Science of Aphasia 5 Cross Linguistic Aspects of Aphasia ...

Science of Aphasia 5
Cross Linguistic Aspects of Aphasia
POTSDAM
SEPTEMBER 16 TH -21 ST , 2004
ABSTRACTS
Scientific Committee
Prof. Dr. Roelien Bastiaanse, Groningen, The Netherlands (chair)
Prof. Dr. Ria de Bleser, Potsdam, Germany (vice-chair)
Prof. Dr. David Howard, Newcastle, UK
Prof. Dr. Matti Laine, Abo, Finland
Prof. Dr. Claudio Luzzatti, Milano, Italy
Prof. Dr. David Poeppel, Maryland, U.S.A.
Prof. Dr. Riitta Salmelin, Helsinki, Finland
Prof. Dr. Cynthia K. Thompson, Evanston, U.S.A.

Applying a connectionist model to German naming data
Stefanie Abel 1 , Walter Huber 1 , Gary S. Dell 2
1 Neurolinguistics at the Department of Neurology, University Hospital, Aachen University of Technology
(RWTH), Germany, 2 Language Production Laboratory, Beckman Institute, University of Illinois at Urbana-
Champaign, USA
Introduction
The connectionist model of Dell and coworkers (Foygel and Dell, 2000; Dell et al. 1997) simulates
normal and impaired lexical access in picture naming and thereby assesses aphasic naming disorders. The lexical
access involves two steps, retrieval of the abstract lexical information (lemma) from semantics and retrieval of
the corresponding phonological form. There are two distinct deficit theories: The weight/decay (WD) model
attributes the impairment to globally weakened activation transmission (altering connection weight parameter w)
and/or to a global loss of activation (altering decay parameter d). In the semantic-phonological (SP) model, the
disorder is attributed to reduced spread of activation between semantic and lexical levels (lexical-semantic
weight s) and/or between lexical and phonological levels (lexical-phonological weight p). The internet program
WebFit (URL: http://langprod.cogsci.uiuc.edu/cgi-bin/webfit.cgi) uses the distribution of naming error types for
a patient to assign parameter values to each deficit model. As articulatory-phonetic processing is beyond the
model’s scope, nonfluent aphasics were originally excluded in order to avoid speech motor deficits.
The model’s network structure preserves the error opportunities of the English lexicon (originated in the
neighborhoods of English words). Under the condition of random output (i.e., without lexical influences but
following phonotactic rules), the model produces a proportion both of words vs. nonwords and of the various
kinds of word errors similar to English error opportunities.
In this abstract, we explore if the connectionist diagnosis can be applied to German speaking, fluent and
nonfluent aphasic patients. Therefore, we analyze and compare characteristics of the English versus German
Lexicon. We show the fitting of these patients’ naming data to both deficit models to compare goodness of
fitting.
Methods
Subjects
In an ongoing therapy study, patients of the Aachen aphasia ward exhibiting post-acute or chronic aphasia
and a naming disorder without a severe speech motor impairment were selected for pre-testing. Six patients
exhibited a fluent aphasia (four Wernicke’s, one transcortical-sensory and one unclassifiable fluent aphasia) and
three patients a nonfluent aphasia (two Broca’s and one global aphasia). Patients were tested for speech fluency
(articulatory diadochokinesis). Four fluent aphasics and all three nonfluent aphasics showed a mild or moderate
speech motor deficit (speech apraxia/dysarthria).
Material
The material consists of 160 pictures from Snodgrass & Vanderwart (1980) validated for the German
language (Mottaghy, 1998).
Procedure
We produced random pairs of the German corpus with a two-fold goal. First, word/nonword opportunities
were analyzed by exchanging phonemes. Second, opportunities for different word errors were derived by
examining the linguistic relation (semantic, formal, mixed, unrelated) between words of each random pair (see
Dell et al., 1997). Aphasic responses were classified according to Dell and colleagues (1997), except that
obvious speech problems were ignored and correct responses included synonyms. (This was justified because
Genzel et al.,1995, found higher name variability in German than in English for the Snodgrass & Vanderwart
pictures). Each error pattern is analyzed by Webfit, using both deficit models (Independence SP- and WDmodels,
20% mixed errors, minimize χ 2 ).
RESULTS
Lexicon comparison revealed that error opportunities for the German lexicon are similar to those for the
English lexicon and the computer model (see Fig. 1). This holds true for both aspects: the likelihood of
word/nonword occurrence and of error types. As the German naming corpus entails more compounds (23%) than
the English naming corpus (Philadelphia Naming Test; 9,1%), we analysed sound exchanges creating new words

in compounds separately. Word proportions for both languages were closest if these compounds were analysed
as words. Therefore, analysing aphasic naming data, semantic neologism were classified as words. (For example,
the semantic neologism Gummihammer (rubber hammer) in relation to target Glocke (bell) constitutes a formal
error.)
Fig. 1. Opportunities for words versus nonwords (left side) and different word errors (right side). As words tend to clump in
phonological space, estimated nonword opportunity for computer model simulation is set to 80%. The remaining 20% word
opportunity is divided among word error types.
Fitting the patients' naming data to the SP-model, five patients with a fluent aphasia showed a
predominantly semantic (sp but s and p less than .002
apart). Using the WD-model, all patients with a semantic lesion and the patient with a mixed phonological lesion
in the SP-model had a decay lesion (w>.055). The remaining two patients with a predominantly phonological
lesion revealed a weight lesion (w

Deficit-specific speech therapy: er-fMRI evidence from phonologicallycued
training in an anomic patient
Jubin ABUTALEBI * °, Paolo VITALI * ^, Marco TETTAMANTI *, Daniela PERANI
*, & Stefano F. CAPPA *
* Centre for Cognitive Neuroscience, University Vita-Salute San Raffaele and HSR,
Milan, Italy
° Interdisciplinary Center for Cognitive Studies, University of Potsdam, Germany
^ Institut Universitaire de Gériatrie de Montréal, Université de Montréal, Québec, Canada
Clinical neuroscientists have long known that some recovery of function is
possible after injury to the brain but the nature and exact mechanisms of language
recovery after brain damage are still controversial although a strong relationship between
neuroplasticity and recovery is postulated (Blomert, 1998). Indeed, when stroke patients
regain some of their language abilities, their behavioral improvements may result from
either of two mechanisms (Cappa, 2000). Recovery may be achieved by adopting novel
cognitive strategies for function performing (i.e., recruitment of uninjured cerebral areas
which are usually not necessary for the lost function in the intact brain). On the other
hand, recovery may be achieved because of the involvement of homotopic (i.e.
homologous) areas of the controlateral hemisphere that may have compensatory function.
However, how speech therapy may directly influence this cerebral re-organization is a
further matter of debate.
The repeatability of fMRI over time allows one to investigate how patterns of cerebral
activity may change during the course of recovery and to observe consequently the
influence of treatment on brain functioning.
In the present study we investigated the effects of anomia-specific speech therapy
on cerebral re-organization in an anomic patient, G.R.. Using event-related functional
Magnetic Resonance Imaging (er-fMRI), we scanned patient G.R. twice, before and after
specific speech therapy for anomia (phonological cued naming training). The rationale
for concentrating on phonologically cued naming training was inferred by the fact that
G.R. was able to name when cued with the beginning syllable.
A set of pictures that G.R. could not spontaneously name was selected for intensive
speech therapy. After acquisition of the first er-fMRI session, training was intensively
administered on a daily basis by a speech-pathologist, until a 50% correct naming
performance (at least) on the training picture set was achieved. During the pre- and posttraining
er-fMRI acquisition G.R. had to overtly name visually presented pictures of the
trained set and of an un-trained control set. Furthemore, fMRI data were also collected
during a third condition in which G.R. was phonologically cued during naming.

The er-fMRI results during non-trained spontaneous successful naming (A), successful
training-induced naming (B) and pre-training phonologically cued-naming (C) are
displayed in figure1.
In G.R. spontaneous (non-trained) naming was mainly associated with activations in
the non-dominant hemisphere (right inferior and middle frontal gyrus and right insula).
Successful training-induced naming resulted, at the brain level, in a significant shift to
left hemisperic perilesional areas (left inferior frontal gyrus and left supramarginal
gyrus). However, also the right homologue of Broca’s area was found to be activated
probably due to the fact that his original brain lesion extended into the left Broca'a
proper. Pre-training phonologically cued naming was mainly associated with significant
brain activity in left-sided perilesional areas (left middle frontal gyrus, left supramarginal
gyrus, left posterior parietal cortex).
This present finding suggests that perilesional areas of the dominant hemisphere may
be crucial for effective recovery. In the absence of the left Broca's area, successful
naming may be mediated by its right homologue area. It is also woth underlining that
anomia-specific rehabilitation, based on phonological cueing, was associated with
activation in brain areas classically ascribed to phonological processing, phonological
working memory and lexical retrieval (left inferior frontal gyrus and supramarginal
gyrus); such an activation pattern could express a facilitated access and retrieval of
lexical items from the mental lexicon by enhanced and trained sub lexical, phonological
strategies, as demonstrated by the brain activity pattern during the phonologically cued
naming condition.
Figure 1.

Neuroanatomical correlates of selective impairments of nouns and verbs
Silvia Aggujaro 1 , Davide Crepaldi 1 , Caterina Pistarini 2 , Mariangela Taricco 3
and Claudio Luzzatti 1
1 Department of Psychology, University of Milano-Bicocca
2 S. Maugeri Foundation, Montescano Medical Center, Pavia
3 Rehabilitation Unit, G. Salvini General Hospital, Passirana, Milano
INTRODUCTION
Theoretical linguistics suggests that the representation of verbs (V) and nouns (N) is independent, reflecting the
functional difference of these word classes in syntax. Observation of brain damaged patients demonstrates the
existence of double dissociations with disproportionate impairments of either Vs or Ns (e.g. Berndt et al. 1997).
However, the mechanisms underlying N-V dissociations are not yet well defined; some authors have considered
them to be proof that the mental organization of Ns and Vs is separate, while others (Bird, Howard and Franklin,
2000), maintain that Ns and Vs do not dissociate on the basis of a true grammatical class effect, but on that of an
artifact due to the different kind of underlying semantic knowledge.
The primary aim of the present study was the identification of the anatomical loci of the lesions causing
selective impairments of Vs and Ns. In addition, an attempt was made to disentangle the effect of imageability
from a possible effect of the degree of motor functional mental images underlying verbs (actionality), and,
finally, to analyze the relation of these variables and the site of the underlying cerebral lesion.
MATERIALS AND METHODS
Subjects
Fifteen aphasic patients suffering from disproportionate impairment of verbs (DIV) and five patients suffering
from disproportionate impairment of nouns (DIN) participated in the study.
N-V dissociations were identified by evaluating naming abilities of Ns and Vs through a picture-naming task.
Disproportionate impairment on either Vs or Ns was identified by a logistic regression analysis. The majority of
the DIN cases also showed a word frequency effect; all 15 DIV cases showed an imageability effect.
Task
The picture naming task used to test the naming ability of the participants is composed of 30 pictures of objects
(15 natural objects and 15 artifacts) and 40 pictures of actions (of which 12 verbs with low degree of actionality,
14 with an intermediate and 14 with a high degree of actionality).
The major lexical-semantic variables (oral word frequency, age of acquisition, familiarity with the underlying
concepts and imageability) were included in the experimental design.
Imageability and actionality ratings
Each of the items included in the naming task was scored by a group of 20 adult controls for visual imageability
using a 1-to-7 rating scale (Luzzatti et al. 2002).
The degree of fine motor planning of hand movements (actionality) underlying the action denoted by each of the
verbs elicited by the naming task was tested by a group of 25 healthy undergraduate students that estimated the
degree of actionality using a 1-to-7 scale.
Mapping of brain lesions from NMR & CT scan pictures
Lesions were reconstructed and mapped onto a lateral template of the brain.
RESULTS
Localization of the brain lesions causing DIN and DIV
In a first part of the study the areas of the brain involved in the storage of lexical representations of Vs and Ns
were investigated by localizing the brain lesions causing disproportionate impairment to the retrieval of these
lexical categories.
In the majority of the DIN patients, the lesion involved the medial part of the middle and inferior left temporal
gyri (Broadmann area (BA) 21, 20, 37). On the contrary, the lesions found in DIV patients clustered in two major
subsets: in the first group of patients, the focal brain damage involved the posterior part of left temporal lobe and
the inferior parietal gyrus (BA 21 posterior, 37 superior, 39, 40); in the second group, the lesion involved the left
frontal, temporal and parietal perisylvian territory (BA 44, 40, 22, 41, 42), causing complete destruction of the

left hemisphere language areas. In a third group the lesion involved the basal ganglia. Isolated frontal lesions
were not present in any of the dissociated cases.
Disproportionate N- or V-impairment and imageability
This analysis addressed the issue whether the verb naming deficit is caused by damage to semantic features
underlying lexical entries, and whether noun superiority emerges exclusively from an increased sensibility to
imageability.
None of the five DIN patients has a selective naming impairment of natural objects, while none of the fifteen
DIV patients had selective impairment of artifacts.
The grammatical class effect remained in two patients after bivariate regression analysis (grammatical class by
imageability), which would seem to indicate that, at least in some cases, DIV may be due to true lexical damage.
Disentangling imageability from actionality
This analysis investigated whether, with respect to Vs, the imageability effect depends on the degree of
underlying motor functional mental images (actionality).
Results show a high degree of independence between visual imageability and actionality ratings.
Is the TP area which causes DIV an interface area for action verbs or a critical area for all types of verbs?
DIV is caused by a TP lesion in most fluent aphasic patients. This analysis tested whether this region represents
a locus of interface between the motor and lexical representation of action verbs or whether it is a functional unit
specifically involved in processing verbs.
Two of the four patients with posterior temporal and inferior parietal lesions could not name any of the low or
high actionality verbs. The remaining two patients showed identical damage to both the high- and lowactionality
verbs. There was no significant difference between the number of correct responses produced for the
two verb sets; in fact, there is a trend in the opposite direction, as low-actionality verbs (LAV) are slightly more
impaired than high-actionality verbs (HAV).
DISCUSSION
The brain areas involved in the lexical retrieval of verbs and nouns were investigated in a series of aphasic
patients by localizing the lesions causing disproportionate impairment to the retrieval of either of these lexical
categories. Impaired naming of objects mostly followed damage to the middle and inferior temporal lobe;
impaired naming of actions often followed damage to the posterior temporal and inferior parietal lobes. Isolated
pre-motor, pre-frontal lesions were not present in any of the dissociated DIV cases.
The data collected with this study are in conflict with the results obtained from functional neuroimaging studies
that have reported a specific activation to the frontal lobe in retrieving verbs, and are in favor of a critical role of
more posterior temporo-parietal areas.
Bird et. al’s predictions based on the imageability assumption were not confirmed. None of the 5 DIN patients
suffered from a disproportionate naming deficit of natural objects and none of the 15 DIV patients suffered from
a disproportionate naming deficit of artifacts. The two patients in which imageability did not partial out the
grammatical class effect seem to dissociate because of a true lexical damage of verbs.
The degree of visual imageability and of actionality underlying verbs are almost independent.
Naming of verbs with high and low actionality ratings is equally impaired in the four patients with TP lesion.
The TP area would therefore appear to play a critical role in the retrieval of verbs, irrespective of their degree of
actionality.
REFERENCES
Berndt, R.S., Mitchum, C.C., Haendiges, A.N., Sandson, J. (1997). Verb retrieval in aphasia. 1. Characterizing
single word impairments. Brain and Language, 56, 68-106.
Bird, H., Howard, D., Franklyn, S. (2000). Why is a verb like an inanimate object?: Grammatical category and
semantic category deficits. Brain and Language, 72, 246-309.
Luzzatti, C., Raggi, R., Zonca, G., Pistarini, C., Contardi, A., Pinna, G.D. (2002). Verb-Noun Double
Dissociation in Aphasic Lexical Impairments: The Role of Word Frequency and Imageability. Brain and
Language, 81, 432-444.

Spelling impairment in Italian dyslexic children during the early phases of literacy acquisition
Paola Angelelli 1,2 , Alessandra Notarnicola 1 , Anna Judica 2 , Donatella Spinelli 2, 3 , Pierluigi Zoccolotti 2, 4 and
Claudio Luzzatti 5
1 2 3
Dipartimento di Psicologia, Università di Bari; Centro Ricerche, IRCCS Fondazione Santa Lucia, Roma;
Istituto Universitario Scienze Motorie, Roma; 4 Dipartimento di Psicologia, Università di Roma “La Sapienza”;
5
Dipartimento di Psicologia, Università di Milano-Bicocca
Background
Very few studies have documented developmental spelling disorders in languages with shallow orthography. In a
recent study on spelling deficit in Italian dyslexic children (Angelelli et al., 2004), a predominance of surface
dysgraphia was found (i.e.phonologically plausible spelling errors on words with unpredictable transcription, and
much better performance on regular words and nonwords). Moreover, the comparison of the reading and spelling
performance showed a fully coherent pattern of impairment, with most children suffering from surface dyslexia
and surface dysgraphia.
Both theoretical work and experimental data (see Tressoldi et al., 1996 for Italian) suggest a later acquisition of
the lexical route with respect to the sub-word-level one. Consequently, the phenomenology of developmental
dysgraphia might assume different features at different ages. In particular, we assume that, in the early phase of
spelling acquisition, developmental dysgraphia may be characterized by delayed acquisition and fragility of the
sub-word level routine, together with much more severe and long-lasting deficit of the orthographic lexical
acquisition.
Aim of this study is the description of the writing impairment of dyslexic children during the early phases of
literacy acquisition. To the extent that a prevalent but inefficient use of the sub-word-level routine characterizes
the initial phases of writing acquisition, we predict, in young subjects, a spelling performance reflecting
inefficient use of both the sub-lexical and lexical routes (i.e, the presence of a general impairment in all tasks
with distributed rate of phonologically plausible and phoneme-to-grapheme conversion errors). In order to test
the hypothesis of a progressive evolution to a more selective damage of the lexical route (surface dysgraphia),
we studied the writing performance in a further group of dyslexic children attending the 5 th grade and compared
their spelling performance to that of both age-matched controls and 3 rd grade dyslexic children.
Methods
Twenty-one dyslexic children were studied: 13 attending the 3 rd grade and 8 attending the 5 th grade. Writing was
investigated by means of a spelling test that included regular words with one-sound-to-one-letter
correspondence, regular words requiring syllabic conversion rules, words with unpredictable transcription. The
spelling task included nonwords in order to examine the efficiency of the sublexical route. The dyslexics’
spelling performance was compared to that of two groups of age-matched normal readers: 53 subjects attending
the 3 rd grade and 50 attending the 5 th grade.
Results
The 3 rd grade dyslexic participants performed poorly, with respect to the age-matched control sample, in all
sections of the spelling test regardless of orthographic regularity (p

On the contrary, control subjects showed significant reduction of both phonologically plausible and syllabic
conversion errors with age (p

How autonomous is orthographic encoding? The role of syntax
Tobias Bormann & Gerhard Blanken
Psycholinguistik, University of Erfurt
Introduction
Based on detailed studies of patients, it has been suggested that written production may occur independently of
phonological processes. Firstly, patients have been observed with preserved written naming despite severe oral
naming difficulties (e.g., Blanken, 1990; Bub & Kertesz, 1982; Hier & Mohr, 1977). Secondly, patients have
been reported who make different semantic errors in subsequent oral and written naming tasks (Miceli et al.,
1997, 1999; Rapp et al., 1997). From this observation of superior written naming of single words, the idea of
autonomous orthographic production processes has been generalized to the production of phrases and sentences
(e.g., Caramazza, 1997).
A review of the literature, however, demonstrates that this strong concept of orthographic autonomy receives
weak support at best. Only two cases have been reported with better preserved written production of phrases and
sentences. The majority of patients with superior written naming hardly demonstrate any preserved syntax in the
written modality. Moreover, EB (Levine et al., 1982), who has been claimed to have lost all inner phonology and
still be able to write sentences, is now considered “severly dysarthric” (Caplan & Waters, 1995, p. 192). The
other patient, GG (Miceli et al., 1983), was tested on a very limited set of items only, and few details are
reported.
Patient Background
We will present data from a patient with superior written naming: AB, a former journalist born 1958, suffered
from a CVA and presented with fluent aphasic speech and very mild signs of apraxia of speech. His written
naming was much better than his oral naming (55/60 versus 30/60 correct). He was almost completely unable to
write nonwords to dictation but did not show semantic errors or particular problems with verbs and function
words. In writing as in speaking, he named actions as well as objects, and correctly wrote 25 of 30 function
words to dictation. AB thus did not suffer from deep agraphia.
Methods and Results
AB was tested on four tasks of phrase and sentence production: description of the Cookie theft-picture,
production of sentences, noun phrases and idioms. For the Cookie Theft-picture, AB’s oral production was
paragrammatic: a total of 267 words include all syntactic categories of words and numerous complete sentences.
A high rate of phrases were correct, verbs appeared in second position in main clauses and sentence final
position in sub-clauses. In contrast, his written production was limited to 13 nouns lacking words from all other
syntactic categories.
In a sentence production task, AB produced 8 grammatically correct sentences for 10 pictures in speaking. In
writing, he seemed to copy his previous oral response. Nevertheless, he produced only two correct sentences
which bore great similarities with his oral response.
In a task of naming a picture with the word’s determiner, AB chose the wrong determiner for six words. In a
subsequent written response, he used the wrong determiner in five of these instances. Thus, he again seemed to
copy his previous oral response instead of writing independently.
Finally, AB was dictated idioms to write. He was unable to do so writing only nouns although he acknowledged
that he knew these idioms.
Discussion
In neither task did AB show better results in the written production at the level of phrases and sentences. Instead,
whenever possible, AB seemed to copy his oral response. This is unexpected if we assume complete
independence of orthographic production processes from phonology. AB showed superior written to oral lexical
access and used a pencil in his spontaneous communication frequently. Written superiority was, however,
limited to concrete nouns. His written production was extremely poor in syntactic organization in contrast to his
paragrammatic oral output. This fits with previous reports of patients with preserved written naming.
Thus we suggest that the notion of completely autonomous orthographic processes is premature and not
supported by the available data. Instead, we will claim that syntactic processes as employed in writing sentences
may depend on intact phonological output processes and that written autonomy may be restricted to single word
production, an idea advanced earlier by Bub and Kertesz (1982; cf. Margolin, 1984; Tainturier & Rapp, 2000).

Pronoun-antecedent dependencies in Broca’s aphasia:
Evidence for delayed syntactic structure building
Petra Burkhardt, Esther Ruigendijk, Sergey Avrutin, & Maria M. Piñango
Yale University, UIL/OTS Utrecht University
We present a study investigating the implementation of pronoun-antecedent relations in Dutch Broca’s aphasia
patients during online sentence comprehension. Two main findings are reported: i) Broca’s aphasia patients are
capable of forming pronoun-antecedent dependencies and ii) they successfully build syntactic structure, but in a
delayed manner.
Pronouns (her, him) are referentially dependent and must establish a link with another entity (their antecedent) to
be fully interpretable. It is proposed that this link is established at the level of discourse. One model that captures
the formation and maintenance of discourse entities and the establishment of pronoun-antecedent dependencies
is the Syntax Discourse Model (Avrutin,1999; Burkhardt,2004). Within this model, two levels of representation
are involved in pronoun interpretation: syntax and discourse. The syntactic structure is built, and the presence of
a Determiner-node triggers the creation of an information unit in discourse. A pronoun projects a referentially
deficient discourse unit and must hence form a dependency with its antecedent to receive referential content.
Crucially, during comprehension, syntactic structure must be fully formed for a discourse referent to be
introduced.
Avrutin (2004) suggests that, in principle, the language system allows for an alternative, non-syntactic way of
introducing discourse referents. Specifically, the conceptual structure can function as the source for discourse
referents. Normally this option is highly restricted and possible only in so-called ‘special registers’ (colloquial
speech, newspaper headlines) because the syntactic system provides the most economical, automatic way of
performing this function. However, if due to brain damage, syntactic structure building is delayed (e.g. the
‘Slow-Syntax Hypothesis’ (Piñango,2002)), the alternative, non-syntactic way of introducing discourse referents
may be used. This serves to explain previously reported behavior of Broca’s aphasia patients: online, they show
aberrant priming for coreferential pronouns immediately after the pronoun (cf. Love et al.,1998), and offline,
they often allow for a reflexive interpretation of pronouns (cf. Grodzinsky et al.,1993; Piñango,2002; Ruigendijk
et al., in press).
The erroneous offline performance is explained by the availability of two competing sources of interpretation:
syntax-driven or extra-syntactic. The Syntax Discourse Model assumes that syntactic information must be in
place for a discourse referent to be formed. However, in the presence of slow syntactic structure formation,
syntactic information is not readily available to rule out incorrect antecedents and discourse referents are then
introduced from an extra-syntactic source, leading to the incorrect interpretation. The online performance is
explained by positing that in the absence of a timely syntactic structure, coreference information (introduced
extra-syntactically) allows for an otherwise incorrect antecedent.
We address the issue of slowed-down syntactic structure formation by investigating interpretation of pronouns
with quantified and coreferential antecedents. These pronouns surface in the same structural position, but the
nature of their antecedents yields distinct pronoun-antecedent dependencies, which has consequences for
sentence processing: the availability of a coreferential antecedent like the hairdresser in (1) allows for a
dependency with a uniquely identifiable individual; the availability of a quantified expression – like everyone –
as antecedent (2) demands a quantified interpretation that requires access to a set:
De kapperi zegt dat de klant hemi bezoekt als het buiten koud en stormachtig is.
“The hairdresser says that the client visits him when it is cold and stormy outside.”
Iedereeni zegt dat de klant hemi bezoekt als het buiten koud en stormachtig is.
“Everyone …”
We examine pronoun interpretation using the cross-modal lexical decision interference paradigm. This paradigm
measures processing cost at discrete points in time as comprehension progresses. Subjects listen to a sentence
and perform a lexical decision task: at a predetermined point, a letter string appears on a computer screen in front
of the subjects. Subjects must decide whether or not the string represents a real word of Dutch by pressing a
response button. The reaction time (RT) to this decision is measured.
In a previous study (Burkhardt et al.,2003a), unimpaired Dutch college-aged subjects showed a significant
increase in RT for the quantified over the coreferential condition immediately after the pronoun (*p

indicates that quantified interpretation of a pronoun in Dutch – which requires access to a set – is more costly
than coreferential interpretation.
Based on the Slow-Syntax Hypothesis, we predict that Broca’s aphasia patients show the same pattern
(quantified>coreferential) as unimpaired subjects but in a delayed manner. We created minimal pairs as in (1) &
(2) and measured interference at three positions: 400msecs before the pronoun (control position), and 100msecs
and 600msecs after the pronoun. The timing of this later point follows from studies of long-distance
dependencies (Love et al.,2001; Burkhardt et al.,2003b) and reflexive interpretation (Burkhardt &
Piñango,2002). We hypothesize that unlike unimpaired controls, Broca's aphasia patients should not show an
interference effect at 100msecs after the pronoun, but further downstream, at the point where syntactic structure
is hypothesized to be formed in the Broca’s system.
We tested 10 age-matched control subjects and two Broca’s patients in the three experimental positions
(additional participants are currently being tested). The results obtained confirm our predictions. Both groups
show no significant difference in the control position. The control subjects pattern with college-aged subjects:
they show a higher RT for quantified interpretation over coreferential interpretation at 100msecs after the
pronoun. In contrast, the two Broca’s patients do not show such an effect at 100msecs. However, they show an
interference effect at 600msecs, where the control group does no longer show a significant difference. This
indicates that syntactic structure has been integrated at this point in the Broca’s system.
Table 1. Paired comparisons for mean-RTs to quantified vs. coreferential interpretation
400msecs before 100msecs after 600msecs after
Controls NS *p

Negative Markers and Bilingual Aphasia:
Evidence from Italian and Northern Italian Dialects
Paolo Chinellato
University of Venice
This paper deals with agrammatic production of negation in five agrammatic patients. Before the injury they
spoke both Italian and a Northern Italian dialect (Northern Vicentino, Venetian). The present study describes the
performance of the patients on several tasks:
- delayed repetition: patients were asked to repeat a sentence with a preverbal and a postverbal negative
marker in both languages:
(1) It. Gianni non beve mica vino
(John not drinks not wine)
- sentence completion: patients were asked to complete a sentence in both languages:
(2) It. Gianni vive a Londra, ma è italiano; Gianni vive a Londra ma ____ è inglese
(John lived in London, but is Italian; John lives in London, but ____is English)
- guided production: patient were asked to produce a sentence with at least one negative marker.
- sentence anagram ordering tasks: patients were asked to order sentences with two negative markers
Moreover a comparison between sentence negation and small clause negation has been investigated:
(3) Non ritengo Gianni degno di fiducia
(Not consider-1 st sing. John trust-worthy)
(4) Ritengo [Gianni non degno di fiducia]
(Consider – 1 st sing. [John not trust-worthy])
In the presentation I will show that:
- preverbal sentence negation (it. non /dial. no) is always omitted in both languages.
- postverbal sentence negation (it. mica/ dial. mia) is always spared in both languages.
- negation of small clause (it. non / dial. n�) is always spared in both languages.
These results are consistent with previous studies on negation production in aphasia (Bastiaanse, Rispens & van
Zonneveld (2000)) and support linguistic theories on the different status of the two negative markers: the clitic
status (X°) of the preverbal negative marker (which is omitted as subject and object pronominal clitics) and the
XP status of the postverbal negative markers (Pollock 1989, Belletti 1990, Zanuttini 1997, Cinque 1999 among
others).
Moreover, negation of the small clauses, although semantically similar, is structurally different from the sentence
negation. Our data support linguistic proposals (Cardinaletti & Guasti 1993, among others) on such a different
status: the latter projects a NegP higher than TP whereas the former is expressed by an AdvP.

