Université de Montréal

Noun and verb forms in Algerian Arabic: A neuropsycholinguistic study

Département de iinguistique et de traduction Faculté des arts et des sciences

Thèse présentée à la Faculté des études supérieures en vue de l'obtention du grade de Philosophiæ Doctor (Ph.D.) option linguistique

Juin 1997

O Zohra Mirnouni, 1997

Acx(uisiins and Acquisitions et Bibliographic S e ~ ~ c e s senrieés bibiiographiques 395 weangm Street 395, Wellington OthwaON K 1 A W Canada

Obwa ON K1A ON4 Canada

The author has granted a nonexclusive licence allowing the National Library of Canada to reproduce, loan, distriiute or sell copies of this thesis m microfonn, paper or electronic formats.

L'auteur a accordé une licence non exclusive permettant à la Bibliothèque nationale du Canada de reproduire, prêter, districbuer ou vendre des copies de cette thèse sous la forme de microfiche/fïIm, de reproduction sur papier ou sur format élecîronique.

The author retains ownership of the copyright in this thesis. Neither the thesis nor subsîantial extracts fiom it may be printed or othefwise reproduced without the author's permission.

L'auteur conserve la propriété du droit d'auteur qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation.

Université de Montréal Faculté des études supérieures

Cette thèse intitulée :

Noun and verb forms in Algerian Arabic: A neuropsycholinguistic study

présentée par :

Zohra Mimouni

a été évaluée par un jury composé des personnes suivantes:

SUMMARY

In the present research, we addressed the issues of lexical representation, morphological relatedness, and modes of access in Arabic. a language that does not conform to the prototypical iinear *xation

models of the kind found in agglutinative

languages. Our investigation. based on the linguistic theoretical mode1 of lexical

representation developed by McCarthy (1979) for Arabic, was carried out in two consecutive phases. In the fint part. using a senes of off-line tasks, we collected data on the performance of three agramm&c aphasic patients whose native language is A1gena.n Arabic. Even though most of the deficits observed pacalleled those reported in the iiterature, the pattern of omissions and substitutions of bound grammatical morphemes appeared to largely refiect the structure of the language. More specifically, in Algerian

Arabic, where omissions of verb suffixes and prefixes may lead to the production of nonwords, the patients' performance did not display such occurrences. Furthemore,

most of the erroneous forms stemming from omissions resulted in the masculine 3rd singular perfective.

In the second phase of this research we used an on-line task which consisted of an auditory morphological priming experiment. We investigated the process of word recognition of singular and plural nouns in the performance of 24 non brain-damaged subjects and two aphasic patients, a l l native speakers of Algerian Arabic. The critical feature of the plural system of Arabic is that most of the plural f o m do not exhibit regularity in their patterns. Pluralization rnostiy involves stem-intemal changes as in the broken plurals. Plural suffixation, as in the sound plurals, exists but to a much lesser extent. In addition, some f o m may take both broken and sound plurak. Our findings confmed the theoretical distinction between idiosyncratic and non idiosyncratic plural forms. This was reflected in the differentiai processing of the two plural f o m indicating whole word access for broken plurals and decomposition into word and suffix for sound plurals. The priming obtained from morphologically related pairs suggested for Algerian Arabic an architecture of the lexicon compatible in most of its features with the family-like lexical models proposed by linguistic (McCarthy, 1979) and psycholinguistic (Segui & Zubizaretra i 985) theones. Combined features of the mode1 of McCarthy (1979) and the mode! of Segui & Zubizaretta (1985) were proposed to account for the overail data. Morphologically related words are a i i Listed in the lexicon in a linking relationship dominated by the discontinuous root which acts as the head of the morphological family. Members of a morphologicai family are accessed via the head which may be the root or the word (for irregular forms). In the cases of AA verbs, access to verb forms takes place fmt through the discontinuous root, then through the masculine 3rd singular perfective stem. With respect to singular and plural nouns, irregular plurals are accessed through the singular forrn, whereas

regular plurals are accessed fmt through the root, then through the singular.

Assuming this organization of the lexicon. we suggested for the aphasic data in off-

line tasks chat omissions rnay be due to the patient's inability to access the lexicon. Online tesùng showed that the patients' processing abilities were welI preserved.

Raymond Queneau. W dans . l'instant fatal, Paris. 1948.

Ces deux dernières décennies ont été marquées par un foisonnerrient de recherches portant sur L'architecture lexicale, aussi bien sur le plan théorique dans le domaine de la linguistique, que sur le plan empirique comme en témoignent de nombreux travaux en psycholinguistique et neurolinguistique. Parmi les questions sur lesquelles les chercheurs se sont penchés, nous retrouvons celles de l'unité de représentation lexicale, de l'organisation interne du lexique. ainsi que des processus d'accès et de traitement du mot. Une revue de la littérature présentée dans le chapitre premier de cette thèse fait état des différentes hypothèses proposées sur ces questions et nous révèle surtout on domaine très complexe comme l'illustre i'absence de consensus au niveau des modèles qui nous

sont offerts.

vii

En ce qui concerne la première question. linguistes et psycholinguistes ont tour à

tour proposé soit le morphème, soit le mot, soit les deux comme unités de représentation Iexicale. L'organisation lexicale se prête à la même controverse. Dans leur tentative de rendre compte des relations lexicales, les linguistes nous offrent trois approches différentes. chacune reflétant le modèle de représentation lexical adopté. D'aucuns vont exprimer ces relations en termes de règles de formation de mots. Dans les modèles qui prévoient que le morphème est l'unité lexical de base. racines, radicaux et mots vont être reliés entre eux

par des règles de formation de mots. A l'opposé, les modèles qui excluent le morphème du lexique vont se démarquer en restreignant les relations morphologiques aux mots. qu'ils soient idiosyncratiques ou Le pmduit de règles de formation. D'autres soutiennent que seules des règles de redondance peuvent rendre compte des relations entres les items lexicaux. Selon cette approche, étant donné que tous les mots sont listés dans Le lexique, les règles de formation de mots se iimitent à la production de nouveIIes formes.

Par ailleurs. selon d'autres modèles. ces mêmes relations vont prendre La forme de connections phonologiques et sémantiques que les mots développent entre eux. Les relations lexicdes résulteraient du renforcement en parallèle de ces connections. Dans la littérature psycholinguistique, nous trouvons des modèles d'organisation lexicale incorporant la notion de famille morphologique p u r rendre compte des relations lexicales. Par exemple, l'hypothèse de l'organisation en satellites (the 'satellite-entries hypothesis' ) postule que tous les mots sont listés, chacun dans une entrée lexicale séparée.

Les items morphologiquement apparentés sont tous reliés 3 une seule forme qui joue le rôle de noyau.

Une autre variante de la famille morphologique propose. par ailleurs, que tous les mots dérivés sont listés et reliés entre eux. dominés par une racine commune. Libre ou liée, qui va représenter la tête de la famille. Les modèles d'accès Lexical vont à leur tour nous offrir des propositions adverses selon qu'il s'agisse d'un lexique basé sur le morphème ou sur le mot. Dans les modèles qui défendent la configuration d'un lexique morphémique, l'accès aux mots morphologiquement complexes (dérivés et fléchis) se fait par le biais du radical, laissant supposer une décomposition en morphèmes avant l'accès lexical. Un autre point de vue consiste à postuler un lexique qui ne comprendrait que les mots. et par conséquent impliquerait un accès direct de chaque mot en tant qu'unité. Ces solutions ne faisant pas l'unanimité, les chercheurs ont alors recow à des modèles hybrides qui combineraient les

deux possibilités d'accès mais dans des conditions particulières: le processus d'accès au mot en tant qu'unité s'appliquerait aux items connus, alors que le processus de décomposition ne toucherait que les néologismes ou items nouveaux.

Le domaine de la pathologie du langage, et plus particulièrement. le syndrome de I'agrammatisme a également donné lieu à des controverses sur l'architecture et le fonctionnement du lexique. Dans leurs efforts à comprendre les causes sous-jacentes aux omissions des marqueurs morphologiques libres et liés qui caractérisent I'agrammatisme, pour ultimement déterminer si ces déficits linguistiques sont le résultat d'une atteinte sélective d'une des composantes de la grammaire ou éventuellement si les processus impliqués sont les mécanismes d'accès aux représentations de ces composantes, les neurolinguistes nous offrent différentes interprétations. Ainsi, les travaux effectués dans des cadres théoriques linguistiques ont tour à tour proposé l'hypothèse d'un déficit phonologique, morphologique, syntaxique, ou lexical et pst-lexical.

Nous remarquons donc, à la lumière des propositions qui s'offrent à nous que, en fait. plutôt que de trancher. les données existantes invitent à retenir à la fois i'une ou l'autre des différentes hypothèses sur la nature des représentations lexicales et de leur mode d'organisation et d'accès. Noue intérêt pour ces thèmes découle directement de cette ambiguité. Malgré les résultats contradictoires qui les caractérisent. les nombreux travaux sur la structure du lexique 0r.t au moins un point en commun, à savoir. ils ont tous porté essentiellement sur des langues aux processus d'affixation linéaire qui se retrouvent dans les langues du type agglutinatif comme l'anglais, le français ou Malien. Nous nous proposons donc dans cette thèse. dans laquelle nous adoptons comme point de départ le modèle de représentation lexicale proposé par McCarthy (1979). d'étendre la recherche sur les trois questions soulevées plus haut relatives au lexique à une langue telle que l'arabe qui se distingue par une morphologie essentiellement nonconcatenative. Nous nous penchons plus particulièrement sur les processus de reconnaissance et d'accès des noms singuliers et pluriels en arabe algérien. Dans le but d'obtenir une vision aussi complète que possible de ces processus, nous avons incorporé deux types de tâches-'off-line'

(sans contraintes temporelles) et 'on-line' (en temps réel)-que nous

avons adressés à deux types de populations de sujets-non-cérébrolésé et pathologique. Cintérêt pour les formes du pluriel découle des proprietés exceptiomelles de leur structure interne. Deux modes de formation des pluriels caractérisent l'arabe: un pluriel dit 'brisé' qui se traduit par un changement dans la structure interne du singulier (ex.. kursi

'chaise'lkrasa 'chaises') et un pluriel dit 'sain' qui se manifeste par un processus de suffixation (ex.. Ibos 'robe'llbesat 'robes'). Les chapitres 2 et 4 présentent une

description de la structure interne du mot arabe et de la représentation lexicale du singulier et pluriel des formes nominaies.

Un postulat du modèle de McCarthy (1979) concerne la différence théorique au niveau de la représentation lexicale des deux formes du pluriel, ce qui nous a amené aux deux hypothèses de travail suivantes: 1)

Les formes du pluriel sain et du pluriel brisé en arabe algérien donnent heu à un traitement différentiel au cours du processus de reconnaissance, tant chez le sujet noncérébrolésé que chez les sujets agrammatiques.

2)

Cette différence découle des traits spécifiques de ces formes et de leur organisation structurale dans le lexique. Ce travail de recherche s'articule en deux volets. Le premier, qui est consigné dans

le chapitre 3, est consacré aux expériences 'off-line' destinées à analyser les déficits

morphologiques dans la performance de deux sujets agrammatiques dont la langue maternelle est l'arabe dgérien. Les données recueillies font ressortir. à certains égards, un tableau du syndrome de l'agrammatisme similaire à ce qui se retrouve dans d'autres langues. Cependant, lorsqu'il s'agit d'omissions de morphèmes morphologiques liés, les erreurs produites semblent refléter la structure de la langue arabe. Contrairement des langues comme l'anglais où I'omission de flexions verbales entraînent la production de radicaux qui sont égaiement des mots de la langue (ex., look-ed), en arabe, elles aboutissent parfois à des radicaux qui ne correspondent à aucun mot de la langue. Nous remarquons alors que ces omissions ne sont pas produites chez les patients arabes. Ce que nous relevons cependant. c'est que toutes les formes erronées qui résultent de ces omissions ne se produisent pas de façon aléatoire, mais suivent un patron bien particulier:

les sujets utilisent la 3ème personne du singulier accompli et ce, quelle que soit la forme de la cible. Dans une deuxième étape. nous examinons le processus de reconnaissance des formes du singulier et du pluriel dans une tâche 'on-line' de décision lexicale et d'amorçage. Une description détaillée de ces techniques en temps réel est présentée aans le chapitre 2. La population testée regroupe 24 sujets normaux et deux des trois sujets

aphasiques qui ont participé à la première étude. Les résultats obtenus c o n f i e n t nos deux hypothèses, à savoir. la distinction théorique entre les formes idiosyncratiques et

celles qui ne le sont pas se reflète dans le processus différentiel d'accès observé entre ces deux formes. Par ailleurs, la facilitation obtenue entre les items morphologiquement reliés semble conforme aux modèles linguistique (McCarthy, 1979)et psycholinguistique (Segui & Zubizaretta, 1985) qui postulent une organisation lexicale en famille morphologique.

Ces modèles permettent aussi d'expliquer les erreurs morphologiques qui caractérisent la performance des sujets agrammatiques.

Une version finale combinant des propriétés de ces deux modèles prévoit pour l'arabe algérien, un lexique où tous les mots seraient listés, reliés les uns aux autres, à I'intérieur d'une famille morphoiogique, dominées par une forme ou 'tête' par laquelle l'accès serait possible. Dans le cas des verbes, I'accès se ferait par le radical libre. d o n

que dans les cas des noms, la racine discontinue senrirait de porte d'accès pour les formes du pluriel sain, la forme du singulier pour les formes du pluriel bris&. En nous basant sur cette organisation lexicale, et considtrant les capacités bien préservées des patients dans les tâches 'on-line*, nous suggérons d'attribuer les omissions qui caractérisent leur performance dans les tâches 'off-line' à des diff~cultésd'accès lexical.

xii

TABLE OF CONTENTS

List of rabIes ..........................................................................................................

xvi

List of Figures ........................................................................................................

xvii

1.1 Linguistics and the LeXicon .......................................................................

8 8 Il

1.1.1 The Lexicon and Listedness ............................................................ 1.1.2 The Lexicon and ReIatedness .......................................................... 1.1.3 The Lexicon and the Locus of uiflection ......................................... 1.1.4The Lexicon and Productivity ..........................................................

14

14

1 .2 Psycholinguistics and the Lexicon ............................................................ 1.2.1 The Mental bxicon and Listedness ................................................. 12.2 The Mental Lexicon and Relatedness ............................................... 1.2.3 The Mental Lexicon and the Locus of Mection ..............................

22

1.2.4 The Mental Lexicoa and Frequency .................................................

22

1.3 Agrammatism and the Mental Lexicon ...................................................... 1.3.1 Early Descriptions of Agrammatism ................................................ 1-3-2Linguistic Approaches to Agrarnmatism .......................................... 1.3-3 Morphological Approaches to Agpmmtkm ................................... 1.3-4Cross-Linguistic S ~ d i e of s Agrammatism ......................................

23 33 25 26 27

17 17 20

xiii

31

2.1 The Arabic Morphological System ............................................................ 2.1.1 The Eariy Arab Grammarians' Approach: The Binya Asliya H y p o ~ e s i s...................................................................................... 2.1.2 The S trucruralist Approach: The Root-and-Patiem Hypothesis ...... 1.1-3 The Generative Approach: McCarthy's Autosegmental Approach ........................................................................................ 2 .1-4 nie Arabie Lexicon: McCarthy's Hypothesis .................................

34 36

2-4 Issues under Investigation .........................................................................

47

2.6 Conclusion ................................................................................................

48

32 33

Chapter t h e : Study one : Agrammatic aphasia in Arabie ................................................................................

50

3 -4 General Discussion ...................................................................................

73

Appendix 3.1 Examples of structures used in repetition reading alouci, md oral comprehension ............................................................

81

.

Appendix 3.2 Sarnple of spontaneous speech for subject NB (Picnü.e description fiom Puadis, 1991 ..............................................

82

xiv

Chapter four: Study two : The mental representation of singular and plural nouns in Aigerian Arabic as revealed through auditory priming in agrammatic aphasic patients........................

4.2 Scope of the Smdy ....................................................................................

1.2.1 Language Background and Theoretical Framework ........................ 4-22 p h a l F~~ati0L-l in

...................................................................

42.3 Objectives of the Study ....................................................................

4.4 Results and Discussion ............................................................................. 4.4.1 Morphologically Related vs. Unrelated Pairs .........................*........ 4.4.2 Sound vs. Broken PIural Pairs ........................................................ 4.4-3 R - P h d F ' k ................................................................................. 4.4.4 PluralSingular vs. Singular-Plural Pairs ........................................

4.4.5 IIkgdy Sufixed Words .................................................................

Appendix 4.1 List of stimuli: Broken plurals (Monosyllabic stimuli) ..................

Appendix 4.2 List of stimuli: Broken plu& (Bisyilabic stimuli) ....................... Appendix 4.3 List ofstimuli: Sound/suffixed plurais ......................................

5 2.4 Conclusion ...................................................................................... Lexical Representation and Processing of Singular and Plurai Forms in .........................................................................................................

5.3.1 Morphologically Related vs. Unrelated Words ...............................

5.3.2 Broken vs. Sound Plural Forms ..................................................... 5.3.3 Singdar vs- F%md F o m ............................................................... 5.3.4 Bi-Plurai Pain ................................................................................. 5.3-5Illegally-Suffmed Words ................................................................. 5.3.6 Conclusion ...................................................................................... Generai Discussion ................................................................................... 5.4.1 The Nature of Lexical Representations for AA ................................ 5.4.2 The Internai ûrganization of the Lexicoo for A A ............................ 5.4.3 The Role of F ~ q u e n c yin the Lexicon ............................................

Implications for F h e r S ~ d i e ............................................................... s Conclusion ................................................................................................ .............................................................................................................

Appendix 1

List of Balnd Stems (Study One) .......................................................

Appendix II Sentences used in the Tasks of Repetition Reading Moud ax-ld Comprehension (Smdy One) ................................................

Appendix III Sentence used in Repetition and Reading Moud Tasks (Study Onel ......................................................................................... Appendix IV Picnires used in Oral Comprehension (Study One) .............................

Appendix V Cumulative Results of Riming Experiment (Snidy Two)...................

LIST OF TABLES Pages Table I Table II

Table üi

Table IV TabIe V

Table Vi

Table W Table Vm Table IX Table X

Table XI

xvii

LIST OF FIGURES Pages

Figure 1

Halle's mode1 of the Iexicon ..............................................................

Figure 2

Bybee's ( 1988)nework mode1 ..........................................................

13

Figure 3

A subpart of the colle 'glue' family (Segui & Zubizarreta, 19851.......

21

Figure 4

The Arabic word according to the binyu %sliys hypothesis (Bohas, 1984) .....................................................................................

32

Figure 5

The basic word in Arabic according to Cantineau ...............................

34

Figure 6

McCarthy's ( 1979) intemal structure of a word in Classicai

Figure 7

McCarthy's ( 1979) intemal structure of a prefixed word in Classical Arabie ..............................................................................

Figure 8

McCarthy's ( 1979) sttuctured lexical entry for Classicai Anbic .........

Figure 9

Lexical representation of an irregutar plural form ...............................

Figure 10 Taft and Forster's ( 1975)mode1 ........................................................ Figure 1 1 McCarthy's ( 1979) structured lexical entry for Classical Arabic (sh0l-tversion) ....................................................................................

Figure 12 Access procedure for the word ya-kteb 'he writes'.............................. Figure 13 Access procedure for the word kat bat 'she wrorc' .............................. Figure 14 Access prowdure for AA verbs .......................................................... Figure 15 Sound plural smcture ..............................................*..........................

Figure 16 Lexical representation ofsound plural forms ...................................... Figure 17 Lexical representation of broken plural forms ....................................

Figure 18 Bi-pIud f~~

...................................................................................

Figure 19 k x k a l rep=senmtion of bi-plural fmns ............................................

Figure 20 Comparative mean reaction times normal vs.aphasic subjects ...........

9

Figure 2 1 Lexical representation of idiosyncratic forms

--.---..-.-..-. ...

.-- L 12-132

A-..---.

Figure 22 Lexical representation of sound and broken plural f o m ...............----

133

xix

ABBRE VIATZONS

SYMBOLS USED FOR ARABIC TRANSCRIPTION I

j

3

J 6 4

t J

A glottal stop long

B

long u long i

SYMBOLS USED IN THE THESIS

I

-7

3 acc bisyl. BP

cv

f Ind

fint penon second person third person accusative bisyiiabic broken plural consonant vowel ferninine indefrnite

rn

masculine

monosyl N/A Neg nom O

monosyiiabic not avdab1e negation nominative Object zero element object verbiditic pronoun subject plural P~W present clitic pronoun subject singular singular sound plural subject verb object verb verb subject object

0

OVpd

P PLUR pres Pro S

S

SING

SP SV0

v vso

ACKNO WLEDGMENTS

1 wish to thank the many people who helped me dong in this chaiienging enterprise.

First of dl, 1 am mostly indebted to my two thesis supervisors Professors Gonia Jarema and Yves-Charles Morin. Prof. Jarema stimulated my interest in Neurolinguistics. 1 had just started my doctoral studies when she offered me the privilege to engage in research in agramrnatism and leam through her keen guidance and enthusiasrn the many aspects of testing, compiling and anaiyzing data, and above ail, to lead the exciting experience of working with agrammatic patients. 1 am most grateful for her perspicacious comments and continuing positive encouragement throughout the course of this thesis.

I should iike to w d y thank Prof. Mono who helped me further rny knowledge of morphological theory. His vast knowiedge of linguistic theory has been a continuous source of inspiration and enlightenment. His invaluable comments on several venions of this thesis have made me reaiize that perfection is a never-ending process. 1 am particularly grateful for his generosity with his time during our discussions, when suggesting ideas. or handing out references.

xxii

1 would like to express specid thanks to Eva Kehayia for her support and advice

when most needed. Her enthusiasm and keen interest for rny research have been most encouraging.

I would aiso like to thank Professors Glyne Piggot. Michel Paradis and Daniel Bub for suggestions and assistance at earlier stages of this research. 1 am indebted to Leila Ait-Mesbah, speech therapist, and tc Prof. Abdelaziz Bendib

of the Department of Neuro-Traumatology at the Mustapha Hospital of Algien, for kindly

providing access to their patients as well as hospitai facilities. 1 am most grateful to the three patients and ali the subjects who cheerfully participated in this study. 1owe special thanks to Paule Samson, Franche Giroux and Mathew Decter for their

superb assistance with editing. statistics and experimental design.

I wish to express my gratitude to Dr. André Roch Lecours, and to al1 the memben of the Centre de recherche du Centre hospitalier Côte-des-Neiges for integrating me and making me feel as one of theû own. Finally, I am grateful to my famdy here and in Algeria, and to the many friends everywhere who have k e n rny second family. This study was supported by scholarships from FCAR and the University of

Montreal for which I am grateful.

INTRODUCTION

Over the

1st

two decades, an increasing amount of theoretical and empiricai

research in linguistics. psycholinguistics and neurolinguistics, has been conducted to investigate the interna1 structure of the lexicon. More specificaliy, researchers have k e n

concemed with the following issues: (i) the unit of representation in the lexicon. (ii) the intemal organization of the items making up the lexicon, and (iii) the modes of lexical access and processing during normal word recognition. Several claims have been made but each has been chailenged by other snidies. To date, a consensus has not been reached with respect to any of the above issues. With respect to the fmt issue, theoretical linguistic and psycholinguistic accounts

have proposed either the morpherne or the word as the basic unit of representation of the

lexicon. Compromise solutions where both morphemes and words are lexically represented have also been put forward.

Regarding the issue of the intemal organization of the lexicon, three main approaches prevail in theoretical linguistics. each one reflecting the mode1 adopted with respect to the unit of representation. The f i t one attempts to express lexical relations in tems of word formation rules: in morpheme-based models, morphological relations are present between stems and affixes as well as between words. whereas in word-based models, morphologicai relations are restricted to words formed from aiready existing words. A second appmach holds that ail words are listed and that morphological relationships are expressed in terms of redundancy d e s . According to a third approach. lexical relatedness is captured by the relative strengths of the semantic and phonologicai connections developed between lexical entries. Morphological relations mn in parallel to these semantic and phonological connections.

In the psycholinguistic literanire. two fundamentai views have been put forward. The fmt one, referred to as the satellite-entries hypotùesis, holds that aU morphologically related words cluster unifonnly in a satellite relationship around a fom which acts as a nucleus. This hypothesis was first proposed for the intemal organization of SerboCroatian nouns in which the nucleus comsponds to the nominative singular form. The second hypothesis whereby morphologically derived words are Listed as autonomous but related entries suggests a family-like organization. Al1 entries for morphologicaUy related complex words are linked to each other and constitute a morphological family. The common root, whether bound or free, acts as the head of the family. Finally, with respect to the mode of access of mental representations, existing psycholinguistic proposals reflect to a large extent the mode of representation of lexical items. For those models supporting a word-based lexicon, each word has its own lexical

entry and c m be accessed in a direct manner as a whole. In contrast, in morpheme-based

models of the lexicon, cornplex words are accessed via their stems, suggesting that these items are decomposed into their constituent morphemes pior to access. Hybrid models proposing two parallet access routes to the lexicon have dso been put forward. In these models, known words are accessed as whole units whereas novels

foms are decomposed into their morpheme constituents. Research on agrammatic aphasia has provided us with invaluable insights in the comprehension of the functional and architectural aspects of the lexicon. Agrammatism is a pathologicai syndrome characterked by a speech output that is typicaiiy noduent and syntacticdly reduced in terms of both the range of grammatical structures produced and the omission and substitution of morphological markers (e.g.. plurais and tense marken) and free-standing words (e.g., articles and prepositions). One critical issue has been to detennuie whether such language deficits reflect selective impaiments to the components of the linguistic grammar, such as phonology, morphofogy. syntax and semantics. and to

their representations, or aitematively, to the processes involved in accessing these components. The various linguistically-based hypotheses that have been offered suggest that agrammatic disorden reflect either a phonologicai impairment related to phonologicai stress, a fundamental impairment in syntax, a morphologicd impairment of the items that do not belong to major categories, or an impairment at both the lexical and post-lexical levels. When reflecting on the studies and proposais reviewed up to now. one notes that most of the research addressing the issues of lexical representation, organization, and access has actually focused on languages which exhibit word formation processes based

on concatenative affixation of the type found in English, i.e., words are made up of

sequences of one or several morphemes that are put together in a linear order. In an effort to fil1 the gap created by the absence of data on nonconcatenative languages, and at the

same time validate the existing models of lexicai access and representation, research on the lexicon is extending its scope to typologically different languages. The goal of the present research is to contribute an additional step towards this

general effort. Adopting as a point of departure the lexical representation proposed by McCarthy (1979) for Arabic, we investigate the process of word recognition of Arabic singular and plural nouns in the performance of twenty-six native speakers of Aigerian Arabic. two of whom are agrammatic aphasic patients and twenty four are non brain-

damaged subjects. Our methodological design incorporates both on-line and off-line tasks.

Unlike languages like English or Italian. where the prototypical process of word

formation is linear affmation, i-e.. in which words are made up of sequences of one or more segments or morphemes that are concatenated in a linear order (e-g., the English word truns-form-ut-ion),Arabic exhibits a nonconcateaative morphological structure

characterized by a word formation process expressed mostly through an infixation process or a change intemal to the word itself (e.g., k i t a e b 'book'/kut ub 'books').

Most

theoreticai account. of Arabic morphology overtly or implicitly make the assumption that the word formation process is essentially root-based, in that words are generaily derived

from a discontinuous consonantal root (e.g., katab 'he wrote' is denved from the discontinuous root /k-t -hi). Furthermore, words are made up of three morphemes-the discontinuous root, the vocalic basis. and a CV prosodic template.

