linguistics-debut volume 2 number 2

8/2/2019 Linguistics-Debut Volume 2 Number 2

1/51

Dbut: the Undergraduate Journal ofLanguages, Linguistics and AreaStudies

Volume 2, Number 2 (2011)

Editor: John Canning

Published by LLAS Centre for languages, linguistics and area studies with thesupport of the Higher Education Academy.

ISSN: 2044-7256


2/51


3/51

51

www.llas.ac.uk

Contents

Editorial: A new look for

Dbut

John Canning 53

The Reduced Relative Clause:

A Misnomer?

Tom Stanton

55

On the narratives' credibility

concerning the disease and the

fatal end in Frulein Else and

Effi Briest

Matthias Stark 67

A modular account of language

change in Alzheimers disease

Claire Cordella 75

Searching for the Source of

Right-Headed Definiteness in

the Balkan Sprachbund

Sean A Guynes 89

Instructions for Authors 99


4/51

www.llas.ac.uk

Editorial: A new look for DbutJohn Canning

LLAS Centre for languages, linguistics and area studies, Faculty ofHumanities, University of Southampton, Southampton, UK.

[email protected]

In my last editorial I briefly wrote about some of the changes which were taking place at the Subject

Centre for Languages, Linguistics and Area Studies. On 1 August 2011 we relaunched asLLAS Centre

for languages, linguistics and area studies, regular readers ofDbutwill notice that the journal now

has a new look.

We continue our work with the UKs Higher Education Academy and

this edition of Dbut forms part of the HEAs and LLAS ongoing

commitment to enhancing the learning experience of students in UK

higher education. I am delighted with the four high quality articles

which appear in this edition of Dbut with contributions from the UK,

the USA and Austria.

My last editorial put forward a number of options for Dbut which might present alternatives to the

review process. Thank you to all who offered their comments. The feedback I received indicated that

students value having their work reviewed by academics outside their university and there was no

evidence of any appetite for change.
mailto:[email protected]:[email protected]:[email protected]


5/51

54

You can also check out our website for studentsstudying languages at university

www.studyinglanguages.ac.uk. It contains advice

on study skills for learning languages, advice on

working and studying abroad and articles about

preparing for life after leaving university.

I hope you enjoy this issue of Dbut.

The team at LLAS Centre for languages, linguistics and area studies.
http://www.studyinglanguages.ac.uk/http://www.studyinglanguages.ac.uk/http://www.studyinglanguages.ac.uk/


6/51

55

Dbut: the Undergraduate Journal of Languages, Linguistics and Area Studies 2.2 (2011)

www.llas.ac.uk

The Reduced Relative Clause: A

Misnomer?Tom Stanton

University College London, London, UK

[email protected]

Tom Stanton is a postgraduate student currently studying for a MRes in Speech, Language

and Cognition at University College London. He gained a first class honours degree in

Linguistics at the same institution, where he developed his interest in syntax and semantics.He is planning to go on to study for a Doctorate in Linguistics, and from there to do

postdoctoral research.

Abstract

In this paper I argue that the term reduced relative is a misnomer. 1 I will begin by discussing

the origins of the term, and the analysis that it suggests. Then I will look at the shortcomings

of that analysis and describe an alternative analysis. The alternative analysis will require that

there be an absence of the types of structure that are essential to relative clauses. I will alsodescribe what may and may not be found within reduced relative clauses. Based upon all of

this I will conclude that the term reduced relative clause is indeed a misnomer.

Introduction

The term reduced relative is used to refer to a particular syntactic construction which has

many parallels with relative clauses. Both constructions contain a noun which is modified in

some sense. The similarities between reduced relatives and relative clauses are

demonstrated in (1) and (2).

(1) a. The horse raced past the barn fell.

b. The horse that was raced past the barn fell.

(2) a. [[[The horse] raced past the barn] fell].

b. [[[The horse] that was raced past the barn] fell].

1A reviewer informs me that in The Cambridge Grammar of the English Language (2002), Huddleston and

Pullum do not use the term. While it has been recognized by some to be a misnomer the term still remains in use,

and I shall argue that the term suggests an analysis which is not sustainable.
mailto:[email protected]:[email protected]:[email protected]


7/51

56 Stanton: The Reduced Relative ClauseRoss (1972) provides an analysis in which reduced relative clauses are argued to have the

same syntax as regular relative clauses. This analysis survives in the term reduced relative

itself. I will argue that the term reduced relative is a misnomer; that reduced relative clauses

have a different syntax to regular relative clauses; and that the similarities between the two

constructions are superficial.

Here I shall outline the analysis presented in Ross (1972). The sentences in (3) are argued

to be derived from the sentences in (4). The motivation given for this claim is sparse, relying

on semantic equivalency and what is most plausible (The most plausible sources for [(3a)

and (3b)] would be [(4a) and (4b)]. Ross 1972). Although this option may seem the most

plausible it is not the only option worthy of investigation.

(3) a. Men sharpening knives were leering at us.

b. Men sharpening knives leer at us.

(4) a. Men who were sharpening knives were leering at us.

b. Men who sharpen knives leer at us.

Two rules are given to describe the transformative processes shown above. The process that

begins with (4a) and results in (3a) is referred to as WHIZ Deletion, which states that when

a relative complementizer and verb be appear next to each other in an embedded clause

they may both be deleted. WHIZ Deletion is taken to be a process similar to

Complementizer Deletion as demonstrated in (5). The deleted segments remain in the

syntactic representation but are not pronounced as part of the phonetic realization.

(5) a. I know that I put it somewhere safe.

b. I know that I put it somewhere safe.

The process that derives (3b) from (4b) is given the name Stuff-ing. When the subject of an

embedded clause is relativized the embedded subject is deleted and the verb takes the form

that incorporates the present participle (its -ing form).

(6) Stuffing

X [NP NP [S NP V Y ]S ]NP Z

SD: 1 2 3 4SC: 1 3#ing 4

(SD = structural description, SC = structural change)

Both WHIZ Deletion and Stuff-ing are surface constraints having no internal motivation, nor

resulting from the mechanisms of the syntax but rather being applied externally. Such

qualities are undesirable in any framework that has the aim of being minimal.

The Problem with Rosss Account

A number of issues with Rosss account of reduced relatives were raised by Hudson (1973).In that account the assumption that reduced relatives are derived from regular (unreduced)


8/51

57


www.llas.ac.uk

relatives was challenged and Hudson described a way in which reduced relatives behaved

differently to their unreduced equivalents. I shall now discuss these issues, and with ideas

from studies on the calculation of tense in syntax propose that there is a more satisfactory

alternative.

(7) a. Books published since 1950 are mostly paperbacks.

b. The people living here are dead.

These two sentences are the seeds from which this discussion will grow. Taking the first,

Hudson argues that this sentence is a well formed sentence which has a reduced relative

clause within it. If reduced relatives are in fact derived from unreduced equivalents, as Ross

claims, then there will be an unreduced equivalent for (7a), and indeed there is. Sentence

(8a) may be said to be an accurate unreduced equivalent to (7a).

(8) a. Books that have been published since 1950 are mostly paperbacks.b. Books having been published since 1950 are mostly paperbacks.

c. *Books that are/were published since 1950 are mostly paperbacks.

However, the two rules given by Ross for deriving reduced relatives cannot be applied to (8a)

in order to derive (7a). WHIZ Deletion cannot apply and Stuff-ing gives us (8b) rather than

(7a). If we were to apply WHIZ Deletion backwards to (7a), taking the ouput of the procedure

and calculating the input, we would get an ill formed structure (8c). It is clear that either we

need more (or more complicated) rules for deriving reduced relatives from their unreduced

counterparts or that there are some reduced relatives that are underived, which would

encourage us to question whether all reduced relatives are underived.

Now we will return to the second seed. According to WHIZ Deletion and Stuff-ing, sentence

(7b) has a number of equivalent unreduced relatives. Some of these are presented in (9),

limiting the tenses to present and past for the sake of simplicity.

(9) a. The people who are living here are dead.

b. The people who were living here are dead.

c. The people who live here are dead.

d. The people who lived here are dead.

The first thing to note about these possible unreduced equivalents to the reduced relative

being considered is that they are not equivalent to one another. This in itself is not a problem

as it is not the relation between the different possible unreduced equivalents that is important

but the relation between the reduced relative and the unreduced counterparts. It should be

possible for (7b) to be interpreted as having any of the tenses of (9a-d).

