum, a number of innovative IPC uses are most easily explained as the outcome

of a process of analogical extension, and these extensions seem to assert IPCs as a

(mildly) productive syntactic pattern. At the same time, it is certainly not the case

that analogy has caused the use of IPCs to have become fully systematic and pre-

dictable.

2.3. Diachronic overview

To summarise the preceding discussion, Table 10.6 lists some of the different IPC-

constructions examined with their most likely historical source and the mechanisms

by which they arose. As Table 10.6 shows, developments start out with IPCs appear-

ing with various more or less isolated predicate types, primarily as a result of syntac-

tic reinterpretation, and in addition through the unrelated loss of a-prefixes in some

gerundial constructions and similar local developments. This means that the early

distribution of IPCs has largely been determined by a factor lying outside the system

of complementation, namely the changing use of adverbial participle clauses, and by

developments that are quite unrelated to this. The following step, however, consists

in speakers recognising certain regularities and extending these on the basis of se-

mantic and/or paradigmatic analogy. The shift in mechanisms is familiar from other

diffusional changes (see especially the diffusion of gerundial ing-complements, ex-

amined in Chapter 9), though, in the case of IPCs, this process is still far from estab-

lishing a predictable and synchronically principled distribution for IPC-

complementation.

At the same time, the diffusion of IPCs differs from the diffusion of other

non-finite complement types, in that it has relied much more strongly on reinterpre-

tation. That is, while for...to-infinitives and gerundial ing-complements primarily

entered the system of complementation through forms of paradigmatic analogy,

IPCs entered the system primarily through syntactic reinterpretation. Indeed, the

lack of a strong paradigmatic analogue whose distribution IPCs can replicate with-

out having to compete may be responsible for the marginal status of IPCs in the

grammar of English.144 As a diachronic phenomenon, the emergence of IPCs instan-

144 The most obvious paradigmatic analogue to IPCs is the gerundial construction with in, but apart from the fact that the paradigmatic variation between the two pat-terns is relatively limited, the two patterns are also in competition, which means that a paradigmatic tie is as much an obstacle as a backing to the diffusion of IPCs, espe-cially in the absence of a functional advantage of one pattern over the other.

FIRST ATTESTATION IPC-CONSTRUCTION MOST PLAUSIBLE HISTORICAL SOURCE

16th century Have done + IPC Have done + gerund

17th century Be busy + IPC Reanalysis from be busy + disjunct

Spend TIME + IPC Reanalysis from spend TIME + disjunct

Be long + IPC Be long + gerund with a-prefix

19th century Be tired + IPC Reanalysis from get / be tired + adjunct

Have difficulty / no business Reanalysis from have difficulty / no business + adjunct

Be done + IPC Analogical extension from have done + IPC

Take time + IPC Analogical extension from be / take long + IPC and/or spend TIME + IPC

19th/20th century Be happy + IPC Reanalysis from be happy + adjunct

Be late + IPC Reanalysis from be late + adjunct or analogical extension from be long + IPC

20th century Have no right + IPC Analogical extension from have no business + IPC

Difficulty + IPC Analogical extension from have difficulty + IPC

Succeed + IPC Analogical extension from have success + IPC

Hesitate + IPC Analogical extension from have no hesitation + IPC

Tire + IPC Analogical extension from tired + IPC

Table 10.6. Overview of IPC-constructions and their possible historical sources.


tiates a shift in the way a subordinate clause is conceptually related to its matrix

clause, either in terms of figure-ground construal or in terms of e-site elaboration.

Although adverbial and complement clauses are conceptually distinct, there are at

least two reasons why the boundaries between the two are not too strict. On the one

hand, e-site elaboration is not an all-or-nothing affair. Because the requirement of a

complement is dependent on the expectation of an extra participant linked to the

complement-taking head, not all complements are ‘obligatory’ to the same degree

(Langacker 1987; Keizer 2004) and e-sites can remain unelaborated. On the other, it

has been observed that “adjunct-less intransitive clauses are rare” (Quirk et al. 1985:

506) – in other words, intransitive predicates are unlikely to occur without any fur-

ther adverbial modification. In combination, implicit e-sites and the relative obliga-

toriness of adverbials form a breeding-ground for the emergence of new comple-

ment-types. The change from adverbial to IPC is possible because predicates have

an elaboration site that is sufficiently weak to be left implicit but sufficiently strong

to assert itself and attract a new complement-type when the occasion arises.

Let us finally consider the grammatical consequences of the development of

IPCs. It has been suggested that grammars change because they leak (Denison

2004). True though this may be, the opposite is probably also true: grammars leak

because they change. The change from adverbial to IPC creates ‘leakage’, in that a

new set of constructions emerges whose categorial status confuses existing cate-

gorial boundaries. The categorial boundaries confused, in this case, are those of ing-

complementation as a nominalisation. Taking up again the argument of Chapter 7, if

we assume that speakers generalise a pattern of ing-complementation (at whatever

level of linguistic salience), the emergence of IPCs must be seen as another step in

the denominalisation of ing-complements, assaulting one of the last strongholds of

ing-complements’ original nominal character, namely their distribution. Gerundial

constructions developed internal clausal syntax and clausal grounding strategies

(Chapter 8); gerundial complements, though remaining associated with the distribu-

tion of noun phrases, became less closely integrated in the grammar of the noun

phrase by losing touch with the bare abstract nouns under which they originally re-

sorted (Chapter 9). Finally, partly as a result of the emergence of IPCs, the nominal

distribution of ing-complements comes under pressure.

To be sure, the actual impact of the emergence of IPCs on the status of ing-

complementation remains hard to estimate. Apart from superficial formal identity,

there are no signs that IPCs and ing-complements are seen by speakers as a single

category. There is no telling to what extent the existence of an abstract category of

ing-complementation may have furthered the reinterpretations that gave rise to IPCs.


Conversely, there is no evidence that IPCs did not arise independently of gerundial

ing-complementation. What is clear is that the cognitively most prominent categorial

representations – namely those responsible for analogical extensions – are highly

local regularities, specific to the domain of IPCs. In this light, the emergence of

IPCs can only be a first step in the denominalisation of ing-complementation.

3. Conclusions

This chapter has examined the use and history of non-nominalised subject-controlled

ing-complements or IPCs. As such, it serves as a counterpoint to the discussion of

nominalised (gerundial) subject-controlled ing-complements in Chapter 9. The syn-

chronic discussion of IPCs, in addition to revealing a fuzzy area between comple-

mentation and adverbial modification, has shown that the use of IPCs is organised

around a cluster of fairly specific constructions. These constructions (whose seman-

tic relations cut across predicate types) give rise to disparate distributions. For ex-

ample, in the domain of adjective complementation, IPCs are found with emotive

adjectives such as tired or happy, but also with adjectives like busy, late or even

successful. Synchronically, then, IPCs represent a complement-type whose use is

impossible to capture in a simple functional characterisation.

Diachronically, however, the development of IPCs resembles that of other

complement-types in showing a stage of very local innovations that occur more or

less independently of one another, followed by a stage of growing coherence during

which small constructions are recognised and extended. The development of IPCs

only differs from that of other complement-types in that it initially relies more heav-

ily on syntactic reinterpretation rather than some form of paradigmatic analogy. On

the one hand, this may explain the relative lack of success of IPCs. On the other, it

illustrates another mechanism in the emergence of complement constructions, in the

form of a pragmatically induced shift between different ways of conceptualising the

relationship between two events, either as a relation of e-site elaboration, or as a

figure-ground relation.

