LANGUAGE BASES ... DISCOURSE BASES
Pragmatics & Beyond New Series Editors: Jacob L. Mey (Odense University) Herman Pa...
31 downloads
921 Views
11MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
LANGUAGE BASES ... DISCOURSE BASES
Pragmatics & Beyond New Series Editors: Jacob L. Mey (Odense University) Herman Parret (Belgian National Science Foundation, Universities of Louvain and Antwerp) Jef Verschueren (Belgian National Science Foundation, University of Antwerp) Edito rial A ddress : Linguistics (GER) University of Antwerp (UIA) Universiteitsplein 1 B-2610 Wilrijk Belgium Editorial Board: Norbert Dittmar {Free University of Berlin) Bruce Fraser {Boston University) John Heritage {University of California at Los Angeles) David Holdcroft {University of Leeds) Catherine Kerbrat-Orecchioni {University of Lyon 2) Beatriz Lavandera {University of Buenos Aires) Marina Sbisà {University of Trieste) Emanuel A. Schegloff {University of California at Los Angeles) Paul O. Takahara {Kobe University) Sandra Thompson {University of California at Santa Barbara) Daniel Vanderveken {University of Quebec at Trois-Rivières) Teun A. van Dijk {University of Amsterdam)
17 Gilberte Piéraut-Le Bonniec and Marlene Dolitsky (eds) Language Bases ... Discourse Bases Some aspects of contemporary French-language psycholinguistics research
LANGUAGE BASES ... DISCOURSE BASES SOME ASPECTS OF CONTEMPORARY FRENCH-LANGUAGE PSYCHOLINGUISTICS RESEARCH
edited by
GILBERTE PIÉRAUT-LE BONNIEC and MARLENE DOLITSKY
JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA 1991
Acknowledgement The editors would like to extend their heartfelt thanks to Colgate-Palmolive France for their help in preparing the manuscript.
Library of Congress Cataloging-in-Publication Data Language bases...discourse bases : some aspects of contemporary French-language psycholinguistics research / edited by Gilberte Piéraut-Le-Bonniec and Marlene Dolitsky. p. cm. -- (Pragmatics & beyond, ISSN 0922-842X ; new ser. 17) Includes bibliographical references and index. 1. Language acquisition. 2. Discourse analysis. I. Piéraut-Le-Bonniec, Gilberte. IL Dolitsky, Marlene. III. Series. P118.L388 1991 401'.41-dc20 91-6685 ISBN 90 272 5027 8 (Eur.)/l-55619-283-5 (US) (alk. paper) CIP © Copyright 1991 - John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher.
Contents Introduction
1
G. Piéraut-Le Bonniec Speech Bases Phylogeny and Ontogeny of Languages F. Bresson Prenatal Familiarization J.P. Lecanuet, C. Granier-Deferre and M.-C. Busnel
11 31
Initial Equipment for Speech Perception J. Bertoncini
45
Target-Language Influences on Prespeech B. de Boysson-Bardies
57
Prosodic Cues in Very Young Children's Speech
73
N. Bacri Basic Discourse Capacities Cohesion: Syntactic Organization Leading to Discourse Verbs as Sentence Organizers G. Piéraut-Le Bonniec and J.P. Pille
93
Pronoun Assignment in the Processing of Locally Ambiguous Sentences / . Seguy and M. Kail
109
Conjunctions: Developmental Issues M. Kail and J. Weissenborn
125
CONTENTS
Children's Production of Textual Organizers J.-P. Bronckart and B. Schneuwly The Development of Discourse Cohesion: Some Functional and Cross-Linguistic Issues M, Hickmann
143
157
Coherence: Language as it Underlies and Organizes Knowledge An Opaque Text as a Mobilizer of Knowledge M. Dolitsky and G. Piéraut-Le Bonniec
189
The Development and Role of Narrative Schema Storytelling E. Esperet
209
Stories A Psycholinguistic and Ontogenetic Approach to the Acquisition of Narrative Abilities M. Fayol
229
The Development of Argumentative Discourse G. Piéraut-Le Bonniec and M. Valette
245
References
269
Index
Introduction G. Piéraut-Le Bonniec
For any collection of articles, the question a potential reader might ask is, "Why have these particular papers been bound together?" The interest of gathering the ar ticles to be found in this book is three-fold: 1. The contributions that make up this volume would seem to bring out clearly how cognition and language are, per force, interrelated. On the one hand, cognitive functions play a fundamental role in the development of an individual's linguistic system. On the other hand, the ability to form concepts and to deal with logical relations is, for a great part, dependent on language, once its bases have been acquired. However, the problem that the languagecognition relation poses is always in the presupposed wings of any study either on language or on cognition. Because the relation between language and cognition is so involuted, it cannot be studied head-on, but must be at tacked by the study of topics