References
Bastiaanse R., Rispens J., van Zonneveld R., (2000). Verb retrieval, verb inflection and negation in agrammatic
aphasia. In Bastiaanse R. & Grodzinsky J. (eds.) Grammatical disorders in aphasia, A neurolinguistic
perspective. Whurr, London /Philadelphia.
Belletti, A., (1990). Generalized Verb Movement. Rosemberg & Sellier, Torino.
Cardinaletti, A. & Guasti, M.T. (1993). Negation in epistemic small clauses. In Probus 5: 39-61
Cinque, G., (1999). Adverbs and Functional Heads. A Cross-Linguistic Perspective. Oxford University Press,
New York – Oxford.
Pollock, J.-Y., (1989). Verb Movement, UG, and the Structure of IP. In Linguistic Inquiry, 20, pp 365-424.
Zanuttini, R. (1997). Negation and Clausal Structure. A Comparative Study of Romance Languages. Oxford
University Press, New York /Oxford.

WORD ORDER AND FINITENESSIN PARKINSON’S DISEASE:
A comparison with Broca’s aphasia.
Katrien S.F. Colman, AZ Maria-Middelares-Sint-Jozef, Ghent, Belgium.
Prof.Dr.Roelien Bastiaanse, University of Groningen,
Dept. of Linguistics, The Netherlands.
Introduction:
� In addition to prominent motor symptoms, Parkinson’s disease (PD) patients exhibit affected linguistic processing
(Lieberman, 2000; Lieberman, et al., 1990, 1992; Grossman, 1999; Grossman et al., 1991, 1992, 1993, 2002a, 2002b, among
others)
� Aim of this study: gain a better understanding of the grammatical abilities in PD.
Linguistic background:
� This study addresses two grammatical aspects and were studied in Dutch:
1. the influence of word order on comprehension and sentence construction.
2. inflection for tense and agreement in verb production.
� Government and Binding theory (GB, Chomsky, 1981).
� The canonical word order in Dutch is S-O-V (Koster, 1975):
De jongen kust het meisje.
The boy kisses the girl.
Agent Theme
� Non-canonical sentence structure (i.e. passive) derived by a grammatical transformation rule:
Het meisjei wordt
The girl is
(The girl is kissed by the boy.)
door de jongen ti by the boy
gekust.
kissed.
Theme Agent
� Canonical word order S-O-V and the Verb Second rule:
Matrix clause without model verb or auxiliary
De jongen aaiti de hond ti. The boy strokes the dog.
Matrix clause with modal verb
De jongen wili de hond aaien ti. The boy wants the dog (to) stroke
(The boy wants to stroke the dog.)
Research question:
Do PD patients exhibit a selective linguistic impairment as is the case with Broca’s aphasic patients?
Methods:
1. Subjects:
6 Dutch speaking non-demented, right handed PD patients (all on-medication), 13 Dutch speaking Broca’s aphasic
patients (participated in the study of Bastiaanse and Edwards, in press), 40 Dutch speaking non-brain-damaged speakers
(participated in the study of Bastiaanse and Edwards, in press).
2.
Results:
Materials: subtests from the Dutch test battery WEZT:
(1) Sentence comprehension, (2) Sentence anagrams with pictures, (3) Filling in finite verbs and infinitives
In general, the subjects with Broca’s aphasia performed worse than the healthy subjects of the control group and
the PD patients on each of the three tasks. The scores of the latter groups (PD versus controls) differ minimal for
sentence comprehension and sentence anagrams and do not differ for filling in verbs (see table of results).
Discussion:
� The Broca’s aphasic subjects chose the reverse role, as would be predicted from the literature on agrammatic
comprehension. Although the sentence comprehension in patients with PD was poorer than the abilities of the
normal subjects, it did not approach, in terms of severity the comprehension impairment associated with Broca’s
aphasia. PD patients performed equally on all of the sentence types on the comprehension task..
� On the sentence anagram task, the Broca’s aphasic patients encountered more problems with passive than with
active sentences, while reversibility did not play a role. The PD patients scored equally well on passives as on
actives.
� The Broca’s aphasic subjects and the PD patients made more semantic errors than expected in the task for filling in
infinitives. However, the Broca’s aphasic patients performed as expected in the task for filling in finite verbs, they
produced predominantly grammatical errors. The PD patients on the other hand, do not encounter problems with

the inflection of verbs in sentence context, but do have problems with the semantic processing of verbs in sentence
context.
Conclusion:
� Relative to Broca’s aphasia, PD patients do not evidence a selective linguistic dysfunction. The language
abnormalities in PD seem to be non-specific and are relative to controls.
� A question that remains to be answered is why PD patients score better than Broca’s aphasic patients, but still
perform poorer than healthy control subjects?
References:
Bastiaanse, R. & Edwards, S (in press) Word order and finiteness in Dutch and English
Broca’s and Wernicke’s aphasia.
Chomsky, N. (1981) Lectures on government and Binding. Dordrecht: Foris.
Grossman, M., (1999), Sentence processing in Parkinson’s disease. Brain and Language,
40, 387-413.
Grossman, M., Carvell, S., Gollomp, S., Stern, M.B., Reivich, M., Morrison, D., Alavi, A., &
Hurtig, H.I. (1993) Cognitive and physiological substrates of impaired sentence
processing in Parkinson’s disease. Journal of Cognitive Neuroscience, 5, 480-498.
Grossman, M., Carvell, S., Gollomp, S., Stern, M.B., Vernon, G., & Hurtig, H.I. (1991)
Sentence comprehension and praxis deficits in Parkinson’s disease. Neurology,
41, 1620-1628.
Grossman, M., Carvell, S., Stern, B., Gollomp, S., & Hurtig, H. (1992) Sentence
comprehension in Parkinson’s disease: The role of attention and memory. Brain and
Language, 42, 347-384.
Grossman M., Lee C., Morris J., Stern M.B., & Hurtig H.I. (2002b) Assesing resource
demands during sentence processing in Parkinson’s disease. Brain and Language, 80,
603-616.
Grossman M., Zurif E., Lee C., Prather P., Kalmanson J., Stern M.B., & Hurtig H.I. (2002a)
Information processing speed and sentence comprehension in parkinson’s disease.
Neuropsychology, 16(2), 174-181.
Koster, J. (1975) Dutch as an SVO language. Linguistic Analysis, 1, 111-136.
Lieberman, M.D. (2000) Intuition: A social cognitive neuroscience approach. Psychological
Bulletin, 126:1, 109-137.
Lieberman, P., Friedman, J., & Feldman, L.S. (1990) Syntax comprehension in Parkinson’s
disease. Journal of Nervous and Mental Disease, 178, 360-366.
Lieberman, P., Kako, E., Friedman, J., Tajchman, G., Feldman, L., & Jiminez, E. (1992)
Speech production, syntax comprehension, and cognitive deficits in Parkinson’s disease. Brain and Language, 43,
169-189.
Appendix:
TABLE of results:
The means and standard deviations for the Control subjects,
the Broca’s aphasic and PD subjects on the three subtests.
Max Cutt-off Control group Broca’s aphasia PD patients
Sentence comprehension 40 39 39,77
(0,42)
Sentence anagrams 20 20 20
(-)
Filling in verbs 20 18 19,22
(0,50)
29,92
(8,57)
13,15
(5,03)
13,08
(3,52)
Note Value is the mean, value in parenthesis is the standard deviation.
37,67
(1,64)
19,84
(0,41)
18,67
(0,52)

Selective Impairment of Nouns and Verbs in Aphasia: a Word Retrieval Task in a Sentence
Context
Davide Crepaldi 1 , Silvia Aggujaro 1 , Lisa Saskia Arduino 2 , Giusy Zonca 3 ,
Graziella Ghirardi 4 , Maria Grazia Inzaghi 5 , Mariarosa Colombo 6 ,
Gennaro Chierchia 1 , and Claudio Luzzatti 1
1 Department of Psychology, University of Milano-Bicocca
2 Department of Psychology, University of Urbino
3 Maugeri Foundation, Montescano Medical Center, Pavia
4 Rehabilitation Unit, G. Salvini General Hospital, Passirana
5 Rehabilitation Unit, Villa Beretta, Costamasnaga
6 Rehabilitation Unit, Legnano General Hospital
INTRODUCTION
Several authors described cases of dissociated impairment in naming nouns and verbs. There are four accounts of this dissociation:
(i) patients may have purely lexical damage, which selectively affects verbs or nouns at a late stage of the linguistic processing
(phonological or orthographic lexicons) (Rapp & Caramazza, 2002); (ii) the damage affects a lexical device, either at an ortographicphonological
modality-specific level (the lexeme; Levelt et al., 1999) or at a unitary lexical-syntactic level (the lemma) (Berndt et al.,
1997); (iii) N-V dissociation arises from semantic damage, causing increased sensibility to imageability (Bird, Howard & Franklin,
2000); (iv) N-V dissociation is due to syntactic damage (Friedmann & Grodzinsky, 2000). Moreover, many other aspects of verbspecific
impairment have been discussed in literature (Bastiaanse et al., 2002).
To disentangle imageability and grammatical class effects, a new task was developed allowing to elicit nouns and verbs with
identical imageability ratings in a sentence context. The results obtained will permit to address the following three questions: Does
imageability play a role in determining N-V dissociation? If so, is imageability the unique cause of dissociation? If there is
additional damage, at which level of linguistic processing does it take place?
METHODS
Subjects
Twelve Italian aphasic patients and 11 normal controls participated in the study.
Tasks
Nouns and Verbs Retrieval in a Sentence Context (NVR-SC). Forty-five pairs of sentences denoting the same event, either using a
noun or the corresponding verb (e.g. the evasion/to evade) were used. The first sentence was presented in complete form, while a
gap was left in the second sentence, to be completed with the target word. For each pair of sentences two different conditions were
employed, one triggering a verb and one triggering a noun.
E.g. V→N: The prisoner was dreaming to evade
The prisoner was dreaming the ………
N→V: The prisoner was dreaming the evasion
The prisoner was dreaming to ………
It is important to observe that 27 of the 45 nouns used were morphologically derived from (and hence more complex than) verbs.
The performance in the NVR-SC task was compared to that obtained on a classic picture naming task (50 nouns and 50 verbs).
Statistical methods
Logistic regression analysis (LRA) was applied to the profiles of the patients, making it possible to study the effects of the lexicalsemantic
variables in univariate and multivariate linear models.
RESULTS
Picture naming task
All patients with predominant verb deficit also have an imageability effect. In eight patients the grammatical class effect was no
longer significant after introducing imageability in the statistical design (bivariate LRA).
NVR-SC task
Two of the verb-impaired patients in the picture naming task maintained the predominant verb deficit also in the NVR-SC task. In

eight patients, the difference between nouns and verbs was no longer significant. In two patients, a paradoxical dissociation (V>N)
emerged.
Group analysis: performance on nouns and verbs across naming tasks (Figure 1)
Patients named actions in the NVR-SC task better than in the picture naming task (58% correct versus 37%; p

Verb second in Dutch sentence production: an fMRI study
D.B. den Ouden, Y.R.M. Bastiaanse, J.M. Hoogduin, R.P. Maguire & L.A. Stowe
School of Behavioural and Cognitive Neurosciences, University of Groningen
As reported by Bastiaanse et al. (2002), agrammatic speakers (Broca’s aphasia) have difficulty with the
production of finite verbs in main clauses, in which these finite verbs are placed in the second constituent
position in the sentence. This has been argued to result from (non)canonicity of word order, or more directly
from the extra cost of a movement operation which derives the main clause word order from a more basic
structure. In Dutch, the base position of verbs is argued to be sentence final (Koster, 1975). This ordering is
maintained at surface level for nonfinite verbs, as in (1), and for finite verbs in embedded clauses, as in (2):
1) De jongen wil een boek lezen
(lit.) The boy wants a book to read
2) Ik zie dat de jongen een boek leest
(lit.) I see that the boy a book reads
For Dutch main clauses, verbs move from this base position to the second position in the sentence (Den Besten,
1977), as in (3).
3) De jongen leest een boek
(lit.) The boy reads a book
We have conducted an fMRI study into production of Dutch main clause and embedded clause word order, thus
capturing the neural correlates of ‘verb movement’ and the particular difficulty it poses for agrammatic speakers.
We had two hypotheses about the results: 1) production of main clauses with finite verbs (verb second) will
correlate with increased brain activation, relative to production of embedded clauses with finite verbs; and 2) this
activation will show up in regions related to syntactic processing in other work, in particular Broca’s area (BA
44/45; see Caplan, 2001), a region traditionally claimed to be involved in agrammatic aphasia (Grodzinsky,
2000).
19 Healthy native speakers of Dutch performed sentence completion tasks, covertly, on the basis of
pictures that they were shown, depicting transitive actions. In alternating blocks, verb-second (V2) main clauses
(4a) were elicited after the cue Ik lach want ... [picture], while verb-final (Vf) embedded clauses (4b) were
elicited after the cue Ik lach omdat ... [picture]. The two complementizers want and omdat both mean ‘because’,
but they are subcategorized for different clause types, coordinate and subordinate, respectively.
4a) Ik lach want de jongen leest een boek (coordinate)
(lit.) I’m laughing because the boy reads a book
4b) Ik lach omdat de jongen een boek leest (subordinate)
(lit.) I’m laughing because the boy a book reads
Scanning was carried out on a 1.5 T Siemens Vision scanner, with TR 3 seconds, slice thickness 3.5 mm and 38
slices/volume. A 30 second block consisted of 9 trials. Between blocks, a fixation point was shown for 30
seconds. Each participant performed 3 runs of 6 blocks per experimental condition. Data were normalized to a
template from the Montreal Neurological Institute, based on 152 brains, and smoothed with a 10 mm filter.
Preprocessing and statistical analyses were carried out with SPM99.
Figure 1 shows the result of the subtraction of Vf activation from V2 activation, uncorrected for
multiple comparisons (p < 0.001). There is increased brain activation in the verb-second condition, compared to
the activation in the verb-final condition. By contrast, there are no brain areas that show more activation in the
verb-final condition than in the verb-second condition. This confirms our hypothesis 1.

Figure 1 Cortical activation associated with production of finite main clauses (V2)
minus production of finite embedded clauses (Vf).
With respect to our second hypothesis, we do not find the expected activation increase in Broca’s area,
related to the V2 condition. We do find activation at 7 mm distance from left hemisphere BA 45 (coord. -46, 24,
12), but this is only significant at the voxel-level, uncorrected for multiple comparisons (z = 3.48; p = 0.000).
The two areas that show the strongest activations are in the right hemisphere inferior temporal lobe (BA 21;
coord. 64, -12, -6; z = 4.73) and the left hemisphere prefrontal region (BA 8; coord. -30, 30, 42; z = 4.53).
Following correction for multiple comparisons, these contrasts are significant at p < 0.05, at the cluster-level.
BA 21 has been reported to be involved in the parsing of relatively complex syntactic structures (e.g.
Just et al. 1996), and in detection of semantic anomaly (Ni et al. 2000). This activation is usually bilateral, which
is indeed what we see in the uncorrected analysis (figure 1; coord. -62, -10, -18; z = 3.42; p = 0.000 at voxellevel).
Left hemisphere BA 8 has been shown to be activated during detection of syntactic anomaly (Ni et al.
2000).
In conclusion, these results show that production of verb-second sentences in Dutch is indeed correlated
with increased brain activation, relative to production of verb-final sentences. This sheds light on the problems
agrammatic speakers have with the verb-second construction. The activation is not directly linked to Broca’s
area, or at the least not restricted to this site, but distributed more widely, involving other areas related to
syntactic processing.
REFERENCES
Bastiaanse, R., Hugen, J., Kos, M. & Zonneveld, van R. 2002. Lexical, morphological, and syntactic aspects of verb
production in agrammatic aphasics. Brain and Language, 80, 142-159.
Besten, den H. 1977. On the interaction of root transformations and lexical deletive rules. Ms., MIT/Universiteit
van Amsterdam. Published in Studies in West-Germanic Syntax, 1989, dissertation, University of
Amsterdam.
Caplan, D. 2001. Functional neuroimaging studies of syntactic processing. Journal of Psycholinguistic Research,
30, 3, 297-320.
Grodzinsky, Y. 2000. The neurology of syntax. Language use without Broca’s area. Behavioural and Brain
Sciences, 23, 1-71 (with open peer commentary).
Just, M. A., Carpenter, P. A., Keller, T. A., Eddy, W. F. & Thulborn, K. R. 1996. Brain activation modulated by
sentence comprehension. Science, 274, 114-116.
Koster, J. 1975. Dutch as an SOV language. Linguistic Analysis, 1, 111-136.
Ni, W., Constable, R. T., Menci, W. E., Pugh, K. R., Fulbright, R. K., Shaywitz, S. E., Shaywitz, B. A., Gore, J. C.
& Shankweiler, D. 2000. An event-related neuroimaging study distinguishing form and content in
sentence processing. Journal of Cognitive Neuroscience, 12, 120-133.

The privilege of the first position in agrammatism, child language acquisition and headlines
Francesca Foppolo, Guasti, Maria Teresa, , Claudio Luzzatti & Claudia Caprin
Psychology Department – University of Milano-Bicocca
Our goal is to show that omission of articles in agrammatic and children speech and in headlines follow a
selective pattern either guided by the grammar of the language or by some option made available in natural
languages and not competing with language particular rules. By looking at the use of articles in Italian
agrammatism and child speech and in Italian headlines, we show that articles are mainly dropped from the first
clausal position. This privileged status of the first position constitutes an option that every languages can avail
itself of (see Rizzi, 2002 for subject omission for first position) and made available by the non c-commanded
status of the first clausal position (see also Baauw, De Roo, Avrutin, 2002).
The data. Our investigation is based on the spontaneous speech of 5 Italian agrammatic patients, on a corpus of
1500 Italian headlines from different newspapers, on speaker judgments about possible headlines in Italian, and
on the available data from two studies on early Italian (Caprin, 2003; Guasti, Gavarrò, De Lange, & Caprin,
2004).
Methods. The spontaneous production data (from agrammatic patients and children) have been transcribed using
the CHAT system of CHILDES (MacWhinney, 1997) and analyzed. We have counted the omissions and the
presence of articles in obligatory contexts. Two types of contexts were separately examined: utterances lacking a
verb and utterances containing verbs. For verbal utterances, we have counted separately presence/omission of
articles in front of Nouns occurring in preverbal and post verbal position. Headlines have been collected and
classified according to the same criteria; judgments from Italian speakers have also been collected.
Results
Our first finding is that article omission is more frequent in utterances without a verb (M=.52; SD=.118) than in
verbal utterances (M=.16, SD=.08= (t-test= 6,54, p=.007) in the speech of agrammatic patients. The same holds
true for children speech (Х 2 = 7.037 p=0.03, Caprin, 2003). Moreover, it is fairly common for Italian headlines
to lack a verb (cervello in mostra “brain on exhibition”; viaggio tra i corpi senza nome “journey among the
bodies without name”) and in this case omission of articles is very frequent. When the verb is present, instead,
we find the following possibilities (as already observed by Stowell, 1999 for English headlines):
(1) a. Pitbull ferisce un bambino “Pitbull wounds a child”
b. Pitbull ferisce bambino “Pitbull wounds child”
c. (Milano) salva cucciolo di foca “Milan saves baby seal”
The article is omitted either both from subject and object position or only from subject position. Articles can be
omitted from the object position if the subject is null or is a proper noun. However, it is not possible to omit the
article from the object position if an article is present in the subject position, that is, (2) is not a possible
headline:
(2) *Un pitbull ferisce bambino “A pitbull wounds child”
This subject/object (or, more generally, clause initial/clause internal) asymmetry in the omission of articles is not
peculiar to headlines, but it can also be found in child speech, where the number of omissions tends to be higher
from the first position than from the clause internal position (Wilcoxon z=3,76, p

of working memory, or of the mechanism for the recovery of lexical items or of the articulatory planning system
(see also Kolk, 1995 for a proposal inspired by similar ideas). Limitations in working memory of readers
combined with the need for headlines to be effective can be the driving force for article omission in Italian
headlines. Finally, we also observed that article omission is higher in nonverbal utterances than in verbal ones.
This observation will also be related to limitations of the performance system.
Baauw, S. E. De Roo, S. Avrutin (2002) Determiner omission in language acquisition and language impairment :
Syntactic and discourse factors. Proceedings of the 26 th BUCLD, Sommerville:Cascadilla Press.
Caprin, C. (2003) La grammatica nei primi discorsi: uno studio trasversale sull’acquisizione della morfosintassi
in Italiano. Ph. D. Dissertation, Università di Pavia e Milano-Bicocca.
Kolk, (1995) A time-based approach to agrammatic production. Brain and Language, 50:282-303.
Rizzi, L. (2002) On the grammatical basis of language development: A case study. MS. Università di Siena.
Stowell, T. (1999) C-command effects in newspaper headlines. MS. UCLA.

Task dependence effects: What are you asking for?
A good context for elicitation of Wh questions
Maria Garraffa
University of Siena
Abstract
The aim of this paper is to show how could be relevant to have a good elicitation context for testing production
of aphasic patients. The main point is the comparison between two different contexts elaborated for elicitation
Wh question. The first was elaborated in elicitation production of Wh structures for investigated agrammatic
production (Friedmann, 2002). The second is referred to a task elaborated in acquisition’s research and here
used in agrammatic subjects (Thornton, 1995).
Task dependence effects are a necessary variable to consider in the production, especially in elicitation task, for
valuing if you are eliciting a Wh question. Two Italian agrammatic speakers were tested.
1 Question research
In the investigation of agrammatic production a hypothesis recently advanced is that agrammatic subjects could
not project the entire syntactic tree (Friedmann, 1995). So, functional material is not randomly omitted but there
is a selection guided by structural variables.
This hypothesis is called the Tree Pruning Hypothesis. For testing this hypothesis it was selected a type of
construction crucially involved in the high domain of the sentence, the interrogatives. This test was conducted
first with Hebrews, Arabic and English subjects (Friedmann, 2002). The main asymmetries arise from this article
was the difference between Yes/No construction and Wh construction, and a significant number of substitution
of Wh question for Yes/No question. This preference for Yes/No question was interpreted by the author in the
sense that the Yes/No questions don’t involve movement to CP layer and so they are preserved in agrammatism.
On the contrary Wh questions are disturbed for the request movement of the Wh operator in CP. In this paper we
want to address the problem of the task dependent effects, and the fact that the substitution for Yes/No question
is a preference derivate by this kind of effects and not by the pruning of the structure.
2 Materials and Methodology
In this experiments we are interested in the ability of building interrogative sentence starting from different
elicitation contexts. We tested two kinds of elicitation contests alternativly in two agrammatic subjects .
2.1 Subjects
Pathological subjects : The subject selected, C.D. and M.R. has diagnosticaded an aphasia by standard battery
for italian speakers (B.A.D.A.) (Miceli,e coll.), a focal lesion in the left emisphere in area FTP caused by an ictus
in the zone of the left artery. The most typicall features arised by the battery are the difficulty in production of
verbs, errors in the comprehension of passive sentences with inversion of thematic roles, non fluent production

with omission of functional materials and low scores in working memory tasks. This simptoms are normally
defined as agrammatic aphasia.
Subject Age Edu Post –
On set
Lesion
M.R. 42 18 2 Years FTP
Subject Age Edu Post -
on set
Lesion
C.D. 45 8 5 Years FTP
2. 2 Elicitation contexts
The task was introduced in a elicitation context inviting the subject to produce a question. It is like to create a
scenario of plausibile common life.
a. First context
In the elicitation task used in aphasia, subjects heard a sentence with a missing detail, signified by non-specific
words like “someone” or “something”, and were required to ask a question about the missing detail (see example
(1) ).
You think that your daughter comes back to school and she is telling you about her day.The point is that
she doesn’t tell you so many things and you want to know the missing details.
For example she said to you that someone kissed her and you want to know about this boy.
So you ask...
(1) Experimenter: Danny ate something. You want to ask me about this thing. So you ask….
B. Second Context
Target: ma dani axal?
What Danny ate?
(Friedmann,2002)
In the experiment for children the investigator tell a story to the child; this was the context for elicitation. At the
end of the story the investigator asked the child to ask at the puppet a detail of the story :
(2) Investigator:
“In that story, there was something the spaceman didn’t like.
Ask the snail What ”
(Thornton,1995)
In the task created for the patient Wh elements were inserted, like in children’s context.