In a larger diachronic picture, the emergence of IPCs is one of the small de-

velopments that puts the character of (subject-controlled) ing-complements as nomi-

nalisations under pressure. To be sure, there is no evidence that the emergence of

IPCs is causally related to the diffusion of gerundial ing-complements, as a direct

consequence of the appearance of gerundial ing-complements in an ever-increasing

number of environments (though there is no evidence supporting the opposite con-

clusion either). Moreover, looking at the patterns that are productive in present-day


usage, we can distinguish between a productive fairly schematic construction of

nominalised ing-complements occurring in nominal positions with transitive verbs

(Chapter 9) and a cluster of non-nominal ing-complements occurring with some

light verbs, verbal idioms, adjectives, verbs and nouns. Assuming that the produc-

tive constructions are also the most salient generalisations, this means that there is

little constructional unification between gerundial ing-complements and IPCs. How-

ever, in as far as speakers do generalise to a category of (subject-controlled) ing-

complementation, the development of IPCs eats away at the originally nominal

character of the construction, further cutting the ties between ing-clauses and noun

phrases.

PART V Conclusions

“evermore one change leaves a kind of breach or dent,

to fasten the building of another”

(Niccolo Machiavelli)

CHAPTER 11: Conclusions

The preceding analyses of historical developments have revealed a recurrent pattern

of change. Following its emergence, a new pattern spreads to different environments

at different times, diffusing over the grammar of the language. Diffusional changes

raise a number of questions. How does diffusional change relate to the synchronic

organisation of grammar and vice versa? How does diffusional change proceed, and

why does it occur? Let me recapitulate here how the preceding chapters have con-

tributed to answering these questions.

Obviously, it is crucial to understand how synchronic grammar and especially

the synchronic system of complementation are organised. The view of the system of

complementation argued for here is informed by corpus-based case-studies of syn-

chronic usage (Chapters 3, 6, 7, 10), but also answers to the phenomenon of diffu-

sional change, which must be incorporated in our understanding of grammar and

language use. Figure 11.1 sketches the model envisaged of grammar and language

use. Grammar consists of linguistic knowledge of constructions, which capture lin-

guistic generalisations, including (semi-)productive distributional regularities in lan-

guage as well as recurrent form-meaning mappings (in accordance with Langacker

1987; Goldberg 1995, 2006; Croft 2001). Constructions form a continuum, with

highly schematic constructions at one end (whose use requires the creative assembly

of various linguistic elements), and fully specific constructions at the other (whose

use implies reproduction of stored sequences). Productivity and regularity imply the

existence of schematic patterns; phenomena such as local regularities or local se-

mantic changes and especially blocking imply the existence of less schematic, lexi-

cally more specified constructions (Chapters 3, 4, 9).

A particularly strong illustration of the layered character of our constructional

knowledge comes from the use of ing-clauses (Chapter 7). Evaluating different types

of evidence, there are indications of the cognitive reality of a highly abstract super-

category (corresponding to Huddleston and Pullum’s 2002 ‘gerund-participles’)

which allows for formal features to be sporadically exchanged between the tradi-

tional categories of gerunds and participles. At the same time, the distributional ties

between (traditional) gerundial ing-clauses and noun phrases and between present

and past participles remain in evidence, which means that the basis for the ger-

und/participle distinction has not eroded away. Further still, there is also evidence of

organisation at lower levels of schematicity. For example, the distribution of inte-

492 – CHAPTER 11: Conclusions

Figure 11.1. Grammar and language use.

grated participle clauses (Chapter 10) must be described with reference to highly

specific combinations and cannot be understood exclusively through a general char-

acterisation of ing-clauses or present participles.

The implementation of this construction grammar in language production

requires an evaluation mechanism as well as a feedback mechanism. Language us-

ers’ store of constructions is extremely rich and can be assumed to offer a variety of

ways of expressing a given mental content. This is most straightforwardly apparent

from various instances of historical competition between variants (Chapters 3, 8). In

language production, however, a speaker must select one way of conveying his/her

message (X in Figure 11.1) from the options available (X1, X2, X3, ... Xn). From us-

X

X1

GRAMMAR

SCHEMATIC CON-

STRUCTIONS

SPECIFIC CON-

STRUCTIONS

Creative

Reproductive

X2 X3 Xn

PROCESSING / PRAGMATIC

CONSTRAINTS

EVALUATION

CHAPTER 11: Conclusions – 493

age data it is furthermore apparent that extra-grammatical factors, such as priming,

rhythmic alternation, cognitive complexity, or horror æqui, impinge on the selection

of a construction to be uttered (Chapter 3). In light of this, the selection process can

be thought of in terms of a racing or competition model (Schlüter 2005). Variants

are differently sanctioned by the grammar and the various processing constraints

(not to mention the degree to which they fit the speaker’s communicative inten-

tions), whereby the most strongly sanctioned variant will come out ‘first’ and get

selected. The phenomenon of blocking, for example, can be understood in terms of a

particular combination receiving very strong grammatical sanction as a highly spe-

cific routinised chunk of language, which consequently outcompetes alternative uses

sanctioned by other regularities of the grammar. Similarly, horror æqui can be

thought of as a functional constraint overriding the principled regularities of gram-

mar. How exactly sanctioning works can be left in the middle, but as far as sanction-

ing through the grammatical component is concerned, frequency clearly plays a cen-

tral role, which would be compatible with a spreading-activation model. Thus,

blocking only works for highly frequent patterns and regularities gain productivity

as they grow more frequent (Chapters 4, 9).

Frequency also brings us to the feedback mechanisms in language production.

As Figure 11.1 indicates, each usage event feeds back into the grammar (not only of

the speaker but also of the hearer) and reinforces the constructions – both specific

and schematic – which it instantiates. Feedback is the point where synchronic lan-

guage use gets diachronic implications, as it implies that every time the grammar is

employed (or not employed!) it changes a little. Thus, feedback may be the only way

in which the architecture of grammar can come to reflect functional or pragmatic

constraints (Haspelmath 1999; Chapter 8). It is also the process that is central in

understanding the hypothesised phenomenon of analogical snowballs (Chapters 4, 9;

see below).

Turning from this to diffusional change, two questions need to be addressed.

One is what gives diffusional change its diffusional character – that is, why does a

new pattern not arise in different environments simultaneously? The other is why a

pattern spreads at all – that is, why can there (apparently) be no status quo? The dif-

fusional character of diffusional change is determined by a variety of factors, which

all have in common, however, that they cause different environments to be suscepti-

ble to change either to different degrees or at different times. In many respects, this

multiplicity of factors is the diachronic correlate of the complicated synchronic

make-up of the grammar, often being at once its cause and its consequence.


First, it is evident that new patterns start out by arising in only a few envi-

ronments because triggering factors are specific to those environments. For...to-

infinitives, for example, were initially associated with adjunct positions (because

that is where one of their ancestors, the forto-infinitive, was most common) and with

extraposed subject positions (where they were tied up to organic for...to-sequences)

(Chapter 5). Similarly, integrated participle clauses arose in contexts where a poten-

tially ‘transitive’ predicate (i.e. a predicate with an implicit e-site) combined with a

participial adverbial clause that could be reinterpreted as a complement. Thus, a first

batch of integrated participle clause constructions appeared through reinterpretation

of participial disjuncts; a second batch through reinterpretation of participial ad-

juncts (Chapter 10). It is, from this point of view, understandable that an innovative

pattern does not appear in all environments at the same time. Linking this to the pic-

ture of synchronic grammar, it is language users’ ability (and inclination) to work

with highly specific constructions that allows this primary stage of diffusional

change to occur.

Second, it has been shown that, synchronically, the use of a pattern can be

determined by a number of different constructional generalisations which may partly

overlap (Chapters 3, 6, 7). This complex state of affairs can be a reflex of a dia-

chronic chain of analogical extensions, with one extension opening up new possibili-

ties for another extension (though at the same time, the diachronic development is

itself a reflex of language users’ synchronic ability to abstract and exploit multiple

regularities). In such analogical chains, speaker’s recognition of one regularity in-

vites changes that give rise to new regularities and so on. For example, for...to-

infinitival complements occurred early on with volitional verbs such as wish and

long, presumably leading to the generalisation that volitional verbs could take

for...to-infinitives, including a verb such as like. This subsequently led to for...to-

infinitives appearing with verbs such as love or hate, which resemble like as co-

members of the emotive class (Chapter 6). The clearest example of an analogical

chain comes from the diffusion of gerundial ing-complements, with the mechanism

underlying diffusion shifting from narrow paradigmatic analogy to semantic analogy

and eventually to indirect and broad paradigmatic analogy (Chapter 9). In analogical

chains, the diffusional character of change is explained by the fact that one change

only becomes possible as a result of a previous change.