that circumscribe it. In this work the articles that have been brought together do not treat this relation explicitly but im plicitly. The first part offers data on the initial equipment the child possesses in order to apprehend the world of language. In the second part, the structure of the world, and the relations between its elements as set up by the child, are assumed to pre-exist his/her attempt to assert them. The collection of these particular articles into a single loosely-woven volume is an attempt to present the language-cognition relation by skimming over it from a number of points of view. 2. Given this theoretical perspective it becomes necessary to consider the development of language over a large age span. For a long time, researchers in psycholinguistics concentrated their efforts on the acquisition of syntax, as lead by Bellugi (1967), Brown (1973) and Sinclair (1971), or on semantic development and the comprehension of word meaning, a perspective first ex plored by Rosch (1973) and E. Clark (1973). This work was mainly carried out on children between age two and age seven. At that time, researchers did
2
G. PIÉRAUT-LE BONNIEC
not have at their disposition the necessary sophisticated techniques to allow them to study the very beginnings of language, And at the other end of the scale, they were not very interested in text discourse. Struck by the ease with which children learn their native language (a point often brought out and em phasized by Chomsky), they accepted all too readily that children arrived at mastery of their language by acquiring rules of transfomation to be applied to an SVO structure. Most often, the studies concentrated on sentences in context-free situations and textual discourse was taken for granted, or at best considered to be a result of schooling. Furthermore, the use of language by school-aged children did not interest psycholinguists as they felt that the metalinguistic approach to language as taught by the schools distorted children's spontaneous use of language. Nevertheless, it should be noted here that the experimental conditions in which subjects were placed created a situation which elicited from the child a metalinguistic approach to language rather than a spontaneous one. But research tendencies changed in the eighties, especially among those working in Europe in the French-language community. First of all, more and more sophisticated recording techniques were developed which allowed for the study of perceptual abilities in new-bom babies and even in the fetus in utero. Such studies attempted to take the study of language as far back as possible to its very source. It would seem that if infonnation is being processed by the fetus in utero, research should be carried out on those aspects of language that may be part of the cognitive abilities possessed by a child at birth and even before. Besides recording techniques, there also exist more sophisticated techniques allowing for the analysis of new-borns' bab bling. Thus, it has become possible to study how babies select sounds for their own productions from the language of their environment. Moreover, at the other end of the language learning spectrum, researchers have become interested in how children go beyond "simply" constructing sentences to become users and producers of textual discourse. Once children have acquired basic language skills, language and cognitive processes (con ceptual skills and logical thinking) develop dialectic ally. Therefore, these authors cannot agree with those who believe that language learning stops at around age ten. It is precisely at this age, when children have acquired the basic rules of speech, that they attack the problem of discourse as a text, i.e. a stretch of written or spoken language (Crystal 1985). This, thus, explains
INTRODUCTION
3
the importance given to text linguistics as a branch of study, particularly in Europe. 3. So it is happening that in the French-language community in Europe a num ber of researchers are now focusing on the very beginnings of language, while others are more interested in the acquisition of basic textual capacities. Unfortunately, the results of their research does not seem to be sufficiently well known. Furthermore, the vectors of research followed and the points of view taken by these researchers often differ from what is published in the English-language press. Thus, one of the aims of this book is to offer to the English reading public some of the most important work that is being done in linguistics by French-writing researchers.