(3) First Elicitation task:
“Gaia ate something. You want to ask me about this thing, so you ask …”
(4) Second elicitation task:
a. “Gaia ate something. You want to ask what. So you ask...
b. “Gaia went in some place . You want to ask where . So you ask...”
In (4) the Wh element is already present in the context and the subject doesn’t have to search in the lexicon for a
Wh element. The subject has to create the interrogative sentence substituing the indefinite of the first sentence
with the Wh and moving it in CP.
3 Results
The most relevant difference in the data obtained regards the substitution of Wh question with Yes/No question.
This typology of error is mostly present in one context and not in the context taking from acquisition research.
Substitution for Yes/No Friedmann context % Thornton context %
C.D. 18/24 75 9/24 37,5
M.R. 10/36 27,7 0/72 0
Total 28/60 46,6 9/96 9,3
Tab.1 C.D. and M.R. data on the production of Wh question
Also the control subject didn’t produce all Wh question with the first kind of context.
4 Discussion
So the production of Yes/No question instead of Wh question is not a problem of structure but a problem of task.
If we want to test the movement of the operator in the CP domain, this was the reason for selecting thie
interrogatives, is important to have the right context.
In the first type of context we are asking for an A’ dependeces , in the second context we are asking specifically
for the construction of a Wh dependecy, so we don’t have (or with have only marginally) a substitution with Yes
/No question. This data is quite important for hypothesis that has his empirical background in the differents
between Wh and Yes/No question.
In our opinion in Friedman’s test we are not eliciting the movement of the Wh operator in CP. Subjects answered
to the task producing an A’ depencies, so also a Yes/No question .
Bibliography
Friedman,N.
2002 Question production in Agrammatism: The tree Pruning Hypothesis. Brain & Language 80:160-187
Rizzi,L.
1997 The fine structure of the left periphery . In Elements of grammar, edited by Haegmann L., Dordrecht:
Kluwer

1999 On the position Int(errogative) in The Left Perifery of the clause. Manuscript University of Siena
Thornton, R..
1995 Referentiality and Wh-movement in Child English: Juvenile Delinkuency. Language acquisition 4.

QUESTION PRODUCTION IN AN ITALIAN AGRAMMATIC SPEAKER:
MINIMALITY EFFECTS
Maria Garraffa, University of Siena
garraffa@studenti.unisi.it
Antonino Grillo, University of Utrecht and University of Siena
Nino.Grillo@let.uu.nl
Research question
The goal of this presentation is twofold. Firstly we will present the results of a test on production of WH-
questions by an Italian agrammatic aphasic. Secondly we will discuss the results in the light of an approach to
agrammatic aphasia developed in Grillo (2003). The basic claim is that, due to a processing resource limitation,
aphasic’s syntactic representations could be impoverished of some features necessary for the normal
computation of “long distance” 1 chains.
Design of the study
a. The Subject
The subject selected, M.R., has diagnostic an aphasia by standard battery for Italian speakers (B.A.D.A.) (Miceli,
& Coll.), a focal lesion in the left hemisphere in area FTP caused by an ictus in the zone of the left artery. The
subject shows non-fluent production with omission of functional materials, difficulty in production of verbs,
errors in the comprehension of sentences with inversion of thematic roles. These symptoms are commonly
defined as agrammatic aphasia.
b. Material and methods
The task elaborated for the elicitation of question is taken from acquisition’s research (Thornton, 1995) and here
readapted for pathologies. The subject M.R. was invited to produce an interrogative sentence after a context
created for the elicitation.
Elicitation task:
a. “Gaia ate something. You want to ask what. So you ask...
b. “Gaia went in some place . You want to ask where . So you ask...”
In this context the Wh element is already present in the context and the subject has to create the interrogative
sentence substituting the indefinite of the first sentence with the Wh and moving it in CP.
Results
A series of clear asymmetries are registered in the M.R.’s performance with subject vs. object Wh-.
WH – Questions N Correct %
Who- subject 12 9 75
Who – object 12 0 0
What - object 12 9 75
TOTAL 36 18 50
Tab.1 Interrogatives produced by M.R. in the task
REFERENTIAL Wh- questions N Correct %
1
By “long distance” we refer to chains generated by movement of an NP over another one. In the case discussed
here Subject vs. Object WH-.

Tab.2 Interrogative with Which- operator produced by M.R.
i) While she shows close to normal performance with subject wh- (75/ 85% correct), her
performance degrades with the latter, reaching 0% of correct productions with who-objectquestions.
ii) Subject’s performance varies considerably also between who-object questions (0% correct) and the
same construction with what (75% correct).
Discussion of the data
Subject 24 20 83,3
Object 24 13 54,1
Tot 48 23 47,9
According to the Relativized Minimality (RM) approach (Rizzi, 1990), in a configuration like (1), a local
structural relation cannot hold between X and Y if Z is a potential bearer of the relevant relation and Z intervenes
between X and Y.
(1). . . X . . . Z . . . Y . . .
So for instance the local relation linking a moved phrase to its trace holds in (2), but not in (3)b 2 : the Wh
operator who intervenes between how and its trace in (3)b, since who is a potential bearer of the antecedent
relation, the required antecedent-trace relation fails, and the question is agrammatical.
(2)a How did you solve the problem?
b How did you solve the problem t?
(3)a I wonder who could solve the problem in this way
b * How do you wonder who could solve this problem t?
Rizzi (2001) shows that the possibility to form a chain between X and Y depends on the nature of the elements
involved and more specifically on the set of morphosyntactic features associated to each element.
(4 ) . . . X . . . Z . . . Y . . .
[ α,β,γ,δ, ε ] [α,β,γ,δ] [ α,β,γ,δ,ε]
In (4) a particular set of morphosyntactic features is associated with every node. Given this configuration, RM
allows the formation of a relation Σ between X and Y: the presence of the element ε suffices for RM to see the
difference between X and Z and therefore to authorize the movement of Y over Z (it is necessary to think about ε
as the distinctive feature of the particular head and the relevant relation we are considering, i.e. a [wh-] feature in
the head of a FocP, which if missed would imply a change of “class” of the relevant set from Argumental to
Quantificational).
Grillo (2003) claims that, a cause of their limited processing capacity, aphasics are not always capable to
maintain active the representation of all the morphosyntactic features normally associated with syntactic nodes.
Such impoverishment can involve some of the features needed for the computation of structures involving
dependencies over possible interveners.
2 Examples from Rizzi (2001).

(5) . . . X . . . Z . . . Y . . .
[α, β, γ] [α, β, γ] [α, β, γ]
X
This leads to expect an asymmetry in aphasic’s treatment of dependencies: those involving the “crossing” over
an NP by another one should be more problematic than those that don’t imply such crossing. The asymmetry
between aphasic comprehension of object (which implies crossing of the subject NP) vs. subject movement
(which doesn’t imply crossing of any NP) confirms this hypothesis on the comprehension side. The present study
indicates that a minimality effect of the kind described above can be at the base of the asymmetries found. The
main difference between the constructions tested here is related, in fact, to the presence or absence of a potential
intervener between the moved element and its trace. Subject questions are correctly predicted not to be
problematic, no NPs intervenes between the moved subject and its trace. No minimality effects are expected
even if the syntactic representation is impoverished.
(6 ) Subje ct wh - :
A.
Normal Representation
[ N , + Animate , φ s,nom , θ 1 , + Wh] [N, + animate,φs,nom,θ1 ,+Wh]
Chi < chi > ti ha baciato?
B.
Agrammatic Representation
[ N, + Animate] [N, + Animate]
Chi < chi > ti ha baciato?
This situation is reversed with object wh-, these construction require the establishment of a dependency over an
intervening NP. Under the present approach, in case of an underspecification of the morphosyntactic features we
will expect a minimality effect and a failure in the attempt to produce the sentence.
(7 ) Object wh - :
Normal Representation
A. [ N , + Animate , φ s,acc , θ 2 , + Wh ] [ N, + animate,φs,nom,θ1] [N,+ Animate , φ s,acc , θ 2 , +
Wh ]
B.
Chi pro hai baciato ?
Agrammatic Representation
[N,+ animate [N,+ animate] [N,+ animate]
Chi pro hai baciato < chi > ?
X
Given the agrammatic representation schematized in (7b) RM will block object movement over the intervening
pro subject. The presence of the sole categorical and semantic features does not suffice for RM to “see” the
difference between the elements involved.

However this could not be the case with the third kind of constructions we had taken into consideration: Objectwhat-questions,
which don’t seem to be problematic. It is natural to think that the presence of the [-animate]
feature associated with the object opposed to the [+animate] carried by the subject suffice for RM to distinguish
between the two NPs and “authorize” the crossing of the subject.
Normal Representation
C. [ N, ¬ Animate , φs,acc , θ 2 , + Wh ] [ N, + animate,φs,nom,θ1] [N,¬ Animate , φ s,acc , θ 2, +
Wh ; ]
References
Che cosa pro hai mangiato < che cosa >?
Agrammatic Representation
D. [ N, ¬ Animato ] [ N, + Animato ] [N, ¬ Animato]
Che cosa pro hai mangiato < che cosa > ?
Friedmann, N. (2002) Question production in Agrammatism: The tree Pruning Hypothesis. Brain & Language
80:160-187
Garraffa, M. (2003) Produzione di interrogative in un parlante italiano agrammatico: alcune asimmetrie. Tesi di
Laurea, Università di Siena.
Grillo, A. (2003) Agrammatic comprehension between processing and representation, Minimality Effects.
Unpublished Tesi di Laurea, Universita’ di Siena
Grodzinsky, Y. 2000a The neurology of syntax. : language use without Broca’s area. Behavioural and Brain
Science 23 (1) : 47-117.
2000 b Syntactic dependecies as memorized Sequences in the Brain. Paper presented at the Tennet XI Annual
Conference on Theoretical and Experimental Neuropsychology. Montreal
Hickok , G. e Avrutin , S.1995 Comprehension of WH questions by two agrammatic Broca’s aphasics. Brain &
Language, 51: 10-26
Miceli, G., Burani, C., Capasso, R. e Laudanna, A. 1994 Batteria per l’analisi dei deficit afasici - B.A.D.A. .
CEPSAG, [ 2 nd edizione
Rizzi, L. (1990) Relativized Minimality, MIT Press.
(2001) Locality and left periphery, In Belletti, A., ed. (2002) Structures and Beyond, OUP
Thornton, R. (1995) Referentiality and Wh-movement in Child English: Juvenile Delinkuency. Language
acquisition 4.

Minimality effects in agrammatic comprehension
Antonino Grillo, University of Siena, Utrecht University
The goal of this presentation is to introduce a new hypothesis on the origin of some of the most
common pattern of comprehension in agrammatic Broca’s aphasia. A large amount of attention has
been paid in the last decades to the so-called capacity limitation approaches, while important critics
have been moved to any representational approach to the topic, especially regarding the “variation”
problem. Less attention has been devoted to the problems raised by “processing” approaches, one this
problems we can call the “complexity” problem: processing approaches state that aphasics
comprehension is poor for more complex constructions, but nothing is said on the nature of such
complexity. If we want build a complete theory of aphasia we need to know also why certain
constructions are more complex than others. We believe that the only way to reach this goal is to
define the possible effects of a processing capacity limitation at the representational level getting rid of
any generic definition of complexity.
Our proposal in such sense is that a reduced processing capacity can generate an underspecification at
the syntactic level, more exactly the inaccessibility (alternatively the fast decay in the sense of Kolk,
1995) of part of the morphosyntactic features normally associated with syntactic nodes. This
impoverishment can in turn give rise to possible minimality effects, which would explain some of the
most typical asymmetries in aphasic’s comprehension between subject and object constructions and
more generally their poor performance with constructions involving the establishment of a dependency
over a possible intervener.
2. Summary of the data to explain
We will discuss and explain the following asymmetries in agrammatic comprehension:
Subject/object relatives
Subject/object clefts
Active/passives
Passives/unaccusatives
Subject/object control
Datives/double object constructions
Subject/object D-linked questions
SVO/OSV-OVS in Hebrew
3. Theoretical framework
According to the Relativized Minimality (RM) approach (Rizzi, 1990), in a configuration like (1), a
local structural relation cannot hold between X and Y if Z is a potential bearer of the relevant relation
and Z intervenes between X and Y.
(1) ....X .…Z .…Y.…
So for instance the local relation linking a moved phrase to its trace holds in (2), but not in (3)b 3 : the
Wh- operator who intervenes between how and its trace in (3)b, since who is a potential bearer of the
antecedent relation, the required antecedent-trace relation fails, and the question is ungrammatical.
(2) How did you solve the problem t ?
(3)a I wonder who could solve the problem in this way
b * How do you wonder who could solve this problem t ?
Rizzi (2001) shows that the possibility to form a chain between X and Y depends on the nature of the
elements involved and more specifically on the set of morphosyntactic features associated to each
element.
3 Examples from Rizzi (2001).

(4 ) . . . X . . . Z . . . Y . . .
[ α,β,γ,δ, ε ] [α,β,γ,δ] [ α,β,γ,δ,ε]
In (4) a particular set of morphosyntactic features is associated with every node. Given this
configuration, RM allows the formation of a relation Σ between X and Y: the presence of the element
ε suffices for RM to see the difference between X and Z and therefore to authorize the movement of Y
over Z.
3. Main claim
If for any reason some of the features associated with the syntactic nodes are deleted we will expect a
reduction of the possibility to establish relationships over intervening elements, because RM will
block it, as there is no feature mismatch between X and Z.
Our claim is that, a cause of their limited processing capacity, aphasics can’t access the representation
of the features needed for the computation of structures involving dependencies over possible
interveners.
Theory applied
The application is illustrated by the well-known asymmetry in aphasic’s comprehension of subject vs.
object clefts.
Normal Representation
[ N;θ 2;ϕ m,s,acc;Σ ][N;θ 1 ;ϕ f,s,nom][N;θ 1;ϕ f,s,nom][N;θ 2;ϕ m,s,acc;Σ]
6 [It is [the boy j [whoj [[the girl]i [i loved j ]]]]]
Spec
(5) . . . X . . . Z . . . Y . . .
[α,β,γ,δ,ε] [α,β,γ,δ] [α,β,γ,δ,ε]
XP
X
’
X
[ α,β,γ,δ,ε] °
Spec
WP
The formation of the relevant chains between the moved NPs and their traces is authorized by RM in
virtue of the presence of the features [ϕs,acc;Σ] distinct from the features [ϕs,nom]. It is specifically the
presence of the Σ feature (that we can call a “relative” feature), that defines the object as
member of a class distinct from the one to which the subject belongs. Being the Operator’s
class the former and Argumental the latter.
W
’
W° ZP
Z
Spec ’
Z°
[ α,β,γ,δ]
YP
Spec
Y
’
. . .
Y
[ α,β,γ,δ,ε] °

Agrammatic Representation
[N] [N] [N; θ 1 ]
7 This is [the boy] who [the girl] [ loved ]
[N;θ 2
If aphasic patients are not able to access some features their representation of these constructions is
impoverished as showed in (7). Because of the impoverishment of features, the subject NP becomes a
potential intervener between the object NP and its trace and RM blocks the formation of the relevant
chain, as a consequence it is impossible to assign the correct theta role to each argument. The
situation is reversed with subject relatives, which are correctly interpreted by agrammatic patients. In
this case, in fact there is no NP intervening between the moved constituent and its trace, which means
that there could be no RM effects.
8 T his is [the boy]who[ loved the girl]
Even an underspecified representation of a subject cleft allows us to form the relevant chain and to
recover the thematic information: no potential bearer of the relevant relation intervenes between the
moved element and its trace. Following the same line it is possible explain other comprehension
patterns reported in the literature. Comprehension is compromised whenever there are potential
interveners between a moved NP and their traces, whereas it is preserved when no potential antecedent
intervenes. Following the same lines we are able to explain the other asymmetries listed in above. We
will also claim that the same system can apply to passive constructions, in which the subject originally
merged in [Spec,VP] intervenes between the moved object and its trace.
References
Kolk, H. (1995) A time-based approach to agrammatic production. Brain and Language, 50: 282-303.
Rizzi, L. (1990) Relativized Minimality. MIT Press
(2001) Locality and left periphery, In Belletti, A., ed. (2002) Structures and Beyond, OUP
Starke, M. (2001) Move dissolves into merge, a theory of locality, Doctoral dissertation, University of
Geneva.
]

Sentence Comprehension and Working Memory Limitation in Conduction aphasia:
Interaction between type of processing and output and input deficits
Aviah Gvion 1,2 , Naama Friedmann 1
1 Tel Aviv University, 2 Reuth medical center
The current study explored the nature of the relation between verbal working memory deficit and sentence
comprehension, focusing on sentence comprehension in conduction aphasia. The two main comparisons that
were made were between sentences that require semantic-syntactic reactivation vs. sentences that require
phonological reactivation, and between input buffer conduction aphasia (repetition) and output buffer conduction
aphasia (reproduction) (Shallice & Warrington, 1977; Shallice, Rumiati, & Zadini, 2000).
We compared two types of processing required during sentence comprehension: syntactic-semantic reactivation
and word-form (phonological) reactivation. In both sentence types we increased phonological working memory
load by increasing the number of words between a word and the position of its reactivation.
Participants
Ten conduction aphasics participated in the study. All participants had pre-morbidly full control of Hebrew. All
patients had 12 years of education and above.
Working memory and phonological input-output evaluations
Recall and recognition span tasks were used to measure the phonological WM capacity. Recall tests included
word span: phonologically similar/dissimilar as well as long/short words and non-word span. Recognition tests
included digit and word spans, listening span and 3 versions of N-back, with two different SOAs.
Output buffer tasks included a full transcription of spontaneous speech, repetition of words and non-words,
visual confrontation naming tasks as well as various phonological manipulation tasks such as spoonerism and
sound deletions. Input tasks included auditory rhyming judgment tasks as well as auditory plausibility
phonological judgment tasks.
Results of working memory assessment
a. Very limited recall spans in all conduction aphasic participants.
b. Two patients were near normal in recognition spans suggesting a selective output buffer deficit.
c. Four patients manifested selective input deficit and four patients suffered from both input and output
disorders.
Sentence comprehension experiments
Experiment 1: does gap-antecedent distance in relative clauses interact with WM limitation?
160 Hebrew relative clauses were included in a binary sentence-picture matching task in which the participant
heard a sentence and chose between the matching picture and a reversed-role foil. Number of words between the
antecedent and the gap (2,5,7 or 9 words), and relative-clause type (subject vs. object) were manipulated.
Examples (translated from Hebrew)
1a) 9-word distance object-relative: This is the NURSE with the-white gown, the-round
eyeglasses and-the-merciful smile that-the-girl dresses__.
1b) 2-word distance object-relative: This is the NURSE that-the-girl dresses__.
Results
a. Non of the conduction aphasics whose working-memory was very limited (spans 2-3 words) showed deficit in
subject and object relative-clause comprehension (89%-96% correct).
b. None of the conduction aphasics showed any effect of length.

Friedman and Gvion (2003) conjectured, based on Love and Swinney (1996) and McElree (2000), that
antecedent-gap distance had no effect on comprehension because the processing at the gap position involves
semantic, rather than phonological, reactivation of the antecedent, and therefore distance that is measured by
phonological units and phonological memory limitation does not affect comprehension. This led us to examine
comprehension in structures that require phonological reactivation, which might result in impaired
comprehension.
Experiment 2 – does disambiguation distance interact with WM limitations?
80 Hebrew sentences with ambiguous words that get disambiguated either after 2-3 words or after 7-9 words
were included in a plausibility judgment and paraphrasing task (auditory presentation). 100 additional
semantically implausible or plausible sentences matched for served as fillers (see examples 2ab that use the
ambiguous word “pen” that means either writing instrument or the place were sheep live).
2a) The PEN that the student received from his grandma when he graduated was packed with wooly sheep.
2b) The PEN is always packed with wooly sheep.
This type of structure includes temporary lexical ambiguity, that at some point of the sentence gets
disambiguated to the less dominant meaning, and thus requires the reactivation of the original phonological word
in order to re-access all meanings and allow for reanalysis. Unlike in experiment 1, in these sentences
phonological, rather than semantic reactivation is required, and this type of reactivation might be hampered by
phonological overload, and comprehension should be compromised when phonological WM is limited.
Results
a. All the conduction aphasics that suffer from a verbal WM limitation (input buffer deficit or “repetition
conduction aphasia”) showed a severe deficit in comprehension of long distance disambiguation sentences,
rejecting more than half of the sentences as “implausible”. They showed no comprehension problem when the
same ambiguity was resolved after 2-3 words.
b. The participants without verbal WM limitation (output buffer deficit or “reproduction conduction aphasia”)
did not show any comprehension problem even in this type of sentences.
These results were further replicated using a rhyme judgment and paraphrasing tasks of structures with short and
long distance between two rhyming words.
Conclusions
Our results indicate that in general, individuals with conduction aphasia do not suffer from sentence
comprehension deficit. It is only a subtype of conduction aphasia and a very special type of sentences that yield a
deficit in comprehension in conduction aphasia.
Sentences that require semantic-syntactic deficit do not pose any problem for the 10 individuals with conduction
aphasia who participated in the study.
Sentences in which phonological reactivation was required posed a problem only to the individuals who suffer
from limited WM, with the input buffer deficit, and only in the presence of phonological overload beyond the
patient’s capacity. The same phonological overload did not have the effect when a syntacto-semantic reanalysis
was required. These results suggest that the type of reactivation required by the sentence, as well as the type of
memory overload are crucial in determining the effect of WM limitation on sentence comprehension.
Friedmann, N., & Gvion, A. (2003). Sentence comprehension and working memory limitation in aphasia: A
dissociation between semantic-syntactic and phonological reactivation. Brain and Language, 86, 23-39.
Love, T., & Swinney, D. (1996). Co-reference processing and levels of analysis in object-relative constructions:
Demonstrations of antecedent reactivation with the cross-modal priming paradigm. Journal of Psycholonguistic
Research, 25, 5-24.
McElree, B. (2000). Sentence comprehension is mediated by content-addressable memory structures. Journal of
Psycholonguistic Research, 29, 111-123.
Shallice, T., Rumiati, R.I., & Zadini, A. (2000). The selective impairment of the phonological output buffer.
Cognitive Neuropsychology, 17, 517-546.
Shallice, T., & Warrington, E.K. (1977). Auditory-verbal short-term memory impairment and conduction
aphasia. Brain and Language, 4, 479-491.

Verb retrieval in anomic aphasic speakers: the role of sentence context.
Roel Jonkers & Christien Prak
Department of Linguistics, University of Groningen (NL)
Introduction
In studies on verb processing in aphasic speakers, different patterns of performance have been found for verb
retrieval in isolation and in spontaneous speech. Williams and Canter (1987), Bastiaanse and Jonkers (1998), and
Pashek and Tompkins (2002) showed that there is no significant correlation between action naming scores and
the production or diversity of verbs in spontaneous speech or narrative discourse.
In action naming, a target verb has to be retrieved in isolation, whereas in spontaneous speech verbs are
produced within sentences and there is no target verb. A third possibility to study the production of verbs is the
investigation of verb retrieval in sentence context. In these cases, target verbs have to be produced within a
sentence. Williams & Canter (1987) showed that especially anomic speakers profit from sentence context in
order to find the intended verb. The relation between verb retrieval in isolation, sentence context and
spontaneous speech has, however, never been studied in a group of anomic aphasic speakers.
Methods and Materials
Subjects: 11 anomic speakers were tested (mean age 60, 7 male, 4 female). All aphasic subjects were righthanded
and suffered from aphasia due to a single stroke. Spontaneous speech data of 10 age-matched non-braindamaged
control subjects were used for comparison.
Materials and procedure: Action naming and verb retrieval in sentence context were tested with the tests
described by Jonkers (1998). Both tests contain 60 items. For action naming a picture was shown, after which the
subject was asked to tell in one word which action was depicted. For verb retrieval in sentence context a picture was
shown and the subject was asked to tell in one sentence what was happening in this picture. For both tests, an item
was scored as correct, if the target verb was produced.
A spontaneous speech sample was obtained by asking the subjects to tell about their speech problems, their work,
and their hobbies. The spontaneous speech was analysed with the method of Vermeulen, Bastiaanse, and Van
Wageningen (1989). 300 Words per sample were used for this analysis. The total number of verbs and the diversity
of the verbs, expressed by the type-token-ratio (number of different verbs divided by the total number of verbs), were
calculated.
Statistics: Verb retrieval in isolation and in sentence context were compared using a paired t-test. Comparisons
between the number of verbs and the type-token-ratio of the verbs produced by the anomic speakers and the
control subjects were made with an independent t-test. Correlations between verb retrieval scores in isolation,
sentence context and the type-token-ratio for verbs in spontaneous speech were calculated with the Pearson test.
Results
The scores for verb retrieval in isolation, sentence context, and in spontaneous speech are depicted in table 1.
Table 1:
A. Mean and standard deviation for the scores of the anomic speakers in verb retrieval in isolation and sentence
context (max.= maximum score)
B. Mean and standard deviation for the total number of verbs and the type-token-ratio for verbs in the
spontaneous speech of the anomic speakers (n=11) and the control subjects (n=10).
A. test results anomic speakers max.
verb retrieval in isolation 28.8 (12.3) 60
verb retrieval in sentence context 33.6 (12.0) 60
B. spontaneous speech anomic speakers control subjects
number of verbs 40.5 (4.5) 34.8 (5.6)
type-token-ratio 0.54 (0.15) 0.74 (0.10)
No significant difference was found between verb retrieval in isolation and in sentence context (t(10)=1.80,
p>0.05). A significant correlation was found between the scores for both tests (rho=0.73, p

No correlation was found for the anomic speakers between verb retrieval in isolation and the diversity of the
verbs in spontaneous speech (rho=0.37, p>0.05). However, verb retrieval in sentence context, significantly
correlated with verb diversity in spontaneous speech (rho=-0.66, p

PERCEPTION OF SIGN-LEXEMES VERSUS NON-VERBAL MOVEMENT OF HANDS
AND ARMS IN DEAF SIGNERS AND HEARING NON-SIGNERS: AN FMRI-STUDY
Juliane Klann (*#§), Frank Kastrau (*), Walter Huber (#)
*: Interdisciplinary Centre of Clinical Research \"CNS\", University Hospital Aachen; #: Neurolinguistics at the
Dept. of Neurology, University Hospital Aachen; §: Dept. of Linguistics, University of Cologne
Introduction:
Sign languages make extensive use of hand- and arm-movements in three-dimensional space. This implies, that
production and perception of sign language requires systems, whose functions are represented in different neural
networks, namely the planning respectively the analysis of hand- and arm-movements, that is known to be
represented in parietal cortex and the processing of language, that is represented in a perisylvian network mainly
concerning frontal and temporal cortical regions. This well established dissociation of neuro-anatomical
correlates leads to the assumption, that deaf signers may recruit both systems when they process sign language,
whereas hearing non-signers should activate parietal regions only, since they are not familiar with sign language
as a natural language system.
Lesion data as well as imaging studies document that deaf signers do not recruit parietal cortex for the
processing of sign lexemes. In order to pursue these issues further, we compared the processing of verbal and
non-verbal hand- and arm- movements in deaf signers and hearing non-signers. To gain an insight in anatomical
localization, functional Magnetic Resonance Imaging (fMRI) was conducted.
Subjects:
The participants were 11 prelingually deaf, age ranging from 22 to 43 years, with primary competence in
German Sign Language (DGS = Deutsche Gebärden Sprache). The control-group consists of 11 hearing,
matched for sex, age and education, with no competence in DGS. All subjects were right-handed.
Methods:
The same task was offered to all participants, hearing as well as deaf, in DGS and written German. Both
conditions were presented separately, each condition in one block. The first block included written nouns,
pseudo- and non-words. The second block contained videotaped single signs, pseudo-signs and non-verbal hand-
/arm-movements and non-signs. The participants were instructed to hit a button when they identified a known
sign or word.
The stimuli were linguistically parallel in both tasks. We chose 60 sign-lexemes, representing objects, 30
pseudo-signs and 30 non-signs. For the written material condition we presented 60 nouns, 30 pseudo-words and
30 non-words. All signs and words consisted of a linguistically simple structure (no compounds) and represented
high frequent nouns. The length of written words, pseudo- and non-words varied between four and nine letters.
The stimuli were presented pseudorandomized in an event related design. Imaging is done with a Philips 1.5
Tesla Gyruscan NT with standard bird-cage headcoil using a multishot T2* weighted gradient echo EPI
sequence. Image analysis and statistical evaluation will be carried out using SPM2 &#61472;(Welcome
Department of Cognitive Neurology).
Preliminary Results:
Data analysis is not yet finished. Inspection of nine deaf individuals demonstrates indeed activation of temporal
regions with no activation of parietal cortex.

