UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The use of cohesive devices by school-age language impaired children Ruthven, Lucille Katherine 1989

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1989_A6_7 R87.pdf [ 10.74MB ]
Metadata
JSON: 831-1.0097565.json
JSON-LD: 831-1.0097565-ld.json
RDF/XML (Pretty): 831-1.0097565-rdf.xml
RDF/JSON: 831-1.0097565-rdf.json
Turtle: 831-1.0097565-turtle.txt
N-Triples: 831-1.0097565-rdf-ntriples.txt
Original Record: 831-1.0097565-source.json
Full Text
831-1.0097565-fulltext.txt
Citation
831-1.0097565.ris

Full Text

THE USE OF COHESIVE DEVICES BY SCHOOL-AGE LANGUAGE IMPAIRED CHILDREN By LUCILLE KATHERINE RUTHVEN B.A.(Hons.), The University of Winnipeg, 1983 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE i n THE FACULTY OF GRADUATE STUDIES THE FACULTY OF MEDICINE THE SCHOOL OF AUDIOLOGY AND SPEECH SCIENCES We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA August 1989 ( c ) L u c i l l e Katherine Ruthven, 1989 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. S c h o o l o f A u d i o i o g y and Speech S c i e n c e s The University of British Columbia Vancouver, Canada DE-6 (2/88) A b s t r a c t F r e q u e n c y and d i s t r i b u t i o n o f c o h e s i v e d e v i c e s u s e d i n t h e n a r r a t i v e d i s c o u r s e o f s i x l a n g u a g e i m p a i r e d and s i x n o r m a l l a n g u a g e c h i l d r e n aged 9;1 t o 11 ;8 were i n v e s t i g a t e d i n o r d e r t o g a i n a g r e a t e r u n d e r s t a n d i n g o f t h e d i s c o u r s e d i f f i c u l t i e s e x p e r i e n c e d by l a n g u a g e i m p a i r e d c h i l d r e n . R e f e r e n c e , s u b s t i t u t i o n / e l l i p s i s , c o n j u n c t i o n , and l e x i c a l c o h e s i o n were a n a l y z e d u s i n g an a d a p t a t i o n o f H a l l i d a y & H a s a n ' s (1976) d e s c r i p t i o n o f c o h e s i o n i n E n g l i s h . D i s c o u r s e r e f e r e n c e was f u r t h e r i n v e s t i g a t e d u s i n g 1) r e t r i e v a l a n a l y s i s ( M a r t i n 1977 , 1 9 8 3 ) , 2) f o r m / f u n c t i o n a n a l y s i s (Bamberg 1987) and 3) s e n t e n c e p o s i t i o n a n a l y s i s . The d a t a a n a l y z e d f o r t h e p r e s e n t r e p o r t c o n s i s t e d o f t h r e e s t o r i e s t o l d by e a c h c h i l d (a t o t a l o f t h i r t y - s i x s t o r i e s ) , w h i c h were o b t a i n e d on t h e f o l l o w i n g t a s k s : 1) p i c t u r e book n a r r a t i o n , 2) r e t e l l o f t h e same s t o r y w i t h o u t p i c t u r e s , and 3) v i d e o r e c a l l . L a n g u a g e i m p a i r e d c h i l d r e n u s e d s i g n i f i c a n t l y f e w e r c o h e s i v e d e v i c e s p e r u t t e r a n c e t h a n n o r m a l l a n g u a g e c h i l d r e n , a f i n d i n g t h a t a p p e a r e d t o be r e l a t e d t o s h o r t e r u t t e r a n c e l e n g t h on t h e p a r t o f l a n g u a g e i m p a i r e d c h i l d r e n , f e w e r a t t e m p t s on t h e i r p a r t t o p r o v i d e c o h e s i v e l i n k s ( p a r t i c u l a r l y l e x i c a l o n e s ) , and a g r e a t e r number o f f a i l e d c o h e s i o n a t t e m p t s . R e t r i e v a l a n a l y s i s r e v e a l e d a g r e a t e r u s e o f a n a p h o r i c and e s p h o r i c r e f e r e n c e ( f o r w a r d r e f e r e n c e w i t h i n t h e n o m i n a l g r o u p ) by n o r m a l l a n g u a g e c h i l d r e n t h a n by l a n g u a g e i m p a i r e d c h i l d r e n . R e s u l t s o f t h e f o r m / f u n c t i o n a n a l y s i s i n d i c a t e d t h a t t h e l a n g u a g e i m p a i r e d c h i l d r e n u s e d more d e f i n i t e ( v e r s u s i n d e f i n i t e ) f o r m s t o i n t r o d u c e p a r t i c i p a n t s , more p r o n o m i n a l s t o s w i t c h r e f e r e n c e , and more n o m i n a l s t o m a i n t a i n r e f e r e n c e t h a n t h e n o r m a l l a n g u a g e c h i l d r e n . The r e f e r e n t i a l d e v i c e s o f b o t h g r o u p s o f c h i l d r e n v a r i e d as a f u n c t i o n o f t h e p r o t a g o n i s t ; b o t h i i groups showed evidence of a thematic subject strategy. Task differences tended to hold for one group or the other, suggesting that the two groups were responding differently to changes in contextual configuration. The findings suggest that the difficulty faced by language impaired children in the realm of discourse may not only be the general one of discovering a systematic relationship between discourse-sensitive forms and their functions, but also that of discovering appropriate form-function pairings specific to a given context. The results underline the importance of discourse analysis to clinical assessment and intervention. TABLE OF CONTENTS ABSTRACT i i LIST OF TABLES vi LIST OF FIGURES v i i i ACKNOWLEDGMENTS ix Chapter 1. THE DISCOURSE DISORDERS OF LANGUAGE IMPAIRED CHILDREN 1 2. METHOD AND PRELIMINARY DESCRIPTION OF THE DATA 10 Method Subjects Data Collection Statistical Procedures Preliminary Description Results Discussion 3. COHESION: A REVIEW OF THE LITERATURE 28 Theoretical Framework Systemic Grammar The Nature of Text Cohesion Related Studies The Development of Cohesion Cohesion and Language Impaired Populations Purpose of the Present Study 4. COHESION: ANALYSIS 74 Method of Analysis Results Amount of Cohesion Distribution Discussion iv 5. REFERENCE: RETRIEVAL ANALYSIS 109 Introduction Reference Network Previous Findings Purpose of the Present Analysis Results Discussion 6. REFERENCE: FORM/FUNCTION ANALYSIS 128 Discourse Reference Theoretical Background The Development of Referential Devices Reference and Language Impaired Populations Results Discussion 7. CONCLUSIONS 179 BIBLIOGRAPHY 192 Appendix 1. SUMMARY OF STORY CONTENT 198 2. SAMPLE NARRATIVES 199 3. MEAN PROPORTIONS OF COHESION CATEGORIES 206 4. CONJUNCTIVE ITEMS USED BY LI AND NL CHILDREN 207 5. PHORIC NOMINAL GROUPS INVOLVING BRIDGING 208 6. CHECKLIST OF COHESIVE DEVICES 209 v LIST OF TABLES Table I. II. III. IV. V. VI. VII. VIII. IX. XI. XII. XIII. Page Summary of Subject Characteristics 12 Mean Frequency of Words, Independent Clauses, and Words per Independent Clause across Tasks (T1,T2,T3) and Lan-guage Groups (LI and NL) 19 Summary of Halliday & Hasan's (1976) Inventory of Cohesive Devices 39 Examples of Cohesion Categories from the Data 76 Frequency of Cohesive Ties Used by LI and NL Children on TI, T2, and T3 78 Proportions of Referential Ties falling into the Subcate-gories of Personals, Demonstratives, and Comparatives for LI and NL children on TI, T2, and T3 81 Proportions of Additive versus Nonadditive Conjunctive Ties for LI and NL children on TI, T2, and T3 85 Frequency (and Proportion) of Conjunctive Ties by Individual Subjects on T2 86 Mean Proportions of Lexical Cohesion Subtypes (out of Total Lexical Ties) for LI and NL Children on TI, T2, and T3 88 Frequency (per Independent Clause) of Lexical Ties and of Two Lexical Subtypes (Repetition with Identity and Verbal) for LI and NL Children on TI, T2, and T3 89 Subtypes of Incomplete Devices used by LI and NL Children on TI, T2, and T3: Frequency, Percentage of Utterances containing each Subtype, and Ratio of these Percentages (LI/NL) 91 Mean Percentages of Cohesion Categories in Narratives: Results of the Present Study Compared to the Adults of Rochester & Martin's Study (1979) 95 Frequency of Phoric Nominal Groups per Independent Clause for LI and NL Children on TI, T2, and T3 117 vi XIV. Mean Proportions of Retrieval Categories for LI and NL Children on Tl, T2, and T3 118 XV. Number of Each Type of Introductory Device (First Mention) Used by LI and NL Children on Tl, T2, and T3 153 XVI. Percentage (and Absolute Number) of Nominal and Pronominal Devices used by LI and NL Children to Switch versus Maintain Reference to the Passengers and to the Ape: (a) on Tl and (b) on T2 155 XVII. Percentage (and Absolute Number) of Nominal versus Pronomi-nal Devices used by LI and NL Children to Svitch Reference to the Passengers and to the Ape: (a) on Tl and (b) on T2 158 XVIII. Percentage (and Absolute Number) of Nominal versus Pronominal Devices used by LI and NL Children to Maintain Reference to the Passengers and to the Ape: (a) on Tl and (b) on T2 158 XIX. Percentage (and Absolute Number) of Nominal and Pronominal Devices in Subject versus Nonsubject Position as used by LI and NL Children to Refer to the Passengers, Ape, and Minor Characters: (a) on Tl and (b) on T2 160 XX. Distribution of Focus in Subject Position: Percentage (and Absolute Number) of Referential Devices in Subject Position that Refer to the Passengers, Ape, or Minor Characters as Used by LI and NL Children: (a) on Tl and (b) on T2 162 XXI. Distribution of Focus in Nonsubject Position: Percentage (and Absolute Number) of Referential Devices in Non-subject Position that refer to the Passengers, Ape, or Minor Characters as Used by LI and NL Children: (a) on Tl and (b) on T2 163 vi i LIST OF FIGURES Figure 1. Place of Cohesion within the Semantic Stratum of Systemic Grammar 30 2. Distribution of Cohesion Categories for LI and NL Children on TI, T2, and T3 80 3. Frequency of Conjunctive Ties per Independent Clause by LI and NL Children on TI, T2, and T3: a) a l l Conjunctions (Intra-utterance and Inter-utterance), b) Intra-utterance Conjunctions ; c) Inter-utterance Conjunctions 83 4. Reference Network i n English 110 v i i i ACKNOWLEDGMENTS Grateful acknowledgment is extended to my thesis supervisor, Dr. Carolyn Johnson, for her valuable comments at each stage of the research, for the many hours she spent examining the text and providing constructive feedback, and for being a constant source of encouragement. Thanks are also extended to Dr. John Gilbert for his helpful suggestions and for reviewing the text. I also wish to express thanks to the many members of the Richmond School District who made collection of the data possible: the administration for granting me permission to obtain subjects from their schools, the speech/ language pathologists for recommending children for the study, the principals for providing quiet rooms and video equipment, and the teachers for recommen-ding control subjects and for allowing me to collect data during school time. Special thanks are also due to the parents who kindly consented to the inclusion of their child in the study, and to the children themselves for their enthusiastic participation and delightful stories. I am also grateful to Lisa Kam from the Department of Statistics for her suggestions, and to Margaret Lahey for her permission to reproduce a cohesion checklist which appeared in Language Disorders and Language Development (MacMillan Publishing Co.,1988). Finally, sincere thanks are extended to friends and family members for their encouragement. In particular, I would like to thank my husband, Lloyd, for the many hours he spent instructing me in computerized word processing, for his patience with my ignorance of such matters, and for the financial and moral support which enabled me to concentrate on this research. ix CHAPTER ONE INTRODUCTION THE DISCOURSE DISORDERS OF LANGUAGE IMPAIRED CHILDREN The recent application of functional grammars to the study of language acquisition and language impairment has resulted in a shift in method from the analysis of single sentences (and of their constituents) to examinations of continuous stretches of language and contextual variables. This development is particularly significant in the domain of education where considerable value is attached to the production of coherent discourse, and conversely, where the production of incoherent discourse is stigmatized. Teachers are not alone in their demand for coherence; a child's classmates may be just as exacting: "those who interrupt, who f a i l to give feedback, or whose remarks are ambiguous or non sequiturs are not likely to be sought-after conversa-tional partners" (Donahue 1985: 97). Unfortunately, there is no concensus among either educators or linguists on what makes for coherence. According to Brown & Yule, "human beings do not require formal textual markers before they are prepared to interpret a text. They naturally assume coherence, and interpret the text in the light of that assumption" (1983: 66). For these authors, accounting for coherence means describing those things in the listener (or reader) which enable him or her to arrive at the speaker (or writer's) intended meaning, things such as the listener's ability to compute communicative function (speech act), to use general socio-cultural knowledge, and to make inferences. Although insight-1 2 ful, Brown & Yule's interpretation leaves two questions unanswered: why do human beings "naturally assume coherence" and why must this strong assumption sometimes be abandoned? The authors support their conclusions with examples of coherent discourse in which there are no formal connections between utterances. However, the majority of texts contain an abundance of these markers. Within a functionalist perspective, the purpose served by such markers requires explanation: The assumption of coherence can be sustained so well because human language has the resource for indicating coherence, while the nature of language as a resource has developed in a particular way because i t has had to serve the needs of the community. Our task is to understand the specific nature of these resources - not simply to hide behind the mind and the intention of particular speakers and listeners. (Halliday & Hasan 1985: 96) Accounting for coherence with reference solely to listener processes is possible only when the normal expectation holds true, i.e. when the producer is speaking coherently. Incoherent discourse that leads to the abandonment of this assumption may be due to the speaker's inability to organise relevant meanings in logical relation to each other and/or to linguistically mark this organization formally. The onus for coherence in such cases does not uniquely rest on the listener. Educators and language clinicians must have some means of evaluating those speakers whose discourse is consistently perceived as "incoherent". The purpose of the present study is to examine some of the linguistic devices available to a speaker that contribute to the production of coherent discourse. Accurate evaluation requires that the formal properties of successful discourse be known and specified. What is increasingly apparent, however, is that "talk" is complex. Part of the "complexity" is due to the current state of research on discourse disorders. As McTear (1985) noted, confusion over 3 terminology, for example, reflects a general lack of a coherent theoretical framework. Much of the complexity is due to the nature of talk itself. Johnston described the act of conversation, for example, as "the most complex of human constructions": Participants assess, secure, and maintain each other's attention; they take and signal turns; they collaborate in establishing mutual reference and in spinning the topical thread; and they monitor the health of the conversation, asking for clarification or offering confirmation as needed. (1985: 83) As research reveals the variety of skills required in the normal speaker, the existence of a number of different discourse symptoms and underlying disorders also becomes plausible. According to Fey & Leonard, "the population of SLI [specific language impairment] children is no more homogeneous with respect to pragmatic skills than i t is with respect to semantic, syntactic, or phono-logical abilities" (1983: 77). A number of explanations have been proposed to account for discourse problems in children (see Johnston 1985). A socio-affective disturbance appears to underlie restricted use of illocutionary force (communicative function served by an utterance). Coherence in discourse requires that one speech act logically follow upon another. In the normal child, basic communi-cative functions (e.g. establishing mutual attention and directing action) emerge early and are rapidly expanded upon. This richness of illocutionary force is in striking contrast to the discourse of some children with affective disturbances. Illocutionary force in autism, for example, has been found to be restricted primarily to directing listener actions; many autistic children f a i l to use language to inform or comment. Discourse symptoms have also been attributed to a general lack of social knowledge as a result of limited conversational experience. In the studies of 4 Donahue, Pearl & Bryan (1980) and Donahue (1981), learning disabled children were less likely to request clarification of ambiguous utterances than age-matched normal language peers. The requests of the learning disabled subjects indicated that the linguistic skills required to request clarification were within their production ability. The investigators concluded that the learning disabled children demonstrated "social naivete" in that they were reluctant to question the adequacy of information provided by an experimenter whom they believed to be trustworthy. The authors suggested that the social rejection experienced by such children results in fewer opportunities in which they might practice conversational skills. The finding that the learning disabled subjects used their most polite language with their "best friends" (Donahue, 1981) led the investigator to conclude that such children show deficits in their understanding of social relationships. However, several investigators have questioned the notion that discourse disorders arise from a lack of social knowledge. Johnston, for example, interpreted the politeness of the learning disabled subjects of Donahue's study (1981) as indicating that they understood their social world all too well (Johnston 1985: 89). Gallagher & Darnton (1978) found that language disabled children were sensitive to communication breakdown and that, like children with normal language, they responded by revising their utterances. An analysis of their revisions, however, indicated that language impaired children were less likely to paraphrase or use synonyms than MLU-matched younger children. These findings suggest that there may be a formal linguis-tic basis for discourse problems. Gale, Liebergott, & Griffin (1981)1 examined requests for clarification and found that not only did language *As reported by Friel-Patti & Conti-Ramsden (1984: 171). 5 impaired children use more nonverbal requests than language-matched normals, but their verbal requests demonstrated the use of a restricted number of linguistic forms. Donahue, Pearl & Bryan (1982) examined the syntactic ability of language impaired children during a task in which the subjects were required to convey information to a listener. The results led the investiga-tors—the same researchers who had previously proposed a social deficit explanation—to conclude that productive language deficits may significantly interfere with even informal conversation. J In the above studies, a linguistic hypothesis was favored over a social one since the language disabled children in each case appeared to be socially on par with younger MLU-matched children. One may question, however, whether language disabled children who only demonstrate the conversational skills of younger children are, in fact, socially "normal". In a study by Fey & Leonard (1982), the conversations of language disabled 6-year-olds were evaluated for adjustments to the needs of various partners (age peer, adult, toddler) and the results compared to adjustments made by both MLU- and age-matched con-trols. The language disabled children performed like their age peers in discourse adjustments involving illocutionary force and requests for repair but unlike their age peers in that they did not adjust the linguistic com-plexity of their utterances when speaking to toddlers. These results conflict with those of Donahue et al. (1980) and favor a linguistic (as opposed to social) explanation for discourse disorders in language disabled children. Results supporting a similar conclusion were obtained in a study by Leonard (1986) which investigated the conversational replies of children with specific language impairment (aged 2;10-3;6) and of those developing language normally (aged 1;5-1;11). Language impaired children replied more often to 6 adult initiations and used a greater variety of conversational replies. They were more likely to answer questions and less likely to imitate when asked a question than the normal language children. They showed a greater percentage of nonobligatory conversational replies, such as affirming and expanding, and were more likely to deny or to provide alternatives to information contained in adult utterances than the controls. Leonard concluded that "language impaired children can serve as responsive conversationalists when the conver-sational behaviors under investigation do not require considerable syntactic s k i l l " (1986: 117). He attributed the superior performance of the language impaired group to greater world knowledge and conversational experience, and to better comprehension abilities. His study highlights the importance of the method used to select controls to the conclusions that are drawn. The nature of the formal deficit implied in many of the above studies is unclear. Does it consist uniquely of a difficulty with syntax, i.e. with sentence-level constraints? Johnston (1985) suggested that there may be children who have special difficulty with rules incorporating discourse categories, a difficulty which she refers to as a true pragmatic disorder. Such children err in their use of certain linguistic forms that are more highly context-sensitive than others (e.g.topicalization and reference). Since the functions of certain forms at the sentence level are not identical to the functions served by those forms at discourse level, the problem space of discourse may indeed be qualitatively different from that of the sentence. A number of explanations have been proposed to account for discourse problems in children: an affective disturbance, lack of social knowledge, a formal linguistic deficit, and what Johnston calls a true pragmatic disorder. It is evident that in any of these cases, assessments restricted to units 7 expressed in terms of individual sounds, words or intrasentential relations are insufficient to identify and describe the nature of the "talk" of language impaired children. They may, in fact, be misleading; the selection of a particular form in any given sentence may have been governed by choices made in previous sentences. The general purpose of the present study is to examine these kinds of forms, i.e. those that appear to be dependent on lexicogram-matical selections made in previous utterances. According to past reports, the language impaired children included in the present study a l l had problems with form at the sentence level sometime in the past. The questions that were asked were whether these children, whose individual sentences appeared grammatically correct 2, would demonstrate difficulties at the discourse level, and if so, what the nature of these difficulties would be. Although both narrative and conversational samples were collected, only the results from the narrative data will be reported here. Narratives, in fact, appear to be the preferred genre of investigations into the discourse of school-age children (see Johnston 1982, Westby 1984, Scott 1988, and Lahey 1988 for reviews), with the majority of investigators adopting a story grammar approach (Hansen 1978, Weaver & Dickenson 1979, Graybeal 1981, Feagans & Short 1984, Sleight & Prinz 1985, Roth & Spekman 1986, Merritt & Liles 1987, and Ripich & Griffith 1988). The results of these studies suggest that the narratives of school-age language impaired children are structurally very similar to those of normal language children (from a story grammar standpoint) with the only agreed upon difference being that they are shorter in length. Despite their failure to shed light on differences between the two 2A detailed examination of sentential construction was not done and this is recognized as a limitation. There were few agrammatical sentences in the narrative data. 8 populations, story grammar analyses continue to dominate research in this area. Their continued use has been criticized on other grounds. According to Brown & Yule (1983), decisions regarding the content of the texts that story grammarians analyze are arbitrary and subjective: "The illusion that the decisions are non-subjective is mainly fostered by the extreme simplicity of the texts investigated. The texts are so constructed so as to be context-neutral, free of potential ambiguity and composed of mainly non-complex sentences" (pp.120-21). The same simple stories are used repeatedly in these studies. If an attempt is made to use more elaborate or spontaneously-generated stories, numerous problems arise in assigning propositions to categories. There is perhaps no reason to expect a child's anecdotes to conform to traditional story grammars of folk-tales. Pappas (1985) criticized story grammar approaches for overemphasizing form isolated from content, and for attending mainly to events and not to participant development. The use of alternative methods of analysis is overdue. A procedure is needed that will take account of the actual nature of stories as produced by children, that can be used for many different kinds of discourse samples, that will go beyond form to permit an evaluation of a child's ability to organize relevant meanings in relation to each other, and that will assess participant development. Text cohesion analysis as proposed by Halliday & Hasan (1976) appears to be more appropriate to the above aims. It is particularly suitable for the purpose of the present study, that of examining discourse-sensitive forms. Theoretical background and the empirical findings of previous studies on cohesion will be presented in chapter three followed by results of the present study in chapter four. In addition to an evaluation of amount and distribution of various cohesive devices, one device, namely reference, has 9 been selected for more detailed examination. Reference has been chosen for the following reasons: l ) i t i s perhaps the most discourse-sensitive of the devices analyzed; 2) i t i s frequently noted by educators and c l i n i c i a n s as an area of concern i n reports on older language impaired children; 3) some theoretical and a n a l y t i c a l issues have already been addressed (though not resolved) by l i n g u i s t s and c h i l d language researchers. A reference r e t r i e v a l analysis, as developed by Martin (1977, 1983), w i l l be presented i n chapter f i v e ; an analysis of devices used to create, maintain, and switch reference (based primarily on Bamberg 1987) w i l l be presented i n chapter s i x . Method and a preliminary description of the data w i l l be provided i n the next chapter. CHAPTER TWO METHOD AND PRELIMINARY DESCRIPTION OF THE DATA Method Subjects Twelve C a u c a s i a n boys between the ages of 9;1 and 11;8 participated: six of the boys had been diagnosed as language impaired by speech/language pathologists; the remaining six were recommended by classroom teachers as having normal language skills. The language impaired children were recruited with the assistance of speech/language pathologists solicited through an urban school board. The clinicians were asked to nominate children who satisfied the following criteria: - early documented history of specific language impairment as manifested by (oral) language production difficulties; - language comprehension within (or near) normal limits; - scores within the normal range on nonverbal tests of intelligence (for example, as measured by performance scores on the WISC-R); - monolingual speakers of English; - normal vision and hearing; - between the ages of 9 and 11 years. They were asked to exclude children whose language disorder was known or suspected to be secondary to disabilities such as mental retardation, cerebral palsy, hearing loss, autism, or acquired aphasia due to head injury. From those nominated, seven children were selected who best met the above 10 11 criteria. A partental consent form was not returned for one of the children, the only girl nominated by clinicians, leaving six boys to be included in the study. Individual subject descriptions are summarized in table I. All six had been referred for speech/language assessment sometimes in the past and had been subsequently diagnosed as language impaired based on results from standardized tests and spontaneous language sample analysis. Reports indica-ted a mild to moderate language production delay or disorder in each case; the specific nature of the language production impairment varied. An early documented history was not available for al l of the children; two of the children had apparently not been referred for speech/language assessment until grade four (see table I). Psychological test results, though not recent in some cases, showed performance scores within or near the normal range on the WISC-R. The size of the discrepancy between verbal and performance scores varied. Verbal scores were approximately equal to performance scores in four cases and significantly reduced in two cases (S3,S4). All of the children were receiving speech/language therapy at the time of testing. In general, therapy appeared to be directed at the formulation of complex sentences and at discourse skills. Length of time in therapy varied considerably; some had been seen since preschool (S1,S2) while others were just beginning a therapy program (S3,S6). The latter children had, however, received extra help from learning assistance teachers in previous grades. Four of the boys had repeated grade one. At the time of testing, two were attending a class for learning disabled children. The remaining children were enrolled in regular classes; one (S2) had attended a learning disabilities class during the previous year. 12 Table I.—Summary of Subject Characteristics Language Impaired Children Subject Age Current Class grade type Grades Repeated First Ref'd to SLP Therapy SI 11;4.1 5 ID8 reg.ld/wk gr.l preschool regular basis since preschool S2 9;8.1 3 regular (LD:gr.2) gr.l preschool regular basis since preschool S3 10;4.16 5 regular — gr.4 just beginning (LAb: gr.4) S4 11;8.12 5 LD gr.l gr.2 regular basis since gr.2 S5 9;1.2 3 regular gr.l gr.l regular basis since gr.l S6 10;0.4 M= 10;4 5 regular gr.4 just beginning (LA: gr.2,3) Normal Language Children Subject Age Grade SCI 9;3.1 4 SC2 10;1.12 5 SC3 10;4.16 5 SC4 11;2.11 6 SC5 11;3.11 6 SC6 9;2.20 M= 10;3 4 aLD: Learning disabled class bLA: Learning assistance The language impaired children were readily intelligible and, except for occasional problems with irregular verb inflection, the majority of utterances were grammatically acceptable at the sentence level (each utterance considered on its own). Six age and gender matched controls were drawn from the same four local schools as the language impaired children. They were chosen by classroom 13 teachers according to the following criteria: - matched in gender and six-month age interval to one of the language impaired subjects; - monolingual speakers of English; - normal vision and hearing - intelligence deemed to be within normal range - academic performance in median range. None of the control subjects had repeated a grade or been referred for learning assistance or speech/language therapy. Since psychological testing had not been done on any of these students, judgments of normal intelligence were made by their teachers. Teachers were asked to nominate students performing at average level academically and to avoid recommending students performing at either the low or high extremes of whatever academic scale was being used. All children, both language impaired and those with normal language skills, were cooperative and enthusiastic on a l l tasks. Data Collection The data were collected over a three-week period. Children were seen during regular school hours in the school they attended. Specific location within the school varied (speech/language therapy room, learning assistance center, nurse's room, office, or empty classroom). Each subject participated in tasks designed to obtain narrative and conversation samples. Narrative tasks were administered first and consisted of the following: personal experience, picture narration, and recall of an action story (video). Order of these tasks was counterbalanced across subjects. Data from a single subject, both narratives and conversation, were collected on the same day. 14 Sessions usually lasted 75 to 90 minutes excluding recess or lunch breaks if they occurred. None of the children seemed bothered by the presence of the recorder and microphone; on the contrary, they appeared to enjoy using the microphone. The experimenter was present for a l l tasks. In addition, a listener was present for certain tasks, or portions thereof. The listener was a classmate and friend of the subject being tested and was nominated by the teacher or subject himself. He or she not only fulfilled the role of unin-formed listener but his or her presence also helped to relax the subject. The personal experience narratives and conversation data will not be reported here. The speech contexts of the remaining narratives (picture-based and video recall) are described below. The narratives were chosen as a starting point for analysis primarily because the boundaries of the text are easily defined. A narrative stands as a complete text. Second, the burden of producing a cohesive text lies entirely with the child, who must exploit his or her own grammatical sytem in order to ensure that what is expressed in one portion of the text relates to what is expressed in another. From amoung the narratives, the picture-based and video recall tasks were chosen because the structure provided by such tasks made the data more comparable across sub-jects; the stories based on personal experience varied greatly both in length and content. In addition, the cohesion and reference analyses used in this research were judged to be more easily applied to contexts expressing primari-ly third person reference (i.e. he, she, it, they). Picture narration: A picture-based task was used to make the data more comparable, and to encourage children to produce relatively long texts by minimizing memory constraints. There were two reasons for seeking to obtain stories that were longer than those typically studied by researchers. First, 15 a sufficiently long text permits an examination not only of local intersenten-tial relations but also of any larger thematic organization that may be present. Second, i t seemed more appropriate to use stories somewhat more comparable to those these children come in contact with every day in school. Two stories were used in the following order: The Great Ape (Krahn 1978) and Arthur's Adventure in the Abandoned House (Krahn 1981). Only the first was used in the analyses reported in the present study. The pictures comprising The Great Ape were mounted on thirty-six 8.5 x 11 inch cards. A summary of the story can be found in appendix 1. Experimenter, subject, and listener were present. To familiarize the subject with the task, another wordless picture book was shown to the child. The experimenter demonstrated how the pictures could be used to tell a story and then had the child continue the story using two or three more pictures. All the children caught on immediately. The subject was then given the cards comprising The Great Ape and told that the cards made up a story. He was told that in a few minutes he would be a radio broadcaster for a children's show and that he could use the pictures to help him tell the story. He was then told to look over a l l the pictures on his own in order to familiarize himself with the story. The listener was seated so that at no time could he or she see the pictures. When the subject was ready to tell the story, he was told that his classmate would pretend to be a child listening at home on his or her walkman. The classmate listened over headphones while the subject told the story. Before each subject began, he was reminded that the people listening to the show over the radio could not see the pictures. When the subject had completed his story, he was asked to retell it without using the pictures. The subject was told to pretend that he was in a 16 different city recording another radio show and that he no longer needed the pictures because he had told that story many times before. He was asked to pretend that his classmate was a child living in that city who had never heard the story before and who had just turned on his or her Walkman to listen. Video recall: Only the subject and experimenter were present at the beginning. The subject was told that he would be shown a video and that after it was over, he would te l l the story to his listener. The video I wasn't scared (National Film Board of Canada 1977) was shown to the subject with the experimenter in attendance. The main protagonists of the twenty minute video are a boy and girl approximately the same age as the subjects themselves. The video is action-packed and suspenseful; there was never question of a child's attention wandering. The story is summarized in appendix 1. None of the subjects had previously seen the video. Once the video was over, the experi-menter told the subject to think about how he would tell the story while she, the experimenter, went to get the child's listener from his of her classroom. Each subject told the story to his listener, who did not verbally interject; changes in facial expression indicated attentiveness on the listener's part. None of the subjects appeared to have any difficulty understanding the content of the picture story or the video. The data was audio tape recorded and later transcribed. For purposes of data presentation, the first telling of the picture story will be referred to as Task 1 (TI), the second telling (without pictures) as Task 2 (T2), and the video recall as Task 3 (T3). For examples of narratives produced by a language impaired child and a normal language child, see appendix 2. 17 Statistical Procedures Two assumptions of the parametric statistical model did not appear tenable, namely normality and equal variance. Given the small sample size and nonrandom method of selection, there was no way of guaranteeing that the observations were drawn from normally distributed populations. Preliminary analyses indicated that equal variance between the two groups could not be assumed with respect to a l l variables examined. Nonparametric statistical tests, therefore, were judged to be more appropriate than parametric ones. The Mann-Whitney U test 1, a nonparametric alternative to the t-test, was used to compare the language impaired and normal language groups. To test dif-ferences across the three tasks, the Friedman two-tray analysis of variance by ranks 2 was used. If a significant difference was indicated by the latter, multiple comparisons between conditions were performed to assess specifically which tasks differed from each other. For the sake of consistency, two-tailed tests were used in a l l cases-, even when the possibility of predicting would have permitted a one-tailed test. A consequence of this decision is that stricter criteria had to be met before observed differences were considered significant. A .05 level of significance was used; if results were signifi-cant at a .01 level, a note was made of this in the presentation of results. Preliminary Description of the Data Before more specific analyses could be performed on the data, i t seemed imperative to compare the samples with respect to story and utterance length. *For procedure and tables, see Roscoe (1975). 2For procedure and tables, see Siegel (1988). 18 The following null hypotheses were tested: (1) Given the same task conditions, language impaired children do not differ from normal language children in narrative length as measured by a) the number of words per story and b) the number of utterances per story; (2) Given the same task conditions, language impaired children do not differ from normal language children in utterance length. (3) Mean utterance length for either group does not vary as a function of context. The above hypotheses were investigated by determing for each of the three contexts (picture narration, retell without pictures, and video recall), the number of words, independent clauses, and words per independent clause. Results Means, standard deviations, and ranges for a l l three measures as well as Mann-Whitney U results are presented in table II. Results of the Friedman two-way analysis of variance by ranks will be noted in the text. Words On the picture narration task, normal language (NL) speakers used approximately 523 words to t e l l the story whereas language impaired (LI) speakers used 400. On retell of the same story without the pictures (T2), the difference between groups was nearly significant {U=6, p<.10)3; the mean number of words used by NL speakers dropped to 418 words and that by LI speakers to 282 words. A significant difference was found between the two groups on the video recall task {U=3, p<.05); NL speakers told the story using 3For a difference to be considered significant at the .05 level, the value of U had to be equal to or less than 5. The smaller {/was, the more significant the difference. 19 Table II.—Mean Frequency of Words, Independent Clauses, and Words per Independent Clause across Tasks (T1,T2,T3) and Language Groups (LI,NL) TI (picture-based) T2 (retell,no pictures) T3 (video recall) LP NL LI NL LI NL Words M 400 523 282 418 291 558 SD 109 113 124 108 113 136 range 279-535 335-642 166-474 286-537 191-482 333-678 Independent clauses M 51.3 59.0 34.2 44.7 40.7 72.0 SD 11.8 15.8 15.2 12.1 15.6 16.0 range 38-71 41-84 17-60 28-60 26-69 46-87 Words/independent clause M 7.74 9.05 8.33 9.45 7.16 7.68 SD .87 1.56 .93 1.11 .68 .27 range 6.9-9.2 7.6-10.0 7.6-9.8 7.6-10.5 7.6-10.5 7.2-8.0 aN= 6 for each group Mann-Whitney lvalues TI T2 T3 Words 9.0 6.0 3.0* Independent clauses 13.0 8.5 2.0** Words/indep. clause 7.0 9.0 7.0 *p(.05 **;X.01 almost twice as many words as LI speakers (mean of 558 compared to 291). Hypothesis (1), as tested by the number of words per story, was, therefore, rejected for T3. Independent clauses Because length of the stories differed across contexts and across groups, 20 it seemed imperative to identify some unit that would permit frequency comparisons relative to story length. Since the cohesion analysis used in the present study had been developed within a systemic grammar framework (Halliday & Hasan 1976), it seemed appropriate to adopt a unit of analysis consistent with this framework, namely the independent clause. As defined by systemic grammarians, an independent clause is a unit which selects independently for mood, i.e. it can stand by itself as a declarative, imperative, interrogative or exclamatory structure." The procedures of Martin (1977), Rochester & Martin (1979), and Pappas (1981) were used to segment the data. As a unit of text segmentation, an independent clause consists not only of a single independent clause in the traditional sense, but also of any subordinate clauses grammatically related to i t . These subordinated clauses may be relative, complement (nominal or verbal), or adverbial clauses. Each of the following examples from the data was considered an independent clause (containing a relative, complement, and adverbial clause respectively): 2.1 a. she told the cops everything that had happened b. and she told them that he took off the propeller c. he stood there for awhile while his sister was getting the policeman In addition to while, the subordinating conjunctions of if, because, after, before, and when also introduced adverbial clauses. In the case of dialogue, a clause that identified the character who was talking was not considered a separate unit; 2.2, for example, was considered one independent clause: 2.2 and then the boy says, "I wasn't scared" Conjoined structures with subject ellipsis as in 2.3 were considered one unit 4The system of mood, in systemic grammar, expresses the speech function of the clause (see Halliday 1985a,b). 21 since, when a subject is missing, mood is frozen: 2.3 and then the brother turned around and went back to the bomb All other conjoined (or coordinated) structures were considered distinct units. Two units are present in each of the following examples: 2.4 a. they were riding their bike down a road / and then they came to big fence b. his sister was gonna hit him / so he started running c. and his sister told him to stop / but he wouldn't Each independent clause was numbered to facilitate analysis. Repetitions, false starts, and "asides" were bracketed and not included in frequency counts. In terms of the number of independent clauses per story, the LI children once again tended to be less productive on a l l tasks than the NL children (see table II), although the difference between the two groups varied considerably depending on the task. NL children used an average of eight and eleven more clauses on Tl and T2 respectively than the LI children to t e l l the same story. On T3, however, the difference jumped to over thirty clauses. This difference was highly significant (U=2, p<.01). Hypothesis (1) was rejected for T3. Words per independent clause NL speakers tended to produce longer clauses than LI children on all tasks. The difference between groups is nearly significant on Tl and T3 (U=l, p<.10). The longest independent clauses tended to occur on T2 and the shortest on T3. Friedman two-way analysis of variance by ranks indicated that the difference between T2 and T3 reached significance for the LI children 22 (fr=7.0, p<.05, T2>T3)5. Hypothesis (3) was rejected, therefore, in the case of the language impaired children. Smaller standard deviations for both groups on T3 suggested less individual variation on this particular task. Discussion The tendency for the language impaired subjects to use fewer words and fewer independent clauses than the normal language subjects on all three tasks support the results of an increasingly large number of studies suggesting reduced productivity on the part of language disabled children. Various factors have been cited in an attempt to explain this tendency for language disabled children to produce shorter stories. Graybeal (1981) found that language impaired and normal language children (aged 7 to 9) differed primar-ily in the amount of information recalled when asked to retell stories read to them. The groups did not differ on the types of story nodes recalled, on ordering of information, on amount of plausible information added, or on number of errors in recall. Graybeal concluded that the reduced productivity of the language impaired subjects was due to inadequate memory for content, i.e. to a gist recall deficit. Differences in memory capabilities between the two groups may have contributed to shaping the results of the present study: The Great Ape consisted of 36 pictures; the video was 20 minutes in length and included several embedded episodes. SI, in particular, had difficulty retelling The Great Ape without the pictures; he began at what corresponded to the twelfth picture. sFor the difference to be considered significant at the .05 level, the value of Fr had to equal or exceed 7.0. The greater Fr, the more significant the difference. 