This is not the case however, suggesting that the interpretation of tense for a reduced

relative is more constrained. It is clear that (7b) only allows a contradictory interpretation,

thereby demonstrating that (9b) and (9d) are not possible equivalents of it. Hudson argues


9/51

58 Stanton: The Reduced Relative Clausethat reduced relatives behave differently to unreduced relatives when it comes to the matter

of the calculation of tense. Reduced relatives have tense that, while not marked, is

interpreted either deictically (that is, relative to the moment of speaking) or derivatively

(that is, relative to the tense of the matrix verb). In (7b) this means that the living being done

by the people discussed is either in process while the sentence is being uttered, or while

the people are being dead. Since are dead can only be interpreted relative to the momentof speaking, the people must also be being dead while the sentence is being uttered. This

is what rules out (9b) and (9d) as possible interpretations. It is assumed that present

participles convey an uncompleted action and so are interpreted as occurring simultaneously

with the time referred to by the matrix verb.

Unreduced relatives can only be interpreted deictically. This is why sentences such as (9b)

and (9d) are perfectly acceptable. Relative to the moment of speaking the people must

have beenliving while simultaneously being dead. These two events do not contradict and

so the interpretation is fine. The point, however, is that this interpretation is not available toreduced relatives.

Hudson describes these differences but does not offer an explanation for them. However, an

explanation may be found in work done on Double Access Reading (DAR) in Italian. In her

recent work Giorgi (2010) the notions of relative to the moment of speaking and relative to

the tense of the matrix verb are discussed and associated with certain syntactic elements.

Giorgi argues for the existence of a syntactic element within the C- layer of a sentences

syntactic structure that encodes the speakers temporal location and a syntactic element in

the I-layer that encodes the temporal location of the subject of Verb Phrase (VP). Once theevent described by the VP has been checked against the temporal position of the subject of

that event it is checked against the temporal position of the speaker too. It is the encoding of

the speakers temporal coordinates within the syntax that allows for the tense of the

embedded verb to be interpreted relative to the moment of speaking. Normally the event

described by the VP of an embedded clause cannot escape to a level higher than its C-layer

and so it can only be checked against the temporal coordinates of the embedded subject and

of the speaker. However, some structures allow it to transcend its C-layer and be checked

against the temporal coordinates of the matrix subject, resulting in Double Access Readings.

Giorgi argues convincingly that this is not an epiphenomenon at the interface between syntaxand some interpretive module but is actually an independent element within the syntax itself.

The point of note here is that it has been argued that there is a connection between the

sequencing of syntactic elements and the computation of tense, with the C-layer being

relevant for interpreting tense as relative to the moment of speaking. The mechanics that

Giorgi employs do not need to be assumed for the argument at hand, simply that a

connection has been seen.

It is possible to see a viable alternative to Rosss account of reduced relative clauses. As

Hudson has demonstrated, there exist a number of unreduced relatives that do not have areduced counterpart, and that even those that do, do not display the same syntactic


10/51

59


www.llas.ac.uk

behaviour. Reduced relative clauses may have the event described by them interpreted

relative to the tense of the matrix verb or relative to the moment of speaking. However,

unreduced relatives may only have the events they describe interpreted relative to the

moment of speaking. Ross argues that the difference between reduced relative clauses and

their unreduced counterparts is that the complementizer (which resides in the C-layer) andthe auxiliary be (which resides in the I-layer) go unpronounced. Giorgi provides the

evidence that there is a connection between these two layers and the tense effects that

Hudson notes.

It is possible to argue for there being no I-layer or C-layer in reduced relatives. Straight away

the absence of complementizers and auxiliaries are readily explained, there is nowhere in

the structure for them to reside. Not only that, but the absence of all auxiliary types such as

the have been seen in (8) above is explained. Also, by disposing of the I-layer and C-layer

we also dispose of the temporal reference points associated with them. For the eventdescribed in VP to have its tense interpreted it must be checked against the I-layer or the C-

layer of the matrix clause. The tense of reduced relative clauses is interpreted relative to the

tense of the matrix verb or relative to the moment of speaking. This also accounts for the

difference between reduced and unreduced counterparts. Unreduced relatives do have an

independent I-layer and C-layer and as such do not have access to the temporal reference of

the matrix verb and can only have their event interpreted relative to the moment of speaking.

Reduced Relatives (A Description)

In this part of the paper I will discuss what may and may not be found within reduced

relatives. I will begin by looking at the various predicate types that may or may not be found

within reduced relatives but I will move onto more abstract elements such as semantic

effects, etc. Sag (1997) in the framework of Head-Driven Phrase Structure Grammar lists the

various predicate types that may be allowed to appear in reduced relatives. The list permits

VPs with present participles, VPs with past participles, VPs with passive participles,

prepositional phrases (PPs) and adjective phrases (APs) to appear in reduced relatives and

bans finite VPs, infinitival VPs, or predicative Noun Phrases (NP). The sentences in (10)

have reduced relatives that contain those elements listed as acceptable and the sentences in

(11) contain those elements listed as unacceptable.

(10) a. The person standing on my foot is not observant

b. The prophet descended from heaven gave a sermon

c. The orders given by the General must be obeyed

d. People in Rome do as the Romans do

e. The committee happy with the proposal agreed to it


11/51

60 Stanton: The Reduced Relative Clause(11) a. *The person stands on my foot is not observant

b. *The prophet descend from heaven gave a sermon

c. *The orders instructions must be made

Keeping in mind that the C-layer and the I-layer are taken to be absent from reduced

relatives then the inability for finite and to some extent infinitival VPs is to expected. Finite

verbs require the presence of an I-layer in order to be tensed and bare-infinitives require the

same structural elements. The absence of the I-layer in reduced relatives does not allow

finite verbs and to-infinitives to appear in reduced relatives. That predicative NPs may not

appear in reduced relatives is a point I will return to.

Turning now to the predicate types that are described as being permitted to reside in reduced

relatives, I highlight a problem with the description. While it seems that VPs with present

participles, predicative PPs and predicative APs can always appear in reduced relatives, for

VPs with past participles this is not always the case. The nature of the verb within the past

participle VPs effects its acceptability within a reduced relative. The sentences in (12)

demonstrate this effect.

(12) a. *A man given his friend a gift is a generous man (transitive)

b. *A man run regularly shows he has self-discipline (unergative)

c. ?A vegetable grown quickly has an evolutionary advantage (unaccusative)

d. A man given a gift by his friend is grateful (passive)

While it is clear that transitive (12a) and unergative (12b) verbs are not acceptable within a

reduced relative, the status of (12c) is a little murky. The reading we are looking for is one

where the vegetable grows but is not grown. The sense we are after is the unaccusative

sense, not the passive sense and, to my ear at least, (12c) requires there be some agent

causing the growth of the vegetable. This is the passive reading which is shown in (12d) to

be perfectly acceptable within reduced relatives. The similarity between the unaccusative

past participles and passives makes it hard to mark a distinction between the two in reduced

relatives.

(13) a. The door opened slowly and revealed that no-one had opened it

b. *The door opened slowly revealed that no-one had opened it

The pair of sentences in (13) are an attempt to tease apart the two types. The verb 'opened'

is ambiguous between unaccusative and passive. If 'opened' in (13a) is interpreted as

passive then the sentence is a clear contradiction, but if it is interpreted as unaccusative, with

the door opening of its own accord, no such contradiction arises. That (13b) can only have acontradictory reading I take to be evidence that 'opened' does not have access to the


12/51

61


www.llas.ac.uk

unaccusative reading and that unaccusative verbs with past participles may not reside within

reduced relative clauses.

This result is problematic however, as it has long been argued that unaccusative past

participle verbs can exist in reduced relatives. For instance, Marvin (2000) in a discussion ofpast participles in reduced relatives assumes that unaccusatives are permitted in English

reduced relatives and gives (14) as evidence for this.

(14) ?The leaf fallen from the tree is red

However, the acceptability of (14) is dubious to me. At the best I would give (14) a question

mark status. I certainly do not feel that all unaccusative past participle verbs are perfectly

acceptable within reduced relatives. The sentences in (15) have a range of pure

unaccusatives in reduced relatives and it seems clear that a number of them are definitely

unacceptable.

(15) a. *The guests appeared at the party are late

b. ?The parcel arrived at the office is in the inbox

c. *The man died in the hospital will be missed

d. *The countries existed for a long time have more history

e. *The events happened regularly can be easily predicted

That some unaccusatives are more unacceptable than others here may suggest that the

group of pure unaccusative verbs may be divided into two groups, those that have definite

judgements of unacceptability and those that have indefinite ones.

Horvath & Siloni (2005) argue for a difference between two types of unaccusative verbs to be

recognised based upon their syntactic distributions. They argue that pure unaccusatives may

be divided into 'decausative' and 'underived' unaccusatives. They come to this proposal

based upon a cross-linguistic investigation into adjectival passives. Adjectival passives areadjectives that are formed using the passive form of certain verbs. They note that

'decausative' unaccusatives can form adjectival passives despite having no verbal passive

equivalent, while 'underived' unaccusatives cannot form adjectival passives. The group of

'decausative' unaccusatives is also referred to as the arriveclass, counting the verbs 'arrive'

and 'fall' among its number, while the group of 'underived' unaccusatives is named the

appearclass.