ICONIC SIGNS IN GERMAN SIGN LANGUAGE: AN FMRI-STUDY
Juliane Klann (*#§), Frank Kastrau (*), Walter Huber (#)
*: Interdisciplinary Centre of Clinical Research, University Hospital Aachen; #: Neurolinguistics at the Dept. of
Neurology, University Hospital Aachen; §: Dept. of Linguistics, University of Cologne
Introduction:
Sign language differs from sound language respectively written language in its linguistic use of space and in
iconicity (transparency). This leads to the assumption that the cerebral representation of sound/written and sign
language does not completely overlap. This view is at variance with recent lesion data and imaging studies
which show left localization of sign language in much the same way as for sound language. Therefore both
language types might share the same neural substrate, thereby implying a supra-modal representation of
language functions in the brain [1-4].
In order to pursue these issues further, we investigated the functional neuroanatomy of processing transparent
and non-transparent signs conducting fMRI.
Subjects:
The participants were 11 prelingually deaf, age ranging from 22 to 43 years, with primary competence in
German Sign Language (DGS = Deutsche Gebärden Sprache). The control-group consists of 11 hearing,
matched for sex, age and education, with no competence in DGS. All subjects were right-handed.
Methods:
In order to prevent the application of conscious strategies the participants were not informed about the purpose
of the study and their focus was directed to a lexical decision task. The detection of possible differences in the
processing of transparent versus non-transparent signs was guaranteed through a systematic variation of the sign
stimuli in respect to the degree of transparency.
The stimuli were presented pseudo-randomized in an event related design. Imaging is done with a Philips 1.5
Tesla Gyruscan NT with standard bird-cage head-coil using a multi-shot T2* weighted gradient echo EPI
sequence. Image analysis and statistical evaluation will be carried out using SPM2 (Welcome Department of
Cognitive &#61472;Neurology).
Results:
In accordance with former empirical data and in disagreement with the expectations derived from linguistic
differences, the activation patterns in the deaf participants show recruitment of the left hemisphere\'s temporal
network, including Wernicke’s area for processing non-transparent as well as transparent signs. The network
activated involves perisylvian areas within the superior temporal gyrus. Elsewhere this was interpreted as an
recruitment of primary auditory cortex [5]. Contrary to expectations we found no activation of Broca\'s are in
deaf participants. This area is consequently documented to be involved in sign language processing in deaf
signers as well as in sound and written language processing in hearing non-signers [3]. Even in the present study
our hearing participants activated Broca\'s area. Therefore the non-activation of this region in deaf subjects
seems to mirror an modality- and task-specific effect.
In the sign language task hearing non-signers show the expected bilateral activation within occipito-parietal
regions.
References:
[1] Poizner G, Klima ES, Bellugi U: What the Hands Reveal about the Brain. Cambridge: MIT Press (1987) [2]
Leischner A: Die Aphasie der Taubstummen. Ein Beitrag zur Lehre der Asymbolie. Arch. Psychiatr. Nervenkr
1943 (115): 469-548 [3] Rönnberg J, Söderfeldt B, Risberg J: The cognitive neuroscience of signed language.
Acta Psychologica 2000 (105): 237-254 [4] Klann J, Kastrau F, Kemény S, Huber W (2002): The
Neuropsychology of signed and written language: an fMRI-study. Cortex 38: 874-877 [5] Finney EM, Fine I,
Dobkins KR (2001): Visual stimuli activate auditory cortex in the deaf. Nature Neuroscience 4 (12): 1171-1173.

Preliminary analyses of the first patient data of a study on the relation between phonological
input and output disorders in aphasia
Julia Klitsch & Roelien Bastiaanse
Rijksuniversiteit Groningen (RuG),
Graduate School of Behavioral and Cognitive Neuroscience (BCN)
Previous research on the (nature of the) relation between phonological input and output disorders in aphasia has
been controversial. While some researchers assume that impairments in the two modalities are associated and cooccuring,
others observed dissociating performance patterns in the two modalities.
The research question of this project is if and how phonological input and output disorders are related. Are
phonological disorders in one modality predictive of disorders in the other or are input and output dissociable?
Another question is if there are phonetic features that are more detrimental to aphasic speech perception than
others.
To assess these questions, spontaneous speech is collected and the patients are administered a number perceptual
tasks from the Dutch PALPA test battery (Bastiaanse, Bosje, & Visch-Brink, 1995) that tap different processing
stages on the perceptual side of a Logogen-like model (Morton, 1968; Patterson, 1988): Auditory discrimination
of words and non words, auditory lexical decision, auditory-written word matching, and auditory word-picture
matching. Furthermore, expressive data are elicited with an oral naming task, using the target pictures of the
auditory word-picture matching task as stimulus material, and with a repetition task. In the latter, the same words
and non words have to be repeated which are also used in the auditory lexical decision task.
Preliminary analyses of the data of the first four patients showed a lexical superiority effect in both auditory
discrimination tasks, in auditory lexical decision, and in repetition, suggesting involvement of the lexical
processing routes. A frequency effect could be observed in the repetition task only (&#967;2(1)=5.6, p

noise; the more severe the aphasia, the stronger the observed effect, as reflected by the performance patterns of
the two groups.
With respect to the relation between phonological input and output disorders there is no clear evidence for
neither point of view (association vs. dissociation) yet.

Age of acquisition and phonological neighbourhood affect two different recovery patterns from
anomia: two case studies.
Marina Laganaro, Marie Di Pietro and Armin Schnider
Service de Reeducation, Geneva University Hospital
Background: Although word-finding difficulties have been largely studied in aphasia from a theoretical and a
rehabilitation point of view, recovery mechanisms are still not fully understood.
Aims: In this study we investigated which psycholinguistic variables affect the word-finding recovery pattern
during a computer assisted therapy in two anomic subjects (AH and PG).
Study 1. Words and pictures (390) for baseline and post-test naming assessment were selected from Alario et al.
(in press) and Bonin et al. (2002). The following variables were available for all the items: name agreement,
image agreement, concept familiarity, image complexity, words age of acquisition (AoA), lexical frequency,
word length, syllable token frequency, and phonological neighbourhood (PH-NEIGH). For each participant, two
matched sets of 50 words from those they had failed to name at baseline were selected for therapy. Each set was
trained successively during therapy with an unsupervised computer-assisted therapy. The program provided
detailed information on which words were retrieved and which words needed a cue at each learning session.
Despite the fact that AH and PG had a similar quantitative and qualitative anomic pattern at baseline with AoA
affecting their naming performance, they showed quite different recovery patterns. The computer-assisted
therapy was effective for both patients, but PG needed twice more sessions than AH to name 80%of the items
without help. The psycholinguistic variables predicting the recovery curve differed between patients. AoA
predicted recovery in AH, i.e. early-acquired words needed less item presentations. PH-NEIGH predicted
recovery in PG, i.e. words with a higher phonological neighbourhood density and higher image agreement were
recovered earlier.
Study 2. Following the observation in study 1of a sublexical effect in recovery from anomia in PG, we analysed
new words learning curve in PG and AH as well as in healthy controls. Twenty new words controlled for
sublexical variables were associated to abstract drawings. Phonological neighbourhood predicted new words
learning speed in controls and in anomic subjetcs, i.e. new words with many phonologically similar real words
were learned easier than new words without phonologically similar words.
Conclusion. Our results show two different patterns of recovery: the first one is affected by a lexical variable
(AoA), the second one is affected by a sublexical variable, which also affects new words learning. We suggest
that the effect of lexical variables (age of acquisition) reflects recovery from an impaired access to intact lexical
representations, while an effect of a sub-lexical variable (phonological neighbourhood) on recovery reflects
partial impairment of the lexical representation.

Introduction
Effects of metrical and segmental cues on word retrieval in aphasia
Antje Lorenz 1,2 , Wolfram Ziegler 1 , Ria De Bleser 2
1 Clinical Neuropsychology Research Group (EKN), City Hospital Bogenhausen, Munich, Germany
2 Institute for Linguistics, Potsdam University, Potsdam, Germany
According to current theories of speech production metrical and segmental information of the target word form
play comparable roles in the generation of single words (Levelt et al., 1999). On these grounds the (pure)
metrical structure of a target word can be expected to be an appropriate cue in anomia treatment. With regard to
segmental cues considerable evidence has been provided in favor of a high (immediate) effectiveness of partial
phonological information in several anomic patients (e.g. Best et al., 2002). Furthermore, the use of orthographic
cues in the facilitation of picture naming produces strong effects in some patients (e.g. Best et al., 2002; Howard
& Harding, 1998) but not in others (e.g. Bruce & Howard, 1988). Different assumptions have been made
regarding the relationship between sublexical processing abilities for written materials and the effectiveness of
orthographic cues (e.g. Best et al., 2002; Howard & Harding, 1998). Several authors have pointed out a direct
relation between preserved sublexical reading and effects of orthographic cues (e.g. Best et al., 2002) whereas
others have found significant effects of orthographic cues in patients without sublexical transcoding abilities for
written materials (e.g. Howard & Harding, 1998).
Method
In this study, the effectiveness of metrical and segmental prompting was examined in a picture naming task with
ten aphasic patients. Severe word retrieval deficits formed a major symptom of each patient’s aphasia.
Extensively low-pass filtered target words were used as metrical cues (Lorenz et al., 2000). In the segmental
condition the immediate effectiveness of initial phonemes and initial graphemes were compared. Various
background assessments, including lexical and sublexical processing of written and spoken materials as well as
picture naming, were conducted with each patient. On these grounds the relationship between underlying
functional deficit(s) in the patients and the effectiveness of specific types of word form specific information on
word retrieval were analysed.
The stimulus set consisted of 40 pictured objects. All target words were monomorphemic nouns with one, two,
or three syllables.
When a picture could not be named accurately within five seconds one type of cue was presented and the
immediate effects on naming accuracy were investigated. With spontaneous correct responses the next picture
was presented right away.
In each run, half of the pictures were presented with metrical cues, the other half with segmental cues. Both
phonemic and orthographic information was used as a segmental cue. Each picture appeared with each type of
cue in different runs. The combination of cue types and pictures was well balanced. Each picture was presented
in randomized order once or twice per session. For each patient the results were cumulated over eight
consecutive sessions.

Results
TABLE 1
Proportion of correct responses in different background assessments
N-Items Subtests
LEXICAL DECISION
RA BR MH GE GG EB JK BF HW ZU
80 auditory 83.8 96.3 86.3 93.8 91.3 92.5 97.5 97.5 93.8 98.8
80 visual
REPETITION
86.3 88.8 80 73.8 87.5 96.3 97.5 90 88.8 96.3
40 pseudowords 72.5 15 90 25 77.5 62.5 77.5 97.5 70 92.5
40 words
READING
90 67.5 100 55 87.5 90 95 92.5 97.5 100
20 letters nt nt 0 nt 10 80 95 90 5 15
40 pseudowords 0 0 0 0 5 77.5 70 47.5 0 0
40 words (regular)
PICTURE-NAMING
2.5 0 12.5 10 70 92.5 100 75 72.5 77.5
240 spoken 1.25 10 17.1 15 52.1 47.9 12.5 55.4 34.6 29.2
20 written 0 0 30 0 5 75 55 40 5 0
nt = not tested
Spoken picture naming was examined in a set of 240 pictured objects obtained from various sources. All other
subtests were taken from LeMo (De Bleser et al., 1997).
Figure 1: Proportion of (additional) correct responses after different types of word form related information
Metrical cues were at least as effective as initial phoneme cues in eight patients. Only two patients profited
significantly more from initial phonemes than from the metrical structure of the target word (patients BF and
ZU: p < .05; χ 2 -Test, two-tailed; cf. Figure 1).
The effectiveness of initial phoneme cues correlated significantly with the ability to repeat pseudowords
(Spearman-Rho, two-tailed: .65, p< .05) but not with the ability to repeat words (Spearman-Rho, two-tailed:
.561, p = .09). Patients with less impairment in the repetition of pseudowords benefited most from initial
phoneme cues. No significant correlations were obtained between repetition accuracies and the effectiveness of
initial grapheme cues or metrical cues.
Initial grapheme cues affected word retrieval significantly in six patients. Only three of them were (partially)
able to transcode written materials (pseudowords; letters) on a sublexical processing level (patients JK; EB; BF).
The other three patients who had profited from initial grapheme cues significantly were not able to name (or
sound out) initial letters or to read pseudowords (GG; HW; ZU). For the whole group the effectiveness of
orthographic cueing correlated with lexical reading (Spearman-Rho, two-tailed: .71, p < .05), but not with

(sublexical) reading of pseudowords (Spearman-Rho, two-tailed: .349, p = .32). No significant correlations were
obtained between reading accuracies and the effectiveness of initial phoneme cues or metrical cues. Furthermore,
overall naming accuracy was not correlated with the effects of any type of cue.
Main Conclusions
i. Metrical cues were at least as effective as initial phoneme cues in most of the patients. These results are in
line with current psycholinguistic theory which predicts comparable roles of metrical and segmental
information in the generation of single words (Levelt et al., 1999).
ii. Initial phoneme cues seem to be processed sublexically, since their effectiveness correlated with the
repetition of pseudowords, but not with the repetition of words.
iii. The effectiveness of initial grapheme cues correlated with lexical reading ability, but not with sublexical
reading ability. This result contradicts the data from Best and colleagues (2002) and points to a directlexical
mechanism of effectiveness of orthographic cues at least in some patients of this group. The
implications for models of word production will be discussed.
References
Best, W., Herbert, R., Hickin, J., Osborne, F., Howard, D. (2002). Phonological and orthographic facilitation of
word d-retrieval in aphasia: immediate and delayed effects. Aphasiology, 16, 151-168.
Bruce, C., Howard, D. (1988). Why don't Broca's aphasics cue themselves? An investigation of phonemic cueing
and tip of the tongue information. Neuropsychologia, 26, 253-264.
De Bleser, R., Cholewa, J., Stadie, N., & Tabatabaie, S. (1997) LeMo, an expert system for single case
asessment of word processing impairments in aphasic patients. Neuropsychological Rehabilitation 7: 339-365.
Howard, D., Harding, D. (1998). Self-cueing of word retrieval by a women with aphasia: why a letter board
works. Aphasiology, 12, 399-420.
Levelt, W.J.M., Roelofs, A., Meyer, A.S. (1999). A theory of lexical access in speech production. Behavioral
and Brain Sciences, 22, 1-38.
Lorenz, A., Ziegler, W., Artinger, F., Zierdt, A. (2000). Metrical information as a cue in picture naming. Journal
of Neurolinguistics, 13, 307-310.

DISCRIMINATION OF SIMILAR ENVIRONMENTAL SOUNDS FOLLOWING STROKE:
LEFT HEMISPHERE PATIENTS FAIL TO USE SEMANTIC FACILITATION
Alessandra LORENZI, Luigi A. Vignolo
University of Brescia
Auditory nonverbal agnosia (auditory agnosia sensu strictiori) consists in the selective inability to recognize
nonverbal sounds. Carl Kleist (1928) maintained that the inability to perceive isolated sounds or noises
(perceptive Geräuschtaubheit) should be distinguished from the inability to understand the meaning of noises
(Geräuschsinntaubheit). Quantitative behavioural studies carried out since the 1960s (see Vignolo, 1982;
Schnider et al., 1994) confirmed this dichotomy and investigated its anatomical bases. They led to define the
existence of two types of auditory agnosia: an acoustic-discriminative one, preferentially associated with lesions
of the right hemisphere and a semantic-associative one, specifically associated with lesions of the left
hemisphere and aphasia. The former consists of imperception for the acoustic structure of sounds. The latter
consists of defective recognition of the meaning of sounds and may be considered to be a cognitive rather than a
perceptual disorder, involving one or several sensory modalities and consisting in the inability to put together
different aspects of the same concept.
The problem with all these studies was that acoustic discrimination and semantic recognition were investigated
by tests involving different sensory modalities. Discrimination was tested by a purely auditory same/different
test, while identification was tested by an auditory-visual multiple choice test. This difference made it difficult to
compare results.
The present study was undertaken in order to verify the right-left difference employing same/different
discrimination tests strictly confined to the auditory modality. We wished to investigate the ability to
discriminate pairs of acoustically similar environmental sounds, with or without semantic help. We assumed that
the discrimination between two acoustically similar sounds should be facilitated when they originate from two
different semantic sources ( e.g. miaowing cat– crying baby ) as opposed to when they come from two sources
belonging to the same semantic group ( e.g. cat 1 – cat 2 ). In the first case the semantic difference adds to the
acoustic difference in order to permit the discrimination, while in the second case no such facilitation is present.
If this is true, patients with right hemisphere damage (who can identify the meaning of sounds) should be able to
take advantage of the semantic help, like normal controls, while patients with left hemisphere damage (who have
problems in identifying the meaning) should not be able to make good use of this help.
A new test, the Environmental Sounds Discrimination Test, was administered to 20 controls and 12 unilateral
hemispheric stroke patients (5 RBD [right brain damaged], 7 LBD [left brain damaged]) harbouring a small or
medium sized lesion (diameter less than or equal to 3 cm) on the CT scan. Exclusion criteria were: a history of
hearing difficulties before stroke, multiple lesions, trauma, neurodegenerative diseases. The test consists of 40
items, each made up of a pair of meaningful environmental sounds. 14 items are pairs of acoustically similar
sounds, originating from different semantic sources (e.g. applause-tape-writer): they constitute Task 1. 14 items
are pairs of acoustically similar sounds, originating from sources belonging to the same semantic category (e.g.
thunder 1 – thunder 2; wind 1 – wind 2): they constitute Task 2. The remaining 12 items are pairs of identical
sounds (e.g. hen-hen). The patient was told that he was going to hear two familiar noises one after the other. He
was asked to listen to them carefully and to say whether they where the same or different.
We expected that a. controls would perform both tasks well, but Task 1 (with semantic help) better than Task 2 (
without semantic help ); b. RBD patients would perform both tasks much worse than the other groups but, like
controls, Task 1 better than Task 2 (because both controls and RBD use the semantic help); c. LBD patients
would perform both tasks better than RBD, without significant differences between Task 1 and Task 2 (being
unable to use the semantic help).
Results confirmed our expectations. The total number of errors made by RBD patients were double with respect
to the LBD and more than triple with respect to controls. Both controls and RBD patients could easily
discriminate among similar sounds originating from different semantic sources (Task 1). Virtually all errors
made by these two groups concerned the discrimination of sounds produced by the same semantic source (Task
2). This indicates that in both controls and right brain damaged, the semantic help makes acoustic
discrimination significantly easier. By contrast, LBD patients were the only group in which a remarkable
number of errors in Task 1 was found. This indicates, in our view, that semantic help is less efficient in LBD than
in RBD and controls.
In conclusion, this study, carried out with a test specifically confined to the acoustic sphere, provided further
evidence of the semantic impairment typical of the left damaged subjects.

REFERENCES
Kleist, C. (1928) Gehirnpathologische und lokalisatorische Ergebnisse über Hörstörungen, Geräuschtaubheiten
und Amusien. Monatsschr Psychiat Neurol, 68: 853-860
.
Schnider A., Benson F.D., Alexander D.N., Schnider-Klaus A. (1994) Non-verbal environmental sound
recognition after unilateral hemispheric stroke. Brain, 117:218-227.
Vignolo L.A. (1982) Auditory agnosia. Phil Trans R Soc Lond Biol, 298:49-57

A Probabilistic Model of Functional Category Production in Agrammatic Narratives
Lisa H. Milman 1 & 2, Michael Walsh Dickey 1,2 & Cynthia K. Thompson, 1,2,3
Aphasia and Neurolinguistics Research Laboratory 1 Communication Sciences and Disorders 2 and Neurology 3
Northwestern University
Introduction
One account of agrammatism developed over the last decade posits that agrammatic errors reflect the
hierarchical structure of syntactic trees. (Hagiwara, 1995; Friedmann, 2001). According to this hypothesis,
accessibility to various functional projections is contingent on the relative severity of the agrammatism. Thus,
the prediction is that in cases of mild agrammatism, only the highest level projection, CP, is affected. In more
severe cases, however, more general impairments should be evident, affecting not only CP but lower projections
as well. Results from several studies support these predictions for Japanese patients as well as for Hebrew &
Arabic-speaking patients. Nonetheless, this account has been challenged since it does not account for all of the
available data within and across languages (Reznik, 1995). It is worth noting that this criticism is not specific to
this theory. In fact, the heterogeneity observed across patients has led to the much broader assertion that patient
variability defies any general description (Miceli, Silveri, Romani, & Caramazza, 1989).
The question addressed here is how to reconcile the fact that the hierarchical model may explain much of the
data, yet, as might be expected given the variability observed in patient performance, not explain all of it. One
possible solution tested here is to recast the question in probabilistic terms. Thus, given the hypothesis that the
hierarchy of impairment observed in agrammatism reflects the structural properties of syntactic trees, we might
predict the following two performance patterns (where 0 indicates impaired performance; and 1 indicates spared
performance):
CP – TP – AgrP – NegP – AspP – VP – V’ – XP
(1) 0 1 1 1 1 1 1 1
(2) 0 0 0 1 1 1 1 1
However, given the variability in performance observed across patients, we might also expect to see patterns like
(3) and (4):
CP – TP – AgrP – NegP – AspP – VP – V’ – XP
(3) 0 0 1 0 1 1 1 1
(4) 1 1 1 1 0 0 0 0
Although pattern (3) departs somewhat from expectations, overall it seems to approximately fit the model.
Pattern (4) on the other hand, is exactly opposite of what we would expect based on the hypothesis. Evaluating
the hierarchical structure hypothesis in a probabilistic framework makes it possible to distinguish between cases
like (3) and (4). Thus the model accommodates some variability in the data, but importantly, statistically defines
what this variability should like, if our model is in fact true. In order to evaluate the viability of reformulating
the hierarchical structure hypothesis in probabilistic terms, two related research questions are addressed:
1. Is the hierarchical ordering of impairment reported for Japanese, Hebrew, and Arabic-speaking
agrammatic patients also evident in the performance pattern of English-speaking agrammatic patients?
2. If this pattern is observed, can any observed variability be accounted for within the context of a
probabilistic model?
Methods
Cinderella narratives were analyzed for 18 agrammatic patients. Each transcript was checked for the occurrence
or omission of eight syntactic elements: C’, T’, Agr’, Neg’, Asp’, SpecVP, V’, and XP (where XP=any internal
argument of the verb). Narratives in which at least two occurrences of a particular element were noted received
a score of “1”; narratives in which there were no occurrences of a particular element were scored “0”.

These data then were analyzed using Item Response Theory (IRT) (Linacre, 2000). IRT is a probabilistic model
used in the field of psychometrics to estimate the difficulty of test items. IRT was used here to provide a
statistically valid estimate of difficulty for each of the syntactic projections, and to test whether the difficulty
scores could adequately account for the variability observed in the performance patterns of each patient. The IRT
analysis computes the likelihood that a subject with ability level x will be successful on an item of difficulty y.
Essentially the model compares observed scores with predicted scores and calculates goodness of fit.
Results
Results of the IRT analysis are illustrated graphically in Figure 1. Mean square error (MSE) for all subjects and
items was less than 2, which is considered acceptable (Linacre, 2000).
Difficulty Score (Logits)
6
4
2
0
-2
-4
C'
T'
AGR'
Fig. 1. IRT difficulty scores for each functional projection
NEG'
ASP'
VP V' XP
Discussion
In general, the hierarchical ordering of impairment previously reported for Japanese, Hebrew, and Arabicspeaking
agrammatic patients was also found for the English-speaking patients in this study. However, when the
performance pattern of individual patients was examined, there were six patients whose performance departed
somewhat from the predictions of our model. For example, consider the response patterns of PR and HW:
CP – TP – AgrP – NegP – AspP – VP – V’ – XP
PR 0 0 1 0 1 1 1 1
HW 0 0 1 1 0 1 1 1
In spite of this variability, when the data were analyzed in terms of probabilities, the resulting difficulty
estimates for each syntactic constituent were able to account for the performance pattern of all patients on all
syntactic elements. Results of this study provide strong preliminary support for the use of probabilistic models,
such as IRT, to identify consistent findings that are still based on individual performance patterns.
References
Friedmann, N. (2001). Agrammatism and the psychological reality of the syntactic tree.
Journal of Psycholinguistic Research, 30 (1), 71-90.
Hagiwara, H. (1995). The breakdown of functional categories and the economy of
derivation. Brain & Language, 50, 92-116.
Linacre, J.M. & Wright, B.D. (2000). A user’s guide to Winsteps, Ministep, Bigsteps:
Rasch Model Computer Programs. Chicago: Mesa Press.
Miceli, G., Silveri, M., Romani, C., & Caramazza, A. 1989. Variation in the pattern of

omissions and substitutions of grammatical morphemes in the spontaneous speech of
so- called agrammatic patients. Brain and Language, 36, 447-492.
Reznik, M. (1995). Functional categories in agrammatism. Brain & Language, 50,
117-133.