23 A memory deficit, however, cannot sufficiently account for all of the data. Even when memory demands were minimal, i.e. on the picture narration task (TI), the normal language subjects s t i l l used an average of 123 more words per story than the language impaired subjects. On retell of the same story without pictures, both groups reduced story length by similar amounts: the language impaired subjects by 118 words or 17 clauses, and the normal language by 105 words or 14 clauses). If memory were the sole factor accoun-ting for story length differences, then the recall stories of the language impaired children should have been reduced by a considerably greater amount than those of the normal language children. Finally, a memory deficit does not account for the fact that both groups used longer utterances on T2 than on TI, i.e. on the recall task as opposed to the picture-supported task. A second factor that has been proposed to account for differences in story length is knowledge of story grammar. Weaver & Dickenson (1979), for example, cited inadequate story schemata as an important source of poorer recall (less information) by reading disabled children with low verbal ability. Merritt & Liles (1987), using story generation and retell tasks, found that language impaired children used fewer complete story episodes, lower frequency of story grammar components, and fewer number of clauses overall than normal children aged 9 to 11. Unlike Graybeal, these authors concluded that their language impaired subjects had adequate storage and recall capability but weak command of overriding story structure. Roth & Spekman (1986), on the other hand, reported relatively intact story structure on the part of their learning disabled subjects aged 8 to 13 on a spontaneous story generation task: the learning disabled subjects used all category types in approximately the same order of saliency as their normal 24 peers. Learning disabled subjects, however, used fewer propositions per story. Because the information omitted tended to relate to character back-ground and story setting, the authors attributed shorter story length to impaired role-taking skills, i.e. to a lesser ability to make appropriate inferences regarding shared knowledge. Sleight & Prinz (1985) also attributed less detailed story recall descriptions on the part of language disordered subjects (aged 11 and 12) to a problem in evaluating listener need. Since knowledge of story schemata and awareness of listener need were not specifically targeted for analysis by the present research, neither can be ruled out as potential contributing factors to variations found in story length. However, as Roth & Spekman noted, the similarities in story structure between language disordered and normal children outshadow the dissimilarities. With respect to role-taking skills, the results of Liles (1985) conflict with the conclusions of Roth & Spekman and Sleight & Prinz. Liles found that language disabled and normal subjects made similar changes when describing a film to listeners in shared versus unshared conditions (uninformed versus naive listeners), one of these changes being the use of significantly more sentences for the uninformed listener. Liles concluded that the language disabled children were able to take on the role of listener and adjust their narratives accordingly. Memory, knowledge of story grammar, and role-taking skills have all been suggested as factors contributing to variations in story length as measured in terms of absolute frequencies, i.e. total number of words or propositions. Variation in utterance length, however, is not readily explained by these factors. Recourse, in the latter case, is usually made to syntactic language skills. In a study comparing comprehension and recall skills, Feagans & Short 25 (1984) found that reading disabled children (aged 6 and 7) appeared to comprehend the narratives as well as normally achieving children but produced fewer action units, complex sentences, and words than normal subjects. The authors attributed these results to less sophisticated syntactic skills. Klecan-Aker & Hedrick (1985) began with a similar hypothesis in their developmental study of the syntactic skills of normal grade six and grade 9 students as demonstrated in spontaneous story generation. Their unit of analysis, the T-unit (minimal terminal syntactic unit) appears to be a similar measure to the independent clause unit as defined by the present report.6 The authors reported significant differences between the grade six and grade nine students in T-unit length and in clause length.7 Based on their findings and those of previous researchers, the authors concluded that T-unit analysis was a reliable tool for the measurement of syntactic development in middle school-age children. A cautionary note by Scott (1988) is in order here. After reviewing the literature on average utterance length, Scott concluded that the developmental increases that occur are relatively small (from 7.5 to 10.0 words over a 9-year period), too small perhaps to be clinically useful. Research on the relationship between mean length of utterance (MLU) and chronological age underlines the need to exercise caution when attempting to use utterance length as an index of language delay, particularly when MLU exceeds 4.5 (see Miller 1981 for a review). The present results suggest that normal language and language impaired children differ in narrative utterance length. The results also provide at 6They defined T-unit as one main clause plus any subordinate clause or non-clausal structure that is attached or embedded in i t . 7 Clause here refers either to dependent or simple independent units, i.e. to minimal clause units. 26 least a preliminary answer to another question raised by Klecan-Aker & Hedrick, namely whether T-unit length would change as a function of the task. As noted earlier, independent clause unit lenth varied with the task for both groups of subjects, significant differences being found between Tl and T3 for the language impaired subjects. The results indicate that not only is utterance length sensitive to contextual factors, but also that normal and language impaired may show differential responses to changes in context. The language impaired subjects, for example, appeared to be more affected by the difficulty of the video recall task than the normal language subjects. For utterance length to be useful as a clinical measure, the following factors would have to be taken into account: small size of developmental changes with time; effect of task conditions and story content; possibility of differential responses elicited by changes in task conditions; and specific unit of analysis adopted. Having found a difference between the grade six and grade nine students in T-unit and clause length, Klecan-Aker & Hedrick sought the means by which the older subjects were extending their utterances. To test the hypothesis that the difference was due to the use of syntactically more complex forms by the older subjects, the investigators examined clause usage (number of clauses per utterance) and verb extensions (structures occurring after the main verb were categorized as simple, complex or infinitive, and as adverbial, object or complement). Failing to find differences on these measures, they looked at one more factor: the use of personal pronouns. They found that sixth graders showed a higher incidence of inappropriate personal pronoun usage, and suggested that the inclusion of an explicit referent for a pronoun may partially explain the longer utterance length of older subjects. A finding by 27 Weaver & Dickenson (1979) i s relevant here. Their normal readers added s i g n i f i c a n t l y more words that explicated temporal and causal relations to their r e t e l l i n g s than did reading disabled subjects with low verbal a b i l i t y . The l a t t e r omitted twice as many existing markers as good readers. These speculations r e l a t i n g utterance length to the use of personal pronouns and of markers of temporal and causal relations raise the p o s s i b i l i t y that utterance length may be influenced by l i n g u i s t i c phenomena that go beyond sentence boundaries. In the next section, discourse l e v e l relations alluded to by Klecan-Aker & Hedrick and by Weaver & Dickenson w i l l be explored more f u l l y . CHAPTER THREE COHESION: A REVIEW OF THE LITERATURE Theoretical Framework Systemic Grammar Systemic grammar is a theory of meaning as choice. Language, within this view, consists of networks of interlocking options; a choice in one system leads to a set of options in another. It is a functional grammar in three respects (see Halliday 1985a): 1. its aim is to account for how language is used, not only as i t unfolds in a particular time and place, but also how it has evolved to satisfy human need; 2. its major components (metafunctions) correspond to the general purposes underlying al l uses of language (to understand the environment and to act on others in i t ) ; 3. each element is explained by its function in the linguistic system. Halliday's work is not a rejection of "formal" grammar; his rather long Short Introduction to Func-tional Grammar (1985a), for example, is al l about form. However, language is interpreted, not as a system of forms to which meanings get attached—as in traditional syntactic approaches—but as a system of meanings accompanied by forms to express them. The forms of language are viewed not as the end in themselves but as means to an end, as realizations of certain choices made in order "to mean." Halliday's approach, then, appears particularly suitable to the task at hand: to examine the discourse of children whose "special point of vulnerability would lie exactly at the intersection of form and function" (Johnston 1985: 91). 28 29 Within the framework of systemic grammar, language is a seaiotic system i.e. a resource which enables humans to exchange meanings. Other semiotic systems include art form, modes of dress, family structures, etc. All these systems of meaning interact and constitute human culture. The "social" component of Halliday's approach lies in the fact that his main concern is with relationships between language and social structure.1 The linguist's task becomes that of describing the relationships between the organization of language and the social environment or context of situation. For Halliday, the appropriate model for the latter will explain the success with which people communicate. As listeners, we make predictions about what meanings are being exchanged and those likely to be exchanged, predictions which result in large part from information gleaned from the context of situation. The kinds of contextual information used by speakers and listeners relate to field (what is happening), tenor (who is taking part), and mode (what part the language is playing and medium of communication). Characterizing how language is organized means describing the purposes for which people use language, i.e. the types of meanings expressed. Halliday distinguishes three of these, which he calls metafunctions. 2 The ideational metafunction is the use of language to understand the environment. It consists of experiential meaning (representation of reality) and logical lHalliday does not claim to be describing, by his grammar and analytical procedures, processes of the human mind; his is not a psychological model in this sense. Brown & Yule's criticism (1983,ch.6) must be seen in this light; Halliday's "substitution view" was not proposed as a processing model but as an analytical procedure specifically for examining anaphoric reference. 2Halliday adopts a traditional stratificational view of the linguistic system in which there are three levels or strata (semantic, lexicogrammatical and phonological). The metafunctions described here comprise the semantic level. Each stratum and each component within i t is described as a network of options, of interrelated choices. components (expression of logical relations). The interpersonal metafunction is the use of language to act on others in the environment; it encompasses those meanings that arise from interaction between speaker and listener. The third organizational component, textual meaning, "breathes relevance into the other two" (Halliday 1985a: x i i i ) ; it comprises the resources language has for cohering within itself and with the context of situation, i.e. for relating any given item to both the verbal and nonverbal context. These functions are schematically presented in figure 1. Fig.l. Place of Cohesion within the Semantic Stratum of Systemic Grammar. Based on Halliday & Hasan (1976) and Halliday (1979). Learning "to mean" in a language involves learning the linguistic system but also learning how to use i t appropriately from one context to another. The task of the linguist, and that of the child, is to learn how features of the context of situation (field, tenor, mode) relate to the functional semantic stratum ideational experiential logical 31 components of the semantic system (experiential, interpersonal, and textual respectively), and in turn how these functional meanings receive lexicogram-matical and phonological expression. The term register is used to refer to variation in language that goes with variation in context of situation. The Nature of Text It is Halliday's third metafunction, the textual component, that is relevant to the present study. The textual component consists of those semantic systems which create text. Any passage, spoken or written, that forms a "unified whole" is known as a text. Though not easily defined operationally, the notion of text is crucial to functional grammar, where it has become the most important unit of analysis. In broad terms, a text is "any instance of living language that is playing some part in a context of situation" (Halliday & Hasan 1985: 10). Although typically longer than a single utterance, text is not defined by its length. A No Smoking sign and a long novel, for example, are both texts; they are communicative events, not supersentences. In more specific terms, texture requires three ingredients: generic structure 3, textual structure and cohesion. The first of these, generic structure, refers to the global structure of a communicative event. The macrostructure of nearly every communicative event could be described very generally as consisting of a beginning, middle, and end. Genre, according to Halliday & Hasan (1985), is short for generic semantic potential, the set of 3According to Halliday (1977), generic structure, while important to texture, actually lies outside the linguistic system, belonging to a higher level semiotic system. The text-forming resources of the linguistic system, in his view, are textual structure and cohesion. meanings usually associated with a particular contextual configuration. The generic structure of a shopping situation, for example, might consist of the following (from Halliday & Hasan 1985: 59): sale request, sale compliance, sale, purchase, purchase closure. Terms other than genre that have been used in similar attempts at representing socio-cultural knowledge include frames, scripts, scenarios, and schemata (see Brown & Yule 1983, ch.7 for a review). There have been relatively few attempts at describing the structure of conversations (e.g. Martin 1985, McTear 1984). Narrative structure has been studied in more detail using one of various story grammar models, a l l of which have attempted to describe "general"4 generic structure for narratives. The latter research has revealed the importance of story grammar to both compre-hension and recall of stories, but has not succeeded in distinguishing language impaired from normal populations on story grammar usage. The second ingredient required for texture within the framework of systemic grammar is textual structure (also called structural cohesion). Unlike generic structure, and, as we shall see, unlike cohesion, textual structure contributes to the derivation of sentential structure. It consists of both thematic and informational components. The thematic system refers to the speaker's organization of the clause as a message. The basic structure is that of theme/rheme as expressed through the order of the elements. The theme is speaker oriented; it is the first element in the clause, the speaker's chosen point of departure for his or her utterance. Rheme simply refers to the rest of the utterance, to whatever is not part of the theme. The informa-tional component of textual structure organizes the discourse into message 4general, because each narrative establishes its own contextual configu-ration and, therefore, the actual meanings expressed will vary with story content blocks called information units. The internal structure of each unit is based on a distinction between given and new information. In English, each informa-tion unit is encoded as a tone group. The new element is marked by the use of tonic prominence and usually coincides with the final lexical element in the tone group, the element the speaker wants the listener to attend to. 3 The given element is that which is treated by the speaker as recoverable to the hearer from the context. Although related, the thematic and informational systems are independent of each other; the thematic system is speaker-oriented, and the informational system is listener-oriented. In the unmarked case, a speaker selects the theme from within what is given and locates the new within the rheme. Marked6 patterns, however, allow the two to combine in a vast number of different ways. For example, by stressing oldest in 3.1, the speaker has located the new within the theme: 3.1 The oldest horse won the race. Parallelism, the recurrence of a grammatical structure throughout one or more successive clauses of a text may also be considered a type of textual structure in that i t contributes to the derivation of grammatical structure and at the same time serves to link utterances, i.e. has a textual effect. Bennett-Kastor (1986) identified three types of parallelism in the oral narratives of children aged 2 to 5: structural parallelism referred to the repetition of a verb phrase structure with different lexical content; lexical 9As Brown 4 Yule (1983, ch.5) have pointed out, i t is frequently d i f f i -cult, if not impossible, to identify tone groups and tonic prominence reliably in spontaneous speech. 6As defined by Radford, "an unmarked phenomenon is one which accords with general tendencies in language; a marked phenomenon is one which goes against these general tendencies, and is hence 'exceptional' in some way" (1981: 29). 34 p a r a l l e l i s m o c c u r r e d when b o t h l e x i c a l c o n t e n t o f t h e v e r b p h r a s e and i t s s t r u c t u r e were r e p e a t e d ; global p a r a l l e l i s m d e s c r i b e d i n s t a n c e s where s t r u c -t u r e and l e x i c a l c o n t e n t o f t h e e n t i r e c l a u s e were r e p e a t e d . The f o l l o w i n g d e v e l o p m e n t a l f i n d i n g s were r e p o r t e d : a d e c r e a s e i n g l o b a l p a r a l l e l i s m , a s m a l l r i s e i n l e x i c a l p a r a l l e l i s m , and a s h a r p r i s e i n t h e use o f g r a m m a t i c a l p a r a l l e l i s m . W i t h a g e , c h i l d r e n a l s o v a r i e d t h e t y p e o f s t r u c t u r a l l i n k a g e so t h a t " t h e p o t e n t i a l i t i e s f o r p r o v i d i n g c o h e s i o n t h r o u g h p a r a l l e l i s m i n c r e a s e t r e m e n d o u s l y a t age 5 " (p .366). Cohesion The p r e c e d i n g r e m a r k s were i n t e n d e d t o s i t u a t e c o h e s i o n i n t h e l i n g u i s t i c s y s t e m as c o n c e i v e d w i t h i n s y s t e m i c grammar. As d i s c u s s e d a b o v e , and as f i g u r e 1 i l l u s t r a t e s , t e x t u r e i s n o t a c h i e v e d by c o h e s i o n a l o n e . The l a t t e r i s j u s t one s e t o f r e s o u r c e s a t a s p e a k e r ' s d i s p o s a l f o r e n s u r i n g t h a t h i s o r h e r u t t e r a n c e r e l a t e s , i n a l i n g u i s t i c way, t o t h e s u r r o u n d i n g c o - t e x t . C o h e s i o n c a n be d i s t i n g u i s h e d f r o m o t h e r t e x t u a l m e a n i n g s by i t s n o n -s t r u c t u r a l p r o p e r t i e s ; i t i s i n d e p e n d e n t o f s e n t e n t i a l s t r u c t u r e . I t c o n s i s t s o f " t h e s e t o f l i n g u i s t i c r e s o u r c e s t h a t e v e r y l a n g u a g e h a s (as p a r t o f t h e t e x t u a l m e t a f u n c t i o n ) f o r l i n k i n g one p a r t o f a t e x t t o a n o t h e r " ( H a l l i d a y & H a s a n 1 9 8 5 : 4 8 ) . A s p e a k e r h a s s e v e r a l t o o l s , o r c o h e s i v e d e v i c e s , a t h i s o r h e r d i s p o s a l f o r t h e p u r p o s e o f c r e a t i n g t e x t , f o r m a k i n g t h e u n i t y o f m e a n i n g i n t h e t e x t as he o r s h e p e r c e i v e s i t e v i d e n t t o h i s o r h e r l i s t e n e r . B e f o r e d e s c r i b i n g t h e n a t u r e o f t h e s e d e v i c e s , i t may be u s e f u l t o h i g h l i g h t t h e i m p o r t a n c e o f c o h e s i o n t o e v e r y d a y c o n v e r s a t i o n . The f o l l o w i n g i s a t e x t i n v o l v i n g a 5 - y e a r - o l d ( f r o m H a l l i d a y 1 9 8 5 a : 2 7 7 ) : 35 3.2 Child: Shall I tell you why the North Star stays s t i l l ? Parent: Yes, do. Child: Because that's where the magnet is, and i t gets attracted by the earth. But the other stars don't. So they move around. If cohesion is removed and unmarked options in the textual systems selected, the following "deconstructed" text results (p.315): 3.3 The magnet is at the North Star. The earth attracts the North Star. The earth does not attract the stars which are not the North Star. The stars which are not the North Star move around. If cohesion were removed and the options of other textual components (thematic and informationsal systems) were chosen at random (instead of unmarked options), the text in (a) below would end up something like (b): 3.4 a. On this job, Anne, we're working with silver. Now silver needs to have love, [yea.] You know - the people that buy silver love i t . [Yea - guess they would.] Yes, mm - well naturally, I mean to say that it's got a lovely gleam about i t , you know; and if they come in, they're usually people who love beautiful things. b. With silver, we, Anne, are dealing in this job. What needs to have love is silver. Silver is loved by the people that buy silver. It is silver that silver has a lovely gleam about. The people who love beautiful things are usually people if people come in. (Halliday 1985a: 315) Halliday's deconstructed texts are revealing in two respects: 1) they illustrate the importance of the textual component to coherent discourse and 2) they model how a selective language disturbance, one restricted to a single metafunction (in this case, the textual component) might manifest itself. Cohesion occurs when the interpretation of one element (the presupposing) requires recourse to another (the presupposed). A single occurrence of cohesion consists of both elements and is known as a cohesive tie. The following example, while intelligible and even semantically decodable, is uninterpretable because the presupposed elements are missing: 3.5 Because he said so. If, instead, a listener overhears the following exchange, because, be, and so are readily interpretable: 3.6 X: How do you know Grant Fuhr is tired of playing for the Oilers? Y: Because he said so. This form of linking, where the presupposed elements are be found in the accompanying verbal context, or co-text, is known as endophora. In other cases, the presupposed element may not lie in the text but in the nonverbal context, a link known as exophora. If, for example, a child reaches for a broken piece of glass, an adult standing by might say the following: 3.7 Don't touch i t ! In this case, it has no verbal presupposed element; its source of interpreta-tion is an element in the environment. According to Halliday & Hasan (1976), only endophoric linking is an instance of cohesion, i.e. where both elements of the tie can be found in the verbal context. This follows from the goal of their particular analysis: to examine those formal elements in the speaker's or writer's discourse that contribute to coherence. However, a point raised by Brown & Yule (1983) is relevant here. Having defined cohesion as the set of semantic relations between elements in a text, Halliday & Hasan (1976) then proceed to call "cohesive" only those links that are verbally expressed. Brown & Yule's point is well made, and in adopting Halliday & Hasan's cohesive analysis, one must keep in mind that what is being analyzed is only a subset of underlying semantic links in the text, namely those that appear in the verbal record. There are times, perhaps many, when the speech/language clinician's task is precisely this, to determine what he or she can know about a referred indivi-dual based on the evidence of verbal productions. A single analysis cannot cover a l l bases simultaneously; this need not be a reason for abandoning i t . A second and more serious error would be to infer from this analysis that systemic grammar is not concerned with nonverbal elements. The grammar predicts, in fact, that genres and populations may differ with respect to their dependence on exophoric versus endophoric reference (Martin 1977,1983). Halliday & Hasan's cohesive analysis, then, applies only to ties whose elements are realized verbally. A second restriction is that the elements occur in different sentences. This decision is based on the assumption that cohesive ties have greater force when they occur between utterances because of the absence of structural (syntactic) links there. This restriction was modified slightly in the present study to include ties across subordinated clauses and across adjoined clauses with subject ellipsis (see chapter four.) Cohesive devices can be distinguished according to the specific type of semantic relation involved. Three relations have been proposed (Halliday & Hasan 1985). In a co-referential relation, the two elements of a tie are identical in reference:7 3.8 once upon a time there was a l i t t l e g i r l and she went out for a walk In 3.8, the presupposing item she and the presupposed item girl both refer to the same individual. In co-classification, the two elements belong to the same class but each refers to a distinct member of this class: 3.9 X: Look at my picture. Y: I'm making a bigger one. In 3.9, the substitute one forms a cohesive tie with picture. The two 7In the examples below, the presupposing element is italicized and the presupposed item (antecedent) is underlined. 38 elements, though both members of the class of pictures, refer to distinct entities. A third semantic relation is that of co-extension in which both elements of a cohesive tie refer to the same general field of meaning: 3.10 X: I ' l l wash i t , Mom. Y: No amount of scrubbing will fix that. The semantic field of wash and scrubbing are sufficiently close so that the use of scrubbing has a strong cohesive effect. In a l l of the above cases, the elements of the cohesive tie are components of or part of a message. In addition to these componential meaning, the semantic relations underlying the use of certain cohesive devices, for example conjunction, involve organic relations, those in which the elements of the tie are whole messages.8 These semantic relations are realized by lexicogrammatical means. Although there is not a neat one-to-one mapping between the semantic relation and its realization, there is a tendency for co-reference to be realized by pronouns, the definite article, and demonstratives; and co-classification by substitution or ellipsis. Relations of co-reference, co-classification or co-extension may be realized by purely lexical devices. The categories of cohesion as presented in Halliday & Hasan (1976) and as adopted by most cohesion researchers are based primarily on linguistic form, i.e. on the lexicogrammatical resources that realize the semantic relations described above. This classification comprises the following: reference, substitution, ellipsis, conjunction, and lexical cohesion. Unfortunately, there is some confusion over the nature of these categories in some studies; they are often presented as a classification based on semantic relations. This may partly be due to the ambiguity of their status as presented by Halliday & Hasan (1976). 8For an account of the clause as a phoric unit, see Martin (1977). An attempt to distinguish between underlying semantic relation and lexicogram-matical realization is presented in table III (adapted from Halliday & Hasan 1985). Description of the cohesive devices as used in the present study follows. Table III.—Summary of Halliday 4 Hasan's Inventory of Cohesive Devices NONSTRUCTURAL COHESION CCMPCNENTIAL RELATIONS ORGANIC RELATIONS GRAMMATICAL COHESIVE DEVICES Device A: Reference 1. Prcoccdnals 2. Demonstratives 3. Definite article 4. Comparatives B: Substitution & Ellipsis 1. Nominal 2. Verbal 3. Clausal LEXICAL COHESIVE DEVICES Repetition Synonymy Antonymy Meronymy Collocation Typical tie relation >co-reference >co-classification co-reference, co-classification, or co-extension A: Conjunctive ties Adjacency pairs e.g. question/answer offer/acceptance order/compliance Continuatives {e.g. still, already) STRUCTURAL COHESION Parallelism Theme-Rheme Development Given-New Organization Note: Adapted from Halliday & Hasan (1985: 82). Reference: Reference will be discussed more fully in chapters five and six. A summary of Halliday & Hasan's (1976) treatment of reference as a cohesive device is presented here. 40 The lexicogrammatical forms comprising this category (personals, demon-stratives, and comparatives) have been grouped together because in each case what is presumed is a semantic representation, a participant or a semantic construct as opposed to a particular piece of wording (see substitution/ellip-sis). A consequence of this is that such items are relatively unconstrained formally; elements of a tie can occur at varying distances and in various grammatical roles (see 3.11 below). Two of the subcategories of reference, personals and demonstratives, realize co-referential relations. The subcate-gory of personals (or pronominals) consists of personal pronouns, possessive determiners, and possessive pronouns: 3.11 Your instructor's name is Joan. She teaches the advanced level. That medal on the wall is hers. You'll like her. Personals are highly presuming; she, hers, and her in 3.11 can only be inter-preted by reference to Joan. They have a strong cohesive effect; the listener must retrieve something that has preceded. Demonstratives locate a referent on a scale of proximity: this, these are used to refer to an entity that is in some sense "near" and that, those to entities somewhat removed. The former often occur cataphorically, i.e. when the source of interpretation is found following the presupposing item: 3.12 You're not going to believe this. I won a thousand dollars. A distinction in proximity is also made between adverbial demonstratives: here, now ("near") versus there, then ("far"). The latter refer, not to an entity, but to a process in space (here, there) or time (now, then): 3.13 Someone's on the verandah. Go and see who's there. The definite article is also included in demonstrative reference. Unlike personals and the other demonstratives, the definite article has no content of its own. It signals only that the noun i t modifies has a specific referent 41 and that the information needed to identify i t is recoverable. Given the restrictions that the elements of a tie must be expressed verbally and occur intersententially, many cases of the are not considered cohesive. In the following examples, the is cohesive only in 3.14d:9 3.14 a. Look at the snow! b. The sun provided light by day and the moon by night. c. I finally found the house that I've been dreaming of. d. A boy and his sister went fishing. The boy wanted to stay longer. There is a third type of reference, comparative, whose underlying relation is that of co-classification (indicating another member of the same class). Like personals and demonstratives, comparatives refer back to a semantic representation, but in this case elements of the tie do not refer to the same individual: 3.15 One man came over to help. Another man phoned the police. The presupposed element one man provides a basis for comparison for another. Ellipsis/Substitution: Ellipsis and substitution may be collapsed into one category and viewed as manifestations of the same phenomenon. In both cases, cohesion is achieved by means of positive omission, i.e. by what is left out. In order to make up the sense, the listener must retrieve a piece of wording from a preceding utterance. The speaker may leave the structure unfilled (ellipsis) or insert a "place-holding" element (substitution). Some grammatical environments allow for either: 3.16 X: So, you'll buy the red dress? Y: Actually, I prefer the green {one). 9The types of reference involved in 3.14 a,b,and c are exophoric, homo-phoric, and esphoric respectively. These will be discussed in chapter five. 42 Dress must be retrieved by X given either of Y's responses. The substitute one in Y's response is a placeholder. In ellipsis and substitution, what is presumed is a lexicogrammatical representation. This contrasts with those devices grouped under reference where the presumed element is a semantic representation. The relationship between elements of a cohesive tie belonging to reference is a semantic one. According to Halliday & Hasan (1976), the relationship between the elements of a cohesive tie belonging to ellipsis or substitution is lexicogrammatical. While this captures certain regularities, there remains some ambiguity. The relationship between dress and 0 or one can also be described by a semantic relation: that of co-classification. The relation between the elements of a cohesive tie may perhaps be described in semantic terms in both categories, i.e. in both reference and ellipsis/substitution. It remains the case, however, that the elements involved in the latter are more constrained formally. In reference, as noted, the elements can be separated by several intervening sentences, and readily take on various grammatical functions. Ellipsis or substitution may be clausal, verbal, or nominal depending on what has been omitted. So, do, and one are the substitutes for clausal, verbal, and nominal substitution respectively. Examples of a l l three types for both ellipsis (E) and substitution (S) follow: 3.17 (clausal) Is he coming? (E) Yes. [he is coming] (S) I believe so. (verbal) I ' l l call tomorrow. (E) Will you? [call tomorrow] (S) How will you do that? (nominal) Have a piece of cake. (E) There isn't any. [cake] (S) I ' l l have a small one. Verbal and clausal ellipsis overlap; clausal ellipsis often entails verbal ellipsis. A modified classification can be found in Bernstein (1981). Conjunction: In conjunction, the terms of the cohesive tie are 43 whole messages. The role of the conjunctive is simply to make explicit, or "stamp" the logical relationship which holds between the two messages. A conjunctive may not even be present: 3.18 I'm going to bed. I!m totally exhausted. Such implicit conjunction obviously contributes to texture and several studies on conjunction have included i t (e.g. Clancy et al. 1976, Martin 1983). Unfortunately, in the absence of an explicit conjunctive, it is not at all clear what criteria might be used to decide whether a conjunctive relation is present in certain cases and not in others, and if so, which kind of conjunc-tive relationship is present. Only instances of explicit conjunction were analyzed in the present study. A second distinction that has been made is between external versus internal conjunctive relations. External relations are connections between external phenomena, i.e. phenomena about the "real world". Internal relations are those that link internal phenomena. Examples follow: 3.19 a. (external) First of a l l Freddie grew pale. Next he started to shake. Finally he passed out. b. (internal) First of a l l I'm not a nurse. Next I'm not paid to care for sick children. Finally I need to be home by four o'clock. The conjunctive relations expressed in both (a) and (b) above are temporal ones. However, in (a) the conjunctions connect events whereas in (b) they connect steps in an argument. Distinguishing between external and internal conjunction is not usually as clearcut as these examples might suggest. For this reason, and since internal conjunction was so rare in the data reported here, external and internal conjunction were collapsed in the present study. Various classifications of the logical relations of conjunction have been proposed. The present study adopted that of Halliday & Hasan's (1976), which 44 divides conjunction into the following form-based categories: additive, adversative, causal, temporal, and continuatives. Examples from the data follow: 3.20 a. (additive) the ship went back to New York/ and the ape went back to the island b. (adversative) the ape wanted the banana/ but he couldn't reach it c. (causal) the ape wanted the banana/ so he tried to grab it d. (temporal) the girl waved goodbye to the ape/ then she started to cry e. (continuative) the girl was running after the boy/ and he ran into this big open area/ and the girl was still looking where the trees were The category of continuatives consists of a closed set of items in English which are not governed by the same syntactic constraints as conjunctions; they can occur, for example, in various sentence positions: of course, veil, anyway, now, surely, after all, again, already, still, too, even. Such items nevertheless relate clauses to each other, thereby performing a cohesive function in a text. A more detailed table of the above relations and their conjunctive markers can be found in Halliday & Hasan (1976: 242-3). For modifications of this system, see Martin (1977, 1983) and Halliday (,1985a). As in other cohesive categories, Halliday & Hasan (1976) distinguish between intrasentential and intersentential conjunction. Their decision to consider only intersentential conjunctive relations has not been followed here. As in Martin's study (1977,1983), subordinating conjunctions and those occurring in conjoined structures with subject ellipsis were also included. There were two reasons for this decision. First, as Halliday & Hasan them-selves point out, the sentence is a very indeterminate category (1976: 232). 45 Cues provided by intonation and pausing do not always satisfactorily support one segmentation over another. A second and related reason was the fact that a considerable amount of information on conjunction would otherwise have been lost. A child who uses conjunction to combine individual propositions into clause complexes may be demonstrating more advanced cohesive skills than a child who speaks in simpler sentences that a l l begin with the same conjunction (e.g. and). By Halliday's decision, the discourse of the advanced child would appear to contain fewer conjunctive links than that of the second child. Because of the tendency for many children to begin each utterance with the same conjunction, i t may be that intrasentential conjunctions are more revealing as far as conjunctive complexity is concerned. The use of inter-versus intra-sentential conjunctions may itself be an important variable. In the present study, instances of each were tabulated separately in order to examine whether this was a variable across groups and contexts. Lexical cohesion: Another resource used by speakers to create cohesive discourse is vocabulary selection. Reference, ellipsis/substitution, and conjunction have often been described as grammatical sources of cohesion in that they involve closed systems. In lexical cohesion, items of the cohesive tie are content words and, therefore, there is a far greater number of options available to a speaker. There are different ways of grouping subtypes of lexical cohesion, a l l of which describe the same kinds of phenomena. The most common subtype is repetition of a lexical item. The repeated item may or may not show identity of reference with the previous occurrence: 3.21 a. (co-reference) X: We finally bought a new car. Y: Yeah, I saw the car in your driveway. b. (co-classification) X: We finally bought a new car. Y: Ooh, cars are expensive these days. 46 An item showing identity of reference with the preceding occurrence is often accompanied by the demonstrative the which signals to the listener that he or she knows which referent is intended (see 3.21a). In such cases, there are two cohesive ties, referential {the) and lexical (car). The cohesive items need not be morphologically identical: 3.22 The winner of tonight's game will play against Boston. Whoever wins that match will proceed to the finals. The elements of a lexical tie may also be related by synonymy or its variants (antonymy, meronymy10, and use of a superordinate or general word). As in the case of repetition, the tie may or may not involve identity of reference. Examples with identity of reference are the following: 3.23 a. (synonym) She did nothing to stop the thief. Later she described the robber to the police. b. (superordinate) Your dog's made a mess on the carpet again. Get that stupid animal trained! Examples without identity of reference are the following: 3.24 a. (meronym) I'd like to go by climbing on birch tree, And climb black branches up a snow-white trunk, Toward heaven. (Robert Frost, Birches) b. (antonym) Inch by inch, life's a cinch. Yard by yard, it's hard. In the above examples, the two elements occurred in adjacent clauses for brevity's sake. Like reference, however, lexical cohesion is highly uncon-strained and a lexical tie may span long passages of intervening discourse. In addition to repetition and synonymy, there is a third lexical type known as collocation in which the elements of the tie are related by a "co-occurrence tendency." Collocation is cohesive in that the occurrence of one element sets up a strong expectation of the occurrence of an associated iopart-whole relationship (e.g. door/house, top/mountain). 47 element. Examples of pairs which often co-occur are: pipe/smoke, thunder/ lightening, nightmare/scared. Though an important cohesive device, colloca-tion has not been analyzed in cohesion studies due to the difficulty of ensuring coding reliability. It has not been included in the present study. Related Studies I am suggesting that irrespective of whether i t is a dialogue or a monologue, as soon as you have said one word, you have created an environment. The more that is said, the greater the limitations on what can be said relevantly and sensically. (Halliday & Hasan 1985: 115) Learning to follow upon a previous utterance "relevantly and sensically" apparently requires a long period of time, possibly an entire lifetime. Literacy likely has an important role to play in sharpening these skills. Tying individual messages or their components utilizes what is called phoric means. Phoric structures are those which presume information which the speaker assumes his listener knows. The acquisition of phoricity involves both knowing what information can and cannot be presumed, and using linguistic resources appropriately to mark this distinction. While it is difficult to examine directly a child's knowledge of what can or cannot be presumed, one can, nevertheless, examine his or her productions for the use of the linguis-tic resources of phoricity. Describing the linguistic devices that are used to t e l l listeners what they do not know and at the same time to remind them of the things they do know involves the textual metafunction. A subset of the resources used to realize textual meaning, namely cohesive devices, have been described in the previous section. The next section will present a summary of what is know about cohesion from the developmental literature and from investigations on language disordered subjects. 48 The Development of Cohesion There i s some confusion i n the l i t e r a t u r e as to when texture develops. This confusion appears, i n part, to be a result of an imprecise d e f i n i t i o n of text. As discussed e a r l i e r , a text i s not defined by i t s length but as a unity of meaning i n some context. Lahey (1988) suggested that just as children move from communicating i n single words to clauses and sentences, so do they eventually move from sentences to sequences of sentences: "Eventually (Post Phase 8 1 1 ) , these sequences form a connected unit, and thus they can be considered a form of discourse, or text ( i . e . a unit of connected language that i s larger than a sentence)" (p.266). While i t i s true that development of texture extends well into the school years, i t begins much e a r l i e r , namely as soon as children begin to make any kind of l i n g u i s t i c connection between their utterances and preceding ones (either their own or their conversational partner's). Halliday (1979) found evidence of texture i n his son Nigel's protolan-guage as early as age f i f t e e n months. In a picture-naming game, for example, Nigel would use a special intonation pattern to indicate a second move; a change i n the intonational pattern of a single-word utterance indicated to his li s t e n e r that his current utterance was a follow-up to the f i r s t exchange. A second early interaction that had texture also involved intonation: i n reparatory acts, Nigel adopted the strategy of repeating, slowly and c l e a r l y , an item misunderstood by his l i s t e n e r . The tr a n s i t i o n from Nigel's protolan-guage to his mother tongue (approximately the second half of his second year) proved to be a c r i t i c a l period for development of text-forming resources. During that time, the demonstratives that and it emerged i n an exophoric 1 1Lahey's Post Phase 8 appears to correspond to the school years. 49 context and gradually became the f i r s t r e f e r e n t i a l items to be used anaphor-ically. The substitute one and the conjunction but also emerged as cohesive items at this time. In addition to developing intersentential relations of cohesion, Nigel was also learning to control textual structure. He learned very early, for example, to use tonic accent to distinguish between given and new information i n order to relate a current utterance to a previous one. Halliday provides additional examples to i l l u s t r a t e Nigel's acquisition of a th i r d component of texture, namely generic structure. The f i r s t generic types to emerge were dialogue and narrative: dialogue out of pragmatic acts (those requiring a response), and narratives out of mathetic ("reality-creating") acts of meaning. According to Halliday, the a b i l i t y to construct text begins early and develops alongside the acquisition of ideational and interpersonal meanings. Based on data collected from young Polish children aged 1;7 to 2;0, Shugar (1978) came to a similar conclusion. The author found evidence of jo i n t text construction (adult-child) at the stage i n l i n g u i s t i c development when the children could only produce one or two words at a time. Of these early "dialogues", Shugar noted: "The c h i l d engages i n these enterprises early i n the game, at the one-word stage, and, i f we examine the process, we find that he places his utterance i n systematic and meaningful relations to the adult utterances" (p.229). Keenan & S c h i e f f e l i n (1976) reported evidence of topic collaboration i n the conversations of an English-speaking mother and her daughter aged 16 to 34 months. While i n the one- and two-word stages, the c h i l d demonstrated an a b i l i t y to establish discourse topic by securing the attention of the l i s t e n e r , a r t i c u l a t i n g the utterance c l e a r l y , and providing enough information for the l i s t e n e r to i d e n t i f y references i n the discourse 50 topic. Both verbal and nonverbal means were used. In a longitudinal study of young English children, Bloom, Rocissano and Hood (1976) examined adult-child discourse i n order to determine how children use information from adults' input sentences to form contingent responses. Linguistically-contingent speech was present i n the data of even the youngest children and developed s i g n i f i c a n t l y over the time period studied (21 to 36 months of age). Keenan 4 Klein (1975) also reported evidence of "text construction" i n their observations of c h i l d - c h i l d discourse (twins boys., aged 33 months). Not only were the children able to attend to and understand one another's utter-ances but they also demonstrated an a b i l i t y to provide relevant responses by means of formal modifications of a previous utterance. Early modifications were f a i r l y p rimitive, consisting mainly of sound play i n which the children altered the phonological shape of each other's utterances. According to the authors, sound play may be a way of developing the following a b i l i t i e s : to attend to and acknowledge a partner's utterances, and to is o l a t e and manipu-late formal elements of a previous utterance. In r e f e r e n t i a l discourse, the strategies of repe t i t i o n and formal modification were once again used to respond relevantly. The l a t t e r involved incorporating a l e x i c a l element of a previous utterance into a current one, and substituting an element with a constituent of the same grammatical status. The authors also provide examples of pronouns used anaphorically by the children. The development of texture, then, begins at a very young age, by at least sometime i n the second year. At this time children begin to demonstrate an a b i l i t y to l i n g u i s t i c a l l y relate their utterances to previous ones, either their own (Halliday 1979) or those of others, whether adults (Shugar 1978, Keenan 4 S c h i e f f e l i n 1976, Bloom et a l . 1976) or peers (Keenan 4 Klein 1975). 51 The development of texture i s not somehow cued by the development of complex sentences, as Lahey (1988) appears to suggest; i t begins at the one-word stage. Children, therefore, take part i n conversations and express themselves in coherent monologue long before the school years. These, too, are forms of discourse or text. Although i t begins early, development of texture occurs over a long period of time. In Bloom et al.'s study (1976), for example, l i n g u i s t i c con-tingency, although present i n the data of a l l the children, s t i l l accounted for less than half of the adjacent speech the children produced at three years of age. 1 2 The course of development of text construction i s not well under-stood. Host information comes from studies that have examined a single device and the majority of these involve young children. Findings on reference w i l l be presented i n chapters f i v e and s i x ; studies on conjunction and substi-t u t i o n / e l l i p s i s are summarized below. 