information structure as an independent word ordering factor in old and middle englishinformation...
TRANSCRIPT
1
Information structure as an independent word ordering factor
in Old and Middle English
1 INTRODUCTION
The relationship between word order and principles of information structure has received
increasing attention in English historical linguistics over the past two decades, and
information structure is also a well-established concept in linguistics in general. However,
there is relatively little research on the relationship between information structure and
morphosyntactic weight, a relationship the present paper specifically seeks to address.
In the following sections I will first look at the concept of information structure and how it
has been understood in the literature. Next, the relationship between information structure and
morphosyntactic weight is discussed, before I outline my understanding of what information
value is, ie the pragmatic analysis on the clause element level. Finally, I will present a recent
quantitative study exploring the relationship between an element’s information value, its
weight, and its position in the clause in Old and Middle English. The object of study is the
word order patterns SXV (verb-final) and SVX (verb-medial) in subordinate clauses.1 There
are two main research questions to be answered:
1 SXV: This is the so-called verb-final pattern, in which the finite verb occupies absolute clause-final
position. Consequently, in complex verb phrases the non-finite verb must precede the finite verb. One
or more X elements may precede the subject, and, most importantly, there must be at least one element
separating the subject and the finite verb.
SVX: Clauses belonging to the SVX pattern are first and foremost characterised by having the subject
immediately before the finite verb, which in turn must be followed by one or more X elements. The
subject may be preceded by one or more X elements. If the verb phrase is complex, it has to be
contiguous, ie the non-finite verb must follow the finite verb immediately.
2
1) Does information structure have an effect on word ordering in Old and Middle
English?
2) If so, does that effect change over time as a result of the fixation of the SVX order
characteristic of Present-day English?
2 BACKGROUND
2.1 Word order in Old and Middle English subordinate clauses
The two word order patterns under investigation here, SXV and SVX, are distinguished by the
relative position of the X element and the verb. These patterns are central in the history of Old
English and Middle English subordinate clauses. The literature on Old English word order
contains numerous references to OE subordinate clause word order, many of which point to a
strong tendency for subordinate clauses to display verb-final order (SXV), ie final position for
the finite verb.2 However, the variation between verb-final and other orders, especially SVX,
is frequently commented on as well, 3
and both Pintzuk (1999) and Heggelund (2009)
demonstrate that there is competition between the two orders. During the course of the Middle
English period, the SVX order gradually becomes predominant, and by the year 1500 the
SXV pattern has nearly disappeared from prose texts (Swieczkowski 1962: 75, Kohonen
1978: 90). We thus go from a situation in Old English where objects, adverbials and
predicatives are found frequently both before and after the verb to a situation where
2 Cf eg, Quirk & Wrenn (1957), Mitchell (1964, 1985), Vennemann (1974, 1984), Lightfoot (1991,
2006), Traugott (1992), Fischer & Van der Wurff (2006).
3 Cf eg Barrett (1952), Kohonen (1978), Allen (1980), Bean (1983), Davis & Bernhardt (2002), Taylor
and Pintzuk (2012).
3
postverbal position is more or less syntactically fixed for these elements. Early Middle
English (1100 – 1300) may be seen as particularly interesting because it is a transitional stage
in which the fixation of SVX syntax is in progress, but by no means completed.
2.1 Information structure
According to Gundel (2003: 124), ‘information structure is a cover label for a number of
different, though partly overlapping, concepts that have often been conflated in the literature’.
She argues that although the concepts normally involved all relate to the distinction between
given and new information in one way or another, one must distinguish between two senses of
givennessnewness, viz referential on the one hand and relational on the other. The former is
what the present paper is primarily concerned with, and it is defined in the following way by
Gundel (2003: 124):
a relation between a linguistic expression and a corresponding non-linguistic
(conceptual) entity in (a model of) the speaker/hearer’s mind, the discourse, or
some real or possible world, depending on where the referents or
corresponding meanings of these linguistic expressions are assumed to reside.
Relational givennessnewness, on the other hand, has to do with the division of a sentence
into two parts, one representing what the sentence is about, and one adding to what the
sentence is about. The parts in this two-way division have received a myriad of names in the
literature, including the pairs theme/rheme, topic/focus, presupposition/focus and
topic/comment. These are extensively covered and also variously defined; to mention but a
few studies, Firbas (1966, 1992), Halliday (1967) and Daneš (1974) discuss theme/rheme,
4
while topic and/or focus are treated in eg Chomsky (1971), Jackendoff (1972), various articles
in Li (1976), Comrie (1981), Gundel (1985), Quirk et al. (1985), Lambrecht (1994), Rizzi
(1997), Birner & Ward (1998) and Erteschik-Shir (2007), among others.
The idea that pragmatic factors may influence word order has existed for a long time.
Weil (1978 [1887]: 29) distinguishes between a sentence’s point of departure, ie the speaker’s
and hearer’s common ground, and its goal of discourse, where new, important information is
presented. Weil compares the movement from the first to the second part of the sentence with
the movement of ideas in the mind of the speaker. Another early statement on discourse and
word order is found in Behaghel’s (1932: 5) second law, which predicts that old concepts
precede new ones. This tendency in language to arrange a clause so that given information
precedes information which is new is often referred to as the information principle (Halliday
1967: 205, Breivik 1989: 31, Biber et al. 1999: 896), or the principle of end focus (eg Quirk et
al. 1985: 1357). Various definitions of given and new information are found in the literature,4
but the one used by Clark & Haviland (1977: 4) nicely captures two important aspects of
givenness/newness. They define given information as ‘information the speaker considers
given – information he believes the listener already knows and accepts as true’. In opposition
to this stands ‘information the speaker considers new – information he believes the listener
does not yet know’. We see that givenness is decided by the speaker, but that there may a
difference between what the speaker considers given/new and what the speaker thinks the
listener considers given/new.
4 Some of the most central ones include Halliday (1967), Chafe (1976, 1994), Kohonen (1978), Prince
(1981, 1992) and Firbas (1992).
5
Two scholars that have particular relevance to the analysis applied later in this paper are
Chafe (1976, 1994) and Prince (1992). Chafe (1976: 30), following Halliday (1967), makes a
binary distinction between given and new information, and defines these notions as follows:
Given (or old) information is that knowledge which the speaker assumes to be
in the consciousness of the addressee at the time of the utterance. So-called
new information is what the speaker assumes he is introducing into the
addressee’s consciousness by what he says.
Importantly, it is up to the speaker to decide what should be regarded as given or new
information. The decision must be made on the basis of either the linguistic or extralinguistic
context, the latter involving the possibility of some shared knowledge which is not explicitly
expressed in the context (1976: 31). In his 1994 book, Chafe extends his definition of
givenness into a three-way distinction between given, accessible and new information (1994:
72). The term accessible is used about information which is on the periphery of the
consciousness, but which still has not left the consciousness and is thus recoverable.