Suffixation and prefixation are also typical operations in Arabic word formation (e.g.. me -kt ab8 'library'lrna -kt sbe- t 'libraries' ). These two fixation processes are

present in Arabic plural formation, the morphologicd process we set out to investigate in the present study. PluraIization in Arabic involves either suffiation as in the sound plural

(e-g.. Ibos 'dress'/lbas~'dresses'), or stem-intemal changes as in the broken plural (e-g., Qrsl 'chair'lkrgç~'chairs*). A theoretical analysis of the Arabic morphologicai

system and noun plural formation is presented in Chapters two and four.

In Chapter one, we present an overview of the studies on the lexicon. with a special emphasis on those addressing the issues of lexical representation, organization, access and processing, as well as to closely related issues such as the locus of idection, and the

role of frequency and productivity. The prominent part played by neuropsycholinguistic research in the understanding of the complex architechire of the lexicon is also stressed. Chapter two is concemed, in its Fust section, with the central constructs of Arabic morphology. Three different approaches to the Arabic morphologicd system are outlined: the early Arab grammarians' view, the suucturalist root-and-pattern hypothesis. and the

generative autosegmental approach. In the following sections, we present the issues under investigation and the hypotheses underlying our study. These will be preceded by a description of the lexical decision and the prirning techniques. Chapter three is a study on the breakdown patterns of morphemes in Algerian Arabic. It investigates the ways in which agrammatism is manifested in Algerian Arabic, and how it compares with the existuig descriptions of the syndrome. The aphasies' error patterns are contrasted across different modalities with those described in the literature,

then the role of language-specific features in the patterns observed is examined. In Chapter four, we present our second study in which we conducted an auditory morphologicai priming experiment on twenty-four non brain-damaged subjects and two

agrammatic aphasic patients. The experimentai design, which involves essentially Arabic

singular and plural noun forms. airns at exarnining the representation and access of such

f o m in the lexicon. A comparative analysis between normal and pathological data is then performed.

The fint part of Chapter five is devoted to a surnmary of the findings. The second part discusses the subjects' overall performance, within and across tasks.

An

interpretation in the light of current Linguistic and psycholinguistic theones on the lexicon

is proposed.

CHAPTER ONE THE LEXICON: R E U T E D ISSUES

The term lexicon is often used in the literanire with systernatic ambiguity, referring either to the component of the grammar where lexical facts are hande& or to a simple list of words or dictionary. Furthemore, depending on whether we are deding with the

linguist's or the psycholinguist's conception of the lexicon, it is refened to either as the lexicon or the mental lexicon. Psycholinguists use the latter labeling in their attempts to

develop models of language use and access. In Chomsky (1970),as in much of the Linguistic literature, the part of the grammat which deds with iexical relations is refemd to

as the lexicon. For the sake of consistency and to avoid confusion, in the present research, the term rnorphofogicalcomponent is used interchangeahiy with the t e m Iexicon to refer to that component of the grammar which handles lexical relations.

1.1 LINGUISTICS AND

TBE LEXICON

1.1.1 The Lexicon and Listedness

In linguistic theory, the lexicon is usually described as consisting of lexical envies

which are presumed to indude information about the pronunciation. the meaning and the morphological as weil as the syntactic properties of the linguistic elements they represent. Among the issues related to the lexicon, the yet unresolved ones are about the nature and

the organization of the items which make up the lexicon. The approach taken by the

structuralists is to include in the Lexicon a List of morphemesl and those items that are idiosyncratic, adopting thus Bloomf~eld'sd e f ~ t i o nof the lexicon: "A complete description of a language wîii list every forrn whose function is

not determined either by structure or by a marker [... 1; it will include. accordingly, a lexicon, or List of morphemes, which indicates the fom-class of each morpheme. as weii as Lists of al1 complex forms whose function is in any way irregulai' (Bloomfield, 1933:269).

On this view, al1 predictable nonarbitrary redundancies are excluded from the lexicon. The morpheme becomes then the basic linguistic unit in the lexicon around which non idiosyncratic words are constructed. This compositional view of the lexicon wiil be taken later by some of the generative morphologists, who also include in their models a set o f word formation d e s (WFRs) for constructing words out of morphemes.

Halle (1973) is one of the earliest generative linguists to propose a morphological mode1 within a generative frarnework. He proposed a lexicon which comprises a list of the morphemes of a language and a set of word formation rules for combining them. These rules provide information on the order in which morphemes are to appear in words. The outputs of the WFRs go fmt through a filter whose functions, among othen, are to -

-

In the Ameflcan srructuralist view. the morpheme is defined as the 'minimal meaningful elemtd.

add idiosyncratic information to words and to mark accidental gaps2 with the feature [-lexical insertion] to prevent them from king incorporated into a syntactic structure. For example, a noun like arrivai which is formed by combining the verb arrive with the f l i x -al, normdy means 'the act of V-ing'. However, a word such as recital does not fit into this pattern. This exceptional meaning of recital must be recorded in the fdter. The outputs of the filter form the dictionary of the !anpage which would then serve as inputs

to lexical insertion transformations. The dictionary contains ody and all infiected fomis of the words of a language. With respect to the list of morphemes, it should be noted that

Halle proposes that each morpheme be specified syntacticaiîy for major grammatical categones, stems, and &xes, both derivational and inflectional, as well as for additional grammatical properties. The model which Haiie proposes c m thus be schematized as follows:

+

Formation

+

Fiter

+

Dictionary (

Figure 1 Halle's mode1 of the lexicon

The problern with Halle's theory is that his model actuaily contains two listings: the

List of Morphemes which, together with the WFRs, defines the potential words of a language. and the dictionary, where only actual words of a îanguage are stored. Aronoff (1976) dispenses altogether with the notions of morphemes, filter, and the two listings. While maintaining Halle's view of a separate component of the grammar

*

In Halle's model, potentiai but not anested words Iike *anïvarion are e x p t e d to M produceci by WFRs; they are considered as accidentai gaps-

t and thus can be

which houses morphological formation, he proposes, instead of a compositional lexicon, a lexicon that has the form of a dictionary where only words may be entered as independent. fully specified items. WFRs apply to existing words3 and not to morphemes. He rejects the notion of the rnorpheme as the minimal lexical unit of semantic representation. arguing that certain rnorphemes "have no rneaning which can be assigned independently of each of the individual words in which they occui' (Aronoff, 1976:9-IO), and gives the word hi11 authority. This is fomuiated in the Word-Based Hypothesis. "Al1 regular word-formation processes are word-based. A new word is

forrned by applying a regular rule to a single existing word. Both the new word and the existing one are members of major lexical categories" (Amnoff. l976:2 1). A consequence of Aronoff's morphologicd theory is that words Like permit and

submit. even though polymorphemic, are not produced by mie. Since morphemes are excluded from this model, affixes are introduced by WFRs. In other words, the representation of an affix is the WFR which attaches the &x

to its base. The word

becomes then the basic unit from which complex words are constnicted. However, it is noteworthy to point out that this approach does not support the hypothesis of a full listing that may be drawn from the word-based lexical organization proposed by the author. In

fact, not d words are listed in the lexicon as Aronoff clairns that the output of productive WFRs as well as regularly inflected items are not listed4. The fmt daim stems from the application of a lexical principle which the author refers to as blocking. Blocking. which

is defined as "the nonoccurrence of one form due to the simple existence of another"

(Aronoff,1976:43) prevents such words as *derival and *arrivafion from k i n g listed since their synonymous forms derivation and arriva1 aiready exist in the lexicon.

In Aronoffs (1995) view. the word is a lcxetne, Le.. an uninflected entity wirhout case. nurnber. or gender. Note. however. that in his theory. inrgularly inflected foms are listed in the lexicon.

However, if applied systematicdy, the blocking principle would not allow the formation of a word like gloriousness since it has a kted synonymous form glory. In order to allow the output of productive WFRs such as -ness in spite of blocking, Aronoff (1976:45) suggests that such foms are not entered in the lexicon. Aronoffs lexicon wili, thus, include the idiosyncratic words of a language. the words produced by the non-productive

WFRs and onlv those words created by productive WFRs which carry idiosyncratic features (the notion of productivity wili be discussed in the section on the lexicon and productivity below). 1.1.2 The Lexicon and Relatedness

Different proposais have been made regarding the interna1 organization of the lexicon. Ways in which relations between lexical items are accounted for depend on the model proposed. More specifcaily, in trying to characterize the types of relations which

hold for example, between pairs of English words such as arriveAarrival, deci&/decision, or refer.prefer, three approaches prevail. One is taken by Halle (1973) and Aronoff (1976), who attempt to express lexical relations in terms of word formation rules.

In Halle's model, relations are present between stems and affixes as well as between words. For example, in the word vacant the WFR [STEM+ant]& which attaches an

affix to a bound stem to form an adjective links. the stem vac- to its affjix -ant. Other words. such as the pair arrive/arrival, are related duough the WFR W R B + al]^, which combines affixes to words. In Aronoffs model, a relation between stem and affix such as the one found between vac- and-ant in Halle's model does not exist As generative niles. WFRs apply only to derivational morphology. thus estabiishing relations between pairs such as arrive/arrival and happy/unhappy.

However. WFRs act also as redundancy rules to

express relations between already existing words such as Vregularly iafiected forms (e.g.. d m e n ; take/took).

The other approach with respect to the ways relations are expressed in the lexicon is adopted by Jackendoff (1975). Jackendoff is the first linguist to provide us with a morphological approach explicitly based on morphologicai relations expressed in terms of lexical redundancy rules. In his model, al1 words of a Ianguage have separate but linked lexical entries; a set of redundancy d e s will serve to signal, for a given lexical entry, any redundant information aiready present in a related lexical item. According to the FuiI-entry theory proposed by Jackendoff (1975). words iike decide and decision have distinct, f d y specified, lexical entries. The two lexical entries are related by a redundancy d e which "designates as redundant that information in a lexical entry which is predictable by the existence of a related lexical item; redundant information will not be counted as independent" (Jackendoff. 1975643). The originality of this theory lies in the role of the redundancy d e s , which are not used as generative d e s but only as evaluation d e s . Along the same lines, at least as far as relatedness is concemed, Ford and Singh (199 1) propose a unified theory of morphology essentially based on the formal relationships between words. A word, whose raison d'être is determined by the use the speaker of a language makes of it, entertains relationships with other morphologicaliy related words. Formal differences between morphologicaUy related words are accounted for by a set of morphological strategiess which, according to the authors, are part of the speaker's cornpetence. In this theory, the speaker's intemalized lexicon is viewed as containing all words he has learned whether they are idiosyncratic or not.

In addition to expressing relations between two words. morphological strategies may be actively used in the production or comprehension of new forrns.

Finally. the last. and least common, approach in theoretical linguistics has k e n put fornard by Bybee (1988). who proposes a lexicon where lexical relatedness is captured in terms of direct semantic and phonological connections between lexical entries and the relative suengths of those connections. One of the basic principles of this model is that morphologicd relahons are defmed in terms of semantic and phonological connections that

run in parallel.

l

1

k

z

t

m æ t s

k æ t

s

Figure 2 Bybee's (1988) network mode1

Figure 2 represents a network in which the word cats shares both semantic and phonologicd connections w ith the singular form cat. The word cats is also connected to other plural forms such as mats, rats, mps, etc., on the basis of the plural feature and the final fricative. The other theoretical construct that characterizes Bybee's mode1 is the

concept of lexical strength. Lexical entries have varying degrees of lexical strength due primarily to their token6 frequency. According to the author, "fiequently used fonns gain lexical strength and foms that are not used Iose lexical suength" (Bybee. 1995131). Thus, lexical strength is used to expfain why English irreguiar forms are usually of high

token frequency. Bybee's model further postulates varying degrees of relatedness among

6

Token frequency refers to the frequency of particular instances of a given type. Type fiequency refers to the number of different forms occuring with a particular affix (Baayen & Lieber. 199 1).

words determined by the nature and number of shared features. For example, walk is more closely related to waiked than deceive is to decepton. 1.1.3 The Lexicon and the Locus of Inflection

It is generally assumed that the pmcesses which operate on the structures of words

are of two types. denvational and inflectional. The formal difference between the two is

somewhat diffïcuIt to establish. Informaily, one cm Say that the fmt one is regarded as producing a new lexical item, ofien with a Merent syntactic category (e.g.. the process allowing an Adjective to form a Noun as in happy-hrrppiness), whereas the second changes the grammatical form of an item (e.g., Noun singular into Noun plural as in catcats). Furthemore, in the formation process, inflectional affixation most often applies

after derivation (e.g., employ+ee+s). Even though this theoretical distinction between inflection and denvation is not universaiiy supported by bguists, the locus of inflection remains the most debated issue in morphology. giving rise to two opposing fronts. those (Halle, 1973; Jackendoff. 1975; McCarthy, 1979) who place both derivational and

inflectionai processes in the morphological component (Strong lexicalist hypothesis) and those who claim that inflectional operations should be excluded from the lexicon (Weak lexicalist hypothesis) and should be handled either by the syntax (Aronoff, 1976) or by both syntax and phonology (Anderson, 1982). 1.1.4 The Lexicon and Productivity

Productivity is the term used by a nurnber of Linguists for various constraint effects on the content of the lexicon. The notion of productivity was brought back to theoretical scrutinity by Aronoff (1976). who used it as the crucial factor in determinhg the contents

of the lexicon. It is important to indicate that Aronoff restricted the notion of productivity to denvational morphology. Even though the term productivity has never been given a

clear d e f ~ t i o nit. has been widely associated with the active use of some of the rules of a language. The more words and new words a mie produces. the more productive it is. Studies of the Engiish intlectionai system show a highly productive regular pattern by which the suffur -ed is attached to the verb base. The nile attaching the affix -ness to

an adjective as in happiness is also considered to be productive in that it is actively used in the language and may attach to any adjective. However, the rule attaching the affix -th as in wannth is considered unproductive: it is found only in a very limited number of words

and cannot be used to fonn new words.

Aronoff (1976) argues that productivity cannot be recognized by the mere cornputation of the number of forms a WFR creates, and that there is no procedure for computing productivity. Furthemore. a number of factors interplay in the productivity of a WFR. One of hem is semantic coherence, i.e., each time a WFR produces a word whose meaning is predictable by the mie, we are dealing with a productive rule. In

English, for instance. al1 #ness derivatives of the form Xousness (e.g.. callousness: the fact that/the extent to which Y is Xous) have more or less the same meaning relationships with their base. It is therefore possible to predict the meaning of any noun of the form Xousness from that of the adjective. The WFR attaching #ness to Xous adjectives is thus considered semanticdiy coherent and productive. In contrast, the +ity derivatives (e.g., variety) of words of the fom Xous (e.g., various) have severai different meaning

relationships w ith their base, and therefore, iack semantic coherence (Aronoff, 1976:38). Productivity may also affect listedness. Under Aronoff's view of the lexicon, whereby "only those words which are exceptional, Le., arbitrary in at Ieast one of theu various features will be entered in the lexicon" (Aronoff. 1976:43), it foilows that words such as varieV rnust be listed whereas words like callousness, which are the products of

productive rules. need not be Listed. Apart from this lexical distinction between productive and non-productive niles. Aronoff ( 1976) does not establish forma1 criteria for establishing whether a WFR is productive or not.

In subsequent research, Anshen and Aronoff (1981, 1988) introduce the frequency factor to account for the productivity of derivational affixes. The authoa designed experiments based on the English denvational affixes -ness and - i in~ order to investigate the relationship between the productivity of a morphological pattern and the number of

tokens that a speaker produces on that pattern. In their 1981 study. they report that their subjects, when presented with possible but non-occkng words formed according to various Iegitimate patterns. more readily accept words like *stimulutivenessor *edibility that correspond to the patterns found in a relatively large nurnber of dictionary occurrences

than words like *stimdativiîy or *ediblenesswhose patterns do oot. Productivity is thus defmed in tenns of the type of morphological patterns new f o m will take, and no longer in terms of the number of existing forms.

Along the same lines, Bybee (1988. 1995) daims that frequency is an important dimension in the lexicon and proposes a rnorphological mode1 in which type frequency of

a morphological pattern is an important determinant of productivity. The author argues that high type frequency contributes to productivity. Furthemore, in order for a given

pattern to attain full productivity, it is necessary that no restrictions-phonological, sernantic or morphological-apply

to it. The more open the pattern7, the greater its

productivity. This is the case of the English past in -ed or plural in -S. Baayen and Lieber (1991). in a corpus-based study on English derivation, c o n f m other researchers' claim that word type counts can be misleading. The authors argue that Bybee (1995:430) uses the word schernas to refer to these patterns. She defines them as "emergent generalizations creatcd when two sets of words with similar patterns of sernantic and phonological connections reinforce one another".

"a rneasure of productivity based on the token frequencies of types, specificdy on the number of hapax legomena for a given H i in a corpus, cornes very close to according with Our intuitions about productivity" (Baayen & Lieber. 1991:801). This proposed rneasure is then used by the authon to compare the degrees of productivity of various English affixes.

1.2 PSYCHOLINGUISTICS AND THE LEXICON

1.2.1 The MentaI Lexicon and Listedness Even though psycholinguistic models have been primarily concemed with lexical access and processing and have not addressed explicitly the question of listedness. they,

however, reflect to a large extent the internal structure and organization of the lexicon.

Three major hypotheses prevail in the literature.

The Decompsitional Hvpothesis Early psycholinguistics shidies (Mure11 & Morton, 1974; Snodgrass & JarveUa, 1972) on access to the internal lexicon suggest that words are stored in their constituent

components (stems and affixes).

However, the fmt f o d i z e d theory was the one proposed by Taft and Forster (1975). The main claim made by the authors is that a process of morphological

decornposition takes place pnor to lexical access. Taft and Forster used experimental evidence collected on words which were assumed to have a decompositional intemal stmcnire. For example, a word like rejuvenate was assumed to consist of a reai prefix rewhich contributes meaning to the word as a whole, and a bound stem -juvenate, as opposed to a word like repertoire which was assumed not to have an internai structure (re-

is thus considered here as a pseudo-prefix in that it does not contribute any meanhg to the word). Their main concern was to investigate whether a word like rejuvenate wouid be stored and thus accessed as a whole word, or stored under the entry juvenate-even though the stem juvemte does not stand as a word on its own-and

therefore. accessed in

a decompositionai manner. Even though their study was based essentially on prefixed words and pseudo-prefixed words, the authors make the general daim that "a morphologicai analysis of words is attempted prior to lexical search" (Taft & Forster, 1975:643). They are, however, very cautious when it cornes to the issue conceming the form in which words are stored. In this respect. the authon present us with two possible interpretations. both compatible with the decompositionai hypothesis. As a fust possibility, one may assume that for rejuvenate, it is the bound stem -juven- that is actuaiiy Listed whereas for the word repertok it is the whole form that is listed. One could also take a different stand and postulate that rejuvenate is üsted as a whole word together wiih its intemal structure re(juven)ate)).

In subsequent research (Taft,1979; Taft & Forster, 1976). the authors introduce substantial changes to their initiai model. One important contribution in the revised version to the hypothesis of an atomic lexicon is the explicit assumption by the authors that prefixes are listed in the lexicon.

The Fuii-Listin~Hy~othesi~ In a 1977 study comparing lexical access to affixed words (words consisting of two morphemes: a root morpheme and a suffi) and non-affixed words (words consisting of a single morpheme). Manelis and Tharp found no difference between the two types of

words. which Ied them to favor an independententry hypothesis according to which ail words have separate lexical entries. One of the main proponents of this view is

B utterworth ( 1983) who proposes the Full Listing Hypothesis (FLH). B utterworth envisions the lexicon as containing all words a speaker knows, be they regular, h g u l a r , derived or inflected. Under this view, each word has its own lexical entry and can be accessed in a direct manner as a whole. B is assumed that parsing is not needed in normal word recognition but is resorted to as a fall-back procedure in the case of novel words or previously unheard forms. Approaches favoring Listing of ail words have been adopted by other mearchers (Bradley, 1980; Lukatela et al., 1978; Segui & Zubizarreta, 1985). Wvbrid Models

Stanners et ai. (1979) found evidence that access to the lexicon did not operate according to a unitary principle for regularly inflected (e.g., pours-pour), irregularly inflected (e-g.. hung-hang), and derived (e.g., selective-select) forms. The former appeared to involve direct access of the base implying a prior decomposition into base and suffm whereas the latter two types of items were found to be accessed directly as single units. Even though the authors did not adopt nor propose a specific mode1 to account for

their results. they suggested in their concluding comments that theories of lexical access and processing include more than a single mechaoism.

One mode1 fonnulated to account for the hypothesis that two types of lexical storage

may exist is the Augmented Addressed Morphology model (Caramma et al., 1985; Laudanna, Badecker & Cararnazzg 1989). This processing model proposes two parallel routes to access morphologically complex words: (i) a whole-word address procedure used for known words and, (ii) a morpheme addnss procedure used for novel f o m . Two important principles characterize this model: (i) morphological decomposition takes place at the level of activation, that is, for known words a letter string simultaneously

activates both whole-word representation and the representation corresponding to the

morphemes that make up the word. and (ii) access to a whole-word representation is faster

than access to the morphemic representation. 1.2.2 The Mental Lexicon and Relatedness

Evidence that morphological relatedness plays a key role in word access and recognition was fmt formulated in psycholinguistic processing Literature by Stanners et al. (1979). Following Taft and Forster (1975), the authors explained the results they found for inflected pairs like lending-[end as reflecting decomposition into base fonn and affix. However. failure to observe simila. results for derived word pain iike select-selective led them to suggest separate, but linked, lexical entries for derived forrns and their corresponding base forms.

Other models have explicitly incorporated morphological relations in the lexicon. The satellite-entries hypothesis proposed by Lukatela et al., (1980), based on experiments

on Serbo-Croatian, holds that (i) every inflected noun has its own lexical entry, and (ü) all morphologically related words cluster uniforrnly in a satellite relationship around the nominative singular which acts as a "nucleus".

The morphological-farnily hypothesis defended by Segui and Zubizaretta (1985) was put forward on the b a i s of experimental data obtained on French derivationallyaffixed words. The morphologicd family consists of a set of independent but linked lexical entries each representing a morphological relative. Two crucial features characterize the morphological family : (i) a rwt (whether bound or free) that ail members of the family share functions as the "head" of the family, (ü) the morphological Links that

memben of the family entertain together are directionally8 related so that they could reflect denvational background (Fig. 3). The authors believe this would prevent recunivity h m taking place in the lexieon.

+ +V

able

décollable

Figure 3 A subpart of the colle 'glue' family

(Segui & Zubizamta, 1985)

The concept of links between morphologicdy related items is also found in the connectionist models of the lexicon. In DeU's (1986) network modelg, the Lexicon consists of interconnected nodes which represent Ievels such as the word. the morpheme, the syiiable, the syllable constituents, the phoneme and the phonetic features. Words have

separate representations but are connected to words sharing any or most of the above levels. Therefore, morphologically related words constitute distinct nodes connected to each other via the cornmon stem-morpheme node as well as to cornmon syllable and

phoneme nodes for the shared stem. 1.2.3 The Mental Lexicon and the Locus of Inflection

Even though linguists have been debating for the last two decades the issue of the theoretical distinction between inflection and derivation and more specifically the locus of their respective processes. psychoiinguists seem to be less concerned with this issue.

ûell's mode1 has k n designed for production.

While being aware of the existence of two distinct affvration processes. they generally assume that denvational and inflectionai processes take place in the lexicon.

In a study on three German-speaking agrammatic subjects on various tasks, de Bleser and Bayer (1988) notice that the patients show a good retention of inflected forms

as well as al1 types of morphological forms in the face of a strong inability to access syntactic representations. Such a picture of the agrammatic deficit led the authors to claim that the only theoretical framework that could account for data From agrammatic aphasics is the one that favoa that inflections be part of the morphological operations that should be

included in the lexicon. 1.2.4 The Mental Lexicon and Frequency

While ignoring the issue of the locus of infiection and its effect on the lexicon, psycholinguistic studies have stressed the importance of another variable, namely frequency. Frequency has been shown to be a deterininhg factor on the way words are organized and accessed in the lexicon and coosequently on the modeling of word recognition processes. Taft (1979) reported a stem frequency effect for regularly inflected fonns rnatched for frequency on a whole form basis but which ciiffer in the frequency of the stem to which they are related. It has also b e n demonstrated in several experiments (e.g., Bradley, 1980; Cole. Beauvillain, & Segui. 1989) that the recognition of a morphologically cornplex word depends on its surface frequency as well as on the frequency of the other members of the morphological family.

Stemberger and

MacWhinney (1988) assert that high-frequency regulariy inflected forms are stored whereas low-frequency regularly inflected f o m are not.

1.3 AGRAMMATISM AND THE MENTAL LEXICON Studies of acquired language impainnents have given researchers invaluable insights into the understanding of the functioning of the language system. More specifically,

rnorphological deficits observed in agrammatic patients have been investigated in order to validate existing linguistic theories about the architecture of the lexicon. 1.3.1 Early Descriptions of Agramrnatism

Agramrnatism has been described traditionaily in the Iiterature as a syndrome characterized by a marked quantitative and qualitative reduction in Broca's aphasies' speech, manifested in the production of short utterances showing omissions and substitutions of grammatical morphemes, whether fiee or bound, and the ~ i m p ~ c a t i of on syntactic structures beyond the omission of closed-class items (Goodglass et ai., 1972). Originally described as one of the symptoms of the larger syndrome of Broca's aphasia, agrammatism has evolved over the numerous studies that have been conducted on language disorders into a syndrome with its own specific features.

The word

"agrammatism" was coined by Kussmaul in 1878 in a work on language disorders in which he distinguishes between two types of grammatical impairments in aphasia: one

involving word order and the other. word inflections. According to Tissot, Mounin. and Lhermitte (1973), Kussmaul suggested, even though not clearly. to use the term "agrarnrnatism" to refer to the word-order disturbance and the word "akataphasia"l0 for the loss of inflections, thus considering agrammatism as one aspect of the larger group of

grammatical disorden found in Broca's aphasia

Io

Frorn the Greek word "akatalictos" rneaning without ending.

Even though the early linguistic descriptions of agrammatism reported the occasionai absence of pronouns and of verb inflectional marken, they mostly dealt with the internai organization of sentences. Pitres (1898) observed an absence of Iinks in the

sentences produced either orally or in wrïting by his patients and attributed the phenornenon to a memory disorder associated with the construction of sentences. In one

of the f i t comprehensive descriptions of agrarnmatism published in Pick's (19 13) book

.

Die agrammatischen Sprachstorungenl 1 Kleist established a distinction between agrammatism and paragrammatism. a dichotomy which reflects the distinction between grammatical deficits as wanifested in expressive aphasia and receptive aphasia. Furthemore, while insisting on a possible disorganization of words within sentences as one of the feanires of agrammatism. Kleist (reported in Tissot et al., 1973)also noted the absence of punctuation in writing, the use of nouns in their nominative form. of verbs in the infinitive and participial forms, the omission of articles, prepositions and linking elements, as weii as the already described telegraphic style characterizhg agrammatism. Along the same lines, and following Kleist's definition of agrammatism, Najouanine ( 1956)claimed that impainnents of grammatical organization alone should not be taken as the landmarks of agrammatism, since they cm be found in other types of aphasias as weii, and therefore "true agrammatism", which he associated with Broca's aphasia in its severe forms of reduced phrase length, should cover other aspects such as the marked presence of uninflected nouns and of infinitive verbs. and the omission of gender. nuinber. and case markers.

I I Translation: Agrammatic language disotden.