1 3 Conjunction: Studies investigating conjunction i n young children have looked at s p e c i f i c conjunctions used, l o g i c a l relations encoded, and the discourse environment they occur i n . Based on findings from their cross-1 2Bloom et a l . (1976) used a rather involved and questionable d e f i n i t i o n of contingency. I t consisted of three main types: expansions (the c h i l d added information about the topic of a prior utterance), alternatives (the c h i l d added information by opposing an aspect of the topic i n the prior utterance), and expatiations (the c h i l d added information to the prior utterance and introduced another related t o p i c ) . A d i s t i n c t i o n was also made between semantic contingency based on s i t u a t i o n a l context only {contextual contingency) and that based on formal c r i t e r i a as well as possible s i t u a t i o n a l context ( l i n g u i s t i c contingency). The l a t t e r was further subdivided into intraclausal relations (information was added by the c h i l d within the same clause structure and with the same verb r e l a t i o n as i n the adult utterance) and interclausal relations (information was added with another clause that was grammatically subordinate to the clause of the adult utterance). 1 3To my knowledge, there have not yet been any developmental studies reported i n the l i t e r a t u r e that have examined l e x i c a l cohesive devices. l i n g u i s t i c study, Clancy, Jacobsen, & S i l v a (1976) suggested that children f i r s t relate their utterances to the immediate nonlinguistic context, and then to previous utterances by simple juxtaposition. Additive and temporal relations appear to be encoded before causal or adversative (Clancy et a l . 1976, Ervin-Tripp 1978, Bloom, Lahey, Hood, L i f t e r & Fiess 1980). The e a r l i e s t conjunction to appear i s and, which i s used to encode various semantic relations (Bloom et a l . 1980). Both the variety of forms used and meanings encoded increases with time. There are c o n f l i c t i n g views as to whether a c h i l d f i r s t uses conjunctions to relate his or her utterances to an immediately preceding adult utterance (Clancy et a l . 1976) or whether conjunc-tions f i r s t appear within the child's own utterances (Ervin-Tripp 1978, Bloom et a l . 1980; Eisenberg 1980). The answer i s no doubt to some degree dependent on the s p e c i f i c conjunction (Eisenberg 1980). In any case, there i s evidence for the use of conjunction both within and across speaker turns i n children as young as 2-years-old. Young children's use of e x p l i c i t connectives i n narratives was analyzed by Bennett-Kastor (1986) 1 4. Rank order for a l l age groups combined (ages 2 to 5) was the following: and, then, so, but, first, and when. The 2-year-olds used the smallest number and variety of e x p l i c i t connectives; and and then were used most frequently and approximately equally. The 3- and 4-year-olds showed considerable experimentation, each group using seven types of connec-t i v e s . The use of i n t e r c l a u s a l connectives decreased, however, between the ages of 3 and 5, a decrease which the author attributed to a greater a b i l i t y of older children to ensure cohesion i n a variety of ways other than by 1 4Bennett-Kastor's findings on p a r a l l e l structure based on the same data were reported e a r l i e r . conjunction alone. The most common connective used by the 4- and 5-year-olds was and. The few studies that exist on the use of conjunction i n older children suggest that the acquisition process i s long and involved. McClure & Geva (1983) investigated the a b i l i t y of students of grades four, s i x , and eight to choose between the adversatives but and although. Findings from adult data supported the authors' proposal that the choice would be related to focus, but being used to introduce foreground information and although to introduce background information. Although a l l children demonstrated an awareness of the syntactic contraints governing the use of these conjunctions and an a b i l i t y to use them to mark an adversative relationship, most (including the grade eight students) did not demonstrate knowledge of the role of these conjunctions i n marking focus. As the authors noted, such a rule of focus would be p a r t i c u l a r l y important i n written text where par a l i n g u i s t i c and nonverbal contextual cues are absent. Their conclusion depends c r u c i a l l y on the v a l i d i t y of the d i s t i n c t i o n they draw between but and although based on their notion of focus. Another effort to c l a r i f y the continuing development of "connectivity" i n older children i s that of Scott (1984). She examined play and interview samples of 114 children (aged 6 to 12) for instances of adverbial conjuncts and disjuncts. According to the grammatical system she used {Grammar of Contemporary English by Quirk,Greenbaum, Leech & Svartvik 1972), conjuncts and disjuncts operate on the periphery of clause structure and can be d i s t i n -guished from adjuncts which are integrated within clause structure. Based on this same system, Scott defined conjuncts as items which signal a l o g i c a l r e l a t i o n between sentences (e.g.in addition, so, then, otherwise, anyway. 54 therefore, however) , 1 9 and disjuncts as items which convey a speaker's attitude toward the form or content of a sentence (frankly, wisely, really, unfortunately).16 Scott found that conjunct use was just emerging at the age of s i x , at which time i t was limited to a few l o g i c a l relations expressed by only one l e x i c a l item per r e l a t i o n : i n f e r e n t i a l then, result so, concessive though and t r a n s i t i o n a l now. With development, there was an increase in the frequency of these conjuncts, i n the number of different l o g i c a l relations, and i n the number of l e x i c a l items used per r e l a t i o n . Disjunct use was rare at a l l ages, limited for the most part to really and probably. Scott also examined the frequency of the sentential coordinating conjunctions and and but and found that connectivity via such elements was more prominent at a l l ages than connectivity v i a conjuncts. Although the increase i n the number of l o g i c a l relations and l e x i c a l items to express them no doubt does increase with age, a few cautionary notes are i n order. F i r s t , the data are presented i n frequency of occurrence with no control for length. This i s problematic because the samples of the 10- and 12-year-olds were longer than those of the 6- and 8-year-olds and because, for each age group, the range i n number of utterances was substantial. With such widely d i f f e r i n g sample lengths, s t a t i s t i c a l analysis would be next to impos-s i b l e . Second, the greater variety of l e x i c a l items supposedly used by older children was not, i n fact, t y p i c a l of the majority of the older children: for any one of these "sophisticated" l e x i c a l items, a maximum of only four 1 3 S c o t t ' s conjunct category did not include sentential coordinating conjunctions such as and and but; she examined these separately i n a subse-quent analysis reported i n the same study. 1 6 D i s j u n c t s are not usually considered to be cohesive devices by most researchers. 55 children, out of the t h i r t y 12-year-olds, actually used i t . Third, conjunct use was rare, occurring only twice i n 100 utterances for the 6-year-olds and four times i n 100 utterances for the 12-year-olds. If average length of the samples i s taken into consideration, i t becomes evident that the conclusions Scott drew were based on very few exemplars per subject (between two and seven for the interview and play data combined). Fourth, i n her comparison with adult data, Scott used a rather crude method to make the samples comparable: from the number of conjuncts adults used i n a 40-minute sample, Scott computed the number predicted to occur i n 45 minutes (recording time for the children's data). She found that the 10- and 12-year-olds would need to produce a three-f o l d increase over their observed frequency and the 6-year-olds eight times their observed frequency to equal the adult rate. There was considerable overlap i n the variety of l e x i c a l items and l o g i c a l relations encoded between the adult and older children's data. Because of the crudeness of method used to equate the samples (recording time), the results are d i f f i c u l t to i n t e r -pret. F i n a l l y , the v a l i d i t y and r e l i a b i l i t y of Scott's d i s t i n c t i o n between conjuncts and sentential coordinating conjunctions i s questionable; she herself admits to considerable d i f f i c u l t y i n distinguishing the two uses, especially when the same l e x i c a l item i s used for both. The above findings indicate that conjunction continues to develop during the school years and probably into adulthood. One aspect of t h i s development, according to McClure & Geva (1983), involves increasing knowledge of an intersentential rule of focus. Scott (1984) described the continuing develop-ment of "connectivity" as one of increased frequency of conjuncts, including increased number of l o g i c a l relations and l e x i c a l items to express them. For Martin (1977,1983), l i n g u i s t i c development i n the school years can best be 56 described as the development of register, i . e . learning to say the "right" thing i n the "ri g h t " place at the "ri g h t " time rather than with learning ways of saying things. As Martin explains, people have different coding orienta-tions: "given the same lexicogrammatical resources and placed i n the same situa t i o n type, speakers may produce texts that vary systematically according to who the speaker i s " (1983: 4). He invokes this notion of code to account for the different kinds of texts produced i n the same situation by children of different ages, who appear to be drawing on the same text-forming resources (as far as the variable under consideration i s concerned). Systemic grammar, because i t describes language as a system of options, appears to be the most appropriate model to use when studying this v a r i a t i o n . Within this grammar, pr o b a b i l i t i e s are assigned to choices on the bases of 1) inherent p r o b a b i l i -t i e s (e.g. active voice i s more usual than the passive i n English regardless of context); 2) register reweightings (passive i s more l i k e l y to be used i n s c i e n t i f i c writing than i n other contexts) and 3) code reweightings (influence of age on the use of the passive voice i n s c i e n t i f i c w riting). Martin examined both reference and conjunction i n the spoken narratives of 90 children aged 6 to 11 years. Data from three s t o r y - t e l l i n g tasks were analyzed: picture book narration (task 1), r e t e l l of the same story without pictures (task 2), and r e t e l l of a fable read to the children (task 3). Martin's findings on reference w i l l be discussed i n chapter f i v e . With respect to conjunction, stories by the younger children (6- and 7-year-olds) differed from those of the older children i n two respects. F i r s t , the younger children treated task 1 more as a picture-describing task than the older children: they used fewer e x p l i c i t conjunctions to mark relations between clauses i n that task. Second, the younger children r e l i e d more on temporal 57 relations than on consequential and comparative relations i n task 3; they made fewer adjustments that would indicate s e n s i t i v i t y to the fable genre than the older groups. From this (together with his findings on reference), Martin concluded that the younger children were using a different coding orientation than the older children i n the contexts examined. Although a l l children had been using additive, temporal, and consequential relations for years, the younger children appeared to use this resource d i f f e r e n t l y given a spec i f i c context: "the coding orientation of the youngest group i s more context dependent and less adapted to fable genre than i s that of the older groups" (1983: 34). Clearly, what a c h i l d says depends not only i n a general way "on the context" but on a child's particular coding orientation with respect to the context he or she finds him/herself i n . An orientation that remains context dependent and insensitive to certain genres would have certain drawbacks i n the educational system, where nonverbal and verbal contexts are increasingly distanced and where students are expected to be increasingly sensitive to generic structure. Ellipsis/Substitution: The category of e l l i p s i s / s u b s t i t u t i o n remains largely unexplored; three studies, a l l involving young children, w i l l be summarized here. Ervin-Tripp (1978) examined changes that occurred over time i n both c h i l d and adult speech i n early child-adult (experimenter) conversa-tions. She found that by Brown's Stage IV, there was a s h i f t i n the conversa-t i o n a l cohesive s k i l l s of the children from "speech that i s merely associated to the partner's speech to speech that requires l i n g u i s t i c processing and presupposes particular features of the partner's speech systematically" (p.361). Amoung the features comprising t h i s change was the appearance of 58 reply e l l i p s i s i n which the subject (often pronominalized) and the au x i l i a r y remained, and the rest of the predicate was absent (adult: "Who came?" c h i l d : "Ellen did."). This occurred i n the conversations of children i n Brown's Stage IV or V (children aged 2;6 to 3;6 i n Ervin-Tripp's study). In a study of e l l i p s i s i n mother-child dialogues, Bernstein (1981) found that the categories of e l l i p s i s developed i n step with Mean Length of Utter-ance. Question responses and polar e l l i p s i s (yes/no replies) appeared f i r s t followed by clausal (propositional and modal) and f i n a l l y nominal e l l i p s i s . Changes that occurred i n the mothers' speech over time were observed i n mother-child (as opposed to mother-mother) sequences, the most important change being an increase i n the variety of e l l i p t i c a l types used. The children participated i n e l l i p t i c a l relations even before they acquired syntax, i . e . at the one-word stage. The adult's embellishment of these texts appeared to lead the c h i l d into a f u l l e r p a r t i c i p a t i o n . As Bernstein noted, "the c h i l d i s lured further into language by virtue of using language" (p.139). Bernstein also found that use of e l l i p s i s emerged i n between-speaker sequences. Dialogue seemed to assist the c h i l d i n producing l i n g u i s t i c relations that would i n t i t i a l l y be beyond his spontaneous a b i l i t y . The finding that e l l i p s i s emerged i n between-speaker utterances supported Clancy et al.'s findings on conjunction (1976) and Eisenberg's finding on the conjunction but (1980). I t confli c t e d , however, with those of Ervin-Tripp (1978) and of Bloom et a l . (1980), and with Eisenberg's findings on conjunc-tions other than but. In view of this unsettled issue, Roberts (1987) examined the development of the verbal substitute, do, i n adult-child dyads. The f i r s t uses of do were imitated or exophoric (the deleted information was available from nonverbal context). Cohesive do then appeared, functioning i n 59 a variety of ways for each of the children. This variety increased with MLU. Like Bernstein, Roberts found that early instances of cohesive do referred back to a previous adult utterance. Within-utterance verbal cohesion occurred l a t e , appearing f i r s t i n tag questions, then as a device to refer to a preceding utterance of the same turn or a previous turn of their own, and f i n a l l y integrated i n a single, usually complex sentence. An example of a child's use of do to refer to a previous turn of her own i s the following (p.4): 3.25 Sarah (37 months) Sarah: You l i k e fish? Adult: No. Sarah: I don't either. Similar examples were provided by Halliday (1979). Such examples suggest that discourse l e v e l relations by the young c h i l d are not en t i r e l y dependent on an immediately preceding adult utterance, i . e . the young c h i l d i s capable of providing both elements of the cohesive t i e . Cohesive Harmony: A recent development i n the study of cohesion i s the analysis of cohesive harmony, which involves ide n t i f y i n g not only cohesive devices but also the chains they may or may not enter into, and determining the extent of interaction between these chains. Advocates of cohesive harmony analysis (Pappas 1981,1985, Hasan 1984) suggest that cohesive harmony i s a r e l i a b l e predictor of coherent text. The f i r s t stage i n the analysis involves id e n t i f y i n g the cohesive devices used (reference, s u b s t i t u t i o n / e l l i p s i s , and le x i c a l ) and then " l e x i c a l l y rendering" the text, i . e . replacing each cohesive device with i t s f u l l or presupposed form (e.g. a pronoun would be replaced by i t s nominal antecedent). Those items referring to the same entity (co-reference} are grouped to form identity chains (IC); those related by co-c l a s s i f i c a t i o n or co-extension comprise similarity chains (SC). Items entering into either of these chains are called relevant tokens and the rati o of relevant to t o t a l tokens i s computed to y i e l d a cohesive density index. In the th i r d stage, the interaction between these types of chains (IC and SC) are ascertained. For two chains to interact, at least two members of one chain must stand i n the same re l a t i o n (e.g. actor-action, action-location, etc.) to two members of the other chain. The following has been taken from an analysis of a narrative text (Pappas 1985: 178): utt.# Chain X Chain Y Chain 4 g i r l go-out 5 g i r l go-out 16 g i r l run 37 woman run place 37 daughter go 39 g i r l run-out house The interaction of chains X,Y, and Z are said to express an actor-action-location r e l a t i o n . Tokens which enter into chain interactions ( a l l of the tokens l i s t e d i n 3.26) are called central tokens. The r a t i o of the number of central tokens to the number of relevant tokens yelds a cohesive harmony index. According to Halliday & Hasan (1985), a text has a greater chance of being judged coherent i f : 1) i t contains a low proportion of peripheral tokens (those not entering chains) to relevent ones; 2) i t contains a high proportion of central (tokens entering into chain interactions) to non-central tokens and 3) there are few breaks i n the interaction network. Pappas (1981, 1985) analyzed oral and written stories by grade one children and found that children's narrative c a p a b i l i t i e s , as measured by cohesive density and harmony indices, varied depending on the context. The cohesive harmony index, for example, was highest i n the oral r e t e l l i n g context and lowest i n a task requiring a c h i l d to write an o r i g i n a l story. Hasan (1984) found the cohesive harmony index to successfully correlate with informal judgments of coherence by readers of stories produced by 6- and 7-year-olds. Cohesive harmony may prove to be a more revealing measure than an analysis of cohesive devices alone. It requires that the " b i t s " of language that enter into cohesive t i e s "march i n step" (Pappas 1985: 184). This analysis, however, i s only beginning to be developed and several important a n a l y t i c a l problems remain. Decisions as to which items may be grouped i n a s i m i l a r i t y chain i s one; handling ambiguous or incorrect references (in anticipation of the data from the present study) i s another. For these reasons, cohesive harmony analysis was not attempted i n the present study. Cohesion and Language Impaired Populations Adults The f i r s t comprehensive cohesion analysis was that of Rochester & Martin (1979), who examined the discourse (interviews and narratives) of normal and schizophrenic (thought-disordered and non thought-disordered) adults. The investigators found that, for schizophrenic speakers, the two measures of cohesion used (amount and d i s t r i b u t i o n into cohesive type) were highly sensitive to context differences but only modestly sensitive to differences among the three groups. With respect to contextual differences, a l l speakers depended more on cohesion i n narratives than i n interviews. Referential t i e s were most sensitive to context differences followed by l e x i c a l and e l l i p t i c a l t i e s : a l l groups used smaller proportions of r e f e r e n t i a l t i e s and greater proportions of l e x i c a l and e l l i p t i c a l t i e s i n interviews than i n narratives. Conjunction did not d i f f e r across contexts. Group differences indicated that the schizophrenic subjects used fewer cohesive t i e s per clause than normals p a r t i c u l a r l y i n narratives. In addition, the thought-disordered group r e l i e d more on l e x i c a l cohesion than did the non thought-disordered group i n i n t e r -views, although neither group differed s i g n i f i c a n t l y from normals. The discourse of fi v e of the s i x thought-disordered patients showing high l e x i c a l cohesion i n interviews also f i t the description of "semantic glossomania": the choice of words based on conceptual associations apparently unrelated to the conversational topic. In narratives, l e x i c a l cohesion was used to a greater extent by normal subjects than by schizophrenics. Cohesion analysis has also been applied to the discourse of individuals with senile dementia of the Alzheimer type (SDAT). Ripich & T e r r e l l (1988) compared the topic-directed interviews of SDAT patients (aged 68 to 85) with those of normal elderly adults. The SDAT patients demonstrated increased verbosity, using more t o t a l words and conversational turns. Between group comparisons of cohesion (frequency and distribution) did not reveal s i g n i f i -cant differences between the two groups, although types of errors d i f f e r e d . Absence of a clear referent occurred four times as often i n the speech of SDAT subjects. Judgments of coherence by four l i s t e n e r s showed s i g n i f i c a n t differences between the two groups. Instances of l i s t e n e r incoherence were related p a r t i c u l a r l y to one subtype of cohesion disruption: missing element (absent piece of discourse that should provide the relationship between preceding text and that which follows, e.g. "My husband, he had a [missing element]. We were there.") There i s no indication i n Ripich & T e r r e l l ' s study of how the texts were demarcated. A second problem with the analysis concerns the measure used for 63 frequency of cohesion: t o t a l words spoken by each group divided by t o t a l occurrences of a l l cohesive devices, both appropriate and disrupted. Such a measure obviously obscures differences i n individual performances. The use of the word as a unit of analysis i s i t s e l f questionable. If i t were true that an attempt to be cohesive required more words, however s l i g h t the increase required, such an attempt would be masked by Ripich & T e r r e l l ' s measure. A f i n a l problem concerns the investigators' conclusion that "SDAT results not only i n the impairment of l i n g u i s t i c a b i l i t i e s but also i n the impairment of discourse a b i l i t i e s that contribute to Alzheimer patients' incoherent speech" (p.8). The separation of discourse a b i l i t i e s from " l i n g u i s t i c a b i l i t i e s " i s t h e o r e t i c a l l y questionable. No attempt was made to relate the impairment revealed by their study to other features of language dissolution i n Alzhei-mer's disease. Their suggestion that incoherence results from a loss of a b i l i t i e s to take the l i s t e n e r ' s perspective i s purely conjectural. Incoher-ence was related to one feature only, namely missing element. I t i s possible that this type of error may also be due to word r e t r i e v a l d i f f i c u l t i e s . Cohesion cannot be separated from other " l i n g u i s t i c a b i l i t i e s " . Accor-ding to Mentis & Prutting, "cohesion i s a prime example of how competence i s achieved through the integrated contributions of pragmatics, semantics, and syntax" (1987: 95). The authors examined cohesion i n the conversation and "narratives" (description and procedural) of three normal adults and three adults with closed head injury (CHI) and compared these results to pragmatic and syntactic a b i l i t i e s . Each of the CHI subjects had communicative d i s a b i l i -t i e s as measured by a protocol examining a number of pragmatic behaviors. Syntax was judged to be " r e l a t i v e l y well preserved"; a l l showed evidence of possessing the grammatical structures required to formulate a cohesive t i e . 64 CHI subjects used fewer cohesive t i e s than the normals i n the "narrative" conditions but not i n conversation. The groups differed with respect to e l l i p s i s , l e x i c a l cohesion, and incomplete t i e s . Normal adults used a higher percentage of l e x i c a l t i e s than CHI p a t i e n t s . 1 7 Mentis & Prutting suggested that lower l e x i c a l cohesion may be related to word r e t r i e v a l problems. CHI patients also used a higher percentage of e l l i p t i c a l t i e s than two of the three normals. Since more of these occurred across communication partners for the CHI patients, the authors speculated that the CHI subjects may have been using their partners' discourse as a scaffold i n order to relieve the stress placed on their impaired l i n g u i s t i c systems. The CHI subjects also used previous l e x i c a l items supplied by a conversational partner to a greater extent than normals. CHI subjects used a r e l a t i v e l y high percentage of incomplete t i e s whereas no incomplete t i e s were observed i n samples from normal adults. Most incom-plete t i e s were reference errors. The authors related reference problems to poor scores on topic maintenance and turn-taking contingency on the pragmatic protocol. They also implicated word r e t r i e v a l problems: unspecified pronouns may have been used i n place of a s p e c i f i c l e x i c a l item as a strategy to prevent complete breakdown i n communication. Use of cohesion was related to syntactic a b i l i t i e s i n the case of one of the CHI subjects who showed a more re s t r i c t e d syntactic p r o f i l e and who also was more r e s t r i c t e d i n the range of ti e s used i n the categories of conjunction and l e x i c a l cohesion. In summary, the authors concluded that the cohesive patterns of the CHI subjects were related to reduced l i n g u i s t i c processing a b i l i t i e s , l imited 1 7 T h i s finding i s i n agreement with Rochester & Martin's (1979) results for narratives but not for conversation. pragmatic a b i l i t i e s , and compensatory strategies (e.g. use of partner's speech as scaffold for one's own). Cohesive patterns also differed according to discourse type. While both groups showed s e n s i t i v i t y to d i f f e r i n g discourse conditions (both showed higher percentages of e l l i p t i c a l and r e f e r e n t i a l t i e s i n conversation and higher percentage of l e x i c a l and conjunction t i e s i n narratives), the normals showed one important difference: they used more cohesive t i e s i n narratives compared to conversation; the CHI did not d i f f e r i n amount of cohesive t i e s across contexts. Two cautionary notes are i n order. F i r s t , Mentis & Prutting did not use s t a t i s t i c a l tests, no doubt because of the small number of subjects. This must be kept i n mind when interpreting comparative terms such as more, less, lower, higher, clearly differentiated, etc. A few of their findings for a given group applied to only two out of the three members. Second, sp e c i f i c ways i n which results on the pragmatic and syntactic measures related to cohesion results were not c l a r i f i e d . The pragmatic protocol examined a wide range of behaviors: speech acts, topic maintenance, turn taking, l e x i c a l selection, p a r a l i n g u i s t i c aspects, nonverbal aspects, etc. The intent of the authors i n r e l a t i n g the cohesion patterns of the CHI subjects to reduced pragmatic a b i l t i e s i s unclear. Is one a source of problems for the other, and i f so, which causes which? Or are the two evaluating the same phenomena? School-age children An extensive cohesion analysis was done by L i l e s (1985,1987) on the story r e c a l l s of 20 normal and 20 language-disordered children (aged 7 to 10). Following the viewing of a f i l m with the examiner present, the c h i l d was asked to t e l l the story to the examiner (shared condition). The c h i l d then retold the story to an adult l i s t e n e r who had not seen the f i l m (unshared condition). After the r e t e l l i n g s , the examiner asked the c h i l d two sets of questions: set I was intended to evaluate factual information and set II questions were designed to tap the subject's knowledge of story grammar (how, why questions). When the data were compared across l i s t e n e r conditions, differences were found in the number of sentences used, use of cohesive categories, and adequacy of cohesion. Both groups produced more sentences, more personal reference and conjunctive t i e s , and a greater proportion of complete t i e s with the unin-formed l i s t e n e r . A l l children, then, viewed the task as a communicative one and were sensitive to l i s t e n e r needs. With respect to group differences, the normal children used a greater number of sentences and a greater percentage of personal reference t i e s than the language disordered subjects, while the l a t t e r used a greater percentage of demonstrative reference and l e x i c a l t i e s . As L i l e s noted, proper use of personal reference requires ide n t i f y i n g the referent i n varying contexts and results i n a complex integration of ideas. The language disordered c h i l d , she suggested, was perhaps less able to "chain", to id e n t i f y the same referent across several events. An alternative she proposed was that the disordered children were constrained at the sentence l e v e l i n the grammatical use of pronouns. A l l of her disordered subjects had a history of late acquisition of pronominal use; they may have developed at an early age a compensatory strategy of replacing pronouns with demonstratives or l e x i c a l items at the le v e l of discourse. Adequacy of cohesion also distinguished the two groups; the language disordered subjects used a greater percentage of both incomplete and erroneous t i e s . Regarding comprehension, both groups did well on set I questions but the disordered subjects did more poorly than the normals on set 67 II questions. L i l e s concluded from this that the disordered children did not have as great a knowledge about story grammar. Age did not correlate with any of the variables measured and L i l e s concluded that s t a b i l i t y of use of cohesion i s attained by age 6. In the absence of adult data, however, this conclusion i s not well supported by her study. L i l e s has obtained interesting r e s u l t s , perhaps the most revealing being that both groups of subjects altered their use of cohesion as a function of li s t e n e r condition i n the same ways. Other results and/or their interpreta-tion require further scrutiny. L i l e s ' finding of greater use of l e x i c a l cohesion on the part of language disordered subjects versus normal speakers c o n f l i c t s with the results of Rochester & Martin (1979) and Mentis & Prutting (1987). Both of the l a t t e r studies found a lower proportion of l e x i c a l cohesion by their language disordered subjects than normal speakers i n narrative contexts. What makes L i l e s ' findings p a r t i c u l a r l y open to question i s her description of how l e x i c a l cohesion was analyzed (see her appendix B, p.133). As described i n the previous section, l e x i c a l cohesion i s achieved through vocabulary selection. According to Halliday & Hasan (1976), repeti-tion represents the most important subcategory of l e x i c a l cohesion since i t t y p i c a l l y occurs far more frequently than other l e x i c a l devices. In the following pair of utterances, balloons i n (a) and (b) would be considered to comprise the elements of a cohesive t i e : 3.27 a. The children bought balloons, b. Then the balloons popped. Such an example, i n fact, would represent the "prototypical" form l e x i c a l cohesion would take. L i l e s , however, used t h i s example to i l l u s t r a t e instan-ces that were excluded. Her rationale was that the use of balloons i n 3.27 68 "does not necessarily require the reader to look outside the sentence for additional information" (p.133). L i l e s ' decision ignores the c r u c i a l d i s t i n c -tion between the rather closed set of "grammatical" cohesive devices on the one hand (reference, e l l i p s i s / s u b s t i t u t i o n ) and the more open class of l e x i c a l cohesive devices on the other. While "grammatical" devices require a li s t e n e r to seek information elsewhere, l e x i c a l devices are semantically s e l f - s u f f i -cient. They nevertheless play a cohesive role. The decision to exclude items which did not require the l i s t e n e r to look outside the sentence led to poor r e l i a b i l i t y , as L i l e s herself noted. It also led to a lower frequency of cohesive t i e s o v e r a l l , a smaller proportion of l e x i c a l cohesion and i n turn higher proportions of other cohesive categories than might have been expected. A second c r i t i c i s m concerns L i l e s ' conclusion that the poorer scores by language disordered subjects on set I I of the comprehension questions r e f l e c -ted poor grammar knowledge. Set I I questions were no doubt more d i f f i c u l t and required longer answers, but i t i s questionable whether any kind of question-answer task can d i r e c t l y tap story grammar knowledge. The examiner, with his or her question, provides the subject with that particular category of the story's grammar. Not a l l of the language disordered group did poorly on set II questions; f i v e obtained a perfect score. The cohesive adequacy of these f i v e was s t i l l poorer than the normals; one, i n fact had the highest percen-tage of erroneous and incomplete tying. L i l e s concludes that although poor story grammar (as supposedly shown by the lower mean on set I I questions of the disordered group) i s a predictor of poor cohesion, good story grammar knowledge i s not s u f f i c i e n t for good cohesion. Unfortunately, neither of these statements i s supported by her experiment since story grammar knowledge was not adequately assessed. 69 Ripich & G r i f f i t h (1988, see also G r i f f i t h , Ripich & Dastoli 1986) analyzed story structure, propositions, and cohesion i n the discourse of normal and learning disabled children aged 7 to 12. According to standardized test r e s u l t s , the learning disabled appeared to be l i n g u i s t i c a l l y competent at the sentence l e v e l . Three subgroups were formed based on vocabulary age. Data were obtained from three story r e t e l l tasks (which differed according to re l a t i v e d i f f i c u l t y based on number of events), and from a fourth task involving picture book narration. Nondisabled subjects used more cohesive devices than disabled subjects, including more reference items and conjunc-tions. These differences were also found between older and younger children. Nondisabled subjects did not d i f f e r s i g n i f i c a n t l y from disabled with respect to cohesion errors. Disrupted cohesion varied across contexts, however, with the story of medium d i f f i c u l t y showing the most reference disruptions. The authors attributed t h i s finding to the fact that this story contained two female protagonists who were sometimes confused. Results and interpretations are once again compromised by questionable a n a l y t i c a l decisions. Ripich & G r i f f i t h omitted the category of l e x i c a l cohesion altogether; no reference i s made to i t so i t i s impossible to know why the researchers chose to exclude i t . Their conclusions that e l l i p s i s was used i n the same way by both groups of children i s not well supported. No figures were provided for e l l i p s i s i n their tables. Since very short stories were used as s t i m u l i , i t i s l i k e l y that their conclusions were based on a small number of tokens of e l l i p s i s , perhaps one or fewer per story. A more serious problem concerns a lack of control for story length. Although they claim to determine "frequency and percentage of cohesion devices" at the outset of their analysis section, no attempt was made to actually do t h i s . In 70 spite of the fact that the learning disabled children used fewer propositions i n their story than the normal children, use of cohesion by the two groups was compared i n absolute terms only. A t h i r d study of discourse cohesion i n school-age children was carried out by Harris, Schaeffer, 4 Lougeay-Mottinger (1987). They examined conversa-tion and monologues ( r e t e l l and procedural) of language impaired children aged 7 to 9, age and gender matched normals (CA), and younger controls matched on expressive language a b i l i t i e s (LA). Use of cohesive devices differed more across contexts than across groups. A l l children used a greater proportion of contextually-based devices, i . e . exophoric as opposed to endophoric, i n the play task (conversation). The CA children were most sensitive to context; they used more demonstratives and e l l i p t i c a l devices i n play to refer to shared physical context 1 8 and more conjunctive devices i n monologues. According to the authors, although the language impaired and language matched children had the devices i n their l i n g u i s t i c repertoire, they apparently possessed less knowledge as to how to vary the use of devices across contexts. The language impaired children, unlike the other groups, did not d i f f e r i n their use of e l l i p s i s across contexts; their continued high use i n non contextually-supported tasks suggested problems i n al t e r i n g use of cohesive devices according to l i s t e n e r perspective. Complete devices were infrequent; a l l children used more incomplete t i e s on the story r e c a l l task, a finding which suggested that the children had more d i f f i c u l t y when no physical context was shared. Language impaired children used a greater proportion of incom-plete devices on the story task than the CA children. Unfortunately, Harris 1 8Note that such devices, being exophoric, would not be considered cohesive by Halliday 4 Hasan (1976) and by those who have adopted their analysis. 71 et a l . , l i k e Ripich & G r i f f i t h (1988), did not examine l e x i c a l cohesion nor provide a reason for this omission. I t i s also d i f f i c u l t to know how they would have distinguished between devices used l i n g u i s t i c a l l y and those used contextually during the play task since the referents (a variety of toys) were physically present. Purpose of the Present Study Many of the older children i n the above studies appeared to perform "adequately" on standardized tests tapping sentence-level phenomena. Teachers continue to refer such children to language c l i n i c i a n s , who, unfor-tunately, lack adequate diagnostic tools. The problem, i f there i s one, i s usually f e l t to l i e somewhere beyond sentential borders, and the solution i s often to describe, i n rather anecdotal fashion, a sequence of "pragmatic behaviors". Pragmatic inventories have not proved helpful i n this respect. They have lumped together a wide variety of behaviors using categories that are rarely r e l i a b l y coded. In spite of the length of some, much of the text remains unanalyzed. A more principled account of discourse l e v e l relations i s required, one that looks more closely at language, i . e . at the l i n g u i s t i c resources operating at the l e v e l of discourse. The aim of cohesion analysis as developed within the framework of systemic grammar i s precisely t h i s . Unfortunately, attempts to analyze cohesion i n the texts of language impaired school-age children have been, for the most part, unsatisfactory. Findings of certain studies remain inconclu-sive due to a lack of consideration of story length (e.g. Ripich & G r i f f i t h 1988), misinterpretation of certain cohesive categories (e.g. L i l e s 1985), omission of important cohesive categories (e.g. Ripich & G r i f f i t h 1988, Harris 72 et a l . 1987), and unsupported theoretical assumptions (Ripich & T e r r e l l 1988). In view of these deficiencies and of the c o n f l i c t i n g results i n the few studies that have been undertaken to date, additional research on cohesion i n the discourse of language impaired individuals i s required. The overall purpose of this study w i l l be to characterize some of the ways i n which language impaired children create, or f a i l to create, cohesive discourse. The questions addressed are the following: 1. Does the use of cohesion by language impaired school-age children d i f f e r quantitatively and q u a l i t a t i v e l y from i t s use by age-matched normal language children? 2. Does the use of cohesion d i f f e r depending on context? Are some categories more sensitive to contextual effects than others? Do the groups d i f f e r with respect to the pattern of cohesion used across contexts? The following n u l l hypotheses were tested: Amount of cohesion (4) Given the same task conditions, language impaired children do not d i f f e r from normal language children i n the frequency of cohesive t i e s as measured by a) the number of t i e s per story and b) the number of t i e s per independent clause. (5) The frequency of cohesive t i e s (per independent clause) for either group does not vary as a funcion of context. Distribution of cohesive ties (6) Given the same task conditions, language impaired children do not d i f f e r from normal language children i n the d i s t r i b u t i o n of cohesive types. S p e c i f i c a l l y , the two groups do not d i f f e r i n the proportion of t i e s belonging to the following categories or their subtypes: a) reference, b) e l l i p s i s / s u b s t i t u t i o n , c) conjunction and d) l e x i c a l cohesion. (7) The d i s t r i b u t i o n of cohesive types for either group does not vary as a function of context. S p e c i f i c a l l y , each of the following catego-ri e s and their subtypes occur i n the same proportions across contexts: a) reference, b) e l l i p s i s / s u b s t i t u t i o n , c) conjunction and d) l e x i c a l cohesion. 73 Cohesive adequacy (8) Given the same task conditions, language impaired children do not d i f f e r from normal language children i n the proportion of incomplete t i e s as a function of t o t a l cohesive t i e s . (9) Cohesive adequacy for either group does not vary as a function of context. CHAPTER FOUR COHESION: ANALYSIS Method of Analysis Cohesion was analyzed using an adaptation of Halliday & Hasan's (1976) method. Three narratives from each child, obtained from TI, T2, and T3, were analyzed, yielding a data base of 662 independent clause units for TI, 473 for T2, and 676 for T3, for a total of 1811 independent clause units. Each independent clause was examined for links to previous utterances.1 Presumed items were noted. For example, they, in 4.1 was counted as a single cohesive element and father and daughter were noted as the presumed items: 4.1 one day there was a father and a daughter that was going on a trip/ and they went to an island The scanning of previous independent clauses was in keeping with Halliday & Hasan's decision to study cohesive ties between sentences only. However, it immediately became clear that a considerable amount of information relevant to cohesion was being excluded (see discussion under conjunction in chapter three). Therefore, a referring item was counted as a cohesive device not only if its linguistic referent could be found in a previous independent clause, but also i f i t could be found in a subordinate clause or conjoined clause with subject ellipsis (see Martin 1977, and Pappas 1981, for similar departures from Halliday & Hasan's analysis). For example: 4.2 and a l l the people gathered in and rowed down to the island 1The term utterance, when used, refers to an independent clause as defined within sytemic grammar (see chapter three). 74 75 According to the procedure adopted here, i t a l i c i z e d and and the zero anaphor that immediately follows i t were included i n the analysis, i . e . counted as cohesive devices. A note was made of such instances when they occurred. Each l i n k was coded for type of cohesive device used: reference, s u b s t i t u t i o n / e l l i p s i s , conjunction, l e x i c a l , and incomplete. Definitions and examples of the f i r s t four types were provided i n chapter three (see table IV for examples from the data). A f i f t h category, incomplete, was added to allow coding of instances where a referring item was used i n the absence of a prior referent. Erroneous and ambiguous uses were also placed i n t h i s category. Devices were considered ambiguous i n the following cases: 1) i f i t was impossible for the experimenter (who knew the story well) to determine which referent was intended or 2) i f there was a f i f t y percent chance or greater that the l i s t e n e r would misinterpret the device. The l a t t e r involved cases judged to be ambiguous from a l i s t e n e r ' s standpoint. If there was any way a device was more l i k e l y to be interpreted correctly than inc o r r e c t l y (based, for example, on a l i s t e n e r ' s use of world knowledge or of what had been predicated so f a r ) , the speaker was credited with a complete device. Group and task effects were tested using the Mann-Whitney U and Friedman two-way analysis of variance by ranks respectively. Two-tailed tests were used throughout. Mann-Whitney U results have been included i n i n the tables. For a difference to be considered s i g n i f i c a n t at the .05 l e v e l , the value of U had to be equal to or less than 5. The smaller U was, the more s i g n i f i c a n t the difference between the sum of the ranks for each group. Results of the Friedman two-way analysis of variance by ranks have been noted i n the text. To be considered s i g n i f i c a n t at the .05 l e v e l , the value of Fr had to equal or exceed 7.0. The greater Fr, the more s i g n i f i c a n t the difference. 76 Table IV.—Examples of Cohesion Categories from the Data Category Subcategory Examples Reference Pronominal Demonstrative Comparative Ellipsis/ Nominal Substitution Verbal/clausal Conjunction Lexical Incomplete Additive Temporal Adversative Causal Continuative Repetition (with identity) Synonymy Repetition (without ident.) Verbal Additive Erroneous Ambiguous the story was about two kids/ and they were riding down and then they came to a big suspension bridge/ they went down that they come to an island / and they send a motor boat out to get to the island quicker and one of the bomb people took the g i r l / .../ and the other guy asked Todd i f he could get the backpack off one of them wasn't really sure about i t / but the other one really wanted to go (S) and he wanted to go to get his backpack again / but they didn't (S) the boy Todd said that he wasn't scared / but a l l the rest were (S) they get the ape to follow them / and they run off they went down that / then they went through a big tunnel and then the g i r l showed them where the time bomb was / but i t was gone she didn't want him to do i t / so they started riding their bikes back home they saw an ape swinging on a swing hanging on the log /.../ then he started swinging again and then they see a village / and in the village they see a l l these indians and they found this big t r a i l on a mountain /.../ and the people followed the big path and he found a bomb just sitting on the ground /.../ and he called into this bomb service or something like this he opened i t /.../ then he opened i t more and more and the g i r l f e l l off a c l i f f (no previous mention of a girl) once upon a time there were some people, well, three boys and a l i t t l e g i r l / but they were trying to find the legend of the Great Ape (conj. error) so the man helped him [the boy] take off his knapsack / and then he put i t down gently Note: See chapter three for definitions. 