Based upon this I claim that the question mark judgements for (14) and (15b) are due to the

predicates not being verbal but rather adjectival. As mentioned above adjectives may residewithin reduced relatives without a problem. The outright unacceptability of the rest of the


13/51

62 Stanton: The Reduced Relative Clausesentences in (16) is assumed to be due to those verbs being 'underived' unaccusatives and

their having no adjective passive form to be ambiguous with. Following this I assume that

past participle unaccusative verbs may not reside inside reduced relatives, that those that

appear to do so are 'decausative' unaccusatives and are actually adjectival passives.

Kayne (1994) discusses adjectives and relative clauses, claiming that prenominal participialconstituents as in (16a) are verbal and result from preposing the verbal element of a reduced

relative (16b).

(16) a. The recently sent book

b. The book recently sent...

Kayne claims that these two sentences share an underlying syntactic form. However,

Sleeman & Verheugd (1998) analyse Kayne's argument and conclude that the sentences in(16) are structurally different, (16a) being a simple adjective, and (16b) being a reduced

relative. It is argued that in (16a) 'recently sent' can describe an event but cannot have any

argument structure, whereas in (16b) 'recently sent' does have argument structure,

suggesting that the syntax of the two sentences is sufficiently different to reflect this. Kaye's

idea of preposing reduced relative clauses to a prenominal position extends to prenominal

adjectives. Parallels are drawn with French as support for the notion that prenominal

adjectives are derived from a postnominal position. However, Sleeman & Verheugd also note

a semantic difference between prenominal and postnominal adjectives that they take to

indicate different structures. In (17a) 'visible' indicates a continuing and ongoing property of'the stars', but in (17b) 'visible' denotes a temporary state or occasion.

(17) a. the visible stars...

b. the stars visible...

I draw attention to this preposing account to demonstrate not only that adjectival passives

and verbal passives ('sent' in (16) above) overlap conceptually to a certain extent and

keeping them separate is a tricky task but also that finding the difference between them is

important. In Kayne's work, for the two sentences in (16) to share an underlying structure,

reduced relatives require that they have as a part of them a Spec-CP, and therefore a C-

layer. However, if the sentences in (16) are taken to have different structures, as Sleeman &

Verheugd argue, then that Spec-CP is not essential and may be dispensed with in accord

with what has already been discussed.

Returning to the main issue of what predicate types are acceptable within reduced relatives,

we may reformulate the list, factoring in the result of the discussion so far.

(18) Predicate Types Found in Reduced Relatives

Present Participal VPs: transitive, unergative, and unaccusative


14/51

63


www.llas.ac.uk

Passive Participal VPs

Predicative PPs

Predicative APs: including adjectival passives

I argue that this list of predicates is a natural class as far the syntax is concerned. The NP

modifying constituent that is referred to as a reduced relative may only contain as its

predicate something from this natural class. However, I have not yet defined what it is about

the syntax of reduced relatives that results in this particular natural class and no other.

Iatridou, Anagnostopoulou & Izvorski (2000) give a generalization (19) for reduced relative

content which they claim holds for all the Indo-European languages.

(19) a. A Reduced Relative can contain a past participle if the missing auxiliary is be

b. A Reduced Relative cannot contain a past participle if the missing auxiliary is have

Considering this generalization for English we see that since of all the past participle

constructions only the passive has auxiliary 'be' then only the passive may appear in a

reduced relative. This is parallel to what we have just argued for. Extending this

generalization to all predicate types (20) is tempting, but ultimately ineffective.

(20) a. A Reduced Relative can contain any predicate that is selected by be

b. A Reduced Relative cannot contain any predicate that is selected by have

It seems primae facie that this is an effective definition; present participal VPs, predicative

PPs, and predicative APs cannot be selected by auxiliary have. However, this generalization

would rule in predicative NPs, which are ruled out for reasons to be discussed in the next

part of this paper, and it rules out certain reduced relative structures (21) we would want to

rule in.

(21) John having been there before knew what to do

In (21) 'having been there before' is a reduced relative. However, 'having' is an auxiliary that

cannot be selected by any form of auxiliary 'be'. Attempts to do so (22) result in ill-formed

structures. The extended generalization fails to include valid reduced relatives such as (21).

It is clear that the extended generalization in (20) is ineffective at defining the natural set.

(22) *John who was having been there before knew what to do


15/51

64 Stanton: The Reduced Relative ClauseThere is a peculiar effect that can be observed in reduced relatives that concerns purely

stative verbs (that is, verbs such as 'stand' that refer to a persisting state or situation). In

matrix clauses purely stative verbs are not acceptable with present participles. The sentence

(23a) is judged to be acceptable while (23b) is not.

(23) a. The wall surrounds the city

b. *The wall is surrounding the city

c. The wall surrounding the city has withstood a number of attacks

However, in reduced relatives (23c) the present participle form is perfectly acceptable. This

provides further evidence against an analysis with unpronounced C and I nodes. Under such

an analysis 'surrounding' should always be accepted or never accepted. This is not observed

to be the case.

A Sketch of the Structure Required for Reduced Relative Clauses

Something similar to Moros (2000) analysis of small clauses may be called for in order to

account for the pattern we see with reduced relative clauses. In the previous section we saw

that defining the type of constituent that may appear within reduced relatives as being the

kind of constituent that is selected by auxiliary becannot be said to work. The existence of

progressive stative verbs within reduced relatives is counter-evidence to such a claim.

However, it is not inconceivable that a sub-constituent of the kind of constituents selected by

auxiliary be may work. The existence of morphological elements such as ing, and ensuggest that what may be found within reduced relative clauses may be big enough to

incorporate the functional heads associated with those morphemes, which are found low

down in the I-domain. A point which is crucial to the argument at hand. (24) serves as an

approximation of the type of structure that may be found within a reduced relative clause.

(24)

NP FxP|

the door Fx FyP

Fy VP|-en V tNP

|open

While I am unable to say for certain which functional head Fx may be said to represent here,

I am able to say that it would need to be one sufficiently low down in the structure to be

below the positions where Tense is assigned.


16/51

65


www.llas.ac.uk

Conclusion

In this paper, I have argued that reduced relatives should not be analysed as having the

same structure as regular relatives. I have suggested that the behaviour described in Hudson

(1973) can be easily accounted for if reduced relatives lack C-layer and I-layer in theirstructures.As such, I argue that the term reduced relative clause is a misnomer, that the

syntax of this type of clause is drastically different to the structure of regular relative clauses.

References

Friedmann, N. (2006) Speech production in Broca's agrammatic aphasia: Syntactic tree

pruning in Y. Grodzinsky & K. Amunts (Eds.), Brocas Region. (Oxford: Oxford University

Press).

Giorgi, A. (2010) About the Speaker: Towards a Syntax of Indexicality. (Oxford: Oxford

University Press).

Gueron, J. and T. Hoekstra (1995) The Temporal Interpretation of Predication, in A.

Cardinaletti and M. T. Guasti (eds.), Syntax and Semantics28, 77-108 (Orlando: Academic

Press)

Horvath, J. and Siloni, T. (2005) Adjectival Passives: Active Lexicon. Available at

http://www.tau.ac.il/~siloni/AdjectivalPassives2005.pdf[accessed 30/07/2011]

Huddleston, R. and Pullum, G. K. (2002) The Cambridge Grammar of the English Language.

Cambridge University Press

Hudson, R. A. (1973) 'Tense and Time Reference in Reduced Relative Clauses', Linguistic

Inquiry, Vol.4, No.2, pp 251-256

Iatrido, S., E. Anagnostopoulou and R. Izvorski (1999) Some observations about the form

and meaning of the perfect. (Ms., University of Crete, MIT, University of Georgetown)

Kayne, R. (1994) The Antisymmetry of Syntax. (Cambridge MA: MIT Press)

Marvin, T. (2000) Past participles in reduced relatives at

http://www.hum2.leidenuniv.nl/pdf/lucl/sole/console9/console9-marvin.pdf [accessed

30/07/2011]

Moro, A. (2000) Dynamic Antisymmetry(Cambridge: MIT Press)

Reichenbach, H. (1947) Elements of Symbolic Logic(New York: Free Press)

Ross, J. R. (1972) 'Doubl-ing', Linguistic Inquiry, 33 (2), 61-86

Sag, I. (1997) 'English Relative Clause Construction', Journal of Linguistics33 (2), 431-483

Sleeman & Verheugd (1998) How Reduced are Reduced Relatives?, Linguistics in the

Netherlands 1988, van Bezooijen, R. and R. Kager (eds.), 187199.