1.3.2 Linguistic Approaches to Agrammatism Though several mearchers emphasized the need to use linguistic principles for the analysis of aphasic impairments (Goldstein. 1948; Jackson, 1958; Luria, 1947), the early accounts of agrammatism were Iargely descriptive, lacking theoretically-based interpretations. The first linguistic approach to agrarnmatism was offered by Roman Jakobson in his seminal work (1956) on the phonemic disturbances in the utterances of aphasic patients which he compared to the speech of children. relating both to more

general aspects of phonology. Even more important was the fundamental distinction that he made between disorders of "similarity" or selection. and disorders of "contiguity" or

combination in aphasic speech (Jakobson. 1964). According to Jakobson, in

agrammatism, "the most typical display of efferent proca's] aphasia", telegraphic speech and syntactic disturbances reflect impairment of the cootiguity component, Le., agrarnrnatic subjects have difficulty with the processing underlying the temporal concatenation of words into meankgfiil sequences. Subsequent research in neurolinguistics bas followed the lines of Jakobson's thinking, Le., "that [...] the pathology of language. far from being a random disturbance, obeys a set of rules; and that no rule underlying the regression of language can be elicited without the consistent use of linguistic techniques and methodology. The disorders of language display their own peculiar order and require a systematic linguistic cornparison with our nonnal verbal code1* (Jakobson, 196422). For example. in an early study on the availability of the English grammatical inflectional endings in agrammatism. Goodglass ( 1968) and Goodglass et al. ( 1972) found that Broca's aphasics showed a distinctive order of difficulty manifested in a higher

number of omissions of the non-syllabic than the syliabic allomorphs of the respective plural and past markers /s/ and /dl.

The first interpretation of agrammatism undertaken within the theoretical Framework

of Generative Transformational Grammar was proposed by Kean (1977) who attributed

the structure reduction of agrammatic speech to a phonological deficit despite the fact that it appean to be a syntactic disorder on the surface. She argued that the elements omitted

by agrammatic patients are actuaiiy the phonologicai clitics (determinen, auxiliaries,

monosyllabic prepositions, strong word-boundary affixes). What is spared in the speech of agrammatic patients are the phonological words which she defines as "the string of

segments, marked by boundaries, which function in the assignment of stress to a word [in English]." (Kean. 1977:22). For example, a word Like definitive where the affuc -ive plays a role in the assignment of stress will be retained whereas a word such as defrziteness where the a f f i -ness is neutral to stress assignment wiii not. A syntactic hypothesis conceming agrammatism has also been put forward by

several researchers. Several studies have suggested that agrammatism reflects an underlying syntactic deficit (Bemdt & Cararnazza, 1980; Saffran, Schwartz, & Marin, 1980; Zunf & Caramazza, 1976). Others have attributed the deficit to the agrammatics' inability to interpret thematic roles (Caplan & Hildebrandt, 1988) or to their failure to coindex trace (Grodzinsky, 1984, 1990).

1.3.3 Morphologieal Approaches to Agrammatism One of the strongest criticisms of Kean's (1977) phonological interpretation of agrammatism came from Lapointe (1983) who suggested instead lookùig at the deficit from a morphological point of view. Lapointe rejected as too strong Kean's contention that her description provides the best explanation for the distinction between impaired

versus retained elements. He reformulated her theory in terms of morphological levels of word formation, i.e., an approach based on a unified theory of morphology which stipulates that both inflectiond and derivational &xes are generated within a single morphologicai component which interacts with the lexicon. Within the mode1 outlined by Lapointe, the elements that are likely to be spared in agramrnatic speech are "those stemlevel items (of major categories) that are inserted into morphosyntactic stnictures during

lexical insertion" (Lapointe, 1983:24).

In an anempt to re-evduate the issues put forth in Kean (1977). Kehayia et al., (1 984) examined the relative retention of complex words. Working within a theoretical

framework where both derivational and inflectional processes are part of the morphological component, the authors noted a correlation between the patients' ability to repeat words and the distinction between listed and non-listed words on the one hand, and the different levels of word formation on the other.

They observed that while

idiosyncratic words were largely retained, complex words yielded difficulties cornmensurate with the level of word formation. They,therefore, suggested the existence of different levels of processing, and the possibility of a distinct impairment of the morphological component of grammar. 1.3.4 Cross-Linguistic Studies of Agrammatism

A large body of studies have k e n conducted, each investigating one or some of the

features charactenzing agrammatisrn, and various theoretical interpretations have been offered. However, most of the studies have k e n limited to English speakers. Bates,

Friederici, and Wulfeck (1987) have suggested that the investigation of linguistic impairments in aphasia should be broadened to strucnirally different languages in order to have

a better picture of the mechanisms underiying language production and

comprehension. The authon investigated morphological impairrnents in English-. Italian-, and German-speaking aphasic patients. A comparative analysis of the results showed that overail, the erron produced by the patients were not violations of their

grammar, which led the authors to conclude that brain darnage does not affect the major rules goveming the well-formedness of lexical items, but appears to affect either the patients' ability to process morphologically complex words, or their ability to access the lexicon. Cross-linguistic studies have shown the effect of the varying role of languagespecific features across a number of stxucturaily distinct Ianguages such as Hebrew (Grodzinsky, 1984), English, Dutch, German, Icelandic, Swedish, French. Italian, Polish, Serbo-Croatian, Hindi. Finnish, Hebrew. Chinese and, Japanese (the contributon to Mem & Obler, 1990). These snidies contributed to a better understanding of one of the Ratures of the syndrome, namely omissions. They showed that the

omission of bound grammatical morphemes is observed only when the resulting form is an existing word of the language. In a language like English, the omission of an afflix usually leads to a stem that is typicaiiy a phonologicaily and morphologicaiiy well-formed

word, a pattern comrnonly reported on in the literature on Engiish-speaking agrammatic subjects. However. in studies on languages like Italian (Miceli & Mazzucchi. 1990), Icelandic (Magnusdottir & Thrainsson, 1990). and Greek (Kehayia. 1990) where omissions of bound morphemes may lead to the production of bound stems'*,

agrammatic patients are reported to substitute affixes and never produce bound stems.

l 2 'Bound stem* is taken here to mean a stem bat does not comespond to a word.

1.4 CONCLUSION

Reflecting on the above literature review on the lexicon, one may outline the following points: 1)

Even though the representation and processing of morphologically complex words

have k e n the focus of a large number of theoreticai and empirical studies. they stiii remain debated issues in linguistic, psycholinguistic, as well as in neuropsycholinguisùc

Literature.

2)

Theories and models proposed in the various fields of investigation Vary in severai

aspects. However, one notices a common feature throughout. narnely. the emergence of morphology as an important part of the grammar of a language. Linguistic theorïes have acknowledged the existence of a separate morphological component of grammar (e.g., Anderson, 1982; Aronoff, 1976; Chomsky, 1970; Haile. 1973; Jackendoff, 1975; McCarthy, 1979). Psycholinguistic studies of the lexicon have provided evidence of the role of morphology in word recognition and processing (e.g., Kenderson, 1985;

MacKay, 1978; Mure11 & Morton, 1974; Stannen et al., 1979; Taft & Forster, 1975). Aphasic data have k e n shown to sornetimes reflect morphological impaiments (e.g., De Bleser & Bayer, 1988; Jarerna & Kehayia, 1992; Kehayia, 1990; Miceli & Caramazza,

1988). 3)

It has become increasingly apparent that the integration of data from different

languages on the one hand. and different fields of language research on the other hand, is needed to have a better understanding of lexical representation and processing. Linguists have been collecting evidence from data on a myriad of typologicdly different languages and, more recently. from different fields such as psycholinguistics and aphasiology.

Likewise. psycholinguists and neurolingiuists have been increasingly using linguistic

concepts in their investigations in order to account for data on language processing and

language pathology.

CHAPTER TWO THEORETICAL AND METHODOLOGKAL TOOLS

Before proceeding to the next sections concemed with the on-line tasks, the issues under investigation in this work, and the accompanying testing hypotheses, a sketch of the Arabic morphological system is presented here-particularly those aspects that are of

interest to us-both

in its traditional aspect and as formaiized by structuraiist and

generative linguists. 2.1 THE ARABIC MORPHOLOGICAL SYSTEM

Most accounts of Arabic morphology overtiy or implicitiy make the assurnption that it is essentiaily root-based. Words are taken to be formed on the basis of a lexical

discontinuous root between which are inserted sets of vowels.

2.1.1 The Early Arab GrammarianslApproach: The Binya Asliya Hypothesis

According to the approach taken by some of the thirteenth century Arab g r m a r i a n s l (e-g.,Ibn y a 9 B. in Bohas. 1984)- the morphology of Arabic is organized

around a set of basic nominal and verbal representations or binye 'esliye from which a i i other Arabic words are denved. A basic form or 7osl is described as consisting of a sequence of consonantal positions, or slots. connected to a string of vocaiic slots bearing the grammatical function of the word. The consonantai rwt is a two-sided element; it has

an acoustic aspect represented by the consonants, and a semantic one or r n a ~ n ewhich l~ is comrnon to dl words derived from it (Bohas, 1984). This description is illustrated in the foiiowing: maf na [Pt]

%sl-root k-t-b 'write'

binya kat sbs 'he m t e *

Figure 4 The Arabic word according to the binya 7asliya hypothesis (Bohas, 1984)

The basic nominal and verbal foms may have three. four or five consonants. Bohas (1993) reports ten basic nominal forms and three verbal foms for three-

consonantai rwts. The three threeconsonant basic verbal forms are given in (1).

l 3 We arc presenting here only one of the several views that prevailed in traditional grammars-the one that Bohas (1984) refers to as "Basri". It is also the view that prevails in most of the grammar books. l4 Bohas (1984) points out the existence in traditional morphology of Arabic of two meanings for the word mayna: one is used to express the rneaning of the mot throughout the paradigm, and the orher one to carry the sernanticesyntactic featurcs of the vocdic base.

CaCaCe

ka tsba

'he wrote*

CaClCa

raiima

'he learned'

CaCuCa

garufa

'khonoted'

Inflectional and denvational processes apply to these basic nominal and verbal word foms. For example. the plural &ut ub 'books' is derived from the singular kitssb 'book* which is in tum generated from the basic form ketebe 'he wrote'. Likewise. keatebe 'he

corresponded*is the result of an affixation process attaching the vowel a to the basic form k8tab8.

However, even though Arab grammarians and Arabic grammars have g e n e d y

assumed the bulk of Arabic words to be derived from a Iimited set of basic word forms, they have never been clear about the formation process of these basic forms from the disconiinuous root. According to Bohas (1993) who studied and analyzed sorne of the early Arab grammars, for the Basri traditional grarnmar schools the mnsdar, one of the classes of nominal forms, is the only one which denves directly from the root. Thus. all other derived words corne either from the binya ?as?iyeor from other derived forms. 2.1.2 The Structuralist Approach: The Root-and-pattern Hypothesis

The root-and-pattern morphological system was fmt formalized by Cantineau (1950) who claimed that most of the Semitic vocabulary is the result of an association between two morphological systems: the root system and the pattern system. Within this framework, every word is identified by its root and its pattern. Roots and patterns are found on bvo perpendicular planes. A word results from the association of a pattern with a root. As an example, the words naze1 'he came down', yenzil 'he cornes down', and msnzil 'house' are the results of

an association between the common root NZL and the

respective patterns CaCaC, yaCCiC and maCCiC corresponding to each word (Fig. 5).

Lists of patterns

Figure 5 The basic word in Arabic according to Cantineau (in Bohas, 1993:46)

2.1.3 The Cenerative Approach:

McCarthy's Autosegmental Approach

Recentiy, the root-and-paîtem description has been formalùed into what has corne CO be

known as The Prosodic Theory of Nonconcatenative Morphology developed by

McCarthy (1979, 1981) and, McCarthy and Prince (1990) who applied to Classical

Arabic the pnnciples of Autosegmental Phonology proposed by Goldsmith (1976) for tone in African languages. Goldsmith's autosegmental theory proposes that the underlying phonological representation is not a linear string of segments, but a multi-tiered representation consisting of segments and autosegments which are associated according to a set of universal conventions15.

.

B y apply ing this muitilinear approach to Semitic morphology McCarthy was able to fomalize the notion of a discontinuous morpheme arranged in a three-dimensional space. Crucial to McCarthy's theory is the idea that the consonant and the vowel sequences that make up the word in Arabic are represented on separate planes or tiers. Another basic tier of the representation of an Arabic word is the prosodic (CV)template. l5

For a more cornprthmsive description of autosegmental approaches. s e Goldsmith (1976. 1990).

In word formation, the phonemic segments that make up the consonant and vowel tiers have to be associated with the C and V slots of the CV template by means of a set of

principles (e.g.. the Weil-fonnedness Condition; Spreading; Erasure) as s h o w in Figure 6 for the word kitaab 'book'.

vocalic

ternplate

prosodic

/l\

template

consonantal

i i\

c

template

v

c

I

v

v

c

I

I

k

Figure 6 McCarthy's (1979) internai stmcture of a word in Classical Arabic

AU aff'ixal rnaterial is treated in the same fashion in accordance with the Morphemic Tier Hyphesis as shown in Figure 7 for the example nketab 'subscribe'. p

p

p

p

p

p

p

p

p

p

p

p

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

Figure 7 McCarthyls(1979) intemal smc&

of a prefïxed word in Classicai Arabic

-

-

-

-

-

The assumption that emerges from this theory is that the root the vocalic basis, the

CV template, and any other possible linear affm constitue distinct morphemes that make up the word in a Semitic language. The word nkatab 'subscribe', for instance, embodies the following morphemes: the discontinuous consonantal root lk-t -b/ which specifies the

semantic family of related lexical entries, the discontinuous set of vowels le-al which specifies its functional/grammatical categories. the prefix In-/, and the prosodic template

CCVCVC. 2.1.4 The Arabic Lexicon: McCarthy's Hypothesis

Following Halle (1973). McCarthy assumes for Arabic a lexicon "fuiiy specified with al1 forms, including inflections" (McCarthy, 1979:388). Given McCarthy's morphological theory for Arabic. a " f d y specified" lexicon in the fashion of Halle would be taken to include roots. morphemes, stems, words. and rnorphological rules.

However, only fully-inflected forms are subject to lexical insertion. McCarthy furiher postulates that morphological rules have the form of context-sensitive rewriting and redundancy rules. The other basic claim is that the lexical entry of a given form is structured as a rooted n-ary branching me. A representation of a lexical entry tree for /kt -b/ appean in Figure 8.

At the head of the lexical tree is the consonantal root which acts as a root node. Lexical relationships are expressed in ternis of domination and immediate domination. As illustrated in Figure 8, lk-t -b/ acts as the root node of the lexical entry. AU other nodes of the me are derived from lk- t -b/, though not imrnediately. For example, tekee t eb is derived from kee t eb which is itself derived from kat eb and Ik- t -hl, though most irnrnediately from ke t ab.

1

maktab 'O fice'

kattab

kast ab

'he caused to write'

'he corresponde&

ktatab kitaabst 'he wrote, 'act of writing'

I

copied'

takaaiab 'they kept up a correspondence' b

Figure 8 McCarthy's( 1979) structured lexical entry for Ciassicai Arabic

Another important feature of this lexical organization is that all idiosyncratic information relevant to a lexical entry is limited to itr, root node. Each time a nonrwt node

has to bear idiosyncratic information, it will have to appear with ali its daughters in a separate lexical entry where it wiU act as a r w t node. This wili be the case, for example, for the word kitaab 'book' in Figure 8 whose plural form is imgular. In order for kitaab to be specified for its plural irregularity, it has to be Listed twice. once within

the lexical

entry dorninated by the root l k - t -b/ as in Figure 8 and once as a root node dominating the plural form kutub and other daughters if any (Fig.9).

I

-

kit bab

[+irrepular plural]

I

'book'

I

kutub 'books'

I

Figure 9 Lexical representaaon-ofan imgular plural fom

The three theoretical approaches to Arabic morphology that have just been presented share the assumption that the discontinuous root constitutes the basic unit of representation. They differ, however, in the way morphological items are organized in the lexicon. The suucturalists propose that al1 morphologically related words cluster around the common root and that word formation involves an association between a word's root and its pattern. For the Arab grammarians and the generativists, not al1 words are linked directly to the root. For the former, only a timited set of specific basic forms is denved from the root. For the latter, Iexical relations are determined by a basic

CV template that a set of morphologicdy related words have in common; links to the root are made through this basic pattern.

In this study, Algerian Arabic (henceforth AA) refers to Arabic as used in Algeria both in its standard and vemacular form. Historically, the Arabic language has always been viewed and described as a confusing and complex language in that it exhibits two fonns-a

standard one and a vernacular one-which

show differences at al1 levels of the

discourse ranging from the phoneme to the syntactic structure. These düparities v a r from one country to another, and sometimes even from one region to another within the

same country. standard16 Arabic is universally used throughout the Arab world in written texts, newspapen, radio and television, and in goverment and educational institutions. The vemacular f o m on the other hand prevails in daily life and is essenfially spoken.

One of the fundamental feanires distinguishing the various Arabic diaiects is the reduction of vowel length. The long vowels of Classical Arabic /ad,/W., and /'J have l6

Ciassicoi and Litcrary are two other adjectives usal in the litcrature CO refer to the form of Arabic which is the closest to that of the Koran.

been shortened in some, if not al1 North African diaiecü. In some regions. short vowels have been reduced to a schwa /a/or have syncopated. Below are some examples of vowel reduction found across Arabic dialects:

Standard Arabic 1. kitsab

2,

katab

Arabic dialects kt eb

a) katab

'book'

'he wrote'

b) ktab 'she m e '

The relatively less reduced forms (2)2.a) and (2)3.a) are usuaily found in the Middle-Eastern countries, whereas more reduced forais such as (2)2.b) and (2)3.b) are fiequent in North Africa. Another feature characterizing Arabic diaiects is the absence of the classical case markers which have totally disappeared from almost every Arabic-speaking country. The three case markers-nominative,

genitive, and accusative-are

usuaily attached to the

noun in a s u f f i position. In (3) a) the noun wulad 'boy' carries the nominative marker -u whereas the word herf 'word' carries the accusative marker -an. These morphological

marken have disappeared fiom a i i Arabic didects as illustrated in (3) b) and (3) c).

(3)

Standard Arabic a)

k8 f abs 81- ~ 8 l 8 d - uh8r f -an he-wrote the-boy-noma-letter-a==

'the boy m t e a leaer'

Arabic dialects b)

kat ab alwalud harf

(usudy found in the Middle East)

C)

ktab alwald haff btheboy wrote a letter'

(usuaiiy found in North Africa)

These examples show that vowel reduction and case deletion do not result in semantic change nor do they affect the syntactic structure of a given sentence. There are several other features that may bring about changes at the semantic and syntactic levels. We are not, however, introducing funher details about these or any other aspects of Arabic dialects since they are not relevant to Our study. A €ist objective of this brief

comparative description of Standard Arabic and Arabic dialects is to give a background overview of the language investigated in the present work. A second objective is to point out that some of the differences that exist between Standard Arabic and Arabic dialects are not crucial enough to require separate theoretical approaches. Dialect differences are not significant enough to entail changes to the Arabic morphological system as a whole. Consequently, the general theoretical frarnework for Arabic applies as well to our data on Algerian Arabic. Thus. when we refer to Algenan Arabic, we mean the Arabic language used in Aigeria both in its standard and vemacular f o m . The linguistic items and structures

tested in this research corne from the Standard as weU as from the vemacular fonns. We tried-as

much as the language permits-to

control for linguistic and sociolinguistic

variations and to select stimuli that are uniform across language varieties and across subjects.

2.3 ON-LINE TASKS Generally, research on language disorders, and more specifically agrammatic aphasia, has largely relied on off-line tasks @ut see Kehayia, 1993; Kehayia & Jarema,

1994; Matthew & Kehayia. 1994; Prather, Zurif, Stem, & Rosen. 1992; Tyler, 1992;

Tyler et al.. 1995). To have as comprehensive and complete a picture as possible of the nature of lexical representations and processes. however, we incorporated both off-line

and on-line testing in our experirnental design. Off-line testing is conducted in the fmt study for the generai assessrnent of agrarnmatism in the performance of two Arabic-

speaking aphasic patients. The various tasks selected towards this end are presented in Chapter three. On-line tasks focus on the real-time analysis of the processes and representations involved in speech production and perception. In aphasia, they are used to provide the bais for a detailed examination of the ways in which these processes and representations break down. A number of on-line techniques have proved usefui tmls in studying lexical representations and their access. In Smdy two, we used an auditory lexical decision task associated with a morphological prunuig paradigm.

The lexicd decision task In studying variables that affect the speed of lexical access, psychologists have relied heavily upon lexical decision tasks. In such tasks. the subject is presented with a list of items (words and nonwords) and requested to decide as quickly as possible whether the letter string displayed on a screen or the utterance heard is a word or not (that is, is a nonword).

This research technique makes the crucial assumption that lexical access involves some matching of the features extracted from the stimulus with an intemal representation. When the manipulation of a variable is responsible for changes in response times, it is assumed that the variable affects the way in which the string of letters or the utterance is recognized.

Taft and Fonter ( 1975) used a lexical decision task to investigate whether prefixed words were analyzed into their constituents morphemes before lexical access. In the

experirnent, they had the subjects made lexical decisions on two types of nonwords: (i) real stems like juvenate denved from words like rejuvenate in which the prefix conuibuted to the meaning of the word (re- in rejuvenate means again) and. (ii) pseudo stems like gulote taken from words which begin with a pseudo prefix that does not

contribute meaning as in regulate. The subjects took longer to reject bound stems as nonwords. The authon interpreted the shorter reaction times for the fmt category of nonwords as an indication that real stems are perceived as king more word-like than pseudo stems, and may therefore be represented in the Iexicon. in cases like gulate. the item is rejected

as a nonword faster as there is no lexical representation for that item.

h a subsequent experiment whose main objective was to c o n f m the previous results. and consequently the decomposition hypothesisl7,the subjects' reaction times to juvenate and gulate were compared to reactions times to pseudo prefixed words like dejuvenate and degulate. As predicted by the experimenters, the real stem pseudo

prefixed nonwords (e.g., dejuvenate) took longer to reject than the pseudo stem nonwords (e.g., degulate). Taft and Forster explained these results as reflecting a rnorphological decomposition of words prior to lexical search. They proposed the following mode1 of recognition to account for their results:

j7

We are reprting herc information relevant to this study (sec Taft & Forster. 1975 for details).

Lener String

1

I . Is item divisible into prefix and

2. Search for stem in lexicon. Kas entry correspondhg to stem been located?

No

b

4. Search for whole word in Iexicon. Has envy corresponding to whole word been located?

5 . 1s item a

3. Can the prefix be addcd to fonn a

I

k e fonn?

No

Figure 10 Taft and ~ o r s c d smode1 (1975)

According to the modei, the search for juvenate involves five steps before the final

recognition as a real word or not: Step one:

Identification for the presence of prefut and stem in the itemjuvenate.

Step four:

Once this has k e n checked, a search begins for the entry juveme.

Step five:

A lexical representation of the stem is found. The lexical entry is then

examined to check whetherjuvemte is a fiee form. Step four:

Search will c o n t h e to make sure there is no lexical entry for another f o m

that could exist as a free forml8. Step seven: The item is rejected as a aonword. 18 ln English. this is a pssibility as sorne items may be fouod as free f o m (e.g.. vent) and as bound stems (e.g.. vent in prevenr).

However, classification of pseudo stems like gulnte requires four steps only (steps

one, two. four, and seven), which wiii then explain the shoner reaction times for these items. In the second experiment, the search for the pseudo prefixed real stem (e-g., dejuvenare) involves the same five steps as for juvenate with an additional check on

whether the prefïx de- is legitimate or not. Four steps oniy are required to reject the item deguiare, hence the shorter reaction times.

The Pnming Technique

Several experiments in psycholinguistics provide evidence that members of a morphological family are strongly related in the mental lexicon (e-g., Hendeaon, Wallis, & Knight, 1984; Kempley & Morton. 1982; S t m e r s et al., 1979). It has been shown

that the prior presentation of a word corresponding to one of the members of the morphological farnily facilitates the subsequent recognition of the other members. Stanners et al. (1979) based their experiments on the phenornenon of repetition or identity priming whereby the prior visual presentation of a given word results in a substantially reduced decision latency when presented the second time. Repetition (or identity) priming is commonly constmed as the result of the facilitating effect on the

access procedure from the repeated activation of the lexical representation of a target word.

In Stanners et al. (1979), this approach was adopted to investigate the relationships between English base verbs (e-g., le&; shake). and (i) regular and irregular verb forrns

( e g , lended; shook); (ii) third person singular inflections (e.g., lenàs, shakes); (iii) gerunds (e.g.. lending; shaking); (iv) adjective derivatives (e-g., seleetive from the base form select). For example, in the first experiment, the targets were base verbs whereas

the primes were divided into two groups: (i) base verbs which served as control primes,

and (ii) inflected and derived forms which served as cntical primes. The subjects were

requested to make a lexical decision. fint on the word LENDING (critical prime), then on the word U N D (critical target). then on the word POUR (control prime). and finally on

POUR (control target). The stimuli were presented visually. A lag of 10 items was used between the occurrence of a prime and a target. A comparative analysis of reaction times within condition (primed vs. non primed) and across the two conditions (repeated vs. inflected prime) shows a significant priming effect in both instances. Accessing POUR for the second time was significantly faster. Likewise, LWDING facilitated recognition of LEND. These results were taken as evidence that recognition is facilitated when a word has k e n preceded by another word that shares the same stem. Morphological priming effects have aiso been reported in auditory word recognition. Using an auditory morphological priming experimental design, Emmorey ( 1989)

conducted three expenments to investigate how morphologicai structure is

represented in the Iexicon. Primes and targets were presented auditorily. The stimuli consisted of English word pairs of the following type: (i) morphologically related words that were not semantically related consisting of distinct prefixes foliowed by a common stem (e-g., submit-permit) as evidenced by the existence of a common morphophonemic rule in the derivation process (e.g., submission-permission); (ii) phonologicaily related words (e-g., balloon-saloon); and (iii) words sharing either an inflectional suffix (e.g.,

joking-zyping), a derivational suffrx (e.g.. blackness-shormess), or fmal segments (e-g., tango-cargo). Words like submit used as primes were found to facilitate recognition of

morphologically related words like permit which were not semanticaiiy related but shared the saine stem (e.g., mit for submit and permit). However, no priming was found for phonologically related pairs or pain sharing the same suffix. The absence of priming between suffixed words led the author to argue in favor of a word-based processing

model. Words that are morphologically related through their stem prime each other because they are related in the lexicon.

Kempley and Morton (L982)found that prior presentation of words with regular inflection (e-g.. reflected) strongly facilitated auditory recognition in noise of morphologicaily related infiected words (e-g., reflecting), but the prïor presentation of an irregularly inflected word as in knelt-heeling did not. These results were explained by proposing a morpheme-based lexicon. Priming was thus interpreted as the lowering of recognition thresholds for lexical units due to previous activation of the stem. In chis model. morphologically related words prime each other because they share a single representation.

In contrast to Kempley and Morton's results, Fowler, Napps, and Feldman (1985) found that auditorily and visually presented irregularly inflected and derived words do

prime their morphological relatives (e.g., drive-driven; heal-health). Fowler et al. ( 1985) impute the differences between their results and Kempley and Morton's results to the long delay between prime and target in the expriment by Kempley and Morton. The study also showed that the results with auditory and visuai presentations were sirnilar.

The Primin~Ex~erimentaiDesien Used in the Present Research Despite conflicting results arnong some experiments, priming has proved an efficient diagnostic tool for analyzing the structure of the mental lexicon and the mechanisms of access. We are. therefore, using this technique in Our investigation of the lexical representation of Arabic singular and plural words. As in Kempley and Morton (1982), Emmorey (1989). and Fowler et al. (1985). the stimuli are presented auditorily.