A variant of analogical chains arises from the fact that different environments

sanction a given pattern to different degrees. The key to understanding this point is

that a use is often sanctioned by more than one grammatical regularity – a notion

that is in line with the synchronic characterisation of grammar given above (cf. Lan-


gacker 1987; Denison 2001). For example, for...to-infinitives are sanctioned in a

great number of environments by their paradigmatic tie to to-infinitives. The fact

that for...to-infinitives nevertheless spread to different environments at different

times is explained in part by additional points of resemblance between source and

target environments of diffusion. Thus, for...to-infinitives spread to non-extraposed

subject positions in copular clauses and extraposed subject positions in non-copular

clauses before they spread to non-extraposed subject positions in non-copular

clauses, despite the fact that paradigmatic analogy would sanction their use equally

in all three environments. Presumably, this is because the former two environments

more closely resembled the more common use of for...to-infinitives in extraposed

subject positions of copular clauses than the latter (Chapter 5).

Third, environments also show differences whose significance to diffusion is

purely incidental to the pattern that is spreading. Thus, it is conceivable that regu-

larities are in some environment systematically suspended or reinforced by inde-

pendent cognitive-functional factors such as horror æqui, cognitive complexity or

rhythmic alternation (possible examples are the use of ing-complements with at-

tempt and bother; see Chapters 9 and 10 respectively). More clearly evident from the

preceding analyses, blocking, which depends on the incidental presence of a com-

peting variant, can hinder or delay diffusion in specific environments, as in the use

of for...to-infinitives with want (Chapter 6), or the use of ing-complements with be-

gin, or non-passival clausal ing-complements with need and want (Chapter 9). Since

blocking correlates with frequency, the extent to which it can delay diffusion to a

specific environment is inevitably variable.

The most important ‘incidental’ factors, however, are even more specific. In

the diffusion of for...to-infinitives the potential occurrence of ambiguous for...to-

sequences in a given environment proved a catalyst of diffusion, with for...to-

infinitives being primarily attracted to environments where their appearance could

be prepared by ambiguous for...to-sequences and thus remained relatively ‘dis-

guised’ by the presence of the homonymous structure (Chapters 5, 6). In a similar

vein, the ground for gerundial ing-clauses as well as clausal ing-complements was

prepared by bare nominal gerunds and other action nominals (Chapters 8, 9). It

seems in these cases as though a primarily statistical (i.e. quantitative) change,

makes a more dramatic qualitative change less conspicuous. This is not to say that

one change is meant to prepare the ground for the latter (in a teleological sense); it is

more likely that both changes are part of the same macro-change but that the spread-

ing pattern advances the faster the less noticeable its progression is (Naro 1981;


Warner 1982). Again, then, diffusion can be linked to synchronic differences in

sanctioning.

Fourth, the environments to which a diffusing pattern can spread are them-

selves subject to change and may, as a consequence, undergo shifts that make them

more or less compatible with the spreading pattern. This is especially true at the

level of lexical changes and the diffusion of a pattern to new lexical environments.

The clearest example here is again the diffusion of gerundial ing-complements,

whose progression can be seen to have been determined to a considerable extent by

independent lexical changes (Chapter 9).

Turning to the question why diffusional change occurs, answers become more

speculative, yet a number of factors can be picked out as potentially relevant. Func-

tional motivations have been most explicitly addressed in this study with respect to

the rise of verbal gerunds at the expense of nominal gerund constructions (Chapter

8). The discussion indicates that where diffusion is replacive – that is, where the

spreading pattern competes with and replaces an older variant – it may be driven by

selectional pressures that consistently favour the same variant. In the case of verbal

gerunds and nominal gerunds, the former were consistently favoured as at once

more economic and more iconic. What happens in such cases is that through the

feedback mechanism outlined above, every functionally motivated choice reinforces

the spreading pattern at the expense of the old variant, which gradually gets re-

placed. This of course does not explain the prior rise of bare nominal gerunds – the

older variant. Regarding the latter, it has been suggested that the replication of for-

eign linguistic patterns may have been a driving force of change.

Where no selectional pressures on variation can be invoked, answers to why

diffusion takes place are more complex. Part of the solution must lie in the factors

that determine the success of an individual innovation in a specific environment. In

all likelihood, these factors are to a good degree language-external, bringing into

play the social dynamics of language change and also perhaps the more mysterious

variable of innovations’ intrinsic usefulness. Still, even if it is difficult to explain

their subsequent success, it is, in view of the preceding, not all that hard to see why

innovations occur. From a language-internal point of view, at least two factors inter-

act: analogical snowball effects on the one hand, and the inevitability of change on

the other.

As indicated above, changes can give rise to new regularities. For example,

the appearance of for…to-infinitives created a paradigmatic variant to the to-

infinitive (Chapters 5, 6); similarly, the appearance of verbal gerunds created a para-

digmatic variant to noun phrases (Chapters 8, 9). In the model of synchronic usage


outlined above these regularities sanction innovations in other environments even if

the innovative uses are not selected in actual language production. Crucially, how-

ever, since there is a correlation between frequency and productivity (Chapter 4), it

can be hypothesised that mere quantitative developments in one environment can

increase the likelihood of change in other environments. That is to say, as the

spreading pattern gains currency in one environment, the pressure increases on other

environments to follow, as it becomes more likely for the innovative patterns to get

selected for actual linguistic output, setting in motion an analogical snowball.

For example, the diffusion of for…to-infinitives to subject-complement posi-

tions can be interpreted in this way: the regularity that sanctions this particular ex-

tension – i.e. paradigmatic analogy to the to-infinitive – remained the same through-

out the diffusion of for…to-infinitives, except that it became stronger, as it was

borne out in more environments and as for…to-infinitives themselves became more

current (Chapter 5). Similarly, the regularity that sanctioned the appearance of ge-

rundial ing-complements with verbs such as enjoy or try – i.e. broad paradigmatic

analogy – had been inherent in the use of gerundial ing-complements throughout

their development, but only turned productive in the latter stages, presumably as it

had gained sufficient salience (Chapter 9). Incidentally, the case of integrated parti-

ciple clauses offers an interesting counterpoint: the appearance of integrated partici-

ple clauses offered no perspectives on a powerful generalisation, which may explain

the continued marginal status of the pattern (Chapter 10).

On top of this, there are additional reasons why, as analogical pressure grows,

change is practically inevitable. One reason is that in some environments the thresh-

old for a spreading pattern to appear is extremely low. This is particularly so when

an environment already selects patterns that closely resemble the innovative spread-

ing construction (e.g. ambiguous for…to-sequences, indeterminate nominal gerunds;

see Chapters 5, 6, 9). In such environments, diffusion is halfway before it has actu-

ally begun. The other reason lies in the processing and other constraints that are ex-

ternal to grammar proper (thought of as stored linguistic knowledge based in experi-

ence) but that nevertheless influence language production. Many of these factors are

highly context-dependent (e.g. priming, horror æqui, etc.), which means that a pat-

tern is never sanctioned in the same lexico-grammatical environment to the same

degree. This means that the moment a pattern is minimally sanctioned in some envi-

ronment through the grammar of the language, some occasions are bound to occur

when functional and pragmatic factors conspire to give the additional boost neces-

sary for the pattern to get selected as actual linguistic output in the environment in

question.


In conclusion, it has been argued in this study that the system of complemen-

tation is organised around a great number of constructions, which, moreover, are

layered in terms of schematicity. This means that the matching of complement types

to predicates and the choices between complement types in specific discourse con-

texts are motivated, yet also that there is no neat one-to-one relation between organ-

ising principles and complement types. This in itself makes the architecture of the

system of complementation highly complex, yet to complicate matters further, it is

not only grammar proper that determines complement choice, but also a set of addi-

tional pragmatic and cognitive-functional constraints.