Williams, E. (1975). Small clauses in English, in J. P. Kimball, (ed.), Syntax and Semantics,

vol. 4, 249-273. (Orlando: Academic Press).
http://www.tau.ac.il/~siloni/AdjectivalPassives2005.pdfhttp://www.tau.ac.il/~siloni/AdjectivalPassives2005.pdfhttp://www.hum2.leidenuniv.nl/pdf/lucl/sole/console9/console9-marvin.pdfhttp://www.hum2.leidenuniv.nl/pdf/lucl/sole/console9/console9-marvin.pdfhttp://www.hum2.leidenuniv.nl/pdf/lucl/sole/console9/console9-marvin.pdfhttp://www.tau.ac.il/~siloni/AdjectivalPassives2005.pdf


17/51


18/51

67


www.llas.ac.uk

On the narratives' credibility

concerning the disease and the fatalend in Frulein Elseand Effi BriestMatthias Stark

University of Vienna, Austria

[email protected]

Matthias Stark studies German Studies and History at the University of Vienna. He attendedthe University of Birmingham (UK) for one academic year where he was a student of German

Studies. His main area of interest is German literature from the 18th century onwards and its

socio-historical and philosophical contextualisation.

Abstract

The turn of the 20th century was the time when the world witnessed the birth of

psychoanalysis and when the psyche gained centre stage. The literature of the time was

closely linked to the development of this emerging scholarly discourse, for the men of

letters seemed to understand the human emotional life considerably better than medical

scholars; hence, Freud, besides pursuing medical approaches, resorted to literature to

originate his theory of the human mind. This paper works in the reverse direction and

aims to scrutinize the narrative of (mental) disease that is woven around the female

protagonists; as Schnitzlers Frulein Else is considered a textbook representation of

the contemporary medical views on female hysteria and Fontanes Effi Briest features

subtle indications of hysteria. The narratives of disease in both texts include implausible

elements and appear unreliable; and this question of credibility is the object of research.

The findings are a subtle social critique in Effi's case and mirror an incredible reality with

Frulein Else.

Article

Somewhere deep down inside, we really think theyre faking it.

Deborah N. Black, University of Vermont, on hysteria patients

(Kinetz 2006, p. 2)

A narrative of disease is woven around the eponymous female protagonists in the texts

under consideration which end with both protagonists being dead. On several occasions,

though, the narration of the disease appears unreliable; or it is hinted that the characters
mailto:[email protected]:[email protected]


19/51

68 Stark: Disease and fatal end in Frulein Else andEffi Briest

feign their illness. This essay tries to detect and highlight those situations where credibility

can be disputed and to frame the ensuing questions, i.e. concerning the problematic nature

of clear thoughts in an unconscious state in Frulein Elseor the trustworthiness of a patient

who obviously pretended to be ill previously in Effi Briest. As both Effi and Else suffer from a

disease which is rooted in psychology, one inevitable difficulty is to determine if the depiction

of the disease seems unbelievable because the disease itself has the connotation ofincredibility in reality, thus making these overtones of pretending a mere representation of

medical facts (or rather non-facts). Alternatively the implausible elements may be hints of the

narrator and if so, what does this hint towards? In the case of Effi, the meaning of her death

and its connection to the prior illness have to be looked at, for this is a crucial point to the

story and its assertion. This paper analyses the texts separately and forgoes a comparison.

***

Fontane's Effi Briest(1894-5) is written in an elliptical narrative style that is distinctive for

the book. It is a story of untold stories: the story of the Chinaman, the love affair between

Effi and Crampas, or the black hen of Frau Kruse (Greenberg 1988, p. 771-4). The

heterodiegetic narrator provides repeated insight into Effi throughout the book; whereas

Innstetten's inner self only comes to the fore from the point of finding his wife's secret

love letters (Effi Briest, C. 27) onwards. This insight is achieved explicitly in dialogues

rather than by representation of untold thoughts or feelings. Although the narrative voice

affects to be objective for long periods, it reveals itself as biased from time to time. This,

for instance, becomes blatantly obvious in the scenes in Berlin where Effi feigns to be ill

and the narrator pays Effi a compliment for her acting performance (ibid, C. 23, p. 200).

Effi's medical record is not mysterious and the credibility of her fatal disease is not

narrowed by her prior simulation in Berlin. Feigning illness is depicted as the only way

out of the claustrophobic domestic sphere and the only acceptable excuse not to pay the

obligatory visits that social protocol dictates. The narration acknowledges this and is on

Effi's side. This is primarily embodied in the character of Dr. Rummschttel, who realizes

Effi's Komdie (ibid) and who knew ... da dergleichen auch mal zu respektieren sein

knne. Denn gab es nicht zu respektierende Komdien, war nicht die, die sie selber

spielte, eine solche? (ibid).

Hisjudgementafter his third visit is cataphoric and anticipates the social critique of the

text: 'Hier liegt etwas vor, was die Frau zwingt, so zu handeln, wie sie handelt.' (ibid, p.

201).

It is society and the Gesellschafts-Etwas (ibid, C. 27, p. 236; Thesz 2010, p. 22) that

makes the domestic sphere in Effi's prison and it is the legal situation of the time that

entitles the husband to suppress his wife (Greenberg 1988, p. 772). The text criticizes, in

its portrayal of Effi's life,that women have to feign illness in order to gain a little freedom

and, as illustrated in the aforementioned biased nature of the narrator, it defends Effi'saction. Furthermore, the story reflects scepticism towards the conventional treatments


20/51

69


www.llas.ac.uk

and the Kur in particular as it is the fresh air of the prescribed walks that make Effi ill.

Her Luftbedrfnis (Effi Briest, C. 35, p. 283) leads to her fatal illness and she dies of

tuberculosis (Greenberg 1988, p. 772). Thus, she is ultimately made a victim of society

as she is forced into feigning illness by social circumstances (Thesz 2010, p. 22) and

becomes a victim of the medical treatments of the time which, ironically, make her trulyill.

Throughout the novel there are hints towards hysteria as Effi is described as nervous on

more than one occasion (Effi Briest, C. 2, p. 18; C. 6, p. 48), sexual deprivation is

implied (ibid, C. 13, p. 102). Sidonie calls her nervenkrank (ibid, C. 19, p. 157) and Dr.

Rummschttel prescribes keine geistigen Anstrengungen (ibid, C. 23, p. 200). After the

birth of Annie there is a tendency towards illness and unrest (Thesz 2010, p. 26) in Effi

and [i]ndeed, hysteria was linked to a crisis in motherhood and to sterility (ibid). The

hysteria depiction culminates in Effi's nervous breakdown after Annie's visit.

All these can be taken as indicators for hysteria, but Effi is nervous in situations where

young, nave, and inexperienced girls are likely to be nervous. Sidonie's statement is

more or less an insult, and according to the narrator, Dr. Rummschttel is participating in

the comedy. The rest remains but does not prevail; it rather contributes to the narrative

of disease being a deciding reason to the course of the disease and its treatment.

Society's morality and its mechanisms lead to her death. According to Krause (2010),

hysteria depicted as a female disease in Victorian literature can be read as a protest

against patriarchy (Krause 2010, p. 32-3) and shows female resistance and malesuppression (ibid, p. 442). This study of English literature supports the reading of the

narrative of disease in Effi Briest as protest, thus making the hints of hysteria a

subordinate part of the greater narrative which aims to criticize the pressures of society

and medical treatments of the time.

Arthur Schnitzler's Frulein Else (1924), although written some thirty years after

Fontane's Effi Briest, is set in the same time period and provides insight into a few hours

in young Frulein Else's life. Written as an interior monologue the Erzhldistanz

(Aurnhammer 1983, p. 501) is nullified. It is the distinctive character of this narrativesituation that everything succumbs to subjectivity. Thus everything is only reliable insofar

as it is the protagonist's view of things, but it is not objectively reliable. Although the affair

of Paul and Cissy is proven as existent in their behaviour at Else's bedside ( Frulein

Else, pp. 74-6), it was not clear up to that point, for Else might as well just have imagined

the affair. Hence, the narration concerning the outer world is not reliable as it is not

verifiable for the reader and is in part merely based on Else's presumptions. The

typographically denoted lines of the other characters offer other perspectives which are

to the same extent subjective as Else's own perspective. Arguably, these lines of direct

speech that seem to intrude on Else's mind can be viewed as distorted, for it is Else'sperception the readers have to rely on and mishearing or misunderstanding cannot be


21/51


excluded; nevertheless there is not one instance that hints to this lack of credibility. In

another regard, the text is completely reliable: as a protocol of her thoughts that is not

designed by her, every thought is her thought and the text is a glimpse at the world

through her eyes.