Interpretable vs. uninterpretable features: evidence from six Greek-speaking agrammatic
patients
Vicky Nanousi, Jackie Masterson, Judit Druks, Martin Atkinson
University of Essex
Introduction
The objective of the present study was to evaluate the claims of the Tree Pruning Hypothesis (TPH, Friedman &
Grodzinsky, 1997) in the light of evidence derived from Greek-speaking agrammatic patients. The categories
under investigation were tense, agreement and aspect features. According to the TPH the locus of the breakdown
in agrammatism is the T node on the syntactic tree, and functional categories that are above it are impaired while
those below it are spared. TPH was originally based on Hebrew, in which Tense is above Agreement (Ouhalla,
1990). This organisation of the syntactic tree could account for the performance of Hebrew-speaking agrammatic
patients whose production of tense features was impaired while agreement features were preserved. In Greek,
the organisation of the syntactic tree is different: Agreement is thought to be above Tense, and Aspect is located
low, closest to the stem (Horrocks, 1982; Belletti, 1990; Alexiadou, 1997). Therefore the prediction from the
TPH for Greek agrammatic production would be that aspect and tense features are preserved, and agreement
impaired. A second prediction, based on Friedman & Grodzinsky (1997), is that the dissociation between tense
and agreement is confined to the production modality.
The experiments tested the availability of functional categories using single word and sentential repetition,
transformation and judgement tasks. Six Greek agrammatic patients with LIFG lesions following CVA
participated in the study.
Repetition tasks
There were 90 single verb forms inflected for tense (simple present, simple past, past continuous, simple future
and present perfect), aspect (imperfective, perfective and perfect) and agreement (three persons and two
numbers), and 90 sentences with verbs inflected for the same features. The sentences were in S-V-O, V-O-S &
V-S-O order and were presented for immediate and delayed repetition. There were no errors in the immediate,
and very few errors in the delayed repetition.
Production tasks eliciting tense, agreement and aspect in single words
A trial consisted of an example verb illustrating the transformation in feature that was required, followed by five
test verbs. The same five tenses that were explored in the repetition tasks were used here. For tense
transformations agreement and aspect were held constant. During agreement transformations tense and aspect
were held constant, and when person changes were required number was held constant and vice versa. Finally,
for aspect transformations, tense and agreement were held constant.
Production of tense, agreement and aspect within sentences
Pairs of sentences were presented assessing tense and agreement. The first sentence contained a verb form
inflected for tense and agreement while in the second sentence the verb was missing. In the second sentence the
change required was indicated either by a temporal adverbial (for tense elicitation) or by an anaphoric pronoun,
overt pronominal person (for agreement elicitation). The same tenses and agreement inflections examined in the
previous tasks were used here. Aspect was tested in two tasks. For the first task there were sentence frames in
which a verb was missing. Sentences were presented with a choice of verb forms each expressing the
imperfective, perfective and perfect aspect. The correct choice was determined on the basis of an adverbial of
frequency. For the second task, the perfective and imperfective aspect were examined using triads of sentences.
The first sentence contained a verb appropriate for the temporal adverbial while in the second and third sentence
the verb form was omitted and the required change was indicated by a different temporal adverbial and adverbial
of frequency.
Results
All three verbal features were equally impaired in the single word tasks, while in sentences tense and aspect were
more impaired than agreement. Performance was better with the sentences than the single words, suggesting that
a sentence frame facilitates the production of the verbal inflectional features.
Grammaticality judgment tasks
Three tasks were used. These were a free, a contrastive and a forced choice sentence completion task. In all three
tasks there were violations in tense, agreement and aspect.

Results
Agreement was better preserved and tense & aspect were equally impaired. The error rate was lower than that
found in the production tasks.
Discussion
The Greek patients’ data provides evidence against the TPH, insofar that Tense and Aspect, that are lower in the
Greek syntactic tree, were found to be better preserved than Agreement. Instead, the results are interpreted
within recent formulations of Minimalism that distinguish between interpretable (tense, aspect) and
uninterpretable (agreement) features and the morphophonological evaluation operations associated with them.
Agree and its sub-process of copying deals with the morphophonological valuation of the uninterpretable φfeatures
of verbs, and Spell-out and a similar sub-process of copying for the valuation of the interpretable
features. The data show that either the patients are better at producing semantically empty material than material
with semantic content, or that the operation of Agree functions better than Spell-out. At present our data cannot
distinguish between these two possibilities. Better performance in sentences than in single words is explained by
arguing that in processing sentences a syntactic frame for the verb is readily available, whereas in single verb
production, the syntactic frame has to be created by ‘projecting’ a sentence frame for the target word. Since the
patients have problems with morphosyntactic representations and/or the relevant operations, they cannot perform
the virtual projection, and this is why the production of inflected single verbs is more impaired than that of verbs
incorporated into sentences. Finally, our results also conflict with the claim of Friedman & Grodzinsky that the
tense deficit is apparent only in production, since the patients exhibited impaired performance, although to a
lesser degree than in production, in grammaticality judgement tasks too.
In conclusion, two intuitively unexpected results were observed in the present study, the first is that features
devoid of semantic content are better preserved than semantically meaningful features, and the second is that
performance with sentences is better than with single words.
References
Alexiadou, A.,1997. On the properties of some Greek word order patterns. In Alexiadou, A.,
Horrocks, G. and Stavrou, M. (eds). Studies in Greek syntax, Kluwer, Dordrecht.
Belletti, A. 1990. Generalised verb movement. Torino: Rosenberg and Sellier.
Friedmann, N & Grodzinsky, Y. 1997. Tense and agreement in agrammatic production:
pruning the syntactic tree. Brain and Language, 56, 397-425.
Horrocks, J. 1982. The order of constituents in MG. In Gazdar, G., Pullum, J. and Klein, E.
(eds.). Order, concord and constituency. Dordrecht: Foris, 95-111.
Lightfoot, D. 1981. Explaining syntactic change. In Hornstein, N. and Lightfoot, D. (eds.).
Explanation in Linguistics. London: Longman.
Ouhalla, J. 1990. Functional categories, agrammatism and language acquisition. Ms. Queen
Mary and Westfield College, London University.

Production and Comprehension of wh-questions
in German Broca's aphasia
Eva Neuhaus & Martina Penke
Heinrich-Heine-University Duesseldorf
Introduction
A long standing issue in the investigation of agrammatism is whether the agrammatic deficit is due to an
impairment of syntactic competence that affects comprehension and production in parallel, or whether
comprehension and production are affected by independent disorders. We will tackle this issue by investigating
wh-question production and comprehension in German agrammatic Broca’s aphasics. Wh-questions are a good
test case since (1) they are dealt with in recent theories on agrammatic production and comprehension and (2)
they can be easily tested in both modalities.
For agrammatic speech production, Friedmann and Grodzinsky have advocated the “Tree pruning Hypothesis”
(TPH; Friedmann & Grodzinsky 1997, Friedmann 2002) according to which the agrammatic deficit causes a
“pruning” of the syntactic tree minimally resulting in the loss of the CP layer which is critical for the formation
of wh-questions. This proposal predicts that agrammatic aphasics are not able to produce complete wh-questions
with a wh-phrase in SpecCP and a finite verb in the COMP position.
Concerning language comprehension Grodzinsky (1990, 2000) has proposed that the agrammatic deficit leads to
the deletion of movement traces in syntactically derived sentences. According to his proposal aphasic subjects
are impaired in the comprehension of wh-object questions because the object theta role cannot be assigned to the
moved object due to deletion of the trace between the base-generated and the derived object position. Via a
default strategy, the thematic role ‘agent’ is then assigned to the moved object.
According to these two approaches the deficits in language production and comprehension are caused by two
distinct disorders and are thus not parallel. A deficit affecting grammatical competence, such as the loss of
functional categories proposed by e.g. Ouhalla (1993), would in contrast predict a parallel impairment in the
production and comprehension of wh-questions.
Method
We present results from three experiments testing the production and comprehension of wh-questions in nine
agrammatic German Broca's aphasics.
In an elicitation task, subjects had to transform a given main clause in which one information was replaced by a
non-specific indefinite term into a wh-question (1). For each subject, 42 argument questions (subject and object
questions) and 12 adjunct questions were tested. In a repetition task, subjects were asked to repeat an auditorily
presented wh-question as accurately as possible (2). 20 argument questions (subject and direct object) and 10
adjunct questions were tested in this experiment.
(1) Petra hat jemanden gesehen. (“P. has seen somebody”)
Klaus fragt:______________? (“K. asks:___________?”)
target: Wen hast du gesehen? (“whom have you seen?”)
(2) Wo versteckt er den Schatz? (“where does he hide the treasure?”)
In the comprehension task, subjects saw pictures depicting semantically reversible actions performed by two of
three persons. In the following example a man brushes a woman who in turn brushes a man. Subjects were then
asked to answer a wh-argument-question such as Wer bürstet die Frau? (“who brushes the woman?”) by
pointing to the corresponding person in the picture (3).
(3)
60 wh-questions (30 wh-subject and 30 wh-object questions) were presented to each subject. Questions were
formulated such that each of the three persons (A, B, C) was the correct target in one third of the items.

Results and discussion
The following table presents an overview of the correct reactions in these three experiments.
In the elicitation task we counted all wh-questions as correct that were introduced by a wh-pronoun and
contained a finite verb placed in V2 position. In the repetition task a wh-question counted as "correct" if it was
accurately repeated.
Correctness scores in %
subjects
production comprehension
elicitation repetition picture pointing
M.B. 51/54(94.4%) 30/30 (100%) 50/60 (83.3%)
G.B. 54/54 (100%) 21/30 (70%) 54/60 (90%)
I.K. 54/54(100%) 30/30 (100%) 26/60 (43.3%)
R.M. 47/54 (87%) 25/30 (83.3%) 15/60 (25%)
W.R. 50/54 (92.6%) 21/30 (70%) 26/60 (43.3%)
W.W. 48/54 (88.9%) 18/30 (60%) 23/60 (38.3%)
P.B. 40/54(74.1%) 5/30 (16.7%) 19/60 (31.7%)
M.J. - 29/30 (96.7%) 32/60 (53.3%)
E.S. 13/54 (24.1%) - 46/60 (76.7%)
∑ 357/432 (82.6%) 179/240 (74.6) 291/540 (53.9%)
As can be seen in the table, the performance of two of our aphasic subjects (M.B. and G.B.) was unimpaired both
in comprehension and production. The majority of our subjects (I.K. to P.B.) generally show a good performance
in the two production tasks, whereas their performance in the comprehension task was only at chance level. Only
subject E.S. displays the opposite pattern i.e. above chance performance in comprehension and a severe disorder
in wh-question production.
These finding suggest that
1. wh-question production is generally not impaired in German agrammatism. This provides evidence that
the CP layer can still be projected in German agrammatism – contrary to the assumptions of the “Tree
Pruning Hypothesis”.
2. comprehension problems such as predicted by Grodzinsky’s trace deletion account cannot be observed
in every agrammatic aphasic (see M.B. and G.B.); however
3. deficits in wh-question comprehension are common. A closer analysis of the error patterns reveals
different error patterns in our subjects. Whereas the results for subject E.S. are in line with
Grodzinsky’s most recent proposal according to which the agrammatic deficit only affects which N
object questions but not who-questions, the data of I.K. provides evidence for the classical tracedeletion
account according to which all wh-object questions are impaired whereas wh-subject questions
are not.
4. production and comprehension are not affected in parallel. This suggests that agrammatism is not due to
a deficit in syntactic competence leading to the loss of all or some functional projections. However,
both in comprehension and in production, we find a clear asymmetry between wh-subject and wh-object
questions: wh-object questions are more difficult to produce and understand than wh-subject questions.
This indicates that the canonical linear order of arguments in sentence structure is a crucial factor
determining the performance of agrammatic aphasics.
References
Friedmann, N. & Y. Grodzinsky (1997). Tense and agreement in agrammatic production: Pruning the syntactic
tree. Brain and Language, 56, 397-425.
Friedmann, N. (2002). Question production in agrammatism: the tree pruning hypothesis. Brain and Language,
80(2), 160-187.
Grodzinsky, Y. (1990). Theoretical perspectives on language deficits. London: MIT Press.
Grodzinsky, Y. (2000). The neurology of syntax. Behavioral and Brain Sciences, 23(1), 1-71.
Ouhalla, J. (1993). Functional categories, agrammatism and language acquisition. Linguistische Berichte, 143, 3-
36.

Syntactic Impairments in Persian
Reza Nilipour
University of Welfare and Rehabilitation Sciences
And
School of Cognitive Sciences (IPM), Tehran, Iran.
Abstract
This paper presents clinical syntactic impairments in adult native speakers of Persian subsequent to brain
lesion. Persian is introduced as an Indo- European language, belonging to the Iranian branch of the Indo-Iranian
group. Typologically, it is an SOV language with a rather rich morphology. It is an atypical SOV language with
a rather loose syntax.
The source of the presented syntactic violations of Persian is from two groups of aphasics and two different
etiologies. The patients include both mono-lingual as well as bilingual native speakers of Persian with different
site and size of lesions and the etiologies include both CVA and trauma.
The agrammatic aphasics and their syntactic violations in Persian are based on systematic assessments with
the same standard aphasia test (The BAT, 1987). The pathological data are extracted from syntactic sub-tests as
well as different propositional connected speech samples.
The presented grammatical violations are compared with the subject control data as well as cross-linguistic
data. The syntactic analysis of the clinical data is based on the descriptive framework of the CLAS ( Menn &
Obler, 1990).
The data on Persian syntactic impairments are characterized into two major groups; general syntactic
impairmments versus language-specific violations.
The general syntactic violations include evidence of breakdown of VP and /or NP structure, substitution or
deletion of grammatical morphems and simplification of syntax.
In the domain of language–specific syntactic violations, three major groups of syntactic violations are
discussed:
Tendency towards substitution of infinitive or its truncated form for the proper inflected form of the verb
substitution of a filler verb (to be) for all sorts of verbs, including transitive verbs
tendency to omit NP or VP internal linking elements.

Reading strategies in a case of pure alexia: Evidence from word length
Papathanasiou Ilias, Kavvada Aimilia,
Department of Speech and Language Therapy, Technological Educational Institute Patras, Greece Konstantinou
Demetrios Department of Neurosurgery, Medical School, University of Patras, Greece
Pure alexia is an acquired reading disorder that occurs secondary to a lesion in the left occipitotemporal junction.
Pure alexia is referred as an inability to identify all the letters of a word as a single unit and to access the
orthographic lexicon (Hillis, 2002). Patients with pure alexia do not have a problem of a purely perceptual
nature. That is, single written letters are perceived and read properly. Therefore, they often spontaneously
compensate for their deficit by naming each letter of the word aloud or silently, which allows them to recognize
the word. Their strategy forms the distinctive feature of this deficit which is a word- length effect; the naming
latencies of patients increase dramatically with increasing numbers of letters in the word. However, the
commonest deficit associated with pure alexia is a right homonymous field defect; an impairment that may, by
itself, interfere with single-word reading because of inability to see the letters towards the end of a word.
Case History: A 42-year-old right-handed woman, XO who is a high school English teacher suffered two
separate episodes of leukoencephalitis during the last five years. Two years prior to this study, an MRI had
shown a remaining proencephalic cyst in the left occipital lobe. Language and reading were described as normal
prior to those two episodes. She was an avid reader, both professionally and for pleasure. She was referred to us
by her neurosurgeon.
Assessment: XO presented a complete right hemianopia. For the initial language evaluation, the Boston
Diagnostic Aphasia Examination was used which revealed a mild pure alexia and no aphasia.
For further assessment of XO&#8217;s orthographic processing abilities a non-standardized test was used. This
involved reading aloud, spelling and recognizing orally spelled words tasks that varied along two parameters;
length and frequency, as well as pseudowords. Specifically, the test contained 80 nouns, which were controlled
for frequency and length , and 40 pseudowords based on these nouns. The selected words and the pseudowords
were chosen to have the same visual outline. In order to rule out the possibility that her right hemianopia was
responsible for her reading errors, all stimuli were four and five letters long, so the subject was able to read most
of the single-word presentations without having to make a rightward saccade. This task was also administered to
a normal age-, sex- and educational level- matched control subject. XO was recorded while reading. Single-
word reading speeds were measured through the Nero Burning Rom program. Statistical analysis was performed
using Micrososft Excell programme.
Results: In comparison with the control data, XO demonstrated no word frequency effect during reading.
However, there was a significant difference in latency of reading between four letter and five letter words, as
well as between real words and pseudo words with the same letter length. During the assessment, XO was
tending to make educated guesses after she had identified the first few letters of a word and therefore her
response time was increasing.
Discussion: This case demonstrated that in pure alexia, reading is not only influenced by letter length but also
there are lexicality effects as she is reading real words better than pseudowords. Friedman et al (2000), has also
reported these lexicality effects as a result of word frequency. However our case has clearly demonstrated no
word frequency effect. Our findings are important in understanding the nature of the reading strategies used by
cases of pure alexia.
References
Hillis (2002) : The handbook of adult language disorders, Psychology Press, New York
Friedman R & Lott SN (2000): Rapid word identification in pure alexia is lexical bt not semantic. Brain &
language, 72(3), 219-227

The neural processing of naming nouns and verbs with special reference to
age-of-acquisition. An event-related fMRI study.
Jenny Postler 1 , Ria De Bleser 1 , Christina Kauschke 2 , Simone Reinders 3,4 ,
Cornelius Weiller 3 & Christian Büchel 3
1 University of Potsdam, Department of Cognitive Neurolinguistics
2 Free University of Berlin, Department of General and German Linguistics
3 University of Hamburg, Department of Neurology
4 Groningen University Hospital, Department of Psychiatry
Abstract
Introduction. The neural representation and processing of nouns and verbs is currently a controversially
discussed issue in neurolinguistics. The assumption that there are different neural representations for nouns and
verbs arose from lesion studies with word class specific impaired patients. These studies suggested that noun
processing is associated with temporal lobe lesions whereas verb processing is linked to left frontal lobe lesions
(Druks, 2002 for a review). Recent neuroimaging studies could not consistently confirm these findings (Perani et
al. 1999; Sörös et al. 2003; Tyler et al. 2001, 2003, 2004; Warburton et al. 1996). Although selective
impairments in noun and verb processing often occure in naming tasks, only one MEG study (Sörös et al. 2003)
used a naming paradigm to verify the putative anterior-posterior dissociation between noun and verb processing.
The authors found a similar neuronal network underlying noun and verb retrieval.
Picture naming is always influenced by a variety of linguistic and visual parameters. Recent behavioral studies
emphasize the influence of age-of-acquisition on word processing in particular for object naming (Morrison et al.
1992). This observation could lately be confirmed in an fMRI study by Fiebach et al. (2003) for auditory and
visual lexical decisions on nouns.
In the present study, we used event-related fMRI to investigate the neural processing of German nouns and verbs
with a picture naming paradigm. The aim of the study was first to test the anterior-posterior dissociation between
noun and verb processing. Second we examined the influence of age-of-acquisition (AoA) on naming.
Methods. We used concrete nouns and verbs presented as black-white line drawings (De Bleser & Kauschke,
2002). Nouns were counterbalanced between animate and inanimate items, verbs between transitive and
intransitive forms. The word classes were matched for frequency, productive age of acquisition (according to a
parents survey, De Bleser & Kauschke, 2003), visual complexity and name agreement. In addition a control
condition was developed consisting of scrambled images of the original noun and verb pictures in order to
control for visual processing and motor activation.
18 subjects (9 female) participated in the study. They were instructed to name noun and verb pictures internally
and react with "ok" in the case of scrambled pictures. Simultaneously they had to press a button with the left
hand. To ensure adequate performance all subjects were intensively instructed and trained prior to the fMRI
scanning. Name agreement was examined after the fMRI scanning with a written naming task.
Data acquisition. Functional images were acquired with a Siemens Symphony 1.5 Tesla scanner using a T2*
sensitive gradient-echo echo-planar sequence (32 axial slices; 1 mm gap; voxel size 3 x 3 x 3 mm; TR 2.56 s;
TE 40 ms; flip angle 90°; FOV 19.2 cm; matrix 64 x 64 voxels). The data analyses were performed with SPM2.
Functional data were preprocessed including correction for slice timing delays and motion artifacts,
normalization to a standard stereotactic space and smoothing with a Gaussian kernel of 12 mm FWHM for group
analysis.
Data analysis. The subject specific data sets were analysed with a factorial-parametric statistical model
containing the conditions: nouns, verbs, scrambled nouns, scrambled verbs and age-of-acquisition for nouns and
verbs, respectively. Age of acquisition was modeled as a parametric regressor. On the second level noun and
verb differences were analysed using within-subjects Anova. Age-of-acquisition effects on naming were tested
with a one-sample T-Test. Four Regions of interest were defined to identify AoA effects: the left inferior frontal
cortex including Broca's area (LIPC), the left superior temporal gyrus including Wernicke's area, the left
angular/supramarginal gyrus and left basal temporal gyrus (BTLA). Data were considered to be significant at a
statistical level of p < .05 (corrected for multiple comparisons).
Behavioral results. The mean reaction time for naming nouns (602.85 ms) and verbs (665.24 ms) differed
significantly (Anova F(1,51) = 72.09; p < .000, post hoc Scheffè test p < .000) whereas no significant differences

were found between the control conditions (post hoc Scheffè p = .52). Name agreement was comparable between
grammatical categories (nouns: 92.88 % correct; verbs 91.41% correct; Anova F(1,17) = 2.26; p < .12).
fMRI results. The comparison of each word category with the corresponding control condition resulted in a
similar activation pattern. In both contrasts we found activation of the LIPC, the fusiform gyrus and the dorsal
frontal gyrus bilaterally. Further activation was observed in the temporo-occipital region up to superior parietal
areas in both hemispheres. Naming verbs relative to nouns identified activation in the left inferior frontal gyrus
and bilaterally in the temporo-occipital region, in the fusiform gyrus and in the inferior parietal sulcus. The
reverse contrast showed no significant voxels.
Increasing signal for the naming of late learned words was detected in the LIPC (BA 47) and the left BTLA (BA
37). There were no suprathreshold voxels for the naming of early learned words, however subthreshold
activation indicates participation of the precuneus (BA 7).
Discussion. We found a very similar activation pattern for the retrieval of nouns and verbs, although verbs
produced a significantly stronger hemodynamic response. Thus, our data do not support the anterior - posterior
dissociation between noun and verb processing but rather suggest similar neuronal processing mechanisms for
the retrieval of these word classes. Furthermore, we demonstrated that the hemodynamic signal for naming
nouns and verbs is modulated by age-of-acquisition. Later learned words causes more activation in areas which
are associated with semantic processing (LIPC BA 47) and with lexical retrieval (BTLA BA 37), thus indicating
stronger effort in lexico-semantic retrieval for these words.
References
De Bleser & Kauschke (2003). Acquisition and loss of nouns and verbs: parallel or divergent patterns?
Journal of Neurolinguistics, 16, 213-229
Druks (2002). Verbs and nouns - a review of the literature. Journal of Neurolinguistics, 15, 289-315.
Fiebach et al. (2003). Distinct brain representations for early and late learned words. Neuroimage, 19, 1627-1637.
Morrison et al. (1992). Age of acquisition, not word frequency, affects object naming, not object recognition.
Memory & Cognition, 20, 705-714.
Perani et al. (1999). The neural correlates of verb and noun processing. A PET study. Brain, 122, 2337-2344.
Sörös et al. (2003). Naming action and objects: cortical dynamics in healthy adults and in an anomic patient with a dissociation in
action/object naming. Neuroimage, 19, 1787-1801.
Tyler et al. (2001). The neural representation of nouns and verbs: PET studies. Brain, 124, 1619-1634.
Tyler et al. (2003). Objects and their actions: evidence for a neurally distributed semantic system. Neuroimage, 18, 542-557.
Tyler et al. (2004). Neural processing of nouns and verbs: the role of inflectional morphology. Neuropsychologia, 42, 512-523.
Warburton et al. (1996). Noun and verb retrieval by normal subjects. Studies with PET. Brain, 119, 159-179.

CONTEXTUAL PRIMING IN SEMANTIC ANOMIA: A CASE STUDY
K. Renvall 1 , M. Laine 2 , N. Martin 3
1 Department of Psychology, University of Turku, Finland
2 Department of Psychology, Åbo Akademi University, Finland
3 Center for Cognitive Neuroscience, Temple University School of Medicine, USA
Introduction
Contextual priming was developed as a procedure to manipulate aphasic naming by contextual relationships
(semantic, phonological, unrelated) and by intensive repetition priming (Laine & Martin, 1996; Martin & Laine,
2000). The first case studies cited above indicated that contextual priming has at least short-term facilitative
effects on target naming and suggested an interaction between the patient's functional deficit and the naming
context. Since then several additional studies have explored the effects of contextual priming. In the first fullscale
treatment study (Renvall et al., 2003) we found generalisation of the treatment effects in the semantic
context and also documented a longer-term treatment effect than in the previous studies. In the facilitation study
of Martin, Fink, Laine & Ayala (in press) in which 11 aphasics were tested an immediate interference effect but
short-term facilitation was found. The current suggestion is that the contextual priming procedure is most
effective when semantic processing of words is relatively spared (Martin, Fink & Laine, in press). The present
case study explores further this suggestion by conducting a treatment study with one Finnish-speaking aphasic
suffering from semantic anomia.
Method
PH is a 73-year-old right-handed male who suffered from a vast left arteria cerebri media infarction (affecting
nucleus lentiformis and the periventricular white matter) and subsequent right-sided hemiparesis over 2 years
prior to this study. Spontaneous speech is fluent but rambling and lacking in content due to constant problems in
word retrieval. In the formal background testing PH exhibits a moderate naming difficulty with semantic errors,
circumlocutions and nonwords while repetition and oral reading is intact. In tasks tapping lexical-semantic
processing of words PH has difficulties in distinguishing between close semantic relatives although the access to
the approximate semantic domain is often successful. These results suggest that the primary source of PH's
word-finding difficulty is at the semantic level.
The treatment design included 3 baseline naming measurements, 12 treatment sessions (carried out within 5
weeks), 2 post-treatment naming measurements right after the treatment period, and 1 follow-up measurement
1.5 months after the treatment had been finished. The stimuli used in the naming measurements included 120
hard-to-name items (60 to-be-trained and 60 untrained control pictures), and 30 easy-to-name items. The trained
items were selected so that they formed 15 picture sets (4 pictures in each) where the items were either unrelated,
semantically related or phonologically related to each other (5 picture sets for each relatedness condition). The
picture sets were cycled so that each picture set was treated in 9 treatment sessions. The training protocol
included repeated cycles of spontaneous naming attempts and repetition of the target names after the examiner
(see Martin & Laine, 2000).
Results and Discussion
McNemar tests of change were used to analyse the naming success before and after treatment. They revealed that
there was a significant change in the naming success on the trained items between the third baseline
measurement (3/60 correct) and the first post-treatment measurement (18/60 correct; McNemar test with
continuity correction, χ 2 = 11.53, df=1, p< .0001) and also between the third baseline measurement and the
second post-treatment measurement (14/60 correct; McNemar test with continuity correction, χ 2 =7.69, df=1, p<
.001). However, no change was anymore observed in the trained items when the third baseline measurement was
compared with the follow-up measurement administered 1,5 months after the treatment (8/60 correct). In the
control items, no change was observed in any of the measurements. Thus, the improvement in naming the trained
items was only short-lived. Interestingly, when the conditions were analysed separately, the semantic condition
was the only condition in which we found significant improvement in naming the trained items. This occurred
between the third baseline measurement (0/20 correct) and the first post-treatment measurement (McNemar's
test with the continuity correction, χ 2 =5.14, df=1, p< .05). This indicates that the treatment had some temporary
benefit for PH.