77 Results Amount of Cohesion Amount of cohesion was evaluated i n terms of t o t a l number of t i e s used and number of t i e s per independent clause. Means, standard deviations, and ranges for each group are provided i n table V along with Mann-Whitney U results. Normal language (NL) speakers used s i g n i f i c a n t l y more t o t a l cohesive t i e s on Tl and T3 than language impaired (LI) speakers. Given task and group differences i n story length (see chapter two), the t o t a l number of t i e s per speaker was divided by the number of independent clauses each speaker produced (see table V). Frequency of cohesion i n terms of ti e s per independent clause differed s i g n i f i c a n t l y across groups on Tl (U=A, p<.05) and highly s i g n i f i -cantly on T3 {U=2, pi.Ol). The difference on T2 was nearly s i g n i f i c a n t (U=6, p<.10). Hyothesis (4), therefore, was rejected. Although the size of the difference between the two groups varied depending on the task, each group considered on i t s own performed i n a similar way across tasks with respect to number of t i e s per independent clause: LI subjects used an average of 3.39, 3.59 and 3.18 t i e s per independent clause on T l , T2, and T3 respectively while NL subjects used an average of 4.36, 4.44, and 3.90 t i e s on the same tasks. Hypothesis (5), therefore, could not be rejected based on the data obtained. In order to determine the effect of the decision to include instances of cohesion where one element of the t i e occurred i n a main clause and the other i n a subordinate clause or conjoined clause with subject e l l i p s i s , these instances were excluded and the data reanalyzed. Total t i e s and t i e s per independent clause were, as expected, reduced for a l l subjects on a l l tasks. 78 Table V.—Frequency of Cohesive Ties used by NL and LI Children on Tl, T2, and T3 Tl T2 T3 (picture-based) (retell,no pictures) (video recall) LP NL LI NL LI NL Cohesive ties Ties/story M 173.2 256.3 125.0 200.3 131.5 283.0 SD 37.3 72.8 63.5 64.7 58.6 74.6 range 132-217 146-348 64-224 101-274 82-230 162-356 Ties/independent clause M 3.39 4.36 3.59 4.44 3.18 3.90 SD .29 .72 .41 .62 .39 .26 range 3.00-3.69 3.55-5.30 3.11-4.19 3.61-5.27 2.65-3.79 3.50-4.12 aN= 6 for each group Mann-Whitney U values Tl T2 T3 Ties/story 4.0* 6.0 1.0** Ties/independent clause 4.0* 6.0 2.0** *p(.05 **p<.01 However, the size of the differences between groups was comparable to the findings of the f i r s t analysis. Sums of ranks were almost i d e n t i c a l to those previously found. For these reasons, i t appeared be n e f i c i a l to include such instances i n order to take into account as much of the data judged relevant to cohesion as possible. In the following analyses, the data include these instances of cohesion unless otherwise stated. Distribution of Cohesion Categories For each subject, the proportion of t i e s used i n each cohesion category was computed for each context. The d i s t r i b u t i o n of t i e s i n the categories of 79 reference (R), conjunction (C), l e x i c a l cohesion (L), and incomplete (I) are depicted i n figure 2. The category of s u b s t i t u t i o n / e l l i p s i s was not included since i t represented one percent or less of t o t a l cohesive t i e s for both groups on a l l tasks. Means, standard deviations, and ranges are presented i n tabular form i n appendix 3. There i s only one category, that of incomplete, i n which the groups diffe r e d s i g n i f i c a n t l y on a l l tasks. The difference was highly s i g n i f i c a n t on each task (U=0, pi.01) with NL children showing a much smaller proportion of incomplete t i e s than LI children. Hypothesis (8) was, therefore, rejected. A highly s i g n i f i c a n t difference was also found between the two groups i n the category of l e x i c a l cohesion on T2 (U=l, pi.01); NL children depended more on l e x i c a l cohesion i n th i s task than LI children. Hypothesis (6d) was rejected, therefore, for T2. Similar tendencies were evident on Tl and T3; the d i f f e r -ence on T3 was nearly s i g n i f i c a n t (U=l, p<.10). The overall order of category use (excluding incomplete) for NL speakers on a l l narrative tasks was the following: R > L > C > S/E. For LI speakers, the order was the same as NL children on T l but differed s l i g h t l y on T2 and T3 where l e x i c a l and conjunction categories were used i n approximately the same proportions, i . e . R > L = C > S/E. The only s i g n i f i c a n t within group context effect was the use of l e x i c a l cohesion by NL subjects (Fr=l, pi.05), who used a greater proportion of l e x i c a l cohesion on T2 than on T3. Hypothesis (7d) was, therefore, rejected for NL children. There were other tendencies which nearly reached s i g n i f i -cance: 1) proportion of conjunction by LI children was greater on T2 than on T l ; 2) proportion of conjunction by NL children was greater on T3 than on T2; 3) proportion of incomplete t i e s by NL children was greater on T3 than on T l . 80 * * * * * « * « * * « * * * * * * * - * ; * « « « « * « * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * x x: 4) O * * * * « * « « * * * * * « * « « « * « * « * * * < N l * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * « * « « * * « « * * « * * « * * * « * * * * • 11 <U • W H O Mean Proportion of Total Cohesive Ties 81 The cohesion categories of reference, conjunction, lexical, and incom-plete were further subdivided in order to investigate more fully the nature of the cohesive ties used. These results are presented below. Reference: As figure 2 shows, the proportion of total referential ties did not differ across groups or contexts. Referential ties were further subdivided into the following categories: personals, demonstratives, and comparatives (see chapter three for definitions and examples and table IV of this chapter). Results are shown in table VI. Comparative reference was not compared statistically since i t was occurred too infrequently. The most striking result is the evident similarities between the two groups in their use of personals and demonstratives. There is a slight tendency for NL subjects to use a greater proportion of demonstratives and a correspondingly smaller proportion of personals than LI subjects on T2. Table VI.—Proportions of Referential Ties falling into the Subcategories of Personals, Demonstratives, and Comparatives for LI and NL children on TI, T2, and T3 • • • • • • I I I . • II I • •». ... ..—•.••H...P Ill llll. — I— » TI T2 T3 (picture-based) (retell,no pictures) (video recall) LP NL LI NL LI NL Referential ties Personals M .599 .616 .628 .568 .701 .676 SD .119 .068 .113 .046 .073 .060 range .418-.746 .492-.667 .423-.737 .512-.645 .569-.780 .596-.752 Demonstratives M .389 .376 .368 .409 .288 .307 SD .122 .068 .113 .044 . .074 .070 range .254-.582 .321-.500 .263-.5T7 .336-.463 .195-.415 .231-.394 Comparatives M .012 .008 .003 .022 .011 .017 aN= 6 for each group 82 With respect to context effects, the NL children used a s i g n i f i c a n t l y higher proportion of demonstratives and lower proportion of personals on T2 than on T3 (Fr=8.33, p<.05). Hypothesis (7a) was, therefore, rejected for the subcategories of demonstratives and personals for NL children. Since there were r e l a t i v e l y few demonstratives of proximity used, this result amounted to a s i g n i f i c a n t l y higher proportion of d e f i n i t e a r t i c l e plus noun constructions. Conjunction: Only e x p l i c i t conjunctions were included i n the analysis. The two groups used almost i d e n t i c a l proportions of conjunction on T l and T3, whereas on T2, the NL children used a s l i g h t l y smaller proportion although the difference was not s i g n i f i c a n t . The data were converted from proportional to frequency data i n order to determine whether the groups diff e r e d with respect to the number of conjunctions used per independent clause. The results are displayed i n figure 3a. A s i g n i f i c a n t difference between the groups was found on T3 {U=i, j?<.05). Thus, although the proportions of conjunctive use on this task are almost i d e n t i c a l for NL and LI subjects (.254 and .250 respectively), the NL children are actually using more conjunctions per independent clause than the LI subjects on T3. It was noted e a r l i e r that a l l conjunctions were analyzed, including those introducing subordinate clauses and conjoined clauses with subject e l l i p s i s . The data were reanalyzed to determine whether the groups diff e r e d i n their use of these "intra-utterance" conjunctions (those introducing subordinate clauses and conjoined clauses with subject e l l i p s i s ) . The frequency of int r a - u t t e r -ance conjunctions per independent clause i s shown i n figure 3b; that of i n t e r -utterance conjunctions (those l i n k i n g independent clauses) i s shown i n figure 3c. Although the differences i n use of intra-utterance conjunctions by the two groups did not d i f f e r s i g n i f i c a n t l y , there was a tendency for NL children 1.2 1.0 .8 .2 .990 .818 TI T2 T3 LI .969 .885 / / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * TI 1 .028 .992 * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * T2 T3 NL .211 .208 .085 TI . 134 .121 / / / * * T2 LI T3 / / -/ / • / 126 734 TI T2 T3 NL .857 / / / .764 / / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * T2 T3 LI .758 a) Tota l Conjunctions b) Intra-ut terance Con junct ions TI c) In ter -u t t erance Con junct ions . 901 * * . 784 * * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * / * T2 T3 NL P i g . 3 . Frequency of Conjunctive Ties per Independent Clause by LI and NL C h i l d r e n on T l , T 2 , a n d T3 : (a) a l l Conjunctions ( Intra-ut terance and I n t e r - u t t e r a n c e ) ; (b) In tra -u t t erance Conjunctions (those in troduc ing subordinate clauses or conjoined clauses with subject e l l i p s i s ) ; (c) I n t e r - u t t e r a n c e Conjunctions ( t h o s e , l i n k i n g independent clause u n i t s ) . CO CO 84 to use more of them than the LI subjects on T l : the three highest ranks were assigned to NL children and the three lowest to LI children. When the frequency of inter-utterance conjunctions only was compared, the groups differed s i g n i f i c a n t l y on T3 (U=4, p<.05) with NL children using a greater frequency than LI children. The higher frequency of conjunctions by NL children than by LI subjects on T3, therefore, does not appear to have been due to a greater use of intra-utterance conjunctives (as on Tl) but to an increased number of inter-utterance conjunctions. Conjunctive t i e s were further c l a s s i f i e d into the subcategories of additive and nonadditive, based on the conjunction used {and versus a l l others); the results are presented i n table VII. NL children used a s i g n i f i -cantly greater proportion of additive conjunctions than LI children on Tl (tf=5, p<.05). When intra-utterance conjunctions were excluded, the difference between the two groups on Tl was no longer s i g n i f i c a n t (although the tendency was s t i l l present), indicating that at least part of the greater additive proportion used by NL subjects was due to internal conjoining of clauses with subject e l l i p s i s . The nonadditive category consisted of temporals, adversatives, causals, and continuatives. 2 In cases of compound conjunction (e.g and then, and so), L i l e s ' (1985) procedure was followed: the most "complex" conjunction was selected for coding according to the following hierarchy (most to least complex): causal, adversative, temporal, additive. The sp e c i f i c conjunctive t i e s used by the two groups and their c l a s s i f i c a t i o n are presented i n appendix 4. No s i g n i f i c a n t differences were found between the two groups with respect to proportions of each subcategory as a function of t o t a l conjunctive t i e s 2See chapter three for d e f i n i t i o n s . 85 Table VTJ. —Proportions of Additive versus Nonadditive Conjunctive Ties for LI and NL Children on TI, T2, and T3 TI T2 T3 (picture-based) (retell,no pictures) (video recall) LP NL LI NL LI NL Conjunctive ties Additive M .429 .630 .407 .564 .462 .500 SD .155 .129 .179 .217 .178 .097 range .250-.676 .472-.773 .143-.647 .282-.844 .197-.692 .348-.643 Nonadditive M .571 .370 .593 .436 .538 .500 SD .155 .129 .179 .217 .178 .097 range .324-.750 .227-.528 .353-.857 .156-.718 .308-.803 .357-.652 aN- 6 for each group Mann-Whitney U values TI T2 T3 Additive 5.0* 10.0 14.0 Nonadditive 5.0* 10.0 14.0 *p<.05 used, nor were significant contextual effects found. Statistical testing was difficult, however, due to the few observations per subject for several of the nonadditive subtypes. This was particularly true of the LI children, many of whom made no use of certain subtypes in any of their stories. The adversative category, for example, was used by only two LI speakers in any given story; each of the NL speakers, on the other hand, used i t in all three stories except for one speaker, who used i t in two out of three of his stories. A second tendency that prevented reliable use of summary measures was individual predilection for a certain conjunction. Individual subject results on T2 are presented in table VIII by way of 86 illustration. As the data show, there are more unrepresented categories on the part of LI subjects than NL subjects. In addition, the maximum proportion of conjunctive ties in any given nonadditive subtype for any NL child is .38 (SC3, temporal) whereas for LI subjects, several exceed the halfway mark. One is as high as .86 (S2, continuatives), a result of a predilection on the part of this child to begin his utterances with now or and now (which accounted for 21 out of his total of 28 conjunctives used on T2). Then and and then were the favorite conjunctions of S3, S4, and S5. Table VIII.—Frequency (and Proportion) of Conjunctive Ties by Individual Subjects on T2 Conjunctive ties Subject Total Conj. ADDrnvE NONADDITIVE Temp. Advers. Causal Contin. SI 17 11 (.65) 5 (.29) - - 1 (.06) S2 28 4 (.14) - - - 24 (.86) S3 48 18 (.38) 27 (.56) - - 3 (.06) S4 32 14 (.44) 16 (.50) - 2 (.06) -S5 52 15 (.29) 28 (.53) 3 (.06) 6 (.12) -S6 24 13 (.54) 7 (.29) 1 (.04) 3 (.12) -SCI 26 19 (.73) 4 (.15) - 3 (.12) -SC2 36 15 (.42) 12 (.33) 2 (.06) 5 (.14) 2 (.06) SC3 39 11 (.28) 15 (.38) 6 (.15) 4 (.10) 3 (.08) SC4 41 18 (.44) 14 (.34) 3 (.07) 2 (.05) 4 (.10) SC5 64 54 (.84) 3 (.05) 1 (.02) 2 (.03) 4 (.06) SC6 61 41 (.67) 18 (.30) - 1 (.02) 1 (.02) Lexical Cohesion: It was noted earlier that NL children tended to use a greater proportion of lexical ties than LI subjects on a l l tasks, the differ-ence being highly significant on T2. These ties were subdivided to determine whether a particular type of lexical tie was responsible for the difference 87 observed between the two groups. The following subcategories were used: 1) repetition with i d e n t i t y of reference; 2) synonymy with i d e n t i t y of reference (including meronyms and use of subordinate or general term); 3) repetition without i d e n t i t y of reference; 4) Verbal (including repetition and synonymy). The f i r s t three categories consisted of nouns and modifiers; verbs were kept separate and not further subdivided. For examples, see table IV. The d i s t r i b u t i o n of l e x i c a l t i e s i n these subcategories i s summarized i n table IX. On T l , s i g n i f i c a n t differences were found between the ranks of the two groups on two subcategories: re p e t i t i o n with i d e n t i t y , and verbal. Hypothesis (6d) was, therefore, also rejected for these two subtypes of l e x i c a l cohesion on T l . LI subjects showed a greater reliance on repetition with i d e n t i t y than did the NL speakers; the difference i n sum of ranks was highly s i g n i f i c a n t {U=0, p<.01). NL speakers, on the other hand, used a s i g n i f i c a n t l y greater proportion of verbal l e x i c a l t i e s than the LI subjects (#=5, p<.05). NL speakers also tended to use a greater proportion of synonymy than LI subjects; the difference was nearly s i g n i f i c a n t (tf=6, p<.10). Similar tendencies were present i n the data of T2 although the differences do not reach significance. The above data were presented i n terms of proportion of l e x i c a l t i e s used. When the number of t i e s per clause was examined, there was an over-whelming tendency for NL speakers to use a higher frequency of a l l types of l e x i c a l t i e s . Even where LI subjects used a greater proportion of a given type of l e x i c a l cohesion than NL subjects, the NL children s t i l l showed a greater frequency of use of that subtype per independent clause. Frequency data are given i n table X for t o t a l l e x i c a l t i e s and for the two major subtypes: repetition with i d e n t i t y , and verbal. Significant or near s i g n i f i -88 Table IX.—Mean Proportions of Lexical Cohesion Subtypes (out of Total Lexical Ties) for LI and NL Children on TI, T2, and T3 TI T2 T3 (picture-based) (retell,no pictures) (video recall) LP NL LI NL LI NL Lexical cohesion subtypes Repetition with identity M .728 .587 .708 .634 .541 .531 SD .047 .051 .095 .095 .111 .090 range .667-.785 .530-.645 .541-.812 .528-.739 .450-.759 .404-.685 Synonyms (and meronyms) M .051 .095 .047 .070 .063 .058 SD .039 .031 .040 .038 .053 .023 range .022-.127 .050-.139 0 -.105 .023-.118 0 -.158 .022-. 077 Repetition without identity M .039 .063 .078 .072 .050 .026 SD .034 .036 .056 .050 .042 .021 range 0 -.087 .027-.128 0 -.164 •016-.142 0 -.100 0 -.054 Verbal lexical cohesion M .182 .255 .167 .224 .346 .385 SD .052 .051 .061 .032 .072 .083 range .109-.250 .206-.345 .095-.250 .182-.264 .222-.409 .233-.471 aN= 6 for each group Mann-Whitney (/Values TI T2 T3 Repetition with identity 0.0** 8.0 15.0 Synonyms (and meronyms) 6.0 11.0 18.0 Repetition without ident. 11.0 17.0 11.0 Verbal lexical cohesion 5.5* 8.0 10.0 *p<.05 **JX.01 89 cant differences were found in the ranks of the LI and NL children, with LI subjects showing lower frequency of use in each case. Table X.—Frequency (per Independent Clause) of Lexical Ties and of Two Lexical Subtypes (Repetition with Identity and Verbal) for LI and NL Children on Tl, T2, and T3 Tl (picture-based) T2 (retell,no pictures) T3 (video recall) LP NL LI NL LI NL Lexical ties Lexical devices/independent clause M 1.03 1.47 1.02 1.55 .86 1.13 SD .14 .42 .31 .32 .24 .08 range .90-1.26 .88-1.86 .59-1.42 1.21-2.04 .64-1.29 .99-1.22 Lexical subtypes Repetition with identity (devices/independent clause) M .75 .87 .71 .98 .48 .60 SD .10 .27 .18 .22 .25 .07 range .65-.91 .49-1.16 •48-.94 .64-1.30 .29-.98 .48-.68 Verbal lexical devices/independent clause M .19 .37 .18 .35 .29 .44 SD .06 .12 .10 .10 .06 .12 range .11-.28 .22-.51 .08-.30 .29-.54 .23-.39 .23-.56 *N= 6 for each group Mann-Whitney Uvalues Tl T2 T3 Lexical devices 7.0 3.0* 6.0 Repetition with identity 12.0 5.0* 6.0 Verbal lexical devices 3.0* 2.0** 5.0* *|X.05 **p<.01 90 Only one significant context effect was found: LI subjects used a greater proportion of verbal lexical ties on T3 than on T2 (fV=9.33, p<.05). Hypothesis (7d) was, therefore, rejected for one subtype of lexical cohesion (verbal) for LI children. A similar difference between T3 and TI was nearly significant so that T3 > T2 = TI. A similar tendency was evident, though not significant, in the data of the NL children. NL speakers also tended to use a higher frequency of repetition with identity on T2 than on T3. Incomplete: As noted, LI children showed a significantly higher propor-tion of incomplete ties than NL children on a l l three tasks: 8 to 11 percent of cohesive attempts on the part of LI children were incomplete whereas only 2 to 4 percent of NL children's attempts f e l l into this category. Incomplete ties were subcategorized into the major cohesion categories: reference, substitution/ellipsis, conjunction, and lexical. Incomplete referential ties represented the largest proportion of incomplete devices for both groups: .684, .714, and .702 for LI subjects on TI, T2, and T3 respectively, and .837, .830, and .718 for NL subjects on the same tasks. These proportions did not differ significantly across groups or contexts. Referential errors were further subdivided into the subcategories of addition (lacking an antecedent), ambiguous, and erroneous. These will be discussed in chapter five. Summary measures and statistical testing were difficult to perform on the remaining subcategories of incomplete cohesion due to the small number of observations, particularly for the NL subjects. The number of occurrences for each group on each task as well as total occurrences summed across tasks are given in table XI. To account for differences in length, the data have been converted to percentages of total independent clauses containing a given error type; the ratio of these percentages, LI versus NL, is also given for each 91 Table XI.—Subtypes of Incomplete Devices used by LI and NL Children on TI, T2, and T3: Number of Errors, Percentage of Utterances Containing each Subtype, and Ratio of these Percentages (LI:NL) NUMBER OF ERRORS TI T2 T3 LP NL LI NL LI NL Subtypes of incomplete devices Referential 66 28 44 26 59 41 Conjunctive 22 - 14 2 15 11 Ellipsis/substit. 11 3 3 3 9 11 Lexical 5 2 0 1 5 0 Percentage of independent clauses attaining error (all tasks) and ratio of these percentages: LI NL LI:NL Referential 22.3% 9.0% 2.5:1 Conjunctive 6.7 1.2 5.6:1 Ellipsis/substit. 3.0 1.6 1.9:1 Lexical 1.3 0.3 4.3:1 aN= 6 for each group subtype. Conjunctive errors were the second most frequent error type for the LI children; they made over five times as many conjunctive errors as NL children. However, many of these errors (30 out of the 51) were made by one LI subject in particular: 16 of his 30 errors were made on TI due to a predilection for beginning his utterances with but. Of the few substitution/ ellipsis errors that were made, the majority occurred in portions of the narrative delivered at a fast rate; many involved inappropriate subject ellipsis. Lexical errors (e.g. the girl when the ape was intended) were rare for either group. 92 Discussion Group Differences in the Use of Cohesion The first question addressed by the present study concerned whether or not the cohesive skills of language impaired children differed from those of normal language peers. The results support an affirmative response for narratives; both quantitative and qualitative differences were found. With respect to quantitative differences, language impaired children used fewer cohesive ties per independent clause than normal language children on a l l tasks; the difference was significant for the picture narration task (Tl), nearly significant for recall without pictures (T2), and highly significant for the video recall task (T3). Rochester & Martin (1979) and Mentis & Prutting (1987) also found that their normal language subjects used a greater frequency of ties per sentence than language disordered subjects in narrative conditions. The results are not as easily compared to studies targeting school-age language impaired populations. Neither Liles (1985) nor Ripich & Griffith (1988) presented their results in terms of frequency of ties per utterance. Harris et al. (1987) failed to find significant group differences in frequency of cohesive devices, but, like Ripich & Griffith's study, their analysis did not include lexical cohesion. Had they included i t , their conclusion may have been different, in light of the differences in use of lexical cohesion by the two groups in the present study. Several factors may have contributed to the lower frequency of cohesive ties found for language impaired subjects. Two of these possibilities relate to utterance and story length: a reduced amount of cohesion may be a result of lower productivity on the part of language impaired subjects. With respect 93 to utterance length, this argument would hold that a tendency on the part of language impaired subjects to express themselves i n fewer words would result in fewer opportunities for scoring cohesive l i n k s . There are several problems with this proposal, however. If a tendency to use shorter utterances were the sole cause of a reduced frequency of cohesive t i e s , one might have expected a closer correlation between amount of cohesion and words per independent clause. However, as seen i n chapter two, although there was a tendency for normal language subjects to use longer utterances than language impaired, the difference did not reach significance on any of the tasks. Given the s i g n i f i -cant differences between the groups regarding amount of cohesion, one would have expected utterance length of the two groups to have diff e r e d to a greater extent. A second problem for the proposal i s i t s f a i l u r e to achieve explana-tory adequacy. The question remains as to why language impaired subjects tend to express themselves i n fewer words. None of the subjects appeared reticent to speak at the time of testing. On the contrary, they were relaxed and eager to begin their s t o r i e s . The lower productivity of language impaired subjects may, i n fact, partly result from a lesser concern for cohesive text or from a reduced a b i l i t y to construct i t . A second productivity-related factor i s story length. As a story gets longer, one might expect that increasingly more of what i s said w i l l relate to something i n the preceding text. I t was noted i n chapter two, for example, that the story length of the two groups differed s i g n i f i c a n t l y on the video r e c a l l task, a task i n which the amount of cohesion between the two groups also differed the most. However, story length does not account for the s i g n i f i c a n t differences i n cohesion observed between the groups on the picture narration task where story length was not s i g n i f i c a n t l y d i f f e r e n t . One would 94 also have expected a greater difference in amount of cohesion on the retell task without pictures since a greater difference in story length was present. The reverse, however, was true: the speakers were less different with respect to amount of cohesion on the retell task, yet more different with respect to story length. Story and utterance length may contribute to differences in amount of cohesion but they do not appear to suffice as sources for these differences. Two additional possibilities may be entertained: language impaired children may not be attempting cohesion as frequently as normal language children and/or they may be failing in some of their attempts. These will be discussed further below. The quantitative difference in amount of cohesion indicates that the discourse of the language impaired subjects was not as cohesive as that of the normal language children. Cohesion encompasses several phenomena, however, and it is necessary to specify which are responsible for the observed differ-ence between the groups. Distribution of cohesion into the subtypes of reference, substitution/ellipsis, conjunction, and lexical cohesion provided additional, more qualitative information which revealed both similarities and differences in the cohesive patterns of normal language and language impaired children. Both groups used a higher proportion of referential ties than any other cohesive type. Substitution/ellipsis represented the smallest propor-tion: 1 percent or less. It occurred on the video recall task more than on the other tasks due to a higher use of reported speech on this task. It appears to be a feature of conversational discourse and will not be discussed further here due to the few instances of use. Proportion of conjunction and lexical cohesion types f e l l between these two extremes with normal language 95 children using a greater proportion of lexical cohesion than conjunction, and language impaired children using similar proportions of each. The individual proportions of the normal language children are very similar to those of the normal adults of Rochester & Martin's study (1979) on a narrative task (see table XII). Table XII.—Mean Percentages of Cohesion Categories in Narratives: Results of the Present Study Compared to the Adults of Rochester & Martin's Study (1979) Rochester & Martin Present Study (ages 8 to 11) (normal adults) NL LI Cohesion categories Reference 46.5% 42.0% 40.4% Lexical 30.3 31.4 25.7 Conjunction 21.1 23.2 24.2 Ellipsis/Subst. 2.0 0.6 0.5 Incomplete ?» 2.8 9.2 aProportion of total incomplete devices was not provided. Percentage of definite nominal groups involving addition (lacking antecedent) was 2.0% in the narrative condition. The reference analysis employed here failed to reliably differentiate the two groups: both used similar proportions of appropriate referential ties on all tasks and similar proportions of the major subcategories, namely personal and demonstrative reference. These findings conflict with those of Liles (1985), who found a greater percentage of personal reference ties and a correspondingly lower percentage of demonstrative reference by normal children than by language disordered children. The opposite tendency, in fact, occurred on T2 of the present study, where normal language children tended to use a greater proportion of demonstrative reference and smaller proportion of personals than the language impaired children. A third category of referen-96 tial ties, comparative reference, was used relatively infrequently by normal language children and rarely by language impaired children. Although too few instances occurred to permit statistical analysis, the difference in absolute occurrences suggests that comparative reference may be a more complex device. In personal and demonstrative reference, the same entity is referred to directly, whereas in comparative reference, the antecedent serves as a frame of reference for a new participant. This relation of contrast may represent a more complex form of reference in that it requires simultaneous consideration of two referents and specification of the relation between them. As reported earlier, language impaired and normal language children used almost identical proportions of conjunctive ties, proportions that are highly comparable to Rochester & Martin's findings for normal adults (see table XII). Frequency analyses, however, revealed differences between the two groups on individual tasks. On the picture narration task (Tl), the tendency for normal language children to use a greater frequency of conjunctions, specifically intra-utterance conjunctions, suggests that the normal language children, more so than the language impaired children, were treating the task as a story-telling task rather than as an individual picture describing task. They immediately connected events depicted by the individual pictures, incorpora-ting them into complex units and making these connections explicit for their listeners. Martin (1977,1983) reported a similar result on his picture narration task: the younger children (aged 6 and 7) used fewer explicit conjunctions to mark relations between clauses than the older children. On retell of the story in the present study (T2), the language impaired were able to add these connectives so that the frequency of conjunctions between the two groups was almost identical on this task. On the video recall task (T3), the 97 normal language children once again were able to make connections between events and to do so e x p l i c i t l y on a f i r s t t e l l i n g , resulting i n a greater frequency of conjunction than that demonstrated by the language impaired children. However, the normal language children tended to use fewer i n t r a -utterance conjunctions on t h i s task than on other tasks, i . e . they were less able and/or w i l l i n g to conjoin events into complex utterances on this task. Their utterances also tended to be shorter on T3. The r e l a t i v e lack of structure provided on t h i s task may be an important factor. On T l , the indi v i d u a l pictures made i t r e l a t i v e l y easy for the c h i l d to i d e n t i f y story components, whereas on T3, the c h i l d had to himself deduce the narrative structure and reproduce i t i n an organized manner. Therefore, although the normal language children made more overall e x p l i c i t connections than the language impaired on the video r e c a l l task, they did not show a greater frequency of intra-utterance conjunctions than the language impaired children. It i s d i f f i c u l t to explain the tendency for the normal language children to use a greater proportion of additive versus nonadditive conjunctions on T l . It may be p a r t i a l l y related to the greater o v e r a l l frequency of conjunctions by normal language children on t h i s task, and partly, as noted e a r l i e r , to the greater frequency of intra-utterance conjunctions by normal language children than by language impaired children on t h i s task. Developmental studies have found sim i l a r r e s u l t s . Bennett-Kastor (1986), for example, found that the proportion of additive conjunctives increased developmentally i n her study of the narratives of normal children aged 2 to 5. Whereas her 2-year-olds tended to use and and then equally as often, her 4 - and 5-year-olds showed a strong preference for the connective and. There was a tendency for the language impaired children of the present study to use a greater proportion of temporal 98 conjunctions than normal language children. Martin (1977,1983) also found a greater reliance on temporal conjunction on the part of his younger subjects compared to older children on a story r e t e l l task. It may be the case that normal language children have more sophisticated l i n g u i s t i c means at their disposal to establish and s h i f t temporal reference (e.g. through the use of more sophisticated markers of tense and aspect) and are not as dependent on conjunction to f u l f i l l these functions as language impaired children may be.1 The use of nonadditive categories by language impaired speakers appeared r e s t r i c t e d : certain categories were overrepresented due to a predilection for a s p e c i f i c conjunction while other categories, adversatives for example, were completely neglected. By comparison, normal language children showed a more evenly distributed p r o f i l e . They also appeared to use a greater variety of conjunctions. I t i s d i f f i c u l t to predict, however, whether language impaired subjects would have used as many different conjunctions as normal language children given i d e n t i c a l story lengths. The r e l a t i v e l y small number of conjunctions i n certain subcategories made appropriate s t a t i s t i c a l analysis impossible. While i t i s essential to examine conjunction i n situations approximating natural conditions, i t may also be useful to devise more contrived tasks such as those used by McClure & Geva (1983; see chapter three of the present study) which could be used together with language sampling. Such a measure might prove to be a more sensitive guage of subtler knowledge and use of conjunction by language impaired and normal language children. The decisions made i n favor of r e s t r i c t i n g the analysis to an examination of e x p l i c i t connectives and to their categorization •See Bamberg (1987) for a developmental study of "temporal reference" i n narratives. 99 based on form imposed additional l i m i t a t i o n s on the present study. The rationale behind these decisions was to avoid undue speculation on the part of the analyzer as to whether or not a l o g i c a l r e l a t i o n was intended by the speake, and, i f so, which type of r e l a t i o n was intended. I t i s evident, however, that information i s lost as a result of these r e s t r i c t i o n s . The mere juxtaposition of utterances, for example, may be cohesive, and a given conjunction may encode a variety of l o g i c a l r e l a t i o n s . The high use of and by normal language speakers on the picture narration task ( T l ) , for example, surely encoded more than a simple additive r e l a t i o n i n many cases. As table XII shows, language impaired children differed most from normal children and adults i n their use of l e x i c a l and incomplete t i e s . With respect to l e x i c a l cohesion, previous studies on language disordered adult populations have reported similar findings: Rochester & Martin (1979) and Mentis & Prutting (1987) found a greater proportion of l e x i c a l cohesion by normal adults i n narrative conditions than by schizophrenic and head injured adults respectively. There are no cohesion studies with school-age children with which the results found here might be compared. Ripich & G r i f f i t h (1988) and Harris et a l . (1987) did not include l e x i c a l cohesion i n the i r analyses. L i l e s ' (1985) r a d i c a l departure from Halliday & Hasan's (1976) analysis of l e x i c a l cohesion makes i t impossible to compare the two sets of r e s u l t s . The results on l e x i c a l cohesion support one of the factors proposed e a r l i e r to account for the reduced frequency of cohesive t i e s on the part of language impaired children. If shorter utterance and/or story length were the only two factors responsible for the l a t t e r finding, one would have expected a l l types of cohesion to have been reduced by a similar amount. This was not the case, since l e x i c a l cohesion was more greatly reduced than other cohesive 100 categories. Since lexical errors were so rare, the reduced proportion and frequency was not due to failed attempts at lexical cohesion. Rather, the language impaired subjects did not appear to be attempting lexical cohesive ties as frequently as normal language children. Subsequent analysis of proportions of lexical subtypes revealed that on the picture narration task (TI), normal language children showed a significan-tly greater reliance on verbal lexical cohesion than language impaired children. Similar tendencies were present on T2. Although the use of verbs as a lexical cohesive device remains unexplored, its role in story coherence has already been alluded to: Bennett-Kastor, for example, described an incoherent story in the following way: Noun phrases are introduced, only to be dropped with the next clause and never appear again; or verb phrases are constantly altered, and show neither literal nor structural repetition, so that there is no continuity from one clause to the next. (1983: 138) In addition to structural VP coherence or parallelism (see chapter three), Bennett-Kastor (1986) examined percentages of predicate types (actives, passives, statives, copulatives), verb structure in terms of number of argu-ments accepted (valency), and semantic classes of verbs. Although her results are not directly relevant to the general analysis adopted here, her study provides a direction for further investigations into the use of verbs in discourse and how this use reflects sensitivity to contextual demands. Cohesive harmony analysis (see chapter three) represents another promising approach in that i t tracks both participants (reference) and events (verbal cohesion), and examines the interaction between the two. In addition to lexical cohesion, the groups differed reliably in the proportion of incomplete ties, with language impaired children showing a higher proportion than normal language children on all three tasks. This 101 supports the fourth factor proposed earlier in an attempt to account for the reduced amount of cohesion shown by language impaired children: not only are they not attempting as many cohesive ties, but they are often failing in the attempts they do make (8 to 11 percent of their attempts were incomplete). Similar findings in othe language disordered populations were reported in the previous chapter (Rochester & Martin 1979, Mentis & Prutting 1987, Ripich & Terrell 1988, Liles 1985, Ripich & Griffith 1988, Harris et al. 1987). By far the majority of incomplete ties in a l l these studies involved reference. Similar findings have been reported in developmental studies: Klecan-Aker & Hedrick (1985), for example, found a higher incidence of inappropriate personal pronoun usage on the part of grade six students when compared to grade nine students on a spontaneous story generation task. Their findings suggest that the development of appropriate cohesive tying had likely not yet plateaued even in the oldest normal language children of the present study. Further reduction in the number of incomplete ties may be expected with time. The finding that language impaired children show a higher proportion of incomplete ties than normal language children—incomplete ties involving reference in particular—is not a surprising one; speech/language clinicians have been reporting similar problems for years. Further study is obviously needed to pinpoint the source of these difficulties. The results of the present study point to at least one contributor. A consequence of a greater use of lexical cohesion by normal language children is that greater lexical support is available to the listener for interpreting referential devices. Lexical items are high in semantic content and carry most of the communicative weight. Lexical support is provided not only by explicit nominals but, as discussed, i t also entails a greater use of verbal cohesion. The referent of 102 a p o t e n t i a l l y amibiguous r e f e r e n t i a l device was often e a s i l y inferred from what was predicated of th i s referent. 2 Normal language children appear to be better able to support their r e f e r e n t i a l devices with l e x i c a l information, p a r t i c u l a r l y l e x i c a l information carried by the verb. Differences in the Use of Cohesion due to Context The second question addressed by th i s study was whether or not the use of cohesion by either group would be sensitive to contextual changes. This question was formulated when the o r i g i n a l intent was to examine both narrative and conversational data. Since only narrative data was analyzed, t h i s question became a secondary one; task effects were expected to be minimal. The lack of such effects was i t s e l f revealing. Amount of cohesion i n terms of frequency per utterance, for example, showed no s i g n i f i c a n t changes across tasks. If frequency of cohesive devices were highly dependent on story length or memory factors, one would have expected task effects to have surfaced. A memory factor, for example, would have distinguished the picture narration task (TI) from the r e c a l l tasks (T2, T3). Task comparisons of the cohesion categories revealed that normal language children used a s i g n i f i c a n t l y higher proportion of l e x i c a l cohesion on T2 than on T3. Because of the lesser amount of structure provided, the video r e c a l l task (T3) was probably the most demanding i n terms of degree of self-genera-tion required. Lexical cohesion appears to have been the cohesive device that was most sensitive to increased task complexity. On T2, the normal language children no doubt benefited from t e l l i n g the story twice; they tended to 2The importance of predication i n interpreting speaker reference w i l l be discussed further i n chapter s i x . 