As Frulein Else is one of Schnitzler's most famous texts, there is a wide academicdiscourse about it. Commonly, it is viewed as a Freudian novella with a medical focus

depicting hysteria (Aurnhammer 1983, p. 500) including interpretations based on the

Oedipus complex or for example Else being abused by her father (Lange-Kirchheim

1998). Although, these readings make Else's hysterical bout seem more comprehensible,

there is no immanent proof in the text.

Regardless of the causes for Else's bouts, it is the depiction or narration of the bouts

themselves and their incredibility that this essay revolves around. After exposing herself

naked to the guests in the music room of the hotel, Else collapses and is carried to herroom (Frulein Else, pp. 70-3). Lying on her bed Else is in a supposedly unconscious

state as the surrounding characters state: 'Es ist ein Ohnmachtsanfall' 'Du siehst

doch, Mama, dasie ohnmchtig ist.' (ibid, p. 71, emphasis in the original). Yet, Else

still has clear thoughts, perceives what is going on around her and claims not to be

unconscious: Sie halten mich alle fr ohnmchtig. Ich bin nicht ohnmchtig (ibid). The

question at issue is the incredibility of the narration that includes a conscious mind in an

unconscious body; only a detailed analysis can clarify if this is a hint towards something

or what it is and means.

Cissy labels what happens as ein hysterischer Anfall (ibid, p. 76) and she is the one

who thinks Else is feigning everything, thus branding her as a fake in a manner that was

customary with hysteria patients (Lange-Kirchheim 1998, p. 285). Lange-Kirchheim calls

Else's state teilohnmchtig (ibid, p. 286) and argues that the hysterical protagonist

speaksthe truth for she exposes Cissy and Paul as Schwindelbande (Frulein Else, p.

74), thus exposing them as liars. This argument, though, is not valid, just because they

kept their affair from the others does not mean Cissy lies about everything else; and their

lie and Cissy's evaluation of Else's medical state are not matters which mutually exclude

each other.

Nevertheless, it is this complex and controversial stigma of hysteria patients that is at

stake (Lange-Kirchheim 1998, pp. 285-6; Kinetz 2010). Apart from Lange-Kirchheim's

incomplete argumentation, Else's self-imposed prohibition to speak (Ich hre, aber ich

schweige. Ich bin ohnmchtig, ich mu schweigen. Frulein Else, p. 76) and her

suddenly regained control over her body when she reaches for the Veronal indicate

simulation.

Other implausible elements, like Else noticing her aunt tiptoeing to her side (ibid, p. 74),knowing who kisses whom (ibid, p. 75), or sensingCissy in front of the mirror (ibid, p. 76),


22/51

71


www.llas.ac.uk

may just as well be imagined for we are bound to Else's perception without being able to

distinguish her dreams from reality.1In traumhaft-disparatem Erleben reagiert sie zwar

noch auf Berhrungen, Gerche, doch ihre assoziative Wahrnehmungen knnen die

Situation nicht mehr organisieren (Aurnhammer 1983, p. 508).

One attempt at an explanation to clarify the contradictions is to resort to Schnitzler's own

medical (psychological) conception. Contrary to Freud, Schnitzler did not believe in the

dichotomous concept of conscious and unconscious but added what he called the

Mittelbewusstsein (Tweraser 2003, p. 152), a permeable stratum between

consciousness and unconsciousness that is influenced by both of the aforementioned

(ibid). Furthermore, he viewed the psyche as an essentially open system (Herzog 2003,

p. 235) partly due to his experience with hypnosis. Although the hypnotic state is not an

unconscious one, it is this concept of the psyche and its always accessible quality that

makes the post-breakdown depiction of Frulein Else more comprehensible.

Finally, the protocolof Else's mind reveals more to the reader than is accessible to the

protagonist herself. The text contains her first dream of her suicide and her dead body

lying on a bier (Frulein Else, pp. 43-4), but this dream is afterwards concealed to

herself or can only partially be recalled: Was hab ich denn getrumt? Ich glaube ich war

schon tot (ibid, p. 44). Therefore, there is another layer of narration that is accessible to

the reader, but only partially to the protagonist. By implication, we cannot conclude

whether, or to what extent, Else is conscious in the final scene. It is external ascriptions

that allege her to be unconscious, whereas she claims to be conscious. It is possible thatthe mind processes the various stimuli, but if Else gained consciousness again she

would, similar to the first dream, not be able to remember this information. The fact that

Else is not able to move (for most of the scene) and not able to speak, reverts to typical

attributions of female hysteria and hysterical paralysis.

The narration after Else's exhibition remains unreliable, but this is due to the

unbelievable nature of the depicted bouts which are rooted in the psyche. Although the

suspicion is legitimate and the author's conception of the psyche matters, the narration

does not suggest that Else is feigning; rather, it depicts the complex nature of hystericalbouts. The search for the cause is ein zuweites Feld (Effi Briest, C. 36, p. 296) and

cannot be determined based on the text alone.

***

Both female protagonists are young and pushed into adulthood through life-changing

decisions made by their parents; in the course of their stories they try to cope with their

new situations, but in the end both of them are dead. It is due to the idiosyncratic

1 In fact, as Aurnhammer unveils in his article, reality and dreams merge in the course of the text and Else herself

cannot distinguish them (Aurnhammer 1983, pp. 503-8).


23/51


narrations that one becomes suspicious and looks for indications of a hidden meaning.

The general gaps in Effi Briest make the narration appear unreliable or at least

ambiguous. The narrative of disease is taken up again and again, loosely arranged with

long passages in between, thus making it more a persistent theme. Due to this

casualness of the recurring theme and the fact that Effi is feigning illness in Berlin, her

real disease and her sudden death seem implausible. Only on closer inspection is thecausal relation clear and reveals the social critique. The narration does not include

unreliable elements, rather it is a subtle representation of society's pressure and its fatal

implications.

In Frulein Elsethe depiction of the protagonist after her nervous breakdown is riddled

with unreliable elements regarding Else as unconscious. This notion is nourished by the

other characters who believe Else to be unconscious and state that. Else herself claims

otherwise, but she cannot move and is not able to speak even when she tries to. The

great difficulty of the text is that from a certain point onwards we cannot distinguishbetween reality and dream, because Else herself cannot tell them apart anymore. As the

first dream shows, the narration includes more than is accessible to Else's conscious

mind. Thus, even the clear thoughts in her unconsciousstate might appear as conscious

thoughts, but may be dream sequences or thoughts triggered by her surroundings. For

all we know, they might not be accessible to Else later. Although the sudden ability to

move her hand is incredible, the whole representation is just a mirror of reality; for the

medical world still struggles with this renamed but not less mysterious disease and the

scholarly world is torn between believing and denying.

It is the incredibility that all patients of such hysterical bouts are confronted with and

have to fight against. Despite it seeming incredible, one, even from a modern point of

view, has to accept the fact that their actions are not based on their free will, but are

psychologically triggered.

Bibliography

PrimaryFontane, T. (1998), Effi Briest(Mnchen, dtv), Original 1894

Schnitzler, A. (2002), Frulein Else(Stuttgart, reclam), Original 1924

SecondaryAurnhammer, A. (1983), '"Selig, wer in trumen stirbt: Das literarisierte Leben und Sterben

von Frulein Else', in: Euphorion: Zeitschrift fr Literaturgeschichte, (77:4), pp. 500-10

Greenberg, V. D. (1988), 'The Resistance of Effi Briest: An Untold Tale', in: PMLA 103 (5). pp.

770-81


24/51

73


www.llas.ac.uk

Herzog, H. H. (2003), 'Medizin ist eine Weltanschauung: On Schnitzler's Medical Writings,

in: A Companion to the Works of Arthur Schnitzler, ed. by D. Lorenz (Rochester, Camden

House), pp. 227-241

Kinetz, E. (2006), 'Is Hysteria Real? Brain Images Say Yes', New York Times September 26,2006. Available from:http://tinyurl.com/66k2brvaccessed 20 March 2011.]Krause, E. H. (2003), 'Eclectic Affinities: Fontane s Effi and Freuds Dora', in: Women's

Studies, 32:4, pp. 431-454

Kuttenberg, E. (2003), 'Suicide as Performance in Dr. Schnitzler's Prose'. In D. Lorenz (ed.)