In Prince’s (1992) discussion of information structure, only referents evoked by noun
phrases are considered, but her observations are arguably applicable on a more general level,
too. She does not find the simple dichotomy between given and new information entirely
satisfactory, and makes a distinction between context-dependency and hearer-dependency,
which captures the fact that context-independent information is not necessarily new to the
hearer/writer. Thus, an item that has not been mentioned in the previous context is labelled
discourse-new, but may nevertheless be hearer-old, eg if the referent is a person or some other
item of which the hearer already has knowledge. In other words, an entity’s discourse-new
6
status does not reveal anything about its hearer-status. On the other hand, a discourse-old
entity is by necessity hearer-old, ‘since hearers are expected to remember what they have been
told’ (1992: 303).5
For the present purpose, Prince’s notion of ‘inferrability’ is especially interesting. A piece
of information is ‘assumed to be inferrable by the hearer on the basis of some trigger entity,
itself discourse-old, in combination with some belief the hearer is assumed to have which says
that entities like the trigger have associated with them entities like the Inferrable’ (1992: 307).
Prince uses the following example to illustrate this:
(1) He passed by the Bastille and the door was painted purple
The definite NP the door appears to be both discourse-new and hearer-new, but it may be
classified as inferrable if the hearer is assumed to infer the existence of the door from the
basis that buildings are generally associated with a particular door. It needs to be emphasised
that inferrable, as is the case with hearer-old and hearer-new, is a category which rests upon
the speaker’s assumptions about the hearer’s knowledge (cf Clark & Haviland and Chafe
above) and reasoning ability, not the hearer’s actual knowledge and reasoning ability. Thus,
the door is treated ‘as though it were already known to the hearer’ (1992: 305).
Mention must also be made of Taylor and Pintzuk’s (2012) work on information
structure and object position in Old English. They code objects on a five-point scale, ie new,
new-anchored, accessible-anchored, accessible and given. However, in the statistics provided
the first two are both treated as new and the last three as given.
5 The distinction between context-dependency and hearer-dependency is further developed in Birner
(2006).
7
2.2 Information structure in Old and Middle English
Some studies on information structure and word order in Old and Middle English deserve
mention here. Firbas (Firbas 1966, 1992) discusses the word order of OE and Present-day
English (PDE) in relation to the theory of FSP (Functional Sentence Perspective), which is
concerned with language from a communicative point of view. The principle of FSP predicts
that ‘sentence elements follow each other according to the amount (degree) of communicative
dynamism (CD) they convey, starting with the lowest and gradually passing on to the highest’
(1966: 240). In turn, communicative dynamism is described as ‘the extent to which the
sentence element contributes to the development of the communication, to which, as it were,
it “pushes” the communication forward (1966: 240).
According to Firbas, an important determinant of an element’s degree of CD is the context
(1992: 10). The contextual factor in FSP has to do with ‘the retrievability/irretrievability from
the immediately relevant context’ (1992: 21), ie whether an element is contextually dependent
or not. A contextually dependent element carries a lower degree of CD than an element which
is contextually independent.
As far as Old English is concerned, Firbas finds its word order to be relatively free (1992:
127ff.). Therefore, Firbas claims, OE is very susceptible to the principle of FSP, more so than
PDE, where word order to a larger extent is determined by grammatical principles. In other
words, the clause elements in OE are typically arranged in accordance with the basic
distribution of CD. Even though later research has demonstrated that OE word order is not as
free as Firbas may suggest, but rather conforms to a limited set of patterns (eg Pintzuk 1999,
Davis & Bernhardt 2002), Firbas’ main point should still be valid: The more syntactic
freedom, the greater the influence of FSP.
8
Bech’s (2001) doctoral dissertation investigates word order in Old and Middle English
main clauses from a pragmatic perspective, more specifically, whether and to what extent
pragmatic factors played a role in the typological shift that English underwent. Bech’s point
of departure is that Old English had some kind of V2 constraint, but that pragmatic factors
could override this constraint.
Bech uses the term information value (IV), which is based primarily on the theories of
Firbas and Chafe. She makes a binary distinction between elements with low and high
information value, and in order to determine the IV of an element, Bech relies on Firbas’ two
primary factors: the context and the semantic structure.
Bech’s results show that although there are a large number of V2 clauses in OE, the word
order in this period is quite heterogeneous. This is taken to be a result of competition between
syntactic and pragmatic factors. Evidence is found particularly in the XSV pattern, which is
pragmatically motivated since the majority of subjects have low information value. In the
XVS pattern, on the other hand, the IV of the subjects is more variable, signalling that this
word order is motivated both by syntactic and pragmatic principles.
Bech’s ME data reveal a much more homogeneous situation, with SVX and XSV as the
dominant patterns. Bech argues that XSV word order has become the unmarked, productive
pattern, governed increasingly by syntactic constraints. V2 clauses are still relatively frequent,
something which is explained by pragmatic constraints: ‘XVS order becomes used in
environments where pragmatic pressure is so strong as to force the subject into post-verbal
position, ie, primarily in existential sentences’ (2001: 195). Since existential sentences
typically have the function of introducing new subjects with high information value, XVS
order makes such sentences adhere to the information principle.
9
Bech’s final conclusion is that pragmatic factors did have an impact on the shift from V2
to V3 (2001: 197). Her study demonstrates that the study of information structure can shed
light on the word order development of English.
3 WEIGHT VERSUS INFORMATION STRUCTURE
It is well known that the weight of an element may influence that element’s position in the
clause, as pointed out already by Behaghel (1909). The principle of end weight may be
defined as ‘the tendency for long and complex elements to be placed towards the end of the
clause’ (Biber et al. 1999: 898). The implication is then that short and structurally simple
elements will tend to be placed early in the clause. The definition by Biber et al. suggests that
both length and complexity are related to weight, and length, in terms of the number of
syllables or words, is fairly unproblematic. Grammatical complexity, on the other hand, is in
itself a hazy concept which is hard to define (see eg Dahl 2004). On any account, it is
important to keep in mind that weight is a relative and not an absolute concept (Wasow &
Arnold 2003: 121), which it is practically impossible to define in absolute terms; a pronoun-
headed noun phrase tends to be lighter than a full NP, while a short version of the latter is
lighter than a long version. Weight is probably best defined separately in different phrasal
categories, such as noun phrases, adjective phrases and adverb phrases, and that is the
approach chosen in this paper (see section 5.2).
We have seen that various aspects of information structure are extensively treated in the
literature. However, the relationship between weight and information structure has not
received a lot of attention,6 despite the fact that morphosyntactic and discourse-functional
6 In the words of Wasow & Arnold (2003), ‘[i]t is surprising that, despite the voluminous literatures on
both weight and information structure, those two literatures are nearly disjoint’.