103 refine i t by increasing the proportion of l e x i c a l t i e s . A related finding was a s i g n i f i c a n t l y greater proportion of demonstrative reference (versus pronomi-nal) by normal language children on T2 than on T3. As noted, the category of demonstratives consisted primarily of the use of de f i n i t e a r t i c l e plus noun constructions. The higher proportion of l e x i c a l cohesion on T2 appears to have been a result of the increase i n use of these repeated noun phrases. In support of this was the higher frequency of the l e x i c a l subcategory of repeti-tion with i d e n t i t y (nouns/modifiers) on T2 than on T3; the difference was almost s i g n i f i c a n t . The reduced proportion of the l e x i c a l subtype of repe t i t i o n with i d e n t i t y on the video r e c a l l task resulted i n a correspondingly higher proportion of verbal l e x i c a l t i e s on T3 compared to T2; the difference was s i g n i f i c a n t for the language impaired children and also present, though not s i g n i f i c a n t , i n the data of the normal language children. The actual frequency of verbal cohesion i n t i e s per utterance did not r i s e s i g n i f i c a n t l y on the video r e c a l l task for either group. This suggests that the children may not be using s i g n i f i c a n t l y more verbal l e x i c a l t i e s on th i s task but that their reduced frequency of e x p l i c i t noun phrases (on the part of both groups) i n favor of pronouns means that the l i s t e n e r l i k e l y had to depend more on l e x i c a l cues provided by the verbs. This may be t y p i c a l of action stories i n general, p a r t i c u l a r l y of those involving s t r i c t l y human protagonists. No s i g n i f i c a n t task effects were found with respect to conjunction. Rochester & Martin (1979) and L i l e s (1987) also f a i l e d to find differences i n the proportion or accuracy of conjunction across tasks. Two tendencies were evident i n the present study, however. F i r s t , language impaired children tended to use a greater proportion of conjunction on T2 than on T l . Frequency 104 analyses reflected an increase i n the number of conjunctions per clause, p a r t i c u l a r l y "inter-utterance" conjunctions. It may be that the actual turning over of pictures on TI functioned as a marker of an additive r e l a t i o n for the language impaired children i n place of e x p l i c i t l i n g u i s t i c marking. Another p o s s i b i l i t y i s that the language impaired children may simply have been better able to be e x p l i c i t about relations encoded on a second t e l l i n g . The second tendency that was observed i n the use of conjunction across tasks occurred i n the data of the normal language children. They tended to use a higher proportion of conjunctions on T3 than on T2. Since the actual frequency did not r i s e i n this case, the higher proportion may simply have been a result of using a proportional analysis. As noted, the normal language children used a s i g n i f i c a n t l y higher proportion of l e x i c a l cohesion on T2 than on T3. Frequency analyses supported this finding. The reduced proportion and frequency of l e x i c a l cohesion on T3 would have caused other categories, l i k e conjunction, to be represented i n higher proportions. This underlines the need for and rationale behind using frequency comparisons i n addition to proportional ones. In the i r examination of reference, conjunction, and e l l i p s i s i n stories and converstion during play, Harris et a l . (1987) found that age-matched normal children were more sensitive to contextual changes than language disordered children. L i l e s (1985), however, found that language disordered subjects altered their use of cohesive types i n ways similar to normal children across l i s t e n e r conditions (uninformed versus informed). Some evidence was found i n the present study to suggest that both groups were making changes i n the cohesive devices they used as the task changed, i . e . both showed s e n s i t i v i t y to a change i n contextual configuration. These 105 changes were not always i n the same di r e c t i o n , however. Whereas normal language children used a s i g n i f i c a n t l y higher proportion of demonstrative reference and l e x i c a l cohesion on T2 than on T3, the language impaired children did not s i g n i f i c a n t l y a l t e r their use of these devices across tasks. The language impaired children, for their part, increased their use of conjunction on T2 compared to Tl whereas normal language children showed similar frequencies of conjunctions across tasks. The data suggest that the language impaired and normal language children may be responding somewhat d i f f e r e n t l y to changes i n contextual configuration. Task d i f f i c u l t y may be a factor leading to these d i f f e r e n t i a l responses. The results underline the need to examine more contrasting conditions i n order to determine the extent of these d i f f e r e n t i a l responses. Rochester & Martin (1979) found that the frequency of cohesive t i e s distinguished schizo-phrenic from normal language speakers i n narratives but not i n conversation. Mentis & Prutting (1987) reported a similar finding with head injured patients. The l a t t e r study also reported a greater proportion of l e x i c a l cohesion than reference by a l l subjects, a finding that i s l i k e l y related to the fact that the "narratives" used did not involve s t o r y t e l l i n g but rather procedural discourse. Statements about general a b i l i t i e s c l e a r l y cannot be based on findings from any one s p e c i f i c context when describing discourse-related s k i l l s . In view of the above comments, the findings reported here may generalize only to similar task conditions. On picture narration, r e c a l l without pictures, and video r e c a l l one might expect to find language impaired children using a lower frequency of cohesion than age- and gender-matched normal language subjects. Four factors were cited i n accounting for t h i s observa-106 t i o n . The f i r s t two were related to length; a tendency to use shorter utterances and to t e l l shorter s t o r i e s , a r i s i n g perhaps from a general tendency away from verbal encoding, may influence frequency of cohesive t i e s per utterance. Both, however, were shown to be inadequate on their own to account for the reduced instances of cohesive t i e s on the part of language impaired children. A t h i r d factor made the simple claim that language impaired children were not attempting cohesive devices as frequently as normal language children. Support for this was found i n the greater proportion and greater frequency of l e x i c a l t i e s on a l l tasks and greater frequency of conjunctions on T3 by normal language children than by language impaired children. Support was also found for a fourth factor, namely that language impaired children more often f a i l e d i n their attempts at cohesion than normal language peers did. The proportion of incomplete cohesive t i e s (missing antecedent, amibiguous or erroneous device) used by language impaired children was s i g n i f i c a n t l y higher than that of normal language children on a l l tasks. The lower reliance on l e x i c a l cohesion and higher incidence of incomplete t i e s on the part of language impaired children would p a r t i a l l y account for pragmatic d i f f i c u l t i e s often described anecdotally, for example, topic maintenance problems. But what accounts for the lower frequency of l e x i c a l cohesion and the higher incidence of incomplete ties? I t was suggested e a r l i e r that the two may be related, s p e c i f i c a l l y that the higher incidence of incomplete t i e s , at least the ambiguous ones, might be due to reduced l e x i c a l support for the interpretation of r e f e r e n t i a l devices. A reduced amount of l e x i c a l cohesion has also been related, i n the case of head injured subjects, to word r e t r i e v a l problems (Mentis & Prutting,1987). A review of the vocabu-lary s k i l l s of the language impaired children used i n the present study as 107 reported i n case h i s t o r i e s indicated that three were functioning i n the average range and three i n low average range (based on PPVT, EOWPVT, TOLD-P re s u l t s ) . None of the children were described as having s i g n i f i c a n t word r e t r i e v a l problems. Further assessment would be required to investigate this p o s s i b i l i t y . The problem, i n any case, i s not l i k e l y to be solely one of accessing. A l l children included verbs of some kind i n their utterances. The normal language children, however, more frequently chose verbs that had a cohesive function. Although the language impaired children had h i s t o r i e s of l i n g u i s t i c processing d i f f i c u l t i e s at the word or clause l e v e l , they a l l formulated grammatically appropriate sentences with r e l a t i v e ease. This i s not to suggest that sentence-level d i f f i c u l t i e s had e n t i r e l y disappeared; more detailed investigation would probably show this to be an incorrect assumption. However, the main problem appeared to l i e i n using their l i n g u i s t i c resources for discourse-level functions. The implications for assessment and therapy w i l l be discussed i n chapter seven. None of the cohesive studies carried out to date has managed to come up with a convincing explanation for language impaired children's cohesive d i f f i c u l t i e s . The present study, too, was not designed to determine cause but to c l a r i f y whether or not the use of cohesive devices by language impaired children di f f e r e d from those of normal language children and, i f so, i n what s p e c i f i c ways they d i f f e r e d . The explanation for the problem i s l i k e l y to involve more than one source. Nor i s a single explanation l i k e l y to hold for different language disordered populations. The language impaired children of this study, for example, did not show p r o f i l e s that were i d e n t i c a l with those of the alzeimer patients of Ripich & T e r r e l l ' s study (1988), the schizophrenic subjects of Rochester & Martin's study (1979) or the head injured of Mentis & 108 Prutting's study (1987). F i n a l l y , given the breadth of the analysis, i t i s not l i k e l y that a single explanation could account for problems i n a l l areas of cohesion; use of reference no doubt requires different l i n g u i s t i c s k i l l s than the use of conjunction or l e x i c a l cohesion. One of these phenomena, namely reference, w i l l be examined i n more d e t a i l i n the following chapters. CHAPTER FIVE REFERENCE: RETRIEVAL ANALYSIS Introduction As Brown & Yule (1983) noted, a sentence grammarian describes r e g u l a r i -t i e s i n his or her data i n s t a t i c terms, i . e . independently of speaker and receiver behavior. Those researchers turning to the study of texts are immediately faced with the inadequacy of tools they have brought with them from "synoptic generative models" (Martin 1985: 272). Moving from the analysis of sentences and/or their constituents to the analysis of discourse i s not easy. Brown & Yule's solution i s to adopt an exclusively "discourse-as-process" view, an approach that may not be p a r t i c u l a r l y helpful for c l i n i c i a n s seeking a means of care f u l l y assessing recorded "outputs" or the products of children or adults referred to them. A more useful approach would examine a text both as product (as an output that can be recorded and studied) and as process (as an interactive event). Existing analyses vary i n how " s t a t i c " versus "dynamic" an approach i s taken. While both perspectives must be kept i n mind i n any analysis, some l i n g u i s t i c phenomena may be more suitably assessed by one approach as opposed to another. Martin (1985) suggests, for example, that conversational struc-ture may be a more prominently synoptic system than reference, and conversely, reference a more dynamic system. Both approaches were used i n the present study. Cohesion was analyzed using what Brown & Yule (1983) described as a "text-as-product" approach. In the next two chapters, reference w i l l be 109 110 analyzed from a "discourse-as-process" perspective. The recorded sample remains the basis of analysis. However, the observed l i n g u i s t i c phenomena are approached as re a l i z a t i o n s of dynamic processes taking place i n time. Reference Network An example of a "dynamic" approach to reference i s an account of the retrieval processes speakers and lis t e n e r s make use of when ident i f y i n g participants. A schematic representation of the reference network (adapted from Martin 1983,1985) i s given i n figure 4.1 Martin himself i s aware of i t s inadequacies: " i t describes i n what w i l l probably turn out to be only a crude and p a r t i a l way the operations a l i s t e n e r might perform i n locating the referent of a phoric nominal group" (1985: 271). Reference, as defined by this approach, i s the semantic system which enables participants to be i d e n t i f i e d i n discourse; i t i s that part of the semantics of English concerned with how nominal groups are structured to create text. Reference— •Presenting •Context of culture (homophora) within nom.groups JoUowing (esphora) (cataphora) -between nom.groups Presuming [-Verbal (endophora) •Preceding (anaphora) C^ontext of situation ^Nonverbal (exophora) (bridging) Fig.4. Reference Network in English. Adapted from Martin (1983,1985). ^or a more detailed representation, see Martin (1977). I l l The reference network distinguishes between participants introduced to the text (presenting items) and those the speaker assumes his or her listener knows (presuming items). Presenting items usually take the form of in d e f i n i t e nominal groups whereas presuming (or phoric) items are t y p i c a l l y demonstra-ti v e s , the de f i n i t e a r t i c l e , and pronouns. Phoric items instruct a li s t e n e r to "search elsewhere" for their interpretation. The location of "elsewhere" may be the context of culture (homophora) or the context of si t u a t i o n . Homophora involves reference to an entity which i s known to members of a community (e.g. the sun). A family may i t s e l f comprise a "community": 5.1 (one spouse to another) Where's the brat now? In 5.1, the brat i s readily i d e n t i f i e d by the l i s t e n i n g spouse to refer to their rambunctious son. Information may also be presumed from the immediate context i n which the utterance occurs. A speaker uses exophoric reference to refer to a p a r t i c i -pant who i s present i n the nonverbal context and endophoric reference to refer to information supplied by the verbal context. In the l a t t e r case, the presumed information usually precedes the phoric item (anaphora), but i t may also follow (cataphora). Cataphora within a nominal group i s known more precisely as esphora, for example, the attainment and the group i n 5.2: 5.2 Tile attainment of their goal suddenly seemed impossible to the group of t i r e d climbers. A second d i s t i n c t i o n within endophora i s made between reference i n which the i d e n t i t y of a participant i s retrievable from the e x p l i c i t verbal context (direct) and that which requires the l i s t e n e r to infer from the i m p l i c i t verbal context (bridging): 5.3 I hurt myself playing tennis. I f e l l against the net and somehow managed to twist my ankle. 112 The id e n t i t y of the net i s inferrable from tennis. It would, i n fact, be odd to use an in d e f i n i t e a r t i c l e here; the l i s t e n e r would be inclined to wonder what kind of net was being referred to. A bridged item may be a necessary component of i t s antecedent as the net i s of tennis i n 5.3 or i t may be highly plausible, as the trail i n the following example from the data: 5.4 and the people with the big banana ran down the mountain/. . ./ and the l i t t l e people are s t i l l running down the trail Previous Findings Martin included r e t r i e v a l analysis i n his investigations into the discourse of normal children (1977,1983) and of schizophrenic adults (Roches-ter & Martin 1979). His findings on conjunction i n children and on cohesion i n schizophrenic and normal adults were reported i n chapter three. His findings on reference, p a r t i c u l a r l y as used by normal school-age children, w i l l be reported here. To review, Martin (1977,1983) examined the spoken narratives of 90 children aged 6 to 11 years. His results on conjunction led him to conclude that the youngest children (6- and 7-year-olds) were using a different coding orientation than the two older groups (8- and 9-year-olds; 10- and 11-year-olds) i n a picture narration task (Task 1) and i n r e c a l l of the same story without pictures (Task 2). His findings on reference supported this conclu-sion. The 6- and 7-year-olds showed greater dependence on presenting groups on Task 1 than the older children: 1) they introduced a number of p a r t i c i -pants that were not central to the story and which were not l a t e r presumed by phoric nominal groups and 2) they f a i l e d to consistently i d e n t i f y a p a r t i c i -pant i n one picture with the same participant i n another ( i . e . use of i n d e f i -n ite a r t i c l e s on subsequent mentions of a protagonist). The youngest children 113 were more dependent on exophora i n a l l contexts than the two older groups (18 percent of nominal groups versus 5 percent i n older c h i l d r e n ) . 2 Although a l l children tended to use more exophora on Task 2, the youngest children i n particular were less w i l l i n g to treat this task as a d i s t i n c t context (27 percent of their nominal groups were exophoric on Task 2 compared to 9 percent on Task 1). Older children used a greater percentage of anaphora than young children; a l l children used more anaphora on Task 1 than on Task 2. As the use of anaphora increased, dependence on presenting or exophoric groups decreased and vice versa. Age and task effects were also found for bridging: the youngest group depended less on bridging than the older groups did. Martin concluded that older children adjust the use of their lexicogram-matical resources i n response to changes i n context i n ways that d i f f e r from those of younger children. Coding orientations i n the use of reference have also been distinguished on the basis of s o c i a l class. Hawkins (1973) examined the use of a r t i c l e s and pronouns i n the narrative and descriptive discourse of two groups of B r i t i s h children (aged 5): 124 were children of middle-class parents and 139 were children of working-class parents. Both tasks used were picture-based. He found that middle-class children used more nouns and associated modifiers than working-class children whereas the l a t t e r r e l i e d more heavily on pronouns, p a r t i c u l a r l y exophoric ones. Hawkins concluded that the discourse of working-class children was more closely t i e d to the context i n which i t occurred. Francis (1974), however, i n a study comparing 96 B r i t i s h children of 2Phoric nominal groups i n Task 2 ( r e c a l l of the same story without pictures) were considered exophoric when participants were introduced by means of d e f i n i t e forms since this forced l i s t e n e r s to retrieve information from Task 1. This procedure was also followed i n the present study. 114 middle-class parents versus 96 of working-class parents, found no differences in the use of nominal groups or exophoric reference on a story reproduction task which did not involve pictures. 3 In l i g h t of Hawkins' r e s u l t s , Francis concluded that children from the two s o c i a l classes were responding differen-t l y to contextual configuration. On tasks involving p i c t o r i a l support, the two groups differed i n the selection of language forms for reference (Hawkins' r e s u l t s ) : working-class children tended to make use of the nonverbal context provided whereas middle-class children abstracted the task from the immediate context. On a task lacking p i c t o r i a l support, the groups performed s i m i l a r l y (Francis' r e s u l t s ) . However, since the same children were not used i n both studies, Francis' conclusion remains speculative. In addition to examining cohesion i n the discourse of normal and schizo-phrenic adults, Rochester & Martin (1979) carried out several analyses on reference using the reference network described above. To summarize, the thought-disordered schizophrenics, l i k e the younger children i n Martin's study (1977,1983), and perhaps l i k e the working-class children of Hawkins' study, did not produce self-contexualizing discourse to the same extent as other subjects. In task-based contexts, thought-disordered speakers used a greater proportion of exophoric reference, either to pictures or to a previous t e l l i n g of the narrative. They also used f i v e times the normal proportion of personal references to themselves or to the l i s t e n e r ( i . e . I or you references) i n the narrative and twice the normal proportion i n a cartoon task. These findings supported a strong dependence on the part of thought-disordered subjects on the surrounding contextual configuration. Rochester & Martin suggested that 30ther differences were found involving certain a r t i c u l a t i o n patterns, marking and use of tenses, use of were/was, and structural complexity. 115 the common element i n such speakers (thought-disordered and younger children) may be a "lack of elaborate encoding i n the verbal context" (1979: 184). According to Martin (1983), this difference i n orientation may not make a text more or less communicative; self-contextualizing texts just happen to correspond to c u l t u r a l and educational ideals. However, i n the case of language disorder associated with schizophrenia, for example, a high propor-tion of unclear reference (see also Bartolucci & Fine 1987) suggested a lack of control of the presumptional system. Instances of missing or ambiguous referents no doubt contributed to some extent to judgments of "thought disorder" i n the f i r s t place. How a speaker handles reference phoricity i n English may have important implications i n the educational context and for successful communication i n general. Purpose of the Present Analysis Retrieval analysis may prove to be a f r u i t f u l way of examining reference i n language impaired children. Since i t can be used for any type of text, i t may be p a r t i c u l a r l y useful i n examining adjustments made with changes i n context. I t should also provide general information on speaker control, or lack thereof, of the presumptional system i n English. Researchers have apparently not used r e t r i e v a l analysis on the discourse of language disordered children so fa r . For these reasons, i t seemed appropriate to attempt such analysis on the data of the present study. The general goals proposed i n chapter three with respect to cohesion are relevant here, namely the i d e n t i f i -cation of any group and/or context effects. The purpose of the r e t r i e v a l analysis i s to obtain more information on the qu a l i t a t i v e nature of the differences revealed by the cohesion analysis. Based on findings from 116 Martin's studies, one might expect to find the following: 1. a higher proportion of exophoric reference on the part of language impaired children and a correspondingly lower proportion of anapho-r i c reference; 2. a lower proportion of bridging by language impaired children; 3. a higher proportion of unclear (ambiguous and erroneous) reference by language impaired children. The following n u l l hypotheses were tested: (10) Given the same task conditions, language impaired children do not d i f f e r from normal language children i n the frequency of phoric nominal groups per independent clause. (11) The frequency of phoric nominal groups for either group does not vary as a function of context. (12) Given the same task conditions, language impaired children do not d i f f e r from normal language children i n the d i s t r i b u t i o n of phoric nominal groups into the following r e t r i e v a l categories: a) endophora (anaphora, cataphora), b) homophora, c) bridging, and d) incomplete (exophora, ambiguous, erroneous). (13) Proportions of r e t r i e v a l categories for either group do not d i f f e r as a function of context. Results The r e t r i e v a l categories depicted i n figure 4 served as the basis for the c l a s s i f i c a t i o n scheme. Each presuming (phoric) nominal group was assigned to one of the following categories depending on the source of the presumed information: homophora (cultural context), anaphora (preceding verbal co-te x t ) , cataphora (subsequent verbal co-text), bridging ( i m p l i c i t verbal co-text) , and incomplete. The l a t t e r was subdivided into the categories of addition (exophora), ambiguous, and erroneous. Since the li s t e n e r did not know the story and could not see the pictures on TI, any exophoric reference (to the pictures) was considered incomplete and categorized as a form of 117 addition. T l and T2 were analyzed as independent contextual configurations and, therefore, nominal groups on T2 that presumed information only available from Tl were coded as requiring addition. The data base consisted of 1437 phoric nominal groups for the language impaired children and 2288 for the normal language children for a t o t a l of 3725 phoric nominal groups. S t a t i s t i -c al comparisons (Mann-Whitney U and the Friedman two-way analysis of variance by ranks) were made when possible to i d e n t i f y any group and/or task effects. Means for the frequency of phoric nominal groups per independent clause are presented i n table X I I I . These figures include phoric nominal groups of a l l categories, including incomplete items. There i s a tendency for NL children to use a greater number of phoric nominal groups per utterance than LI children on a l l tasks, although the differences do not reach significance. Significant contextual effects were not found, although the NL children tended to use fewer phoric items per clause on T3 than on T l or T2. Hypotheses (10) and (11) could not, therefore, be rejected based on the data obtained. Table XIII.—Frequency of Phoric Nominal Groups per Independent Clause for LI and NL Children on Tl, T2, and T3 Tl T2 T3 (picture-based) (retell,no pictures) (video recall) LP NL LI NL LI NL Phoric nominal groups/independent clause M 1.88 2.30 1.98 2.31 1.86 1.98 SD .18 .49 .22 .30 .16 .12 range 1.68-2.12 1.78-3.05 1.63-2.24 1.96-2.74 1.70-2.12 1.80-2.12 aN= 6 for each group Mann-Whitney Uvalues Tl T2 T3 Phoric NGs/independent clause 8.0 6.5 9.5 118 Table XIV presents mean proportions of phoric items assigned to each retrieval category. These results are described below. Table XIV.—Mean Proportions of Retrieval Categories for LI and NL Children on TI, T2, and T3 Retrieval Categories Hcmoph. Endophora Total Bridging Incomplete Total Phoric Grp. Task anaphora cataph. endoph. addition ambig/err. incompl. NGs LP 1 .009 .848 .016 .864 .007 .066 .054 .120 581 2 .029 .826 .011 .837 .021 .076 .037 .113 407 3 .035 .793 .010 .803 .016 .049 .097 .146 449 NL 1 .024 .880 .031 .912 .026 .013 .025 .038 803 2 .018 .860 .050 .910 .024 .020 .028 .048 624 3 .017 .8% .028 .924 .008 .022 .029 .051 .861 aN= 6 for each group Mann-Whitney values Homoph.b Endophora Anaphora Cataph.b Total Bridging* Incomplete addition ambig/err total TI 11.0 5.5* 10.0 5.5* 5.0* 0.0** 3.0* 0.5** T2 18.0 11.0 3.0* 9.0 14.0 4.0* 14.5 6.5 T3 12.0 1.0** 7.0 0.0** 4.0* 13.0 0.0** 2.0** b Caution must be exercised in interpreting these U values due to the small proportions obtained in these categories in the data of either group (5% or less). *p<.05 **j*.01 Endophora (anaphora and cataphora): As table XIV shows, the presumed information can be located in the explict verbal co-text in an overwhelming majority of instances. NL children demonstrated a greater reliance on endophoric reference than LI children on a l l tasks. The difference was significant on TI (tf=5, p<.05) and highly significant on T3 (lf=0, /><.01); the tendency was also present on T2. No significant context effects were found. 119 Most instances of endophoric reference were also anaphoric. Group differences follow the same pattern as endophoric reference: the NL children showed greater reliance on preceding verbal co-text than LI children, the difference being significant on TI (U=5, p<.05), highly significant on T3 {U=l, p<.01), and present, though not significant, on T2. Hypothesis (12a) was, therefore, rejected for TI and T3 with respect to anaphora. Once again, no significant task effects were found. NL children also tended to use a greater proportion of cataphora than LI children: the difference was significant on T2 (#=3, p<.05) and nearly significant on T3 (U=l, p<.10). Caution must be exercised here given the relatively small proportions of cataphoric reference by either group. Most instances involved esphora, i.e. forward reference within the nominal group. The following are examples from the data: 5.5 a. the cliff that the girl is sitting on b. rAe mountain where the great ape was c. the dad of this girl The information needed to identify which cliff, hole, and dad were intended are available from what follows within the nominal group. In 5.5, the esphoric structures served to introduce a new entity. A second function is to ensure that the referent of a phoric item is not confused with another. In the latter cases, i t usually occurs in conjunction with anaphoric reference: 5.6 a. the people watching the turtles race b. the people holding the banana c. the people who were on the boat Esphoric reference was rare in the stories of the LI children; two of them never used i t , and only one LI child used i t on T3. 120 Homophora: References to the cultural context were infrequent for both groups of children, accounting for 1 to 3 percent of phoric NGs. The follow-ing were considered instances of homophora on Tl and T2: the Empire State building, the United States, Nev York, the Big Apple, the Statue of Liberty, the Atlantic Ocean, the Pacific, the air, the ground. On T3, the following instances occurred: die police, the cops, the army, the ground, the time. Proper names invented by a child for a character in the story were not coded as instances of homophora but as anaphoric nominal groups once a referential chain was created. Very few children actually made use of proper nouns to refer to a protagonist. The small number of homophoric nominal groups made statistical testing difficult. The proportion of homophora by LI children on T3 appeared to exceed that of the NL children. This may stem from the use of proportional analysis: NL children used a greater proportion of endophoric reference on T3 than LI children, the difference being highly significant. This would have resulted in lower proportions in other categories. Bridging: Phoric items presuming information inferable from the verbal co-text were also used infrequently by a l l children (1 to 3 percent), too infrequently for reliable statistical comparison. The groups appeared to differ most on Tl: only five instances of bridging occurred in the data of LI children compared to twenty by NL children. Story length alone would not account for the size of this difference since story length was most similar on this task. The five instances of bridging by the LI children occurred in the data of only two children, whereas each of the NL subjects made use of bridging in two or more of their phoric nominal groups. On T2 and T3, the groups performed more like each other with respect to proportional data. Bridged items are listed in appendix 5. A greater variety of bridged items 121 occurred i n the data of the NL children on a l l tasks. Incomplete: These results paralleled those found for incomplete cohesive devices when a l l cohesive subtypes were considered (see chapter four). LI children used a greater proportion than NL children on a l l tasks: the difference was highly s i g n i f i c a n t on TI (tf=0, p<.01) and T3 (U=2, p<.01) and nearly s i g n i f i c a n t on T2 (U=6, p<.10). Hypothesis (12d) was, therefore, rejected. As i n the case of anaphora, homophora, and bridging, the difference between the groups i s smallest on T2. When the performances of a given group across tasks were compared, no si g n i f i c a n t context differences surfaced. LI children showed a greater reliance on addition on TI and T2 than the NL children; the difference was highly s i g n i f i c a n t on TI (#=0, p<.01) and sig n i f i c a n t on T2 (U=A, p<.01). On these tasks, the LI children involved addition i n 7 to 8 percent of their phoric nominal groups compared to 1 to 2 percent on the part of NL children. Although the means on T3 as presented i n table XIV appear to d i f f e r , ranks according to the Mann-Whitney U test did not s i g n i f i c a n t l y d i f f e r . Significant task effects were not present for either group. Ambiguous and erroneous instances were collapsed into a single category due to the small number of erroneous items. A highly s i g n i f i c a n t group difference was found on T3 (tf=0, p<.01), where LI children used an ambiguous or erroneous referent i n approximately 10 percent of their phoric groups compared to only 3 percent on the part of NL children. A s i g n i f i c a n t difference was also found on TI UMi, p<.05). On T2, the groups appeared to use similar proportions. Once again, no s i g n i f i c a n t task effects were found for either group. Hypothesis (13), therefore, was not rejected for any of the r e t r i e v a l categories. 122 Discussion Phoricity is concerned with what a speaker assumes his listener knows. It is structured into English in several ways: through reference, conjunction, substitution/ellipsis, clefting, and tonicity (for further discussion see Martin 1977). Reference, conjunction, and substitution/ellipsis were analyzed in chapter four. The retrieval analysis of the present chapter permitted a more dynamic and comprehensive investigation of reference: dynamic in its attempt to identify where a listener might retrieve the presumed information, and comprehensive in that all phoric nominals were examined regardless of the location of the presumed information. In the cohesive analysis, only those referential devices were included whose antecedent was verbally explicit and located in a clause other than that of the phoric item itself. Group and task comparisons revealed 1) a tendency for the utterances of normal language children to contain more phoric nominal groups than those of language impaired children on a l l tasks and 2) a tendency for normal language children to use fewer phoric items per utterance on the video recall task than on other tasks. These group and task tendencies did not reach significance and they are no doubt related to the greater utterance length of normal language speakers than language impaired speakers across tasks, and to the tendency for normal speakers to use shorter utterances on the video recall task than on other tasks. Normal language speakers tended to use fewer intra-utterance conjunctions on T3, i.e. they did not conjoin propositions into complex utterances to the same degree on this task as on other tasks. As discussed in chapter four, i t is difficult to know whether the reduced utterance length resulted in fewer opportunities to code referential items or 123 whether the use of fewer phoric items resulted i n shorter utterances. A d i s t r i b u t i o n a l analysis based on r e t r i e v a l categories revealed, not surprisingly, that, i n narrative contexts, over 90 percent of phoric nominal groups for normal language children presumed information that could be retrieved from preceding verbal text. The r e s t r i c t i o n to narratives i s important: the normal adults of Rochester & Martin's study (1979), for example, r e l i e d on direct verbal context 86 percent of the time i n narratives but only 46 percent of the time i n interviews. There are two possible explanations for the greater use of anaphoric reference by normal language children (on TI and T3): the language impaired children may be relying less on preceding verbal co-text than the normal language children and/or the language impaired children may be attempting as many anaphoric references as the language impaired children but f a i l i n g i n their attempts. In either case, the l i s t e n e r ' s task becomes more d i f f i c u l t ; he or she cannot see the pictures and i s , therefore, highly dependent on information presented verbally. Cataphoric reference was infrequent i n a l l stories but p a r t i c u l a r l y i n those of the language impaired children. Esphora (forward reference within the nominal group) accounted for the majority of cataphoric references and was used both to introduce new participants to the discourse or to c l a r i f y old ones. Both are discourse functions; both contribute to cohesive text. As Martin (1977) notes, by r e s t r i c t i n g the analysis of cohesion to inte r - c l a u s a l relations only, Halliday & Hasan's (1976) analysis leaves certain texture-creating devices unanalyzed. The use of t h i s p a r t i c u l a r device appeared to distinguish the two groups but the limited number of observations render this conclusion tentative at best. I t was suggested i n chapter four that greater l e x i c a l support on the part of normal language children may result i n fewer 124 ambiguous referents on their part. A greater use of esphora may be another. A more obvious factor responsible for the higher proportion of incomplete reference by languge impaired children was their greater reliance on addition (exophora) in the picture narration task and on retell of the same story without pictures. On both tasks, addition involved introducing new partici-pants using a phoric item, usually a definite article plus noun construction. The use of a pronoun to introduce participants occurred only three times, each time by a language impaired child. The greater use of exophora by language impaired children than by normal speakers on the picture narration task indicated a greater reliance on the nonverbal context, i.e. on the pictures that were provided. The normal language children were better able to take into account the fact that the listener could not see the pictures and/or to reflect this knowledge in the use of indefinite versus definite forms. Although the listener was the same for the two tasks, each subject was instructed on T2 to pretend he was in another city speaking to a new audience who had never heard the story before. The normal language children seemed more able to treat the second telling as a distinct contextual configuration requiring that participants be introduced in nonpresuming ways. On both tasks, then, normal language children self-contextualized their stories to a greater degree than language impaired children, showing less dependence on the situational context, whether i t be that of pictures or of a prior telling of the story. In the video recall task, there was s t i l l a tendency for language impaired children to use definite forms to introduce, reflecting perhaps dependence on a previous context (their viewing of the film), but this tendency was reduced. These differences parallel those found between younger and older children 125 in developmental studies. As reported earlier, Martin (1977,1983) found both a greater percentage of anaphora and a smaller percentage of exophora in the data of his older children in similar tasks as to those used in the present study. Similar findings were reported in comparisons between socially advantaged and disadvantaged children (Hawkins 1973, Francis 1974) and between normal and schizophrenic adults (Rochester & Martin 1979, Bartolucci & Fine 1987). These investigators explained their results with reference to the notion of code, the predilection to select certain options and not others given a particular contextual configuration. The ability to choose the "acceptable" forms for a given context develops over time and, based on the findings of this study, i t develops to a different extent in language impaired and normal language children. The normal language children, like the older children of Martin's study (1977, 1983) were better able to produce the self-contextualizing stories valued in western culture, particularly in the school setting. The different pattern of results found when the two groups were compared on addition and on ambiguous/erroneous instances underlines the importance of subdividing inappropriate use of reference: addition does not involve the same kind of "inappropriateness" as the ambiguous/erroneous category. The high proportion of ambiguous items by language impaired children on the video recall task may have resulted from the greater use of pronouns on this task, many of which referred to protagonists of the same grammatical number and gender. There were four males in the story, each with a different role to play in advancing thematic progress. Language impaired children had more difficulty providing sufficient clues to enable the listener to identify the appropriate referent. 126 It was also predicted that normal language children would use bridging to a greater extent than language impaired children. This seems to have been the case on the picture narration task, where only five instances of bridging occurred in the data of language impaired subjects compared to twenty in the normal children's data. A greater variety of bridged items was demonstratd by normal language children on a l l three tasks. The normal language children used a smaller proportion of bridging (2 percent) than children of approxi-mately the same age in Martin's study (6 to 7 percent).4 The discrepancy may be due to differences in story content or to criteria used for coding bridged items. The proportions of bridging in both sets of data (those of the present study and Martin's) were less than that of normal adults in Rochester & Martin's study (1979). The discrepancy may once again be due to differences in story content or analytical procedure. Alternatively, i t may be due to developmental differences. Increasing world knowledge may permit a speaker to more easily discern what can or cannot be inferred by the listener. Given the small proportion of bridging found in the present study, a similar problem arises as in the analysis of cataphora and homophora: there are too few observations available upon which to base one's conclusions. Low proportions, in particular, are highly vulnerable to a change in a single observation. The problem stems from a general one associated with natural language sampling: the few instances that occur naturally make meaningful comparisons difficult. The analysis of cataphora, homophora, and bridging may require data obtained from more contrived experimental conditions. "Martin's figures are presented as proportions of total nominal groups. For comparison with the results of the present study, these figures were calculated as proportions of phoric nominal groups only. The values of 6 to 7 percent derive from this conversion. 127 Friedman two-way analysis of variance by ranks f a i l e d to reveal s i g n i f i -cant context effects. There was a tendency for language impaired children to use a smaller proportion of addition on the video r e c a l l task than on the other tasks, due largely to the higher proportion of exophoric reference on TI and T2 as noted e a r l i e r . With respect to group and task interaction, the language impaired children tended to perform more l i k e the normal language children on T2 than on other tasks (except i n the case of addition for reasons discussed above). This suggests that language impaired children may derive considerable benefit from a second t e l l i n g , a finding that may have implica-tions for assessment and therapy. If a child's best performance i s sought, a second t e l l i n g of a story may prove more revealing. The l a t t e r may also prove be n e f i c i a l for assimilation of a new l i n g u i s t i c behavior. To summarize, r e t r i e v a l analysis revealed that normal language children used a greater proportion of anaphoric and exphoric reference and a smaller proportion of incomplete (exophoric, ambiguous, and erroneous) reference than language impaired children. Insufficient data made conclusions concerning the use of cataphora, bridging, and homophora d i f f i c u l t ; the normal language speakers appeared to make greater use of esphora and bridging than language impaired speakers. The most s i g n i f i c a n t finding was the higher proportion of exophora and corresponding lower reliance on verbal encoding (endophora) by language impaired children than normal language speakers. This appears to r e f l e c t a greater reliance on the part of language impaired children on the s i t u a t i o n a l context (pictures or a prior t e l l i n g ) when i t was available. They appeared to resemble younger normal speakers i n this respect. Phoric items w i l l once again be targeted i n the following chapter, where the functions served by anaphoric expressions w i l l be explored i n more d e t a i l . CHAPTER SIX REFERENCE: FORM/FUNCTION ANALYSIS Discourse Reference Theoretical Background The analysis presented in chapter four examined a number of different cohesive devices used in discourse. Use of one of these devices, reference, was further explored using retrieval analysis in chapter five. The purpose of the present chapter is to narrow the field of investigation s t i l l further by focusing on one type of reference, namely anaphora. As in the case of retrieval analysis, the approach is a "dynamic" one: the focus is not on the presence or absence of certain forms but on the function such forms appear to be ful f i l l i n g for the child, i.e. on discourse-as-process unravelling in time. Recent investigations into anaphoric reference have tended to side with one of two theoretical positions. According to the "structural" approach, anaphoric elements—pronouns in this case—have no inherent meaning. In order to be interpreted, they must be "substituted" with a preceding nominal expression. Early forms of transformational grammar described pronouns as surface structure substitutes derived from an underlying ("deep") structure which was structurally and lexically identical to the underlying form of the pronoun's antecedent. This transformation required two conditions: 1) the existence of an explicit linguistic antecedent showing agreement with respect to number and gender and 2) an antecedent in the same sentence as the pronoun. Coreference within this syntactic view was a matter of indexing coreferential 128 129 items in deep structure with the same subscript. The technical device of indexing has no intrinsic value of its own, and the methodological flaws associated with its use have led to theoretical revisions. In recent Chomskyan theory, indexing has been removed; pronouns are no longer syntactically derived but are generated directly. The problem of explaining how pronouns refer, however, remains unresolved and has been superceded by a different issue: the specification of anaphoric constraints. Unfortunately, even the best of these constraints, namely constituent command (c-command), fails to account for many anaphor-antecedent relations even at the sentential level (for examples, see Bosch 1983, and Wiese 1983). As Bosch (1983) points out, the substitute/transformational view is limited both theoretically and empirically. It is limited theoretically by i its avoidance of discussion on how pronouns refer, and empirically by the small number of pronoun occurrences i t can account for syntactically. It excludes intersentential anaphoric relations, those with no explicit linguis-tic antecedent, those with antecedents which do not agree in number and gender (e.g. split antecedents1), and any other intrasentential relations that violate the constraints formulated so far. In other words, this view may not account for the majority of pronoun occurrences in English. These inadequacies have led to the adoption, by some linguists, of a non-structural approach. According to this view, a l l grammatical constraints on 1 According to the transformational approach, a pronoun must be lexically and syntactically identical to its antecedent. The split antecedent Bill and his brother for they in the simple example below clearly does not meet this identity condition: B i l l told his brother that they had to be home by noon. 130 pronoun-antecedent pairs will f a i l , given certain discourse conditions. "Nonstructuralists" have largely abandoned syntactic sentence-level analyses in favor of a more psychological or processing orientation. Anaphoric expressions are understood to be performing the pragmatic function of trans-forming isolated sentences into text. A processing explanation as to how anaphors are able to achieve this requires recourse to the notion of speaker reference. A speaker reference account of anaphora holds that the referent of an anaphoric expression (pronoun or definite nominal group) is not the linguistic antecedent per se, but the speaker's intended referent. Successful reference depends, not first and foremost on the satisfaction of grammatical con-straints, but on the listener being able to identify a speaker's intended referent. The selection of a referring expression will depend on a speaker's expectations about and intentions towards the listener. According to Donellan (1978), for example, speakers who expect and intend that their listeners recognize who they have in mind, will start with a definite description: 6.1 (to a colleague) The new guy keeps pestering me with questions. If, on the other hand, a speaker does not expect or intend that the listener recognize the person he or she has in mind, the speaker will introduce the referent using an indefinite form: 6.2 (to a spouse) A new guy at work keeps pestering me with questions. In nonstructural terms, anaphora "covers any expression which the speaker uses in referring on the basis of which the hearer will be able to pick out the intended referent given certain contextual and co-textual conditions" (Brown & Yule 1983: 215). Pragmatically-controlled anaphora does not require a linguistic antecedent and is, therefore, not concerned with the grammatical 131 conditions that hold between pronoun-antecedent pairs. The main goal of the nonstructural approach has been to determine how a listener is able to identify speaker reference (Marslen-Wilson et al. 1982, Wiese 1983, Brown & Yule 1983). In chapter five, pronouns and definite nominals were classified according to the "location" of the referent, whether in shared culture, nonverbal context, or verbal co-text (explicit or implicit). Referring expressions were categorized as anaphoric if a linguistic antecedent was present. In light of the above comments, the reference network was somewhat simplistic and the notion of "retrieval categories" perhaps misleading. An analyst cannot know for sure what information the listener is using to interpret the speaker's intended referent. The linguistic antecedent is certainly one source of information; gender and number of the pronoun used represent another. However, the many cases of missing antecedents or of grammatical mismatch between an antecedent and pronoun suggest that such clues are not the only ones a listener is using. As noted in chapter four, informa-tion gleaned from previous or subsequent predicates, for example, may be particularly important in choosing between antecedent nominals as illustrated in 6.3 and 6.4 below (from Brown & Yule 1983: 219): 6.3 One of our main jobs in the Botanies is writing on the flora of Turkey + they . . . One cannot be sure about the interpretation of they, i.e. of the given element, until the subsequent predicate, the new information, is provided: 6.4 One of our main jobs in the Botanies is writing on the flora of Turkey + they don't have the scientists to do i t . The listener, then, is using a l l sorts of information, explicit and implicit, from both the verbal and nonverbal contexts, to compute speaker reference. And the speaker is guided in decisions as to what kind of 132 information is required based on expectations about and intentions towards the listener. The inadequacy of the "structuralist" view of reference to account for such facts underlines the necessity of moving the study of reference phenomena away from decontextualized, contrived sentences to naturally-occurring discourse. If the choice of a referring expression is to a great extent pragmati-cally controlled, then the functions served by referring expressions may be expected to vary depending on the particular discourse genre examined. As noted above, a speaker chooses a form according to expectations about and intentions toward the listener in view of enabling the listener to identify the referent. Beyond this general purpose, there is perhaps more to be discovered with respect to a speaker's purpose in selecting one referring expression over another in a given descourse genre. In the present study, this means exploring how function governs the choice of form during the course of a narrative. The selection of referential devices by a narrator may be influenced not only by the local context in which the device occurs but also by properties of the narrative itself: its episodic organization and protago-nist development (whether, for example, there is a main protagonist and, if so, his or her relationship to other participants). The choice of referring expressions in narratives may be expected to play a role in the organization of a narrative, specifically as they f u l f i l l the cohesive text-forming func-tions of creating, maintaining, and switching reference. The Development of Referential Devices In narrative contexts, the information provided by a speaker to enable the listener to identify participants is linguistic in nature. A narrator 133 does not usually direct the l i s t e n e r to the nonverbal context of utterance (unless props are provided), but to other l i n g u i s t i c information, whether i n the form of antecedent noun phrases and/or antecedent or subsequent predi-cates. This intralinguistic use of signs has been contrasted with deictic use (signs related to some object i n the nonlinguistic context) i n developmental studies. The entity ultimately referred to i n both cases i s a nonlinguistic one; deixis points to i t d i r e c t l y whereas i n t r a l i n g u i s t i c use {anaphora) does so i n d i r e c t l y through sign-to-sign relationships. Research findings suggest that children f i r s t use pronouns and d e f i n i t e a r t i c l e s d e i c t i c a l l y , i . e . to "point" with their words to nonverbal e n t i t i e s present4 i n the environment. There i s considerable debate as to when exactly children are able to create and maintain discourse through l i n g u i s t i c means alone. Keenan & Klein (1975), for example, concluded from their observations of informal dialogue between twin boys (aged 2;9 at the outset, observed over a one-year period) that the children were able to establish a new topic and then to use pronouns anaphori-c a l l y to maintain reference. However, as Hickman (1980) pointed out, the referents were present i n the si t u a t i o n and, therefore, there was no way of knowing whether the children of Keenan & Klein's study were using pronouns d e i c t i c a l l y or anaphorically. Developmental studies have also explored i n d e f i n i t e versus d e f i n i t e noun phrases and the functions served by these forms. Maratsos (1976) examined both comprehension and production of i n d e f i n i t e and d e f i n i t e a r t i c l e s i n experimental research involving 3- and 4-year-olds. He concluded that children both understood the difference between the two constructions and were capable of using these forms appropriately to distinguish between s p e c i f i c and nonspecific reference. His conclusions have been c r i t i c i z e d on a number of 134 counts (see Warden 1976, Karmiloff-Smith 1979, and Hickmann 1980): 1) the experimenter and not the c h i l d provided the verbal context for the child's subsequent referring expression; 2) the child's use of d e f i n i t e noun phrases may have resulted from a general preference for d e f i n i t e forms regardless of whether the referent had been i d e n t i f i e d previously or not; 3) i n d e f i n i t e a r t i c l e s may have been used merely to name objects as opposed to identify them;2 4) only an advanced subgroup of 4-year-olds performed appropriately when r e f e r e n t i a l knowledge of the l i s t e n e r was taken into account. Warden (1976) studied the a r t i c l e s used to introduce a referent to a context of discourse by children aged 3 to 9 and by adults. Children appeared to master the naming function of the i n d e f i n i t e a r t i c l e before i t s identifying function. Only the 9-year-olds and adults consistently introduced the i r referents using i n d e f i n i t e expressions. The referring expressions of children under 5 were predominantly d e f i n i t e ; children between 5 and 9 years of age used both d e f i n i t e and i n d e f i n i t e a r t i c l e s to introduce referents. Warden concluded that young children f a i l to take account of their audience's knowledge of a referent; the c h i l d f a i l s "to r e a l i z e that his audience w i l l only become familiar with his referent after he has i d e n t i f i e d i t for them verbally" (p.110). However, as Warden himself pointed out, nearly every c h i l d produced at least some ide n t i f y i n g expressions, r e f l e c t i n g at least p a r t i a l awareness of the need to i d e n t i f y referents. Young children may have a 2According to Warden (1976) and Hickmann (1980), the main functions performed by i n d e f i n i t e forms are the following: 1) to mark indefiniteness ("I want a monkey"): reference i s made to any member of the class of monkeys; 2) to name ("That's a monkey"): a particular item i s singled out as one member of the class of monkeys; 3) to identify ("There's a monkey i n the back yard"): a particular monkey i s introduced to the discourse. 