Adding diffusional change to the picture, it has been shown that the progres-

sion of diffusion – and, indeed, the very fact that diffusional change is diffusional –

is partly determined by the complex architecture of grammar and the system of

complementation and partly by factors that are incidental to the structure that is

spreading. The latter set of factors are arbitrary from the point of view of comple-

mentation as a functional system. Regarding the question why diffusional change

occurs, this study has suggested a number of language-internal mechanisms that

cause change to proceed unidirectionally (though no doubt language-external factors

must play their part as well): these include functional pressures on selection, ana-

logical snowballing, and a set of factors that make change practically inevitable.

REFERENCES

Aitchison, Jean. 1991. Language change. Progress or decay? Cambridge: Cam-

bridge University Press. Andersen, Henning. 2001. Actualization and the (uni)directionality of change. In

Henning Andersen (ed.). Actualization. Linguistic change in progress. Ams-terdam: John Benjamins. 225-248.

Anglade, Joseph. 1966. Grammaire élémentaire de l’ancien français. Paris: Colin. Ariel, Mira. 1990. Accessing noun-phrase antecedents. London: Routledge. Behrens, Leila. 2005. Genericity from a cross-linguistics perspective. Linguistics 43:

275-344. Berg, Thomas. 1998. Linguistic structure and change. An explanation from lan-

guage processing. Oxford: Clarendon Press. Bladon, R.A.W. 1968. Selecting the to- or -ing nominal after like, love, hate, dislike,

and prefer. English Studies 49: 203-214. Bolinger, Dwight. 1968. Entailment and the meaning of structures. Glossa 2: 119-

127. Bolinger, Dwight. 1977. Meaning and form. London: Longman. Brems, Lieselotte. 2003. Measure noun constructions. An instance of semantically

driven grammaticalization. International Journal of Corpus Linguistics 8: 283-312.

Bresnan, Joan W. 1979. Theory of complementation in English syntax. New York: Garland Publishing Inc.

Brinton, Laurel. 1988. The development of English aspectual systems. Aspectualiz-ers and post-verbal particles. Cambridge: Cambridge University Press.

Brunot, Ferdinand. 1966. Histoire de la langue française. Des origines à nos jours. Vol. 1. De l’époque latine à la Renaissance. Paris: Colin.

Bybee, Joan. 1985. Morphology. A study of the relation between meaning and form. Amsterdam: John Benjamins.

—2000. The phonology of the lexicon. Evidence from lexical diffusion. In Michael Barlow and Suzanne Kemmer (eds). Usage-based models of language. Stan-ford: CSLI. 65-86.

—2006. From usage to grammar. The mind’s response to repetition. Language 82: 711-733.

—Forthcoming. Word frequency and context of use in the lexical diffusion of pho-netically conditioned sound change. Language Variation and Change.

Bybee, Joan and Dan I. Slobin. 1982. Rules and schemas in the development and use of the English past tense. Language 58: 265-289.

Cappelle, Bert. 2005. Particle patterns in English. A comprehensive coverage. Doc-toral thesis, Department of linguistics: University of Leuven.

—2006. Particle placement and the case for “allostructions”. Constructions. http://www.constructions-online.de/articles/specvol1/683.

Carlson, Greg N. 1977. A unified analysis of the English bare plural. Linguistics and Philosophy 1: 413-57.

Chomsky, Noam. 1980. On binding. Linguistic Inquiry 11: 1-46. Claridge, Claudia. Sine dato. “Life is ruled and governed by opinion”. The Lampeter

corpus of Early Modern English tracts. Manuel of information. http://www.tu

500 – References

-chemnitz.de/phil/english/chairs/linguist/real/independent/lampeter/manual/ pages/contents.html.

Coates, Jennifer. 1983. The semantics of the modal auxiliaries. London: Croom Helm.

Comrie, Bernard. 1976. The syntax of action nominals. A cross-linguistic study. Lingua 40: 177-201.

Comrie, Bernard and Sandra A. Thompson. 1985. Lexical nominalization. In Timo-thy Shopen (ed.). Language typology and syntactic description. Cambridge: Cambridge University Press. 349-98.

Croft, William. 2000. Explaining language change. An evolutionary approach. Lon-don: Longman.

—2001. Radical construction grammar. Syntactic theory in typological perspective. Oxford: Oxford University Press.

—2003. Lexical rules vs. constructions. A false dichotomy. In Hubert Cuyckens, Thomas Berg, René Dirven and Klaus-Uwe Panther (eds). Motivation in lan-guage. Studies in honor of Günter Radden. Amsterdam: John Benjamins. 49-68.

—Forthcoming. Construction grammar. In Dirk Geeraerts and Hubert Cuyckens (eds). The handbook of cognitive linguistics. Oxford: Oxford University Press.

Croft, William, and Alan D. Cruse. 2004. Cognitive linguistics. Cambridge: Cam-bridge University Press.

Curme, George O. 1931. Syntax. Boston: Heath. Cuyckens, Hubert and Hendrik De Smet. 2007. For…to-infinitives from Early to

Late Modern English. In Javier Pérez-Guerra, Dolores Gonzalez-Álvarez, Jorge L. Bueno-Alonso and Esperanza Rama-Martínez (eds). ‘Of varying language and opposing creed’. New insights into Late Modern English. Bern: Peter Lang. 77-103.

Dal, Ingerid. 1952. Zur Entstehung des englischen Participium Praesentis auf -ing. Norsk Tidsskrift forSprogvidenskap 16: 5-116.

Davidse, Kristin. 1999. The semantics of cardinal versus enumerative existential constructions. Cognitive Linguistics 10: 203-50.

Declerck, Renaat. 1981. On the role of progressive aspect in nonfinite perception verb complements. Glossa 15: 83-113.

—1982. The triple origin of participial perception verb complements. Linguistic Analysis 10: 1-26.

—1991a. A comprehensive descriptive grammar of English. Tokyo: Kaitakusha. —1991b. The origins of genericity. Linguistics 29: 79-102. Dekeyser, Xavier. 1986. Romance loans in Middle English. A re-assessment. In Di-

eter Kastovsky and Aleksander Swedek (eds). Linguistics across historical and geographical boundaries. In honour of Jacek Fisiak on the occasion of his fiftieth birthday. Berlin: Mouton de Gruyter. 253-265.

Denison, David. 1993. English historical syntax. Verbal constructions. London: Longman.

—1994. A corpus of Late Modern English prose. In Merja Kytö, Matti Rissanen and Susan Wright (eds). Corpora across the centuries. Proceedings of the First International Colloquium on English Diachronic Corpora. Amsterdam: Ro-dopi. 7-16.

References – 501

—1998. Syntax. In Suzanne Romaine (ed.). The Cambridge history of the English language. Vol. 4. 1776-1997. Cambridge: Cambridge University Press. 92-329.

—2001. Gradience and linguistic change. In Laurel J. Brinton (ed.). Historical lin-guistics 1999. Selected papers from the 14th International Conference on His-torical Linguistics, Vancouver, 9-13 August 1999. Amsterdam: John Benja-mins. 119-144.

—2004. Do grammars change when they leak? In Christian J. Kay, Simon Horobin and Jeremy Smith (eds). New perspectives on English historical linguistics. Amsterdam: John Benjamins. 15-29.

De Smet, Hendrik. 2003. Synchronic and diachronic variation in the non-finite com-plementation patterns of the verbs like, love and hate. Unpublished MA the-sis, Department of Linguistics: University of Leuven.

—2004. Semantics and variation in complement constructions. Gerunds and infini-tives following the verb like. Belgian Journal of English Language and Lit-eratures 2: 247-260.