Words include sound (suffixed) plurals (e.g.. lbasst 'dresses'), monosyllabic broken (infixed) plurals (e.g., wled 'boys'), bisyllabic broken (infixed) plurals (e.g., msemar

'nails*) as well as the singular fomis of the plural stimuli (e-g., 168s 'dress', wald 'boy' and masmar 'nail'). The target immediately follows the prime. a lag of 250 mec. separating the two items. However. we are adding a new feature to the way word pairs are presented. Contrary to most previous experimenü in which a list of items occurs only in the prime position and another List of words in the target position, our stimuli are organized in such a way that the critical items appear both as primes and as targets. For example, a singular form serves as a prime as in the pair wald 'boy' - wlad 'boys' and as a target as in wled 'boys'-wald 'boy'. A cornparison of the amount of priming between the two pairs will allow us to adopt a more appropriate interpretation of the results in terms of how relationships between morphologicaily related words are structured. A detailed description of the stimuli material. the testing procedure, the equipment and of the subjects is given in Study two.

2.4 ISSUES UNDER INVESTIGATION

In light of the above theoretical considerations, and given the typological particularities observed in Arabic, our main objective is to give a psycholinguistic account of lexical access and representation by addressing the following questions: 1) Docs the discontinuous consonantal root serve as a unit of access and representation? 2) Does the discontinuous consonantal root act as the head of a morphological family?

3) How are Arabic words which are morphologicaiiy related by the processes of

inffation,suffixation and prefxation organized in the mental lexicon? 4) What are the implications of these observations on a nonconcatenative language like

Arabic for the existing theories on the lexicon?

2.5 HYPOTHESES

Considering the theoreticaiIy-based distinction between regular and irregular plurals, the following hypotheses underiie Our research:

1) First, a differential processing of regular and irregular plural forms will be observed

during word recognition. 2) Second, this difference is a reflection of the feature specifications these forms carry and of their stnicniral organization in the lexicon.

2.6 CONCLUSION

The first part of this chapter has been concemed with the central constnicts of Arabic morphology. We have seen that the Arabic morphologicd system displays a word

fornation process that does not conform to the prototypical hear affmation models of the kind found in agglutinative languages. This distinguishing nonconcatenative feature of the language is given special focus in the issues we set out to investigate in the present

research. In its second half, this chapter describes one of the experimentai tools we will be using to investigate the organization of the lexicon. Lexical decision and pnmùig are two techniques frequentiy used in psycholinguistics, which have proveo to be efficient tools in word recognition and access.

CHAPTER THREE STUDY ONE

Agrammatic Aphasia in Arabic

Department of Linguistics and Translation, University of Montreal

and Théophile-Alajouanine Laboratocy, Research Centre of the Côtedes-Neiges Hospital, Montreai

Aphasblogy, in press.

Address correspondenceto: Z o h Mimouni, Research Centre, Côte-des-Neiges Hospital, 4565 Queen-Mary Road. Montréal (Québec), H3W 1W5,Canada Tel.(5 14) 340-3540 Fax (5 13) 340-3548

3.1 INTRODUCTION

Traditionally, agrammatism has been described in the literature as a syndrome characterized by a marked quantitative and qualitative reduction in the speech of Broca's aphasies, manifested in the production of short utterances showing omissions and

substitutions of grammatical morphemes, whether free or bound and the simplification of syntactic structures beyond the omission of closed-class items (Goodglass et al., 1972). Although several researchers emphasized the need to use linguistic principles for the

anaiysis of aphasic impairrnents (Goldstein. 1948; Jackson, 1958;Luria, 1947). the early

accounts of agrammatism were largely descriptive, lacking theoretically-based interpretations. The first linguistic approach to agrammatism was offered by Roman Jakobson in his seminal work (1956) on the phonemic disnubances in the utterances of

aphasic patients which he compared to the speech of children. relating both to more general aspects of phonology. Even more important was the fundamentai distinction that he made between disorders of "similarity", or selectioo, and disorders of "contiguity", or

combination, in aphasic speech (Jakobson. 1964). According to Jakobson, in

agrammatism "the most typicai display of efferent proca's] aphasia", telegraphic speech and syntactic disturbances reflect an impairment of the contiguity component, Le., agrammatic subjects have difficulty with the processing underlying the temporal concatenation of words into rneaningful sequences. Subsequent research in neurolinguistics has followed the lines of Jakobson's thinking. Le., "that [...] the pathology of language, far from k i n g a random disturbance, obeys a set of d e s ; and that no rule underlying the regression of language can be eiicited without the consistent use of linguistic techniques and methodology. The disorders of language display their own peculiar order and require a systematic Linguistic cornparison with our normal verbal code" (Jakobson, 196422). For example. in an early study on the availability of English inflectional morphology in agrammatism, Goodglass (1968),and Goodglass et al., (1972) found that Broca's aphasics showed a distinctive order of difficulty manifested in a higher number of omissions of the non-syilabic than the syilabic

allomorphs of the respective plural and past markea ls/ and /d/. A large body of studies have been conducted each investigating one or severai of the

features characterizhg agrammatism. and various theoretical interpretations have been offered in terms of a phonological (Kean, 1977). morphological (Lapointe, 1983), and syntactic (Caplan & Futter, 1986) breakdown. However, most of these studies have k e n conducted on English. Bates, Fnederici, and Wulfeck ( 1987) have suggested broadening the investigation of Linguistic impairments in aphasia to stsucturally different languages in order to achieve a better understanding of the mechanisrns underlying language production and comprehension.

Recently, linguistic investigations have shown the effect of the varying role of language-specific features across structurally distinct languages such as Russian and

Hebrew (Grodzinsky, L984). as well as English. Dutch, German, Icelandic, Swedish. French. Italian. Polish, Serbo-Croatian. Hindi. Finnish, Hebrew, Chinese. and Iapanese (the contributon to Menn & Obler, 1990). These investigators also contributed to a reformulation of one of the features of the syndrome, nameiy, omission of bound grammatical morphernes is observed oniy when the resulting forrn is an existing word of the language. In contrat to languages like English, where the prototypical process of word formation is linear affixation, Le., words are made up of sequences of one or more segments or morphemes that are concatenated together in a Linear order as shown in the word trans-fonn-at-ion, a Semitic language Wre Arabic, the language under study, is characterized by a word formation process expressed mostly through a change intemal to the word itself. According to the traditional treatments of Arabic, simple words are commoniy formed on the b a i s of a lexical root of three or four consonants, between which are inserted sets of vowels as shown in (1) a. In addition to this infixation process,

suffixation and prefmation are also typical operations in Arabic word formation (( 1) b). (1)

a.

kitaab'book'

b . ma-kt8ba 'iibrary'

+

kutub 'books'

mu-ktsbu- t 'Libraries'

Recently, this description has been formalized into what has corne to be known as The Prosodic Theory of Nonconcatenative Morphology developed by McCarthy ( 1979, 198L), and McCarthy and Prince (1990) who applied to Classical Arabic the principles of

Autosegmentai Phonology proposed by Goldsmith (1976) for tone in Afncan languages. Goldsmith's autosegmental theory proposes that the underlying phonologicai representation is not a linear string of segments, but a multi-tiered representation consisting of segments and autosegments which are associated according to a set of

univenal conventions. By applying this multilinear approach to Sernitic morphology,

McCarthy was able to fomalize the notion of a discontinuous morpheme arranged in a

the-dimensional space. Crucial to McCanhy's theory is the idea that the consonants and vowels of Semitic words belong on distinct. morphologically-defined tiers which are associated with the C (consonant) and V (vowel) slots of a CV template. Al1 affixai material is treated in the same fashion in accordance with the Morphemic Tier Hypothesis. The assumption that emerges from this theory is that the root, the vocaiic basis. and any other possible linear affix constitute distinct morphemes that make up the word in a Sernitic language. A form Iike k i l a a b 'book'. for instance, embodies three different morphemes: a discontinuous consonantai root /k-t-6/ which specifies the family of related lexical entries, a discontuiuous set of vowels /i-O/ which identifies the lexical entry and specifies its functional/grammaticalcategories, and a prosodic template CVCWC as shown in Figure 6.

vocalic

prosodic

template

c

template

consonantal

template

I

v

c

I\

v

v

c

\i/b

k

Figure 6 McCarthy's (1979) interna1 sauii~reof a word in Classicai Arabic

An example of representation for a prefixed word (e.g., nketeb 'subscribe') is given

in Figure 7.

Figure 7 McCarthy's (1979) interna1 structurë of a prefixecl woni in Classical Arabic

Given the a b v e theoretical claims, and the characterization of agrammatism in terms of omissions and substitutions of grammatical morphemes. an interesting issue is to look at the breakdown patterns of

morphemes in Arabic.

Our main objective in this paper is to investigate the ways in which agrammatism is manifested in Algenan Arabic, and how it compares with the existing descriptions of the syndrome. The aphasics' error patterns are contrasted across different modalities with those described in the literature. We will then attempt to speciQ the role of languagespecific features in the patterns observed and, finally, propose an interpretation of the subjects' performance in the iight of curent linguistic and psycholinguistic theories on the

iexicon.

3.2 METHODOLOGY

3.2.1 Language In this paper. the term Algerian Arabic (hereafier AA) refea to Arabic as used in Algena both in its standard and vernacular forms. AA encompasses Standard Arabic

which is used in the educational system, in the newspapen and in radio and television

programs, as well as a vemacular used in daily life. The latter form is characterized maidy by a reduction of vowel length which may lead in some cases to vowel deletion, by

the absence of case marken. and by a distinctive syntactic structure for negation. In ternis of word order, AA exhibits a relatively free movement of constituents. In order to obtain

comparable data between wntten and oral production, and in keeping with frequency requirements, we included in our stimuli only structures which are commonly used in the standard and vemacular fonns. More detaiied descriptions of the structural properties of AA will be given in the course of our analysis.

3.2.2 Subjects One fernale and two male right-handed Algerian Arabic-speaking patients were tested. They were diagnosed as Broca's aphasics on the Arabic version of the Ducarne de Ribaucourt aphasia battery (1976). None of the subjects had visual or hearing defects.

NB was 27 years old at the time of testing. She had suffered an ischemic stroke which resulted in aphasia and severe right hemiplegia. One year post-onset, she was able to waik but showed residuai weakness in her right arm.

RB was 35 years old and showed a mild right hemiplegia at the time of testing. He had suffered a cerebrovascular accident which caused lefi-hernisphere brain damage and

aphasia.

OH. 27 yean old at the time of testing, had suffered head trauma resulting in a unilateral left hemisphere front-parietai lesion.

Detailed neurological information was not available at the hospital where testing was conducted. Al1 three subjects were native speakers of Algenan Arabic. They were educated in Arabic and learned French as a second Ianguage in school. They were matched with normal control subjects of the same age, sex, and educational background.

Testing was conducted in Algeria. Table 1 summarizes background data on the three subjects.

CVA

CVA

Trauma

8

4

2

Sex

female

male

male

Pre-onset occupation

studeat

baker

accountant

Years of education in Arabic

11

7

Years of education in French*

9

4

11

nght

right

right

Etiology Years post-onset

Handedness

-

14

'French is taught as a second Ianguage simultaneously with Arabic.

Table I Background data for agrammatic subjects

3.2.3 Testing Material SampIes of oral speech were collected using the following tasks: (a) history of illness (b) the 'Cookie Theft' from the Boston Diagnostic Aphasia Examination (Goodglass & Kaplan. 1972). and (c) recitation of verses from the Koran. The subjects were aiso tested on parts of the Maghrebian version of the Bilingud Aphasia Test

(Paradis, 1991) which included description of a picture, repetition, reading aloud, dictation. copying. writing, and oral cornprehension of single words and of sentences. We aiso constructed a set of 60 simple and complex sentences which were added to the

ones included in the Bilinguai Aphasia Test battery in order to test a wider sample of structures in repetition. reading aloud and oral comprehension. The sentences used were of the SVO. VSO and OVS type. Arabic is a VSO language with an alternative SV0 order (Ouhalla. 1991) and displays a relatively free word order. Object-fint constructions require that a clitic pronoun agreeing in gender and number with the subject be affixed to

the verb.

Examples of sentences tested in repetition, reading aioud, and oral

comprehension are presented in Appendix 3.1.

3.3 RESULTS AND DISCUSSION Apart from a non-significant number of phonetic deviations (e-g., naftsh for mafteh

.

'key') the subjects' performance on single words was accurate, and does not require any further analysis. We will thus focus our discussion on the errors produced in the performance of sentences. Furthemore, the data on the three control subjects. who were matched for age. sex and educational background with the three patients, hcluded very few errors, which led us to conclude that the structures selected met testing criteria for

frequency, complexity, and picturabiiity. 3.3.1 Oral Production An overview of the subjects' spontaneous speech reveals features prototypical of

agrammatism, namely: simplification of syntax to grammaticdy disconnected utterances

which contain mainly content words; omissionsl9 of articles, prepositions and pronouns; omissions ancilor substitutions of inflectionai f i x e s . It must be pointed out that the samples coiiected in spontaneous speech are too small to d o w an appropnate quantitative and qualitative analysis of the patients' performance. The subjects were totally unable to produce connected free speech beyond one or two-word utterances. The spontaneous speech elicited included a maximum of twenty words (see sample in Appendix 3.2). The subjects' performance on the tasks of repetition, reading aloud. and recitation is

essentiaily characterized by omissions and substitutions of verb affixes, and omissions of articles. pronouns. and verbs. Tables II, ïII, and IV surnmarize the omission and substitution patterns found in repetition, reading aloud and recitation. We do not have any data on recitation for subject OH who could not perform the task. even though he indicated that he knew the verse of the Koran he was requested to recite. We wiil discuss in the following sections each of the elernents ornitted or substinited in light of the structure of AA.

In narrative discoune. we countcd as omissions al1 missing obligatory elernents leading to syntacticaily deviant structures.

SENTENCE REPErrTION (percent error) Omissions of articles

180

Omksions of subject pronouns

30

Omissions of relative pronouns

28

Omissions of verb-bound clitics

26

Omissions of negation markers

24

Verb omissions

126

Omissions of verb prefnes

Omissions of verb prefmes leading to tense substitution e.g.,

ya-drab 3ms- hit(pres) 'He hits'

+

drab hit(3rndpast) 'He hit'

Omissions of verb prefixes ieading to tense and gender substitutions e.g.,

ta-drab 3fs-hit(prcs) 'She hits'

+

drab hit(3 mdpast) 'He hit'

Stem substitutions leading to tense andor gender substitutions e-g.,

ya-drab 3ms-hit(pres) 'He hiis'

+

darb-at hit(past)-3fs 'She hit'

Table If Omissions and substitutions in sentence repetition

READING SENTENCES

(percent error)

Omissions of articles

186

Omissions of subject pronouns

30

Omissions of relative pronouns

28

Omissions of verb-bound clitics

28

Omissions of negation markers

24

Verb omissions

126

Omissions of verb prefixes Omissions of verb prefixes leading to tense substitution e-g.,

ya-drab 3ms-hit(prcs) 'He hits'

+

drab hit(3mdpast) 'He hit'

Omissions of verb prefixes leading to tense and gender substitutions e-g..

ta-drab 3fs-hit(pres) 'She hits'

+

drab hit(3mdpast) 'He hit'

Stem substitutions leading to tense ancilor gender substitutions e.g.,

ya-drab 3ms-hit(pres) 'He hi&'

darb-at hit(past)-3fs 'She tiit'

Omissions and substitutions in reading aloud

ALOUD

Omissions of articles

13

Omissions of relative pronouns

1 I

Omissions of verb-bound clitics

1

1

1

Verb omissions

2

7

Table IN Omissions and substitutions in recitation

Artides

In AA. there is only one article /al/ which is prefixed to nouns as well as to adjectives. It is used to express defhteness while indefiniteness is generally signaled by its absence. Another aspect of the Arabic article Lies in the phonologicai assimilation process it undergoes when foliowed by a coronal consonant as iiiustrated in (3). (2)

al+ban t

(3)

al+Sarns

+ +

albant

'the girl'

agams

'thesun'

A closer examination of the results in Tables II. III, and IV shows a high rate of

omissions of articles in the performance of subjects NB (88.8%. 72%. and 53.8%) and

RB (66.796.50%. and 6 1.5%)respectiveiy in repetition, reading aloud of sentences. and recitation. whereas in the third subject's performance articles are preserved, at l e s t in two of the three tasks since OH could not perform the third one. Two out of three subjects tend to drop the article. resulting in a slight semantic error. In the omissions observed. we

do not find a differential performance between coronal venus non coronai f o m of the

assirnilated article. indicating that the required assimilation or lack thereof does not seem to

play a role in the error patterns of Arabic-speaking aphasies. As shown in (4) and (3, article omissions lead to the production of a word in the indefdte form. (4)

al-bant

( 5 ) ag-Sams

+

'the giri'

'the sun'

bant

'agirl'

gams

'asun*

The fact that the article in AA does not bear number. gender and case markea may explain the high omission rate observed. This pattern is similar to the one found in English in which articles are not rnarked for gender, number and case unlike in French (Jarema & Friederici. 1994; Jarema & Kehayia, 1992; Rosenthal & Goldblum. 1989) or in Italian and German (Bates, Friederici, & Wuffeck, 1987), where articles, which are marked for gender and number, tend to be retained in agrammatic speech. Articles are also found to be omitted in structures such as cleft SV0 and cleft OV-

pros (see examples (S3) and (S), Appendix 3.1) where the def~tenessproperty camied by the micle is syntactically obligatory. In these specific cases, omissions yielded

syntacticaiïy deviant productions (6).

(6) *bant hije elli girl she who

ta-drab

wald

3fs-bit

boy

Free-Standine Pronouns

Ln

AA, subject and relative pronouns are free morphernes. Whereas relative

pronouns do not carry any morphological markers. subject pronouns are marked for gender and number. AA being a pro-drop language, the use of subject pronouns is restricted to emphatic and cleft structures. Examples (7), (8) and (9) illustrate the

omission patterns found in the subjects' performance.

(7) huws he 'He writes'

ya-ktab 3ms-write

+

ya-ktab

3ms-write 'Hewrites'

(8)

huw8

alli ya-ktab

he

who 3ms-write

+

'It is he who writes'

(9)

al-wald

huwa

the-boy

he

ït

huwa

ya-ktab

he

3ms-write

'He writes'

alli ya-ktab w ho 3ms-write

+

is the boy who writes'

al-wald ya-ktab theboy 3ms-write 'The boy writes'

As shown in (9). even the simultaneous omission of the two pronouns required in

clefi constructions does not yield asyntactic structures. Omissions of subject pronouns

may be atuibuted to the fact that they are not obligatory in the construction of a syntacticaliy weil-formed sentence, whereas omissions of relative pronouns may be due to

the syntactic complexity of cleft sentences. Verb Inflectio~

Verbs in AA are marked for gender, number, person and aspect. Since the Arabic verb system is highly complex (see McCarthy. 1979, for details), for the purpose of this study, we will only present the intemal structure of triconsonantal verbs in the perfective

and imperfective fomis20. Thus, if we take the verb k t a b 'wnte' (which is the masculine 3rd singular perfective used as a citation f o m siace no infinitive form exists in the language), we obtain the perfective f o m in ( IO). ktab- t21

mte-1s

'1 wrote'

ktab-ti

wrote-2fs

'You wrote'

ktab-t

wrote-2ms

'YOUwrote'

ktab

wrote(3ms)

'He wrotc'

katb-at

wrotc-3fs

'She wrote'

ktab-ns

m t e - 1p

'We wrote*

ktab-tu

wrote-2p

'You wrote'

katb-u

wrotc-3p

'They wrotc'

20 Testing materiais included exclusively perfective and imperfcftive triconsonantal verbs. 21 This form is homophonous with the one in (10) c.

The imperfective verb forms are given in (1 1). na-ktab

'1 write'

ta-katb-i

'You write*

ta-ktap

'YOUWnte'

ta-ktab

'S he writes'

ya-ktab

'He Wntes'

na-kath-u

'We write'

ta-katb-u

'You Wnte'

ya-ka tb-u

'They write'

What is of interest for the purpose of the present study is the distribution of the verb stem in the two aspectuai paradigrns. With the exception of f o m (IO)e and h and (1 1) b. f. g, and h, the verb stem k t a b appears throughout irrespective of aspect, gender,

number or penon. Note that ktab, which corresponds to the citation form, is a reai word in the Ianguage, whereas the other less common stem *kath is not. As shown in Tables II and III, verb morphology in the performance of Our three subjects is charactenzed by tense and gender substitution as well as by verb omission. Substitutions mostly affect prefixes (in repetition and reading sentences aloud), but also to a lesser extent, suffies (in spontaneous speech). However. it is not clear whether these errors should be taken as substitutions or omissions, since they are manifested at the surface level as omissions of bound morphemes while reflecting at the same time substitutions of gender and tense. Compare the foiiowing errors displayed in the patients' production in repetition (examples ( 12) and (14)), and narrative discourse (example ( 13)).

(12) a.

ya-drab 3ms-hit 'He hits*

b.

ta-drab 3fs-hit 'She hits*

+

+ -

drab hit(past) 'Hehit'

drab hit(past ) 'He hit'

-

22 This forxn is homophonous with the one in (1 1) d.

(13)

xadm-at

+

work-3fs 'She worked' ( 14)

ya-hsab

wotk(past) 'He worked'

+

3ms-count 'He counts'

hasb-at counted-3fk 'She counted'

We observe a tendency in the subjects' performance to either omit prefixes as in ( 12)

a and b resulting in a switch to the morphologically unmarked past tense form. or substitute entire verb forms, leading to gender and tense substitutions as in ( 14). As Grodzinsky (1984) pointed out for Hebrew, omissions in the AA system never lead to the production of bound stems. In the present data, each time the omission of either a prefm or a suffix would produce a stem which is not a word in the language, patients revert to a simple morphologicaiIy unmarked form or drop entire verb forms altogether. In (13) for instance, while ornitting the affix /-aV. the subjects never produce a nonword fonn like *xadm. even if it is a phonologicaiiy Iegal form in AA.

Our data dso show that the

omission rate of verb prefmes (84.496 and 86% respectively for NB and RB in sentence repetition: 64% and 88.8% respectively for NB and RB in reading sentences aloud) is much higher than the substitution rate (15.6% and 14.28% respectively for NB and RB in sentence repetition; 36% and 1 1.1 1% respectively for NB and RB in reading sentences aloud). Verb-bound cliticsz Verb morphology in AA exhibits a large variety of verb-bound clitics. Among them are the object pronouns which have been tested in our study. Within the hierarchical order of affixation of the verb-bound clitics, object pronouns must appear as a suffix, and may

23 We are using the term "clitic" here in accordance with cumnt Iiteranire on Arabic syntax (Ouhaila. 1991). We are not, however. making any theoretical c l a h regarding the statu of these items across languages.

be followed only by the negation suffix. Like many clitics in AA, they agree in gender

and number with the CO-referent.as exempMed in ( 15) and ( 16). (15) a. ya-ktab broya 3ms-write letter

'He writes a lener'

b. ya-ktab-ha 3 ms-writes-it(fs)

'He writes it'

(16) a. na-katbubrsye-t 1-write-p Ietter-fp 'We

write letters'

b. na-katbu-hum 1-write-pthem

'We writc hem'

As the summary Tables II and III indicate, verb-bound clitics are the most problematic items in the performance of al1 three subjects. who show a differential treatrnent of these affixes in the tasks of repetition (20%, 100%, and 100% respectively for NB. RB and OH)and reading aloud (lûûa,57.12, and lOq% respectively for NB. RB and OH) as opposed to the task of recitation where their performance is errorless

(Table IV). Instances of pronoun omissions are given in ( 17). (17) a. ya-drab-ha

+

3ms-hit-her(fs) 'He hits her'

b. ya-darb-u 3ms-hit-him(ms) 'He hits him'

ya-drab 3ms-hit

'He hits'

+

ya-drab 3ms-hit

'Hehits'

Our data parallel previous findings reported for Polish and French (Jarema. & Kadzielawa, 1990; Nespoulous. Dordain, Perron. larema, & Chazal, 1990). in which clitics have k e n descnbed as king impaired in the speech of agrammatic aphasies.

Nesation Negation in AA is rnost frequentiy expressed through the use of the discontinuous morpheme /me..3/which attaches to the verb after aii other affiixational operations have applied. Except for some specific cases which we wiil not develop hem. the two elements of the negative discontinuous morpheme are required to appear simultaneously attached

pre- and postverbally ( 18). (18)

al-bant me-ta-drab-5 the-girl Neg-3fs-hit-Neg 'The @ri does not hit the boy'

al-wald the-boy

In the production of the negative marker, the subjects seem to expenence the sarne difficulty as with the bound pronouns. manifested mostiy in the high rate of omission (an average of 85% in repetition and 44.4% in reading aloud) of the two particles of the

negation yielding a construction in the declarative f o m as in (19). or in the systematic omission of any verbs flanked with the negative discontinuous morphemes as in (20).

thus reverting to an asyntactic construction. (19) al-bant

ta-drab 3 fs-hic

the-girl 'The girl hits the boy'

(20) *al-ban t the-girl

al-wald the-boy

al- wald the-boy

We never observed productions containing only one of the two elements of the negative marker. The simultaneous occurrence of both parts is also obligatory in normal speech. Violations of this constra.int would Iead to morphologicdIy and syntactically illformed productions ((2 1) and (22)). (21)

*al-bant the-girl

me-ta-drab Neg-3fs-hit

al-wald the-boy

(22)

*al-ban t the-girl

ta-drab-i 3 fs-hic-Neg

al- w ald the-boy

3.3.2 Written Production Testing of wntten production included dictation and copying of single words, and a free writing task in which the subjects were requested to describe the cucumstances of

their illness. Our findings reveal that. with the exception of the copying task which yielded no enon. the subjects reproduced in their free writing samples of the substitution and omission patterns observed in the oral tasks.

In dictation, in which the limited number of stimuli precludes an in-depth error analysis, we obsented the following interesthg phenornenon. When writing to dictation, two of the patients (NB and OH) tended to produce the pointed target items, i.e.. marked

with symbols used to represent the three short vowels IV, lu/. and /a/ of Arabic, whereas

they did not produce pointed f o m when describing thek illness. These vowel symbols,

which do not nonnalIy appear in written Arabic are, however, introduced in the teaching materiais in the first years of the reading and writing learning process. They disappear gradually from educational texts once the child is able to read. One possible explanation of the use of pointed forms in the subjects' writing to dictation cou!d be inherent to the task

itself, i.e.. writing to dictation using pointed forms is associated with school where these vowel symbols are used as part of the learning process for reading and writing. An adequate interpretation wouid requke further investigation. 3.3.3 Sentence Cornprehension Studies on sentence comprehension disorders in agrammatism give varying picnires of the deficits and yield almost as many interpretations as there are studies. In the fmt detailed work on this topic. presented by Caramazza and Zunf (1976). comprehension is

found to be more severely impaired than production in agrammatic subjects. The authors aiso report that the patients' difficulties were confined to sentences that are semantically reversible with respect to thematic role assignment, and suggested that the patients develop heuristic strategies based on word order. This was the

k

t

experimentally-based

approach to comprehension deficits, and it gave rise to a surge of experirnental investigations on how patients interpret syntactic structures (e-g., Bemdt & Caramazza, 1980; Caplan & Hildebrandt, 1988; Grodzinsky, 1986; Linebarger, Schwartz, & Saffran, 1983; Schwartz, Saffran, & Marin, 1980). However, most of the studies have been carried out on patients who are native

speakers of English, a language which, because of its Limited morphological system, relies heavily on word order which, in turn, does not allow Bexibility of movement of the constituents of a sentence at the surface level. Our study attempted to investigate sentence comprehension in a language with, fust, a nonconcatenative morphology and, second, a fairly flexible word order. Oral comprehension was tested on a sentence-picture matching task. Examples of sentences tested in oral comprehension are presented in Appendix 3.1. The testing materials consisted of four pictures: the target (e.g., the girl hugs the rnother), a syntactic distractor (role reversal: e g , the mother hugs the girl) and two semantic distractors (e.g., the girl hugs the girl and the woman bugs the wuman). Words used were highly familiar and picturable (e-g., boy, man, old man, womm, girl, to hug. to follow, to surprise. to

hold, etc.). Sentence length varied between two and five words. Sentences were semirandomly ordered. Table V summarizes the results obtained in oral comprehension.