With regard to word-level errors during treatment, PH produced lots of contextual word errors (names of other
items in the same picture set) and non-contextual word errors (words that are not included in the picture set but
hold a semantic or phonological relation to the target). Contextual word errors appeared in all conditions but half
of the errors occurred in the semantic condition (semantic 79, phonological 33, unrelated 28) indicating
immediate interference at the level of his deficit. This finding is consistent with that of Martin, Fink, Laine &
Ayala (in press). The great number of semantically related non-contextual word errors (89/112; in the semantic
condition 28 errors, in the phonological condition 34 errors, in the unrelated condition 50 errors) is also
consistent with prior studies, e.g., Martin et al. (1989), and not surprising as PH made many semantic errors in
spontaneous speech and in different naming tasks administered before treatment. The possible changes in the
distribution of the naming error types during the 6 naming measurements are yet to be analysed.
In summary, the present results provide support to the claim that contextual priming might not show longstanding
beneficial effects in aphasics who have a lexical-semantic impairment (Martin, Fink & Laine, in press).
Research is underway to combine contextual priming with methods that would provide item-specific semantic
support for semantic anomia patients.
References
Laine, M. & Martin, N. (1996) Lexical retrieval deficit in picture naming: implications for word production
models. Brain and Language, 53, 283-314.
Martin, N., Fink, R. & Laine, M. (in press) Treatment of word retrieval deficits with contextual priming.
Aphasiology.
Martin, N., Fink, R., Laine, M. & Ayala, J. (in press) Immediate and short-term effects of contextual
priming on word retrieval in aphasia. Aphasiology.
Martin, N. & Laine, M. (2000) Effects of contextual priming on impaired word retrieval. Aphasiology, 14,
53-70.
Martin, N., Weisberg, R.W. & Saffran, E. M. (1989) Variables influencing the occurrence of naming errors:
implications for models of lexical retrieval. Journal of Memory and Language, 28, 462-485.
Renvall, K., Laine, M., Laakso, M. & Martin, N. (2003) Anomia treatment with contextual priming: A case
study. Aphasiology, 17, 3, 305-328.

Movement in agrammatism: is the operation impaired ‘per se`’?
Eleonora Rossi a Roelien Bastiaanse Y.R.M a
a Graduate School for Behavioural and Cognitive Neurosciences (BCN), University of Groningen, Department of
Linguistics, P.O. Box 716, 9700 AS, Groningen, The Netherlands
E-mail address corresponding author: e.rossi@let.rug.nl
Telephone number: 0031503635859
Fax number: 0031503636855
Italian Negative Clauses:
Italian negative clauses are characterized by a negative particle, i.e. non (not) which is expressed pre verbally.
The structure of Italian negative clauses has been very much into debate. According to Belletti (1990) the
negative particle heads the NP which is found between AGRP and TP. Moreover, ‘non’ is considered to be a
clitic therefore undergoing movement to Agr in order to give the constituent order: Subj. non Vfin . According to
the author, following Pollock`s theory, the verb moves to T and Agr to get agreement and tense. Therefore the
‘negation chain’ and the ‘verb chain’ share the same landing site i.e. Head of Agr. In order to have sentences
derived like:
Maria non mangia la mela
Maria does not eat the apple
Maria not eats the apple
the verb is moving from its based position in order to get Tense and Agreement to land in HeadAgr. Note that
this movement is independent from the movement of the negative particle non which is moving with a head to
head movement from HeadNegP to HeadAgrP. The V and the negative particle non have the property to share
the same landing site.
Italian Small Clauses:
According to Stowell (1981) a small clause (henceforth SC) is the minimal predicative structure containing an
argument, a predicate but no tense:
Vedo [il gatto mordere inf il topo]
I see the cat biting the mouse.
In contrast in Italian Matrix clauses (MCs) the verb is always finite and express agreement:
Il gatto mordefin il topo
The cat bites the mouse
Even though the analysis proposed by Cardinaletti and Guasti (1993) about negation in full clauses and small
clauses enlights structural differences, no difference is claimed in respect to the rising of the negative element.
Therefore we could match small and matrix clauses in this respect.
Hypotheses:
First, we expect small clauses to be better produced than matrix clauses, where the verb has to be moved in order
to pick up its features.
Moreover affirmative clauses are expected be better produced than negative clauses where the negative element
non has to move with a head to head movement to AGR, its landing site. In specific: small clauses should be
better produced than matrix clauses and negative matrix clauses should be the worst produced structures.
Material and methods:
So far, one Italian right-handed agrammatic speaker has been tested: R.B. 51 year old suffered a stroke in the left
hemisphere 12 months ago. R.B. is a native speaker of Italian and was classified on the AAT test base as a Broca
aphasic.
Experiment design:
Using a sentence completion task affirmative and negative sentences (108) crossing within the two conditions
i.e. MCs, and SCs, were presented.
The test was prompted by pictures, where the corresponding verb and object had to be filled in. Negative
sentences were prompted with the corresponding affirmative ones and vice versa:
MC: Qui il ragazzo mangia la mela e qui il ragazzo.....
Here the boy eats the apple and here the boy…
SRC: Qui vedo il ragazzo che non mangia la mela e qui vedo il ragazzo che....

Here I see the boy who does not eat the apple and here I see the boy who…
SCs: Qui vedo il ragazzo mangiare la mela e qui vedo il ragazzo....
Here I see the boy eating the apple and here I see the boy…
Results:
Running a Chi 2 test for independence revealed a highly significant association (p=0.000) between structure i.e.
(MC, SC) and finiteness. In 72% of the cases the verb in MCs is infinite. Only 28% presents Tense. Instead, in
the SC condition 86% of the verbs are correctly produced, i.e. in the non finite form and only 14% show Tense
and Agreement.
Regarding negativity, there was no difference between affirmative and negative clauses (both in Matrix clauses
as well as in embedded clauses). The negative particle non was always produced, in the right position in all
structures.
Conclusions:
We found a significant association between structure and finiteness: this result supports our first hypothesis that
in SCs, where the verb has to be in the finite form, is better produced that MCs where the verb has to express
Tense and Agreement, difficult operation for agrammatic patients.
The hypothesis that negative sentences should be worse produced than affirmative ones is not supported.
Nevertheless, because verb movement to check features resulted impaired and the clitic movement of non did not
we can confirm that the two operations are independent. We can therefore support Belletti`s analysis, which is
stating that the two movements are independent.
Furthermore, since verb movement was impaired and the clitic one was not, we can affirm that the operation of
movement per se` is not impaired in agrammatism. The explanation for the difference in the performance
between the two operations has probably to find other explanations, i.e. the difference in the type of movement
or possibly the length of movement. Further investigations to understand how the possible explanations can
operate are needed.
References:
Belletti A. 1990. Generalized verb movement; aspects of verb syntax, Torino:Rosenberg & Sellier,.
Cardinaletti A. and M. T. Guasti. 1993. Negation in epistemic small clauses. Probus 5:39-61
Rizzi L.. 1982 Issues in Italian Syntax, Foris ed.

Does processing of verbs automatically activate action related areas?
Raffaella Ida Rumiati, 1,2 Karsten Specht, 2 Claudio Luzzatti, 3 Chen Qi, 2 and Gereon R. Fink, 2,4
1Cognitive Neuroscience Sector, SISSA, via Beirut 2-4, I-34014 Trieste, Italy, 2 Institute of Medicine, Research
Center Jülich, D-52425 Jülich, Germany; 3 Department of Psychology, University Milano-Bicocca, I-20126
Milan, Italy; 4 Department Neurology, University Hospital, RWTH, D-52074 Aachen, Germany.
Introduction
Neuropsychological and neuroimaging studies report conflicting evidence as to the neural correlates of verb
retrieval and lexical processing. Patients have been described with disproportionate verb impairment not due to
verbs being more prone to language deficits since patients with the opposite pattern were documented. The
majority of patients with selective verb deficits show left frontal lesions (Daniele et al., 1994 for a review).
However patients with disproportionate verb deficit were reported having lesions falling outside the left frontal
cortex (Daniele et al., 1994; Tranel et al. 2001; Luzzatti et al., 2002). In particular, in Luzzatti et al. (2002;
Luzzatti et al in prep.), lesions found in the 15 patients with predominant impairment of verbs clustered in three
subsets: i) left posterior temporal lobe and left inferior parietal cortex; ii) frontal, temporal and parietal
perisylvian territory; and iii) insula, external capsula and basal ganglia. To date, in any of these patients an
isolated frontal lesion was identified. Different loci have been found for the retrieval of verbs: left temporal plus
premotor-prefrontal regions (Perani et al., 1999), left posterior temporal and parietal areas (Damasio et al., 2000;
Martin et al., 1995). Others failed to find any difference in activations as compared to nouns or found that verbs
are associated with a higher degree of activation but with very similar cortical distribution (Warburton et al.,
1996; Teyler et al., 2001). Finally TMS applied to left prefrontal cortex led to an increase in RTs (Shapiro et al.,
2001) in one study but to a facilitation effect in another one (Cappa et al., 2002). Very recently, activations
within the premotor, and primary motor cortices have been reported during silent reading of action verbs (Hauk
et al., 2004). One possible interpretation of this latter finding is that verbs implicitly activate motor related areas.
If the actionality of verbs automatically activates the underlying motor representation, activation of actionrelated
areas should be observed even when the actionality of the verb is irrelevant to the online task such as
deciding whether the sentence contains a verb in the present or in the past tense.
Methods
Twelve healthy subjects (ten males, two females) were scanned while they decided whether a written sentence
contained a verb either in the present or in the past form. They responded with either the right or the left hand,
using the index and the middle finger to press the key. Order of the task (i.e. “does the sentence contain a verb in
the present form?” or “does the sentence contain a verb in the past form”), and the hand used for responding
were counterbalanced across subjects.
Subjects were presented with sixty sentences containing a verb in the present form and sixty in the past form.
Half of them contained an action verb (i.e. “I play/played to the piano”), and half a non-motor verb (i.e. “I
listen/listened to the piano”). Verbs in the two classes were closely matched for frequency. A baseline was used
in which subjects were asked to decide whether a sentence contained a physical or an abstract entity. Nouns
belonging to the two classes were closely matched for frequency. The decision involving the baseline (i.e. “does
the sentence contain a physical entity?” or “does the sentence contain an abstract entity?”), and the hand used for
responding were counterbalanced across subjects. All experimental material was in German.
The paradigm was designed as a hybrid blocked design, by varying parametrically the probability of containing
sentences with an action verb. The analysis was eventually performed as an event-related design, separating the
different sentence types. This single-session fMRI experiment contained 440 EPI images, with 25 slices and a
TR of 2.6 seconds. The slices were positioned according to a high-resolution T1 weighted image, and were
oriented parallel to the AC-PC line.
Eye-movements were recorded throughout the experiment to ascertain whether subjects scanned the sentences
normally.
Results
Imaging data were analyzed using SPM2. In a random effect analysis, when non-action verbs were subtracted
from action verbs, no significant regions were found activated (for K > 10; P = 0.001, uncorrected).
When action verbs were subtracted from non-action verbs, activations were observed in right temporo-parietal
junction and left parietal lobe, as well as bilaterally in the occipital lobe.

In an ANOVA (1 factor, two levels), the areas activated both when processing action and non action verbs were
located in the left premotor cortex, bilaterally in the parietal lobe, and in the associative visual areas bilaterally,
but more pronounced on the left. There was also a bilateral activation in SMA (for K > 10; P = 0.001, corrected).
When the baseline was subtracted from the action verbs, areas of significant differences were detected in left
premotor cortex and left associative visual areas. The same areas, but more significant, were detected, when the
baseline was subtracted from the non-action verbs.
Discussion
Preliminary results do not seem to support the view that actionality of verbs activates motor-related areas in a
task in which this characteristic is only implicitly processed. Since some of these areas seem to be activated
when subjects processing both types of verbs, it is possible that they constitute the neural bases of grammatical
judgment. These findings are consistent with recent neuropsychological reports of patients with damage to the
left parietal cortex whose impairment in naming was equally severe for action and non actions verbs (Luzzatti et
al. in prep.).

Making meaning in an interaction with an aphasic man
Emmanuelle Savarit and Paul Dickerson
Roehampton University of Surrey
This paper will look at the use of qualitative research and more particularly, the usefulness of conversation
analysis in the study of aphasia. In the last decade, researchers from different backgrounds such as linguistics,
speech and language therapists, psychologists etc., have been looking at aphasia from a different perspective,
which now views language as a communicative process.
Conversation is a very important form of everyday social activity. The CA focus of attention is to look at the
detail of local management and the sequential organisation of the conversation and aims to comprehend how one
produces his/her own behaviours in conversation and how he /she understands the interactional behaviour of
others, (Sacks, 1992). Conversation analysis is an empirical approach to the study of naturalistic interaction,
(Hemmersley, 2003). It is a departure point for the exploration of language impaired people in interaction,
(Goodwin, 2003), focusing on participants’ non-verbal pragmatic skills rather than looking at the syntactic or
lexical content of the speech. Applied CA is able to focus attention on the meaning rather that the aesthetic of
speech.
Aphasic conversation is well organised and structured, (Goodwin and Heritage, 1990). CA has been able to
demonstrate that the organisation of aphasic behaviour is much more complex than one would expect. Goodwin,
(1995) shows the complexity of aphasic interaction and it is with the use of conversation analysis that Goodwin
illustrates the conversational success of an aphasic man despite his lexical limitation.
One way to look at the interaction of aphasics is to investigate the structure of talk with aphasic participants.
Such an investigation would look at sequential placement ,e.g. gaze; embodied activities and the talk of others;
sequential organisation, e.g. repair, turn taking; how meaning is built in a local context; or how aphasic people
use physical resources to facilitate, is the co-participant’s access to meaning. A naturalistic setting is one of the
focal points of the observation of aphasic behaviour in everyday conversation. Locally constructed context too,
plays an important role in the understanding of the data, (Tetnowski and Franklin, 2003). These investigations
demonstrate that a person suffering from aphasia may achieve communicative success in interaction with nonimpaired
speakers.
Conversation analysis of microanalytic studies provides a clear picture of the aphasic’s interactional capabilities,
(Perkins, Crisp and Walshaw, 1999). Conversation Analysis may be a useful tool in the assessment of aphasic
people, not taking the place of cognitive neuropsychological assessment, but as an additional resource for the
therapist, to help her make decisions, (Lesser and Perkins, 1999). Lapointe, 2000; Damico et al, (1999)
demonstrate that conversation analysis brings new values (insights) that can be used in therapeutic interventions.
Recent investigations of therapeutic interventions highlight the importance of the conversational partner,
(Cunningham and Ward, 2003, Wilkinson, 1999).
This paper will draw a brief outline of the origins of CA in general terms, as well as its adaptation to the study of
aphasia. Then, developing on previous CA works, it will explore the interaction of an aphasic man which
involves turn taking, repair, gaze, embodied activities, physical environment and talk.
References
Damico, J.S. Oelschlaeger, M. and Simmons-Mackie, N., (1999) Qualitative methods in aphasia research:
conversation analysis. Aphasiology, 13(9-11), 667-679.
Cunningham, R. and Ward, C.D. (2003) Evaluation of a training programme to facilitate conversation between
people with aphasia and their partners. Aphasiology, 17(8), 687-707.
Goodwin, C. (1995) Co-Constructing meaning in conversations with an aphasic man. Research on language and
social interaction, 28(3), 233-260.
Goodwin, C. (2003) Introduction in Conversation and brain damage. Oxford University Press, NY, USA.
Hammersley, M. (2003) Conversation analysis and discourse analysis: methods or paradigms? Discourse and
Society, 14 (6) 751-781.

Lapointe, L.L.; (2003) Functional and Pragmatic Direction in Aphasia Therapy, in The Science of Aphasia:
From Therapy to Theory edited by Papathanasiou, I. and De Blesser, R. Pergamon , Oxford UK.
Perkins, L. Crisp, J and Walshaw, D, (1999) Exploring conversation analysis as an assessment tool for aphasia:
The issue of reliability. Aphasiology, 13 (4/5), 259-281.
Sacks, H.(1992) Lectures on Conversation: Volume I &II. G. Jefferson, ed. Oxford.
Tetnowski, J.A. and Franklin, C.T. (2003) Qualitative Research: Implications for Description and Assessment.
Special forum on Qualitative Research. American Journal of Speech and Language Pathology. Vol.12, 155-164.
Wilkinson, R. (1999) Introduction. Aphasiology, 13(4/5), 251-258.
Wilkinson, R. (1999) Sequentiality as a problem and resource for intersubjectivity in aphasic conversation:
analysis and implication for therapy. Aphasiology, 13(4/5), 327-343).

Treatment of sentence production in agrammatism: Generalization effects and the role of
syntactic hierarchy, complexity and treatment modality
A. Schröder, A. Lorenz, J. Postler, N. Stadie, F. Burchert, & R. De Bleser
University of Potsdam, Department of Linguistics
Introduction
Agrammatic patients have specific problems with the production of movement-derived non-canonical sentence
structures. According to the “Tree Pruning Hypothesis” (TPH) (Friedmann & Grodzinsky, 1997; Friedmann,
2001) this deficit is due to an inability to access hierarchically higher nodes of the syntactic tree. As a result,
movement-derived structures that rely on higher nodes in the syntactic tree cannot be produced correctly by
agrammatic patients.
With respect to the treatment of agrammatic patients and following the TPH, a specific remediation of structures
that rely on higher nodes should result in an improvement of all syntactic abilities and across different sentence
types relying on lower node structures even though no specific treatment has focussed on lower node structures.
This has been shown by Friedmann and colleagues in a Hebrew-speaking agrammatic patient (Friedmann et al.,
2000).
Alternatively, the complexity account of treatment efficacy (CATE, Thompson, Shapiro, Kiran & Sobecks,
2003) predicts that remediation of syntactically complex sentences results in an improvement of syntactically
less complex sentences but not vice versa. Thompson et al. (2003) reported generalization effects from
production of sentences in which movement occurs within an embedded clause (syntactically complex sentences,
i.e. object-relative-clauses) to sentences in which movement occurs in the matrix clause (less complex sentences,
i.e. object extracted who-questions). As syntactically complex sentences rely on higher node structures than
syntactically less complex sentences, the two accounts predict generalization effects in the same direction.
However, Thompson and colleagues predict generalization effects only for structures that rely on the same
movement operations (i.e. wh-movement or NP-movement) as the treated structures. In fact, after treatment of
structures relying on wh-movement (i.e. object-relative-clauses), Thompson et al. (1997) observed an
improvement only in the production of other wh-movement structures (i.e. wh-questions) but not of NPmovement
structures (i.e. subject-raising, passive-sentences) and vice versa.
With regard to sentence comprehension versus sentence production, Jacobs and Thompson (2000) found a crossmodal
generalisation effect in one direction, i.e. from training sentence comprehension to sentence production
but not from production to comprehension. This finding is in line with the assumption of at least partially
identical processing components underlying comprehension and production as proposed by Garrett (1995).
However, it remains unclear why the generalisation effect is only observed in one direction and how it is
compatible with the assumption of distinct processing components for comprehension and production (e.g.
Caramazza & Hillis, 1989).
The aim of the present study was to compare two competing accounts on the treatment of agrammatic sentence
production outlined above (TPH and CATE) and to investigate possible generalization effects in German. Both
TPH and CATE predict a direction for generalization effects in the treatment of agrammatic aphasia, i.e.
remediation of syntactically more complex structures (that rely on higher nodes in the syntactic tree) should
result in better performance of less complex structures (that rely on lower nodes in the syntactic tree) but not vice
versa (Friedmann et al., 2000; Thompson et al., 2003). However, TPH predicts generalization effects to all
syntactic abilities and structures that rely on lower nodes than the treated structure independently of the kind of
movement these structures rely on (Friedmann et al., 2000), whereas Thompson and colleagues only predict
generalization within sentences that are based on the same movement operations (Thompson et al., 1997). With
respect to the assumption of partially identical processing components (Garrett, 1995), a generalization effect
from sentence production to sentence comprehension can be hypothesized, whereas no generalization should be
observable considering distinct processing components (Caramazza & Hillis, 1989).
Design and procedure
We will present the results of an ongoing multiple single case treatment study with seven chronic German
Broca’s aphasics with agrammatism (4 men, 3 women; mean age: 53 years, range 33 – 67 years; mean time post
onset: 9,1 years, range 4 – 16 years). A cross-over design controlled for items and tasks was applied (Barlow &
Herson, 1984; Franklin, 1997).
Two distinct treatment phases were applied in order to train the production of two types of non-canonical whmovement
structures varying in syntactic complexity and hierarchy. Object-relative-sentences were used for
remediation of the production of syntactically more complex sentences, relying on higher nodes in the syntactic
tree; who-questions were used in order to treat syntactically less complex sentences relying on lower nodes.

During training, the patients had to identify the thematic roles and carry out the movement operations required
for deriving the specific sentence types (after Thompson et al., 2003). Within each treatment phase, up to 12
treatment sessions were conducted over a period of six weeks. Baseline testing included the set of trained and
untrained items in production and comprehension tasks as well as the production of NP-movement structures
(passive sentences).
Results and Discussion
Preliminary results of the ongoing study showed that both treatment methods seem to be appropriate to reveal
item-specific learning effects. The results will be discussed within the framework of the two theoretical accounts
on agrammatic production treatment depicted above. In particular, we will refer to the following effects:
(1) a structure specific learning effect when improvement is observed within sentence structure. These
improvement patterns can either be established only for treated item sets, or improvement can be observed for
untreated item sets.
(2) a generalized learning effect when improvement is observed across sentence structure for untrained item sets
within the same modality. This effect can either be established for untrained sentences structures that rely on the
same movement operations (wh-movement), or to untrained sentences structures that rely on different movement
operations (NP-movement)
(3) a cross-modal generalization effect in case of improvement in the untrained modality (comprehension).
Theoretical and therapeutic implications will be discussed.
References
Barlow, D.H., & Hersen, M.H. (1984) The single-case in basic and applied research: A historical perspective. In
Single Case Experimental Designs. London: Allyn & Bacon.
Caramazza, A., & Hillis, A. (1989) The disruption of sentence production: Some dissociations. Brain and
Language, 36, 625-650.
Franklin, S. (1997) Designing single case treatment studies for aphasic patients. Neuropsychological
Rehabilitation, 7, 401-418.
Friedmann, N., & Grodzinsky, Y. (1997) Tense and agreement in agrammatic production: pruning the syntactic
tree. Brain and Language, 56, 397-425.
Friedmann, N., Wenkert-Olenik, D., & Mali, G. (2000) From theory to practice: treatment of agrammatic
production in Hebrew based on the Tree Pruning Hypothesis. Journal of Neurolinguistics, 13, 250-253.
Friedmann, N. (2001) Agrammatism and the psychological reality of the syntactic tree. Journal of
Psycholinguistic Research, 30, 71-90.
Garrett, M. (1995) The structure of language processing: Neuropsychological evidence. In: M.S. Gazzaniga (ed.)
The Cognitive Neurosciences. Cambridge, MA: MIT Press.
Jacobs, B.J., & Thompson, C.K. (2000) Cross-modal generalization effects of training noncanonical sentence
comprehension and production in agrammatic aphasia. Journal of Speech, Language, and Hearing Research, 43,
5-20.
Thompson, C.K., Shapiro, L.P., Ballard, K.J., Jacobs, B.J., Schneider, S.S., & Tait, M.E. (1997) Training and
generalized production of Wh- and NP-movement structures in agrammatic aphasia. Journal of Speech and
Hearing Research, 40, 228-244.
Thompson, C.K., Shapiro, L.P., Kiran, S., & Sobecks, J. (2003) The role of syntactic complexity in treatment of
sentence deficits in agrammatic aphasia: the complexity account of treatment efficacy (CATE). Journal of
Speech and Hearing Research, 46, 591-607.

Endcoding and decoding deficits in a coduction aphasic
Kyriakos Sidiropoulos 1 , Ria De Bleser 2 , Hermann Ackermann 3 , Bruno Preilowski 4
1 Graduate School of neural and behavioural Sciences, University Tübingen
2 Humanwissenschaftliche Fakultät, Patholinguistics, Potsdam
3 Department of Neurology, University Tübingen
4 Department of Psychology, University Tübingen
The research in the fields of auditory agnosia [1] showed that there is a preverbal processing level, in which
acoustic signals are analyzed of their temporal characteristics. A functional loss of this level is accompanied by a
reduction of the ability to perceive verbal and nonverbal sounds. Bilateral lesions in posterior superior temporal
gyrus (pSTG), which extend also laterally in planum temporale, are responsible for it.
The conduction aphasic subject examined in this study showed in the acute stage of his illness a similar
symptomatology to auditory agnosia. His auditory difficulties for nonverbal sounds regressed completely in the
course of the time, however occurred the processing of syllables, words and nonwords incorrectly. These
difficulties are not based on a phonemic disturbance like in word deafness because word-deaf patients are
generally unable to analyze linguistic stimuli and access their phonology.
In order to specify the nature of the disturbances in our conduction aphasic patient, four experiments were
designed: 1. a consonant-vowel distinction task, 2. a gap detection task with wide-band noises, 3. a lexical
decision task and 4. a repetition task.
The results from this study support the hypothesis that in the left pSTG a buffer does exist, which phonetically
analyse [2] and briefly stored [1] all verbal sequences. This buffer is involved in phonological aspects of speech
perception and production. Therefore lesions in the left posterior temporal gyrus (pSTG) not only affect
executive, but also auditory perception functions. Structural factors, like the familiarity and regularity of
phonotactic patterns, the frequency of use of a word and the lexical density affect the strength of storage. Also
the arriving semantic information contributes to the stability of the representation. In syllables and nonwords
arise the interferences faster, because such sequences don’t have a semantic representation. This is why
conduction aphasics with a left pSTG lesion have particularly greater difficulties to keep in mind such verbal
sequences.
In order to confirm the results from this case study, a voxel-based lesion-symptom mapping study is in progress.
1. Buchsbaum, B. R., Hickok, G., Humphries, C. (2001): Role of left posterior superior temporal gyrus in
phonological processing for speech perception and production. Cognitive Science 25, 663-678.
2. Hickok, G., Poeppel, D. (2000): Towards a functional neuroanatomy of speech perception. Trends
Cogn. Sci. 4, 131–138.
3. Poeppel, D. (2001): Pure word deafness and the bilateral processing of the speech code. Cognitive
science 25: 679-693.