10
factors are to a certain extent interdependent and hard to keep apart. For instance, the well-
known tendency for object pronouns to occur preverbally in Old English may either be due to
the fact that they are light elements, or because they are contextually given elements, or to a
combination of the two.7 The same can be said about full noun phrases, even though the
correlation between information value and weight for full NPs is not as straightforward as for
pronouns; full NPs may of course be both contextually given and new, and the relative weight
may also vary considerably. The information principle and the principle of end weight tend to
work in conjunction in the formation of clauses, together with syntactic constraints, as well as
other factors like prosody and rhythm. The potential coextensiveness of the two principles no
doubt complicates matters. According to Wasow & Arnold (2003: 129), it may be difficult to
tell whether the two are distinct factors, or whether ‘one of them only looks like a causal
factor because of its high correlation with the other one’. Their psycholinguistic
experimentation combined with a corpus study concludes that ‘neither the length nor the
discourse status (...) could account for constituent ordering as well as the two combined’. A
similar result is reported by Siewierska (1993: 263) in a study of Polish and the relationship
between short/long and given/new. She concludes that one cannot really say which of the two
principles is stronger.
Hawkins (1991, 1994) does not agree that the information principle has a place in word
order studies. He posits that elements are ordered for maximally easy recognition of syntactic
structures. Among other things, Hawkins looks at modern English ordering of verb–object–
7 Rybarkiewicz (1977: 89) discusses the relationship between and coextensiveness of what he labels
‘the heaviness principle’ and Firbas’ Functional Sentence Perspective with respect to pronominal
placement in OE, and concludes that the position of pronouns is equally well accounted for by both
principles.
11
particle in order to find out which of the dimensions short/long (morphosyntactic) and
given/new (pragmatic) most affect word ordering. According to Hawkins, ‘pragmatics
appears to play no role whatsoever. The [pragmatic] theories proposed add nothing to the
syntactically based predictions of EIC [=Early Immediate Constituents]’ (1994: 240–241).
Recent studies have demonstrated that Hawkins’ dismissal of pragmatics cannot be
upheld. Bresnan et al. (2007) perform a multi-variate analysis to find out which factors
influence the choice between the double object structure (Susan gave the children toys) and
the prepositional dative construction (Susan gave toys to the children). They show that
discourse accessibility (based on Prince 1981), together with factors such as animacy,
definiteness and pronominality, is not reducible to syntactic complexity and has an
independent effect on the dative construction choice (2007: 82).
Taylor & Pintzuk (2012: 62) find that there is some relationship between weight and
information status, but demonstrate that the two are not directly correlated, and that weight
and information status must be considered independent factors. They find that object position
in Old English subordinate clauses is influenced by both information structural factors and
morphosyntactic weight. Interestingly, they also posit that the change from OV to VO in
English is accompanied by a reduction in the effect information structure has on object
positioning.
In this connection, Gries (2003: 149f) makes an interesting observation: even in cases
where morphosyntax seems to have the strongest effect on word ordering, discourse-
functional variables must play at least an indirect role, since an element’s givenness/newness
influences morphosyntax. Morphosyntactic and discourse-functional variables, as well as
phonological and semantic ones are strongly interrelated, Gries argues, and one can therefore
never dismiss discourse-functional factors altogether.
12
4 THE CONCEPT OF INFORMATION VALUE
Information structure as discussed above concerns how the clause is structured information-
wise, eg from given/known to new/unknown information. In order to establish the information
structure of a given clause, we need something that deals with the status of the constituents
within that clause. Information value may be described as an operationalization of information
structure, in the same way that the information concepts given and new have traditionally been
used.
The method of analysis is inspired by Bech (2001: 145ff). As was mentioned in section
2.3, Bech operates with the concepts low information value (low IV) and high information
value (high IV).8 These are related to, but not co-extensive with the more traditional ones,
given and new information. For example, an element which is used contrastively may be
contextually given, but still have high information value. In addition, certain elements which
cannot be classified according to the given/new distinction may be assigned low information
value in the pragmatic analysis (Bech 2001: 152). Examples of such elements are the
anticipatory subjects hit ‘it’ and þær ‘there’, and the OE indefinite pronoun man ‘one’.
Recall from section 2.1 that a distinction may be made between referential and relational
givennessnewness (Gundel 2003: 125). To the extent that the two concepts are possible to
keep apart, the present study deals with the referential aspect. In other words, the clauses in
the corpus are not divided into two main parts (theme/rheme, topic/comment etc); instead it is
individual clause elements that play a pivotal role. Thus, subjects, objects, predicatives and
adverbials are assigned either low or high IV. It must be emphasised that the referential aspect
8 The term information value is also used on several occasions by Quirk et al. (1985: 1357).
13
of givennessnewness cannot be fully detached from the relational aspect, since the
topic/theme will tend to contain mostly given elements while the comment/rheme/focus will
contain mostly new elements.9
As stated above, the context is an important determinant in the assignment of information
value. However, a major problem is constituted by what to regard as relevant context. Firbas
talks about ‘the immediately relevant context’ (1992: 21), but does not define precisely what
that means, but states that ‘it is normal for the retrievability span to be very short (...) due to
the continuous influx of new irretrievable information’ (1992: 2930). In this respect Firbas is
supported by Chafe (1994: 79): ‘the number of different referents that can be active at the
same time is very small (...) any referent, unless it is refreshed, will quickly leave the active
state’.
Keeping in mind that givenness is a status decided upon by the speaker, one might ask
whether studying information structure is worthwhile at all, since direct access to the
speaker’s own assessment is almost never possible. Fortunately, in most cases information
value can be decided upon relatively uncontroversially, based on the preceding context. The
most problematic elements are those that are not strictly contextually given, but nevertheless
have the potential to constitute some shared knowledge, cf Chafe’s (1976: 31) discussion of
extralinguistic context. This ‘shared knowledge’ is extremely hard to define, as is Kohonen’s
(1978: 67) concept of ‘pragmatically known information’. Kohonen mentions unique referents
like ‘sun’, ‘moon’ and ‘heaven’, as well as central biblical characters as examples of such
information, and classifies these as given. Can these elements, although obviously familiar to
9 Quirk et al. (1985: 1362) distinguish between the ‘contextually established’ concepts given and new,
and the ‘linguistically defined’ theme (in terms of position) and focus (in terms of prosody).