135 greater awareness of listener need than Warden allows; their difficulty may lie in using the indefinite/ definite article distinction to express this awareness. Props were present in Warden's experiments (model animals or pictures) and i t is difficult to determine the extent of their influence on the forms that were elicited. For many children, identifying such objects with indefinite forms may have appeared redundant given the nonverbal context. In an attempt to have subjects rely strictly on linguistic means to create reference, Hickmann (1980) had adults and children aged 7 and 10 narrate short films from memory, without the use of props or pictures, to listeners who had not seen the films. A developmental progression was found: only the adults and 10-year-olds used appropriate forms more often than inappropriate ones. Animacy and degree of involvement in speech events had an effect on appropriateness of first mentions: 7-year-olds had greater d i f f i -culty introducing a l l animate referents with indefinite reference and both groups of children had more difficulty when both participants and "nonpartici-pants" (those introduced in reported speech) were animate. Hickmann also examined a few narratives by 4-year-olds for comparison and found a complete lack of indefinite forms to introduce even inanimate nonparticipants; on several occasions, highly presupposing forms (pronouns) were used on first mention. Hickmann concluded that the skills for textual cohesion emerge at approximately 7 years of age, and that children do not consistently create referents for later intralinguistic cohesive relations in discourse until relatively late (10 years of age). However, the assumptions she has made in arriving at her conclusions must be questioned. Her first conclusion, for example, incorrectly assumes that text cohesion is entirely accomplished by 136 the use of indefinite articles for first mentions. In reaching her second conclusion, she has assumed that to create referents for later intralinguistic cohesive relations in discourse means to use an indefinite article for first mention. It may be, however, that the young child is "creating a referent" the first time he or she refers to i t regardless of the form of the accompany-ing article. The most that can be concluded from either Hickman's or Warden's results is that one specific kind of textual device, namely the creation of referents in narratives using the conventional means of indefinite forms, may not be mastered before 9 or 10 years of age. It would be incorrect to also conclude, based on the findings reported so far, that up to that time a child is not making use of prior linguistic context or is not able to take i t into account when evaluating listener need. Bennett-Kastor (1983) reached a different conclusion than Hickmann did based on an examination of noun phrases in the stories of children aged 2 to 5 : she reported evidence for the emergence of textual coherence in children as young as 2 years of age. The discrepancy between her results and those of Warden or Hickmann arises from the type of evidence each investigator accepted as representing children's attempts to relate their utterances. According to Bennett-Kastor, the fact that a noun phrase is repeated from clause to clause is a feature that makes a story coherent. The children of her study reitera-ted primarily animate entities, most of which were expressed as agents. Most noun phrases were reiterated in subject position; even the 2-year-olds were "sensitive to the function of grammatical subjects as elements of focus, and to the predominantly agent-oriented nature of narrative discourse" (Bennett-Kastor 1983: 142). With development, reiterated NPs increased in frequency, variety, and distance between mentions. With respect to grammatical form, 137 i n d e f i n i t e a r t i c l e + noun c o n s t r u c t i o n s were u s e d a t a v e r y y o u n g age t o i n t r o d u c e r e f e r e n t s ; t h e 2 - y e a r - o l d s , f o r e x a m p l e , i n t r o d u c e d 5 4 . 5 % o f t h e i r noun p h r a s e s ( t h a t were l a t e r r e i t e r a t e d ) i n t h i s way. T h i s p e r c e n t a g e decreased o v e r t i m e , f a l l i n g t o 19.4% f o r t h e 5 - y e a r - o l d s . B e n n e t t - K a s t o r p r o p o s e d s e v e r a l r e a s o n s f o r t h e i n c r e a s e d use o f d e f i n i t e r e f e r e n c e t o i n t r o d u c e r e f e r e n t s on t h e p a r t o f t h e o l d e r c h i l d r e n . I n n e a r l y a l l c a s e s , t h e s e d e f i n i t e r e f e r e n c e s i n t r o d u c e d a r c h e t y p a l s t o r y v i l l a i n s ( e . g . the witch, the wolf). O l d e r c h i l d r e n ' s e x p e r i e n c e w i t h n a r r a t i v e s w o u l d e n a b l e them t o d e c i d e w h i c h s o r t s o f c h a r a c t e r s c o u l d be c o n s i d e r e d a r c h e t y p a l . G r e a t e r w o r l d e x p e r i e n c e may a l s o h a v e p e r m i t t e d t h e u s e o f d e f i n i t e r e f e r e n c e f o r e n t i t i e s t h a t a d u l t s a l s o r e f e r t o i n t h e same way ( e . g . the doctor). From h e r r e s u l t s on t h e f r e q u e n c y o f r e i t e r a t e d NPs and on t h e i r s e m a n t i c and g r a m m a t i c a l c h a r a c t e r i s t i c s , B e n n e t t - K a s t o r c o n c l u d e d t h a t f e a t u r e s o f t e x t u a l c o h e r e n c e c o u l d be f o u n d i n t h e s t o r i e s o f a l l t h e c h i l d r e n : P e r h a p s due t o t h e f r e q u e n c y w i t h w h i c h most c h i l d r e n a r e e x p o s e d t o a n a r r a t i v e , t h e y l e a r n as e a r l y as two y e a r s how t o i n t r o d u c e NPs i n t o a s t o r y , how t o r e i t e r a t e them i n s u c c e s s i v e s t o r y c l a u s e s t o r e f l e c t t h e i r g i v e n n e s s o r t h e i r f o c a l i n t e r e s t , how g e n e r a l l y t o c o n v e y t h a t t h e d i s c o u r s e t h e y a r e s p e a k i n g i s , s p e c i f i c a l l y , n a r r a t i v e . ( 1 9 8 3 : 148) A c c o r d i n g t o B e n n e t t - K a s t o r , b e c a u s e t h e 5 - y e a r - o l d s were l i n g u i s t i c a l l y more m a t u r e , t h e y were b e t t e r a b l e t o a c c e s s and e x p l o i t t h e i r g r a m m a t i c a l s y s t e m t o a c h i e v e t e x t u a l c o h e r e n c e . B e n n e t t - K a s t o r ' s f i n d i n g s , p a r t i c u l a r l y t h o s e i n v o l v i n g t h e u s e o f i n d e f i n i t e + n o u n c o n s t r u c t i o n s t o i n t r o d u c e p r o t a g o n i s t s , c o n f l i c t w i t h t h o s e o f b o t h H i c k m a n n (1980) and Warden ( 1 9 7 6 ) . The d i s c r e p a n c y may be due t o t h e age r a n g e o f t h e s u b j e c t s u s e d . As n o t e d , t h e o l d e r s u b j e c t s i n B e n n e t t -K a s t o r ' s s t u d y , t h e 5 - y e a r - o l d s , u s e d a s i g n i f i c a n t l y h i g h e r p r o p o r t i o n o f d e f i n i t e r e f e r e n c e t o i n t r o d u c e t h e i r r e f e r e n t s t h a n t h e 2 - y e a r - o l d s . The 138 4- and 7-year-olds of Warden's and Hickmann's studies respectively also used a high proportion of d e f i n i t e reference, t h i s time i n comparison with older subjects and adults. If both sets of data are v a l i d , one might postulate that the proportion of d e f i n i t e reference i n very young children increases u n t i l approximately 4 or 5 years of age and decreases thereafter. The f i r s t development would be due to the child's increasing use of archetypal, homopho-r i c , and bridged reference. This tendency would l a t e r be overshadowed by a greater pressure, one that pushes the c h i l d i n the direction of the culture-s p e c i f i c ideal of self-contextualization i n narratives. The method used to e l i c i t stories may be another factor underlying the different results obtained across studies. Both Warden and Hickmann provided some form of experimental s t i m u l i to e l i c i t s t o r i e s : pictures and films respectively. Bennett-Kastor used a more spontaneous method; she had her subjects t e l l any story they l i k e d . It may be that i n the presence of experimental s t i m u l i , children show a greater tendency to treat that material as given than when contextual support i s absent and narrators must come up with their own referents. In other words, children may perceive their own invented referents to be i n greater need of introduction than those provided by the experimenter. Karmiloff-Smith*s research seeks to discover not the if and at what age children use a pa r t i c u l a r category, but on the processes involved i n such use. The findings of one of her narrative studies (1981) i s p a r t i c u l a r l y revealing with respect to the processes involved i n the use of pronominalization. Each c h i l d (170 English and French children, aged 4 to 9) was presented with a book of s i x pictures and asked to t e l l the experimenter what was happening as the experimenter turned the pages. There were four kinds of books which differed with respect to the number and c e n t r a l i t y of the main protagonists and 139 tightness.of links between the pictures. One of these stories consisted of the following: a central character (a boy walking along a road [picture 1]), introduction of another character (the boy sees a balloon man [picture 2], who gives the boy a balloon [picture 3]), and return to first character (the boy walks off with the balloon [picture 4 ] , the balloon flies into the sky [picture 5] and the boy starts to cry [picture 6]). The stories of the youngest children were held together by spatial devices (e.g. there) and by paralinguistic gestures (pointing, eye gaze, head movements) accompanying their pronouns. Since the experimenter could see the pictures, these deictic pronouns were unambiguous as long as each utterance was considered separately and no intralinguistic links were attempted. The narratives of slightly older children (approximately 6 years old) showed evidence of anaphoric pronominal use and of sensitivity to the cohesive function performed by such devices. The main character (the boy) was intro-duced and subsequently pronominalized and placed in i n i t i a l sentence position, i.e. children reserved the use of pronouns in utterance-initial position for the "thematic subject" (central protagonist) of the story. Nominal forms were used for other entities and these appeared almost exclusively in nonsubject position. On those rare occasions when a pronoun was used for a nonthematic subject in i n i t i a l position, a nominal was added in postposition (at the end of the utterance) for clarification (French data). In interpreting these findings, Karmiloff-Smith suggested that the children at this stage were imposing a rigid macro-thematic structure onto their narratives. In the narratives of the oldest group of children, participants other than the main protagonists sometimes occurred in i n i t i a l position, reflecting a growing control over the interplay between discourse structure and local sentential 140 relations. Pronominalization of nonthematic subjects was s t i l l rare and usually reserved for use within complex sentences. A pronoun in in i t i a l position was s t i l l most likely to refer to the thematic subject even when several utterances containing other referents had intervened. A thematic subject strategy, then, was evident even in the older children's narratives. In a subsequent experiment, Karmiloff-Smith (1981) took one of the pictures from the story described above (picture 3, in which the balloon man hands a balloon to the boy) and showed i t to 180 French and English children (who were not involved in the previous experiment). Each child was asked to describe what was happening in the picture. Over 80% of these children put the balloon man in subject position and used the verb to give (or to sell). There were no repairs or noun phrase postpositioning. As Karmiloff-Smith noted, the same picture (picture 3) was described very differently in the context of a single utterance than when that utterance was part of a sequence of utterances that made up a story. Based on data obtained from both linguistic and nonlinguistic domains, Karmiloff-Smith has proposed a three-phase processing model of children's problem solving (1983,1984). Phase one, the "procedural phase", is character-ized by data-driven processes. The child uses environmental feedback to solve a problem without making an attempt to organize isolated success and/or failure experiences. In phase two, the "metaprocedural phase", behavior is generated predominantly by a top-down control mechanism. A rigid internal representation is imposed on external stimuli to such an extent that environ-mental information is often ignored. A simplifying procedure is used to link previously isolated procedures. In the final phase, the "conceptual phase", neither data-driven nor top-down processes predominate; a control mechanism 141 modulates interaction between the two. The child is able to apply internal representations and simultaneously take into account environmental feedback. Karmiloff-Smith has also applied this problem-solving model to the development of referential cohesion in narratives (1984). At level l 3 (the procedural phase), the child uses success-oriented procedures to describe each picture in isolation from the others. Individual utterances are syntactically and lexically correct, but appear merely to be juxtaposed, lacking any organizational property. At level 2 (the metaprocedural phase), although the story appears weaker in lexical variety and story details, there is evidence of an overall organization, i.e. of top-down processing. The child uses a "thematic subject" strategy to simplify the data. Level 3 (the conceptual phase) is characterized by greater control over both global structure and local intersentential contraints. Lexical variety is again rich and more story details are included than at level 2, this time coordinated with the top-down control mechanism. Karmiloff-Smith's comprehensive research sheds considerable light on the acquisition of referential cohesion. However, one of her conclusions—that intralinguistic (anaphoric) reference does not appear until the age of five or six years—is not well supported. In an experiment with French children, she found that when natural gender of a referent conflicted with grammatical gender (e.g. the nonce term le bicron in which the article and word ending were grammatically masculine and the extralinguistic referent was feminine), more older children preserved the grammatical gender of the antecedent in their pronouns or made changes that would ensure grammatical continuity (e.g. 3Karmiloff-Smith uses the term level to refer to changes within a specific domain, in this case the development of narratives. 142 by changing le bicroa to Ja bicronne. Younger children tended to use pronouns which corresponded to the natural gender of the extralinguistic referent. Karmiloff-Smith concluded that the older children were using anaphoric and the younger children deictic expressions. However, in conversational French, i t is common, even among adults, to use pronouns that reflect the natural gender of an extralinguistic referent regardless of the grammatical gender of the nominal antecedent. In written French, on the other hand, grammatical gender is usually preserved even when it conflicts with natural gender. The atten-tion to grammatical gender as demonstrated by the older children may reflect their approach to the task as a "test" situation; repairs made in the direc-tion of "anaphoric" reference may have been partially motivated by recollec-tions from grammar classes. It may also reflect their greater experience with the conventions of written language. A second and related interpretation that must be questioned concerns Karmiloff-Smith's claim that the young children in the narrative described earlier (boy and balloon story) were only capable of deictic reference. In a pilot study, she had children t e l l the story without pictures and concluded that the young children were s t i l l using their pronouns deictically. Since the possibility of paralinguistic behavior (e.g. pointing) was removed by taking away the pictures, it is not clear what kind of evidence Karmiloff-Smith used to conclude that the younger child's use of pronouns was deictic. The relevance of discourse factors to early form-function mappings in language acquisition has also been investigated by Bamberg (1987). There are three ways, according to Bamberg, of "moving" through a narration, depending on the anchorpoint one chooses: character, time, or location. In most stories, the three are interwoven. Bamberg examined both character and time 143 in his study; only his results on the character-anchored perspective will be presented here. Within this perspective, a l l clauses are related to a character's inner and outer states. In actual discourse this is done by establishing a character's identity at the beginning of a story and subsequen-tly referring to him or her by reference-maintaining and reference-switching devices. He examined the narratives of 25 German children of three age groups: 3- and 4-year-olds; 5- and 6-year-olds; 8- to 10-year-olds. Adult data was also obtained. The narratives were based on a wordless picture book4, which was first presented to the children in the form of a slide show via question-answer interaction between researcher and children. Four phases of elicitation followed: 1) the child told the story to the experimenter at school; 2),3) a caregiver told the story to the child at home on two occa-sions; 4) the child retold the story to the experimenter at school. The results were based on the adult data from phase 3 and the child's data from phase 4. In a l l cases, both child and adult viewed the pictures together. In his analysis of forms used for first mentions of the main protago-nists, Bamberg found that adults showed a fairly even distribution for a l l characters between indefinite and definite expressions. For children, an even higher percentage (75%) of first mentions were definite. The children also used more pronouns to introduce than the adults did (16% versus 1%). The only developmental difference found in the children's data was the absence of indefinite expressions to introduce on the part of the youngest children (aged 3 and 4). This supports the hypothesis that young children are more "context-dependent" in their selection of forms. 4 Frog, Where are You? by Mercer Mayer. 144 A second analysis involved coding the form (nominal versus pronominal) and function (maintain versus switch reference) of each subsequent mention of the two main characters (boy and dog). Adults were highly consistent: they used nominal expressions to switch reference 95% of the time, and pronominal expressions to maintain reference 92% of the time. The few instances in which a pronoun was used to switch reference appeared to signal to the listener that information that immediatly preceded was to be treated as background informa-tion only, and that the speaker was now continuing with foregrounded informa-tion. The second case, the use of a nominal to maintain reference, usually occurred at the beginning of a new episode and was normally preceded by a long pause and accompanied by new temporal markers. Bamberg concluded: 1) that the predominant tendency to match pronominal forms with maintaining reference and nominal forms with switching reference was due to a bottom-up strategy, i.e. the use of linguistic forms for local cohesion and 2) that the exceptions were due to a top-down strategy aimed at marking global thematic structure in view of lending greater coherence to the narrative. With respect to discourse development, Bamberg predicted an orientation first towards one or the other strategy before a child would be able to integrate both in constructing text. An analysis of the forms used to switch and to maintain reference by the children revealed a number of strategies which varied depending on the protagonist referred to (boy or dog) and on the age of the children. An "anaphoric"8 strategy was used to refer to the dog whereas four different strategies were used to refer to the boy: thematic, "anaphoric", locally contrasting and nominal. The thematic strategy was marked by a preference for 8Bamberg uses the term anaphoric to describe the strategy employed by the adults of his study, namely the use of nominal forms to switch and pronominal forms to maintain reference. 145 pronominal forms to both maintain and switch reference. A locally-contrasting strategy showed no clear separation of a form-function pairing (both pronomi-nals and nominals were used to switch or maintain reference); decisions were made at a very local level. Speakers using a nominal strategy appeared to describe each picture in an isolated manner, reintroducing characters by nominal forms as each picture was turned over. All children showed a preference for using pronominal forms over nominal to refer to the boy and a preference for nominal over pronominal ones for the dog. The latter trend was also present to some extent in the adult data. With respect to age differences, a l l children differed from the adults in the proportion of noun phrases used to switch reference versus maintin reference: the children switched reference twice as often as they maintained it whereas adults showed a fairly even distribution between the two functions. Referential devices to maintain increased somewhat over time. A second developmental trend consisted of an increase in the use of nominals to switch reference to the boy (all children and adults preferred nominals to switch reference to the dog). A thematic subject strategy was the predominant strategy of the youngest and middle age groups but was used by only a few of the oldest children. The "anaphoric" strategy used by adults was not found in the youngest children but was present in a few stories of the middle group; i t dominated the stories of the 9- and 10-year-olds. Bamberg concluded that the developmental process of form-function pairings consists of moving from a global level of discourse organization to a local level, and from there to an integration of the two levels. He proposed two factors motivating these changes. The first influence is the child's experience with written material. A second appears to arise from within the 146 child in the form of a local subcomponent at work within the child's "global" system. In addition to using pronouns to mark the thematic subject, young children also used them to maintain reference to nonthematic actors (the dog). As Bamberg noted: This particular sub-strategy being situated right within the overall system created by the child could metaphorically be described as some kind of explosive, or a "time-bomb" ticking away and awaiting its detonation, (p.198) For a child to relate this local strategy to his global strategy, Bamberg argued, the child must first have recovered "the local viewpoint". Bamberg's study is commendable in its attempt to investigate the internal processes involved in language acquisition and the reorganization of these processes with development. However, his ordering of these processes into three distinct phases—global followed by local followed by an amalgamation of the two—is not well supported by his data. As Bamberg himself points out, a local substrategy appeared to be governing the choice of referential terms for the dog in the first phase, i.e. when a child was also demonstrating use of a global strategy in references to the boy. Nor do his data support a temporary disappearance of a global stategy in phase two and mysterious reappearance in phase three. What his results do suggest is that there are two pressures— global and local—acting on even a very young child's organization of dis-course. The route taken to greater control of these factors appears to be a continuous one, and no doubt varies considerably with the individual. A second cautionary note concerns Bamberg's scope of analysis. Although he claims to have found support for Karmiloff-Smith's thematic subject strategy, he has, in fact, found evidence for only one aspect: the use of pronouns to refer to a thematic subject regardless of the function served (maintain or switch reference). He has not explored the use of sentence position as it 147 relates to a story's thematic subject. With respect to methodology, Bamberg provided well-motivated reasons for the procedures he followed. However, he neglected to discuss the full impact of contextual factors on his results. Each child heard the story three times (once by the experimenter and twice by a caregiver) and had previously told the story before the data used in the analysis was collected. Shared story knowledge, rehearsal, and the fact that both experimenter (or caregiver) and child viewed the pictures together are a few of the factors that certainly influenced the speaker's (both child's and adult's) choice of referential expressions. In less highly presuming situations, different results no doubt would have been obtained and different conclusions drawn. A final criticism concerns Bamberg's failure to provide operational definitions of maintain and switch reference. Reference and Language Impaired Populations The use of referential devices by language impaired populations remains largely unexplored; developmental research appears rich by comparison. The results of three studies will be presented below. Fayne (1981) administered a reading task to learning disabled6 and two groups of nonhandicapped adolescents (academic and nonacademic) from grades ten and eleven. The group of nonacademic students had no specific learning disabilities but required reading materials that were two to three years below their actual grade placement. The task was designed to investigate compre-6Entrance requirements to the learning disabled program included the following: average intellectual functioning, two years' delay in a basic academic s k i l l area, and processing deficiencies (as measured by psychoeduca-tional tests). 148 hension of pronoun-antecedent relationships; each test item consisted of a three-sentence paragraph followed by a multiple choice question which required students to choose the best antecedent for a pronoun underlined in the paragraph (pronoun and antecedent occurred in different sentences); Both nonhandicapped groups performed significantly better than the learning disabled students. Unlike the nonhandicapped adolescents, learning disabled students performed as poorly on easy items as on difficult ones, suggesting that syntactic and semantic complexity of the antecedent were not major factors. The learning disabled subjects appeared to demonstrate a deficiency in their comprehension of discourse reference. Since a l l learning disabled students had histories of reading difficulties, Fayne proposed that earlier decoding difficulties may have led to a word-by-word reading style, a strategy that would prevent a reader from connecting elements of a cohesive tie that were separated by sentential boundaries. One implication is obvious: such students will have trouble following a writer's train of thought in even simple paragraphs. The referential skills of learning disabled adolescents7 and normal language adolescents aged 14 to 16 have also been compared on production tasks. Caro & Schneider (1982) examined referent-creating devices in data obtained from two tasks: a problem-solving activity and recall of two videotaped stories (one recalled orally and the other in writing). Learning disabled subjects used fewer appropriate devices (indefinite article + NP, [some] + plural NP) for a l l first mentions regardless of the task. On the 7Teachers of learning disabled children referred students who had oral expressive language problems. Students described by their teachers as having comprehension problems or primary sensory and emotional problems were exclu-ded. 149 narrative task, the learning disabled subjects used fewer appropriate devices (definite article + NP, and pronouns) for first mentions of human characters in both oral and written texts. Caro & Schneider's findings parallel those reported in the developmental literature (Warden 1976, Hickmann 1980, Bamberg 1987); adolescents with language production problems and young children both experience considerable difficulty when relying solely on linguistic means to introduce participants. Lovett, Dennis, and Newman (1986) examined the use of pronouns to achieve cohesion in the discourse (story retell) 8 of one right and two left hemidecor-ticate adolescents aged 12 to 14. Hemisphere-dependent differences as well as individual strategies were found. Both left hemidecorticate subjects had considerable difficulty planning extended discourse units. One avoided pronominal use in favor of explicit, and often redundant, nominals. He was least willing to rely on exophoric reference or on implicit devices like pronoun deletion. The other left hemidecorticate subject erred in the opposite direction, producing an abundance of both appropriate and ambiguous pronouns. A high proportion of nonpersonal pronouns (e.g. it) reduced narrative specificity. In contrast, pronominal use by the right hemidecorti-cate subject was both more economical and cohesive. He relied more on implicit cohesive elements and on intersentential cohesive relations than did the other subjects, and was able to maintain several intersecting referential chains simultaneously. The authors concluded that these findings were consistent with other descriptions of the subjects' language systems: although a l l three subjects performed similarly on tests examining lexical-8Each child recalled each of the following fairytales after they were read by the examiner: Little Red Riding Hood, The Frog Prince, The Practical Princess, and Goldilocks. 150 semantic competence, the right hemidecorticate subject was significantly superior to the left hemidecorticate subjects on syntactic knowledge. Although the small number of subjects enabled Lovett at al. to examine a large number of variables, i t also renders their conclusions regarding hemisphere-dependent differences somewhat premature, particularly in view of the many individual differences that were found. A second reservation concerns conclusions drawn without discussion of possible contextual or task effects. Based on their retrieval analysis, for example, the authors conclu-ded that a l l three subjects expressed primarily endophoric relations. However, since a l l subjects were free to use puppets in their retellings, it seems probable that the pronouns classified as endophoric also contained an exophoric component - for both speaker and listener. Another important contextual factor was the informed status of the listener, i.e. the experimen-ter who had just read the story to the subject. The fact that the stories were well-known fairytales no doubt also contributed to the shape of the results. All three of the above investigations represent i n i t i a l attempts to discover something about the discourse reference skills of language impaired populations; the territory clearly remains uncharted. The neglect of this area is surprising in view of the frequency of complaints by teachers and language clinicians concerning students whose difficulties appear to be connected to reference skills or lack thereof ( e.g. "unable to maintain topic," "difficult to follow," etc.). Given the increasing demands placed particularly on school-age children to understand and produce longer and more abstract texts as they progress through the school system, i t seems crucial to gain some understanding of what i t means to linguistically create and then to 151 follow up on one's referents. Some information is available from the develop-mental literature. In the research presented below, function-oriented analyses have been borrowed from developmental studies in an attempt to examine the nature of the difficulties language impaired children experience in their use of referential devices to create, maintain, and switch reference. More specifically, the alternation of nominal and pronominal devices will be analyzed to determine what i t reveals about the child's ability both to establish and maintain reference and to cohesively organize thematic pro-gression in narratives. Since there is l i t t l e information available, l i t t l e can be confidently predicted beyond what was suggested by the results of the retrieval analysis presented in chapter five, namely a greater reliance on the part of language impaired children on definite expressions to introduce participants. Based on the findings of Karmiloff-Smith and Bamberg regarding the youngest children in their studies, one might also expect to find the use of a more rigid global or thematic subject strategy in the narratives of the language impaired children than in those of normal language children. Results Devices used by the children to create, maintain, and switch reference were analyzed according to a method devised by Bamberg (1987). The narratives obtained from TI (picture narration) and T2 (retell of the same story without pictures) provided the data. Analysis of the forms used to create reference (first mentions) was restricted to three protagonists: the ship's crew and passengers (hereafter referred to simply as passengers), the ape, and the l i t t l e g i r l . To investigate linguistic devices used to maintain and switch 152 reference, references to the two major protagonists, the passengers and the ape, were analyzed. In addition to Bamberg's procedure, nominal versus pronominal references were also compared as a function of sentence position (subject versus nonsubject position). Analysis of First Mentions Each linguistic device used to introduce the passengers, ape, and girl was assigned to one of two categories: appropriate versus inappropriate. Appropriate forms consisted of the following: indefinite article + noun (e.g. an ape, some people), nonphoric this or these, possessive pronoun + noun, numerative + noun (e.g. three men), and comparative + noun (e.g. "another wierd person with a beard"). There were two types of inappropriate forms that occurred: definite article + noun, and pronouns. The children used various terms for the passengers, including people, men, sailors, and voyageur team. 9 When a collective term was not used (e.g. "a dad and a few sailors"), the individual participants were coded separately. Absolute frequencies of occurrence are provided in table XV. The data for a l l three protagonists have been collapsed. Normal language children used a greater percentage of appropriate devices than language impaired children on both tasks: 95% and 89% of NL children's devices were appropriate on Tl and T2 respectively, compared to 73% and 60% of LI children's devices on the same tasks. The most common appropriate device used by both groups was indefinite article + noun constructions. There were more instances of nonphoric this and these for first mentions in the data of NL children than in the data of LI children on both tasks; half of those used by NL children occurred in the 9Only one child actually used the term passengers in one of his stories. 153 Table XV.—Number of Each Type of Introductory Device (First Mention) used by LI and NL Children on TI and T2. TI T2 (picture-based) (retell,no pictures) LE NL LI NL Introductory devices Appropriate devices: indefin.art.+ noun 10 14 7 9 nonphoric this, these 1 4 3 6 possess.pron.+ noun 2 1 1 2 numerative + noun 2 - - -comparative + noun 1 - 1 -total appropriate 16 19 12 17 percent appropriate 73% 95% 60% 89% Inappropriate devices: definite art.+ noun 5 1 7 2 pronoun 1 - 1 -total inappropriate 6 1 8 2 ( % inappropriate) 27% 5% 40% 11% stories of one subject who introduced a l l three protagonists in both stories in this way. The majority of inappropriate devices involved the use of definite forms. A single instance occurred in the data of NL children on TI and two instances on T2; a l l three involved references to the g i r l . LI children, on the other hand, used definite forms in their introductions to a l l three protagonists. Use of a pronoun occurred in the data of one LI child only; on both tasks, he introduced the passengers by means of they. Switching and Maintaining Reference Each subsequent reference to the passengers and to the ape (in a l l sentence positions) was coded according to its form (nominal versus pronomi-nal) and its function (switch versus maintain reference). Nominal expressions included definite article + noun constructions and proper nouns. Pronominals 154 consisted of third person singular and plural pronouns (he, she, it, and they) and zero form. Possessives were excluded following Bamberg's procedure.10 A referential device was coded as maintaining reference (MR) if no other entity capable of contributing to the thematic progression of the story intervened between the device and its immediately preceding antecedent: 6.5 and the people followed the big path/ and then they came to a big suspension bridge In 6.5, they was coded as a pronominal form serving to maintain reference, since no animate entity intervened between i t and an immediately preceding mention of the same entity (the people). 11 If reference to'a different entity intervened, the device was coded as switching reference (SR): 6.6 then the ape looked up/ and the girl was falling/ and tie ape grabbed the g i r l . In 6.6, the second occurrence of the ape is coded as a nominal serving to switch reference since another entity (the girl) intervened between it and the immediately preceding reference to the ape. Zero forms in coordinated clauses with subject ellipsis were coded as maintaining reference to the preceding subject whether or not an animate entity intervened: 6.7 then the monkey picked her up and [9] put her on his head The zero form in the adjoined clause was coded as maintaining reference to the monkey in spite of the intervening reference to the girl (her). Table XVIa,b (Tl and T2 respectively) summarizes the distribution of 10According to Bamberg, whether an object belongs to one protagonist or to another is irrelevant to the thematic progress of the story. ^Entities capable of contributing to thematic progress were usually animate. There were two exceptions: ship and airplane. In some narratives, the ship carried the people back to their home in New York, and the airplane dropped a message to the ape. 155 Table XVI.—Percentage (and Absolute Number) of Nominal and Pronominal Devices used by L I and NL Children to Switch versus Maintain Reference t o the Passengers and to the Ape: (a) on T l and (b) on T2 (a) Task 1 (Picture-based) L I NL Switch Maintain Sum Switch Maintain Sum % to PASSENGERS Nominal 6.2 (9) - 6.2 (9) N 8.6 (18) - 8.6 (18) Prcncndnal 37.5 (54) 56.2 (81) 93.8(135) P 27.4 (57) 63.9(133) 91.3(190) Sum 43.8 (63) 56.2 (81) (144) Sum 36.1 (75) 63.9(133) (208) % to APE Nominal 31.1 (42) 11.9 (16) 43.0 (58) N 32.5 (51) 7.6 (12) 40.1 (63) Pronominal 29.6 (40) 27.4 (37) 57.0 (77) P 24.2 (38) 35.7 (56) 59.9 (94) Sum 60.7 (82) 39.3 (53) (135) Sum 56.7 (89) 43.3 (68) (157) % t o PASSENGERS + APE Nominal 18.3 (51) 5.7 (16) 24.0 (67) N 18.9 (69) 3.3 (12) 22.2 (81) Prcoaninal 33.7 (94) 42.3(118) 76.0(212) P 26.0 (95) 51.8(189) 77.8(284) Sum 52.0(145) 48.0(134) (279) Sum 44.9(164) 55.1(201) (365) (b) Task 2 ( R e t e l l . without Pictures) L I NL Switch Maintain Sum Switch Maintain Sum % to PASSENGERS Nominal 1.6 (2) 2.5 (3) 4.1 (5) N 10.0 (16) 1.2 (2) 11.2 (18) Pronominal 32.2 (39) 63.6 (77) 95.9(116) P 23.8 (38) 65.0(104) 88.8(142) Sum 33.9 (41) 66.1 (80) (121) Sum 33.8 (54) 66.2(106) (160) % t o APE Nominal 42.0 (34) 7.4 (6) 49.4 (40) N 39.8 (47) 1.7 (2) 41.5 (49) Prcncndnal 12.3 (10) 38.3 (31) 50.6 (41) P 15.3 (18) 43.2 (51) 58.5 (69) Sum 54.3 (44) 45.7 (37) (81) Sum 55.1 (65) 44.9 (53) (118) % t o PASSENGERS + APE Nominal 17.8 (36) 4.4 (9) 22.3 (45) N 22.7 (63) 1.4 (4) 24.1 (67) Pronominal 24.3 (49) 53.5(108) 77.7(157) P 20.1 (56) 55.8(155) 75.9(211) Sum 42.1 (85) 57.9(117) (202) Sum 42.8(119) 57.2(159) (278) 156 referential devices according to function. The observations of individual children have been collapsed. NL children referred slightly more often to the passengers than to the ape on both tasks (57.0% of the referential devices involved the passengers on TI and 57.6% on T2). LI children referred almost as often to the ape (48.4%) as to the passengers (51.6%) on TI in spite of the fact that the ape did not appear until one-third of the way through the story. The passengers received greater focus by LI children on T2 (59.9%) than on TI. NL children and LI children, therefore, performed similarly on T2 with respect to distribution of focus between the two protagonists. When the functions of maintaining and switching reference were compared irrespective of the forms used, both groups maintained reference more often than they switched reference to the passengers on both tasks. This was not an unexpected finding, given the greater number of total references to the passengers as reported above. NL children performed similarly on both tasks. LI children used a smaller percentage of devices to maintain reference than NL children on TI (63.9% to maintain by NL children versus 56.2% by LI children), but increased this percentage on T2 so that the two groups performed identi-cally on T2 (66.1 and 66.2% of devices to maintain reference to the passengers by LI and NL respectively). With respect to the ape, both groups switched reference more often than they maintained i t on both tasks. When nominal versus pronominal forms (to the passengers and the ape combined) were compared irrespective of function, the two groups shared almost identical distributions on both tasks: both groups used a pronominal in three out of four references to the two protagonists (combined). References to the passengers by both groups involved a pronominal form over 90% of the time with NL children tending to use a slightly smaller proportion of pronominals than 157 LI children, particularly on T2 (88.8% by NL compared to 95.6% by LI). A preference for pronominals was also evident, albeit to a lesser degree, in references to the ape by NL children on both tasks and by LI children on TI. On T2, LI children showed an even split between nominals and pronominals to refer to the ape. The distribution of nominal versus pronominal forms according to the function served is summarized in tables XVIIa,b and XVIIIa,b. As table XVIIa indicates, both groups showed a preference for pronominal forms to switch reference to the passengers on both tasks; the preference of LI children was stronger than that of NL children on both tasks. LI children increased their percentage of pronominals to switch reference to the passengers (from 85.7% on TI to 95.1% on T2) whereas those of NL children decreased (76.0% on TI to 70.4% on T2). Group and task differences were also evident when devices used to switch reference to the ape were compared. NL children used a greater percentage of nominals than pronominals to switch reference to the ape on both tasks. LI children showed an even split between pronominals and nominals to switch reference to the ape on TI but used a greater proportion of nominals than pronominals on T2 (51.2% nominals on TI to 77.3% on T2). NL children also increased the percentage of nominals to switch reference to the ape on T2 (from 57.3% on TI to 72.3% on T2). The distribution of nominal versus pronominal forms used to maintain reference is shown in table XVIIIa,b. Both groups used pronominals to maintain reference to the passengers almost 100% of the time on both tasks. Both groups also preferred pronominals to nominals to maintain reference to the ape. NL children used a greater percentage of pronominals to maintain 158 Table XVII.—Percentage (and Absolute Number) of Nominal versus Pronominal Devices used by LI and NL Children to Switch Reference to the Passengers and to the Ape: (a) on TI and (b) on T2 (a) Task 1 (Picture-based) SWITCH REFERENCE % to PASSENGERS % to APE % to PASSENGERS + APE LI NL LI NL LI NL Nominal 14.3 (9) 24.0 (18 51.2 (42) 57.3 (51) 35.2 (51) 42.1 (69) Pronominal 85.7 (54) 76.0 (57) 48.8 (40) 42.7 (38) 64.8 (94) 57.9 (95) Sum 100% (63) (75) (82) (89) (145) (164) (b) Task 2 (Retell, without Pictures) LI NL LI NL LI NL Nominal 4.9 (2) 29.6 (16) 77.3 (34) 72.3 (47) 42.4 (36) 52.9 (63) Pronominal 95.1 (39) 70.4 (38) 22.7 (10) 27.7 (18) 57.6 (49) 47.1 (56) Sum 100% (41) (54) (44) (65) (85) (119) Table XVIII.