What the speaking brain tells us about functional imaging - can language be far behind?
John J. Sidtis
Nathan Kline Institute, New York University
Functional brain imaging with positron emission tomography (PET) and functional magnetic resonance
imaging (fMRI) has become the dominant approach to the study of brain-behavior relationships. In spite of
the growth of this field, the functional significance of data from brain imaging is not well understood
(Sidtis, 2000). Studies on brain activity during speech production and other motor tasks raise questions
about some common assumptions such as whether images obtained during complex behavior can be
decomposed into meaningful subunits, and whether the degree of \"activation\" in a brain region is a
measure of its importance. More generally, these studies suggest that experimental designs that depend on
contrasting imaging conditions may be problematic. Although these issues have been raised in the context
of motor speech studies, the concerns should extend to language studies as well.
A group of 13 right-handed normal volunteers (aged 43 ± 11 years) participated in studies one and two.
A subset of 11 right-handed normal volunteers (45 ± 12 years) participated in study three. All participants
were right-handed, native speakers of English. Images were obtained using [15O]water and PET. All
studies employed the same design: four 90 sec resting scans (R) alternating with four 90 sec task
performance scans (P). Each scanning session was dedicated to a single task.
Tasks were performed for 60 sec. Syllable production (studies 1,2,3) required subjects to repeat the
syllables /pa, ta, ka/ as quickly as possible. Sustained phonation (studies 1,3) required subjects to produce
and sustain the vowel sound /ah/ as long as possible. Lip closure (studies 1,3) required subjects to
repetitively open and close their lips as if they were saying /pa/, without making any sound, as quickly as
possible. Finger opposition (study 3) required subjects to use the thumb of the non-dominant hand to touch
successive digits in sequence (2, 3, 4, 5, 4, 3, 2, 3 ...) at a rate of 1 Hz paced by a tone.
The following regions (left and right) were examined: inferior, middle, and superior portions of
cerebellum in horizontal planes, superior temporal gyrus, transverse temporal gyrus, putamen, caudate,
thalamus, inferior frontal lobe, pre- and postcentral gyri, supplementary motor area.
The first study examined whether syllable repetition could be decomposed into articulatory and
phonatory components by subtracting different constituent tasks. The activity identified when the resting
scans were subtracted from the lip closure scans was compared to the articulatory pattern identified when
phonation scans were subtracted from the syllable repetition scans. Similarly, the activity identified when
the resting scans were subtracted from the phonation scans was compared to the pattern identified when lip
closure scans were subtracted from the syllable repetition scans. The results of these two approaches (task
subtraction and rest subtraction) did not agree, and the results of the resting state subtractions were more
like the known functional anatomy of speech than were the task subtractions. Moreover, activity was
higher in some key brain regions during the constituent tasks than during the syllable repetition task (Sidtis
et al., 1999).
The second study examined the question of whether any form of the imaging data could be used to
predict speech rate in the syllable production task. Using regional data normalized by an estimate of global
flow, a linear combination of an increase in Broca\'s area and a decrease in the right caudate predicted
increased speech rate. However, if the speech production scans were contrasted with resting state data,
performance levels could not be predicted (Sidtis et al., 2003). In addition, it was noted that the blood flow
in Broca\'s area during speech did not represent the highest level of activity in the brain, with or without
subtracting the resting state and blood flow decreased in the caudate.
The third study examined the resting states associated with each task (syllable repetition, sustained
phonation, lip closure, and finger opposition). When the data for the resting states associated with each
task were analyzed, a significant effect of task was observed. Further, when regional blood flows were
correlated, it was found that resting state values from a particular session were more highly correlated with
the task performance values from that session than with the values from resting states associated with other
tasks. Interestingly, the caudate was the only region that responded similarly during task performance and
rest states (Sidtis et al., in press).
Functional imaging is a powerful tool for studying brain-behavior relationships. However, several
assumptions should be approached with caution. The speech studies demonstrate that repeating syllables is
not the sum of the appropriate articulatory and phonatory components. Consequently, task subtraction does
not identify important brain regions. Contrasting speech scans with resting scans does not help to identify
important regions either. Such comparisons eliminate relationships between regional blood flow and
speech rate. Finally, an analysis of resting state data indicates that cognitive set plays a significant role

during the scanning session, further complicating task comparisons. The role of functional imaging in
understanding brain-behavior relationships will grow as we better understand the meaning of these signals.
Sidtis JJ, et al. NeuroImage 5: 490-496, 1999.
Sidtis JJ. Brain and Cognition 42: 75-77, 2000.
Sidtis JJ, et al. NeuroImage 20, 615-624, 2003.
Sidtis JJ, et al. NeuroImage, in press.

Production and comprehension of verbs with alternating transitivity by Greek patients with
Broca’s aphasia
Stavroula Stavrakaki, Zoe Katsarou, Artemis Alexiadou & Sevasti Bostanjopoulou
University College London, Hippokration Hospital Thessaloniki, University of Stuttgart & University of
Thessaoniki
Recent studies on the production of verbs with alternating transitivity (e.g. the boy breaks the glass/the glass
breaks) by patients with Broca’s aphasia have indicated that most of the aphasic patients had more problems with
the intransitive rather than with the transitive construction (Bastiaanse & van Zonneveld 2002).
This study addresses the issue of verbs with alternating transitivity in Broca’s aphasia on the basis of Greek data.
Specifically, we investigate the performance of Greek patients with Boca’s aphasia on the production and
comprehension of unaccusative verbs entering transitivity alternations. In Greek, unaccusative verbs show
evidence for A-movement. However, they do not all show the same morphology. There are unaccusatives with
active morphology (Act), unaccusatives with passive morphology (Nact) and finally unaccusative verbs both
active and passive morphology as shown by the examples in (1-3) below:
(1) i sakula adiase
the bag-nom emptied-Act
'The bag emptied'
(2) i supa kaike
the soup-nom burnt-Nact
'The soup burnt'
(3) a. to ktirio gremise
the building-nom collapsed-Act
b. to ktirio gremistike
the building-nom collapsed-Nact
'The building collapsed'
We suggest that the study of unaccusatives in Greek enables us to investigate the effects that A-movement and
passive morphology have on the aphasic performance.
Two individuals with Broca’s aphasia and agrammatism participated in the experimental study: KS and AM.
They were right-handed and the aphasia was due to an ischemic infarct in the left hemisphere. Two tasks have
been used in this study: an elicited production task (cf. Bastiaanse & van Zonneveld 2002) and a picture-pointing
task. The experimental material included 15 transitive and 15 unaccusative verbs. In the production experiment,
a picture was shown and the verb was orally presented in the first person, as there no infinitival forms in Greek.
The subject was asked to tell in one sentence what was happening in the picture using the verb. The response
was scored as correct when a sentence with the verb in the appropriate transitive or intransitive form was
produced. In the second experiment, the subject was visually presented with three pictures; two of them
corresponded to the transitive/intransitive form of the verb while the remaining one showed a semantic
distracter. The subject was orally presented the transitive or intransitive sentence and should point to the correct
picture.
The results indicated that both patients showed a high level of correct performance on comprehension of both
transitive and unaccusative verbs. However, they did not show homogeneous performance on the production of
unaccusative verbs: on the one hand, KS produced some unaccusative verbs with passive morphology but she
failed to produce unaccusative verbs with active morphology. On the other hand, AM failed to produce
unaccusative verbs altogether. When the patients failed to produce the target unaccusative structure, they
produced the transitive structure in most of the cases. In sum, with respect to the patient KS, the findings show
that unaccusative verbs that appear with passive morphology were retrieved more easily than unaccusatives with
active morphology; consequently, passive morphological marking of unaccusative verbs facilitated verb
retrieval. By contrast, passive morphology was not exploited at all by the patient AM who showed a general
deficit in unaccusative verb production. Finally, the fact that both patients had problems with active unaccusative
verbs indicates difficulties with A-movement in unaccusative structures.

Syllable Structure in Aphasic Jargon
Prisca Stenneken, Roelien Bastiaanse, Walter Huber & Arthur M. Jacobs
Freie Universität, Berlin
Aim of the present study is to further investigate the syllable structure in neologistic jargon aphasia. The frequent
occurrence of phonematic neologisms makes it difficult to identify target words and thus to investigate
correspondences to legitimate words of the standard language. Still, a tool for investigating possible regularities
in aphasic neologisms has been provided by analyses of sonority structure (Christman, 1992). The underlying
concept is the notion of a preferred syllable type (Clements, 1990), reflected by the simple, open consonantvowel
(CV-) syllable with presents a maximal contrast in sonority between its segments. Syllables are regarded
as being less complex, the more they resemble this preferred sonority structure. These theoretical considerations
form the basis of the present comparison of aphasic neologisms to standard German. Analyses of German
syllable structures are considered theoretically interesting, as German contains extraordinarily complex syllable
structures. A general prediction derived from the sonority theory is the frequent occurrence of syllable types,
which are least complex according to their sonority profile, for the aphasic neologisms and for legitimate
German words. Moreover, the preferred syllable structure should be even more prominent when lexical
influences are minimized or when normal phonological processing is blocked; thus a stronger tendency should
be observed in neologisms compared to standard German.
Methods
The neologism data are collected from the spontaneous speech of a jargon aphasic patient, KP, a right-handed,
59 year-old male and native speaker of German. He developed a severe Wernicke's aphasia with neologistic
jargon after a CVA to the left hemisphere with an extensive lesion in the perfusion region of the left arteria
cerebri media. The spontaneous speech is characterized by many phonematic neologisms and few phonematic
paraphasias. Speech production is fluent and does not contain any spontaneous corrections of speech errors or
any signs of lexical search. KP does not produce any phonetically distorted utterances; nor is there evidence for
an articulatory disorder.
The aphasic speech data were collected 7 months post-onset of the CVA in a semi-structured interview.
Utterances presenting phonemic deviations from standard German were categorized as either phonematic
paraphasias or neologisms. Criterion for the classification as a phonemic paraphasia was the occurrence of
phonematic distortions, which allowed identifying a word of standard German and also the process leading to the
distortion, whereas the utterances that – on the level of speech sounds – can not be considered part of the
standard language (Huber, Poeck & Weniger, 1989) were classified as neologisms.
Analyses of the German data rely on the CELEX database for phonological German word forms (Baayen,
Piepenbrock & van Rijn, 1993). In two variants of the analysis, the frequency of the demisyllable types was
computed either as the total number of words containing this demisyllable type, or as the summed frequency of
all words containing this demisyllable. The sonority profile was determined for the neologistic and standard
German syllables according to the sonority theory (Clements, 1990).
Results and Discussion
Results show, that the standard German data and the aphasic neologisms reflect a tendency towards the preferred
syllable type. In both data sets, the most frequent demisyllable types present a maximum sonority contrast in
syllable onsets and a minimum contrast in offsets. A similar tendency is observed in more complex demisyllable
types, though these occurred less frequently and showed the tendency in a less prominent way. Results are
confirmed by additional context-specific analyses, applied to the neologism data, and by comparable findings for
the two variants of analysis of the German data set. Thus, it can be concluded that sonority constrains aphasic
speech even when no lexical target can be identified. Converging evidence for sonority constraints in non targetrelated
aphasic utterances comes from previous analyses of neologisms produced in various language tasks
(Christman, 1992) and of recurring utterances (Code & Ball, 1994).
Comparisons of the sonority structure in neologisms to that in the German language reveal high correlations of
the two data sets, indicating similar distributional patterns. Based on previous findings of a less complex sonority
structure in aphasic speech errors (e.g., Bastiaanse, Gilbers & van der Linde, 1994; Romani & Calabrese, 1998),
the present study investigates whether a similar tendency towards less complex sonority structure is observed in
the non target-related neologisms when compared to legitimate German words. As expected, results confirm a
stronger tendency towards the preferred syllable in the aphasic neologisms, as indicated by increased frequencies
of the theoretically preferred syllable types. These findings are in line with an enhanced influence of the sonority
principle in the impaired language system, not only in phonematic distortions when the target word can be
identified but also in non target-related neologisms. The results of the present study support the view of a default
sonority pattern of segment selection which becomes apparent when phonological processing is impaired.

References
Baayen, R. H., Piepenbrock, R. & van Rijn, H. (1993). The CELEX Lexical Database (CDROM). Philadelphia,
PA: Linguistic Data Consortium, University of Pennsylvania.
Bastiaanse, R., Gilbers, D. & van der Linde, K. (1994). Sonority substitutions in Broca's and Conduction
aphasia. Journal of Neurolinguistics, 8, 247-255.
Christman, S.S. (1992b). Uncovering phonological regularity in neologisms: Contributions of sonority theory.
Clinical Linguistics & Phonetics, 6, 219-247.
Clements, G.N. (1990). The role of sonority in core syllabification. In Kingston, J. & Beckman, M.E. (eds.),
Papers in Laboratory Phonology I. Between the Grammar and Physics of Speech (p. 283-333). Cambridge:
CUP.
Code, C. & Ball, M.J. (1994). Syllabification in aphasic recurring utterances: Contributions of sonority theory.
Journal of Neurolinguistics, 8 (4), 257-265.
Huber, W., Poeck, K. & Weniger, D. (1989). Aphasie. In Poeck, K. (ed.), Klinische Neuropsychologie (S. 89-
137). Stuttgart: Thieme.
Huber, W., Poeck, K., Weniger, D. & Willmes, K. (1983). Aachener Aphasie Test. Göttingen: Hogrefe.
Romani, C. & Calabrese, A. (1998). Syllabic constraints in the phonological errors of an aphasic patient. Brain
and Language, 64, 83-121.

Mental Transformations after damage to the Motor Hand Area:
Neuropsychological Evidence
Barbara Tomasino(1), Miran Skrap (2), Raffaella Ida Rumiati (1)
(1) Programme in Neuroscience, Scuola Internazionale Superiore di Studi Avanzati (SISSA), Trieste, Italy
(2) Unità Operativa di Neurochirurgia, Azienda Ospedaliera S. Maria della Misericordia, Udine, Italy
Recent studies seem to attribute higher cognitive functions to an area that so far has been considered to deal with
lower level motor movements (i.e. primary motor cortex). Hauck et al. (2004) found a somatotopic activation of
the M1 area (and premotor cortex) during a passive reading task of action words referring to face, arm or leg
actions. It is known that imagery for movements of different body parts activate the M1 cortex in a somatotopic
manner according to the body part involved in the simulation (Stippich et al., 2002; Ehrsson et al., 2003) In the
present study, to further investigate the contribution of the M1 cortex to the simulation of movements we studied
Metal Rotation (MR) abilities in patients with lesions invading the M1 cortex. We aimed at investigating
whether the involvement of M1 in MR is stimulus-bound (Exp. 1) or it is strategy driven (Exp.2). We found that
lesions to the hand motor area of the left hemisphere (as opposed to M1 lesions sparing the hand area) impaired
the ability to solve a MR task by imagining a hand grasping a turning movement (i.e. adopting the motor
strategy, Kosslyn et al., 2001; Tomasino et al., in press).
Method
We studied 7 patients with a tumor involving the left (patients PF, GG, OM, JM and GA) or the right M1
(patients LC and HN). The lesion invaded the hand motor area in 5 patients (PF, GG, OM in the left M1 and LC
and HN in the right M1), and spared the hand area in two of them (JM and GA in the left M1).
In Exp.1 we manipulated the type of stimulus. All patients performed MR of hand and of letter shapes. Only
patients with a left M1 lesion invading the hand motor area (as compared with those whose lesion was localized
above or below the hand area representation) failed to mentally rotate hands but succeeded with letters. JM and
GA, were better in mentally rotating hands, than those PF (Wilcoxon, p < .05 and p < .001 respectively), GG
(Wilcoxon, p < .01 and p < .001 respectively) and OM (Wilcoxon, p < .05 and p < .005 respectively). In addition
patients with a right M1 lesion were as accurate as controls in rotating both tasks. LC and HN were better than
PF (Wilcoxon, p < .005 for both), GG (Wilcoxon, p < .005 for both) and as OM (Wilcoxon, p < .05 for both) in
solving the MR of hand task.
But is the involvement of the motor hand area stimulus-bound or rather is it strategy-driven? We therefore
carried out Experiment 2, in which patients were instructed to mentally rotate hands and Shepard and Metzler’s
stimuli using a motor (i.e. MR as a consequence of their own hand turning the stimulus) and a visual strategy
(i.e. imagining the stimulus as rotating in the visual space). Independent of the type of stimulus, patients with the
lesion of the left motor hand area (as compared to those with a lesion sparing the left hand motor area) showed a
dissociation between the two strategies: they selectively failed to use the motor strategy but succeeded in
adopting the visual strategy. In particular in the MR of hand task JM and GA were more accurate than PF, GG
and OM when using the motor strategy (Wilcoxon, p < .001 for GG and OM and Wilcoxon, p < .01 for PF, when
contrasted with patient JM; Wilcoxon, p < .001 for GG and OM and Wilcoxon, p < .05 for PF when contrasted
with patient GA). By contrast all of them performed the MR normally when adopting the visual strategy
(Wilcoxon, all p > .05, n.s.). In addition patients whose lesion involved the right hand area succeeded in
performing both tasks. HN was more accurate than OM, PF and GG when using the motor strategy (Wilcoxon, p
< .001 for PF and OM and Wilcoxon, p < .05 for GG) as it were LC (Wilcoxon, p < .001 for PF and GG and
Wilcoxon, p < .05 for OM). By contrast all of them performed the MR normally when adopting the visual
strategy (Wilcoxon, all p > .05, n.s.). The same significant differences were obtained for the MR of cubes task.
In summary, varying the type of strategy in a MR task of hands and 3D stimuli, we found that,
independent of the type of stimulus: (i) patients with a lesion in the left M1 (and not right M1) selectively failed
to MR stimuli as a consequence of their own action, but succeeded in carrying out the task by imagining the
stimuli rotating in the visual space; (ii) the critical area in the left M1 cortex is the hand representation and (iii)
the strategies are implicitly triggered according to the type of stimulus (Exp 1) but can be adopted during MR of
both kind of stimuli if explicitly required by the experimenter instructions (Exp 2). These results suggest that the
involvement of the M1 cortex in MR is strategy-driven and that the hand motor area supports the simulation of
hand movements. Therefore we argue that areas within the M1 cortex can be selectively recruited according to
the body part involved in the motor imagery process.

The Use Of Tactile-Kinaesthetic Feedback As A Therapeutic Intervention In A Case Study Of
Pure Alexia
Tree, J. J., Skillicorn, B. J. & Kay, J. M.
University of Exeter
The present study reports the case of Eddie, a retired teacher who presents with a pattern of impairment
consistent with pure alexia (letter-by-letter reading) following a stroke. Assessments were carried out to identify
the precise nature of his reading difficulties and revealed considerable deficits in the processing of written letters,
often leading to misidentification of whole words. However, his other language abilities remained substantially
intact. From the analysis of this assessment, therapy was designed and implemented with the aim of improving
Eddie’s accuracy at identifying individual letters to lead to more accurate identification of whole words.
Specifically, a tactile-kinaesthetic facilitation approach (similar to that employed by Lott and colleagues (Lott,
Friedman and Linebaugh, 1994, Lott & Friedman, 1999)) was adopted in which Eddie was taught to trace
visually presented uppercase letters onto his palm. Assessments designed to evaluate the usefulness of therapy
revealed improved performance after therapeutic intervention on uppercase letter identification and reading of
whole words. Interestingly, these improvements did not extend to identification of lowercase letters or to
uppercase visual lexical decision. The potential limitations and implications of the study and areas for further
research are discussed.
Lott, Friedman & Linebaugh (1994) Rationale and efficacy of a tactile-kinaesthetic treatment for alexia.
Aphasiology, 8, 181-195.
Lott & Friedman, (1999) Can treatment for pure alexia improve letter-by-letter reading speed without sacrificing
accuracy? Brain & Language, 67, 188-201.

Syntactic movement and comprehension difficulties in Broca’s aphasia
Ineke van der Meulen (a), Roelien Bastiaanse (b), and Johan Rooryck (a)
(a) University Leiden Centre for Linguistics (ULCL)
(b) Graduate School of Behavioral and Cognitive Neurosciences (BCN)
Introduction
It has often been observed that Broca patients show better comprehension of sentences involving
movement out of the subject position (e.g. subject clefts and relative clauses) than of their object
counterparts (1-3). This subject/object asymmetry is commonly assumed to characterize the
comprehension deficit in Broca’s aphasia and as such forms the basis for many accounts of the
comprehension difficulties of Broca patients.
However, patients’ comprehension of wh-questions does not show a subject/object asymmetry.
English-speaking Broca patients understand both subject and object wh-questions equally well (4-6).
Still, wh-questions also involve syntactic movement. The question thus arises why patients’
comprehension of wh-questions differs from that obtained on other movement-derived constructions.
Does movement of the wh-word not affect patients’ comprehension?
Comprehension of wh-questions has only been examined for English-speaking Broca patients.
Crosslinguistic data are needed to determine whether the intact comprehension of wh-questions is a
general characteristic of Broca’s aphasia. We have conducted a study examining comprehension of
wh-questions in French-speaking Broca patients. The results of these patients will be compared to
those obtained on English-speaking Broca patients. Further, this study aims to answer the question
whether movement of the wh-word affects patients’ comprehension.
Experiment
Comprehension of wh-questions was tested using a picture-pointing task. Each picture represented a
reversible event involving three persons, e.g. a boy hitting a girl, who is in turn hitting another boy.
The experimenter read a wh-question aloud and patients were asked to point to the person representing
the answer to the question. In French object questions, movement of the wh-word is optional.
(1) a. Le garçon frappe qui? (lit. the boy hits who?)
b. Qui est-ce que le garçon frappe?
Both questions in (1) mean Who does the boy hit?. The only difference between these questions is that
in (1a) the wh-word has remained in-situ, whereas it has been moved in (1b). This property of French
makes it possible to examine the effect of wh-movement on patients’ comprehension. If wh-movement
does affect patients’ comprehension, French-speaking Broca patients will have more difficulties
understanding a question of the type in (1b) than that of the type in (1a). The experiment thus
contained subject questions and the two types of object questions illustrated in (1).
Nine monolingual francophone agrammatic patients participated in the experiment (3 men, 6 women,
mean age 51,7 years, mean time post onset 91,9 months). All patients were non-fluent and classified as
Broca patients on the basis of standardized neuropsychological tests. Their aphasia resulted from a
stroke in the left frontal brain regions.
Results
The results are given in table 1.

Table 1
Comprehension of wh-questions (percentage correct)
Patient Subject question Object question
(wh-in-situ)
CA 41,7 83,3 66,7
JD 100 100
AM 77,8 61,1
AR 97,2 94,4 94,4
ST 30,6 83,3 77,8
LD 94,4 100 88,9
MG 27,8 66,7 38,9
SM 47,2 88,9 72,2
BS 33,3 66,7 50
Mean 53,2 84,6 72,2
Object question
(wh-movement)
The results show that patients’ comprehension of wh-in-situ wh-questions is significantly better than
that of object questions involving wh-movement (Wilcoxon: z=-2,41; p=0,016. Secondly, wh-in-situ
object questions are understood significantly better than subject questions (Wilcoxon: z=-2,2;
p=0,028). Thirdly, patients show a tendency towards a better comprehension of object questions
involving wh-movement versus that of subject questions, but this difference is not significant
(Wilcoxon: z=-1,86; p=0.063).
Discussion
French-speaking Broca patients understand wh-in-situ object questions (1a) better than their
counterparts involving wh-movement (1b). This shows that movement of the wh-word affects
comprehension in Broca’s aphasia.
The second purpose of the experiment was to examine whether the findings on wh-questions
obtained on English-speaking Broca patients could be generalized to French. A comparison between
the results obtained on both groups of patients reveals that French-speaking Broca patients score
considerably lower on subject wh-questions than English-speaking Broca patients do (53,2% correct
versus 80% correct respectively). This crosslinguistic difference is unexpected under the standard
characterization of the comprehension deficit in Broca’s aphasia. We will argue that this difference
follows from a difference in the syntactic derivation of subject wh-questions in both languages. In
contrast to French subject wh-questions, English subject wh-questions do not involve movement of the
wh-word, but only of its features. It will be argued that this type of movement has less effect on
patients’ comprehension than movement of the entire wh-word does. This also suggests that the
comprehension difficulties of Broca patients are not related to the position (subject/object) out of
which an element has been moved, but to the type of movement that has taken place (feature/category
movement). A new characterization of the comprehension deficit in Broca’s aphasia will be proposed,
relating patients’ comprehension difficulties to the type of movement through which constructions are
derived. It will be shown that the findings on object questions obtained on French-speaking Broca
patients are consistent with this characterization. This study shows that the use of linguistic tools
offers a new and more fine-grained description of the data of Broca patients.
References
[1] Caramazza, A. & E. Zurif (1976). Dissociation of algorithmic and heuristic processes in language
comprehension: evidence from aphasia. Brain and Language 3, 572-582.
[2] Caplan, D., C. Baker &. F. Dehaut (1985). Syntactic determinants of sentence comprehension in
aphasia. Cognition 21, 117-175.
[3] Grodzinsky, Y. (1989) Agrammatic comprehension of relative clauses. Brain and Language 37,
480-499.
[4] Hickok, G. & S. Avrutin (1996). Comprehension of wh-questions in two Broca’s aphasics. Brain
and Language 52, 314-327.
[5] Thompson, C., M. Tait, K. Ballard & S. Fix (1999). Agrammatic aphasic subjects’ comprehension
of subject and object extracted wh-questions. Brain and Language 67, 169-187.

[6] Avrutin, S. (2000). Comprehension of Discourse-linked and non-Discourse-linked questions by
children and Broca’s aphasics. In: Y. Grodzinsky, L. Shapiro & D. Swinney (eds.), Language and
the Brain: Representation and Processing. San Diego: Academic Press.