14
the addressee, really be expected to be in his or her consciousness at all times? Or more
precisely, is it probable that the speaker expects these elements to be in the consciousness of
the addressee at all times? Again, of course, we cannot look into the mind of the speaker. As
Geluykens (1992: 10) rightly points out: ‘not everything the hearer “knows” can be assumed
to be in his consciousness’. It is thus necessary to distinguish between elements that are
presumed to be familiar to the addressee and elements that can be expected to be in his/her
consciousness. Consequently, elements that are potentially known or shared on the basis of
world knowledge or the wider context have generally been assigned high information value in
the present analysis. That is, even referents such as ‘the earth’, ‘the sun’, ‘God’, ‘Christ’ and
‘the Lord’ are not assumed to be in the consciousness of the addressee at their first mention,
and need to be made accessible first. However, central elements like these are likely to remain
in the addressee’s consciousness for a longer time than peripheral elements before they leave
the mind and become contextually independent, and need only be repeated once in a while to
retain a low information value.
No attempt has been made to establish firm criteria for how long an element remains
context-dependent once mentioned. The reason is, as discussed above, that the decision
depends on a number of things, most importantly the nature of the element in question, how
central it is to the subject matter, how frequently it has been mentioned previously, and what
type of material intervenes between two references to the same element. On the whole, the
present method of analysis can be said to be fairly restrictive with regard to the span of an
element’s givenness. That is, keeping in mind the human brain’s limited capacity for short-
term memory (Jonides et al. 2008), an element is retrievable for a very short period of time
after its first mention. The retrievability span will naturally increase somewhat when an
15
element is mentioned several times in the text, and the span is also likely to be longer for
central elements than for peripheral ones (see eg Poirier & Saint-Aubin 1996).
Psychologically speaking, it seems fair to say that the concepts of givenness and
information value are intrinsically connected with the speaker’s own assessment of the
addressee’s consciousness, as suggested by both Prince (1992) and Chafe (1994).
Nevertheless, most linguists, and certainly all historical linguists, have no access to the
speaker’s mind and must rely entirely on texts. Therefore, in the analysis of information value
carried out here, context and semantic content will be the main deciding factors. Elements that
are context-dependent have been assigned low IV, whereas elements not mentioned in the
preceding context have high IV.10
Context-dependence includes anaphora, deixis, paraphrase,
synonymy and certain cases of antonymy. In addition, and unlike Bech (2001), I have taken
Prince’s (1992) idea of inferrable elements into account. It has not been kept separate,
however, because inferrable elements seem to have much in common with context-dependent
elements. Moreover, a discrete category would be comparatively small and of little use
statistically. Thus, elements regarded as inferrable have been given low information value. In
conclusion, the assignment of information value is the outcome of both contextual and
semantic considerations.
The subsequent sections describe in more detail how the various types of clause element
have been analysed pragmatically. It should be noted that the description is general and covers
all element types that may be analysed for information value, even though the present paper is
only concerned with X elements in the SXV and SVX patterns.
10
For examples of the practical analysis of information value, see section 4.3 below.
16
4.1 Objects and predicatives
Objects and predicatives are analysed as having low IV or high IV. If an element is regarded
as contextually dependent, it is given low IV, whereas a contextually independent element is
assigned high IV.
The analysed objects and predicatives are realised by noun phrases or adjective phrases.
For the most part, the assignment of IV to an element can be done solely on the basis of the
preceding context, but cases of synonymy, paraphrase, contrast and inference also occur in the
data.
Since pronouns are generally contextually dependent, they have been left out of the
present investigation. The exception is NPs with a postmodified pronoun head, which are
classified as full NPs. Moreover, clause elements constituted by a clause have been left out of
the pragmatic analysis, since they typically convey a mix of given and new information.
4.2 Adverbials
Adverbials are also analysed as either low or high IV elements. If the adverbial is realised by
an adverb, contextual dependence determines the IV. However, adverbs are relatively rarely
co-referential with previous context in the same way that full NPs and adjective phrases may
be, apart from cases of deixis (here, there, now). Context-dependence is therefore not as
relevant for adverbs as for other types of elements; rather, a decision must be made in each
case as to whether the semantic content of the adverb contributes with important information
or not. Adverbs which serve to link the clause to the preceding context are considered to have
low information value, such as then, afterwards, however, therefore and yet. OE ær ‘before’ is
a frequent item that also appears to serve a linking function in many cases. Interrogative
17
adverbs (how, when, where, etc) introducing dependent questions are also assigned low IV.
Adverbs that are clearly context-independent are classified as high IV elements, the most
typical examples being manner adverbs, for instance terribly, rightly and thoughtfully.
Temporal adverbs such as before, immediately, always, often and soon, as well as negative
and restrictive elements like never, not and rarely are not usually dependent or independent of
context in the same sense as manner adverbs, and seem to vary with respect to their
informational load. Of course, both manner adverbs and temporal/negative/restrictive adverbs
may be repeated in the context, in which case they are assigned low IV.
Adverbials realised by prepositional phrases are not easy to analyse pragmatically since
they consist of two parts, a preposition and a preposition complement, most typically a noun
phrase. Kohonen (1978: 138ff.) and Bech (2001: 156) propose that only the noun phrase be
taken into consideration in a pragmatic analysis, and this is the method chosen here. Thus, if
the preposition complement has low IV, the whole prepositional phrase is regarded as a low
IV element, and vice versa. The exception is when a preposition is clearly used to express
contrast, in which case the prepositional phrase has high IV. Adverbials in the form of clauses
have been disregarded in the analysis of information value.
4.3 The practical analysis of information value
The following short segment from the OE text Cura Pastoralis (29,23–31,6), together with
the comments accompanying it, may serve to illustrate the current method of analysis.11
The
relevant clause elements are underlined.12
11
An intersubjectivity test performed by Kristin Bech shows a high level of correspondence between
Bech’s and my own analysis, and suggests that the method of analysis is sufficiently objective. From a
total of 2,056 clause elements from two texts, 3.1% (63/2,056) were analysed completely differently,
18
a. Oft ðonne se hirde gæð on frecne wegas, sio hiord ðe
often when the shepherd goes on dangerous ways, the flock which
b. unwærre bið, gehrist. Be suelcum hirdum cwæð se witga: Ge
unwary is, falls. of such shepherds spoke the prophet: Ye
c. fortrædon Godes sceapa gærs & ge gedrefdon hiora wæter
trod down God’s sheep’s grass and you defiled their water
d. mid iowrum fotum, ðeah ge hit ær undrefed druncen. Sua
with your feet, though you it previously undefiled drank. thus
e. ða lareowas hi drincað suiðe hluter wæter
the teachers they drink very pure water
a. Oft. The adverb is analysed as having high IV here, but it is one of several short adverbs of
time and restriction whose IV is hard to determine. For the placement of such adverbs (eg oft
‘often’, ær ‘before’, ærest ‘first’, eac ‘also’ and syþþan ‘later’), it is possible that weight or
simply word class membership is more important than IV.
se hirde. This is another way of saying lareowas ‘teachers, priests’, which is found in the
immediately preceding context. Such cases of synonymy are given low IV.
ie one analyst assigned low IV and the other high IV. 3.1% (66/2,056) were given either low or high
IV by one analyst, but marked ‘uncertain’ by the other. The percentage of elements analysed
identically by the two analysts thus equals 93.7%, or 1,927/2,056.