—Percentage (and Absolute Number) of Nominal versus Pronominal Devices used by LI and NL Children to Maintain Reference to the Passengers and to the Ape: (a) on TI and (b) on T2 (a) Task 1 (Picture-based) MAINTAIN REFERENCE % to PASSENGERS % to APE % to PASSENGERS + APE LI NL LI NL LI NL Nominal - - 30.2 (16) 17.6 (12) 11.9 (16) 6.0 (12) Pronominal 100.0(81) 100.0(133) 69.8 (37) 82.4 (56) 88.1(118) 94.0(189) Sum 100% (81) (133) (53) (68) (134) (201) (b) Task 2 (Retell, without Pictures) LI NL LI NL LI NL Nominal 3.8 (3) 1.9 (2) 16.2 (6) 3.8 (2) 7.7 (9) 2.5 (4) Pronominal 96.2 (77) 98.1(104) 83.8 (31) 96.2 (51) 92.3(108) 97.5(155) Sum 100% (80) (106) (37) (53) (117) (159) 159 reference to the ape than LI children on both tasks (82.4% by NL children compared to 69.8% by LI children on TI, and 96.2% by NL children compared to 83.8% by LI children on T2). The percentage of nominals decreased consider-ably on T2 for both groups, to a negligible amount in the NL data (3.8%). Sentence Position The narratives obtained on TI and T2 once again provided the data for the analysis. The passengers, ape, and minor characters were coded as to whether they occurred in subject or nonsubject position and whether a nominal or pronominal form was used. Minor characters were treated as a group and included any potential actor capable of advancing thematic progression of the story (e.g. the g i r l , her father, the captain of the ship, the natives on the island, and others). Subjects of subordinate and adjoined clauses with subject ellipsis were included in the analysis. Distribution of devices according to sentence position (subject versus nonsubject) and protagonist is summarized in table XIXa,b. Distribution of focus between the passengers and the ape was reported in the previous section. The percentage of references to minor characters combined was highly similar across groups and tasks (27 to 30%). Both groups showed a small preference for nominal forms to refer to minor characters irrespective of sentence position; the preference was stronger for LI children than NL children on both tasks (63.2 and 61.9% on TI and T2 respectively for LI children compared to 54.5 and 55.3% for NL children). No task differences were present. A comparison of subject versus nonsubject positions, irrespective of form, showed that both groups preferred subject position to refer to a l l three "protagonists" (passengers, ape,and minor characters combined). However, the 160 Table XIX. —Percentage Nonsubject Position as Characters: (a) on Tl (and Absolute Number) of Nominal and Pronominal Devices in Subject versus used by LI and NL Children to refer to the Passengers, Ape, and Minor and (b) on T2 % to PASSENGERS Naninal 9.5 (13) Pronominal Total % to APE Nominal Pronominal Total 90.5(124) 91.3(137) 39.8 (45) 60.2 (68) 79.6(113) % to MINOR CHARACTERS Naninal 70.3 (45) Pronominal 29.7 (19) Total 60.4 (64) (a) Task 1 (picture-based) LI Subject Nonsubject Total 7.7 (1) 92.3 (12) 8.7 (13) 62.1 (18) 37.9 (11) 20.4 (29) 52.4 (22) 47.6 (20) 39.6 (42) 9.3 (14) 90.7(136) (150) 44.4 (63) 55.6 (79) (142) 63.2 (67) 36.8 (39) (106) N P Total N P Total N P Total Subject 10.1 (19) 89.9(170) 87.1(189) 38.8 (52) 61.2 (82) 82.2(134) 58.5 (48) 41.5 (34) 56.6 (82) NL Nonsubject Total 25.0 (7) 75.0 (21) 12.9 (28) 65.5 (19) 34.5 (10) 17.8 (29) 49.2 (31) 50.8 (32) 43.4 (63) 12.0 (26) 88.0(191) (217) 43.6 (71) 56.4 (92) (163) 54.5 (79) 45.5 (66) (145) (b) Task 2 (Retell, without Pictures) LI Subject Nonsubject Total % to PASSENGERS Naninal 6.7 (8) Pronominal Total % to APE Naninal Pronominal Total 93.2(111) 93.7(119) 44.1 (30) 55.9 (38) 76.4 (68) % to MINOR CHARACTERS Naninal 60.5 (26) Pronominal 39.5 (17) Total 51.2 (43) 25.0 (2) 75.0 (6) 6.3 (8) 85.7 (18) 14.3 (3) 23.6 (21) 63.4 (26) 36.6 (15) 48.8 (41) 7.9 (10) 92.1(117) (127) 53.9 (48) 46.1 (41) (89) 61.9 (52) 38.1 (32) (84) N P Total N P Total N P Total Subject 14.4 (22) 85.6(131) 92.2(153) 39.2 (40) 60.8 (62) 82.3(102) 58.2 (39) 41.8 (28) 54.5 (67) NL Nonsubject Total 15.4 (2) 84.6 (11) 7.8 (13) 68.2 (15) 31.8 (7) 17.7 (22) 51.8 (29) 48.2 (27) 45.5 (56) 14.5 (24) 85.5(142) (166) 44.4 (55) 55.6 (69) (124) 55.3 (68) 44.7 (55) (123) 161 strength of this preference varied depending on the protagonist. References to the passengers occurred in subject position approximately 90% of the time and in nonsubject position 10% of the time. Those to the ape occurred in subject position approximately 80% of the time. References to minor charac-ters occurred in subject position approximately 60% of the time on Tl. LI children reduced the percentage of subject position references to minor characters from 60.4% on Tl to 51.2% on T2. Distribution of focus between the protagonists in terms of subject position alone is shown in table XXa,b. Both groups reserved subject position for the passengers more often than for the ape, and for the ape more often than for minor characters combined. For LI children, the tendency to reserve subject position for the passengers increased on T2, rising from 43.6% to 51.7% of references occurring in subject position. LI children showed a corresponding decrease in the percentage of subject position references to the ape on T2 compared to Tl (29.6% on T2 compared to 36.0% on Tl). The perfor-mance of the normal language children was consistent across tasks: approxi-mately half of subject position references were made to the passengers, 30% to the ape and 20% to minor characters combined. Distribution of focus as a function of nonsubject position is displayed in table XXIa,b. In general, nonsubject position references involved minor characters more often than the ape, and the ape more often than the passen-gers. The groups performed almost identically with respect to the percentage of nonsubject position references involving minor characters; both groups increased these references from approximately 50% on Tl to 60% on T2. The groups differed, however, with respect to nonsubject references to the passengers and the ape. On both tasks, particularly Tl, nonsubject references 162 Table XX.—Distribution of Focus in Subject Position: Percentage (and Absolute Number) of Referential Devices in Subject Position that Refer to the Passengers, Ape, or Minor Characters as Used by LI and NL Children: (a) on TL and (b) on T2 SUBJECT POSITION (a) Task 1 (Picture-^ ased) % to PASSENGERS % to APE % to MINOR CH. TOTAL LI Nominal 12.6 (13) 43.7 (45) 43.7 (45) 32.8 (103) Pronominal 58.8 (124) 32.2 (68) 9.0 (19) 67.2 (211) Total 43.6 (137) 36.0 (113) 20.4 (64) (314) NL Nominal 16.0 (19) 43.7 (52) 40.3 (48) 29.4 (119) Pronominal 59.4 (170) 28.7 (82) 11.9 (34) 70.6 (286) Total 46.7 (189) 33.1 (134) 20.2 (82) (405) (b) Task 2 (Retell, without Pictures) % to PASSENGERS % to APE % to MINOR CH. TOTAL LI Nominal 12.5 (8) 46.9 (30) 40.6 (26) 27.8 (64) Pronominal 66.9 (111) 22.9 (38) 10.2 (17) 72.2 (166) Total 51.7 (119) 29.6 (68) 18.7 (43) (230) NL Nominal 21.8 (22) 39.6 (40) 38.6 (39) 31.4 (101) Pronominal 59.3 (131) 28.1 (62) 12.7 (28) 68.6 (221) Total 47.5 (153) 31.7 (102) 20.8 (67) (322) by NL children involved the passengers more often and the ape less often than nonsubject references by LI children. With respect to the type of referential device used, pronominals were used to refer to the passengers when they occurred in subject position an overwhelming majority of the time for both groups (see table XIXa,b). The groups performed almost identically on TI, pronominals being used approxi-mately 90% of the time; on T2, NL children used a greater percentage of 163 Table XXI. —Distribution of Focus in Nonsubject Position: Percentage (and Absolute Number) of Referential Devices in Nonsubject Position that Refer to the Passengers, Ape, or Minor Charac-ters as Used by LI and NL Children: (a) on Tl and (b) on T2 NONSUBJECT POSITION (a) Task 1 (Picture-based) % to PASSENGERS % to APE % to MINOR CH. TOTAL LI Naninal 2.4 (1) 43.9 (18) 53.7 (22) 48.8 (41) Pronominal 27.9 (12) 25.6 (11) 46.5 (20) 51.2 (43) Total 15.5 (13) 34.5 (29) 50.0 (42) (84) NL Nominal 12.3 (7) 33.3 (19) 54.4 (31) 47.5 (57) Pronominal 33.3 (21) 15.9 (10) 50.8 (32) 52.5 (63) Total 23.3 (28) 24.1 (29) 52.5 (63) (120) (b) Task 2 (Retell, without Pictures) % to PASSENGERS % to APE % to MINOR CH. TOTAL LI Nominal 4.3 (2) 39.1 (18) 56.5 (26) 65.7 (46) Pronominal 25.0 (6) 12.5 (3) 62.5 (15) 34.3 (24) Total 11.4 (8) 30.0 (21) 58.6 (41) (70) NL Nominal 4.3 (2) 32.6 (15) 63.0 (29) 50.5 (46) Pronominal 24.4 (11) 15.6 (7) 60.0 (27) 49.5 (45) Total 14.3 (13) 24.2 (22) 61.5 (56) (91) nominals than LI children. When referring to the ape in subject position, the groups performed identically on Tl, using pronominals approximately 60% of the time and nominals 40% of the time; on T2, LI children increased their percen-tage of nominals. As noted above, minor characters occurred least frequently in subject position; when they did, nominals tended to be the preferred form. This preference was stronger for LI children than for NL children on Tl, but both groups performed similarly on T2. 164 As noted above, nonsubject position was most frequently filled by minor characters. Distribution of nominals versus pronominals in nonsubject position for minor characters showed a fairly even split except for LI children on T2; they increased the percentage of pronominals from 52% on TI to 63% on T2. Nonsubject references to the ape usually took the form of nomi-nals. On TI, the groups performed similarly, involving nominals in 60 to 65% of their nonsubject references to the ape. NL children performed consistently across tasks; LI children increased their use of nominals from 62% on TI to 86% on T2, resulting in a considerably greater use of nominals for the ape in nonsubject position than LI children on T2. There were too few nonsubject references to the passengers to permit meaningful comparisons. As in subject position, pronominals were preferred in the majority of cases. Distribution of focus in terms of nominal versus pronominal forms is shown in tables XXa,b, and XXIa,b. In subject position, both groups performed similarly on TI (table XXa): nominals in subject position usually referred either to the ape (44% of the time) or to minor characters (40-44%). Pronomi-nals in subject position usually referred to the passengers (60% of the time) and somewhat less frequently to the ape (29-32% of the time). Nominals were rarely used to the passenger in subject position (13-16%), and pronominals were rarely used for minor characters in subject position (9-12%). Group and task differences were found on T2 (table XXb). Nominals in subject position s t i l l referred to the ape and to minor characters more often than to the pasengers, but more of the NL children's nominals in subject position referred to the passengers on T2 than on TI whereas more of the LI children's nominals in subject position referred to the ape. With respect to pronominals in subject position, LI children used a greater percentage than NL children to 165 refer to the passengers and a smaller percentage to refer to the ape. In nonsubject position (table XXIa,b), the percentage of nominals to refer to the passengers was very small on both tasks. Just over half of nonsubject nominals referred to minor characters on TI; this percentage rose somewhat on T2 for NL children. On both tasks, a greater percentage of the nonsubject nominals used by LI children referred to the ape than those used by NL children. With respect to nonsubject pronominals, those of LI children referred more often to the ape than those of NL children on TI. On T2, the groups perfomed identically with approximately 60% referring to minor charac-ters, 25% to the passengers and 15% to the ape. Discussion Creating Reference An analysis of the devices used to create reference to the main protago-nists of a narrative revealed the use of a greater percentage of inappropriate devices on the part of language impaired children than normal language children. Nearly a l l devices judged inappropriate involved the use of definite article + noun constructions. This finding is a subset of a more general one reported in chapter five, namely that of a higher use by language impaired children of what was coded as exophora. Those results were interpre-ted as supporting a difference in coding orientation between the two groups: the language impaired children demonstrated a greater reliance on the nonver-bal context and were less likely to selfcontextualize than the normal language children. The results of chapter five and those reported in the present chapter suggest that the code of language impaired children resembles in some 166 respects that of young children as described in developmental studies (Warden 1976, Karmiloff-Smith 1979, Hickmann 1980, Martin 1983, Bamberg 1987). The results support the findings of Caro & Schneider (1982), who found that adolescents with production problems used fewer appropriate referent-creating devices for first mentions than normal subjects in discourse. The finding of a high use of indefinite constructions to introduce participants by the normal language children conflicts with the findings of Bamberg (1987). His normal adults used definite constructions in approxi-mately half of their introductions. The children used an even higher percen-tage; the youngest never used indefinite forms. However, Bamberg's narratives were gathered in conditions that would lend themselves to a higher use of definite reference: both the child and caregiver could see the pictures, and both were very familiar with the story before the data that was eventually analyzed was collected. Bamberg's results, together with those of the current study, underline the importance of a major tenet of systemic func-tional grammar, namely that the selection of a particular form is highly dependent on contextual factors. Certain forms, such as referent-creating devices, appear to be more highly sensitive to contextual factors than others. The selection of these forms appears to vary according to the age of the child (as reported in developmental studies) and according to linguistic "intact-ness" as suggested by the results of the present study and by the findings of Caro & Schneider (1982). Further research is required which will examine the productions of language impaired versus normal language children in response to varying contextual configurations and which will investigate the nature of individual strategies. One of the normal language children of the present study, for 167 example, introduced most of his referents with nonexophoric this or these: 6.8 there was these people that got on some kind of boat/ and they spot this island/ . . . I and they find this indian tribe/ and they have al l these little huts sitting on sticks/ . . . / and they see this ape swinging on a swing with this big tree that crosses through this canyon (SC5, Tl) According to Halliday & Hasan, the use of nonphoric this or these, as in 6.8, is typical of "highly coded, in-group speech" used by a speaker to convey "a sense of immediacy and also of solidarity with the hearer" (1976: 61), i.e. one of the means narrators have at their disposal for emphasizing common experience with their listeners. A second, related function may be to direct the listener to what the speaker is currently focusing on. In addition to first mentions, many subsequent references in the data of this particular child also contained this or these. These, too, may have been motivated by the speaker's intent to convey shared interest and/or focus to the listener. Switching and Maintaining Reference With respect to distribution of focus among the story's protagonists, the passengers were referred to more often than the ape, which in turn was referred to more often than a l l minor characters combined. The increased number of references to the passengers was no doubt due to the fact that the ape did not appear in the pictures until one-third of the way through the story. There was one exception to the above trend, however; on Tl, the language impaired children referred just as often to the ape as to the passengers. Since a l l speakers familiarized themselves with the story as told by the pictures before actually telling i t , this finding suggests a slightly different planning strategy between the two groups, at least on Tl. On that task, normal language children tended to organize their utterances more around 168 the passengers and to focus less on the ape than the language impaired children. A similar finding was reported by Bamberg (1987) in his developmen-tal study. Despite the fact that the two main characters of his stimulus story were depicted equally as often, both adults and older children tended to refer to the human protagonist (the boy) more frequently than to the nonhuman protagonist (the dog). This difference was less pronounced, however, in his younger subjects. In the present study, the difference in the number of references to the passengers and to the ape may not solely be explained by the number of pictures involving each protagonist but also by the way each character is perceived with respect to potential for advancing thematic progression. The normal language children, like the adults and older children of Bamberg's study and unlike the language impaired children, seemed to render this distinction in the number of references made to each protagonist. On retell, the language impaired children altered their distribution of focus in the direction of the normal language children's performance. The above findings on distribution of focus no doubt relate to the results obtained from a comparison of devices used to switch versus maintain reference irrespective of the forms used. The increase in the percentage of devices used to maintain reference by the language impaired children on T2 parallels their increase in the number of references made to the passengers on that task. Once again the two groups performed almost identically on T2 due to a change on the part of the language impaired children. The normal language children on both tasks and the language impaired children on T2 tended to organize their stories more around the passengers and to t e l l them in such a way that reference was maintained more often than switched. This was not the case for references to the ape: both groups switched reference 169 more often than they maintained i t . The greater percentage of reference-switching devices for the ape may be due both to the distribution of focus in favor of the passengers as suggested above and to story content. In two sections of the narrative—at the outset of the story and after the passengers leave the natives in search of the ape—there are several pictures in which no potential actors other than the passengers are depicted. This results in at least two sequences of several utterances in which reference to the pasengers is maintained due to the absence of other characters. There are no such sequences involving the ape alone. Story content may also account for the greater percentage of devices used to maintain reference by the children in the present study compared to the children of Bamberg's study. The latter switched reference almost twice as often as they maintained it in their references to both the boy and the dog. The adults of his study maintained and switched reference in equal propor-tions. Bamberg's finding of a greater use of reference maintaining devices by the older children and adults compared to the younger children parallels the greater use of devices to maintain reference to the passengers by the normal language children compared to the language impaired children on TI of the present study. When nominal versus pronominal references to the two main protagonists combined were compared, the groups performed almost identically. However, this finding is misleading on its own. When references to the passengers and those to the ape were examined separately, i t immediately became evident that the two characters were being treated differently with respect to the form of the referential device. The children used pronominals to refer to the passengers 90% of the time or more compared to 60% of the time when referring 170 to the ape. Potential for advancing the story's plot may once again be a factor. Bamberg found a similar trend towards a greater use of pronouns for animate characters (the boy) than inanimate particularly in his younger subjects. He concluded that for the young child, pronominal forms were "the prototypical candidates for thematic progression, while nominal forms are adding on information that does not directly contribute to thematic progres-sion" (1987: 96). This kind of explanation can only partially account for the results of the present study since the ape, too, had a significant role to play in advancing thematic progression. The greater use of nominal forms for the ape than for the passengers may indeed be related to an animate/inanimate distinction, but this distinction may not neatly match up with direct/indirect contribution to thematic progression respectively. Another possible factor is grammatical number: it may have been relatively more difficult to come up with a collective term for the passengers than to refer to a singular noun. In addition, each child was told the name of the story, i.e. The Great Ape, so that this character was named for the child; the passengers, on the other hand, were not "named" in any way. The next step in the analysis was to examine the relation between the particular form used and the function i t served. As noted, the adults of Bamberg's study consistently used nominal expressions to switch reference and pronominal expressions to maintain reference. The young children of his study used this "anaphoric" strategy only when referring to the dog; only the oldest children showed evidence of a similar strategy when referring to the boy. The normal language children of the present study performed more like the adults and older children of Bamberg's study than did the language impaired children: they used more nominals to switch reference and more pronominals to maintain 171 reference than the language impaired children. This generalization held for references to either of the main protagonists (passengers or ape). Task effects were also apparent. The normal language group tended to move in the direction of a more adultlike strategy in their retellings (T2). The language impaired children, however, increased their percentage of pronouns to switch reference to the passengers from 86% on Tl to 95% on T2, resulting in a considerably higher percentage on T2 than that of normal language children on the same task (95% versus 70%). The forms used to either switch or maintain reference varied considerably depending on whether reference was made to the passenger or to the ape. All the children of the present study resembled the young children of Bamberg's study in their high use of pronouns to switch reference to the passengers and to a lesser degree to the ape. Even the normal language children on T2, for example, used a pronominal form 70% of the time when they switched reference to the passengers. There are a number of factors that may have contributed to the different results obtained on the two studies. One of these concerns the greater familiarity with the stimulus story on the part of Bamberg's subjects. Some support for this was found in the present study in the types of changes that were made on retell of the same story, changes which, for the normal language children, were often in the direction of a more adultlike strategy. Story content may be a second factor. The stimulus story used in the present study involved more participants and was much longer than Bamberg's story. Number and gender differences of the main protagonists, for example, certainly played a role. In Bamberg's story, the two main protagonists, the boy and the dog, could be referred to by means of the same pronoun {be). In such a story, the 172 need for an increased number of nominals when reference i s switched from one protagonist to the other i s obviously greater than i n the present study i n which the main protagonists differed i n number (passengers [pi] versus ape [sg]) and i n gender (in sections of the narrative involving the g i r l ) . Number and gender information carried by a particular pronoun may have sufficed i n many cases to "switch reference." The fact that the present study used a longer story than that of Bamberg's may have resulted i n a greater use of pronouns as the story progressed, i . e . as the l i s t e n e r ' s store of knowledge about story protagonists grew. This appeared to be the case i n a study by Marslen-Wilson, Levy, & Tyler (1982): their normal adult subject increasingly referred to story protagonists i n less s p e c i f i c ways as the story progressed. In the present study, even when speakers appeared to adhere to an " a d u l t l i k e " strategy, l i s t e n e r s were s t i l l i n some cases required to refer to their growing representation of the story's characters and events i n order to interpret the r e f e r e n t i a l devices used. In the following excerpt from a story by a normal language c h i l d (T2), for example, the speaker used a nominal to switch reference and a pronominal to maintain reference, but the same nominal i s used to switch reference to two different groups of people: the passengers of the ship and the natives of the isla n d : 6.9 and then the people came out and scared the voodoo people away/ and they ran o f f / and they hid/ and the t u r t l e s kept racing/ and then the people went on/ and they found t h i s big t r a i l on a mountain/ and then the people ran back out of their hiding places and kept watching the t u r t l e race/ and the people followed the big path. Bamberg's "anaphoric" strategy, the use of nominals to switch reference and pronominals to maintain reference, would not on i t s own provide enough 173 information to enable the l i s t e n e r to i d e n t i f y referents i n 6.9. However, by t h i s point i n the narrative the l i s t e n e r has already acquired knowledge about the story's characters. This, i n addition to what i s predicated about the participants i n 6.9, enables the l i s t e n e r to readily i d e n t i f y the speaker's intended referents. Since the voodoo people are the ones that were scared away, they are the ones that "ran o f f , " "hid," and who l a t e r "ran back out of their hiding place and kept watching the t u r t l e race." The passengers from the ship are the only possible referent for the people who "came out and scared the voodoo people away," and who "went on" and "found t h i s big t r a i l on a mountain," since they were the ones who had been going from one place to another discovering things on the island. The "anaphoric" strategy s t i l l plays a r o l e : the use of a nominal appears to at least signal a change i n reference. Each time the people i s used, a change i n reference i s intended. The above comments were intended to point out some lim i t a t i o n s of the method of analysis used by Bamberg and adopted i n the present study. Bam-berg's "anaphoric" strategy must not be generalized beyond the s p e c i f i c context i n which i t occurred. The differences between his findings and those of the present study underline the degree to which r e f e r e n t i a l devices are discourse-sensitive and the need for research to examine a greater number of contexts before more general conclusions can be drawn. The analysis w i l l require r e f i n i n g i n order to take into account such c r u c i a l factors as number and gender of major participants. The effect of more "intangible" f a c t o r s -such as the degree to which l e x i c a l information provided by prior or subse-quent predicates influences the speaker's choice of r e f e r e n t i a l devices — w i l l be more d i f f i c u l t to assess. Despite i t s l i m i t a t i o n s , the analysis revealed differences i n the use of 174 r e f e r e n t i a l devices by normal language and language impaired children. The results provided some evidence suggesting that language impaired children were more l i k e the younger children of Bamberg's study and less l i k e the adults of his study than were their normal language peers. Perhaps the most s i g n i f i c a n t finding was the tendency for language impaired children to use more pronouns to switch reference and more nominals to maintain reference than the normal language children on both tasks. I t was suggested i n the previous section, that language impaired children were not using conventional l i n g u i s t i c forms to introduce participants as consistently as normal language children, r e s u l -ting i n the use of many d e f i n i t e forms for both f i r s t and subsequent mentions. A functional analysis of subsequent mentions indicated once again a problem i n mapping form i n a conventional way to functions served. There are l i n g u i s t i c means available to mark given/new information and to indicate that reference i s being switched or maintained. The results suggest that language impaired children are marking these d i s t i n c t i o n s less consistently than normal language children, that they are indeed experiencing d i f f i c u l t y using discourse-sensitive forms to mark discourse-level functions. It i s important to add that the language impaired children of the present study were able to make these d i s t i n c t i o n s some of the time. The oppportunity to r e t e l l the story appeared b e n e f i c i a l i n some respects; the language impaired children performed more l i k e the normal language children on T2 than on TI with respect to d i s t r i b u t i o n of focus among the protagonists and percentage of devices used to maintain versus switch reference to each of the protagonists irrespective of form. However, when form was examined i n conjunction with the function served, the language impaired children did not perform as well as normal language children on either task. In the case of 175 at least one variable, i n fact, they moved in the opposite d i r e c t i o n : whereas normal language children increased their use of nominals to switch reference to the passengers on T2, the language impaired children decreased the number of nominals i n their references to the passengers i n their r e t e l l i n g s . This w i l l be discussed i n the following section. Thematic Strategy As noted above, comparison of the d i s t r i b u t i o n of nominal versus pronomi-nal forms indicated d i f f e r e n t i a l treatment of the two protagonists. Karmi-loff-Smith (1981) and Bamberg reported similar findings i n their developmental studies. The young children of both studies appeared to make use of what the authors ca l l e d a "thematic subject strategy." For Karmiloff-Smith, t h i s involved reserving the use of pronouns i n u t t e r a n c e - i n i t i a l position exclu-s i v e l y for the story's main protagonist; references to minor characters occurred i n nonsubject position i n the form of nominals. Bamberg interpreted the preference, p a r t i c u l a r l y on the part of his younger subjects, for pronomi-nal forms to refer to the boy for both maintaining and switching reference as a manifestation of a similar thematic strategy. His analysis, however, did not compare the use of r e f e r e n t i a l devices as a function of subject versus nonsubject position. The results of such an analysis i n the present study provide additional insight into the notion of "thematic strategy" as i t might occur i n a much longer narrative. Both normal language and language impaired groups appeared to be making some use of a thematic subject strategy, a l b e i t i n a much less r i g i d form than the younger children of Karmiloff-Smith's study. Such a strategy would account for the variation observed i n the use of r e f e r e n t i a l devices as a 176 function of protagonist and subject position. The children referred d i f f e r e n -t l y to major versus minor characters, and also distinguished between major protagonists. Minor characters occupied u t t e r a n c e - i n i t i a l position only 20% of the time whereas over half of nonsubject references involved minor charac-ters. Nominals for minor characters were preferred over pronominals i n subject position and were used as frequently as pronominals i n nonsubject position. The passengers, on the other hand, occupied almost half of utter-a n c e - i n i t i a l references and were involved i n only 10% of nonsubject refer-ences. Subject position was preferred over nonsubject position 90% of the time when references to the passengers alone were considered. Pronominals were overwhelmingly preferred over nominals i n both subject and nonsubject position to refer to the passengers. A pronominal i n subject position referred most often (60% of the time) to the passengers. The ape was treated l i k e the passengers with respect to sentence position: subject position was preferred 80% of the time. Pronominals were preferred i n subject position but to a lesser degree than i n the case of the passengers. In other respects, the children c l e a r l y distinguished between the ape and the passengers i n how re f e r e n t i a l devices were deployed. Only half as many u t t e r a n c e - i n i t i a l positions involved the ape as they did the passengers, and a far weaker preference for pronominals was evident i n th i s position (60% compared to 90% i n the case of the passengers). Whereas pronominals were preferred i n references to the passengers i n both subject and nonsubject positions, nominals were preferred i n references to the ape i n nonsubject position 60% to 70% of the time. The children of the present study, though adhering to an ad u l t l i k e ("anaphoric") strategy to a certain degree i n their references to the ape, 177 showed form and sentence position preferences which varied with the protago-n i s t , i n particular with how central a role he or she was perceived to play i n the story and perhaps with his or her humanness. The ape was seen as having a more central role than any of the minor characters but not as central a role as the passengers. The "humanness" of the passengers may explain the greater preference for pronominals i n their case and the higher percentage of nominals used for the ape. In an attempt to explain similar findings i n his own data, Bamberg suggested that the young narrator "measures along a continuum the importance of the protagonists: an abstract object i s less important for the narrative progress than a concrete object; an inanimate less than an animate; a non-human less than a human" (1987: 95). According to Bamberg, The young c h i l d tends to relate the use of pronouns more to the function of advancing story plot than to the function of maintaining reference. He called this strategy a global one since "the function of the referring device i s estab-lished for the text as a whole" (p.97). With respect to group differences, the results of the analysis of devices used to switch/maintain reference suggested that the normal language children were making greater use of Bamberg's "anaphoric" strategy than the language impaired children. The results of the sentence position analysis indicated that, although both groups showed evidence of a thematic subject strategy, the language impaired children were more r e s t r i c t e d by such a strategy. A stronger thematic subject strategy on the part of language impaired children was p a r t i c u l a r l y evident on T2. The lack of pictures no doubt placed greater demands on the c h i l d to provide his own organization of the story and may have resulted i n recourse to a more global strategy. The strength of a thematic strategy on the part of a l l the children of 178 the present study, including the oldest normal language children, suggests that such a strategy may not disappear with adulthood as Bamberg seems to imply. In their examination of the discourse ( r e t e l l of a comic book story) of their adult subject, Marslen-Wilson et a l . (1982) found sequences of several utterances i n which one of the characters was established by name, placed i n subject position, and subsequently referred to by pronouns or zero anaphora. Spontaneous repairs i n favor of reserving u t t e r a n c e - i n i t i a l position for the thematic subject and for pronominalizing i t there were also found. Brown & Yule (1983) argue for a similar strategy on the part of the adult l i s t e n e r , i . e . for a reference-resolution strategy whereby the l i s t e n e r interprets pronouns as references to the main character. The lack of evidence of such a strategy i n Bamberg's adult data may be due to the s i m p l i c i t y of the story he chose and to the fact that the story had been rehearsed. His "anaphoric" strategy may be more t y p i c a l of rehearsed oral or written narra-tives than of informal spontaneous s t o r i e s . The "global strategy" (thematic) Bamberg attributes to children may also be t y p i c a l of adults i n some contexts. CHAPTER SEVEN CONCLUSIONS The general purpose of the present research was to investigate the nature of the discourse d i f f i c u l t i e s experienced by language impaired school-age children. S p e c i f i c a l l y , the study examined some of the l i n g u i s t i c devices available to a speaker that contribute to the production of coherent narrative discourse. Both group and task differences were found i n the use of the l i n g u i s t i c resources that l i n k one part of a text to another. Language impaired children used fewer cohesive devices per utterance than normal language children, a finding that appeared to be related to the following: 1) shorter utterance length on the part of language impaired children, 2) fewer attempts on their part to provide cohesive l i n k s , p a r t i c u l a r l y l e x i c a l ones and 3) a greater number of f a i l e d attempts on their part. Retrieval analysis indicated a greater use of anaphoric and esphoric reference (forward reference within the nominal group) by normal language children than by language impaired children. The l a t t e r findings, together with the results of an analysis of referent-creating devices suggested a greater reliance on nonverbal context on the part of language impaired children than normal language children and a smaller tendency to self-contextualize. An examina-tion of the devices used to switch and to maintain reference revealed that the language impaired children used more pronouns to switch reference and more nominals to maintain reference than the normal language children. The r e f e r e n t i a l devices of both groups of children varied as a function of the protagonist. Both groups showed evidence of a thematic subject strategy which 179 180 distinguished between major and minor protagonists and between the two major protagonists. There were r e l a t i v e l y few s i g n i f i c a n t task effects. Those that occurred tended to hold for one group or the other, a finding which suggested that the two groups were responding d i f f e r e n t l y to changes i n contextual configuration. Normal language children, for example, used a higher proportion of demonstra-tive reference and l e x i c a l cohesion on T2 (story r e t e l l without pictures) than on T3 (video r e c a l l ) , i . e . they tended to use a greater proportion of nominals on T2 and a higher proportion of pronominals on T3. The fact that the video involved human protagonists only or that i t was highly action-oriented may have accounted for the greater use of pronouns on t h i s task. Although the language impaired children did not show these particular contextual effects, other task-related differences were evident. The language impaired children seemed to benefit more from the opportunity to r e t e l l the picture-based story; they tended to perform more l i k e the normal language children on T2. On TI, the language impaired children used fewer conjunctive t i e s and incorporated events into complex utterances less often than normal language children. On T2, however, the language impaired children increased the frequency of conjunctions to equal those of the normal language children. Retrieval analysis also showed a more similar pattern of results between the two groups on T2 compared to TI. The language impaired children's d i s t r i b u -tion of focus among story protagonists and proportion of devices used to maintain versus switch reference was also similar to those of the normal language children on T2. There was an important exception, however: whereas normal language children tended to employ a more "anaphoric" or a d u l t l i k e strategy on T2 compared to TI, the language impaired children, l i k e younger 181 children of developmental studies, showed a more r i g i d use of a thematic subject strategy on T2. This finding may relate to the increased memory demands of th i s task. Martin's notion of code was invoked i n view of the differences observed between the groups with changes i n context. Like the older versus younger children of Martin's study, the normal language and language impaired children of the present study appeared to "take different paths through their grammar, the same context drawing from each a different register response" (1983: 5). The particular paths that have been explored by th i s research can be characterized as involving ph o r i c i t y . Phoric structures are the l i n g u i s t i c devices available to speakers to distinguish, for the benefit of thei r l i s t e n e r s , between given and new information. The results of the present study suggest that language impaired children do not exhibit the same degree of control over phoric structures as normal language children. They tend to resemble developmentally younger children i n the options they select, or f a i l to select, i n view of making one utterance contingent on a prior one. The language impaired children of the present study did not lack the requisite l i n g u i s t i c forms; their range of l i n g u i s t i c options did not appear to be r e s t r i c t e d . There are at least two possible explanations, which are not mutually exclusive: 1) language impaired children have more trouble d i s t i n -guishing between given and new information, s p e c i f i c a l l y from the lis t e n e r ' s perspective and/or 2) they do not mark th i s d i s t i n c t i o n i n conventional ways as consistently as age-matched normal language children. L i l e s (1985) argued against a problem i n adopting l i s t e n e r perspective based on her finding that language impaired children altered t h e i r use of cohesive devices i n ways similar to normal language children as the lis t e n e r 182 condition changed (naive versus informed). Although not formally tested i n the present study, there was evidence i n the data to support L i l e s ' conclu-sion. Only one language impaired c h i l d used a highly presuming form (pronoun) to introduce a referent and only when i t followed a l e x i c a l item which may i t s e l f have permitted the li s t e n e r to infer the required information (the use of ship before the presuming item they). Additional evidence was supplied by the changes made by the language impaired children on T2 ( r e t e l l without p i c -tures) : they increased the number of d e f i n i t e r e f e r e n t i a l devices to i n t r o -duce participants, ignoring at some l e v e l the experimenter's instruction to pretend that the l i s t e n e r for T2 was someone who had never heard the story. However, the data do not rule out the p o s s i b i l i t y of subtle d i f f i c u l t i e s i n t h i s area. As Maratsos suggested, even adult speakers sometimes use de f i n i t e reference for e n t i t i e s that are "given" for the speaker but not for the l i s t e n e r : "The definiteness favored by the s p e c i f i c i t y of reference for the self sometimes takes precedence over the i n d e f i n i t e reference specified by the c r i t e r i a of calculating the l i s t e n e r ' s knowledge even for adults at various times" (1976: 105). Awareness of l i s t e n e r knowledge and of the relevance of the li s t e n e r ' s viewpoint no doubt continues to develop well into adulthood. What the results did support was a reduced a b i l i t y on the part of language impaired children to l i n g u i s t i c a l l y demonstrate their awareness of their l i s t e n e r ' s knowledge using conventional means. This, too, i s a s k i l l that no doubt develops over a long period of time, even i n normal language individuals. The given/new d i s t i n c t i o n was not marked as often or as well by language impaired children as i t was by normal language peers. Language impaired children appear to have d i f f i c u l t y selecting the appropriate forms 183 according to what a l i s t e n e r can or cannot be assumed to know. The domain of phoric structures i s discourse; the appropriate selection of such devices implies a knowledge of the functions they serve at the l e v e l of discourse. Referential devices do not merely "refer," but also organize the text. A particular device signals to the l i s t e n e r whether reference i s being created, maintained, or switched. Choice of a referring expression may mark episodic organization; the use of a nominal for a referent just mentioned, for example, may signal the beginning of a new episode. Such form-function pairings are obviously context-sensitive. The d i f f i c u l t y faced by the language impaired children i s not only the general one of discovering a systematic relationship between these forms and their discourse-level functions, but also that of discovering appropriate form-function pairings s p e c i f i c to a given context. In Johnston's terms (1985), the language impaired children of t h i s study had d i f f i c u l t y with rules incorporating discourse categories, a d e f i c i t she refers to as a true pragmatic disorder. As noted i n chapter one, i t i s d i f f i c u l t to neatly separate t h i s disorder from a formal one. Lovett et a l . (1986), for example, found that their right hemidecorticate subject performed better than their l e f t hemidecorticate subjects on measures of pronominal cohesion and also demonstrated superior syntactic s k i l l s . Right and l e f t hemidecorticate subjects performed s i m i l a r l y on measures of lexical-semantic competence. Achievement of textual coherence implies that the forms are present i n the child's grammatical system and that they are accessible. A l l of the language impaired children of the present study had d i f f i c u l t i e s with form i n the past. A general lag i n l i n g u i s t i c maturity may have resulted i n reduced f l e x i b i l i t y with t h i s grammatical system and, therefore, i n a more limited a b i l i t y to exploit i t to achieve texture. A tendency away from verbal 184 encoding would be one re s u l t : as noted, the language impaired children of thi s study used fewer cohesive devices per utterance and demonstrated a greater reliance on s i t u a t i o n a l context, when i t was available, than the normal language children. A c h i l d with a history of d i f f i c u l t i e s with form can also be expected to have pragmatic d i f f i c u l t i e s of the kind described i n this study, and the pragmatic d i f f i c u l t i e s of the children i n this study cannot be f u l l y understood outside of an examination of form. The most important implication of these findings, and those of other studies reported i n th i s paper, i s the importance of discourse analysis to c l i n i c a l assessment. Phoric structures are highly context-sensitive; their selection depends on both verbal and nonverbal contextual factors. Pictures or a previous t e l l i n g of the story, for example, w i l l influence the choice of cohesive devices. Referring expressions can be expected to vary depending on whether reference i s being created, maintained or switched. Protagonists are referred to i n different ways depending on such factors as the c e n t r a l i t y of their role i n the story and on their "humanness', i . e . on their potential for advancing thematic progress. Number (singular or plural) and gender of protagonists w i l l influence the degree of l e x i c a l s p e c i f i c a t i o n , as w i l l the parti c u l a r point i n the story's unfolding that the expression occurs (see Marslen-Wilson et a l . 