PERCENT ERROR BY STRUCTURE NB

RB

OH

Clefi S V 0

OV-pros Cleft OV-pros

VSO Negative SV0 Negative OV-pros

Table V Oral comprehension

As can be seen in Table V, OH'S oral comprehension of ail the sûucnires tested is

generally weii preserved. For NB and RB whose comprehension deficit is more marked.

the difficulties experienced seem to be commensurate with the complexity of the structures. Overaii, problerns were encountered in object-fmt, clefi object, and negative

abject-fmt sentences. These structures involve a non-canonical word order and complex verb morphology, which may result in additional processing load. Let us consider

examples (S 1) and (S4) taken from Appendix 3.1. (SI)

al-bant t a-drab the-girl 3fs-hit 'The giri hits the boy'

al- wald ihe-boy

(S4)

ta-darb-u the-boy 3fs-hit-him 'the boy is hit by the girl'

al-bar~t~~ the-girl

al-wald

OVS smcture has a passive rneaning in Arabic. The passive consmiction. which exists in the Ianguage, is highly infrequent and was not included in our stimuli.

24 ï h e

In a canonical S V 0 active sentence such as (S l), the prefm /ta-/ of the verb agrees with the lexical subject al-bant in pre-verbd position. In contrast, in an object-first sentence such as (S4), the same prefix now agrees with the postverbal lexical subject while the verb clitic I-ul agrees with the pre-verbal object al-wald, thus creating a crossover in the linear ordering of agreement. Subjects showed problems with (S4) type. but not with (S 1) type sentences.

The data indicate chat the patients rely both on morphological markers and on canonical word order. When the verb does not exhibit a bound chtic as in the active, cleft

subject and verb-initial active constructions, the percentage of errors is very low if not

equai to zero, whereas in the more morphologicaliy complex structures where the verb prefm does not agree with the noun in subject position and canonical order is absent. the

error rate increases. Furthemore, we notice that syntactic load adds to the processing difficulties encountered by two of the patients. This is refiected in the higher error rate observed in the cleft object and the negative object-fmt as opposed to the other object-fmt sentences. Moreover, the fact that the subjects manifest problems with the canonically ordered SVOpro and negative SV0 structures indicates that canonical word order alone does not facilitate performance (see Caplan & Hildebrandt, 1988). Cruciaily. canonical VSO and SV0 constructions are not problematic unless they exhibit cornplex morphology.

These results point to the role of language-specific features in the performance of aphasic subjects since, at l e s t for AA, morphology and word order play a cruciai role in the parsing capacities of agrammatic patients. Simiiar results were found for subject RB in a cross-linguistic study by Jarema et ai.. (1991) investigating the capacity of a g r m a t i c subjects to interpret canonical. inverted and cleft constructions of active and passive sentences.

3.4 GENERAL DISCUSSION

The most salient and discussed features of agrammatism are the omission and

substitution of grammatical morphemes, whether bound or free, in speech production. Our findings are consistent with this observation since our patients do omit free standing and bound grammatical markers. Cross-linguistic studies have shown that language-specific features determine the pattern of omissions and substitutions found in brain-darnaged subjects. In a language like English. stripping a word of its grammatical affïx leads to another word that

corresponds to the stem, a pattern commody reported in the Literature on English-speaking agrammatic subjects. However, in studies on laquages like Italian (Miceli & Mazzucchi, 1990). and Icelandic (Magrlusdottir & Thriiinsson, 1990) where generally stems are not morphologicaiiy weii-hned words, a&rammaticpatients are reported to substitute affixes and never produce bare stems, even when these forms exist in the paradigm as is the case with the imperative singular form of verbs in Italian and Icelandic. Likewise. in AA where verb stems are not always weU-formed words as shown in katb-at (wrote-3fs)

'she wrote' in which the stem kat& is not a word. bound stems are never produced as a result of omission. In the performance of our patients, the erroneous forms are not randomly selected, but rather are singled out arnong the possible options towards the production of a morphologically related word. For instance, in the omission pattem observed in the verb-bound clitics (see (17) b.), the f o m s resorted to are always morphologically related words totally or partialiy saipped of their bound elements. In the error patterns displayed, the same phenornenon is observed throughout, which leads us to conclude that omissions or substitutions do not operate randody in the speech of our

patients. In this respect. Our conclusion is similar to the one reached by many authors. e.g.. Grodzinsky ( 1984); Kehayia ( 1990): Kehayia. Jarerna. & Kadzielawa, 1990:

Magniisd6tti.r & Thrainsson ( 1990); Miceli & Mauucchi ( 1990). Another important finding of Our study is the frequent omission of linear suffixes

and prefixes as opposed to their substitution. a pattern that was not reported on by Gmdzinsky (1984) for Hebrew-speaking agrammatic subjects. Our data show that the

omission rate is much higher than the substitution rate. With the exception of a small number of single verb affix substitutions (e.g., ta-drab (3fs-hit) 'She hits' + ya-drab (3m-hit)

'He hits') in spontaneous speech samples, substitutions mostly involve entire

verb forrns (e.g., te -dra b (3fs-hit-imperfective) 'She hits'

+ darb-a t (kt-3fs-perfective) 'She

hit' ). Omissions which involve prefixes of imperfective verbs result in the production of the masculine 3rd singular perfective, a form which is morphologicaily Iess marked. The

question that arises here is whether verb prefix omissions which lead to the production of a form from a different aspectual paradigm than the one intended are the result of an

omission or a substitution. For English, the question whether agrammatic subjects are omiaing morphological

markers or sirnply substitutïng a complex form with a 0-marked form remains open. For AA. ihe answer may be found in the morphological structure of the language. which

brings us to our next point, namely, the intemal structure of words. If we assume with McCarthy (1979) that simple words are made up of three discontinuous morphemes. then we are left with the option that for simple words only substitutions may occur in the

speech of agrammatics. Let us retum to the intemal structure of words in Arabic as

shown in Figure 6 and see what happens if omissions take place.

consonantal

template

k

\!yb Figure 6 McCarthy's (1979) internai structure of a word in Classical Arabic

In the above exarnple, omission of the discontinuous vocalic b a i s li-a/ may well lead to an unpronounceable string of segments / k - t - b p . What we observe in the present study is that patients do not rernain speechless; they do utter words encouched in reduced and disconnected sequences.

In fact, patients substitute a bound stem with a

morphologically related free stem26. an option that is allowed by the structure of the Ianguage. Again, if we look at Figure 6, the discontinuous vocalic b a i s may be substituted with another one represented by the vowels lu-u/ yielding the f o m kutub 'books' (see example (1) a.). Such substitutions rnay be accounted for within the theoreticai model of the lexicon adopted by McCarthy (1979).

25 Depending on the phonological anaiysis adopted, o f coune Ik-t-bf could always be realized. e.g.. [ktab], given a late phonetic mle of epenthesis. 26 *Free stem* is w d to refer to a stem that corresponds to a word. In English. for instance. in the word looking, look is a free stem.

1

k-t-b

kit886 'a book'

... etc..

Figure 11 McCarthy's ( 1979) structureci lexical enûy for Classical Arabic (short version)

McCarthy picnues the Arabic lexicon as a set of Lexical mots and words. FoiIowing Haiie (1973). he assumes that the lexicon contains morphemes. words, stems. and a set of morphological rules. Furthemore. the basic unit over which these rules operate is fmt the discontinuous consonantal root. The lexical entry for Arabic is suuctured as a branching tree dominated by the consonantal root h m which al1 other forms are derived.

This conception of word formation was a challenge to existing Linguistic theories on the lexicon. Even though McCarthy adopts Halle's contention that word formation rules operate over morphemes as well as over words, he introduces a novel conception of a morpheme-based lexicon when he assumes the existence in the lexicon of a discontinuous

morpheme, Like the mot / k t -61. According to McCarthy's mode1 of the lexicon, and assuming his word formation

theory, which proposes that an Arabic word is the result of a simultaneous mapping of two discontinuous morphemes and any possible a f i x ont0 a prosodic template, only two steps are required in order to access a word in AA. One has to first access the discontinuous root, then the stem or the word. This is illustrated in Figure 12.

1

SteP 1

d INPUT

consonantai mot

k-t -b

Morphological Rules

word

ya-ktab 'he writes'

Figure 12 Access procedure for the word ya-ktab 'he writes'

The questions that arise here is why an affued fom like ye-ktab should be so

difficult to access for the patients tested in Our study, and why would it preferentially result in the production of the free stem ktab 'he wrote' as observed in our results? As a preliminary explmation. one couid assume that morphemes are stored according to three categories: the consonantal roots, the vocaüc bases, and the suffixes and prefmes. W e could then hypothesize that it is access to this Iast category of morphemes that is impaired in Our patients, whereas access to the other two categories is preserved. Therefore,

according to Figure 12. the production of the free stem ktab can be accounted for if we argue that accessing the suffix lya-/ is costly since it requires extra processing. Patients

are thus producing tmncated lexical items, i.e. items stripped of their prefures ancilor suffixes. Imponantiy, the resulting foms are existing words in the language. The patients are accessing a form whicb is morphologically less marked while keeping within the same semantic field and the sarne grammatical category.

However, our results also show, in the items which underwent omission of suffixes and prefixes, a significant shift towards a stem which is not part of the target. TOexplain

ihis phenomenon. let us look at an another example:

ka tb-a t 'she wroteb

Figure 13 Access procedure for the word katbat 'she wrote'

Difficulty of access to the suffix k a t / here would resuit in the production of the stem kath, which is actudiy a phonologicaily possible nonword. As already stated above. our

data do not show. in any of the tasks conducted. the production of bound stems. aithough a substantial number of the structures tested included lexicai items which exhibited a

bound stem (45%) in their intemal structure. One may argue that these bound stems exhibit phonotactically illegal final consonant clusters which prevent the patients from producing them. An andysis of these bound stems (see Appendix T) reveais that the stem-

fmd consonant clusten appearing in our stimuli are aU possible occurrences in AA. These considerations exclude an alternative explmation in ternis of language-specific phonological constraints applying here and thus preventing the patients from producing such stems. What is of interest in our findings is that the form that the patients systematically

resort to when expenencing difficulties with affues is the masculine 3rd singular perfective (see (10) d.), be this fonn part of the target or not. This gives rise to two important questions: 1) why would the patients avoid producing phonologically wellformed bound stems which are presumed to be listed in the mode1 of the lexicon proposed

by McCarthy?. and 2) why would they resort to a specific form and not pick randomiy

among the other rmt-related forms that are also available in the same lexical entry? Grodzinsky (1984:105) has suggested on the basis of data on agrarnmatic Hebrew-

and Russian-speakers. that mis-selections of inflected forms are made randomly "in cases where the inflectional paradigrn of an item does not include 0-i.e., nonword-".

the 'bare' stem is a

an assumption which does not hold for the present data. unless we dismiss

the restricting notion of "paradigrn" and adopt a wider notion of "family" (Scgui & Zubizarreta, 1985) where verb f o m are not organized on the basis of tense and aspect. Centml to Segui and Zubizarreta's mode1 are the concepts of "family" and of "head". As suggested by the authors, what defines a family is the common root that al1 memben of

the family share. This common root, whether free or bound. may constitute the head via which access is possible. An approach incorporating the concept of morphological farnily was adopted for AA

by Mimouni. Kehayia and Jarema ( 1992). The study. which tested singular and plural forms of nouns in AA through an auditory priming experiment, was conducted on 24 non brain-damaged subjects. The results showed that in AA nouns are organized into families where the head is the singular form for irregular plurds and the consonantai root for regular plurals. The suffixed plural fomi is shown not to be accessed directly. It was suggested that in order to access a plural word Wce Ibes-st 'dress-es', the subjects access fmt the root 1-b-S. and then the singular Ibes 'dress' before accessing the target. An organization of the mental lexicon into families is also most pertinent to the present

data. We propose that the masculine 3rd singular perfective, the form Our subjects most frequently resort to when in diffîculty, represents the head of a verb family in AA, and access to any member of the family is achieved via this fom (Fig. 14). This interpretation would aiso explain why the lowest error rate in the comprehension task concems

structures containing these verb forms or forms which are the closest to them in the

step 1

-

WüT

word (head of the farniiy)

ktab

+ 3 ms perfective 'he wmte'

Morphological Rules

step 2

OUTPUT

wad

ya-ktab 'he wtites'

Figure 14 Access procedure for AA verbs

Even though McCarthy's mode1 of lexical organization includes bound stems togedier with words and roots in the list of items that make up the lexicon. it does not explicitly speciQ whether bound stems can serve as inputs to morphological rules or not ~McCaahydoes, however. state that inputs need not "be words in the (lexicalist) syntactic

sense of this tem." (McCarthy. 1979:390). The present data show that free stems and bound stems do not play the same role in the lexicon. While the results do not provide evidence that bound stems are listed. they do show. at least for verbs in AA. that these items do not serve as inputs to morphological d e s . Rather, it is the masculine 3rd

singular perfective fkee stem which serves as input for al1 verb forms in a lexical entry.

Acknowledgments This study was supported by an FCAR scholarship to the principai author. We are gratefui to Rofessor Abdelaziz Bendib and Leila Ait-Mesbah, speech thcrapist, of the Department o f Neuro-Traumotology, Mustapha Hospital. Algiers. for their valuable assistance. and to the patients who panicipated in this study. We thank Yves-CharlesMorin and Eva Kehayia for fruitful discussions, and the reviewers for their very hirlpful comments. WCwould a h like to thank Paule Samson for her technical assistance.

Symbols used: s: p: f: m:

Singular Plural Ferninine Masculine

pres: Resent Neg: Negaiion h k indefinite

Appendix 3.1 Examples of structures used in repetition. reading aloud and oral comprehension

ai-waid

ai-bant ta-drab the-girl 3fs-hit 'the gir1 hits the boy'

the-boy

hije

ta-darb-u she 3fs-hic-him .she hits him'

Cleft S V 0 hija elli ta-dfab the-girl she who 3fs-hit 'It Is the girl who hits the boy'

6 3 ) al-bant

al-wald t a-darb-u the-boy 3fs-hit-hirn 'the boy is hic by the girl'

al-wald the-boy

al-bant the-girl

Cleft OV-pros (SS) al-wald

huwa

elli

ta-darb-u the-boy he who 3fs-hit-him 'it is the boy who is hit by the girl'

al-bant the-girl

VSO ( S 6 ) ta-drab

al-bant 3fs-hit the-girl 'the girl hits the boy'

al- wald the-boy

Negative S V 0 (S?) al-bant

me-ta-drab-S

the-gir1 Neg-3fs-hit-Neg 'the girl doesn't hit the boy'

al-wald the-boy

Negative OV-pros (S8) al-wald

me-ta-darb-u-9 the-boy Neg-3fshit-him-Neg 'the boy is not hit by the girl'

al-bant the-girl

Sample of spontaneous speech for subject NB (Picnire description from Paradis. 199 1 )

re2al

zswaS

na ?

ZW-

Tah

man@d)

bird(Ind)

'a man*

'a bird'

went up(3dpast) 'he/it went up'

bird-mp 'birds'

fet 1(3ms/past) 'helit fe1I'

Teh

razal

fell(3mdpast) 'heht fell*

man(Ind) 'a man'

nirsh injured(3rns/past) 'he got injured'

Tah rZa 1 felI(3rns/past) leg(Ind) 'hdt fell' 'a leg*

Tïyara sbiTsr pIane(hd) hospitai(1nd) 'a plane' 'a hospital*

railu Ieg-his(ms) 'füs leg'

CHAPTER 4 STUDY TWO

The Mental Representation of Singular and Plural Nouns in Algerian Arabic as Revealed through Auditory Priming in Agrammatic Aphasic Patients

Zohra Mimouni 1.

2. 3. 4.

* ,Eva Kehayia

3 9

4

& Gonia Jarema 1.2

Département de linguistique et traduction,Université de Montréal, Monaéai Laboratoire Théophile-Najouanine, Centre de recherche, Centre hospitaiier Côtedes-Neiges, M o n W Department of Linguistics, Mffiill University, Montréal Research Department, Iewish Retiabilitation Hospital, Montréai

Bruin and Language, submitted

Address correspondence to:

Zohra Mirnouni. Centre de recherche, Centre hospitaiier Côte-des-Neiges, 4565 chemin Queen-Mary. Montréal (Qc). H3W 1WS,Canada

Tel.(514) 340-3540 Fax (5 14) 340-3548

ABSTRACT

Adopting as a point of departure the mode1 of lexical organization developed by McCarthy ( 1979) for Arabic. we address. in the present paper. the issues

of lexical representation. morphological relatedness, and modes of access in

Algenan Arabic in an auditory rnorphological priming experiment. More specifically, we investigate the process of word recognition of singular and plural nouns in the performance of 26 subjects. 24 of whom are non braindamaged subjects and two of whom are agrammatic aphasies. Plurais in Arabic involve either suffixation as in the sound plural (e.g.. l b e s 'dress'llbes~'dresses' or stem-interna1 changes as in the broken plurals (e.g., kursl 'chair'lkras& 'chairs').

Our findings reveal a differential

processing of the two f o m indicating whole word access for broken plurais and decomposition into word and suffx for suffixed plurals. Furthemore.

for Algerian Arabic, the evidence suggest an architecture of the lexicon reflecting a family-like organization taking into account language-specific features.

This study was supported by an FCAR scholarship to the principal author. In preparing this experiment, 1 benefited from helpful discussions with Yves-CharlesMorin, Glyne Piggott and DanieI Bub. 1am grateful to Professor Abdelaziz Bendib and Leila Ait-Mesbah, speech therapist, of the Department of NeuroTraumotology, Mustapha Hospital. Algies. for their valuable assistance, and to the patients who participated in this study. 1thank Mathew Decter who helped in putting together the experirnentd design; Fmcine Giroux and Paule Samson for theu technical assistance.

4.1 INTRODUCTION

A theory of language must contain as a cenaal component a iïst of the elements that

are used in the formulation of meaningful sentences. In linguistic ternis, this component is referred to as the lexicon, Le., our store of information on lexical items which are accessed and manipulated by grammar. In psycholinguistic te-,

this is Our mental

lexicon, i.e., the actual component of human cognition in which an individual's knowledge about lexical items is stored. This component plays a crucial role in the mapping of propositions onto Linguistic utterances. Many authors have claimed that the mental lexicon comprises lists of lexical items. analogous to the lists of words in a dictionary. Research in theoretical linguistics and psychohguistics over the last three decades has demonstrated that the system of lexical storage and retrieval is extremely cornplex. Nevertheless, despite the rich body of literature in the field and the various models proposed, a number of fundamental issues stiU remain unresolved. These issues concem: 1) the identification of the units that serve as eatries in the mental lexicon, 2) the

intemal organization of the mental lexicon and 3) the different modes of access of mental

representations during normal processing. With respect to the first issue, theoreticai hguistic and psycholinguistic accounts

have advanced in parailel ways in proposing that the fundamental unit of the (mental) lexicon is either the morpheme (Halle. 1973; Taft & Forster, 1975) or the word (Aronoff,

1976; Buttenvorth. 1983). There are also those that put forward compromise solutions where both morphemes and words are lexicdy represented (Anderson. 1992; Cararnazza,

Laudanna, & Romani, 1988; DeU, 1986). Regarding the issue of the internai organization of the lexicon in theoretical hguistics, Jackendoff (1975) provides us with a morphological approach explicitiy based on morphological relations expressed in ternis of lexical redundancy d e s which link two morphologically related words. In his model, ail related words of a language have separate, but linked, lexical entries. In the psycholinguistic literature, it has k e n shown that there are phonetic, semantic. and morphological associations between mental representations. Specifically with respect to morphological associations, it is suggested that every word has its own entry (with or without intemal specifications) but all entries for morphologically related complex words are linked either to a base form or nucleus

(satellite-entries hypothesis, Lukatela et al., 1980). or to each other, the common root acting as the head of a given morphological family (morphological-family hypothesis. Segui & Zubizaneta, 1985). This concept of Links between morphologicaliy related items

is also found in the connectionist models of the lexicon (Dell, 1986).

Finally. with respect to the issue of the mode of access of mental representations, existing psycholinguistic proposais reflect to a large extent the mode of representation of lexical items. According to the authors who support the full listing hypothesis, each word

has its own lexical entry and c m be accessed in a direct manner as a whole (Bunenvorth.

1983; Manelis & Tharp. 1977). Under this view. parsing is not needed in normal word

recognition but is resorted to as a fall-back procedure in the case of novel words or previously unheard forms. In contrast, according to the decomposition hypothesis initiaily advocated by Taft and Forster (1975) and refined in Taft (1979), it is claimed that complex words are accessed via their stems. This implies that complex items are decomposed into their constituent morphemes prior to access. Finally. given that both of the above proposais have been vaiidated by various concurrent studies, hybrid models have been put forward. For example, Laudanna and

Burani (1985) and Del1 (1986) have proposed that two parailel routes are available for access of the lexicon: A whole-word access procedure used for known words, and a morpheme address procedure used for novel forms. A large number of the psycholinguistic studies addressing the issues of lexical

representation, organization, and access have initiaiiy focused on English. Recently, there has been a surge of research on other languages (e.g., Finnish, French, Gerrnan, Greek,

Italian. Polish, Serbo-Croatian). A common characteristic across ali of these languages is their concatenative structure. i.e., words are made up of sequences of one or severai

morphemes that are put together in a linear order. In an effort to validate the existing models of lexical access and representation, psycholinguistic research is extending its scope to typologically different languages (e.g., Hebrew: Feldman & Bentin. 1994; Arabic: Mirnouni, Kehayia, Br Jarema, 1992; Chinese: Zhang & Peng, 1992). Research in aphasia has also provided us with invaluable insights in the comprehension of the functional and architectural aspects of the lexicon. Most of the neurolinguistic studies have been carried out using off-line experiments. Recently,

however, a nurnber of researchen (Kehayia, 1993; Kehayia & Jarema, 1994; Matthew, & Kehayia, 1994; Prather. Zurif, Stem. & Rosen, 1992; Tyler, 1992; Tyler, Ostrin, Cooke, & Moss, 1995) have resorted to a series of on-fine experiments, which, together with

other off-line tasks, aim at providing a more comprehensive picture of the deficits observed in aphasic patients.

4.2 SCOPE OF THE STUDY

In the present study we address the issues of lexical representation, morphological relatedness, and modes of access in Algerian Arabic (henceforth AA) in an auditory

morphological pnming experiment. More specifically,we investigate the process of word recognition of singular and plural nouns in the performance of 26 native speakers of Algerian Arabic, 24 of whom are non brain-damaged subjects and two of whom are agrammatic aphasies. 4.2.1 Language Background and Theoretical Frarnework

Arabic, a language with a nonconcatenative intemal structure, exhibits a pattern of

word formation mostiy expressed through a change in the intemal structure of the word itself.

k8t8b

'he wrote*

kitasb

'a book*

kutub

'books*

k88 tab

'he componded'

rnaktab

'office'

ki tssba

'act of writing*

etc.. .

As shown in ( 1), morphologically related words in Arabic share a discontinuous

lexical root of three consonants (e-g.. lk-t -b/) between which sets of vowels are inserted.

In addition to this infixation process. suffixation and prefzation are also typical operations in Arabic word formation (2). (2)

b.

me-ktebe 'library'

[n recent

+

ma-k tsbe- t 'Iibdes'

linguistic accounts of Arabic (McCarthy, 1979, 1983). the word kiteeb

'book'. for instance. is assumed to consist of three difierent morphemes: a discontinuous consonantai root l k - t -b/ which specifies the family of related lexical items, a discontinuous set of vowels If-id which identifies the lexical item and specifies its functionai/grammaticai categones. and a prosodic template CVCWC. The vocalic and consonantal templates are associated with the consonant and vowei slots of the prosodic template by the linking conventions of autosegmental theory (Goldsmith, 1976). According to the Prosodic Theory of Nonconcatenative Morphology developed by McCarthy (1979) for Arabic. suffixes and prefixes are treated as distinct morphemes which make up the word.

McCarthy (1979) pictures the Arabic lexicon as a set of lexical roots. morphemes, stems, words, and morphological rules. Furthemore. the lexical entry for Arabic is stnictured as a branching tree dominated by the consonantal root from which al1 other

forms are derived (Fig. 11).

...etc...

kit asb

k8t8b 'he wrote'

'a book'

Figure 11 McCarthy's (1979) stnictured Iexical entry for Classical Arabic (short version)

In iight of the typological differences observed in Arabic, a psycholinguistic account of lexical access and representation would address the following questions: 1) do discontinuous morphemes such as the consonantai root and the vocalic basis which make up the word in Arabic serve as units of access and representation? 2) How are morphologicaily related words organized in the mental Lexicon? 3) What would be the implications of the answers to questions 1) and 2)? In order to address these questions,

we propose in the present paper to look into the nominal system of AA and more specificdy into plural formation.

4.2.2 Plural Formation in AA Nouns in AA form their plurals either by suffixation, or by an initemai change in the

word. The suffixed plural, also called soundplurul in Arabic grammars (e-g., Blachère & Gaudefroy-Demombynes. 1975). is formed by the affîxation of a plural suffix to the singular fom. (3)

a)

lbas(m)

'dress'

4

!bas&

'dresses'

b)

Tbibs(f)

'doctor'

+

Tbib_pf

'dl~un-~'

C)

mruya (m)

'mimr'

mraYaf.

'&CS'

d)

xObb8z

'baker'

Xt?bb82&

'bakers'

+ -+

The suffix -at(sometimes -r at the surface ievel) which appears in examples (3) a). b). and c) is the most frequent plural suffm in the language. According to McCarthyts

theorv, the singular and plural forms of these nouns are related via a morphologicd rule which attaches the plural suffix to the singular base form.

Figure 15 Sound plural stnicnue

Since these plural forms are produced by regular processes, their singular forms are

hypothesized to be represented in the lexicon dominated by the consonantai root.

1-b-S

I I

lbes 'dress'

lbesat

'dresses'

Figure 16 Lexicai representation of sound plural forms

Aithough the sound/suffixed plural is quite regular and transparent, it is not the most frequently occumng plurai form in the language. Rather, the most fiequent form is the broken plural. a form that exhibits a wide variety of unpredictable stem-internai changes.

'camel'

'cameis'

'grave'

'graves'

'donkey'

'donkeys'

'slave'

'sfaves'

'man'

'men'

'comb'

'combs'

'cup'

'cups*

'chair'

'chairs'

AS can be seen in (4) a) and b), even though both singular forms share intemal structure (CCVC). their plural forms are different. ùi contrast, examples in (4) c) and d)

have the same plural form (CCi C)but exhibit different singular forms. AU other examples

present similar idiosyncratic features. According to McCarthy, broken plural forms are listed separately in a linking relationship of immediate dominance and carry the

idiosyncratic feature [+BPI.

2mal 'camel'

[+BPI

I Figure 17 Lexical represent&on of broken plural fonns

Findy, there are also nouns, referred to as bi-plurals that can take both a sound and a brokeo plural.

Figure 18 Bi-plural forms

These nouns combine feamres of broken and sound/suffixed plurals in their iexical representation. The süigular fonns are assumed to .ave two separate lexical entries.