12 For illustrative purposes, all subjects and X elements are analysed here. In section 5 the analysis is
restricted to X elements in SXV and SVX clauses.
19
on frecne wegas. As mentioned above, prepositional phrases are analysed according to the IV
of the noun phrase complement. Frecne wegas has not been mentioned before, and does not
seem to be retrievable from the context in any way, thus it has high IV.
sio hiord ðe unwærre bið. Even though sio hiord represents a good example of an element
inferrable from the context (in this case based on the mention of se hirde), the postmodifying
relative clause ðe unwærre bið renders the NP subject as a whole context-independent with
high information value. The postmodifier is in turn analysed separately, with a low IV relative
pronoun subject, and a context-independent high IV subject predicative.
b. Be suelcum hirdum. The NP suelcum hirdum is context-dependent, and the whole PP is
thus given low IV.
se witga. The reference here is not clear, but could be to one of the Old Testament prophets.
In any case the NP appears to be context-independent and have high IV.
Ge. The pronoun does not have anaphoric reference, but must be said to have low IV based on
the situational context. The same is the case with the other two instances of ge.
c. Godes sceapa gærs. Even though Godes sceapa is inferrable from the mention of hirde and
hiord, the phrase as a whole is assigned high IV due to the context-independence of the head
noun gærs ‘grass’.
hiora wæter. This NP has not been mentioned before, and thus has high IV.
d. mid iowrum fotum. This element is assumed to be inferrable from fortrædon earlier in the
sentence, and has low IV.
hit. The object pronoun has anaphoric reference to hiora wæter, hence low IV.
20
ær. In quite a few cases, the common adverb ær seems to link to the preceding context. That
link does not seem to be entirely clear here, and this instance of ær has thus not been analysed
with regard to IV.
undrefed. This adjective is deemed to have lexical cross-reference with gedrefdon ‘defiled’,
and is therefore assigned low IV. One might argue that it is used contrastively, in which case
it would have high IV.
Sua. This is a typical linking adverb, with low IV.
e. ða lareowas. This NP is taken to be synonymous with the ‘shepherds’ mentioned in line 2,
and it is therefore assigned low IV.
hi. Here hi is a resumptive pronoun with low IV.
suiðe hluter wæter. This object appears to be a paraphrase of the reference to the undefiled
water above: low IV.
5 EVIDENCE FROM OLD AND MIDDLE ENGLISH
5.1 The material
My corpus comprises 2,466 subordinate clauses in the patterns SXV and SVX. The data are
drawn from a total of 12 texts from early Old English (eOE), late Old English (lOE) and early
Middle English (eME), and the texts are as follows:
eOE: The Old English Orosius (Or), Alfred’s Cura Pastoralis (CP).
lOE: Ælfric’s Catholic Homilies, Second Series (ÆCHom II), The Blickling Homilies
(BlHom), The Old English Apollonius of Tyre (ApT), The C-text of the Anglo-Saxon
Chronicle (ChronC).
21
eME: The Peterborough Chronicle Continuations (PC), Kentish Homilies (Kentish),
Katherine, Margaret, Vices and Virtues (VV), Ancrene Wisse (AW).13
5.2 The results
The comparison in this section between SXV and SVX order has three main objectives.
First, to establish whether there is a different distribution of IV for the X elements between
the two patterns, and second, to find out whether any variation in information structure
between the two patterns simply reflects the morphosyntactic weight of X elements, or
whether information value has an influence on word order independent of element weight.
Thirdly, it is interesting to find out whether the effect of information value is stable or
whether it changes over time.
A distinction has been made between short (1 or 2 words) and long (3 words or more)
NPs, while adverb phrases and adjective phrases have been subdivided into simple and
complex phrases. A simple AdvP or AdjP consists of a single word, ie the head alone,
whereas complex phrases have pre- and/or postmodifiers. Constituents consisting of two or
more coordinated heads are also considered heavy and consequently classified as complex.
Note that all relevant X elements assigned low or high information value are included in
the tables below. Thus, the fact that many clauses contain two or more X elements means that
the total number of elements exceeds the total number of clauses.
13
A wide selection eME texts are used to minimise the influence of dialect differences. It should be
noted that in the present eME data, the intertextual variation between texts from the same dialect
appears to be greater than the variation between dialects.
22
Table 1 presents the results for the SXV pattern in early OE.
low IV high IV Total
n % n % n %
short NP 54 43 71 57 125 100
long NP 9 35 17 65 26 100
simple AdvP 92 77 28 23 120 100
complex AdvP 7 28 18 72 25 100
PP 69 48 74 52 143 100
simple AdjP 9 53 8 47 17 100
complex AdjP 0 0 11 100 11 100
Total 240 51 227 49 467 100
Table 1
The IV of X elements according to type in eOE SXV clauses
The bottom row in table 1 shows that there is a fairly even distribution between low and
high IV X elements. There is no statistically significant difference in the distribution of IV
between short and long NPs.14
However, there are systematic differences within adjective and
adverb phrases, according to whether the head is modified or not: phrases constituted by
heads alone have low IV more often than complex phrases. For AdvPs the low IV rates are
77% (simple) and 28% (complex), for AdjPs 53% and 0%.15
Excluding PPs, which have not
been analysed for length, we see that there is a clear preference for the short versions of NPs,
AdvPs and AdjPs (262/324, 81%).
14
Chi-square contingency table test, short vs long NPs: χ2 = 0.35, p = .55, Cramér’s V = .07.
15 Contingency, simple vs complex AdvPs: χ
2 = 20.43, p < .0001, Cramér’s V = .4; simple vs complex
AdjPs: Fisher Exact Test, two-tailed, p = .004.
23
Simple adjective phrases with low IV (3) and high IV (4), as well as a complex adjective
phrase with high IV (5) and a long NP with low IV (6) are presented below.