1982). A speaker's selection of phoric expressions i s part of the dynamic process of text construction; i t i s dependent on what has preceded, on what the narrator wishes to focus on currently, and on what w i l l follow. The pro-cesses involved i n th i s selection procedure can only be understood i f the unit of analysis i s larger than the sentence. As Karmiloff-Smith has noted, "some aspects of language may be t o t a l l y misconceived i f an explanation i s sought at 185 the isolated sentence l e v e l " (1981: 125). Tests currently available to c l i n i c i a n s for examining school-age populations are obviously inadequate; none assess textual devices i n a systematic way. The analyses used i n the present study—adapted from the research of Halliday & Hasan (1976, 1985), Martin (1977, 1983), and Bamberg (1987)—represent possible ways of investigating these devices i n any number of different contexts. Because of the lack of normative data, these analyses must be used i n conjunction with other forms of assessment. The main goal i n their use would be to obtain q u a l i t a t i v e information that could contribute to a p r o f i l e of a child's l i n g u i s t i c strengths and weaknesses. The above comments are also applicable to intervention. To target the development of cohesive devices at the "isolated sentence l e v e l " would be as "misconceived" as analyses based on t h i s unit. Texture must be targeted using texts, preferably those whose contextual configurations correspond to those the c h i l d t y p i c a l l y encounters i n home and school experiences. The school setting, i n p a r t i c u l a r , requires that children be exposed to a variety of texts, since, with each academic subject, content and associated lexicogramma-t i c a l expression can be expected to vary. Within the overall goal of more coherent text i s that of a greater control over the structures available i n English for distinguishing between what a l i s t e n e r can and cannot be assumed to know. This goal does not consist i n greater economy of expression. According to the results of the present study and those of Bamberg's research, there are other more c r u c i a l factors determining the form of a r e f e r e n t i a l device, for example, whether reference i s being created, maintained, or switched. Given the findings of the present study, s p e c i f i c goals might be the following: to increase the number of cohesive devices used, p a r t i c u l a r l y 186 the number of l e x i c a l t i e s ; to decrease reliance on perceptual (nonverbal) support and increase self-contextualization; to increase use of esphora (forward reference within the nominal group); to increase the use of i n d e f i -nite expressions to introduce referents; and to increase the use of pronomi-nals to maintain reference and nominals to switch reference. Both nominal and verbal anaphoric devices can be targeted. These goals are not a l l l i k e l y to be relevant to every language impaired c h i l d . To assist i n determining individualized goals for intervention, Lahey (1988) has developed an informal method of analysis i n the form of a checklist (reproduced i n appendix 6). The s e n s i t i v i t y of cohesive devices to contextual factors underlines the importance of using a variety of contexts both i n assessment and intervention. An additional goal of therapy suggested by the findings of th i s study i s that of developing the a b i l i t y to make the appropriate adaptations with changes i n context. As the abstractness and scope or variety of topics increases with grade l e v e l , so do the demands on a child's a b i l i t y to make the appropriate adjustments. Narratives may be p a r t i c u l a r l y helpful i n th i s respect since each one represents a different contextual configuration. Story content may be controlled for number and gender of protagonists, the i r r e l a t i v e importance to thematic progress, and degree of humanness. Listeners could be varied whenever possible and may include other children being seen for therapy, classmates, family members, teacher, etc. The s t o r y t e l l i n g task could be varied to include picture narration tasks, r e t e l l of a story read to the c h i l d or of one he or she has previously t o l d , video r e c a l l , personal experience narratives, f i c t i t i o u s story generation, and story completion. The fact that language impaired and normal language children performed more s i m i l a r l y i n several respects on T2, i . e . on r e t e l l of the same story used i n the picture 187 narration task, suggests that language impaired children may benefit consi-derably from rehearsal. New s k i l l s may be more easily acquired on such a task, i . e . on a second t e l l i n g . This recommendation may be applied not only by speech/language c l i n i c i a n s but also by educators who work within the classroom setting on a regular basis. Cohesion s k i l l s must also be targeted using genres other than the narrative. This may include other types of monologue, such as procedural and exposition text, or conversation. However, patterns of cohesion can be expected to d i f f e r across genres and t h i s must be kept i n mind when comparing the p r o f i l e of strengths and weaknesses of a given c h i l d obtained i n different contexts. Another potential context for intervention i s dramatic play. P e l l e g r i n i (1984) compared the effect of dramatic play, discussion, and drawing on the story r e c a l l s of children i n kindergarten, grade one, and grade two (aged 5 to 8). Children of grades one and two who were exposed to the dramatic play condition used more endophora than those exposed to either the discussion or drawing treatments; kindergarten children i n the dramatic play group used less exophoric e l l i p s i s than children i n the discussion group. P e l l e g r i n i suggested that the effectiveness of dramatic play i n producing more cohesive narratives arose from the greater opportunity i n that context to l e x i c a l i z e meaning. The children i n that condition negotiated verbally the play's theme and interpretation of characters even before the play episodes could be i n i t i a t e d , with the result that players' assumptions were made e x p l i c i t . The players had to define e x p l i c i t l y any "pretend" objects that were used, i . e . idiosyncratic symbolic transformations were verbally defined. Children i n the drawing condition did not interact with others; those i n the discussion group merely responded to the experimenter's questions, questions 188 which themselves supplied the relevant phoric structures. Research into the use of cohesive devices i n English i s i n an exploratory phase. Empirical investigations, including the present study, have been limited by the small number of contexts analyzed. Additional contexts are required i n which narrative task and content are varied. Lexical s p e c i f i c a -tion i n creating, maintaining and switching reference, for example, appears to be dependent on such factors as story length, grammatical number and gender of protagonists, episodic structure, nonverbal support, and l i s t e n e r status (naive or informed). Further research i s required into the effect of manipu-l a t i o n of these parameters. In school-age populations, both oral and written modes can be tapped. Contexts other than the narrative (e.g. other forms of monologue and conversation) w i l l provide information on the use of cohesion i n interactive situations. I n i t i a l research attempts have revealed important differences i n the d i s t r i b u t i o n of cohesive devices between narrative and conversational contexts (Rochester & Martin 1979, Mentis & Prutting 1987, Harris et a l . 1987). Although both conversational and narrative data were collected for the present study, only the narrative data were analyzed for thi s report. A comparison of the patterns of cohesion of the same individuals i n both narrative and conversational discourse may be expected to y i e l d a more complete picture of the use of cohesion by these children. Even the narratives analyzed for the present report have not yet been f u l l y explored for the information on cohesion they contain. False s t a r t s , for example, were transcribed, but not analyzed. The following excerpt i l l u s t r a t e s the kind of struggle that i s sometimes experienced by language impaired children i n thei r attempts at r e f e r e n t i a l tying: 7.1 and he caught i t with her, and she, and he caught i t with, and, and she caught, and he [both children laugh], and he 189 caught i t , and he caught her with her, with his toe.(S2T2) A study of errors and self-corrections may y i e l d important q u a l i t a t i v e information on the s p e c i f i c d i f f i c u l t i e s experienced by individual language impaired children. Generalizations from the present study are also limited to a small sector of English speakers, namely male subjects i n the age range of 9 to 11 years. Martin's data (1977) suggest that there may be gender differences i n the use of phoric structures: the g i r l s of his study appeared to be better at introducing information that was l a t e r presumed, and at presuming information known to the l i s t e n e r instead of reintroducing i t . The age range must also be expanded. Adult data would have been p a r t i c u l a r l y useful i n the present study; the normal language children showed evidence of a f a i r l y strong thematic subject strategy and the question remains as to whether such a strategy disappears or whether i t i s retained i n some form. Research i s also required that w i l l involve different kinds of impaired populations. The problems of the language impaired children of t h i s study did not appear to stem primarily from a d i f f i c u l t y i n knowing what was given and new for the l i s t e n e r but from a reduced a b i l i t y to consistently express this d i s t i n c t i o n using conventional forms. A d i f f i c u l t y i n taking l i s t e n e r perspective, however, could not be immediately ruled out i n other language impaired individuals, for example i n the a u t i s t i c population. Information on the r e f e r e n t i a l s k i l l s of language impaired children i s scarce. The results of the present study were compared primarily to developmental research due to the lack of relevant studies on impaired populations. This i s recognized as a l i m i t a t i o n since the d i f f i c u l t i e s of a language impaired c h i l d are not l i k e l y to be simply those of a younger normal language c h i l d . The nature of compen-190 satory strategies, for example, can be expected to be related to the particu-l a r language problem. This raises yet another l i m i t a t i o n of the present study: the data were not examined for individual strategies. Bamberg's i d e n t i f i c a t i o n (1987) of a number of individual strategies amoung his normal language subjects enables one to predict that the same may be true, possibly to a greater extent, among language impaired children. The present study was also limited by an examination of production data alone. A c h i l d who has d i f f i c u l t y expressing the d i s t i n c t i o n between given and new information may also find i t d i f f i c u l t , as a l i s t e n e r , to comprehend the l i n g u i s t i c signals that mark phoricity, and to use such signals to i d e n t i f y speaker reference. There i s some evidence for t h i s i n the area of reading comprehension: Fayne (1981), for example, found that learning disabled adolescents had d i f f i c u l t y i d e n t i f y i n g pronoun antecedents i n simple paragraphs. This type of enquiry needs to be extended to l i s t e n i n g comprehen-sion as well. As i n the case of other l i n g u i s t i c phenomena, the relationship between the comprehension and production of cohesive devices would shed considerable l i g h t on the nature of the d i f f i c u l t i e s experienced by language impaired children. It was noted e a r l i e r that the most important implication of the findings of t h i s study i s the importance of discourse to c l i n i c a l assessment and intervention. One qualifying remark must be added by way of conclusion. Although textual devices cannot be adequately analyzed at the sentence l e v e l , the d e f i c i t s of language impaired children w i l l not be f u l l y described i f d i s -course-level relations are examined i n i s o l a t i o n of sentence l e v e l con-s t r a i n t s . Linguists have long distinguished between reference i n which the anaphoric expression i s s y n t a c t i c a l l y related to i t s antecedent and reference 191 in which the r e l a t i o n between the anaphoric expression and i t s referent i s established pragmatically. I t was noted i n chapter s i x that existing theories of reference have largely been developed by l i n g u i s t s preoccupied with the search for syntactic constraints. It would be equally misguided to err in the opposite d i r e c t i o n , i . e . to interpret a l l r e f e r e n t i a l devices as being constrained solely by discourse-level considerations. Bosch's work (1983) represents a more balanced view; he provides formal constraints for the mechanisms involved i n both syntactic and pragmatic modes of pronominal reference. The present study examined only the l a t t e r . However, both textual and sentential phenomena—and their interaction—must eventually be i n v e s t i -gated. Cohesive devices, reference i n p a r t i c u l a r , may be ideal phenomena for studying the interaction between discourse and sentential constraints i n the process of text construction. Such knowledge would provide a p a r t i a l answer to the more general question of how the major semantic components (experien-t i a l , interpersonal, and textual) are woven together i n a single lexicogramma-t i c a l r e a l i z a t i o n . Only then w i l l texture be understood i n i t s proper l i g h t , i . e . as part of a much larger scheme. BIBLIOGRAPHY Bamberg,M. (1987). The Acquisition of Narratives. B e r l i n : Mouton de Gruyter. Bartolucci,G. & Fine,J. (1987). The frequency of cohesion weakness i n psychiatric syndromes. Applied Psycholinguistics, 8, 67-74. Bennett-Kastor,T. (1983). Noun phrases and coherence i n c h i l d narratives. Journal of Child Language, 10, 135-149. . (1986). Cohesion and predication i n c h i l d narrative. Journal of Child Language, 13, 353-370. Bernstein,L.E. (1981). Language as a product of dialogue. Discourse Processes, 4, 117-147. Bloom,L., Lahey,M., Hood,L., L i f t e r , K . , & Fiess,K. (1980). Complex sentences: acquisition of syntactic connectives and the semantic relations they encode. Journal of Child Language, 7, 235-261. Bloom,L., Rocissano,L., & Hood,L. (1976). Adult-child discourse: developmental interaction between information processing and l i n g u i s t i c knowledge. Cognitive Psychology, 8, 521-552. Bosch,P. (1983). Agreement and Anaphora: A Study of the Role of Pronouns in Syntax and Discourse. London: Academic Press. Brown,G. & Yule,G. (1983). Discourse Analysis. Cambridge: Cambridge Univer-s i t y Press. Caro,D. & Schneider,P. (1982). Creating referents i n text: a comparison of learning disabled and normal adolescents' texts. Proceedings from the Third Wisconsin Symposium on Research in Child Language Disorders. University of Wisconsin, Madison. Clancy,P., Jacobsen,T.,& Silva,M. (1976). The acquisition of conjunction: a c r o s s - l i n g u i s t i c study. Papers and Reports on Child Language Develop-ment, 12, 71-80. Donahue,M. (1981). Requesting strategies of learning disabled children. Applied Psycholinguistics, 2, 213-234. . (1985). Communicative style i n learning disabled children: some implications for classroom discourse. In D. Ripich & F. S p i n e l l i (Eds.), School Discourse Problems. San Diego: C o l l e g e - H i l l Press. Donahue,M., Pearl,R., & Bryan,T. (1980). Learning disabled children's conver-sational competence: responses to inadequate messages. Applied Psycho-linguistics, 1, 387-403. 192 193 Donahue,M., Pearl,R., & Bryan,T. (1982). Learning disabled children's syntac-t i c proficiency on a communicative task. Journal of Speech and Hearing Disorders, 47. 397-403. Donnellan,K. (1978). Speaker reference, descriptions and anaphora. In P. Cole (Ed.), Syntax and Semantics: Pragmatics, vol.9. New York: Academic Press. Eisenberg,A. (1980). A syntactic, semantic, and pragmatic analysis of conjunc-t i o n . Papers and Reports on Child Language Development, 19, 70-78. Ervin-Tripp,S. (1978). Some features of early child-adult dialogues. Language in Society, 7, 357-373. Fayne,H. (1981). A comparison of learning disabled adolescents with normal learners on an anaphoric pronominal reference task. Journal of Learning Disabilities, 14, 597-599. Feagans,L. & Short,E. (1984). Developmental differences i n the comprehension and production of narratives by reading-disabled and normally achieving children. Child Development, 55, 1727-1736. Fey,M. & Leonard,L. (1982). Code-switching i n children with s p e c i f i c language impairment. Proceedings from the Third Wisconsin Symposium on Research in Child Language Disorders. University of Wisconsin, Madison. Fey,M. & Leonard,L. (1983). Pragmatic s k i l l s of children with s p e c i f i c language impairment. In T. Gallagher & C. Prutting (Eds.), Pragmatic Assessment and Intervention Issues in Language. San Diego: Co l l e g e - H i l l Press. Francis,H. (1974). Social class, reference and context. Language and speech, 17, 193-198. F r i e l - P a t t i , S . & Conti-Ramsden,G. (1984). Discourse development i n atypical language learners. In S. Kuczaj (Ed.), Discourse Development. New York: Springer-Verlag. Gallagher,T. & Darnton,B. (1978). Conversational aspects of the speech of language disordered children: revision behaviors. Journal of Speech and Hearing Research, 21, 118-135. Graybeal,C. (1981). Memory for stories i n language-impaired children. Applied Psycholinguistics, 2, 269-283. G r i f f i t h , P . , Ripich,D., & Dastoli,S. (1986). Story structure, cohesion, and propositions i n story r e c a l l s by learning-disabled and nondisabled children. Journal of Psycholinguistic Research, 15, 539-555. Halliday,M.A.K. (1977). Text as semantic choice i n so c i a l contexts. In T. van Dijk and J.Petofi (Eds.), Grammars and Descriptions. B e r l i n : Walter de Gruyter. 194 . (1979). Development of texture i n c h i l d language. In T. Myers (Ed.), The Development of Conversation and Discourse. Edinburgh: Edinburgh University Press. . (1985a). A Short Introduction to Functional Grammar. London: Edward Arnold. . (1985b). Dimensions of discourse analysis: grammar. In T. van Dijk (Ed.), Handbook of Discourse Analysis, vol.2. New York: Academic Press. Halliday,M.A.K. & Hasan,R. (1976). Cohesion in English. London: Longman. Halliday,M.A.K. & Hasan,R. (1985). Language, Context, and Text: Aspects of Language in a Social-Semiotic Perspective. V i c t o r i a : Deakin University Press. Harris,M., Schaeffer,L., & Lougeay-Mottinger,J. (1987). Use of cohesive devices by language impaired and non language impaired children. Paper presented at the 1987 American Speech-Language and Hearing Association Convention, New Orleans. Hasan,R. (1984). Coherence and cohesive harmony. In J. Flood (Ed.), Understan-ding Reading Comprehension. Newark: International Reading Association. Hawkins,P.R. (1973). Social class, the nominal group and reference. In B. Bernstein (Ed.), Class, Codes, and Control, vol.2. London: Routledge & Kegan Paul. Hickmann,M. (1980). Creating referents i n discourse: a developmental analysis of l i n g u i s t i c cohesion. In J. Kreiman and A. Ojeda (Eds.), Papers from the Parasession on Pronouns and Anaphora. Chicago: Chicago L i n g u i s t i c Society. Johnston,J. (1982). Narratives: a new look at communication problems i n older language-disordered children. Language, Speech, and Hearing Services in Schools, 13, 144-155. . (1985). The discourse symptoms of developmental disorders. In T. van Dijk (Ed.), Handbook of Discourse Analysis, vol.3. New York: Academic Press. Karmiloff-Smith,A. (1979). A Functional Approach to Child Language. Cambridge: Cambridge University Press. . (1981). The grammatical marking of thematic structure i n the development of language production. In W. Deutsch (Ed.), The Child's Construction of Language. London: Academic Press. . (1983). Language development as a problem-solving process. Papers and Reports on Child Language Development, 22, 1-22. 195 . (1984). Children's problem solving. Advances in Developmental Psychology, vol.3. Hillsdale,New Jersey: Lawrence Erlbaum Associates. Keenan,E. & Klein,E. (1975). Coherency i n children's discourse. Journal of Psycholinguistic Research, 4, 365-380. Keenan,E. & Schieffelin,B. (1976). Topic as a discourse notion: a study of topic i n the conversations of children and adults. In Charles N. L i (Ed.), Subject and Topic. Academic Press, Inc. Klecan-Aker,J. & Hedrick,D. (1985). A study of the syntactic language s k i l l s of normal school-age children. Language, Speech, and Hearing Services in Schools, 16, 187-198. Krahn,F. (1978). The Great Ape. New York: Viking Press. _. (1981). Arthur's Adventure in the Abandoned House. New York: E.P.Dutton, Inc. Lahey,M. (1988). Language Disorders and Language Development. New York: Macmillan Publishing Company. Leonard,L. (1986). Conversational replies of children with s p e c i f i c language impairment. Journal of Speech and Hearing Research, 29, 114-119. Lil e s , B . (1985). Cohesion i n the narratives of normal and language-disordered children. Journal of Speech and Hearing Research, 28, 123-133. . (1987). Episode organization and cohesive conjunctives i n narratives of children with and without language disorder. Journal of Speech and Hearing Research, 30, 185-196. Lovett,M., Dennis,M., & Newman,J. (1986). Making reference: the cohesive use of pronouns i n the narrative discourse of hemidecorticate adolescents. Brain and Language, 29, 224-251. Maratsos,M. (1976). The Use of Definite and Indefinite Reference in Young Children. Cambridge: Cambridge University Press. Marslen-Wilson,W., Levy,E., & Tyler,L. (1982). Producing interpretable discourse: the establishment and maintenance of reference. In R.J. J a r v e l l a and W. Klein (Eds.), Speech, Place, and Action. Chichester: John Wiley & Sons Ltd. Martin,J. (1977). Learning how to tell: semantic systems and structures in children's narratives. Doctoral di s s e r t a t i o n , University of Essex. (University Microfilms. No. 78-70,018). . (1983). The development of register. In J. Fine and R. Freedle (Eds.), Developmental Issues in Discourse. Norwood, New Jersey: Ablex Publishing Corp. 196 . (1985). Process and text: two aspects of human semiosis. In J. Benson and W. Greaves (Eds.), Systemic Perspectives on Discourse, vol.1. Norwood,New Jersey: Ablex Publishing Corp. McClure,E. & Geva,E. (1983). The development of the cohesive use of adver-sative conjunctions i n discourse. Discourse Processes, 6, 411-432. McTear,M. (1984). Structure and process i n children's conversational develop-ment. In S. Kuczaj (Ed.), Discourse Development. New York: Springer-Verlag. . (1985). Pragmatic disorders: a question of dir e c t i o n . British Journal of Disorders of Communication, 20, 119-127. Mentis,M. & Prutting,C. (1987). Cohesion i n the discourse of normal and head-injured adults. Journal of Speech and Hearing Research, 30, 88-98. Merritt,D. & Liles,B. (1987). Story grammar a b i l i t y i n children with and without language disorder: story generation, story r e t e l l i n g , and story comprehension. Journal of Speech and Hearing Research, 30, 539-552. M i l l e r , J . F . (1981). Assessing Language Production in Children. Austin: Pro-Ed. National Film Board of Canada. (1977). J wasn't scared. 20 min., video. Pappas,C. (1981). The development of narrative capabilities within a syner-gistic, variable perspective of language development: an examination of cohesive harmony of stories produced in three contexts - retelling, dictation and writing. Doctoral dissertation, Ohio State University. . (1985). The cohesive harmony and cohesive density of children's oral and written s t o r i e s . In J. Benson and W. Greaves (Eds.), Systemic Perspectives on Discourse, vol.2. Norwood: Ablex Publishing Corp. Pellegrini,A.D. (1982). The construction of cohesive text by preschoolers i n two play contexts. Discourse Processes, 5, 101-108. . (1984). The effect of dramatic play on children's generation of cohesive text. Discourse Processes, 7, 57-67. Prutting,C. & Kirchner,D. (1983). Applied pragmatics. In T. Gallagher and C. Prutting (Eds.), Pragmatic Assessment and Intervention Issues in Language. San Diego: Co l l e g e - H i l l Press. Radford,A. (1981). Transformational Syntax. Cambridge: Cambridge University Press. Ripich,D. & G r i f f i t h , P . (1988). Narrative a b i l i t i e s of children with learning d i s a b i l i t i e s and nondisabled children: story stucture, cohesion, and propositions. Journal of Learning Disabilities, 21, 165-173. 197 Ripich,D. & S p i n e l l i , F . (1985). School Discourse Problems. San Diego: Co l l e g e - H i l l Press. Ripich,D. & Terrell,B. (1988). Patterns of discourse cohesion and coherence i n Alzheimer's disease. Journal of Speech and Hearing Disorders, 53, 8-15. Roberts,J. (1987). Cohesive DO in children's language. Paper presented at the 1987 American Speech-Language and Hearing Association Convention, New Orleans. Rochester,S. & Martin,J. (1979). Crazy Talk: a Study of the Discourse of Schizophrenic Speakers. New York: Plenum Press. Roscoe,J. (1975). Fundamental Research Statistics for the Behavioral Sciences. New York: Holt, Rinehart & Winston, Inc. Roth,F. & Spekman,N. (1986). Narrative discourse: spontaneously generated stories of learning-disabled and normally achieving students. Journal of Speech and Hearing Disorders, 51, 8-23. Scott,C. (1984). Adverbial connectivity i n conversations of children 6 to 12. Journal of Child Language, 11, 423-452. . (1988). A perspective on the evaluation of school children's narratives. Language, Speech, and Hearing Services in Schools, 19, 67-82. Shugar,G. (1978). Text analysis as an approach to the study of early l i n g u i s -t i c operations. In N. Waterson and C. Snow (Eds.), The Development of Communication. Chichester: John Wiley & Sons. Siegel,S. (1988). Nonparametric Statistics for the Behavioral Sciences. New York: McGraw-Hill. Sleight,D. & Prinz,P. (1985). Use of abstracts, orientations, and codas i n narration by language-disordered and non-disordered children. Journal of Speech and Bearing Disorders, 50, 361-371. Warden,D. (1976). The influence of context on children's use of identifying expressions and references. British Journal of Psychology, 67, 101-112. Weaver,P. & Dickenson,D. (1979). Story comprehension and r e c a l l i n dyslexic students. Bulletin of the Orton Society, 29, 157-171. Westby,C. (1984). Development of narrative language a b i l i t i e s . In G. Wallach and K. Butler (Eds.), Language Learning Disabilities in School-Age Children. Baltimore: Williams & Wilkins. Wiese,B. (1983). Anaphora by pronouns. Linguistics, 21, 373-417. 198 APPENDIX 1 SUMMARY OF STORY CONTENT The Great Ape (Krahn, 1978) A young g i r l and her father are aboard a huge steamboat headed for an island (Picture 1). As they get closer, the passengers and crew members get into a small boat (P2) and come ashore (P3). They spot a v i l l a g e (P4) and find natives watching a t u r t l e race (P5). The captain of the ship speaks to one of the natives (P6) who then leads the passengers to a mountain (P7). The people from the ship begin climbing the mountain (P8), cross a suspension bridge (P9), and continue their climb up a narrow path along the side of a steep c l i f f (P10). They f i n a l l y arrive at the top (Pll) and begin taking pictures of what they see below (P12): an enormous ape s i t t i n g on a swing suspended from a huge f a l l e n tree trunk that spans a steep canyon (P13). Suddenly, the l i t t l e g i r l f a l l s when the rock beneath her crumbles (P14). The ape catches her with his toes (P15), l i f t s her up (P16), places her on his head (P17) and begins swinging (P18). The people from the ship, looking rather concerned (P19), blow up a huge i n f l a t a b l e banana (P20). The ape t r i e s to reach i t (P21) and the passengers and crew start running down the mountain with the banana (P22). The ape chases them down the mountain (P23) and through the v i l l a g e (P24). The passengers board the small boat and head for the ship with the ape following them i n the water (P25). The ape f i n a l l y gets the banana (P26), and the l i t t l e g i r l f a l l s off his head (P27). The ship s a i l s away (P28), and the ape realizes that the g i r l i s no longer on his head (P29). The g i r l s tarts to cry and waves goodbye to the ape (P30). The ship s a i l s back to New York (P31) and the g i r l ' s father takes a picture of her i n front of the Empire State building (P32). The ape, mean-while, i s looking extremely sad back on the island (P33). A plane comes by, with the father and daughter i n i t , and drops an enormous "message" to the ape (P34). The ape unrolls i t (P35) and hugs what he sees: a picture of the l i t t l e g i r l i n front of the Empire State building (P36). I wasn't scared (National Film Board, 1978) 1 I wasn't scared i s a dramatic enactment of what happens when a brother and s i s t e r venture onto a m i l i t a r y base i n search of a better trout stream. Instead of f i s h they f i n d a bomb. Todd, indulging an i l l - a d v i s e d c u r i o s i t y , digs out the bomb and unscrews the arming device. A thin smoke escapes. Shocked and scared, the two children run from the s i t e but Todd returns. He hasn't f u l l y s a t i s f i e d that c u r i o s i t y of h i s . Tension builds as Debbie flags down a car i n a desperate attempt to obtain help. Minutes l a t e r , an Explosives Disposal Team i s racing to the s i t e , but both Todd and the bomb have disappeared, the l a t t e r with a time-delay fuse burning inside i t . Working against an i r r e v e r s i b l e deadline, the Team f i n a l l y spot the boy carrying the bomb i n his knapsack. A near disaster i s averted - t h i s time. One of the team members helps Todd take his knapsack off and get to safety just before the bomb explodes. 2 ••Adapted from a summary provided by NFB. 2The video was stopped at this point. The story continues with a demonstration lecture on explosives. 199 APPENDIX 2 SAMPLE NARRATIVES Task 1, LI c h i l d (S4, aged l l , ^ ) 1 The Great Ape / 2 one day the people were walking along / they were s a i l i n g i n the ocean / then they saw some shore / they went and rowed a boat to shore / then they unloaded the boat and started to get thei r movie cameras out / then they started a fort / they were having races with t u r t l e s / they were talking to the other people who were racing the t u r t l e s / they were walking up to the top of the mountain / they crossed a bridge / and they're crossing another bridge / they were walking up a big mountain / th i s guy he spots a log i n the rock / and they want to f i n d out what the log i s doing there / so they're filming i t / and the g o r i l l a i s swinging on the log back and forth / and the g o r i l l a i s looking up at them by hanging on the rope / he was laughing and having a good time swinging / and the l i t t l e g i r l s l i p s and f a l l s / and the g o r i l l a saves the l i t t l e g i r l from f a l l i n g / and he has her i n hand / and then the g o r i l l a puts her on his head / then he starts swinging again / then they start to get a parachute out / and then they get a big 3 big banana out / the g o r i l l a goes after the banana / the g o r i l l a starts to run after the banana / and then they start to run away with the banana / the g o r i l l a i s chasing them / the g o r i l l a goes through the fo r t / they take the [lemon?] away / then they row away and take the banana with them / the g o r i l l a s i t s on their big boat that they're r i d i n g away i n / the g i r l f a l l s off the g o r i l l a / and the two men on board of the boat catches her / the g o r i l l a i s trying to eat the banana / he scratches his head while he's sinking / the l i t t l e g i r l i s crying cuz she misses the g o r i l l a / then the boat i s going into 1 False starts have not been included i n the following texts. 2Slashes separate independent clause units. 3Stressed words have been i t a l i c i z e d i n the text. 200 the town / the man i s taking pictures of the l i t t l e g i r l i n front of a building on the ship / the g o r i l l a i s hiding i n the mountains / an airplane comes by and drops a r o l l of paper with a bow on to the g o r i l l a / the g o r i l l a opens i t and sees part of a building / the g o r i l l a sees the l i t t l e g i r l i n a building i n a picture / and that's the end Task 1, NL c h i l d (SC6, aged 9;3) here i s our great story for a l l you kids out there l i s t e n i n g on your walkmans and radios / The Great Ape! / once there was a man and his l i t t l e daughter Sarah / and they went on a big ship / and they sai l e d down to a big island / and then they got onto a little rowboat and rowed to that big island / they brought cameras and boxes and bags and a gun and a big camera box / and Sarah was very interested i n a l l the stuff around her / then they came to a big v i l l a g e where a l l these voodoo men were a l l dancing around watching turtle races / they were t u r t l e s with domino-like cubes with l i t t l e dots on them for their number / then they came over to a l l those voodoo indian men and scared them half to death / and they put their hands up i n the a i r / and they ran off / Sarah was very happy / then they saw a big t r a i l that led way up a big mountain / so they followed a l l the way up / and they came to a big suspension bridge / and i t led right into a tunnel / they came through the tunnel and got to a l l these trees and more of the path / they kept on going tray tray up high into the mountain ' t i l they came to a big tree where they looked down / they could see a l l the mountains below them with the big tree going out s t i l l / then right i n the middle of the tree, they saw a big, big, ape on a swing / and he just sat there and looked down / then he looked over / and he saw this l i t t l e Sarah / she f e l l down off the c l i f f / and the monkey swang and swang / and then Sarah grabbed onto his toes, his very big toes / and then he grabbed her i n his hands, brought her up / and she sat on his finger / and they stared eye to eye / and then he put Sarah up on top of his head and swang and swang 201 on the swing / then the people got a big i n f l a t a b l e banana / and they blew i t up and unpeeled i t / and the monkey reached up for i t / but i t was too high for him / they ran off with the banana / the monkey climbed up with Sarah on top of his head / he climbed and climbed the mountain / they went down the t r a i l as fast as they could with that big banana / they ran and ran and ran I and the monkey followed right behind them / they ran by the people watching the t u r t l e s race / they rowed back to the big boat / the monkey followed them a l l that way / then f i n a l l y got to the boat / the monkey got the banana / and lifted i t way up i n the a i r and sat on the boat / and the boat was halfway up in the a i r / i t pulled the anchor right out of the ground / and that monkey, he loved that banana / then the g i r l on his head, Sarah, she was holding on very tight / then she dropped down / and Mr. Pig and Mr. Wigwam captured her / and she was safe / and the monkey ran off with his big banana / then he looked back and scratched his head and wondered why he couldn't eat that banana / and the people looked back at her / and Sarah waved goodbye / and then they got back to their hometown i n the big apple New York and took a picture of Sarah standing by the Empire States building and flew over i n a big plane down to where the monkey was, the giant ape, and dropped him down a big picture / and he opened i t up and saw the Empire States building / and he opened i t up more and more / and Sarah was s i t t i n g i n the picture by the Empire States building / and that's the end of our story Task 2, LI c h i l d (S4) f i r s t of a l l , i t was about t h i s g o r i l l a book / and these people wanna discover t h i s g o r i l l a / so they get i n a boat / and they s a i l away / then they stop the big boat and get into a smaller boat and row into shore / and then they take their stuff out of the rowboat / and they start walking / and then they meet th i s fort / and then they are walking / and then they see these people who are racing t u r t l e s / and then they start talking to them / then 202 they leave that / then they climb the mountain / and then they find the g o r i l l a / then they find a log / they found the g o r i l l a swinging on a log with some s t r i n g t i e d to another piece of log as a swing / then the l i t t l e g i r l slipped and f e l l / and then the g o r i l l a saved her / they wanted the l i t t l e g i r l back / so they pumped up a banana and showed the g o r i l l a i t / and the g o r i l l a went after them / and the g o r i l l a didn't get i n [ i t ? ] ' t i l he sat on the boat / and the l i t t l e g i r l slipped off the g o r i l l a ' s head / and then they went to the boat / and then these two men caught her / then they went back into the c i t y / they sai l e d into the c i t y / then they took a picture / then they flew an aiplane over the g o r i l l a and threw the picture out of the airplane and gave i t to the g o r i l l a / and the g o r i l l a opened i t / and there's a l i t t l e g i r l i n front of a building / and the g o r i l l a ' s there Task 2, NL c h i l d (SC6) The Great Ape! / once there was a l i t t l e g i r l , and her father, named Sarah / they went on a big ship and went down the r i v e r i n the ocean / they went down the ocean / and then the captain caught sight of an island way down south / and they stopped the boat where i t was nice and deep i n the water / and they got a rowboat / and a l l the people gathered i n and rowed down to the island / and they got onto the island and walked and walked a long way ' t i l they came to a l i t t l e voodoo mumbojumbo house where a l l these people were watching turtle races / and they had l i t t l e dots l i k e on dominos or dice for their numbers / i f there's a s i x , there'd be s i x dots on their big plate / and then the people came out and scared the voodoo people away / and they ran off / and they hid / and the t u r t l e s kept racing / and then the people went on / and they found this biff t r a i l on a mountain / then the people ran back out of their hiding places and kept watching the t u r t l e race / and the people followed the big path / and then they came to a big suspension bridge / they went down that / then they went through a big tunnel / and they came to where 203 there was l o t s of trees around the mountain / and they kept following the t r a i l way down / then they came to a big c l i f f where there was a big tree s t i c k i n g out from one c l i f f to another / then they looked down / they saw a big ape s i t t i n g on a swing / and Sarah looked down / and she leaned too far / and she f e l l / and the monkey swang and swang as he watched her f a l l / then she grabbed hold of his toe and climbed up / then the monkey picked her up and put her on his head / and then the people they blew up a big banana from a big box that's big and i n f l a t a b l e / and they brought i t down the t r a i l past the voodoo people watching the race / the race still wasn't over / they kept going and going across the suspension bridge / they got to the rowboat / they rowed, rowed and rowed and rowed / and the monkey kept on following and following / then he came over and sat down on the boat and got the banana and wondered why he couldn't eat i t / and he scratched his head / and the people rowed away i n their big ship / they started going off a l i t t l e ways then stopped / and Sarah waved goodbye to the monkey / and then they got back to their hometown i n the big apple New York and took a picture of Sarah i n the boat by the Empire States building and went i n a big plane, flew over to where the monkey was i n the mountains and threw down a big, big, picture of Sarah by the Empire States building / then the monkey caught i t and looked at i t / he opened i t / he could see the top of the Empire States building / then he opened i t more and more / and i t was a giant picture of Sarah and the Empire States building / and that i s the end of our story Task 3, LI c h i l d (S4) i t was about t h i s boy / these two kids they were going f i s h i n g / and they were playing tag / and t h i s boy he rode behind a tree / and his s i s t e r went past him / and then he took off and tripped over a bomb / and then his s i s t e r went to get help / his brother called her / and then his s i s t e r said: "leave i t , leave i t ! " / and then he took off the propeller / then she went to go get 204 the cops / and they were racing / l i k e they were racing the bikes / and then they went 7 and this g i r l stopped and the man and the car / and then the car went to the police / and the police phoned the army / and the army went and got the bomb off this kid's back cuz he put i t i n his knapsack / and then he put i t i n his knapsack / and the kid couldn't get i t off him / and then the guy said: "I'm coming to help you" / and then he went and helped him / and then he took the knapsack off / and then he put i t down on the ground / then he ran away / and then i t exploded Task 3, NL c h i l d (Sc6) f i r s t there was a g i r l and a boy / they were rid i n g their bikes down a road / and then they came to th i s big fence / and they climbed over i t /and they went to a lake / and they weren't supposed to go there / and they went fi s h i n g / and then they were there for a long time / then they decided to leave / then they started to get i n a fight / then the boy pushed the g i r l over and then ran off / and then the g i r l started chasing him / and then he ran behind a tree / and she didn't notice him / and she ran right by / and he got away from the tree and then went around to th i s big f i e l d with no trees / but there were trees a l l around i t / and he was running down there / and then he tripped / then he came back to i t / and then he started digging i t out with a s t i c k / and then his s i s t e r was c a l l i n g him again i n another part / and his s i s t e r found him / then he told her to come over there / and he dug a bomb out of the ground / and he started unweeling t h i s cap off i t / and he took the propeller off / then smoke came out of i t / then they ran away from the bomb / and then the boy dropped his knapsack there / and he remembered i t when they were climbing over the fence / and then he said they have to go back / then she says: "oh no!" / she didn't want him to do i t / so they started r i d i n g t h e i r bikes back home / and they were racing / then they went along / and then his s i s t e r was behind / and he was up here [uses hands on table to demon 205 strate] / then his s i s t e r got ahead / and then he stopped and then turned round and went back the other way to get his knapsack and the bomb / he wanted to show i t to his friends / so he went back, climbed over the fence / and his s i s t e r kept going the other way and went to get the police / and the boy climbed over the fence / he went down to where his knapsack was / he stood there for awhile while his s i s t e r was getting the policeman / and she drove away in somebody's car / and they found a policeman / then she went over and told the policeman / and while she was doing that, the boy was s t i l l standing there / then he walked up caref u l l y / and he got his knapsack / and he ran back to where he was standing / then he went back to the bomb and took a l l his stuff out of the knapsack and then put the bomb i n / and then he started walking away / and then that g i r l had the policeman phone these people / they were part of the army / and they drove t h i s truck down / and they picked up the g i r l / and she showed where he was / and they came to the place / and the boy started running off down into the woods / and they came / they found where the bomb was / but then she didn't know which way he would have gone / but then the guy found his coat off over here [uses table] / and then they followed off that way / and that's the way he went / and they kept on follow-ing him / then they found him / and they were back i n the f i e l d again / and the boy couldn't take off his knapsack / and the bomb was there / and i t could have exploded / then the g i r l and a bomb squad guys ran way off into the f i e l d and jumped down into the grass / and that other guy kept standing there and told him to take i t off / but he couldn't / but then he ran up to the boy and helped him get i t off / and then he got i t off / they both ran over to where the boy's s i s t e r and the other squad guy was / and they took a jump / then the bomb blew up / and that was the end of the movie 206 APPENDIX 3 Mean Proportions of Cohesion Categories for LI and NL Children on TI, T2, and T3 TI (picture-based) LP NL T2 (retell,no pictures) LI NL T3 (video recall) LI NL Reference M .410 SD .025 range .3T7-.446 Ellipsis/Substitution M .004 SD .004 range 0 -.008 Conjunction M SD range Lexical M SD range Incomplete M SD range .220 .039 .167-.261 .276 .034 .245-.322 .088 .031 •038-.135 .429 .032 .389-.479 .002 .002 0 -.005 .224 .052 .154-.297 .325 .047 .243-.371 .020 .006 .013-.027 .406 .034 .368-.449 .002 .004 0 -.011 .256 .032 •218-.315 .258 .054 .180-.316 .079 .019 .056-.101 .413 .032 .386-.472 .002 .003 0 -.006 .219 .035 .157-.259 .337 .028 .304-.383 .029 .014 .011-.047 .397 .032 .354-.435 .008 .008 0 -.020 .250 .023 .219-.287 .237 .041 .198-.303 .108 .034 .073-.172 .417 .025 .392-.461 .013 .006 .006-.019 .254 .012 .237-.266 .281 .019 .260-.308 .036 .013 .020-.056 *N= 6 for each group TI Mann-Whitney U Values T2 T3 Reference Conjunction Lexical Incomplete 11.5 17.0 9.0 0.0** 16.0 8.0 1.0** 0.0** 13.0 14.0 7.0 0.0** */X.05 **p(.01 207 APPENDIX 4 Conjunctive Items Used by LI and NL Children on Tl, T2, and T3 TASK 1 (Picture-based) LI Additive Nonadditive temporal and (6)» tfien (5),and tfien (b),tthile (2), when (1),after (1) adversative but (2), but then (1) causal so (4),so then (l),cuz (1) oontinuative now (3), still (3) ,and now (1), even (1), too (1), finally (1) TASK 2 (Retell, no pictures) LI Additive Nonadditive Temporal and (6) then (5),and then (A),while (1), when (1), after (1), 'til (1), soon (l),aoce (1) Adversative but (2) ,but then (1) Causal so (4), and so (1) Continuative srilZ (3),again {2),now (1), and near (1), and finally (1), anyway (1) TASK 3 (Video recall) LI Additive Nonadditive Temporal and (6) and then (6), then (5), when (2) Adversative but first (2) ,but then (1) ,but (1) Causal cuz (4),so {A),and so (2) Continuative again (2), and finally (1), too (1) well (1) NL and (6) tfien (6),and then (6),wien (3),as (1) while (1), after that (1), ' t i l (1) but (6),but then (1),instead of (1) so (5),2>ecause/cuz (4),and so (2), so that (l),as i f (1) still (4), asain (3), finally (2), now (1), well (1), eventually (1), anymore (1) NL and (6) then (5),and then (5),when (3),as (1) after (1), 'til (1), this time (1) Aut (4) so (5),cuz (3),and so (l),so tfien (1), if (1) s t i l l (5), and finally (2), anymore (2), finally (1), again (1), everafter (1), yet (1) NL and (6) and tfien (6), tfien (4), when (4), while (2), 'til (1),after (1),first (1) as soon as {l),by that time (1), meanwhile (1) fiut (6),2»t tfien (2),but only (1) so (5),cuz (4),and so (3),so tfien (1) i f (1) still (3), well (3), ffrw77y (2), again (2),anymore (1),anyways (1), eventually (1), either (1) aThe number of different given in parentheses. speakers (for a given group) who used the preceding conjunctive item is 208 APPENDIX 5 Phoric Nominal Groups involving Bridging as Used by LI and NL Children on TI, T2, and T3 TASK 1 (Picture-based) LI: the beach (from shore) a, the water (boat), the back end (boat), the rock (mountain), the trail (mountain) NL: the beach (sea), the water (ship), the rocks (mountain), the top (mountain), the other side (mountain or bridge), the anchor (boat), the bow (ship), the ship (cruise), the ocean (cruise or sailors), the captain (sailors), the jungle (voyageur), the swing (swinging), the race (turtles with cards on them with numbers), the turtles (turtle race), riie rail buildings (New York) TASK 2 (Retell, without pictures) LI: the water (boat car ocean), the captain (ship), the top (mountain), the men (ship), the beach (shore), the land (island) NL: the water (boat or ocean), the captain (ship), the top (mountain), the ocean (ship or cruise boat), the sea (ship), the river (ship), the people (ship), the motor (boat), the back (boat), rite ship (sailing along the ocean), the end (trail or board) TASK 3 (Video recall) LI: the propeller (bomb), the general (army men) NL: fiie propeller (bomb), the trees (fishing), the bushes (fishing), the mods (trees), the grass (field), the sign (a spot where it said...) aThe antecedent serving as the basis for the inference is given in parentheses. 209 APPENDIX 6 : CHECKLIST OF COHESIVE DEVICES Part I V Taking into Account Linguistic Context and Listener (Cohesion)— Form/Use A . D i d the child use reference? If so: 1. Was the original referent easily retrievable? 2. Was the referent usually retrieved: a. From the context (exophoric)? b. From prior text (anaphoric)? c. By shared culture? d. By bridging? B. D i d the child use ellipsis? If so: 1. Was appropriate information assumed and omitted? 2. Could presupposed information be easily retrieved? C. D i d the child use conjunction? If so: 1. Were l inked clauses semantically related? 2. Were appropriate conjunctions used to express the relations? 3. Were conjunctions other than and and and then used? D. If the child used lexical cohesion or reference, 1. Were many different referents tied or thematized with these devices? 2. Were they tied frequently throughout the narrative? 3. Was there ever more than a one- or two-clause distance be-tween original mention and remention? 4. Were rementioned referents members of different grammati-cal categories than the original mention and from each other? F_ D i d the chi ld use Parallel Structures? If so: 1. Was the lexical content and structure of the entire clause the same (global parallelism)? 2. Was the lexical content and structure the same for verb phrase only (lexical parallelism)? 3. Was verb phrase structure similar, but lexical content differ-ent (structural parallelism)? Comments on Use of Cohesive Ties: (Note that types of lexical cohesion are not included above because there is little information on developmental changes in its use. You may want to comment on whether cohesion was evident primarily in repetition of the same word or in use of related words. Various types of reference were not included for the same reason, but you might want to describe whether most were pronouns, demonstratives, or comparatives.) Additional Comments Not Directly Related to Coals—(Include frequency of su-bordinating clauses used throughout, problems with tense markings or shifts, etc.) N o t e : R e p r i n t e r d , by p e r m i s s i o n o f t h e a u t h o r , f r o m Language Disorders and Language Development by M . L a h e y (New Y o r k : M a c m i l l a n P u b l i s h i n g C o . , 1 9 8 8 , f 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0097565/manifest

Comment

Related Items