A Companion to the Works of Arthur Schnitzle(Rochester, Camden House), pp. 325-345

Lange-Kirchheim, A. (1998), Adoleszenz, Hysterie und Autorschaft in Arthur SchnitzlersNovelle Fraulein Else, in: Jahrbuch der deutschen Schillergesellschaft, 42, pp. 265-300

Lersch-Schuhmacher, B. (1998), 'Ich bin nicht mtterlich Zur Psychopoetik der Hysterie in

Schnitzlers Frulein Else', in: Text und Kritik, 138/9, pp. 76-88

Prutti, B. (2004), 'Weibliche Subjektivitat und das Versagen des sanften Patriarchen in

Schnitzlers Fraulein Else'Orbis Litterarum 59, pp. 159-87

Thesz, N. (2010), 'Marie Nathusius' Elisabethand Fontane's Effi Briest: Mental Illness andMartial Discord in the Century of Nerves, The German Quarterly83.1, pp. 19-37

Tweraser, F. (2003), 'Schnitzler's Turn to Prose Fiction: The Depiction of Consciousness in

Selected Narratives', in: A Companion to the Works of Arthur Schnitzler, ed. by D. Lorenz

(Rochester, Camden House), pp. 149-186
http://tinyurl.com/66k2brvhttp://tinyurl.com/66k2brvhttp://tinyurl.com/66k2brvhttp://tinyurl.com/66k2brv


25/51


26/51

75


www.llas.ac.uk

A modular account of language

change in Alzheimers diseaseClaireCordella

Cornell University, Ithaca, NY, USA

[email protected];[email protected]

Claire Cordella is a senior Linguistics major at Cornell University. She comes from

Londonderry, New Hampshire, USA. Within Linguistics, she is especially interested in

Phonetics and Sociolinguistics.

Abstract

The following analysis derives from a larger study of language functioning, in which the

nature of language change in six subjects with mild cognitive impairment (MCI)/probable

Alzheimers disease (AD) is compared to that of nine healthy adult subjects and eighteen

healthy elderly subjects (Cohen Sherman et al 2011). In this analysiswhich uses the same

data setAD-type language change is considered more narrowly, in light of linguisticmodularity arguments and specifically, the declarative/procedural memory model. Attempts

are made first to identify any significant difference between groups (AD/MCI, healthy adult,

healthy elderly) in terms of overall language abilities, as measured by performance on a

Relative Clause Elicited Imitation (RCEI) task. Data are then analyzed according to a subset

of lexical-, structural- and semantic-type measures. In keeping with prior research, significant

differences in general linguistic abilities are found to distinguish AD/MCI subjects from both

healthy adults and the healthy elderly. Of the sub-type measures, only the lexical variable is

significant in distinguishing AD/MCI subjects from their age-matched healthy elderly

counterparts. Results are taken to support the declarative/procedural memory model andmodularity arguments more generally.

Introduction

Alzheimers disease (AD) is a progressive, generalized dementia affecting a broad range of

cognitive functions and leading to near-complete neurodegeneration (Alzheimers Foundation

of America 2010). Common symptoms of the disease are well-documented but definitive

diagnosis proves difficult since brain tissue samples obtained via autopsy are, at present, the
mailto:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]


27/51

76 Cordella: Language change in Alzheimers disease.

only means of confirming beyond any doubt that a patient suffered from AD1. Thus, it

remains valuable to investigate further means of probable diagnostics and/or early indicators

of AD. A hallmark of Alzheimers-type dementia is language degeneration, especially

impairments of semantic fluency (Hodges and Patterson 1995). Increasingly though,

researchers are finding that other types of language irregularities (e.g. lexical, morphological)

may precede declines in semantic fluency. In general, any findings which point to an

asymmetric language decline are valuable tests of linguistic modularity arguments and the

idea that different aspects of language may be controlled by interrelated but semi-

autonomous brain regions which can develop and/or deteriorate at differential rates.

For decades, linguistics has looked to modularity of mind arguments as a way of grounding

certain field fundamentals; modularity arguments have been variously cited to explain

anything from innateness to acquisition processes and ever-so-increasingly (but not

unrelatedly), pathological performance (Fodor 1983; de Haan et al2010). The latter is an

especially useful extension and one with great potential for practical application. At its

theoretical base, it is quite similar to the many modularity arguments to have preceded it,

most of which have claimed syntactic processing to be fundamentally separate from linguistic

processes which are not rule-governed in the same manner (Linebarger 1989). For most

theoretical practitioners, this allows for a relatively neat delineation of syntactic abilities on

the one hand and lexical functioning on the other (Kinsella 2009). To be sure, there are

arguments to be made in dissent of this view (Pustejovsky 1995); it remains a fact however,

that foundational theories in nearly every linguistic sub-field assume, explicitly or otherwise,

the syntactic/lexical distinction.

What is less clear, even to the more dogmatic of modularists, is the relative location of

semantic, phonological, and morphological faculties; to the extent that their reliance on rules

is undetermined or indeterminable, so too is their modular location. Semantic functioning is of

especial concern for the purposes of this analysis and as such, it is helpful to assign a

preliminary modular identity despite the aforementioned uncertainties. In light of current and

past research, semantics will be assumed to group more reliably with lexical functioning as

opposed to syntactic functioning (Duong et al2006; Rebok et al1990).

The more controversial points of linguistic modularity aside, it remains a useful framework

even in its most basic formulation, namely that which articulates only syntactic and lexicalcapacities as definitively distinct in any modular sense. Indeed, this is the version most

familiar to linguists looking to make such a theory practically applicable. The base

simplification allows for contributions and input from disciplines other than linguistics, most

notably cognitive neuroscience. Especially in the case of language pathology, there is an

ever-increasing reliance on neuroimaging as a means of grounding linguistic theory. Of these

1Given this difficulty, some physicians and researchers have introduced the Mild Cognitive Impairment (MCI)

diagnosis as a way of identifying patients presenting with Alzheimers-like symptoms but who, for obvious reasons,

have not been confirmed to have AD. MCI, though a diagnosis all its own and one which is by no means

synonymous with Alzheimers, is nonetheless a helpful diagnostic proxy since MCI patients go on to suffer from

(autopsy-confirmed) AD at rates ten times that of non-MCI groups (Cohen Sherman et al, forthcoming).


28/51

77


www.llas.ac.uk

approaches, the declarative-procedural model is among the most well-established and widely

accepted. It posits the lexical and syntactic faculties to be governed by distinct regions of the

brain associated with declarative and procedural memory, respectively. In this model, syntax

and other grammar-based linguistic processes are carried out in the frontal cortex and basal

ganglia regions while the lexicon is associated primarily with temporal and tempoparietalneural structures (Ullman 2001). With this as a base assumption, the declarative-procedural

model acquires real predictive power. Many current researchers make reference to this

model as a means of accounting for various particularities relating to a host of pathological

and/or degenerative language disorders, AD among them (Ullman 2001; Walenski et al

2009).

With specific reference to AD, much current research aims to evaluate the declarative-

procedural model in light of its ability to accurately predict and/or correlate clinical

manifestations of the disease. Early results suggest, for example, that AD-typeneurodegeneration is associated with significant declines in lexical and semantic abilities but

that syntactic abilities remain relatively unaffected, at least in the early clinical stages

(Kempler et al, 1987). These findings, coupled with the general medical understanding that

AD affects the temporal regions most severely (Scheltens et al. 1992; Killiany et al. 1993),

are often summarized as being generally concordant with the declarative-procedural model.

The analysis to follow is a continuation of previous research, with the goal being to weigh

inempirically rather than speculativelyon the question of linguistic modularity as it relates

to AD in particular. The hypothesis, then, is as follows:

It is expected that (i) Clinical AD/MCI patients will show an appreciable deterioration in

language abilities, as compared to either healthy adults or the healthy elderly and that (ii) this

deficit will be more marked in the case of primarily lexical and semantic performance;

syntactic functioning is hypothesized to be relatively less affected.

Methodology

All data in this analysis is derived from a larger, unprecedentedly comprehensive study into

AD-type language degeneration (Cohen Sherman et al, 2011), the questions and goals of

which are beyond the scope of this small and comparatively limited investigation. Certain

methodological approaches of that larger study do, however, require mention. These include

the three-way grouping of subjects as part of the healthy adult (MIT), healthy elderly (Cornell)

or AD/MCI cohort (MGH), a grouping schema upon which all subsequent statistical analyses


29/51


are based2. Because grouping is done according to combined age and AD characteristics,

three-way comparisons are almost always followed by post-hoc analyses designed to isolate

the AD/MCI effect (i.e. the Cornell-MGH comparison), as this is the variable of primary

concern in this particular analysis.

Though the larger study developed and administered an all-encompassing battery of

linguistic tests (including elicited imitation, 3-word, and picture description tasks, among

others), this sub-investigation deals exclusively with Relative Clause Elicited Imitation (RCEI)

data. In large part, this choice was purely practical: RCEI was, at the time of initial

investigation, the only task completely transcribed and scored. In a sense though, the choice

proved provident in that elicited imitation involves both lexical and syntactic faculties (Lust et

al, 1996), represented in this data set in terms of lexical and structural sub-scores.