(3) Biscepe gedafnað ðæt he sie tælleas. Ðærbufan is geteald hwelc he beon bishop-D befits that he is blameless. Besides is told what he be
sceal, gif he untælwierðe bið shall, if he blameless is
‘It befits a bishop that he is blameless. Besides it is told what he shall be, if he is
blameless.’ (CP 53.10)
(4) 7 norðeweard, he cwæð, þær hit smalost wære, þæt hit mihte beon
and northward, he said, where it narrowest was, that it might be
þreora mila brad to þæm more three miles broad to the waste
‘And northward, he said, where it is narrowest, that it might be three miles broad to the
waste.’(Or 15.28)
(5) Eac is to geðencanne ðæt on ða tiid ðe se biscephad swa gehiered also is to reflect that at the time when the bishop-office so valued
wæs, sua hwelc swa hine underfeng, he underfeng martyrdom was whoever it accepted, he accepted martyrdom
‘One should also reflect that at the time when the bishop office was so valued,
whoever accepted it also accepted martyrdom.’ (CP 53.17)
(6) þonne cymeð se man se þæt swift[ost]e hors hafað then comes the man who the quickest horse has
‘Then comes the man who has the quickest horse.’ (Or 17.22)
The results for late OE results are found in table 2.
24
low IV high IV Total
n % n % n %
short NP 72 58 52 42 124 100
long NP 12 44 15 56 27 100
simple AdvP 94 75 31 25 125 100
complex AdvP 7 30 16 70 23 100
PP 83 52 77 48 160 100
simple AdjP 6 26 17 74 23 100
complex AdjP 0 0 3 100 3 100
Total 274 56 211 44 485 100
Table 2
The IV of X elements according to type in lOE SXV clauses
We see that the overall proportion of low IV elements in this pattern increases from early
to late OE (51% vs 56%), and with that in mind no particular element type stands out in
comparison with early OE. As in eOE, short and long NPs do not differ statistically in the
distribution of IV.16
Complex AdvPs, on the other hand, have a greater likelihood for high IV
than their simple counterparts,17
while there are too few complex AdjP to carry out a
meaningful comparison for that phrase type. The proportion of short elements is 84%
(272/325), quite similar to that found in eOE.
In table 3 below the eME results are presented.
16
Contingency, short vs long NPs: χ2 = 1.16, p = .28, Cramér’s V = .11.
17 Contingency, simple vs complex AdvPs: χ
2 = 16.0, p < .0001, Cramér’s V = .35.
25
low IV high IV Total
n % n % n %
short NP 13 81 3 19 16 100
long NP 7 58 5 42 12 100
simple AdvP 40 89 5 11 45 100
complex AdvP 2 29 5 71 7 100
PP 18 86 3 14 21 100
simple AdjP 4 44 5 56 9 100
complex AdjP 0 0 3 100 3 100
Total 84 74 29 26 113 100
Table 3
The IV of X elements according to type in eME SXV clauses
Again there is a general increase in the proportion of low IV elements (56% vs 74%),18
and the increase is most marked for prepositional phrases, from 52% to 86%. SXV clauses
thus arguably become increasingly pragmatically governed, in the sense that elements with
low IV occur more often and elements with high IV more rarely over time. Interestingly, the
short versions of NPs, AdvPs and AdjPs occur in 76% of the cases (70/92), which is less
frequently than in eOE (81%) and lOE (84%). Thus, while the strengthening of SVX syntax in
Middle English seems to restrict the pragmatic properties of preverbal X elements, that does
not appear to hold for morphosyntax, since there is no increase (but actually a small, non-
significant decrease) in the proportion of short elements.19
18
Contingency, low IV and high IV, lOE vs eME: χ2 = 11.41, p = .0007, Cramér’s V = .14.
19 Goodness-of-fit, short elements in eME vs lOE: χ
2 = 0.41, df = 1, p = .52.
26
Table 4 summarises the results in tables 1–3.
eOE lOE eME
n % n % n %
low 240 51 274 56 84 74 high 227 49 211 44 29 26
Total 467 100 485 100 113 100
Table 4
The IV of X elements in SXV clauses
Let us now turn to SVX clauses for a comparison with the above findings. The results for
eOE are presented in table 5.
low IV high IV Total
n % n % n %
short NP 24 25 72 75 96 100
long NP 4 6 63 94 67 100
simple AdvP 7 58 5 42 12 100
complex AdvP 0 0 6 100 6 100
PP 24 21 88 79 112 100
simple AdjP 2 15 11 85 13 100
complex AdjP 0 0 27 100 27 100
Total 61 18 272 82 333 100
Table 5
The IV of X elements according to type in eOE SVX clauses
27
Within this pattern, long noun phrases are more likely to have high IV than short noun
phrases, and the same applies to complex adverb phrases in comparison with simple ones.20
No such effect can be found for adjective phrases, however. Compared to the eOE SXV
pattern (cf table 1), SVX clauses considerably more often occur with a high IV X element in
short NPs (75% vs 57%), long NPs (94% vs 65%) and PPs (79% vs 52%).21
In other words,
the same type of element is more likely to have high IV when occurring postverbally than
preverbally. The difference for simple adjective phrases is not statistically significant.22
The
short/long ratio for NPs, AdvPs and AdjPs is 121 vs 100, which means that 55% of these
phrases are short. That is considerably less than in OE SXV clauses, and indicates that long
elements are more likely to be placed postverbally than preverbally, in accordance with the
principle of end weight.
20
Contingency, short vs long NPs: χ2 = 8.75, p = .003, Cramér’s V = 0.25; two-tailed Fisher Exact
Test for complex vs simple AdvPs: p = .04.
21 Contingency, eOE SXV vs eOE SVX, short NPs: χ
2 = 7.1, p = .008, Cramér’s V = .19; long NPs: χ
2
= 10.51, p = .001, Cramér’s V = .37; PPs: χ2 = 18.36, p < .0001, Cramér’s V = .28.
22 Fisher Exact Test, two-tailed, p = .06.
28
The late OE findings are presented in table 6.
low IV high IV Total
n % n % n %
short NP 21 43 28 57 49 100
long NP 10 17 50 83 60 100
simple AdvP 5 71 2 29 7 100
complex AdvP 0 0 4 100 4 100
PP 34 27 91 73 125 100
simple AdjP 0 0 5 100 5 100
complex AdjP 0 0 12 100 12 100
Total 70 27 192 73 262 100
Table 6
The IV of X elements according to type in lOE SVX clauses
As in eOE, long NPs have high IV more often than short NPs.23
There is an increase of
low IV elements from 18% to 27%. Only the three categories of short and long NPs and PPs
are sufficiently sizeable to validate statistical testing. Diachronically (eOE vs lOE), these
contain an increasing proportion of low IV elements, in line with the general increase.
However, the difference for PPs is not significant.24
The synchronic comparison with the SXV
pattern shows a much higher rate in SVX clauses of high IV long NPs (83% vs 56%) and PPs
23
Contingency, short vs long NPs: χ2 = 7.85, p = .005, Cramér’s V = .29.
24 Contingency, eOE vs lOE SVX, NPs: χ2
= 10.52, p = .001, Cramér’s V = .20; PPs: χ2 = 0.78,
p = .38, Cramér’s V = .07.