—2005. A corpus of Late Modern English texts. ICAME-Journal 29: 69-82. —2007a. In search of syntactic reanalysis. A usage-based approach. Preprint 261,

Department of Linguistics: University of Leuven. —2007b. For...to-infinitives as verbal complements in Late Modern and Present-

Day English. Between motivation and change. English Studies 88: 67-94. —Forthcoming a. Nominal gerunds in 16th-century English. The function of the

definite article. Folia Linguistica Historica. —Forthcoming b. Functional motivations in the development of nominal and verbal

gerunds in Middle and Early Modern English. Journal of English Language and Linguistics.

—Forthcoming c. Integrated participle clauses. From adverbial to complement. Word.

—In preparation. Frequency and morphological productivity. A study of idiolectal variation.

De Smet, Hendrik and Hubert Cuyckens. 2004. Pragmatic strengthening and the meaning of complement constructions. The case of like and love with the to-infinitive. Journal of English Linguistics 33: 3-34.

—2007. Diachronic aspects of complementation. Constructions, entrenchment, and the matching-problem. In Christopher M. Cain, and Geoffrey Russom (eds). Studies in the history of the English language III. Managing chaos: strategies for identifying change in English. Berlin: Mouton de Gruyter. 187-214.

De Sutter, Gert. 2005. Rood, groen, corpus! Een taalgebruiksgebaseerde analyse van woordvolgordevariatie in tweeledige werkwoordelijke eindgroepen. Doc-toral thesis, Department of Linguistics: University of Leuven.

De Vogelaer, Gunther, Magda Devos and Johan Van Der Auwera. 2006. Voeg-woordvervoeging. Morfologisch of syntactisch verschijnsel? Taal en Tongval 19: 212-230.

Disterheft, Dorothy. 1981. Remarks on the history of the Indo-European infinitive. Folia Linguistica Historica 2: 3–34.

Dixon, Robert M.W. 1984. The semantic basis of syntactic properties. Berkeley Lin-guistics Society 10: 583-595.

502 – References

—1991. A new approach to English grammar. On semantic principles. Oxford: Ox-ford University Press.

Donner, Morton. 1986. The gerund in Middle English. English Studies 67: 394-400. Duffley, Patrick J. 1992. The English infinitive. London: Longman. —1999. The use of the infinitive and the -ing after verbs denoting the beginning,

middle and end of an event. Folia Linguistica 33: 295-331. —2000. Gerund versus infinitive as complement of transitive verbs in English. The

problems of “tense” and “control”. Journal of English Linguistics 28: 221-248.

—2003. Les conditions de production de l’effet de sens ‘imperfectif’ avec la forme en -ing de l’anglais. Langages 149: 86-99.

—2004. Verbs of liking with the infinitive and the gerund. English Studies 85: 358-380.

Egan, Thomas. 2003. Distance and direction. A usage-based study of infinitive and -ing complement clauses in English. Doctoral thesis, Faculty of arts: Univer-sity of Oslo.

Einenkel, Eugen. 1914. Die Entwickelung des englischen Gerundiums. Anglia 38: 1-76.

Ellegård, Alvar. 1953. The auxiliary do. The establishment and regulation of its use in English. Stockholm: Almqvist & Wiksell.

Emonds, Joseph. 1973. The derived nominals, gerunds, and participles in Chaucer’s English. In Braj B. Kachru and Robert B. Lees (eds). Issues in linguistics. Pa-pers in honor of Henry and Renée Kahane. Urbana: University of Illinois Press. 185-89.

—1976. A transformational approach to English syntax. Root, structure-preserving, and local transformations. New York: Academic Press.

Epstein, Richard. 2002. The definite article, accessibility, and the construction of discourse referents. Cognitive Linguistics 12: 333-378.

Erdmann, Peter. 1993. Die for…to-Konstruktion nach dem Verb want. Zeitschrift für Anglistik und Amerikanistik 41: 124-132.

Fanego, Teresa. 1996a. The development of gerunds as objects of subject-control verbs in English (1400-1760). Diachronica 13: 29-62.

—1996b. On the historical development of English retrospective verbs. Neuphilolo-gische Mitteilungen 97: 71-79.

—1996c. The gerund in Early Modern English. Evidence from the Helsinki Corpus. Folia Linguistica Historica 17: 97-152.

—1998. Developments in argument linking in early Modern English gerund phrases. English Language and Linguistics 2: 87-119.

—2004a. On reanalysis and actualization in syntactic change. The rise and devel-opment of English verbal gerunds. Diachronica 21: 5-55.

—2004b. Some strategies for coding sentential subjects in English. From exaptation to grammaticalization. Studies in Language 28: 321-361.

—2007. Drift and the development of sentential complements in British and Ameri-can English from 1700 to the present day. In Javier Pérez -Guerra, Dolores González-Álvarez, Jorge L. Bueno-Alonso and Esperanza Rama-Martínez (eds). ‘Of varying language and opposing creed’. New insights into Late Modern English. Bern: Peter Lang. 161-236.

References – 503

Fidelholz, James L. 1975. Word frequency and vowel reduction in English. Chicago Linguistic Society 11: 200-213.

Fischer, Olga. 1988. The rise of the for NP to V construction. An explanation. In Graham Nixon and John Honey (eds). An historic tongue. Studies in English linguistics in memory of Barbara Strang. London: Routledge. 67-88.

—1991. The rise of the passive infinitive in English. In Dieter Kastovsky (ed.). Eng-lish historical syntax. Berlin: Mouton de Gruyter. 141-188.

—1992. Syntax. In Norman Blake (ed.). The Cambridge history of the English lan-guage. Vol. 2. 1066-1476. Cambridge: Cambridge University Press. 207-408.

—1995. The distinction between to and bare infinitival complements in Late Middle English. Diachronica 12: 1-30.

—2000. Grammaticalisation: unidirectional, non-reversible? The case of to before the infinitive in English. In Olga Fischer, Anette Rosenbach and Dieter Stein (eds). Pathways of change. Grammaticalization in English. Amsterdam: John Benjamins. 149-169.

—2007. Approaches to morphosyntactic change from a functional and formal per-spective. Oxford: Oxford University Press.

Fischer, Olga and Frederike van der Leek. 1983. The demise of the Old English im-personal construction. Journal of Linguistics 19: 337-368.

García, Erica C. 1999. Diachronic viability of syntactic alternatives. Linguistics 37: 65-125.

Garret, Andrew. 2004. On Kiparsky 2004. http://ist-socrates.berkeley.edu/~garrett /DIGS2004.pdf.

Gee, James Paul. 1977. Comments on Akmaijan. In Peter Culicover, Thomas Wa-sow and Adrian Akmaijan (eds). Formal syntax. New York: Academic Press. 461-481.

Givón, Talmy. 1980. The binding hierarchy and the typology of complements. Stud-ies in Language 4: 333-377.

—1990. Syntax. A functional-typological introduction. Vol. 2. Amsterdam: John Benjamins.

—1992. The grammar of referential coherence as mental processing instructions. Linguistics 30: 5-55.

Gleason, Henry A. 1965. Linguistics and English grammar. New York: Holt, Rinehart and Winston.

Goldberg, Adele E. 1995. Constructions. A construction grammar approach to ar-gument structure. Chicago: University of Chicago Press.

—2006. Constructions at work. The nature of generalization in language. Oxford: Oxford University Press.

Goossens, Louis. 1987. Modal tracks. The case of magan and motan. In Anne-Marie Simon-Vandenbergen (ed.). Studies in honour of René Derolez. Gent: Semi-narie voor Engelse en Oud-Germaanse taalkunde. 216-236.

—2000. Patterns of meaning extension, “parallel chaining”, subjectification, and modal shifts. In Antonio Barcelona (ed.). Metaphor and metonymy at the crossroads. A cognitive perspective. Berlin: Mouton de Gruyter. 149-169.

Gries, Stefan T. 2003. Multifactorial analysis in corpus linguistics. A study of parti-cle placement. New York: Continuum.

504 – References

Gundel, Jeanette K., Nancy Hedberg and Ron Zacharski. 2001. Definite descriptions and cognitive status in English. Why accommodation is unnecessary. English Language and Linguistics 5: 273-295.