The larger data set codes also for a semantic variable in the sense that each elicited

imitation task is administered via two distinct sets of batteries, one of which is semantically

meaningful (+SEM) and the other of which is semantically vacuous (-SEM). A prediction of

semantic abilities and the (potentially) differential deterioration thereof was included as part

of the hypothesis in light of the persuasive findings of earlier studies on the matter (Hodges

and Patterson 1995) and a conscious effort to identify the most promising avenues for future

research.

Four variables, then, were subjected to statistical analysis using SPSS. These included

overall RCEI performance (RCEI.total, measured as a composite battery score), lexical

RCEI performance (RCEI.lex, measured in terms of total number of lexical -type errors),

structural RCEI performance (RCEI.str, measured in terms of total number of syntactic-type

errors) and semantic RCEI performance (DIFFplus/minusRCEI.total, a distilled variable

measured as the magnitude of difference in +SEM battery scores as against SEM battery

scores).

For each of the above variables, a one-way analysis of variance (ANOVA) was conducted to

determine significance. Where a significant difference of means was identified3, post-hoc

2The relevant grouping of subjects is according to age and clinical status. For convenience, these cohorts are

identified sometimes by institution, where healthy adult = MIT (subjects in this cohort are MIT undergraduate or

graduate students and all interviews were conducted on campus), healthy elderly = Cornell (subjects were healthy

elderly drawn from the Ithaca, NY area and all interviews were conducted at Cornell), and AD/MCI = MGH

(subjects were patients referred to study by physicians at Massachusetts General Hospital or affiliated

organizations and all interviews were done at MGH).3

Fisher LSD tests were administered in all and only cases where the omnibus F was significant. This stipulation

provides some degree of statistical protection against alpha-error (Sirkin 2006), which is especially important in

light of the general shortcomings of the Fisher LSD (see footnote below).


30/51

79


www.llas.ac.uk

analyses using the Fisher LSD test4 were done in an attempt to determine which of the

comparisons accounted for the overall significance. Especial attention was paid to the

Cornell-MGH comparison, since these age-matched groups differ only in AD characteristics,

thus making it possible to isolate the desired Alzheimerseffect.

Results

A one-way analysis of variance (ANOVA) was conducted to compare the effect of group

membershipa combined proxy for age and AD characteristicson RCEI performance in

healthy adult (MIT), healthy elderly (Cornell) and clinical AD/MCI (MGH) conditions. RCEI

performance was measured in terms of overall composite scores and also in terms of

disaggregated measures relating to lexical and structural sub-scores and semantic differencescores. Overall composite scores were figured as an integer score out of twenty-four, where

twenty-four is the total number of RCEI token sentencesone point was recorded for each

response marked as correct, zero points were given in the case of an incorrect response.

Lexical and structural sub-scores were figured as an error count score across all batteries,

where errors classified as either lexical or structural were summed to yield respective error

scores. The semantic difference

score is calculated as the

magnitude difference between

the +SEM battery score (a twelve

token subset of sentences with

semantic meaning, score ceiling

of twelve) and the SEM battery

score (a twelve token subset of

sentences without semantic

meaning, score ceiling of twelve).

The ANOVA showed that the

effect of age/AD characteristics

on overall RCEI performance is

significant, F(2,30)= 5.312,

4The legitimacy of the Fisher LSD is sometimes debated on the grounds that it allows for unacceptable levels of

alpha error (i.e. >.05), or the chance of identifying a significant result where none exists. It is just this leniency,

however, which makes it an appropriate test for the study at hand; the studyowing to its incredibly small N and

possibly small effect sizes, among other factorsis at incredible risk for beta error, the case where a significant

effect is not detected where one does, in fact, exist (Sirkin 2006). So because the study lacks statistical power

and moreover because comparisons are rather limited in number (maximally three-way, making the risk of alpha-

error only slightly elevated), the choice of a lenient post-hoc seems justifiable. Thus, the standard .05 is taken as

the threshold significance level, even in the case of the Fisher LSD. Should the study be extended in sample size

or otherwise made less prone to beta-error, the applicability and legitimacy of such a test ought to be re-evaluated.


31/51


p=.001. The mean total score is significantly lower in the clinical AD/MCI group condition

(M=8.67, SD=6.713) than in either the healthy adults group condition (M=20, SD=4.301),

p=.003, or the healthy elderly group condition (M=15.44, SD=7.406), p=.037. There is no

significant difference in mean total scores of the healthy elderly as compared to healthy

adults. A graphical representation of this data is given in Figure 1. Overall significance is

shown in Table 1 (first row) and post-hoc pvalues are given in Table 2 (first row).

Table 1. Linguistic Performance by Study Group

Subscore me

Healthy

Adult

(MIT)

(n=9)

Healthy

Elderly

(Cornell)

(n=18)

Clinical AD/MCI

(MGH) (n=6)

p valueb

RCEI Totalc 20.00 14.44 8.67 .011

RCEI Lexical 4.89 10.93 18.33 .003

RCEI Structural 0.22 0.94 2.33 .034

RCEI Semantic 2.00 2.056 2.33 .918

aAs explained in METHODOLOGY, the semantic score reported here is actually a magnitude difference score.

bPvalues are based on the one-way analysis of variance (ANOVA).

cNote that RCEI Total is a composite score which includes more subscores than those disaggregated here.

Table 2. Multiple Comparisons of Selected Variables

(A) Group (B) Group p valuea

RCEI Total Healthy Elderly (Cornell)

Clinical AD/MCI (MGH)

.101

.003

Healthy Elderly (Cornell) Clinical AD/MCI (MGH)

Health y Adult (MIT)

.037

--

RCEI Lexical Healthy Adult (MIT) Healthy Elderly (Cornell)


.037

.001

Healthy Elderly (Cornell) Clinical AD/MCI (MGH)

Health y Adult (MIT)

.028

--

RCEI Structural Healthy Adult (MIT) Healthy Elderly (Cornell)


.235

.010

Healthy Elderly (Cornell) Clinical AD/MCI (MGH)Healthy Adult (MIT)

.052--

apvalues are based on a (protected) Fisher LSD test.


32/51

81


www.llas.ac.uk

Significant differences were likewise found in mean lexical sub-scores among the three

groups, F(2,30) = 7.077, p=.003. Post-hoc analyses using the Fisher LSD test showed that

the average lexical sub-score was significantly higher in the clinical AD/MCI group condition

(M=18.33, SD=8.189) than in either the healthy adults group condition (M=4.89, SD=3.219),

p=.001, or the healthy elderly group condition (M=10.94, SD=7.55), p=.028. The averagelexical sub-score was also significantly higher for the healthy elderly as compared with the

healthy adults, p=.037. In Figure 2, mean lexical sub-scores are graphed by group. Overall

significance is shown in Table 1 (second row) and post-hoc p values are given in Table 2

(second row).


33/51


The effect of age/AD characteristics on structural sub-scores was significant, F(2,30) = 3.779,

p=.034. Post-hoc analyses showed that the mean structural sub-score was significantly

higher in the clinical AD/MCI group condition (M=2.33, SD=1.966) than in the healthy adults

group condition (M=.22, SD=.441). No other group comparisons were significant. Mean

structural sub-scores are graphed by group in Figure 3. Overall significance is given in Table

1, p.4 (third row) and post-hoc p values are listed in Table (third row).

The omnibus F for semantic difference scores was not significant F(2,30) = 0.86, p=.918,

meaning that scores of semantic difference do not differ significantly as a function of age/AD

characteristics. As such, no post-hoc comparisons were obtained. Figure 4, p.6, graphs

semantic sub-score by group. In that figure, a side-by-side comparison is made between

+SEM battery scores and SEM battery scores. Distilled from that data are the mean

semantic difference scores, which are then graphed by group in Figure 5. Overall

significance for the difference scores is given in Table 1 (fourth row).

Discussion

Interpretation of Results

From the preliminary results given above, it would appear as if the data generally supportthe

initial hypothesis, which predicted (i) that linguistic performance would be significantly lower

overall for clinical AD/MCI subjects relative to both the healthy adult and healthy elderly

cohorts and (ii) that this difference in linguistic performance would be significantly more

pronounced in lexical- and semantic-type tasks and less so in structural-type tasks. To

evaluate this hypothesis, it is necessary to consider the post-hoc p values for relevant

dependent variables, RCEI.total, RCEI.lex, and RCEI.str. In a somewhat separate

manner, it is also necessary to consider the non-significance of the semantic variable as

possibly contradictory evidence as regards the hypothesis.