29
(73% vs 48%), while the higher rate for short NPs is not statistically significant.25
In this
period there are more long (76) than short (61) elements overall.
Next, SVX clauses in eME are treated in table 7.
low IV high IV Total
n % n % n %
short NP 59 50 60 50 119 100
long NP 20 19 83 81 103 100
simple AdvP 58 70 25 30 83 100
complex AdvP 0 0 15 100 15 100
PP 127 51 124 49 251 100
simple AdjP 8 44 10 56 18 100
complex AdjP 5 13 35 88 40 101
Total 277 44 352 53 629 100
Table 7
The IV of X elements according to type in eME SVX clauses
Interestingly, there is another increase in the proportion of low IV elements, from 27% in
lOE to 44% in eME.26
Yet again, short and long NPs differ with respect to information value,
the latter being more likely to have high IV. In this period the same applies to both AdvPs and
AdjPs.27
A comparison with table 3 also shows that simple AdvPs, short/long NPs and PPs
25
Contingency, lOE SVX vs lOE SXV, short NPs: χ2 = 2.68, p = .1, Cramér’s V = .14; long NPs: χ
2 =
6.21, p < .0001, Cramér’s V = .3; PPs: χ2 = 18.36, p < .0001, Cramér’s V = .28.
26 Contingency, low vs high IV, lOE vs eME: χ
2 = 22.61, p < .0001.
27 Contingency, short vs long NPs: χ
2 = 20.62, p < .0001, Cramér’s V = .31; simple vs complex AdvPs:
χ2 = 22.87, p < .0001, Cramér’s V = .51; simple vs complex AdjPs: χ
2 = 5.56, p = .02, Cramér’s V =
.35.
30
are more likely to have high IV in SVX than in SXV clauses.28
In eME the short/long ratio is
220 vs 158, or 58% vs 42%.
6 DISCUSSION AND CONCLUDING REMARKS
The data confirm the expected close relationship between morphosyntactic weight and
information value, though with variations across patterns and phrase types. In SXV clauses
NPs do not have significant IV variation along the short/long dimension, while AdvPs and
AdjPs do. The SVX pattern, on the other hand, displays such variation for NPs in all three
periods, but only in two periods for AdvPs and one period for AdjPs. On the whole, though,
we can conclude that the long versions of all three phrase types are more likely to have high
IV than the short versions.
Synchronically, the two word order patterns behave differently: within the same time
period, a certain element type, such as a long NP, has greater likelihood of having high IV
when it occurs after the verb than when it occurs before the verb. This tendency applies in
particular to NPs and PPs, and to a lesser degree to AdjPs and AdvPs, where low numbers
tend to make comparison difficult. All in all, there are strong indications that postverbal
elements are more likely to have high information value than preverbal ones with the same
relative weight. Therefore, information structure can play a role in the ordering of clause
elements independent of morphosyntactic weight.
28
Contingency, eME SVX vs eME SXV, simple AdvPs: χ2 = 4.86, p = .03, Cramér’s V = .21; short
NPs: χ2 = 21.92, p < .0001, Cramér’s V = .33; long NPs: χ
2 = 7.02, p = .008, Cramér’s V = .28; PPs: χ
2
= 8.24, p = .004, Cramér’s V = .19.
31
Diachronically, we have seen that SXV clauses become more restricted to low IV X
elements in eME, but that this is not simply a result of morphosyntax, since there is no
increase in the proportion of short and simple elements. At the same time, SVX clauses go
from having predominantly high IV X elements to allowing a greater mix of information
value for the X elements. This corroborates the findings of Taylor & Pintzuk (2012), but for
all clause elements and phrase types and not only nominal objects. Thus, the gradual syntactic
development towards a fixation of SVX order in English (cf. section 2.1) goes hand in hand
with a gradual change in the information structuring of clauses with both SVX and SXV
order.
The analysis of the current data provides evidence, both synchronic and diachronic, that
information structure influenced the choice of word order in both OE and eME.
32
REFERENCES
Allen, Cynthia L. 1980. Topics in diachronic English syntax. New York: Garland.
Barrett, Charles Robin. 1952. Studies in the word-order of Ælfric's Catholic Homilies and
Lives of the Saints. Cambridge: W. Heffer & Sons.
Bean, Marian C. 1983. The development of word order patterns in Old English. London:
Croom Helm.
Bech, Kristin. 2001. Word order patterns in Old and Middle English: a syntactic and
pragmatic study. Doctoral dissertation, University of Bergen.
Behaghel, Otto. 1909. Beziehungen zwischen Umfang und Reihenfolge von Satzgliedern.
Indogermanische Forschungen 25, 110–42.
Behaghel, Otto. 1932. Deutsche Syntax: eine geschichtliche Darstellung, vol. 4: Wortstellung.
Heidelberg: Carl Winter.
Biber, Douglas, et al. 1999. Longman grammar of spoken and written English. Harlow:
Longman.
Birner, Betty J. 2006. Inferential relations and noncanonical word order. In Betty J. Birner &
Gregory L. Ward (eds.), Drawing the Boundaries of Meaning. Neo-Gricean Studies in
Pragmatics and Semantics in Honor of Laurence R. Horn, 31–51.
Amsterdam/Philadelphia: John Benjamins.
Birner, Betty J. & Gregory L. Ward. 1998. Information status and noncanonical word order
in English (Studies in language companion series 40). Amsterdam: John Benjamins.
Breivik, Leiv Egil. 1989. On the causes of syntactic change in English. In Ernst Håkon Jahr &
Leiv Egil Breivik (eds.), Language change: Contributions to the study of its causes, 29–
70. Berlin: Mouton de Gryuter.
Bresnan, Joan, Anna Cueni, Tatiana Nikitina & R. Harald Baayen. 2007. Predicting the dative
alternation. In Gerlof Bouma, Irene Krämer & Joost Zwartz (eds.), Cognitive Foundations
of Interpretation. Amsterdam: Royal Netherlands Academy of Science.
Chafe, Wallace L. 1976. Givenness, contrastiveness, definiteness, subjects, topics and point of
view. In Charles N. Li (ed.), Subject and topic, 25–55. New York: Academic Press.
Chafe, Wallace L. 1994. Discourse, consciousness, and time: the flow and displacement of
conscious experience in speaking and writing. Chicago: University of Chicago Press.
Chomsky, Noam. 1971. Deep structure, surface structure and semantic interpretation. In D.
Steinberg & L. Jakobovits (eds.), Semantics, an interdisciplinary reader in linguistics,
philosophy and psychology, 183–216. Cambridge: Cambridge University Press.
Clark, Herbert H. & Susan E. Haviland. 1977. Comprehension and the given-new contract. In
Roy O. Freedle (ed.), Discourse production and comprehension, 1–40. Norwood, N. J.:
Ablex.