Formulaic language in spontaneous speech of left- and right hemisphere-damaged subjects
Diana Van Lancker Sidtis & Whitney Anne Postman
New York University, NKI, & NIH/NIDCD
The notion that nonnovel or formulaic language (FL) may be processed in the brain differently from novel
expressions is based on observations in aphasia and from comprehension studies of right hemisphere (RH)
damaged patients. In normal subjects, counting (Van Lancker, et al., 2003), idioms (Kempler, et al., 1999), and
proper nouns (Ohnesorge & Van Lancker, 2001) may be represented in both cerebral hemispheres.
The relationship of production to comprehension among varieties of FL remains unclear. Comprehension of
familiar proper nouns occurs in both hemispheres of normal subjects (Ohnesorge &Van Lancker, 2001) and in
LH damage (Van Lancker & Klein, 1990) suggesting a role of the RH. In contrast, deficient production of
famous names occurs following LH damage (Semenza, et al., 1995).
Interest in FL has grown (Wray, 2002). Our goal was to examine the spontaneous speech of persons with
unilateral brain damage. We questioned whether classes of FL would be selectively preserved, deficient, or
otherwise unusual.
Method
Natural speech samples were collected from 5 patients with aphasia due to LH damage, 5 patients with RH
damage, and 5 neurologically intact normal-control subjects (NC). LH damaged subjects, including one anomic
and four with Wernicke’s aphasia, ranged in age from 49-76 (mean 62.8), average education of 12.6 years; RH
subjects’ age ranged from 41-70 (mean 59.4) with 12.2 years of education; NC subjects were 46-78 years of age
(mean = 63.6), educated at 15.2 years. All were male except for 2 NCs.
Written transcripts of spontaneous speech, in which subjects described their family and work, were searched by
two independent raters. RH patients’ texts contained a mean of 614.6 words (range = 325-1030), with a mean
length of utterance (MLU) of 11 words; LH subjects’ texts averaged 433.6 (range = 257-587), MLU of 8.8; NC
texts = 518 words (range 367-1022 words), and a MLU of 10.2.
Seven FL categories were chosen for this analysis: fixed expressions (speech formulas: \"let’s see\";
conventional expressions: \"as a matter of fact\"), sentence stems (\"I guess…\"), discourse particles (\"well\"),
pause fillers (\"uh\"), dyads/triads (e.g., \"pepper and salt\"), proper name, and numerals. Criteria developed in
Van Lancker-Sidtis and Rallon (2004) aided in identifying the exemplars.
Results
The proportions of words in the texts comprising FL were compared across groups, revealing the fewest for the
RH group: 16.5% for RH, 29.5% for LH, and 24.6% for NC. A one-way ANOVA revealed a significant main
effect of group (F (2,12) = 7.343, p

Semenza, C., et al. (1995). The anatomical basis of proper names processing: a critical review. Neurocase: Case
Studies in Neuropsychology, Neuropsychiatry, and Behavioural Neurology, 1(2), 183-188.
Van Lancker, D. & Klein, K. (1990). Preserved recognition of familiar personal names in global aphasia. Brain
and Language, 39, 511-529.
Van Lancker, D., et al. (2003). PET activation studies comparing two speech tasks widely used in surgical
mapping. Brain and Language, 85, 245-261.
Van Lancker-Sidtis, D. & Rallon, G. (2004) Tracking the incidence of formulaic expressions in everyday
speech: methods for classification and verification. To appear in Language and Communication.
Van Lancker-Sidtis, D. (2004). When novel sentences spoken or heard for the first time in the history of the
universe are not enough: toward a dual-process model of language. IJLC D, 39 (1), 1-44.
Wray, A. (2002). Formulaic language and the lexicon. Cambridge: Cambridge University Press.

The breakdown of functional categories in Greek aphasia
Natalia Valeonti, Alexandra Economou, Maria Kakavoulia, Athanassios Protopapas, Spyridoula Varlokosta
Panteion University, University of Athens, Panteion University, Institute for Language & Speech Processing,
University of the Aegean and University of Reading
Background and research questions
Verbal inflectional errors are among the most prominent characteristics of aphasic non-fluent speech (e.g.,
Goodglass, 1976). However, a number of studies have shown that such impairment is selective and that not all
verbal inflectional morphemes are equally disturbed. Subject-verb agreement is relatively intact (e.g. De Bleser
& Luzzatti, 1994) while tense is severely impaired (e.g. Friedmann & Grodzinsky, 1997). Two types of accounts
have been proposed to explain these findings. Some researchers view the deficit that non-fluent individuals
exhibit as structural and thus attribute verbal inflectional errors to a breakdown of functional categories and their
projections. Specifically, Friedmann & Grodzinsky (1997) and Grodzinsky (2001) have argued that impairment
in agrammatic production can be characterised in terms of a deficit in the syntactic tree. Agrammatic individuals
produce trees that are intact up to the Tense node and ‘pruned’ from this node up. Other researchers have
suggested that difficulties in the production of particular inflectional morphemes are due to processing
limitations (e.g., Crain, Ni, & Shankweiler, 2001). Within such accounts, grammatical representations are intact
but access to them is impaired. The asymmetry in the performance of aphasic individuals on production vs.
grammaticality judgement tasks has been taken as further evidence for such approaches (Linebarger, Schwartz,
& Saffran, 1983).
The purpose of the present study was to investigate functional categories in Greek aphasia across
different tasks. Specifically, in order to contrast structural and processing theoretical approaches to
morphological impairments in aphasia, we investigated (a) the relative sensitivity of functional categories to
brain damage and the systematicity thereof; and (b) the relation between production and comprehension
performance.
Greek language
(Modern) Greek is a null-subject language with rich morphology and relatively free word order. The Greek
verbal paradigm distinguishes six person inflections, and also makes an aspectual distinction between perfective
and imperfective aspect, which surfaces in the verb stem. The aspectual distinction shows up in the past tense
and in the future. There is no aspectual distinction in the present tense, which always uses the imperfective stem.
Given the richness of the Greek inflectional paradigm, several functional categories are instantiated in the
extended projection of the Greek verb (Philippaki-Warburton, 1987). The order of these categories though
remains quite controversial. According to Philippaki-Warburton (1998) and Alexiadou & Anagnostopoulou
(1996), AgrP is higher than TP (but not according to Tsimpli, 1990). Therefore, the likely clause structure for
Greek is:
(1) CP > MoodP > NegP > AgrP > TP > VoiceP > AspectP > VP
Method
Eight individuals (1 female) aged 42–81 years, clinically diagnosed with aphasia, were referred by speechlanguage
pathologists to participate in the study. All had suffered a left-hemisphere cerebrovascular accident 3 or
more months before testing. Their fluency was assessed using a simple speech rate measurement, there being no
standardised testing materials in the Greek language. Each participant was tested for comprehension, using a
grammaticality judgment task, and for production, using a sentence completion task and a picture description
task (cookie theft and a store situation). Grammaticality judgment and sentence completion used the same set of
sentences, in different sessions, to ensure comparability of comprehension and production measurements. The
sentences were constructed to assess three functional categories: (a) subject-verb agreement (32 items); (b) tense
(16 items); and (c) aspect (32 items). The number of trials for the grammaticality judgment was double that for
production in order to include an equal number of incorrect sentences. Eight verbs were used to construct the

sentences, balancing frequency of use (estimated on the basis of subjective familiarity measured on an
independent sample) and regularity of aspectual conjugation.
Results
The table shows the proportion of incorrect responses for each participant, separately for each task and
functional category. The most striking observation is the range of ability exhibited by this participant group,
from flawless performance (Participant 5, especially production) to total inability to perform the task (Participant
6, no production at all). However, in this diverse group, several systematic patterns emerge upon closer
inspection.
Grammaticality judgment Sentence completion
Patient Agreement Tense Aspect Agreement Tense Aspect
1 3.1 0.0 17.2 15.6 12.5 43.8
2 43.8 62.5 43.8 40.6 93.8 81.3
3 17.2 56.3 46.9 28.1 81.3 56.3
4 1.6 50.0 48.4 56.3 68.8 68.8
5 0.0 3.1 1.6 0.0 0.0 0.0
6 23.4 56.3 48.4 – – –
7 3.1 0.0 10.9 9.4 0.0 12.5
8 0.0 28.1 32.8 15.6 12.5 37.5
All 11.5 32.0 31.3 23.7 38.4 42.9
Discussion
Performance on subject-verb agreement tends to be least impaired, accounting for up to 23% of an individual’s
total errors in comprehension and up to 49% in production. Tense and aspect are more impaired, with errors in
aspect ranging from 33% to 85% of an individual’s comprehension errors and from 34% to 61% in production.
Tense errors are particularly interesting, in that they make up only a small proportion of the total individual
errors of the least impaired participants (1, 7, and 8) but a large proportion of the total individual errors, even
more than aspect errors, for the most impaired participants (2, 3, and 4).
Given the clause structure of Greek shown in (1), a structural account such as the tree pruning hypothesis would
predict that aspect would be least impaired while subject-verb agreement would be more (if not most) impaired,
a pattern opposite from the one observed. Therefore, our findings do not support this hypothesis. Taking into
account that we found a higher proportion of production than comprehension errors in almost every case, our
results would seem to be most compatible with some processing accounts.
Our findings are consistent with the conclusions of Stavrakaki & Kouvava (2003), who found that “high or low
tree position in the sentence hierarchy was not the only determinant of the aphasic performance,” and attributed
the aphasics’ difficulties to impaired access rather than impaired grammatical representations. Similarly,
Tsapkini, Jarema, & Kehayia (2001) proposed a computational load deficit to account for their observations
regarding inflectional impairments in the Greek verb on repetition, reading, and elicitation tasks.

Interpretation of pronouns in VP-ellipsis constructions in Dutch agrammatism
Nada Vasic, Esther Ruigendijk, Sergey Avrutin
UiL OTS, Utrecht University
The results of the studies on reference assignment in agrammatism point towards a selective impairment (e.g.
Grodzinsky et al., 1993). Agrammatic patients have fewer problems when interpreting reflexives (herself) in
simple transitive sentences then when interpreting pronouns (her) in the same constructions. The results of our
previous studies (Ruigendijk, et al. in press) indicate that the impairment of pronouns depends on the
grammatical level at which the operations of assigning the referent to the pronoun occur. In the Primitives of
Binding (Reuland, 2001) a hierarchy of levels at which reference can be established is put forth for non-braindamaged
speakers. According to Reuland, syntactic dependencies are the most economical, and therefore, block
the less economical ones, such as semantic and discourse dependencies. Our previous findings show that this
hierarchy is disturbed in agrammatism: syntax is “weakened” and no longer blocks the establishment of
dependencies at other levels.
The main question for this study is:
How is the economy hierarchy in agrammatism different from that in the unimpaired population, which is:
syntax → semantics → discourse?
In order to determine the relative hierarchy of the semantic and discourse operations in agrammatism, we
investigated patients’ comprehension of VP-ellipsis constructions exemplified in (1).
(1) Bill touched his dog and John did too [e].
This sentence has two possible interpretations representing precisely the focus of our study. The semantic
interpretation (derived from the bound variable anaphora) is represented in (2) and treats the pronouns as a
variable that is assigned reference locally, i.e. in the first conjoint the local NP Bill and in the second conjunct
NP John. (Where λ is an abstract semantic operator.)
(2) Bill ( λx (x touches x’s dog)) and John (λx (x touches x’s dog))
‘Billl touches Bill’s dog and John touches John’s dog
The discourse interpretation (derived from coreference) exemplified in (3) implies that the pronoun is
assigned a discourse referent in the first conjunct, NP Bill in this case, and that the whole VP is then
copied in the second conjunct.
(3) Bill ( λx (x touches z’s dog) & z=Bill) and John (λx (x touches z’s dog) & z=Bill)
Bill touches Bill’s dog and John touches Bill’s dog
Thus, constructions of type (1) will allow us to see whether both semantic and discourse dependencies are
equally available in agrammatism, and which of the two is preferred.
Method
We tested four agrammatic patients (for now) and eight non-brain-damaged adults. In a picture selection task the
participants were presented with the target sentence orally and then asked to choose one out of three pictures that
corresponds best to the sentence they heard. The experiment consisted of three conditions with 10 items per
condition and 30 filler sentences (total 60 items). In the first half of the experiment patients were presented with
the following conditions:
BOUND VARIABLE ONLY (BVonly)
The boy touches his dog and the man does too.
Picture 1: boy touching boy’s dog and the touching man’s dog (grandfather standing next to them with his own
dog)
Picture 2: boy touching boy’s dog and man touching grandfather’s dog.
Picture 3: filler

If the bound variable interpretation (semantic dependency) is available Picture 1 should be chosen; otherwise
they could chose action related filler – Picture 2 or an unrelated distracter Picture 3.
COREFERENCE ONLY (COonly)
The girl touches her horse and the woman does too.
Picture 1: girl is touching girl’s horse and woman is touching girl’s horse (grandmother standing next to them
with her own horse)
Picture 2: girls touching girl’s horse and woman touching grandmother’s horse.
Picture 3: filler
If the coreference interpretation (discourse dependency) is available Picture 1 should be chosen; otherwise they
could chose action related filler – Picture 2 or an unrelated distracter Picture 3.
In the second half of the experiment with the condition where they could choose between the two possible
interpretations:
BOUND VARIABLE vs. COREFERENCE (BVCO)
The witch touches her cat and the fairy does too.
Picture 1: witch touching witch’s cat and fairy touching fairy’s cat (woman standing next to them with her own
cat)
Picture 2: witch is touching witch’s cat and fairy is touching fairy’s cat (woman standing next to them with her
own cat)
Picture 3: filler
Results
Table 1 exhibits the overall results for both aphasics and controls. The agrammatic patients score significantly
above chance the BV and CO ONLY conditions (BVonly: chi-square=9.447, p

The Role of the Verb in Dutch On-line Sentence Processing of Dutch Matrix Clauses
F. Wester a , D. de Goede a , R. Bastiaanse a , L. P. Shapiro b , & D. Swinney c
a Department of Linguistics, Graduate School for Behavioural and Cognitive Neurosciences (BCN)
b Department of Communicative Disorders, San Diego State University (SDSU)
c Department of Psychology, University of California, San Diego (UCSD)
Introduction
In declarative matrix clauses in Dutch, the finite verb appears in second position, having, (according to some
theories) moved from its basic sentence final position (Koster 1975). Bastiaanse et al. (2002) conducted an
experiment in which Dutch Broca's aphasics produced clauses with finite verbs either in verb-second position or
in their basic position. Their results show that the patients encountered fewer difficulties with finite verbs in
clause-final position than with finite verbs in verb-second position. Whether these results are due to verb
'movement', problems with 'traces' in general (Grodzinsky 2000) to 'tree pruning' (Friedman 2002) or more
general processing strategies remains to be investigated.
It is now well established that NP’s that are not in their basic position are reactivated on-line at their
basic position in the sentence (e.g. Love & Swinney 1996). If displaced verbs are re-activated in their basic
position like displaced NP’s, then the fact that verb-second is problematic for Broca's aphasics can be examined
more closely by investigating which reactivation pattern displaced verbs show when tested on-line.
Experiment 1 and Experiment 2
In two cross-modal lexical priming experiments (where sentences were presented auditorily and lexical decision
probes were presented visually) we examined the activation pattern for displaced verbs in Dutch. The
experimental sentences in these experiments are Dutch matrix (SVO) clauses. If verb movement and NP
movement are reflected similarly in psychological reality, we should expect to find activation of the verb directly
after the verb (direct priming), deactivation of the verb in between the overt verb position and the gap, and
finally reactivation at the gap (gap-filling). In both experiments the matrix clause ended at the second argument.
In experiment 1, the second clause was an embedded clause (1) and in experiment 2 it was coordinated (2).
1. De wanhopige verslaafden beroven[1] eenzaam wande[2]lende bejaarden [GAP], omdat[3] ...
(The desperate addicts rob[1] lonely stro[2]lling seniors [GAP], because[3] ...)
2. De arme straatkinderen beroven[1] voornamelijk argeloze, chique gekle[2]de toeristen[3][GAP],
want[4]....
(The poor street urchins rob[1] mainly they don’t know how else to unwary stylishly dress[2]ed
tourists[3] [GAP], because[4]....)
Evidence for 'activation' was priming of related vs. control verb probes, where both probe types were matched
for was extensively pre-tested for any possible inadvertent source of priming. The data was analysed using a
Repeated Measures ANOVA (F1 and F2) and paired t-tests (t1 and t2). In this abstract, only the results for t1 are
presented.
Results
Neither of these experiments provided evidence for reactivation of the verb at its basic position. Instead, they
both demonstrated that the verb was activated immediately and was kept active at points 700 ms later (t1 (40) =
2.53, p = 0.008), 1500 ms later (t1 (56) = 2.49, p = 0.008), and on through the offset of the object head noun (t1
(56) = 2.08, p = 0.021). Evidence for activation for the verb ended immediately following the conjunction
linking the matrix to the embedded clause (experiment 1: t1 (40) = 0.81, p > 0.2; experiment 2: t1 (56) = -0.98, p
> 0.15). See Table 1 for (mean) reaction times.
Experiment 3
Because occurrence of a conjunction definitively signals that no more clausal information will be forthcoming, a
third experiment employed an adjunct rather than a conjunction immediately after the second argument (3). This
allowed investigation of whether saturation of the argument structure of the verb was the basis for discontinued
activation of the verb, or whether only evidence that no other possible elements could occur in the clause

employing that verb caused discontinued activation. The same analyses were used as for the first two
experiments.
1. De domme gedetineerden beroven [1] vijftien rijke bejaarden tijdens hun [2] eerste proefverlof [3],
dus…
(The stupid detainees rob [1] fifteen rich seniors during their [2] first parole [3], so ...)
Results
In this study, verb activation was obtained shortly after the matrix verb appeared (t1 (44) = 3.08, p = 0.002), 700
ms after the onset of the adjunct (at least 700 ms after the last argument had been processed) (t1 (44) = 2.35, p =
0.012) and at the end of the clause (t1 (44) = 2.58, p = 0.007). See Table 1 for (mean) reaction times.
Table 1. Mean reaction times, standard deviations (between brackets) and ‘priming effect’ of all probe points in
three cross-modal lexical priming experiments.
Experiment Probe
Type
Probe
Point
Offset
finite verb
700 ms
down
stream
1500 ms
down
stream
Offset
OHN (end
of clause)
700 ms.
into
adjunct
End of
clause
Offset
conjunction
Exp. 1 Control 633 (68) 635 (61) 626 (72)
Related 617 (65) 621 (66) 620 (73)
‘priming’ 16 (41) 14 (33) 6 (47)
Exp.2 Control 663 (94) 671 (99) 668 (95) 666 (88)
Related 662 (91) 657 (84) 654 (101) 672 (103)
‘priming’ 1 (49) 15 (45) 14 (51) -5 (42)
Exp. 3 Control 721 (86) 723 (95) 712 (89)
Related 697 (89) 706 (96) 693 (88)
‘priming’ 24 (53) 17 (48) 19 (48)
Conclusion
Three cross-modal lexical priming studies reveal that displaced verbs remain active during the processing of the
entire clause employing them, regardless of the position of possible gaps or (possible) arguments. In other words,
Dutch SVO sentences require continued activation of the displaced finite verb, whereas SOV sentences do not. It
seems likely, therefore, that the relative difficulties that Broca’s aphasics encounter with displaced finite verbs in
Dutch matrix (SVO) clauses (Bastiaanse et al. 2002) cannot solely be attributed to impaired gap-filling
strategies, but could also be related to the extra ‘activity’ needed to keep the displaced verb active. Further
research will investigate the behaviour of Broca’s aphasics using the cross-modal lexical priming paradigm, and
similar linguistic materials.
References:
Bastiaanse, R., Hugen, J.,Kos, M. & Van Zonneveld, R. (2002). Lexical, morphological, and syntactic aspects of
verb production in agrammatic aphasics, Brain and Language, 80, 142-159.
Grodzinsky, J. (2000). The neurology of syntax: Language use without Broca's area, Behavioural and Brain
Sciences, 23, 47-117.
Friedmann, N. (2002). Syntactic tree pruning and question production in agrammatism, Brain and Language,
83, 117-120.
Koster, J. (1975). Dutch as an SOV language, Linguistic Analysis, 1, 111-136.
Love, T. and Swinney, D. (1996). Coreference processing and levels of analysis in object-relative constructions:
demonstration of antecedent reactivation with the cross-model priming paradigm, Journal of Psycholinguistic
Research, 25 (1), 5-24.

1. Introduction
Relative Clauses and Object Scrambling In Turkish Agrammatic Production
Tuba Yarbay Duman
University of Groningen (graduate of EMCL)
Agrammatic speakers are reported to overuse the canonical word order of their language and avoid
using non-canonical or scrambled forms. The tendency to avoid non-canonical forms has been verified
for SVO languages such as English, Italian, French and Chinese (Bates, Friederici, Wulfeck, & Juarez,
1988), SOV languages such as Turkish (Slobin, 1991) and Dutch (Bastiaanse, Rispens, Ruigendijk,
Juncos-Rabadan, & Thompson, 2002). Cross-linguistic data support the view that there is a strong
relationship between the agrammatic production and syntactic dependency (Bastiaanse & van
Zonneveld, 1998; Friedmann & Grodzinsky, 1997). However, the relationship of syntax to the
agrammatic production lends itself to different kinds of approaches.
In 1997, Friedmann and Grodzinsky proposed the Tree Pruning Hypothesis (TPH), according to
which the nodes that are high in the syntactic tree, CP and TP, are unavailable for functional
projections. It is not just that functional projections are impaired, but as a consequence the movement
of arguments and adjuncts to the functional domains high in the syntactic tree is impaired as well
(Friedmann, 2002). However, Bastiaanse and her colleagues (Bastiaanse, Rispens, Ruigendijk,
Juncos-Rabadan & Thompson, 2002) argue that syntactic dependency, ‘ movement’ operation, is the
only critical factor in agrammatic production regardless of the position in the syntactic tree (high or
low). This is known as Verb Second Hypothesis (VSH).
2. Research Question
1) Why is the canonical word order simpler for agrammatic aphasics?
2) What would be the implications of movement operations low in the syntactic tree for agrammatic
aphasia?
We concentrated on two types of movement operations low (IP) in the syntactic tree: Turkish relative
clauses and object scrambling (Aygen, 2002). A total of 4 agrammatic speakers participated in this
study.
3. Study 1: Relative Clauses
We compared the production of regular sentences with relative clauses (RCs).For subject RCs , the
participle (P) suffix –(y)An is used and the subject is moved to the end of the clause. For object RCs,
the participle suffix –DIK is used and the object is moved to the end of the clause. Genitive suffix is
attached to the subject of the RC and the possessive suffix is attached to the participle in the object
RCs (Dede, 1982). Examples follow;
(1) Regular SOV
adam ekmek-i kes-iyor
man bread-Acc cut-3 rd prs.
‘the man cuts the bread’
(2) Subject RC
[ Øi ekmek-i kes- en ] adami ]

Ø bread-acc cut-SbjP man
[ RC O V [ S]
“the man who cuts / cut the bread”
(3) Object RC
[adam-ın Øi kes-tiğ- i ] ekmeki ]
man-Gen. Ø Cut-ObjP-Poss bread
[RC S V ] O ] ]
“the bread that the man cut/cuts”
The sentence completion test was adapted from Bastiaanse, Hugen, Kos and van Zonneveld, 2002.
There were three conditions:
(a) regular clause that should be completed with the canonical word order, SOV (no movement)
(b) subject RC condition that should be completed with an NP[object- non-finite verb] subject]]
(+ movement)
(c) object RC condition that should be completed with an NP [subject- non-finite verb] object]]
(+ movement). An example follows;
Subject RC Condition (+ movement)
tester: Bu adam domatesi kesiyor ve bu adam ekmeği kesiyor. Yani, bu domatesi kesen adam ve bu
……[patient: ekmeği kesen adam]
tester: This is the man the tomato cuts and this is the man the bread cuts. So, this is the tomato
cutting man and this….[the bread cutting man]
“this man cuts the tomato and this man cuts the bread. So, this is the man who cuts the tomato and this
is the man who cuts the bread”.
Results
We found a significant difference between the production of regular sentences and subject / object
RCs (x 2 = 19.40, df = 2, p< 0.05). The production of subject RC is significantly poorer than the regular
sentence condition (x 2 = 5.2, df = 1, p< 0.05). The production of object RCs is significantly poorer
than regular sentence condition (x 2 = 17.81, df = 1, p< 0.05).There is a higher performance in subject
RCs compared to object RCs, however the difference does not reach to significance (x 2 = 3.38, df = 1,
p ≥ 0.05).
4. Study 2: Object Scrambling
The canonical position of adverbs in Turkish is just before the verb (Wilson & Saygın, 2001).
However, the definite object can move over the adverb. This is called “object scrambling”. The same
methodology was used to test object scrambling.
a. Unscrambled

Results
[S Bu adam [DO kadını [ADV neşeyle [V selamlıyor ]]]
This man woman-acc joy-IC greet-progr.
“This man greets the woman joyfully”
b. Scrambled
[S Bu adam [ADV neşeyle [DO kadını [V selamlıyor]]]
This man joy-IC woman-acc greet-progr.
* “This man joyfully greets the woman”
We found a significant difference between scrambled and unscrambled condition(x 2 = 10.94, df = 1,
p< 0.01), the former being damaged.
5. Discussion
We showed that the production of non-finite subject / object relative clauses and clause internal object
scrambling, which are all NP movements that are low in the syntactic tree (i.e., IP), are impaired. This
is in line with VSH. The error analysis showed that Turkish agrammatic speakers are able to perform
some types of movements that are high in the syntactic tree (i.e., CP): ‘post-verbal scrambling’
contrary to the predictions of TPH. The results pointed to a disassociation among functional categories
that preserves the higher node (i.e., CP) intact while leaving the lower node (i.e., IP) impaired.
Another disassociation was traced among the members of the same category. Contrary to the
predictions of TPH, none of the subjects made tense errors. Hence, the study presents us with
important cross-linguistic data on the production of verbs as well. Turkish agrammatic speakers
produce finite verbs better than non-finite verbs. When they are required to use a non-finite verb, they
use a finite verb with the correct tense and agreement instead. This pattern was also reported by Slobin
(1991): basic verb form for Turkish agrammatic aphasics is the finite verbs (5 tensed forms). Nonfinite
verbs (both infinitives and participles) hardly occur in their speech. We suggested that the
difficulties that agrammatic speakers encounter may not lie in the representation of the tense node.
This was further supported by the fact that the overt subject-verb agreement, as predicted by VSH, was
problematic.
Syntactic movements low in the syntactic tree are difficult for the agrammatic speakers of
Turkish, but not all word order variations are problematic.
References
Aygen. G. (2002). Finiteness, case and clausal architecture. Doctoral Dissertation, Harward
University, Massachusets, USA.
Bastiaanse, R., & van Zonneveld, R. (1998). On the relation between verb inflection and verb
position in Dutch Agrammatic aphasics. Brain and Language, 64, 165-181.
Bastiaanse, R., Hugen, J., Kos, M., & Zonneveld, R. (2002). Lexical, morphological, and
syntactic aspects of verb production in agrammatic aphasics. Brain and Language, 80,
142-159.
Bastiaanse, R., Rispens, J., Ruigendijk, E., Juncos-Rabadan, O., & Thompson, C.K. (2002). Verbs:
some properties and their consequences for agrammatic Broca’s aphasia. Journal of
Neurolinguistics, 15, 239-264.
Bates, E., Friederici, A., Wulfeck, B., & Juarez, L. (1988). On the preservation of word order
in aphasia: Cross-linguistic evidence. Brain and Language, 3, 323-364.

Dede, M.A. (1982). A syntactic and semantic analysis of Turkish nominal compounds. Ann
Arbor: University Microfilms International.
Friedmann, N. & Grodzinsky, Y. (1997) Tense and agreement in Agrammatic production:
Pruning the syntactic tree. Brain and Language, 56, 397-425.
Friedmann, N. (2002) Question production in agrammatism: The tree pruning hypothesis.
Brain and Language, 80, 160-187.
Slobin, D.I. (1991). Aphasia in Turkish: speech production in Broca’s and Wernicke’s
patients. Brain and Language, 41, 149-164