Figure 19 Lexical representation of bi-plural fonns

4.2.3 Objectives of the Study

Working within the theoretical framework proposed by McCarthy (1979). we investigated whether the proposed theoretical daim on the differential representation and organization of sound and broken plurals in the lexicon is reflected in the pattern of word recognition and lexical access of these f o m . In order to test this hypothesis, we conducted an auditory lexical decision task within a morphological priming paradigm.

Much of the empincal evidence relating to lexical representation and access cornes frorn studies using simple and primed lexical decision tasks.

4.3 METHODOLOGY

4.3.1 Subjects

Control Subiects Twenty-four non brain-damaged Algerian subjects took part in the experiment. They were between 27 and 45 yean old; a l i were nght-handed and their level of education ranged from 12 to 22 years. Aphasic Subjects

Two right-handed Algerian agrammatic aphasic subjects were tested. Both had

suffemd a cerebrovascular accident which caused left-hemisphere brain damage. They were diagnosed as Broca's aphasics on the Arabic version of the Ducarne de Ribaucourt Aphasia Battery (1976). None of the subjects had visud or hearing defects. S 1. a female patient. was 29 years old at the time of testing. She had completed 11 years of education

in Arabic and 9 overlapping years in French. She had suffered an ischemic stroke which resulted in aphasia and severe right hemiplegia. One year post-onset, she was able to

walk but showed residual weakness in her right a m . In off-line tasks conducted in an earlier study (Mimouni & Jarema, in press), she showed mild oral comprehension deficits of morphologically complex words and sentences, with severe comprehension problems with the sarne structures and items in their wntten fomis. The second subject (S2) was a 37 year-old male, who had completed 7 years of education in Arabic and 4 overlapping

years in French. At the time of testing. he showed a rnild right hemiplegia His oral and written comprehension of morphologicaiiy complex words and sentences was found to be

severely impaired in off-line tasks. In oral production tasks, morphologicaliy complex forms were found to be more preserved in S2 than in S 1 (Munouni & Jarema, in press). 4.3.2 Equipment and Procedure

We used an auditory lexical decision task within a morphological priming paradip.

The stimuli were recorded using a Farallon Mac recorder sound system. Testing was done using PsychLab v. 1-00-0.84 (Bub & Gum, 1988). The intervai between words in a pair was 250 mec. Subjects were given unIimited tune for response. The experiment

was divided into 8 blocks of about 7 mn each, and each block consisted of 150 pairs. A

training block of 36 pairs was used. Each subject was presented a different randomized version for each block. The control group of 24 subjects and the two agrammatic subjects were tested individuaiiy in a quiet room. Testing was conducted in 2 sessions of 4 blocks each for the controls, and in 4 sessions of 2 blocks each distributed over a period of two weeks for the aphasic subjects. The subjects were instnicted to decide whether the second item (the target) they heard was a real word in AA by pressing the yes-key or no-key assigned by the program. They were requested to answer as fast as possible and to be accurate.

4.3.3 Stimulus Material Since Our objective was to examine lexical access and representation. as well as lexical relatedness. we designed the experiment so that the singular and plural f o m of the same item appear both as primes and as targets. Thus. with this experimental design, the

criticai stimuli included: 1. Singular-plural and plural-singular pairs in the broken plural group. In this group, we

included 12 monosyllabic and 12 bisyllabic pairs. Singular and plural patterns share

only one discontinuous morpheme, the consonantal root. These two monosyllabic and bisyllabic plural patterns have been selected on the basis of their frequent occurrence in the language.

2. Singular-plural and plural-singular pairs in the sound plural group. Here again, 12

pairs of monosyllabic and bisyllabic words taking the plural suffix-at were incorporated. Singular and plural forms have identicai stems consisting of the discontinuous consonantai morpheme and the vocalic basis. e-g.,

lbas tbiba

'dress'

t)

lbssaf

'driesçes'

'doctor'

f)

tbibd

'docto~~'

3. Singular-plural, plural-singular, and plural-plural pairs in the bi-plural group. There

were 12 pain in each condition. Depending on the type of plural target, the prime shares either stem and consonantal root (sound plural) or consonantd mot only (broken plural) with the target.

4. Finaiiy. in order to assess the productivity of the suffixation process in the language

as weil as its effect on processing, we added a set of pseudowords fomed by an iilegal affixation of the s u e -at to singular f o m of the broken plural group. e.g.,

wald

'boy'

t)

f&7&1

k~p'

t)

* walda * fsnigl&

The word-word combinations also included control pairs of the foliowing types: 27 Double armw is used to express b i d i t i o n a l p-ntation

experirncnt.

of stimuli as prime and target in the same

5. Twelve pairs where the prime shares both singular and plural internai structures with the target. In contrat to the stimuli described in 1. above, the cornmon rnorpheme of these forms is the vocalic basis. The use of this type of stimuli is two-fold: (i) they will serve as contmls for the assessrnent of the amount of priming for morphologicaily

related pairs, and (ü) they will serve to test whether two words having a discontinuous vocalic bais as a single common rnorpheme may prime each other.

6. Twelve pairs where the prime shares the singular pattern of the target. but exhibits a

different plural pattern. If there is an effect of the vocalic basis. this group will give us further confimation.

7. Twelve pairs sharing neither the singular nor the plural pattern. These i t e m are

completely unrelated to the target and should not lead to any faciütation. e-g.,

bab

'door'

biben

'doors'

-+

+

i;rna/

'carnet'

.?mal

'camels'

A list of stimuli is presented in Appendices 4.1, 4.2, and 4.3. We also included

194 distractors and 386 nonwords. The distractors were a set of verbs. The nonwords were constnicted by changing the fust a d o r second phoneme of the real words. Al1

items were controlled for Iength (monosyllabic vs. bisyllabic) and for interna1 CV structure. In the absence of a frequency table for Algenan Arabic. we first selected the stimuli on the basis of our intuitions. Then, for each word selected, we asked a group of *8

Single amiw is uscd to express prcscntation of h

t stimulus as prime and second stimulus as

targct

10 AA speakers to tell us whether they thought it was frequent or not. Finally, we

conducted a simple lexical decision task with a group of 12 native speakers of AA- Al1 items that consistently resulted in reaction times (RTs) higher than 1200 msec. were excluded from the experiment. The two frequency ratings led to the replacement of 1.8%

of the stimuli in the fmal version of the experiment.

4.4 RESULTS AND DISCUSSION

Errors and extreme RTs (greater than 2 SDs) were excluded from the analysis. This

eüminated less than 1% of the responses. Error rates are reported only when significant. Staûsticai analyses were performed on data for the group of control subjects. The aphasies' performance showed high variability across and between the two subjects as

well as a high error rate; we are, therefore, presenting mean RTs separately for each subject without statistical analysis, as we found in a prehninary attempt that this could not capture generalizations reliably. We are, however. presenting a graph comparing mean RTs obtained with aphasic subjects to meaa RTs, minimum RTs, and maximum RTs obtained with controls. 4.4.1 Morphologically Related vs. Unrelated Pairs

Results Tables VI and VII present mean RTs obtained in related and unrelated pairs across type of plural and type of prime.

Controis Patient Monosyllabic plurai-singular critical pairs e-g., imsl-ha1 Monosyilabic control pairs e.g., gdm--1 'heele-'camel'

Patient

683

1056

947

75 1

1110

974

702

1 142

987

7 17

1040

977

8 14

1119

1068

755

1086

1066

897

1166

1052

Bisyilabic plural-singuiar critical pairs

e-g.,

fnaial-fanlal 'mp'-'cup'

Monosyllabic singular-plural critical pairs e-g., tmal-2msl 'came1'-'camels' Monosyilabic conml pairs e-g., gdam-l;mal 'heels'-c~teLc' Bisyilabic singuiar-plural CnticaI pairs e.g., fan2al-fna2al

Bisyilabic control pairs

e.g.,

rnfa&h-fn&l 'key s' -'cups'

Tabie VI Broken plurals: Morphologicaily related vs. unrelateci pairs (mean reaction times in rnsec.)

Controls Patient (N=24) S 1 Plurai-Singular critical pairs lbssêf -1bas e .g.,

Patient S2

752

1049

837

1114

1061

979

1173

1021

99 1

'dresses'-'&& Conml pairs

e-g-,

vraf -/bas 'enve1ope'-'M

Singular-plural critical pairs lbss-lbasa

e .g.,

* - ' - ' ~ '

Control pairs e-g., Ylsfa-\basa 'enve1opes'-'Qesses'

Table VLI Sound pfurals: Morphologidy related vs. unrefated pairs

(mean reaction times in msec.)

Repeated measures and t-test for paired samples performed across groups of subjects and across conditions show that singular and plural fonns are accessed ~ i ~ c a n tfaster l y when preceded by a morphologicdiy related prime than when preceded by an unrelated prime (p < 0.001 for plural-singular pairs in the broken plural group;

p < 0.017 for singular-plural pairs in the broken plural group; p < 0.01 1 for plural-

singular pairs in the sound plural group; p < 0.019 for singular-plural pairs in the sound plural group).

Discussio~ The priming effect observed here is in line with previous psycholinguistic studies

that emphasize the importance of morphological relationships during word recognition

(Emmorey,1989; Fowler, Napps, & Feldman, 1985; Henderson, Wallis, & Knight.

1984; Kehayia & Jarema, 1994; Kernpley & Morton. 1982; Stanners et al., 1979). It also

demonstrates that priming is present even for morphological relatives which exhibit reduced morphotactic t r a n ~ ~ a r e n csince y ~ ~ . in our data, the change of vowels as well as the change in the Linear arrangement of segments (CVCCVC vs. CCVCVC)between the singular and the plural in the broken plural pairs (e-g., faniallfneial) did not prevent the occurrence of priming.

Similar results were found by Fowler et al. (1985) who conducted two experirnents on English words in both visual and auditory modalities and observed strong pruning effects even when the orthographie or phonological representation of morphologically related pairs did not fully overlap (e.g., clear/clarifjr). Along the same line, although Stanners et ai. (1979) did fud a priming effect between related irregular verb f o m such as h u n g h n g or shookMake, priming was reported to be of a lesser extent than that for identical word-pain like h a n g h n g . Our fuidings are, however. in contrast with those reported by Kempley and Morton (1982), who found no facilitation between auditorily presented irregularly related words such as manhien. losr/los& held/holding. One should recall that this lack of facilitation in Kempiey and Morton's results has been imputed to the long delay between prime and target (Fowler. Napps. & Feldrnan, 1985). With respect to the aphasic subjects, we observe that, in generai. their performance is accurate but is much slower than that of the control group. This observation is

consistent with the results reported in the Literature by Prather et al. (1992) and Kehayia (1993). who concluded that lexical access is slowed rather than unavailable in nonfluent aphasies. A point worth k i n g pointed out here is the variability in the performance of the

two patients. S 1's performance is characterized by long RTs, but overall, it parallels that 29 Acording to Kilani-Schoch (1988a:1 18) "plus la segmentation est difficile percevoir. moins grande est la transparence morphotactique". Morphotactic tramparency is gradient and can be measured by the degrce of changes brought about by a process of affixation. The more cornplex the changes. the las the degrce of morphotactic transparency.

of the controls. S2 on the other hand, even though generaiiy faster than SI. does not show the great facilitation obtained with other subjects in morphologically related pairs.

more specifically in the singular-plural pairs for bisyilabic broken plurals as well as as for sound plurals. However, when compared to the controls on a subject-basis. S2's overail performance is hardly different from that of the slowest of the subjects (see Fig. 20). The present results also nile out the possibility that, as a distinct morpheme in an Arabic word, the vocalic bais may have an effect on lexical access. Unrelated pairs sharing a vocalic bais only (e-g., mzsrnar 'naii'

+ f M a ! 'cup*)are not found to yield

any priming. 4.4.2 Sound vs. Broken Plural Pairs

Resul~ As indicated in Tables VI and

VII, mean RTs folr morphological relatii

Vary

across types of plurais. Statisticai analyses of the data on rnorphologically related pairs

show that within the set of plurals tested, broken plurals are responded to significantiy faster (p < .û4) than sound plurals. Further, length was not found to be a confounding variable in the conuol subjects'

performance. In order to veriw whether length was a detemiiniag factor in the differential processing of the plural forms. we performed statistical compaxisons of mean RTs for ali bisyllabic pairs. The same latencies (X= 810 msec.) were found for bisyllabic unrelated

pairs like qahwe 'coffee'lmrqa 'mirror' and malteh 'key'/lanh! 'cup', whereas bisyiiabic related pairs like maftah 'key'lmfatah 'keys' h m the broken plural group yielded much faster RTs (X = 755 msec.) than for bisyllabic related pairs like mraya 'mirror'lmreyet

'mirron' from the sound/suffixed plural group (X = 979 msec.). The possibility that length may have affected access is hence mled out.

With respect to the two aphasic subjects, Figure 20 shows, as mentioned above, an

overail pattern which refiects a slower performance than the one observed in the control group. The patients' mean RTs do not, however, deviate importantly from those of the slowest control subject for both types of p l d s .

O

Controls' rnax RTs

Controls' mean RTs S 1's mean RTs S2's mean RTs

Broken plural

Sound plural

Figure 20 Comparative mean reaction times for normal vs. aphasic subjects Singular-plurai pairs (Broken and sound plurals)

Discussioq Cornparhg the impaired and non-impaired subjects' performance on sound and

broken plurals, we note thai the amwnt of priming among the morphologically related words is greater for broken plural than for sound plural forms. In the absence of

frequency and length effects, this fmdiug points towards a differential processing of the two plural forms. More specificaily. if we assume that during the recognition of suffied

plurais, a decornposition of the suffied form into stem + affix takes place, then we may postulate that the processing of both the stem and the s u f i leads to slower reaction times.

In the case of the broken plural, where suffiation does not occur and plurality is

expressed through stem-interna1 changes. reaction times are faster, pointing towards whoie word access. 4.4.3 Bi-Plural Pairs

That there are differential recognition processes between the broken and sound

plurals discussed above is funher supported by the data on the bi-plurals as shown in Table W. Here again, for the controls, access is significantly (p < 0.03) faster for

broken plural forms tbanfor the altanate suffixexi fomis. Controls Patient

Patient S2

(N=24)

S1

Broken Pld-Singular' criticaI pairs e. g. sier-sah "mes"-"tne"

7 10

1048

1008

Sound Plural-Siagular criticai pairs

7 11

1034

988

130

Io00

917

975*

1 100*

998*

(17)

(u)

.

e .g.,

sefr&seirs "mEs'*-a'mm

Singuiar-Broken plurai nitical pairs e.g., safra-sfer e'~a'-"uœs"

Singular-Sound plural critical pairs e .g., s8ffdds8&a

(3

"~C**-"~I#S"

Targct is somctimes rejected as a nonwod Eigurcs betwecn parentheses indicatc emr rates.

Table Vm Bi-plural pain (mean reaction times in mscc.)

Discussion Apart from the effect resulting from the type of plural used. we also observe in this

group of stimuli what appears to be an inhibition effect when either the singular or the broken plural form is used as the prime for the suffixed bi-plural form (e.g., sier

sairet;seira

+

+ safrat). In these cases, non aphasic as weli as aphasic subjects either

rejected the suffixed bi-plural form as a nonword or accepted it with longer RTs than the ones obtained with the rest of the suffixed fonns. Note, however, that the error rate among non-aphasic subjects is not significant and is reported here only for cornparison purposes. Regarding the inhibition effect observed with the bi-plural forms, two explanations may be advanced. First. the aitemate suffied form in the bi-plural group is associated with a specific semantic context which is required in order for this form to be recognized

easily. According to Heath (1987), the occurrences of these lexical items is restricted to enurneration andfor small numbers of objects. In our experiment, lexical items were presented as single units. The second possibility is that these forms, which are of low frequency, are not stored in the lexicon and therefore, have to be generated through a word formation rule. At the same the. the alternate high-frequency broken plural form inhibits the formation of

the suffixed form. 4.4.4 Plural-Singular vs. Singuiar-Plural Pairs

ResuIe Tables IX and X below, which recapitulate data on related pairs, are presented to isolate a finding that we believe crucial for the present research. Across aii conditions, for

controls singular forms are found to be accessed significantly faster than plurd foms (p < 0.009 for broken plurals; p = 0.000 for sound plurals). Furthennon, singulars of broken plurals yielded faster RTs than those of sound plurals. In conuast. in the bi-plural group, singulars show the same RTs whether the prime is a sound or a broken plural form. A

similar pattern is observed in the performance of the two aphasic subjects. Controls Patient

Monosyliabic broken plural-singular pairs

e-g-,

683

1 056

702

1142

710

1048

2mel-imal 'cameis'-'camd'

Bisy llabic broken plural-singular pain e-g., fnazaI- fanisl 'cups'-'cup' Sound plural-singular pairs e.g-, Ibos8 t -lbtS '*'-'&'

Monosyllabic bi-plurai-singular pairs

e.g.,

shr-saira ' ~ ' * ' ~ '

Sound bi-plural-singular pairs

e-g.,

saira t -saira '~'-'tree

Table IX Singular targets (mean reaction times in msec.)

Patient

ControIs RI=24)

Monosy lIabic singular-broken plural pairs e-g., 2maf -2mal 'camel' -'camels' Bisyllabic singular-bmken plural pairs f an2el-fneEal *cup*-'cups'

Patient SI

Patient S2

717

755

cg-,

e-g-,

Singular-sound plural pairs /bas-lb8~8t '&'-'A'

837

SinguIar-broken bi-plural pairs

e-g.,

sa*-siar '='-'-'

Singular-sound bi-plural pairs

e.g.,

975

sah-saZrst ' ~ ' - ' ~ '

Table X Plural targets (rnean reaction times in msec.)

Discussion First. the overall faster RTs for the singular fonns than for the plural forms in the broken and bi-plural groups may be explained if we hypothesize that plural forms carry the extra feature [+plurai] as part of their mental representation; this added feature may

have induced increased processing time. Second. the faster activation of the singulars of broken and bi-plural forms may be accounted for if we adopt McCarthy's (1979) mode1 of the Iexicon for Arabic. As shown in Figure 16, the singular of suffuted plural forms is hypothesized to be listed in a linking relationship dominated by the consonantai root. On the other hand, words that take a

Second, the faster activation of the singulars of broken and bi-plural forms may be accounted for if we adopt McCarthy's (L979)mode1 of the Lexicon for Arabic. As shown

in Figure 16, the singular of suffixed plural foms is hypothesized to be listed in a linking relationship dominated by the consonantai root. On the other hand, words that take a broken plural involve no morphological analysis as such, rather both singular and plural

foms are hypothesized to be listed separately. linked to each other, the singular f o m and not the consonantal root dorninating the plural (see Fig. 17). These inherent structural differences of Arabic are reflected in the results presented here, suggesting that when the subjects are accessing the singular of words exhibiting the broken plural, the point of entry is the singular form; however, when accessing the singular of words exhibiting the suffixed plural, the point of entry is the consonantal rwt. The longer RTs for the singular forms of words exhibiting the suffixed plural fomis, as compared to broken plural forms, point to this effect. 4.4.5 Illegally Suffixed Words

ResuIts Overall, control subjects either took very long reaction t h e s (p < -03)to reject illegaily sufixed words or accepted these nonwords as reai words (38% mon). The two aphasic subjects. on the other hand, accepted ail these nonwords as real words (100% errors).

MonosyIlabic singutar-illegally suffixed singu1a.rpairs

e.g-,

wald-* w a l d a 'boy'

Bisyllabic singular-illegaiiy suffixed singular pairs

Monosyllabic piutal-illegally suffixed singular pairs

wlsd- *w a l d ' 'boys' BisyIIabic plural-illegally suffuted singular pairs

* Nonword target is sometùnes accepted as mil word. Figures in parentheses indicatc e m r rate.

Table XI nlegally sufhed words (mean reaction tirnes in msec.)

Discussion

The point to be raised here concerns the processing of the plural sumx -ai. The longer reactioc times observed in the processing of the Uegdy suffixed words as targets may result from a recognition of the plural suffix. These results are in line with previous

claims (e.g., Laudanna, Badecker, & Caramaua. 1989; Taft & Forster. 1975) that identification of the affix takes place during word recognition. There is aiso contradicting

evidence that has been reported by Manelis and Tharp (1977) and, Henderson et al. (1984) who did not find any difference in RTs between pseudoaffixed and control words

in English. The &xes they used were aü derivational prefixes which may point toward a different processing of these forms as opposed to inflected f o m . However, what we did End in our results that has never k e n documented so far, at least to our knowledge. is the high number of affirmative responses by a large number of subjects to illegally suffixed items. This pattern is M e r observed in the two patients' performance on affmed nonwords. One rnight argue that this is due to a strategy whereby subjects respond affimiatively to any stimulus bearing a potential s u f f i without hirther analysis of the stimulus. This is a plausible explanation in the context of AA, where the -at ending in words is airnost exclusively associated with the suffixed plural.

4.5 GENERAL DISCUSSION

The results of the present study suggest for AA an organization of the lexicon generally compatible with the one proposed by McCarthy (1979). With respect to the question of listedness, McCarthy adopts Halle's (1973) assurnption that the lexicon consists of a Iist of free and bound morphemes to which morphological rules apply. Taking into account the inherent specificities of Arabic. and elaborating on Haile's rnodel. McCarthy includes the discontinuous consonantal mots in the lexical inventory. He further proposes that morphological relations between words are reflected in lexical organization. A lexical entry is thus hypothesized to comprise al1 morphologicaiiy related forms,

and to be headed by the consonantal mot. which carries al1 idiosyncratic information. However, any element within a lexical entry canying idiosyncratic features is stipulated to also have a separate lexical entry and function as the head of dl other morphologically

related forms (see Fig. 2 1).

\ac. / I J

k- t-b

kitaab

...

'a [+BPI book'

..

katab 'he wrote'

&ut tsab 'Koran schoot'

ki taab

[+BPI ..*

'a book'

-

kutub 'books'

6

Figure 2 1 Lexical representation of idiosyncratic f o m

Turning to Our data, we first found an effect of morphologicai relatedness manifested in the faster reaction rimes for singular and plural forms when primed by morphological relatives; if we assume that this facilitation effect beîween morphologically related items is the result of the simultaneous activation of aii members of a family when access to one of the members is achieved. then this may be taken as an indication that morphological relations are represented in the lexicon, and that mernbers of the same morphological family share strong relations. S i d a r conclusions have been reached by psycholinguists (e.g., Henderson, Wallis, & Knight, 1984; Kempley & Morton. 1982; Stannea et ai., 1979) who emphasized the importance of morphological relatedness in the

process of word recognition. Second, the theoretical distinction between idiosyncratic and non idiosyncratic forms is evidenced in the faster RTs observed for the singular and broken plural f o m

when compared to the suffixed plural forms. If the theoretical distinction betweeo broken and suffixed plural f o m is reflected in the mode of access during word recognitionwhole-word access venus decomposition-and

if decomposition takes time. then this

rnay account for the longer RTs of the suffuced plural forms. Consequently. if we assume

that longer access to suffixed forms reflects decomposition of a word into stem and a f f ~ . then our results cannot be accounted for within a theory where complex words of a language, whether suffixed or prefixed. are accessed as whole uni& without parsing (e.g., Buttenvorth. 1983). The morphological decomposition view which assumes that each rnorphologically complex word is decomposed into its base and affixes prior to or

during access (e-g.. Caramaua, Laudanna, & Romani, 1988; Laudanna & Burani, 1985; Taft 1979) seems best suited to account for the differential processing of sound and broken plurals observed in Our experiment. Further evidence of a decomposition process during word recognition of suffixed forms cornes from the long reaction times observed

with illegaiiy suffmed words. Another aspect of McCarthy's mode1 supported by our data concerns the f d y - l i k e morphological organization of lexical entries. According to McCarthy's view. al1 singular fonns are listed within their respective morphological families, dominated by their discontinuous roors. Furthemore. singulars whose plurals are irregular are stored in two separate lexical enuies. once dominated by the root, and a second t h e dominating the broken plural form. The asyrnmetrical organization characterizing the relationships between head and memben of a given morphologicai farnily is displayed in the results obtained. For example. even though the broken plural f o m are listed together with their corresponding singular f o m separate from the discontinuous root, access to these f o m does not yield the same RTs. The singulars are always found to be accessed faster. If we

assume, however. that access to the plural takes place through the singular. then this form must be activated faster than the plural. The assumption that singulars of broken plural forms are not accessed via the discontinuous root is M e r supported by the RTs obtained in the bi-plural group. If the singulars of the bi-plural forms exhibiting suffixed plurals

were accessed through the discontinuous root, then these forms should yield the same

RTs as singulars in the sound plural group. This is not shown by our data. which indicate instead RTs sirnilar to singuiars in the broken plural group.

Accounts supporting a famiiy organization of the lexicon have already been put fonvard in psychoiinguistic literature (Lukatela et al., 1980; Segui & Zubizarreta, 1985). Regarding the question of the organization of morphological relatives, Lukatela et al. (1980) developed the satellite-entries model for Serbo-Croatian, where an inflected word and not the root is proposed as the nucleus to which ali morphological relatives are

u n i f o d y Iinked. Accoràing to this view, one should expect an equal amount of priming between the nucleus and any of its satellites. In our experiment, even though the results

pointed toward the existence of a lexical form acting as a head of a family. the asymmetry observed in the amount of priming between memben of a morphological family contradicts the view of a satellite organization of affixed forms. Rather, our data suggest that lexical items do not cluster around a nucleus in a uniform fashion. An organization more compatible with our data is that proposed by Segui and Zubizarreta (1985). who

suggested that lexical items are accessed via the lexical entry for the common root (where root = word) which acu as the head of the family. This lexical structure is reflected in our results in that singulars and broken plurals seem to be accessed via the head of their respective families. There are. however, some difficulties with adopting this model to account for Our data. The authoa' clairn that suffuted words are listed, headed by a word. does not hold for AA. Whiie broken plural forms seem to be accessed via the singular, i.e., a word, suffixed forms seem to be accessed via the discontinuous consonantal root.

One of Our fmdings still rernains a pualing matter. It concems the inhibitory effect observed in accessing the suffixed form of broken plurals. Two explanations based on linguistic and psycholinguistics considerations may be advanced, but would require further investigation.

Fint. language-specific semantic constraints may be relateci to this type of words.

Heath ( 1987:100) observed that in some of these cases, "there may be a semantic nuance in the choice, whereby the suffixal kat/ variant is associatexi with enmeration and /or with smaii numbers of abjects". This hypothesis is consistent with our own and other AA native speakers' intuition, but it could not be further documented, as Heath's observation on this semantic idiosyncratic feahlre of these plurals is the only instance that could be

found in the Iiterature. Thus, if these forms occur in restncted contexts. thei. frequency

of use is consequently reduced, which leads us to our next solution. The second possibility is that these forms, which are of low frequency, are not stored in the lexicon and therefore, have to be generated through a word formation rule. At the same time, the alternate high-fkquency broken plural form inhibits the formation of the suffixed forrn. In horse-race models (e.g. Anshen & Aronoff, 1988), this would be explained by a cornpetition between the two forms. The high-frequency broken fonn

would prevent the low-frequency suffixed form from k i n g produced. However, in the absence of a frequency table for Arabic, this hypothesis cannot be verified. Furthemore, the mode1 of Anshen & Aronoff has k e n designed for production. Whether it could be

applied here remains open. Finally, comparing control and aphasic data we may conclude that despite the longer RTs observed in the latter group, overall patterns of word recognition are similar

for controls and aphasics across ail experimental stimuli. The present results do not

support the existence of the morphological deficit found in these patients in off-line testing. On-line testing shows that the patients' processing abilities are well preserved.

leading us to conclude that a single-explanatory account may not be sufficient to undentand agrammatic deficits.