Haiman, John. 1980. The iconicity of grammar. Language 56: 515-540. —1983. Iconic and economic motivation. Language 59: 781-819. Halliday, Michael A.K. 1985. An introduction to functional grammar. London: Ed-

ward Arnold. Halmøy, Odile. 1988. Les formes verbales en -ANT dans un roman français du

XIIIème siècle: Le Tristan en prose (tome I). In Kirsten Broch Flemestad, Tove Jacobsen and T. Selmer. Mélanges d’études médiévales, offerts à Helge Nordhal à l’occasion de son soixantième anniversaire. Oslo: Solum Forlag. 64-79.

—2003. Les formes gérondives dans Les .XV. Joies de Mariage et autres textes du XVe siècle. Langages 149: 25-36.

Hantson, André. 1987. Worth + gerund and some related constructions. In Anne-Marie Simon-Vandenbergen (ed.). Studies in honour of René Derolez. Gent: Seminarie voor Engelse en Oud-Germaanse taalkunde. 262-275.

Harris, Alice C. and Lyle Campbell. 1995. Historical syntax in cross-linguistic per-spective. Cambridge: Cambridge University Press.

Haspelmath, Martin. 1989. From purposive to infinitive. A universal path of gram-maticalization. Folia Linguistica Historica 10: 287–310.

—1998. Does grammaticalization need reanalysis? Studies in Language 22: 315-351.

—1999. Optimality and diachronic adaptation. Zeitschrift für Sprachwissenschaft 18: 180-205.

Hawkins, John A. 1978. Definiteness and indefiniteness. A study in reference and grammaticality prediction. London: Croom Helm.

—2004. Efficiency and complexity in grammars. Oxford: Oxford University Press. Hay, Jennifer. 2006. Lexical frequency in morphology. Is everything relative? Lin-

guistics 39: 1041-70. Heine, Bernd. 1993. Auxiliaries. Cognitive forces and grammaticalization. Oxford:

Oxford University Press. Hengeveld, Kees and Gerry Wanders. 1997. On the use of subjunctive and indica-

tive verb-forms in adverbial clauses. In Jadranka Gvozdanović (ed.). Lan-guage change and functional explanations. Berlin: Mouton de Gruyter. 249-272.

Heyvaert, Liesbet. 2003. A cognitive-functional approach to nominalization in Eng-lish. Berlin: Mouton de Gruyter.

—2004. Towards a symbolic typology of -ing nominalizations. Michel Achard and Suzanne Kemmer (eds). Language, culture, and mind. Stanford: CSLI Publi-cations. 493-506.

—2006. On the constructional semantics of gerundive nominalizations. Preprint 258, Department of Linguistics: University of Leuven.

Heyvaert, Liesbet, Hella Rogiers and Nadine Vermeylen. 2005. Pronominal deter-miners in gerundive nominalization. A “case” study. English Studies 86: 71-88.

References – 505

Hogg, Richard. 1988. Snuck. The development of irregular preterite forms. In Gra-ham Nixon and John Honey (eds). An historic tongue. Studies in English lin-guistics in memory of Barbara Strang. London: Routledge. 31-40.

Hollmann, Willem B. 2003. Synchrony and diachrony of English periphrastic cau-satives. A cognitive perspective. Doctoral thesis, Faculty of Arts: University of Manchester.

Hooper, Joan. 1976. Word frequency in lexical diffusion and the source of morpho-phonological change. In William Christie (ed.). Current progress in historical linguistics. Amsterdam: North-Holland. 96-105.

Hopper, Paul. 1987. Emergent grammar. Berkeley Linguistics Society 13:139-157. —1988. Emergent grammar and the a priori grammar postulate. In Deborah Tannen

(ed.). Linguistics in context. Connecting, observation, and understanding. Norwood: Ablex. 117-34.

—1991. On some principles of grammaticization. In Elizabeth C. Traugott and Bernd Heine (eds). Approaches to grammaticalization. Vol. 2. Focus on types of grammatical markers. Amsterdam: John Benjamins. 17-35.

—2001. Hendiadys and auxiliation in English. In Joan Bybee and Michael Noonan (eds). Complex sentences in grammar and discourse. Essays in honor of San-dra A. Thompson. Amsterdam: John Benjamins. 145-175.

Hopper, Paul J. and Sandra A. Thompson. 1980. Transitivity in grammar and dis-course. Language 56: 251-299.

Hopper, Paul J. and Elizabeth Closs Traugott. 1993. Grammaticalization. Cam-bridge: Cambridge University Press.

Houston, Ann. 1989. The English gerund. Syntactic change and discourse function. In Ralph W. Fasold and Deborah Schiffrin (eds). Language change and variation. Amsterdam: John Benjamins. 173-95.

Huber, Magnus. 2007. The Old Bailey proceedings, 1674-1834. Evaluating and an-notating a corpus of 18th and 19th century spoken English. In Anneli Meur-man-Solin and Arja Nurmi (eds). Annotating variation and change ( Studies in Variation, Contacts and Change in English 1). http://www.helsinki.fi/ varieng/journal/volumes/01/huber/.

Huddleston, Rodney and Geoffrey Pullum. 2002. The Cambridge grammar of the English language. Cambridge: Cambridge University Press.

Hudson, Richard. 2007. Word grammar. In Dirk Geeraerts and Hubert Cuyckens (eds). The Oxford handbook of cognitive linguistics. Oxford: Oxford Univer-sity Press. 509-539.

Israel, Michael. 1996. The way constructions grow. In Adele E. Goldberg (ed.). Conceptual structure, discourse and language. Stanford: CSLI. 217-230.

Itkonen, Esa. 2005. Analogy as structure and process. Approaches in linguistics, cognitive psychology and philosophy of science. Amsterdam: John Benja-mins.

Jack, George. 1988. The origins of the English gerund. Nowele 12: 15-75. Jespersen, Otto. 1940. A Modern English grammar on historical principles. Vol. 5.

Syntax. London: George Allan & Undwin. Joseph, Brian. 1983. The synchrony and diachrony of the Balkan infinitive. A study

in areal, general, and historical linguistics. Cambridge: Cambridge Univer-sity Press.

506 – References

—1992. Diachronic explanation. Putting speakers back into the picture. In Garry W. Davis and Gregory K. Iverson (eds). Explanation in historical linguistics. Amsterdam: John Benjamins. 123-144.

Jørgensen, Erik. 1975. Notes on the group ‘for’ + (pro)noun + to-infinitive. English Studies 56: 129-139.

Kager, René. 1999. Optimality theory. Cambridge: Cambridge University Press. Kaltenböck, Gunther. 2003. On the syntactic and semantic status of anticipatory it.

English Language and Linguistics 7: 235-255. Keizer, Evelien. 2004. Postnominal PP complements and modifiers. A cognitive

distinction. English Language and Linguistics 8: 323-350. Keller, Rudi. 1990. Sprachwandel von der unsichtbaren Hand in der Sprache.

Tübingen: Francke. Kempson, Ruth M. and Randolph Quirk. 1971. Controlled activation of latent con-

trast. Language 47: 548-572. Killie, Kristin and Toril Swan. 2006. Pragmatics and the development of adverbial -

ing-clauses (converbs) in English. Paper presented at ICAME 27, Helsinki, 26 May.

Koptjevskaja-Tamm, Maria. 1993. Nominalizations. London: Routledge. Kortmann, Bernd. 1991. Free adjuncts and absolutes in English. Problems of con-

trol and interpretation. London: Routledge. Kranich, Svenja. 2006. The origin of English gerundial constructions. A case of

French influence? In Andrew James Johnston, Ferdinand von Mengden and Stefan Thim (eds). Language and text. Current perspectives on English and German historical linguistics and philology. Heidelberg: Universitätsverlag Winter. 179-195.

Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Lan-guage Variation and Change 1: 199-244.

Kroch, Anthony and Ann Taylor. 2000. Penn-Helsinki Parsed Corpus of Middle English, second edition. http://www.ling.upenn.edu/hist-corpora/PPCME2-RELEASE-2/.