In the case of RCEI.total, mean scores for the clinical AD/MCI cohort were in fact

significantly lower than those of either the healthy adult or healthy elderly populations. That

mean scores differ significantly between the two age-matched groups (Cornell-MGH) is an

especially encouraging result, as it suggests an Alzheimers effect independent of age

effect(s). Indeed, of the between-group comparisons implied by the hypothesisMGH-

Cornell andMGH-MITonly the former is truly useful in any scientifically valid sense. In the

case of the latter comparison, a significant result says nothing of the source of difference (i.e.

whether it is age or AD/MCI characteristics) since healthy adults differ from clinical AD/MCI

patients on at least two levels: age and AD/MCI status.

At this point, then, it is helpful to refine the orientation of the analysis so that the MGH-

Cornell comparison is understood as the primary one of interest. Thus, in evaluating the


34/51

83


www.llas.ac.uk

second extension of the hypothesis, one need only compare the MGH-Cornell post-hoc p

value for RCEI.lex with the corresponding post-hoc p of RCEI.str. Crucially, the first of

these post-hoc ps was significant, meaning that the clinical AD/MCI group had significantly

higher lexical error scores (i.e. they made more lexical-type errors, on average) than did the

healthy elderly group. In the case of structural error scores, no significant difference wasfound in means of the clinical AD/MCI group as against the healthy elderly. A significant

RCEI.lex post-hoc pvalue, as compared to a non-significant RCEI.str post-hoc p,makes it

more certain that there exists a meaningful difference5 (Cornell v. MGH) in mean lexical sub-

scores and less certain that there exists such a difference (Cornell v. MGH) in structural sub-

scores. Frustratingly, the real premise of the hypothesis (i.e. that lexical differences would be

greateror more markedthan structural differences) is largely unanswerable in light of the fact

that RCEI.str returned a non-significant MGH-Cornell post-hoc p, thereby making it

impossible to compare magnitude effects. To get around this, non-significant differences are

sometimes assigned a tvalue of zero and then compared to other non-zero ts (as in, thoseassociated with significant ps); this approach, though, is not entirely sound and for this

reason, is not attempted here. The difference in certainty, discussed above, ought to be

enough to affirm the basic extension of the hypothesis.

Before deciding finally on the overall strength of the hypothesis, one ought to consider the

semantic variable, DIFFplus/minusRCEI.total, and its overwhelming non-significance. Of

the four dependent variables studied, DIFFplus/minusRCEI.total was the only one to return

a non-significant omnibus F in the preliminary ANOVA . This means that there is no apparent

age/AD effect on semantic difference scores. This finding is in contrast to earlier research onthe subject of semantic fluency in subjects with probable AD; most of this research notes

significant AD-related declines in semantic abilities, especially semantic memory (Hodges &

Patterson 1995; Nebes 1989). If valid, the anomalous finding of this analysis with regard to

the semantic variable also poses an implicit challenge to the declarative/procedural model,

given that semantics is thought to be processed in regions of the brain associated with

declarative memory (e.g. anterior temporal lobes) (Martin & Chao 2001) . Of course, the

question remains whether the semantic variable extracted for the purposes of this analysis is

analogous in type to more traditional semantic measures. Manipulating semantic meaning in

the context of elicited imitation (i.e. by devising +SEM/-SEM battery types) is likely not themost direct way of assessing semantic abilities and any variations therein. Fortunately, the

larger study from which this analysis derives includes a number of tasks which lend

themselves more naturally to semantic analyses (e.g. picture description, 3-word task) in the

sense that a semantic variable can be extracted more or less directly, with minimal

abstraction.

5Meaningful difference refers here to the research hypothesis, as statistically defined. In other words, the

difference is meaningful if between-group variation is too great to be attributed to chance and must therefore be

attributed to an independent Alzheimers effect.


35/51


Taken together, the significance results for the variables RCEI.lex, RCEI.str, and

DIFFplus/minusRCEI.total, support the notion of linguistic modularity, as introduced in the

beginning of this analysis. If the language faculty were domain-general in nature, one would

expect language deficits and declines to be uniform across all linguistic categories. Analysis

of just three linguistic categoriessyntax (RCEI.str), semantics

(DIFFplus/minusRCEI.total) and the lexicon (RCEI.lex)reveals just the opposite pattern:

AD/MCI subjects make significantly more lexical-type errors than do the healthy elderly, but

the difference in structural-type errors between the two groups is non-significant, as is the

difference in semantic (difference) scores between those two groups. Moreover, even the

non-significant structural- and semantic-type results differ from one another in their degree of

non-significance, with the latter being highly non-significant and the former only marginally

so. Clearly then, this analysis findings are consistent with the notion of linguistic modularity.

More arguably, the analysis is also consistent with the declarative/procedural model of

memory, which in its essence, is linguistic modularity as applied to neurobiology. Again, the

one caveat to this consistency is the unexpected non-significance of the semantic variable,

which patterned in ways not predicted by the declarative/procedural model. In that model,

lexical and semantic functioning are usually grouped together as part of declarative memory

and expected therefore to pattern together where applicable (Ullman 2001; Tulving and

Markowitsch 1998) . In the case of AD-type language change, lexical as well as semantic

functioning was hypothesized to be relatively more impeded than syntactic functioning, given

that the neurodegeneration characteristic of AD affects the temporal regions most severely

(and the temporal areas are precisely where the lexicon and semantics, like all other aspects

of declarative memory, are processed) (Scheltens et al. 1992; Ullman 2001; Tulving and

Markowitsch 1998). The fact that the semantic variable returned a highly-non significant

result, combined with the comfortable significance of the lexical variable, casts some degree

of doubt on the declarative/procedural model; it does not, however, undermine its validity per

se. Rather than dismantling an otherwise legitimate framework, it is perhaps better to view

the anomalous patterning of the semantic variable as just that, especially in light of the

conceptual problems and uncertainties discussed above in relation to

DIFFplus/minusRCEI.total. More important is the consistent patterning of RCEI.lex as

compared to RCEI.str, since both of these variables are more naturally derived and/or

assessed in the context of elicited imitation.

Anticipated Criticisms

Largely ignored in this analysis was the independent effect of age on language abilities in

each of the three groups. This omission is potentially problematic in that the age effect may

bear importantly on conclusions and connections made in this analysis. It could be the case,

for instance, that age effect patterns prove similar to those identified for the Alzheimers

effect. In other words, it could be that the healthy elderly also lose lexical abilities in a

comparatively more marked fashion than they do structural abilities. To the extent that AD-


36/51

85


www.llas.ac.uk

type language change mirrors that of normal ageing, this analysis fails to fully account for the

data in all its richness.

The second category of criticisms relates to statistical approach. In this analysis, great care

was taken to maintain the integrity of the experimental design (of the original study), in whichall groupings are made according to a combined proxy which takes into account both age

and AD/MCI characteristics. This yields three independent groups (healthy adults, healthy

elderly and clinical AD/MCI patients) and *one* independent variable (IV), hence the choice

of the one-way ANOVA. Post-hoc comparisons are then used as a means by which to

separate out the component IVs (age and AD characteristics) and measure their respective

effects on the DV(s). In theory, this seems a logical approach to data analysis, given the

design of the study in question. In reality though, the multi-step analyses implied by the

ANOVA complicate the statistical picture. Might there exist a simpler, more direct way to

assess significance in this study? In answer to this, one can imagine a scenario in which dataare recoded in binary fashion to a yield a new age variable (i.e. 0 for subjects previously

classified as healthy adults and 1 for subjects previously classified as either healthy elderly

or clinical AD/MCI) and also a new AD/MCI variable (i.e. 0 for all subjects classified

previously as either healthy adults or healthy elderly and 1 for all subjects classified

previously as clinical AD/MCI patients). In this case, a two-way ANOVA with fixed factors

age and AD/MCI and dependent variables RCEI.total, RCEI.lex, RCEI.str, and

DIFFplus/minusRCEI.total could be run. Besides eliminating the need for post-hoc

procedures, the two-way ANOVA is appealing in its ability to derive fairly transparent

interaction effects between age and AD/MCI status, which, though not of particular import inthis analysis, are likely to prove important in future analyses.

Conclusion

This analysis looked at language abilities across three groups: clinical AD/MCI subjects,

healthy adult subjects and healthy elderly subjects. With regards to overall linguistic

competence, as measured by score totals on the RCEI task, significant differences were

found between mean RCEI scores of AD/MCI subjects as compared with those of both thehealthy adult and healthy elderly populations. When AD/MCI subjects were compared to

their age matched, healthy elderly counterparts across lexical-, structural-, and semantic-type

measures, a significant difference was found only in the case of the lexical-type measure.

This finding in particular supports linguistic modularity arguments and somewhat more

arguably, the declarative/procedural memory model.

An

linguistics-debut volume 2 number 2

Documents