Appendix 4.1 List of stimuli: Broken plurals (Monosyllabic stimuli) Targec t,Pnme

Gloss

Conml prime 1

Gloss

Sing H PIur

Sing/Plur

fmal/imul

gdamlgdam

heel

ktaf/ktaf

hbaVhba1

ZbaVdbal

mountain

Sfar/S far

me'= eyelash

d far/d f a r

nail

xbar/xbar

a piece of news

Tdam/Cdam

bone

hnak/hnak

cheek

b YaVb Yal

mule

SbaT/SbaT

finger

w ald/ wlad

habs/hbas

jail

banthnul

boy giri

Tawd/Cwad

horse

Cabd/Tbad

individual

ka?b/klab

dog

si&

Tars/Crss

marriage

kab5/kba8

dayf/dy8f

l9"'st

T a r f /Traf

qant /qna t

corner

Conml prime 3

Gtoss

ianb/inab

Control prime 2 S ing/Plur

Gloss

qbar/qbur

&rave

bab

dooc

dharldhur

back

kas

glas

paiace

ksr

bus

fmt

fer

mouse

ihar/Shur

month

wad

s

bharlbhur

sea

iar

neighbor

danb/dnub

sin

fus

pickax

chicken

nub

letter

Sag

Sing

m

Appendix 4.2 List of stimuli: Broken plurais (Bisyilabic stimuli) -

-

- - -

-

Target H Rime

Gloss

Sing H Plur fanial/fnuial ma f tah/m fatah sarwal/srawal Cangud/fnagad sandug/snadag saiïum/slalam

ma Trah/rnTurah barnus/brenas

ma Trag/m rarag sabba f.sb8baT maskin/msakan mandil/mnadal

kaskas masrnar manguS fakrun mahbul rnayraf sarduk

maz wad barrad barmil Tafri t

Control prime 2 kursi rabsi babur Sadf kabus iamaf' bendir Tadi in

hanu t za waS

Gloss chair boat

monkey gun moque

tambourine Pan store

Surab

bd Ii p

lizar

curtain

Appendix 4.3 List of stimuli: Soundlsuffixed plurais Target

Prime

Gloss

Sing t)Plur mrays/maraya t luYa/fu va t xadma/xadma t ragba/ragbaC kalma/kalma t S ixa/S txat Tùiba/Tbibat dhez/Zhazat Ibas/lbesa t lsan/lsana t Ylaf/Ylafa t hsab/hsabat

rnirror/mix~ors languagdanguages jobljo bs neckfnecks word/words

Control prime 1

Gloss stepfsteps flowedflowers fiiendlfrknds walk/waiks piIgrim/piIgrims neighborlneighbors auntlaunts

teachedteachers

doctor/doctors dowryldowries dress/dresses

tongue/tongues envelopelenvelopes account/accounts -

Cefsa/Safsa t warda/werda t

-

-

box/boxes

slaplslaps wavefwaves tnzudkd

soup/soups Conml prime 2

GIoss small boat~ -

fle yak frayes qhawr rwaql 3wsda mqaIt

dfayar klawi Tbasa zwawaS hwanat mxayad

brides

coffee places windows monkeys fiying pans brai& ùidneys plates birds

shops piIlows

--

Appendix 4.4 List of stimuli: Bi-plurals Target t)Rime

Gloss

Sing t,Sound PIur~BrokenPlur saira/sairet/s2ar Çam Ca/garn Cat/Srna T namlu/namlat/nrnal Jarfa/qarTaV q r aS naxla/naxia t/nxal nahla/nahla t/nhal haiWha2la t/hf al xaSba/xaS&at/xSab garba/garbat/grab

treehea candle/candles antiants bonielbotties p h tfee/trees fxehees pmidge/pamidges w d sticIdstickS flask/flasks

Conml prime 1 haira/hairat/hiar

Gloss stone/stones cow/cows sheet/sheets lofloaves streetlstreets hoiefholes lourdice baskethaskets short tabldtables

Control ~ n m e2

Gloss

dru2

stain stars

nium drus 9dw

dinah riai hyuT sfuf

teeth

cooking pots wings

men walk rows

CHAPTER FZVE GENERAL DISCUSSION

5.1 INTRODUCTION

The overail results of the two snidies provide converging evidence on a number of theoretical points and issues related to the lexicon. These will be reviewed in the general context of the hypotheses underlying the present research, namely: 1 . A differentiai processing of regular and irregular plural forms wiil be observed

dunng word recognition. 2. This difference is a reflection of the feature specificatioas these forms cany and

of their structural organization in the lexicon. We will proceed fust with a summary of the data from Study one, and undencore the major points of morphological breakdown in AA. An overview of the major fmdngs related to the second study will then be presented. Findly, a discussion of the currendy

debated issues of listedness, relatedness and the role played by frequency in the lexicon wiii foilow in light of the findings h m Study one and Study two.

5.2 MORPHOLOGICAL DEFICITS Studies of language deficits have been carried out for various objectives. One of them is to use the patterns of deficits observed to validate existing theories of the normal

language system. Morphological deficits. for example, have been extensively investigated in various tasks and in different languages in order to understand the mechanisms underlying lexical representation and processing. More specificaily, in agrammatic aphasia, morphological breakdown has maidy involved omission and substitution of bound and free tical al

morphemes leading to reduced-length sentences as welI as syntactic ~ i m p ~ c a t i o nThese . general characteristics of the syndrome of agrammatic aphasia have ernerged from various studies across various languages (e.g., Goodglass, 1968; Kean. 1977; Goodglass et al., -

1972; Tissot et . al., 1973; Bates, Friedenci, & Wulfeck, 1987; Grodzinsky, 1984; Contributors to M e m & Obier, 1990)Another feature that emerges from agrammatism accounts concerns oral comprehension of complex structures, which has been reported to be from rnildly to severely impaired (e-g., Caramazza & Zurif. 1976; Schwartz. Saffran, & Marin. 1980; Bemdt & Cararnazza, 1980; Linebarger, Schwartz. & Saffran, 1983; Grodzinsky, 1986; Caplan & Hildebrandt. 1988) . However, given the extent of individual variability across

patients, on the one hand, and the typology of the language. on the other hand, varying degrees of impairment appear to affect some or all of these Linguistic aspects. Let us now consider each point separately in the context of our fmidings.

5.2.1 Free Grammatical Morphemes As reported in the literature, free grammatical morphemes were liable to omissions

in the performance of our three patients. A close examination of the omission patterns of the two types of free Dorammaticaimorphemes elicited in the tasks of sentence repetition,

reading aloud and recitation reveals that subject pronouns are more resistant to breakdown than relative pronouns. We attributed the omission of subject pronouns to the fact that AA

is a pro-drop language. Consequently, even though marked for gender and number, these items are not obligatory in the construction of a syntacticdy weU-formed sentence except in the cases of emphatic and cleft structures. On the other hand. relative pronouns which do not carry any morphological markers are relatively less resistant to omissions. In the sentences tested, these items are mostly part of complex syntactic structures such as the cleft structures, which are found to contribute to the great processing difficulties experienced by the patients.

In addition, the patients' performance in free speech is also characterized by the absence of these two types of free standing morphemes as well as prepositions and conjunctions in obligatory contexts. 5.2.2 Bound Grammatical Morphemes

Omissions of bound morphemes constitute the prominent feature in our data. Articles, verb prefixes and suffixes, clitics, and negation markers are al1 prone to omissions in substantial quantities, with some variations in the error rates across tasks and subjects. In the case of articles, negation marken, verb clitics, and verbs, omissions occurred across the board, in obligatory as well as non obligatory contexts with the

exception of the recitation task. where the four clitics have been retained by two subjects; the third subject could not perfonn this task.

With respect to tense. number, and gender verb markers. we observe three different patterns: Fint, in two of the subjects (NB and OH) omissions of verb prefixes resulted in tense substitutions (e-g., ya-drab 'he hits'

+ drab 'he hit'). The errors represented 38%

and 44% respectively for NB and OH in the reading aloud task, and 60% and 52% in the repetition task.

The second pattern is illustrated in the omissions of verb prefixes leading to tease and gender substitutions (e.g., t a-drab 'she hits'

+ drab 'he hit'). The errors represented

63% and 56% respectively for NB and OH in the reading aloud task, and 69% and 48% in the repetition task.

Substitutions also occur, leading to a third pattern where they are found to affect entire verb forms resulting in an item which is different in gender andor tense (e.g., t a drab 'she hits'

-+ darb-at

'she hit'). However, substitutions occur to a much lesser

extent than omissions (36% and 1 1% respectively for NB and OH in the reading aloud

task, 15% and 14% in the repetition task). The most interesting finding here Lies in the form resulting once omission and/or substitution erroa have occurred. In many cases, it is not clear whether these forms should be taken as substitutions or omissions. since they are manifested at the surface level as omissions of bound morphemes whiie reflecting at the sarne time substitutions of gender and tense. For exarnple, when the subjects produce drab 'he hit' instead of yadrab 'he hits'

.are they omining the prefur ya-, or reverting to another tense.

Furthemore, omissions odand substitutions of verb prefixes and suffixes never Iead to the production of nonwords. In the present data, each t h e the omission of either a prefix or a suffix would result in a stem which is not a word in the Ianguage, patients revert to another fonn or drop the entire verb form aitogether. In AA, the distribution of verb stems in the two aspectual paradigms shows that in IO out of 16 cases, the verb stem is a real word correspondhg to the masculine 3rd singular perfective. However, in the

rest of the paradigms, the stem may be a nonword as in katb-at which. when suipped of its prefix, will give the bound stem *katb. Productions of this type of stem as a result of

sufflx or prefuc omission have never been found in our data.

5.2.3 Oral Comprehension The data indicate that sentence comprehension is less problematic when canonica

word order is present and verb morphology involves only gender and number H i a t i o n . Therefore, VSO , S V 0 and Cleft SV0 structures which involve canouical word order and gender and nurnber verb rnorphology are found to be easier to process. Error rate is very low and varies from O to 14.2%. We have already observed in the tasks of reading aloud

and repetition of sentences, an impairment of verb morphology (verb clitics, tense, gender, oumber and negation verb markers). The same pattern is found in sentence comprehension. where the combination of verb morphology and noncanonical word order in structures such as OVpmS,Cleft OVpmS,NegSVO, and NegOVpmSincreases difficulty of comprehension. The mculties experienced by the patients appear to be cornmensurate with the complexity of the structures, on the one hand, and noncanonical word order, on

the other. However, if we compare the patients' performance in oral production with their performance in oral comprehension, a cornmon problematic feature that emerges in both modaiities is the complexity of the verb morphology. The more complex the verb

morphology. the higher the error rate. This points to the role of language-specific features

in the performance of the three agramatic patients. 5.2.4

Conclusion The theoretically central result that emerges from this study Lies in the pattem of

omission of bound grammatical morphemes in oral production. In contrast with several studies on agrarnmatism, and in parùcular to Grodzinsky's (1984) study on Hebrew. our patients are found to produce more omissions than substitutions in their oral speech. Furthemore. as observed in cross-linguistic snidies, these errors follow a specific pattern whereby omissions and substitutions occur o d y if the resulting form is a word; this happens even when the items undergohg omission yield forms that are phonologicalIy possible in the language. The questions that corne to rnind here are the foliowing: why do omissions occur more frequentiy than substitutions in AA? Why would omissions lead only to real words even in the cases where stripping items of their prefixes would not result in a phonotacticaiiy impossible item? We will attempt to answer these questions in the generai discussion in light of linguistic and psycholinguistic theories on the lexicon.

Another important observation that must be pointed out is the variability among the three patients. Several instances of striking individual ciifferences deserve to be reported. First, as mentioned in Study one. one of the patients did not perform the recitation task as he could not remember any of the verses of the Koran. Second, OH shows excelient oral comprehension of sentences, as opposed to NB and RB, whose individual performance may be characterized as mildly and severely impaired respectively. The same pattern is found in reading aloud, except in the case of verb clitics, which are al1 omitted by OH (100% errors). Finally. within the repetition task, we have a different picture. Here, it is

RB who shows an excellent performance on verb morphology.

Given this variability. one should be cautious in capturing generaiizations from agrarnmatic data. This factor wiii be kept in mind in our general discussion of the overall data.

S. 3 LEXICAL REPRESENTATION

AND PLURAL FORMS IN AA

AND PROCESSING OF SINGULAR

A long-standing debate in linguistics and psycholinguistics concerns the

representation and access of morphologicaily complex words in the lexicon. Researchers have been addressing questions like the foliowing: is an inflected word like cats stored in the lexicon as a whole word together with its corresponding singuiar form cat ? is it stored

as a base f o m cat to which a sufnxation d e attaching the plural s u f f i -s would apply? Would the words cat and cats be related in the lexicon and, if, so, how? How is the inflected word cats accessed? Similar questions have been asked for derivationallyforrned words. There is an obvious absence of consensus on these issues. which highlights the

complexity of the question regarding the representation and access of morphologically complex words. Even though major advances have k e n made in this field, there is still

no strong evidence in favor of one theory or model over another one. The existing theones wiU be examined in the coune of the discussion of the overall f~ndings. But before we proceed. let us go over the main characteristics of the data obtained in the prirning experiment on AA sound (or regular) and broken (irregular) plurals. With

respect to the two aphasic patients, the data coilected on their performance show high variability between the two, which preclude any a reliable statistical anaiysis. Therefore, in order to compare their performance to that of the normal subjects. we took a different

approach. We looked at the trends observed for each aphasic patient and compared them

to the performance of the slowest subject in the control group. We notice that, overail. the aphasic patients follow a pattern that is slower, but which parallels die one characterizing the control group.

5.3.1 Morphologically Related vs. Unrelated Words Singular and plural fonns are found to be responded to significantly faster when the prime is a morphologically related word than when it is not. For example, a singular word like imal *came19is accessed faster when the prime is the comsponding plural imsl 'carnels* than when it is an unrelated word like qdam 'heel*. This is m e for al1 types of stimuli (monosyllabic, bisyllabic, broken and sound plural) and for aphasic as weli as non aphasic subjects. 5.3.2 Broken vs. Sound Plurai Forms

Arnong the morphologicaily related pairs, the data indicate a differential processing

of broken and sound plurals. a finding that c o n f m our fmt hypothesis. Broken plural f o m , which exhibit a change intemal to the stem, show significantly shorter reaction times than sound (suffixed) plurals do. As an example, in the broken plural pair fnefal 'cups'-fanisi 'cup', the singular form is accessed faster than in the sound plural pair l b 8 ~ 8 tdresses'-lbss

'dress*. This effect is also observed among the set of nouns taking

both the broken and sound plural forms. A sunilar trend is also found in the aphasic data. A comparative analysis of stimuli with different lengths niles out the possibility that length

may have had an effect on the results.

5.3.3 Singular vs. Plurals Forms Across al1 conditions and subjects. singular forms are accessed significantly faster

than plural fomis. However, with respect to the singular group, we observe a differential processing among the three types of singulars. First, RTs obtained for singulars of broken plurals are shorter than those obtained with singulars of sound plurals. The second point concems the bi-plurai group where no such d i f f e ~ n c eis observed between the singulars. We will show in the general discussion that this difference in the processing of singular forms is a reflection of the feature specifications these forms cany and of their structural organization in the lexicon. 5.3.4 Bi-Plural Pairs

Most of the results in this group of stimuli support what has been outlined above for the other types of stimuli. The data reveal, however, an inhibition effect with respect to the sufixed fom. The suffixed targets are either rejected as nonwords or exhibit very long reaction times. 5.3.5 Illegally-Suffixed Words

The data on iiiegally suffixed words which consist of mono- and bisyllabic singular foms to which an iiîegal plural suffix -nt was attached (e.g., *waldst), give a quite puviing picture. The long reaction t h e s observed were predictable and consistent with other studies reported in the literature (e.g., Laudanna, Badecker, & Caramaua, 1989:

Taft & Forster. 1975) which led to the hypothesis that suffixes may be identified during word recognition. However, the high number of positive responses to these nonwords was unexpected. More irnportantly, the aphasic subjects showed a 100% error rate on

these items. No convincing explanation has been found to account for these data

5.3.6 Conclusion Three major findings emerge from this study in which we examined the ways

singular and plural nouns in AA are represented and accessed. A diff'erentiai processing is observed between (i) morphologicaliy related pairs and unrelated pairs, (ii) broken and

sound plural forms. (iii) singulars of the broken plural forms and singulars of the sound plural forms.

5.4 G E N E M L DISCUSSION

The overail results of the two studies suggest for AA an organizatioa of the lexicon generally compatible with the one proposed by McCarthy ( 1979). 5.4.1 The Nature of Lexical Representation for AA

First, with respect to the question of listedness. McCarthy adopts Halle's mode1 with a lexicon containing morphemes (roots. stems and voclic bases) as weiI as words: "1 assume. essentially following Halle (1973). that the lexicon is fully specified with ail f o m . including inflections" ( M c C a y l979:388).

.

Applied ?O o u data, this "fuliy specified lexicon" would include a word iike taktab 'she writes' together with its corresponding stem ktab 'he wrote'. Similarly, katbat 'she wrote' would be stond as a fully affixed f o m and its corresponding bound stem * k a t b would also appear in the lexicon. Furthemore, this type of lexicon would aiso contain singular forms as well as irregular and regular plural f o m such as fnaial 'cups'. faniel

'cup' , [bas 'dress', Ibesat 'dresses'.

Retuming to Our first study, the data show that the patients, when experiencing difficulties. produce the less complex or free stem contained in the target word (e.g.. ktab

'he wrote') rather than the more morphologically complex elicited lexical item (e.g., takt ab 'she wrote*). If the two forms are lisred as assumed by McCarthy's model, why

would the aphasic patients tested in Study one tend to produce lexical items stripped of their prefixes and suffixes? One may also wonder why bound stems which are hypothesized to be listed never occur in the patients' performance as a result of prefix and

suffix omissions. Another set of questions may arise when we examine the results of the second study. If regular and irregular plurals are listed, why would they show a differential processing pattern in the task of recognition? Why would a singular word like imal 'camel' be accessed faster than a singular word üke lbes 'dress'?

The answers lie in the properties inberent to the laquage which McCarthy incorporates in his mode1 of the lexicon. In McCarthy's proposai of lexical stmcture for Arabic, there is an intenvoven relationship between the nature of the items making up the lexicon and the way these items are organized, 5.4.2 The Interna1 Organization of the Lexicon for AA

Taking into account the specificities of Arabic, and elaborating on Haiie's model, McCarthy includes the discontinuous consonantal mots in the lexical inventory but daims that ody fully-infiected fonns are subject to lexical insertion. He further proposes that

morphological relations between words are reflected in lexical organization. A simplified version of a lexical entry as proposed by McCarthy is presented in Figure 2 1.

/\ \ k-t-b

katab 'he wrote'

kitaeb

.

[+BPI

kitasb

[+BPI *.*

.m.

kut taab 'Koran school'

..a

'a book'

I

kutub 'books'

'a book'

Figure 2 1 Lexical representation of idiosyncraric forms

A lexical entry is described as comprising ali morphologically related f o m . The

consonantal root, which carries al1 idiosyncratic information constitutes the head of the lexical entry. Therefore, a lexical entry has the shape of a family headed by a common discontinuous root which, in fact, defines the morphologicai family. In Figure 2 1, the words katab 'he wrote', kitaab 'book', kut taab 'Koran school' a l l share a common root and hence are ail memben of the Ik- t -b/ morphological family. However, any element wiihin a lexical entry carrying idiosyncratic features is stipulated to also have a separate lexical entry and function as the head of al1 other morphologically related f o m . This is illustrated in Figure 21. The singular word kiteab 'book', which c e e s the idiosyncratic feature [+broken plural] must start a new family for which it becomes the head. Such a structural organization is reflected in our data. First, with respect to the singular and plural nouns, the effect of morphological family is manifested in the facilitation effect observed when morphological relatives prime each other. In the singular-plural as well as in the plural-singular pairs, singular and plural forms used as

targets are accessed significantly faster when the prime is either a singular or a plural belonging to the sarne morphological family, i.e., sharing a common discontinuous root, than when the prime is an unrelated item, i.e., with a different discontinuous root and a different or similar vocdic basis. The roIe of the discontinuous root as a head of a morphological family is further strengthened by the lack of facilitafion in pairs sharing only a discontinuous vocaiic basis. The fact chat morphological relatedness is represented in the lexicon is funher evidenced in the performance of the agrammatic patients in off-line tasks. The erroneous forms produced are not randornly selected, but singled out among items which are morphologically related to the elicited target. Converging evidence for this view cornes from a number of psycholinguistic studies which outline the importance of morphoIogical relations in the process of word recognition (e.g.. Fowler, Napps, & Feldman, 1985; Henderson, Wailis, & Knight, 1984; Kempley & Morton, 1982; Segui & Zubizaretta, 1985; Stanners et al., 1979) A second point concems the implication of the "headness" status of mots and words

for the representation of broken and plural words in the lexicon. The structural organization of these two types of plurai is iuustrated in Figure 22:

1-b-s

I I

lbes 'dmss'

lbsss t 'dresses'

Figure 22 Lexical representation of sound and broken plural forrns

As indicated in Figure 22. a sound plural form like lbasat is dominated by its

singular form Ibas, which in tum is headed by the discontinuous root 1-b-s, whereas a broken plural form iïke imel is directly dominated by the singular imal. If this theoretical distinction between the singular and the plural forms of the two types of plurals is reflected in the way these forms are accessed, then we should expect a differentiai

processing of these forms.

From the standpoint of the morphoiogical decomposition as initially advocated by Taft and Forster (1975), the longer reaction times for suffixed forms observed in Our data are the result of additionai time needed for the search and identification of the plural suffix.

In the case of the broken plural forms where no suffix is involved, access time is thus shorter. which is evidenced by our data. The theoretical assumption behind Taft and Forster's model, however, is different from the one we have adopted in our research. While Taft and Forster's decomposition hypothesis States as a point of departure that morphologically complex words are stored into their base and constituents. our theoretical framework assumes that words are listed as whole words together kt!their constituents. Let us examine now Taft and Forster's (1975) alternative explanation regarding the

decomposition hypothesis. According to the authors. the decomposition hypothesis does not preclude the possibility that words may be listed as whole words together with their interna1 swcnire. On the contrary, this option is given full consideration, the only difference being in "assumptions about how the prefixes are represented in the lexical entiy". Accordingly, the issue here is not whether a complex word is listed as a whole lexical item or not, but rather whether access involves analysis of a word into its constituents or not.

Hence, the hypothesis that complex words may be listed together with a representation of their intemal structure (e-g., ((1bas)e t)), is compatible with McCarthy's model which provides for an internai anaiysis of words through a set of morphologicai d e s . These rules are used as redundancy rules, and therefore apply "redundantly rather

than generatively, except in the case of neologisms" (McCarthy, 198L:375). Thus. complex words cm be stored in the lexicon and analyzed into their constituents. a hypothesis that has been suggested by Jackendoff ( 1975). Stemberger and MacWhinney ( 1986)have also claimed rhat Engiish high-frequency

regularly inflected f o m are stored

in the Lexicon and analyzed into their constituents whereas low-frequency regularly inflected forms are not, This view seems. then. best suited to account for the differentiai processing of sound and broken plurals observed in the priming experiment. Further support for this interpretation cornes from the processing of illegaiiy suffixed words, which yielded long reaction times, indicating a recognition of the s u f i . On the other hand, our data cannot be accounted for within the Full Listing

Hypothesis (Butterworth, 1983) according to which complex words are listed and accessed as whole units without parsing. Such a theory would predict, in terms of access

time, equal reaction ùmes for broken and sound plurals, a hypothesis that is not confhned by the present results. Another aspect of McCarthy's model that should be discussed concerns the organization of lexical items into rnorphological families. The singulars in the sound plural group are hypothesized to be headed by the discontinuous root whereas the singulars of broken plural forms appear twice. once headed by the root. and another time heading the broken plural fom. Thus, if we assume that access of the singular foms

takes place through their respective heads, i.e., the root, then we should expect equal RTs for d l singular forms. This is not confirmed by our data, which show that singular forms in the broken and bi-plural groups are accessed significantly faster than singular forrns in the sound plural group. This may be explained if we hypothesize that when accessing the

singular of the broken plural, the point of entry is the singular form; however, when accessing the singular of the suffied plural form, the point of entry is the consonantal root. This hypothesis is hinher evidenced by the singulars in the bi-plural group which are found to show similar RTs to singulars in the broken plural group even, though they combine features of sound and broken plurals. Accounts supporting a family organization of the lexicon have already been put forward in psycholinguistic literanire (Lukatela et al., 1980; Segui & Zubizarreta, 1985).

The sateilite-entries hypothesis developed by Lukatela et al. ( 1980) proposes that every word has its own lexical entry, but aU rnorphologically related entries are linked in a uniforrn fashion to an inflected word which acts as the nucleus. On such a view, one should expect an equal arnount of priming between the nucleus and the other membes of the morphological facnily. In Our data, the difference in reaction Urnes among the singular forms and the plural forms does not support a satellite organization. A family organization more compatible with our results is the one proposed by

Segui and Zubizarreta (1985) who suggest that lexical items are ali stored and accessed via the lexical entry for the cornmon root (the mot king a word) which acts as the head of the

farnily. This lexical structure is reflected in our results in that singulars and broken plurals seem to be accessed via the head of their respective families. However, the authors' daim that suffixed words are Iisted, headed by a word, does not hold for AA. As has been shown for AA, while broken plural forms are accessed via the singular, i.e., a word. other forms are accessed via the discontînuous consonantal root.

This farnily organization seems most pertinent to account for the morphological errors produced in verb morph0logy3~by the three agrammatic patients. Lf we assume that aii verb forms are listed as whole uni& and organized into a family headed by the free stem (i.e.. the masculine 3rd singular perfective) through which access takes place, then this may explain the high rate of free stem productions as a result of omissions. In trying to access the target within the morphological family. the patients cannot go beyond the head of the family and produce it instead. In this case, these results would suggest that morphological impairments reflect deficits in lexical access and not lexical representation.

Further support for intact lexical representation cornes fkom the occasional production of entire verb forrns as a result of substitution. We are thus proposing that the masculine 3rd singular perfective (the form our

subjects most frequently resort to when in ciifficulty) represents the head of a verb family in AA, and access to any member of the f a d y is achieved via this f o m The present data show that free stems and bound stems do not play the same role in the lexicon. We believe, however. that, in order to capture generalizatioas, verb morphology in Arabic should be further investigated on a wider range of stimuli. using different types of testing material. Finally, comparing control and aphasic data in on-line testing. we may conclude that despite the longer RTs observed in the latter group, overall patterns of word recognition are similar for controls and aphasics across ail expenmental stimuli. The present results do not support the existence of the morphological deficit found in these patients in off-line testing. On-line testing shows that the patients' processing abilities are well preserved.

30 We resaict our discussion to veh morphology as other erron are beyond the scope of lhis research.

These findings, together with the vaîiability displayed in the patients' performance in off-line testing lead us to conclude that one should be cautious when making generaiizations from agrammatic data 5.4.3 The Role of Frequency in the Lexicon

Even though McCarthy's mode1 is best suited to account for our data, it does not, however, make predictions for the inhibitory effect observed with the bi-plural foms. In this group, broken forms used as primes are found to either trigger long RTs or lead to

rejection of the target. A possible explanation may be found in psycholinguistic models of the Iexicon which incorporate frequency as a major deteminant of lexical organization.

In models for a lexical look-up such as the one developed by Anshen and Aronoff (1988), two word forms compete in the race and the wimer is the most frequent one.