Comrie, Bernard. 1981. Language universals and linguistic typology. Oxford: Basil
Blackwell.
Dahl, Östen. 2004. The growth and maintenance of linguistic complexity (Studies in Language
Companion Series 71). Amsterdam: John Benjamins.
Daneš, F[rantisek]. 1974. Functional Sentence Perspective and the organization of the text. In
F. Daneš (ed.), Papers on Functional Sentence Perspective, 106–28. Prague: Academia.
Davis, Graeme & Karl Bernhardt. 2002. Syntax of West Germanic: the syntax of Old English
and Old High German. Göppingen: Kümmerle Verlag.
33
Erteschik-Shir, Nomi. 2007. Information structure: the syntax–discourse interface (Oxford
surveys in syntax and morphology). Oxford: Oxford University Press.
Firbas, Jan. 1966. Non-thematic subjects in contemporary English. Travaux linguistiques de
Prague 2, 239–56.
Firbas, Jan. 1992. Functional sentence perspective in written and spoken communication
(Studies in English language). Cambridge: Cambridge University Press.
Fischer, Olga & Wim Van der Wurff. 2006. Syntax. In Richard Hogg & David Denison
(eds.), A history of the English language, 109–98. Cambridge: Cambridge University
Press.
Geluykens, Ronald. 1992. From discourse process to grammatical construction: on left-
dislocation in English (Studies in discourse and grammar). Amsterdam: John Benjamins.
Gries, Stefan Th. 2003. Multifactorial analysis in corpus linguistics: a study of particle
placement (Open linguistics series). New York: Continuum.
Gundel, Jeanette K. 1985. Shared knowledge and topicality. Journal of Pragmatics 9, 83–107.
Gundel, Jeanette K. 2003. Information structure and referential givenness/newness: how much
belongs in the grammar? In Stefan Müller (ed.), Proceedings of the HPSG–2003
Conference, Michigan State University, East Lansing, 122–42. Stanford University: CSLI
Publications.
Halliday, M. A. K. 1967. Notes on transitivity and theme in English, part 2. Journal of
Linguistics 3, 199–244.
Hawkins, John A. 1991. Syntactic weight versus information structure in word order
variation. In Joachim Jacobs (ed.), Informationsstruktur und Grammatik/Information
Structure and Grammar, 196–220. Opladen: Westdeutscher Verlag.
Hawkins, John A. 1994. A performance theory of order and constituents. Cambridge:
Cambridge University Press.
Heggelund, Øystein Imerslund. 2009. Word order in Old English and Middle English
subordinate clauses. Doctoral dissertation, University of Bergen.
Jackendoff, Ray. 1972. Semantic interpretation in generative grammar. Cambridge, Mass.:
MIT Press.
Jonides, John, et al. 2008. The mind and brain of short-term memory. Annual Review of
Psychology 59, 193–224.
Kohonen, Viljo. 1978. On the development of English word order in religious prose around
1000 and 1200 A.D.: a quantitative study of word order in context (Meddelanden från
Stiftelsens för Åbo Akademi Forskningsinstitut 38). Åbo: Research Institute of the Åbo
Akademi Foundation.
Lambrecht, Knud. 1994. Information structure and sentence form. Cambridge: Cambridge
University Press.
Li, Charles N. (ed.) 1976. Subject and topic. New York: Academic Press.
Lightfoot, David. 1991. How to set parameters: arguments from language change.
Cambridge, MA: MIT Press.
Lightfoot, David. 2006. How new languages emerge. Cambridge: Cambridge University
Press.
Mitchell, Bruce. 1964. Syntax and word order in The Peterborough Chronicle.
Neuphilologische Mitteilungen 65, 113–44.
Mitchell, Bruce. 1985. Old English syntax, 2 vols. Oxford: Clarendon Press.
Pintzuk, Susan. 1999. Phrase structures in competition: variation and change in Old English
word order (Outstanding dissertations in linguistics). New York: Garland.
34
Poirier, Marie & Jean Saint-Aubin. 1996. Immediate serial recall, word frequency, item
identity and item position. Canadian Journal of Experimental Psychology 50, 408–12.
Prince, Ellen. 1981. Toward a taxonomy of given/new information. In P. Cole (ed.), Radical
pragmatics, 223–54. New York: Academic Press.
Prince, Ellen. 1992. The ZPG letter: subjects, definiteness, and information-status. In S.
Thompson & W. Mann (eds.), Discourse description: diverse analyses of a fund raising
text, 295–325. Philadelphia: John Benjamins.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A
comprehensive grammar of the English language. London: Longman.
Quirk, Randolph & C. L. Wrenn. 1957. An Old English grammar, 2nd edn. London:
Routledge.
Rizzi, Luigi. 1997. The fine structure of the left periphery. In L. Haegeman (ed.), Elements of
grammar, 281–337. Dordrecht: Kluwer.
Rybarkiewicz, W. 1977. The word order in OE prose and the functional sentence perspective.
Studia Anglica Ponsnaniensia 9, 87–95.
Siewierska, Anna. 1993. Syntactic weight vs. information structure and word order variation
in Polish. Journal of Linguistics 29, 233–65.
Swieczkowski, Walerian. 1962. Word order patterning in Middle English: a quantitative
study based on Piers Plowman and Middle English sermons (Janua linguarum, Series
minor 19). 'S-Gravenhage: Mouton.
Taylor, Ann & Susan Pintzuk. 2012. The effect of information structure on object position in
Old English: a pilot study. In Anneli Meurman-Solin, María José López-Couso &
Bettelou Los (eds.), Information structure and syntactic change in the history of English,
47–65. Oxford: Oxford University Press.
Traugott, Elizabeth Closs. 1992. Syntax. In Richard Hogg (ed.), The Cambridge encyclopedia
of the English language, vol. I, 168–289. Cambridge: Cambridge University Press.
Vennemann, Theo. 1974. Topics, subjects and word order: from SXV to SVX via TVX. In C.
M. Jones & J. M. Anderson (eds.), Historical linguistics I, vol. 1, 339–76. Amsterdam:
North Holland.
Vennemann, Theo. 1984. Verb-second, verb-late, and the brace construction in Germanic: a
discussion. In Jacek Fisiak (ed.), Historical syntax, 627–36. Berlin: Mouton de Gruyter.
Wasow, Thomas & Jennifer Arnold. 2003. Post-verbal constituent ordering in English. In
Günther Rohdenburg & Britta Mondorf (eds.), Determinants of grammatical variation in
English (Topics in English Linguistics 43), 119–54. Berlin: Mouton de Gruyter.
Weil, Henri. 1978 [1887]. The order of words in the ancient languages compared with that of
the modern languages (Amsterdam classics in linguistics 14). Amsterdam: John
Benjamins.