Kroch, Anthony, Beatrice Santorini and Lauren Delfs. 2004. Penn-Helsinki Parsed Corpus of Early Modern English. http://www.ling.upenn.edu/hist-corpora/PPCEME-RELEASE-1/.

Krug, Manfred. 2000. Emerging English modals. A corpus-based study of gram-maticalization. Berlin: Mouton de Gruyter.

—2003. Frequency as a determinant in grammatical variation and change. In Gün-ther Rohdenburg and Britta Mondorf (eds). Determinants of grammatical variation in English. Berlin: Mouton de Gruyter. 7-67.

Kuteva, Tania. 2001. Auxiliation. An enquiry into the nature of grammaticalization. Oxford: Oxford University Press.

Kytö, Merja. 1996. Manual to the diachronic part of the Helsinki Corpus of English Texts. Coding conventions and lists of source texts. University of Helsinki: De-partment of English.

Labov, William. 1989. The child as linguistic historian. Language Variation and Change 1: 85-94.

—1994. Principles of linguistic change. Vol. 1. Internal factors. Oxford: Blackwell. Langacker, Ronald W. 1977. Syntactic reanalysis. In Charles N. Li (ed.). Mecha-

nisms of syntactic change. Austin: University of Texas Press. 57-139.

References – 507

—1987. Foundations of cognitive grammar. Vol. 1. Theoretical prerequisites. Stan-ford, CA: Stanford University Press.

—1991. Foundations of cognitive grammar. Vol. 2. Descriptive application. Stan-ford, CA: Stanford University Press.

—1992. Prepositions as grammatical(izing) elements. Leuvense Bijdragen 81: 287-309.

—2000. A dynamic usage-based model. In Michael Barlow and Suzanne Kemmer (eds). Usage-based models of language. Stanford: CSLI. 1-63.

—2001. Langacker, Ronald W. 2001. Topic, subject, and possessor. In Hanne Gram Simonsen & Rolf Theil Endresen (ed.). A cognitive approach to the verb. Morphological and constructional perspectives. Berlin: Mouton de Gruyter. 11-48.

—2003. Constructions in cognitive grammar. English Linguistics 20: 41-83. Lass, Roger. 1980. On explaining language change. Cambridge: Cambridge Univer-

sity Press. —1998. Historical linguistics and language change. Cambridge: Cambridge Uni-

versity Press. Lees, Robert B. 1966. The grammar of English nominalizations. Bloomington: Indi-

ana University. Lightfoot, David. 1979. Principles of diachronic syntax. Cambridge: Cambridge

University Press. López-Couso, María José. Forthcoming. Adverbial connectives within and beyond

adverbial subordination: The history of lest. In Ursula Lenker & Anneli Meurman-Solin (eds). Clausal connection in the history of English. Amster-dam: John Benjamins.

Los, Bettelou. 1999. Infinitival complementation in Old and Middle English. The Hague: Thesus.

—2005. The rise of the to-infinitive. Oxford: Oxford University Press. MacWhinney, Brian. 2000. Connectionism and language learning. In Michael Bar-

low & Suzanne Kemmer (eds). Usage-based models of language. Stanford: CSLI. 121-150.

Mair, Christian. 1990. Infinitival complement clauses in English. A study of syntax in discourse. Cambridge: Cambridge University Press.

—2002. Three changing patterns of verb complementation in Late Modern English. A real-time study based on matching text corpora. English Language and Linguistics 6: 105-131.

—2003. Gerundial complements after begin and start. Grammatical and sociolin-guistic factors, and how they work against each other. In Günther Rohden-burg and Britta Mondorf (eds). Determinants of grammatical variation in English. Berlin: Mouton de Gruyter. 329-345.

Maling, Joan. 1983. Transitive adjectives. A case of categorial reanalysis. In Frank Heny and Barry Richards (eds). Linguistics categories. Auxiliaries and re-lated puzzles. Vol. 1. Categories. Dordrecht: D. Reidel. 253-289.

Markus, Manfred. 1999. Manuel of ICAMET. Vienna: Braumüller-Verlag. Martin, Robert, and Marc Wilmet. 1980. Syntaxe du moyen français. Manuel du

français du Moyen Age. Bordeaux: Sobodi.

508 – References

McClelland, James L. and David E. Rumelhart. 1981. An interactive activation model of context effects in letter perception: Part 1. An account of basic find-ings. Psychological Review 88: 375-407.

Michaelis, Laura A. 2003. Word meaning, sentence meaning, and syntactic mean-ing. In Hubert Cuyckens, René Dirven and John R. Taylor (eds). Cognitive approaches to lexical semantics. Berlin: Mouton de Gruyter. 163-209.

Miller, Garry D. 2002. Nonfinite structures in theory and change. Oxford: Oxford University Press.

Mindt, Dieter. 2000. An empirical grammar of the English verb. Berlin: Cornelsen. Mustanoja, Tauno F. 1960. A Middle English syntax. Helsinki: Société Néophilolo-

gique. Naro, Anthony J. 1981. The social and structural dimensions of a syntactic change.

Language 57: 63-98. Nevalainen, Terttu and Helena Raumolin-Brunberg. 2003. Historical sociolinguis-

tics. Language change in Tudor and Stuart England. London: Longman. Newmeyer, Frederick J. 1998. Language form and language function. Cambridge,

MA: MIT Press. Ney, James W. 1981. Semantic structures for the syntax of complements and auxil-

iaries in English. The Hague: Mouton Publishers. Noël, Dirk. 2001. The passive matrices of English infinitival complement clauses.

Evidentials on the road to auxiliarihood? Studies in Language 25: 255-296. —2003. Is there semantics in all syntax? The case of accusative and infinitive con-

structions vs. that-clauses. In Günther Rohdenburg and Britta Mondorf (eds). Determinants of grammatical variation in English. Berlin: Mouton de Gruyter. 329-345.

—2006. Diachronic construction grammar vs. grammaticalization theory. Preprint 255, Department of Linguistics: University of Leuven.

Noonan, Michael. 1985. Complementation. In Timothy Shopen (ed.). Language typology and syntactic description. Vol. 2. Complex constructions. Cam-bridge: Cambridge University Press. 42-110.

Nyrop, Kristoffer. 1930. Grammaire historique de la langue française. Vol. 6. Co-penhagen: Gyldendalske Boghandel.

Ogura, Mieko. 1993. The development of periphrastic do in English. A case of lexi-cal diffusion in syntax. Diachronica 10: 51-85.

Ogura, Mieko and William S-Y. Wang. 1996. Snowball effects in lexical diffusion. The development of -s in the third person singular present indicative in Eng-lish. In Derek Britton (ed.). English historical linguistics 1994. Papers from the 8th International Conference on English Historical Linguistics. Amster-dam: John Benjamins. 119-141.

Pawley, Andrew and Frances H. Syder. 1983. Two puzzles for linguistic theory. Nativelike selection and nativelike fluency. In Jack Richards and Richard W. Smith (eds). Language and communication. London: Longman. 191-226.

Philips, Betty S. 1984. Word frequency and the actuation of sound change. Lan-guage 60: 320-342.

—2001. Lexical diffusion, lexical frequency, and lexical analysis. In Joan Bybee and Paul Hopper (eds). Frequency and the emergence of linguistic structure. Amsterdam: John Benjamins. 123-136.

Plank, Frans. 1984. The modals story retold. Studies in Language 8: 305-364.

References – 509

Poutsma, H. 1926. A grammar of Late Modern English. Part 2. The parts of speech. Groningen: P. Noordhoff.

Prince, Alan and Paul Smolensky. 1993. Optimality theory. Constraint interaction in generative grammar. New Brunswick Rutgers university: Center for cogni-tive science.

Pullum, Geoffrey, K. 1982. Syncategorematicity and English infinitival to. Glossa 16: 181-215.

—1991. English nominal gerund phrases as noun phrases with verb-phrase heads. Linguistics 29: 763-799.