Applied to the English verb system, the mode1 predicts that hi&-fkquency irregular f o m

block or prevent the competing regular f o m from king produced. Now tuming to our data, if we hypothesize for the bi-plural group that the suffixed plural fonns are not listed

but generated by a word formation rule, and that the alternate broken form for the bi-plural category of words is of higher frequency than the suffixed form, we should expect the inhibition effect we observed. In this parùcular case, the competition between the lexicalized broken form and ihe suffixed fonn may be the cause of the long reaction ùmes the subjects take when they have to make a lexical decision on the suffixed form. The

question that arises here is whether, and if so, to what extent, production modeis may be applicable to our data Unfortunately, in the absence of a frequency table for Arabic. this tentative interpretation is mere speculation and will remain a hypothesis to be verified once frequency counts for Arabic become available.

5.5 IMPLICATIONS FOR FURTHER STUDIES

While working and reflecting on the issues related to this research, severai other interesting questions have corne up. Some have already been mentioned in the process of the data interpretation. We have reported in the literature review that in recent studies productivity has been found to correlate with frequency rather than with regulariq or transparency as a number of authors (e.g., Bybee & Slobin, 1982; Pinker, 1991; Bybee, 1995) claim that certain classes of verbs within the English imgular past-tense forms, even though listed, show

some degree of productivity. This is the case of the Arabic plural forms. The broken plural €omis, despite their high degree of irregularity, show fbil productivity whereas the more regular and transparent sound plural forms do not. Our own knowledge of the langage and more specificdy of Algenan Arabic points to the opposite view that plural suffixation U productive. We have observed that this form has k e n increasingly used in Ioan words and extended to existing words which take broken plural forms. We have started investigating this issue in a probe task. Presentiy, we are in the data collection process. According to Bybee (1995:442), it appears "that its [the sound plural] productivity is extremely limited, due to its low type frequency". Type frequency,

however, will not be integrated in our study since there is not fiequency table for Arabic yet. Another point that we raised in our discussion is verb affmation. In order to probe McCarthy's claim regardhg both the status of the head and the lexical status of the

discontinuous rwt in Arabic, and to further v e m the morphological family hypothesis, a

similar priming experiment to the one conducted in this research should be extended to verbs, which display an even more complex morphology than nouns. if the free stem (the

masculine 3rd singular perfective) is the head of the verb family. then in a priming experiment. we should expect a differential processing between free stem and other prefixed or suffured memben of the family and similas processing between members of the sarne family. Another issue that could be dealt with is whether there is any modality-

specific effect by using a visual lexical decision task. This is even more pertinent for Arabic given the differences between the spoken and written f o m .

5.6 CONCLUSION

The contributions of the preseni research are two-fold. On the one hand, it c o n f i i the view that linguistic theoretical iools are needed to provide for a more adequate account

of data on language processing and language pathology. On the other hand. it underscores the fact that the integration of evidence from different fields such as aphasiology and psycholinguistics and from typologically diverse languages like Arabic into linguistic tbeory offers a better picture of the various mechanisms underlying language use in the normai speaker-hearer.

REFERENCES

Alajouanine, T. ( 1956).Verbai realization in aphasia Brain, 79, 1-28. Anderson, S.R. ( 1982). Where's morphology? Linguistic Inquiry, 13, 57 1-6 12. Anderson, S.R. ( 1992). A-morphous morphology. Cambridge, MA: Cambridge University Press. Anshen, F. & Aronoff, M. (1981). Morphological productivity and phonological transparency. Canadian Journal of Linguistics, 26,63-72. Anshen, F. & Aronoff. M. (1988). Producing morphologically complex words. Linguistic, 26. 64 1-655. Aronoff, M. (1976). Word formation in generative grarnrnar. Linguistic Inquiry, Monograph #1. Cambridge, MA.: MIT EVess. Baayen. H. & Lieber, R. (1991). Productivity and English denvation: A corpus based study. Linguistics, 2 9, 80 1-843. Bates, E., Friederici, A., & Wulfeck. B. ( 1987). Comprehension in aphasia: A crossIinguistic snidy. Brain arzà Languuge ,32, 19-67. Berndt, R.S. & Caramazza, A. (1980). A redefinition of the syndrome of Broca's aphasia: Implications for a neuropsychologicai mode1 of language. Applied Psycholinguistics, 1. 225-278. Blachère. R. & Gaudefroy-Demombynes. M. (1975). Grammaire de l'orabe classique. Paris: Maisonneuve et Larose. Bloomf~eld,L. (1933). Language. New York, NY: Holt.

Bohas. G. ( 1993). Développements récents en linguistique arabe et sémitique. Séminaire tenu au Collège de France. Damas, Syrie: Institut français d'études arabes de

Damas. Bohas, G. & Guillaume, J.P.( 1984).Étude des théories des grammairiens arabes-Tome I Morphologie et phonologie. Damas: Institut français de Damas. Bradley, D.C. ( 1980). Lexical representation of derivational relation. In, M. Aronoff and M. Kean (Eds.), Juncmre. Saratoga, CA: Amma Libri & Co. Bub, D.N.& Gum, T. (1988). PsycNab Software. (unpublished). Buttenvonh, B. (1983). Lexical representation. In, B. Butterworth (Ed.), Language production Vol. 2, (pp. 257-294). London: Academic Press. Bybee, J.L. (1988). Morphology as lexical organization. In, M. Hammond and M. Noonan (Eds.), Theoretical rnorphology: Approaches in modern iinguistics, (pp. 119- 14 1). San Diego, CA: Academic Press. Bybee, I.L. (1995). Regular morphology and the 1exIcon. Lnnguage and Cognitive Processes, 1O(5), 425-455. Bybee, J.L. & Slobin, D.I. (1982). Rules and schemas in the development and use of the English past tense. Language, 58,265-289. Cantineau, J. ( 1950). La notion de «schème» et son altération dans diverses langues sémitiques. Semitica. 3. Caplan, D. & Futter, C. (1986). Assignment of thematic soles to nouns in sentence comprehension by an agrammatic patient. Brain and Language, 27, L 17-134. Caplan. D. & Hildebrandt, N. ( 1988). Disorders of syntuctic comprehension. Cambridge, MA: MTT Press. Caramazza, A., Laudanna, A., & Romani, C. (1988). Lexical access and inflectionai morphology. Cognition. 28. 297-332. Caramma, A.. Miceli, G.,Silveri. M.C., & Laudanna, A. (1985). Reading mechanisms and the organization of the lexicon: Evidence from acquired dyslexia. Cognitive Neuro-Psychology. 2, 8 1- 1 14. Caramazza, A. & ZUI'if, E.B. (1976).Dissociation of algorithmic and heuristic processes in language comprehension: Evidence from aphasia. Brain and Lmguage, 3,572582. Chomsky, N. (1970). Remarks on nominalization. In, R. Jacobs and P. Rosenbaum (Eds.). Readings in English tm.formational g r a m a r , (pp. 184-22 1). Waltham, MA: Blaisdeîi.

Chomsky. N. (1988). Language and problerns of knowledge. l'le Managua lectures. Cambridge, MA: MIT Press.

Cole, P., Beauvillain. C., & Segui. J. (1989). On the representation and processing of prefixed and suffixed derived words: A differential frequency effect. Journal of Memory and Language. 28, 1- 13. de Bleser, R. & Bayer. 1. (1988). On the role of inflectional morphology in agrammatism. In, M. Hamrnond and M. Noonan (Eds.), Theoretical morphology: Approaches in modern linguirtics, (pp. 45-69). San Diego, CA: Acadernic Press. Dell, F. & Vergnaud, .J.-R. (1984). Les développements récents en phonologie: queIques idées centrales. In, F. Dell. D. Herst, and .i.-R. Vergnaud (Eds.), Les formes sonores du langage. Paris: Hermann. Deii, G.S. (1986). A spreading-activation theory of retrieval in sentence production. Psychological Review. 93. 283-32 1. Ducame de Ribaucourt, BI. (1976). Test pour l'examen de l'aphasie. Paris: Centre de psychologie appliquée. Emmorey. K.D. (1989). Auditory morphological priming in the lexicon. Languuge and Cognitive Processes. 4(2), 73-92. Feldman, L.B. & Bentin, S. (1994). Morphological analysis of disnipted morphemes: Evidence from Hebrew. The Quarterly Journal of Experimental Psychology, 47A(2), 407-435. Ford, A. & Singh. R. (1991). Propédeutique morphologique. Folia Linguistica. XXV(3/4)* 549-574. Fowler, C.A.. Napps. S., & Feldman, L.B. (1985). Relations among regular and irregular morphologically related words in the lexicon as revealed by repetition priming. Memory and Cognition. 13(3), 24 1-255. Goldsmith, J. (1976). Autosegmental phonology. Doctoral dissertation. Cambridge, MA: MIT, (published in 1979. Garland, New York). Goldstein. K. ( 1948). Language and language disturbances. New York, NY: Grune & Straton. Goodglass, H. (1968). Studies on the grammar of aphasies. In. S. Rosenberg and J. Koplin (Eds.), Developments NI applied psycholinguistics research, (pp. 177-208). New York, NY: MacMillan. Goodglass. H., Gleason, J.B.. Bemholtz. N.A., & Hyde, M.R. (1972). Some linguistic structures in the speech of a Broca's aphasic. Cortex, 8, 191-212.

Goodglass, H. & Kaplan. E. ( 1972). ï h e assessrnent of aphasia and relured disorders.

Philadelphia, PA: Lea and Febiger. Grodzinsky , Y. ( 1984). The syntactic characterization of agrammaùsm. Cognition, 16, 99- 120. Grodzinsky. Y. ( 1986). Language deficits and the theory of syntax. Brain and Language,

27, 135-159. Grodzinsky, Y. (1990). Theoretical perspectives on language deficits. Cambridge. MA: MIT Press. Halle. M. (1973).Prolegomena to a theory of word formation. Linguistic Inquiry. 4. 316.

Heath, J. ( 1987).Ablaur and ambiguity phonology of a Moroccan Arabic dialect. Albany, NY:State University of New York Press. Henderson, L. (1985). Toward a psychoiogy of morphemes. In. A.W. Ellis (Ed.), Progress in the psychology of Imguage. VoLI. London: Lawrence Erlbaum. Hendenon, L., Wallis, J., & Knight, D. (1984). Morphemic structure and lexical access. In, H. Bouma and D.G.Bouwhuis (Eds.). Attention and pe$onnonce X: Control of language processes. Wsdale, NJ: Lawrence Erlbaum. Jackendoff. R. ( 1975). Morphological and semantic regularities in the lexicon. Languuge. 51(3), 639-67 1. Jackson, J.H. (1958). Selected writings of John Hughlings Jackson. New-York, NY: Basic Books. Jakobson, R. (1956). Two aspects of language and two types of aphasic disturbances. In, R. Jakobson and M. Halle (Eds.), Fundamentals of language. (pp. 55-82). The Hague: Mouton. Jakobson. R. (1964). Toward luiguistic typology of aphasic irnpairments. In, A.V.S. De Reuck and M. O'Connor (Eds.), Disorders of language, (pp. 21-46). London: Churchiil. larema, G. Br Fnederici, A.D. (1994). Processing articles and pronouns in a g r m a t i c aphasia: Evidence from French. Braui and Lmguoge, 46.683-694. Jarema, G. Br Kadzielawa, D. (1990). Agrammatism in Polish: A case study. In, L. Menn and L.K. Obler (Eds.). Agrammatic aphasia. A cross-language narrative sourcebook, (pp. 8 17-893). Amsterdam: John Benjamins. larema, G. & Kehayia, E. (1992). Impairment of inflectional morphology and lexical storage. Brain and Laquage, 43, 541-564.

Jarema, G., Kehayia, E., Mimouni, Z., & Kadzielawa, D. (1991). Interaction between rnorphology and syntax in thematic rote assignment in agrammatism: A crosslinguistic study. Paper presented at the Academy of Aphasia 29th Annual Meeting. Rome, Italy. Kean. M.L. ( 1977). The linguistic interpretatioa of aphasic syndromes: Agrammatisrn in Broca's aphasia, an example. Cognition,5, 9-46. Kehayia, E. (1990). Morphological deficits in agrammatic aphasia: A comparative linguistic study. Unpubiished doctoral dissertation. Montréal: McGili University. Kehayia. E. (1993). Morphological priming of derived and inflected words: A neurolinguistic and psycholinguistic study. Paper presented at the Academy of Aphasia 31st Annual Meeting, Tucson, AZ. Kehayia, E., Caplan, D., & Piggot, G.L. (1984). On the repetition of derivational *xes by English agrammatics. McGill Working Papers in Linguistics, 2 (1). Kehayia, E. & Jarema, G. (1994). Morphologicai priming (or prVn#ing?) of inflected verbs forms: A comparative study. Journal of Neurolinguistics, 8(2),83-94. Kehayia, E., Jarema, G., & Kadzielawa, D. (1990). A cross-linguistic study of morphological errors in aphasia: Evidence from English, Greek and Polish. In, J.-L. Nespoulous and P. Villiard (Eds.), Morphology, phonology, and aphasia, (pp- 140- 155). New York, NY:Springer-Verlag. Kempley. S.T.& Morton, 1. (1982). The effects of priming with regularly and irregularly related words in auditory word recognition. British Journal of Psychology, 73, 441-454. Kilani-Schoch, M . ( l988a). Introduction à la morphologie naturelle. Berne: Éditions Peter Lang. Kilani-Schoch, M. (1988b). Discontinuité ou continuité de la base morphologique en arabe classique et en arabe tunisien? Zeitschriftfur arabische linguistik, 8 1-92. Kussmaul, A. (1878). In, R.J. Tissot, G. Mounin, and F. Lhermitte (1973). L'agrammathw. Bruxelles: Dessart. Lapointe, S.G. (1983). Some issues in the linguistic description of agrammatism. Cognition, 14, 1-39. Laudanna, A., Badecker, W.. & Caramma, A. (1989). Priming homographic stems. Journal of Mernury and Lunguage, 28,53 1-546. Laudanna, A. & Burani, C. (1985). Address mechanisrns to decomposed lexical entries. Linguistics, 23, 775-792.

Lecoun, A.R. & Lhermitte, F.( 1979). L'aphasie. Paris: Rammarion Médecine-Sciences. Chapitres 4 et 8. Linebarger. MC.,Schwartz. M.F.. & Saffran, E.M. (1983). Sensitivity to grammatical structure in so-cded a g m a t i c aphasies. Cognition. 13, 36 1-392. Lukatela, G.. Gligorijevic, B.. Kostic, A.. & Turvey, M.T. (1980). Representation of inflected nouns in the internai lexicon. Memory and Cognition. 8 (3,415-423. Lukatela, G.. Mandic. 2.. Gligorijevic, B., Kostic, A., Savic, M., & Tuwey, M.T. ( 1978). Lexical decision for inflected nous. Lrmguage and Speech. 21. 166- 173. Luria, A R . ( 1947). Traumatic Aphasia. (Reprinted in translation, Mouton, The Hague, 1970). MacKay, DG. (1978). Denvational rules and the internai lexicon. Journal of Verbal Leaming and Verbal Behovior, 17.6 1-7 1. MacWhinney , B. ( 1978). The acquisition of morphophonology. Monographs of the Society for research in child development. Serial no. 174,43(1-2). MagnUsd6tt.h. S. & Thrainsson, H. (1990). Agrammatism in Icelandic: Two case studies. In, L. Menn and L.K. Obler (Eds.), Agrammatic aphasia. A cross-langunge narrative sourcebook. (pp. 443-544). Amsterdam: John Benjamins. Manelis, L. & Tharp, D.A. (1977). The processing of affixed words. Memory & Cognition, 5. 690-695. Matthew, H. & Kehayia, E. (1994). Sensitivity to argument structure in agramrnatism. Paper presented at the Academy of Aphasia 32ndAnnunl Meeting, Boston, MA. McCarthy, J.J. ( 1979). Fomal problems in semitic phonology and rnorphology. Unpubiished doctoral dissertation. Cambridge. MA: MIT. McCarthy, J. J. ( 1981). A prosodic theory of nonconcatenative morphology. Linguistic Inquiry, 12(3). 373418. McCarthy, J.J. (1983). A prosodic account of Arabic broken plurals. la, 1. Dihoff (Ed.), Current trends in African linguistics, I . (pp. 289-320). Fons: Dordrecht. McCarthy, J.J. & Prince, A.S. (1990). Foot and word in prosodic morphology: The Arabic broken plural. Naturai b g t i a g e and Linguistic Theory, 8, 209-283. Mem. L. & Obler. L.K. (Eds.) ( 1990). Agrammatic aphmia. A cross-lunguage narrative sourcebook. Amsterdam: John Benjamins. Miceii, G. & Caramazza, A. (1988). Dissociation of inflectiond and derivational morphology . Bruin and Language, 35,24-65.

Miceli, G. & Mazzucchi, A. (1990). Agrammatism in Italian: Two case studies. In. L. Menn and L.K. Obler (Eds.). Agrammatic aphasia. A cross-language narrative sourcebook, ( p p . 7 17-8 16). Amsterdam: John Benjamins. Mirnouni, 2.& Jarema, G. (in press). Agramrnatic aphasia in Arabic. Aphasiology. Mimouni. 2.. Kehayia, E., Br Jarema, G. (1992). Morphological priming of inflected nouns in a nonconcatenative language: Implications for lexical storage and access. Congress of the Academy of Aphasia 30th Annual Meeting. Toronto. Ontario. Murell. G.A. & Morton, J. (1974). Word recognition and morphemic structure. Journal of Ekperimental Psychology. 102,963-968. Nespoulous, J.-L., Dordain. M., Perron, C.. Jarema, G., & Chazd, M. (1990). Agrammatism in French: Two case studies. In, L. M e m and L.K. Obler (Eds.), Agrammatic aphasia. A cross-language narrative sourcebook, ( p p . 623-716). Amsterdam: John Benjamins. Ouhalla, J. ( 1991). Functional categories and parametric variation. New York, N Y : Routledge. Paradis, M. ( 1983). ReadVlgs on aphasiu in bilingualr and polyglots. Quekc: Didier. Paradis. M. (199 1). Maghrebian version of the Blingual Aphasia Test. Mahwah: Lawrence Erlbaun Pick, A. ( 1913). In, M. Paradis ( 1983). Readings on apharia in bilinguals and po lyglots. Que'bec: Didier. Pinker, S. ( 1991). Rules of language. Science, 253, 5 30-535. Pitres. A. (1898). In, M. Paradis (1983). Readings on aphasia in bilinguals and polyglots. Que'bec: Didier. Prather, P., Zurif, E.. Stem. C.. & Rosen. T.J.(1992). Slowed lexical access in nonfluent aphasia: A case snidy. Brain and Language, 43,336-348. Queneau. R. (1948). Le temps passé. In, L'htant Fatal, Paris. Rosenthal, V . & Goldblum, M.L. (1989). On certain grammatical prerequisites for agrammatic behaviour in comprehension. JounuzZ of Neurolinguistics, 4, 179-211. Saffran. E.M.,Schwartz. M.F.,& Marin, O.S.M.(1980). The word order problem in agrammatism: II. Production. Brain and Language, 10,263-280. Schwartz, M.F., Saffran. E.M..& Marin,O.S.M.( 1980). The word order problern in agrammatism. 1. Comprehension. Brain and h g u a g e , 10, 249-262. Segui, J. & Zubizarreta, M.-L. (1985). Mental representation of morphological complex words and lexical access. Linguistics, 23, 759-774.

Sells, P. ( 1985). Lectures on conternporary syntactic Theories: An introduction to governrnent-binding theory. generali~edphrase structure gramrnar. and lexicalfincrional grammar. Stanford, CA: CSLI. Snodgrass. J.G. & larvella. R.J. (1972). S o m linguistic determinants of word classification tirnes. Psychonomic Science, 27,220-222. Spencer. A. ( 1991). Morphological theory. Oxford, England: Basil Blackwell. Sperber, D. & Wilson. D. (1986). Relevance. Communication and cognition. Oxford, England: Basil Blackweil. Stanners, R.F., Neiser, J J., Hernon, W.P., & Hall, R. (1979). Memory representation for morphologically related words. Journal of Verbal Leaming and Verbal Behavior. 18, 399-412. Stemberger, J.P. & MacWhinney, B. (1986). Frequency and the lexical storage of regularfy inflected fonns. Memory di Cognition, 14(1l), 17-26. Stemberger, J.P.& MacWhinney, B. (1988). Are inflected forms stored in the lexicon ? In, M. Hammond and M. Noonan (Eds.), Theoretical morpholagy: Approaches in modern linguistics, (pp. 101-116). San Diego, CA: Academic Press. Tafi, M. (1979). Recognition of afixed words and the word frequency effect. Memory & Cognition, 7, 263-272. Taft.M. & Forster, K. ( 1975). Lexical storage and retrieval of prefixed words. Journal of Verbal Learning and Verbal Behavior, 14, 635-647. Taft, M. & Forster. K. (1976) Lexical storage and retrieval of polymoiphemic and polysyllabic words. Journal of Verbal Leuming and Verbal B e h i o r , 15, 607-620. Tissot, R.I., Mounin, G. & Lhennitte. F. (1973). L'agrammatisme. Bruxelles: Dessart. Tyler, L.K. (1992). Spoken language comprehension: An experirnental approach to disordered and mmral processing. Cambridge, MA: ha Press. Tyler. L.K.,Osuin, R.K..Cooke, M., & Moss, H.E. (1995). Automatic access of lexical information in Broca's aphasics : Agauist the automaticity hypothesis. Brain and Languuge, 48, 13 1- 162. Zhang. B.Y. & Peng, D.L. (1992). Decomposed storage in the Chinese lexicon. In, H.C. Chen & O.J.L. Tzeng (Eds.),Language processing in Chinese, (pp. 131- 149). London: Elsevier. Zurif, E.B. & Caramaua. A. (1976). Psycholinguistic structures in aphasia: Studies in syntax and semantics. In, H. Whitaker and K.A. Whitaker (Eds.).Studies in neurolhguistics. Vol. I . (pp. 261-292). New York, NY: Academic Press.

APPENDIX I

LIST OF BOUND STEMS (Study One)

Target

Stem

Examples of occurrences of same consonant clusters

t aaat

'she followed'

ta&

wlx

“P-'

î2aat

'she hugged' 'she hit'

Tanp

Tua

'neck'

rien

S

'tenasse*

qa&! ~alall

halu

'w*

m&l

'regret* 'shart' 'season'

nena t

m

aiat %agnat

'she greeted'

fahat Yagat

'she understood' 'she washed'

fah Yaa

sah fasl

2aMat

'she p u U d 'she held'

2aM

w2d

'human*

wd

'she pushed'

9aL?d damr

xam

'human' 'wine'

'she drank'

s'as

YarQ

'we~t'

'she counted*

ha& f'aip

va?

'rush'

sa+

'prison'

gaMat damrat SaBat haspat

W&at

'she worked'

'she lcnead&*

SENTENCES USED IN THE TASES OF REPETITION, READING ALOUD AND ORAL COMPREKENSION (Study One)

The boy hits the girl

The boy kisses the man The girl follows the mother

The girl hits the boy The girl hugs the woman

The man holds the old man The man kisses the boy

The man surprises the woman

The mother follows the girl The old man holds the man

The woman hugs the girl The woman surprises the man

VSO The boy hits the girl

The boy kisses the man The girl follows the mother

The girl hits the boy The girl hugs the woman The man holds the old man

The man kisses the boy The man surprises the woman The mother follows the girl

The old =an holds the man The woman hugs the girl

The woman surprises the man

OAS JP13

The boy is hit by the girl The boy is kissed by the man

The girl is followed by the mother

The girl is hit by the boy

The girl is hugged by the woman The man is held by the old man The man is surprised by the woman The mother is foilowed by the girl

The old man is held by the man The woman is hugged by the girl The woman is surprised by the man

The man is kissed by the boy

It is the boy who is kissed by the

man It is the girl who is hit by the boy It is the girl who is hugged by the woman It is the mother who is followed by the girl It is the girl who is foliowed by the rnother It is the man who is kissed by the boy It is the man who is surprised by the woman It is the woman who is surprised by the man It is the man who is held by the old man It is the old man who is held by the man It is the woman who is hugged by the girl It is the boy who is hit by the girl

SENTENCES USED IN REPETITION AND READING ALOUD TASKS (Study One)

The father does it(m) The mother washes it/him

The boy holds it/him

The girl pushes it/him He does it(f) She washes itfher He holds it/her She pushes it/her

The father explains to him The rnother kneads it(m) The boy drinks it(m)

The girl counts it(m) He explains to ber She kneads it(f)

He di& it(f) She counts it(f)

PICTURES USED IN ORAL COMPREHENSION (Study One)

APPENDZX V

CUMULATIVE RIESULTS OF PRIMING EXPERIMENT (Study two)

Subject cumulative results for critical stimuli Plural-Singular pairs Monosyl-BP

Bisyl.BP

(imel-imal) (fna2al-fanZal)

Bi-plural Bi-plural (siar-sa2ra) (sairat-saira)

Sound Plural (Tbi bat-Tbibe)

Subject cumulative resuits for critical stimuli

Singular-Plural pairs Monosyl. BP cimal-Srnal)

Bisyl. BP (fanfal-fnaial)

Bi-plural (sazra-siar)

Bi-plural

Sound Plurai

(saira-saint) (Tbiba-Tbi bat)

Subject cumulative results for control stimuli

Plural-Plural pairs Monosyt. BP (qdam-imal)

Bisyl. BP Imf atah-fnaial)

Bi-plural (Ylafat-lbasat)

Subject cumulative results for control stimuli

Singular-Plural pairs Monosyl. BP (qdam-Zmal)

Bisyl. BP (maf tah-fanial)

Bi-plurai (Ylaf -1bas)

WORD PLURAL

- WORD

- SINGULAR

- -

MONOSYLLABIC (BP) (N=12)

CONTROL SUBJECTS

BISYLLABIC (BP) (N= 12)

-

* Control singular prime

PATIENT

PATIENT

SI

S2

CONTROL SUBJECTS

PATIENT

PAIENT

(~=24)

S1

(~=24)

S2

1142

987

1 145

1029

11 12

1016

1120

980

- WORD -

WORD SINGULAR

MONOSYLLABIC (BP) (N= 12)

BISYLLABIC (BP) (N=12)

* Control plurai prime

- PLURAL

CONTROL SUBIECTS

(~=24)

CONTROL SUBJECTS

(N=24)

1 PATIENT S1

PATIENT

52

I

WORD

BI-PLURAL

PATIENT

(N=12)

S1

J

I

- WORD

Target sometimes rejected as a non word

PATIENT

S2

1

I

SOUND PLURAL (N=12)

Tbibat

+

Tbiba

xadmu

+

Tbiba

qehw8

+

Tbiba

WORD-

WORD

CONTROL SUBECTS

(N=24)

1

CONTROL SUBJECTS

(N=24) Tmel

Tmal

Tmal Tm81

Ynaial Yanfal

Yanial YneZal

Ybibet

mibut

Ybibe

Tbibat

Ybiba

Tbiba

Ybfbst

Tbiba

PATIENT

S1

PATIENT

S2

189 l

OSL

HL

P88

008 LS8

66L 08L OLL

(

WORD

BI-PLURAL (N=I 2)

COMXOL SUBJECTS

(N=24)

wald

wled

- ILLEGALLY-SUFFIXED WORD 1 PATIENT S2

PATIENT S1

*walda

+

*walda

* Target sometimes accepted as a word.

1

ILLEGALLY-SUFFIXED WORD

- WORD

CONTROL SUBJECTS

(N=24)

1

IMAGE EVALUATION TEST TARGET (QA-3)

-------

A P P L I E i I L M A G E . Inc = 1653 East Main Street -. Rochester. NY 14609 USA Phone: 716/482-0300 fax: 71W268-5989