Quirk, Randolph. 2004 [1965]. Descriptive statement and serial relationship. In Bas Aarts, David Denison, Evelien Keizer and Gergana Popova (eds). Fuzzy grammar. A reader. Oxford: Oxford University Press. 327-349.

Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik. 1985. A comprehensive grammar of the English language. London: Longman.

Raumolin-Brunberg, Helena. 1991. The noun phrase in Early Sixteenth-Century English. A study based on Sir Thomas More’s writings. Helsinki: Société Néophilologique.

Rissanen, Matti. 1999. Syntax. In Roger Lass (ed.). The Cambridge history of the English language. Vol. 3. 1476-1776. Cambridge: Cambridge University Press. 187-331.

Rohdenburg, Günter. 1995. On the replacement of finite complement clauses by infinitives in English. English Studies 76: 367-388.

—1996. Cognitive complexity and increased grammatical explicitness in English. Cognitive Linguistics 7: 149-182.

—2002. Processing complexity and the variable use of prepositions in English. In Hubert Cuyckens and Günter Radden (eds). Perspectives on prepositions. Tübingen: Niemeyer. 79-100.

—2003. Cognitive complexity and horror aequi as factors determining the use of interrogative clause linkers in English. In Günther Rohdenburg and Britta Mondorf (eds). Determinants of grammatical variation in English. Berlin: Mouton de Gruyter. 205-249.

Rohdenburg, Günther, and Britta Mondorf (eds). 2003. Determinants of grammati-cal variation in English. Berlin: Mouton de Gruyter.

Rosenbaum, Peter S. The grammar of English predicate complement constructions. Cambridge, MA: MIT Press.

Ross, John Robert. 1973. Nouniness. In Osamu Fujimara (ed.). Three dimensions of linguistic research. Tokyo: TEC Company. 137-257.

Rudanko, Juhani. 1984. On some contrasts between infinitival and that complement clauses in English. English Studies 65: 141-161.

—1989. Complementation and case grammar. A syntactic and semantic study of selected patterns of complementation in Present-day English. Albany: State University of New York Press.

—1998. Change and continuity in the English language. Studies on complementa-tion over the past three hundred years. Lanham, MD: University Press of America.

—2000. Corpora and complementation. Tracing sentential complementation pat-terns of nouns, adjectives and verbs over the last three centuries. Lanham, MD: University Press of America.

510 – References

—2003. Comparing alternate complements of object control verbs. Evidence from the Bank of English corpus. In Pepi Leistyna and Charles F. Meyer (eds). Corpus analysis. Language structure and language use. Amsterdam: Rodopi. 273-283.

Rydén, Mats. 1984. The study of eighteenth century English syntax. In Jacek Fisiak (ed.). Historical syntax. Berlin: Mouton Publishers. 509-520.

Schachter, Paul. 1976. A nontransformational account of gerundive nominals in English. Linguistic Inquiry 7: 205-241.

Schlüter, Julia. 2005. Rhythmic grammar. The influence of rhythm on grammatical variation and change in English. Berlin: Mouton de Gruyter.

Severs, Burke. 1958. The tale of Melibeus. In W.F. Bryan, Germaine Dempster and Carleton Brown (eds). Sources and analogues of Chaucer’s Canterbury Tales. New York: Humanities press. 560-614.

Sinclair, John. 1992. Corpus, concordance, collocation. Oxford: Oxford University Press.

Smith, Michael B. and Joyce Escobedo. 2001. The semantics of to-infinitival vs. -ing verb complement constructions in English. In Mary Andronis, Christo-pher Ball, Heidi Elston, and Sylvain Neuvel (eds.). Proceedings from the main session in the Chicago Linguistic Society’s thirty-seventh meeting. Chi-cago: Chicago Linguistic Society. 549-564.

Swan, Toril. 2003. Present participles in the history of English and Norwegian. Ne-uphilologische Mitteilungen. 104: 179-195.

Szmrecsanyi, Benedikt. 2005. Language users as creatures of habit. A corpus-based analysis of persistence in spoken English. Corpus Linguistics and Linguistic Theory 1: 113-150.

Tajima, Matsuji. 1985. The syntactic development of the gerund in Middle English. Tokyo: Nan’un-do.

—1996. The common-/objective-case subject of the gerund in Middle English. Nowele 29: 569-578.

—1999. The compound gerund in Early Modern English. In Sheila Embleton, John E. Joseph and Hans-Josef Niederehe (eds). The emergence of the modern lan-guage sciences. Studies on the transition from historical-comparative to structural linguistics in honour of E.F.K. Koerner. Amsterdam: Benjamins. 265-276.

Thompson, Sandra A. and Anthony Mulac. 1991. A quantitative perspective on the grammaticalization of epistemic parentheticals in English. In Elizabeth C. Traugott and Bernd Heine (eds). Approaches to grammaticalization. Vol. 2. Focus on types of grammatical markers. Amsterdam: John Benjamins. 313-329.

Timberlake, Alan. 1977. Reanalysis and actualization in syntactic change. In Charles N. Li (ed.). Mechanisms of syntactic change. Austin: University of Texas Press. 141-177.

Tomasello, Michael. 2006. Construction grammar for kids. Constructions. http://www.constructions-online.de/articles/specvol1/689.

Traugott, Elizabeth C. 1992. Syntax. In Richard M. Hogg (ed.). The Cambridge his-tory of the English language. Vol. 1. The beginnings to 1066. Cambridge: Cambridge University Press. 168-289.

References – 511

Traugott, Elizabeth C. and Ekkehard König. 1991. The semantics-pragmatics of grammaticalization revisited. In Elizabeth C. Traugott and Bernd Heine (eds). Approaches to grammaticalization. Vol.1. Amsterdam: John Benjamins. 189-218.

Van Der Auwera, Johan and Vladimir A. Plungian. 1998. Modality’s semantic map. Linguistic Typology 2: 79-124.

Van der Wurff, Wim. 1993. Gerunds and their objects in the Modern English period. In Jaap van Marle (ed.). Historical linguistics 1991. Amsterdam: Benjamins. 363-375.

Verspoor, Marjolijn. 1993. To-infinitive and NP XP constructions. A cognitive analysis. Language and cognition 3: 229-238.

Verstraete, Jean-Christophe. 2004. Initial and final position for adverbial clauses in English. The constructional basis of the discursive and syntactic differences. Linguistics 42: 819-853.

Visser, Frederikus Theodorus. 1963-73. An historical syntax of the English lan-guage. Leiden: Brill.

Vosberg, Uwe. 2003. The role of extractions and horror aequi in the evolution of -ing complements in Modern English. In Günter Rohdenburg and Britta Mon-dorf (eds.). Determinants of grammatical variation in English. Berlin: Mou-ton de Gruyter. 305-328.

Wagner, Susanne. 2000. “Depends how long you want for it to take”. For/to clauses in present-day spoken British English. Arbeiten aus Anglistik und Ameri-kanistik 25: 191-211.

Wang, William S-Y. 1969. Competing changes as a cause of residue. Language 45: 9-25.

Warner, Anthony. 1982. Complementation in Middle English and the methodology of hystorical syntax. A study of the Wyclifite Sermons. London: Croom Helm.

—1993. English auxiliaries. Structure and history. Cambridge: Cambridge Univer-sity Press.

Wierzbicka, Anna. 1988. The semantics of grammar. Amsterdam: John Benjamins. Willemse, Peter. 2005. Nominal reference point constructions. Possessive and

esphoric NPs in English. Doctoral thesis. Department of Linguistics: Univer-sity of Leuven.

—2006. Esphoric the N of a(n) N-nominals. Forward bridging to an indefinite refer-ence point. Folia Linguistica 40: 319-64.

Wood, F.T. 1956. Gerund versus infinitive. English Language Teaching 11:11-16. Zipf, George K. 1949. Human behavior and the principle of least effort. An intro-

duction to human ecology. New York: Hafner.

diffusional change in the english system of complementation

Documents