Language Bases What is most basic to language is the audio-phonological equipment that allows for the production and perception of speech sounds. Scientific progress has taken us well beyond Herodotes' barbaric experiment of isolating human babies to learn what language their first words were in. We now know they will speak no language at all if no one speaks to them. Children appropriate language from their environ ment, and most generally it is the language their mother uses with them. B. de Boysson-Bardies' work shows us that appropriation takes place very early. Her study of children from 6 to 10 months of age shows that their babbling contains certain general features which are characteristic of the target language and which can even be identified by naïve adults. But the appropriation of a native language is obviously dependent on the percep tual-motor equipment children possess at birth, and this equipment is, as is shown by F. Bresson, the end result of a long evolutionary process in the species. In the first chapter of this book Bresson presents a number of stimulating hypotheses con cerning the origin of the production and perception of language which are based on the most recent work done in the field of the phylogenesis and ontogenesis of lan guage. Many studies have shown that neonates have no problem discriminating lan guage sounds categorically and that their transition points are the same as adults'. J. Bertoncini offers a quick overview of the work that has been carried out in this field during the last fifteen years. She then continues on to present some of the
4
G. PTERAUT-LE B0NN1EC
work she has carried out with J. Mehler and P. Jusczyk in their search for more and more precise knowledge concerning neonates' phonetic categorization skills which they have worked on through the comparison of different vowels and consonants. But there is still the question concerning whether children at birth do have some experience in the perception of speech sounds or not. There are studies being car ried out now on the acoustic environment within the uterus and on the fetus' reac tions to acoustic stimulation which seem to show that the acquisition of auditory skills begins before birth. J.-P. Lecanuet, C. Granier-Deferre and M.-C. Busnel review the work that has been done on new-borns' recognition of their mother's voice, and more generally, on the role that might be played by the linguistic en vironment of the fetus during the last months of pregnancy (this would include both the mother's voice and the voices of any others in contact with her) on the development of the neonate's perceptual skills. Until now most researchers have considered phonemic and syllabic perception to be the most important aspect in the perceptual processing of the speech signal. However, very recent studies have shown that much importance should also be lent to children's ability to identify various parameters, such as voice quality and in tonation. N. Bacri shows here that very early on, children are capable of controlling their voice sufficiently to imitate some intonations, rhythms and melodic contours that occur in the linguistic environment. The result is that at an age when a child's language is still made up of unintelligible vocal signals, i.e. mainly babbling, adults "understand" what the child says, or they believe they do, by using intonation clues.
Basic Textual Capacities Already in 1924 Piaget showed that, contrary to what most adults believe, children do not always understand what is said to them. It is important to under stand how children's discursive competence develops, how they learn to understand and produce a complex text. In fact, lending cohesion and coherence to a textual production is not as easy as it seems. The ability to create or understand sentences one by one does not imply that one is able to construct or comprehend discourse. The second part of this volume is devoted to children's textual capacities, which will be taken from two points of view: the cohesion conditions of texts, i.e. the syntactic organization leading to discourse, and coherence conditions, i.e. the or ganization of knowledge. It must be remembered here that communication is the result of an effort made by both speaker and listener working in cooperation. Each
INTRODUCTION
5
one plays an active role in the creation of the meaning a sentence will take on in a given text. Both speaker and allocutor work to create cohesion in discourse (cf. Seguy and Kail) and make it coherent (cf. Dolitsky and Piéraut-Le Bonniec). On the other hand, the other articles in this section deal with the way in which the speaker or writer gives a text both cohesion and coherence.
Cohesion Conditions (Syntactic Organization of Discourse) At about one year old, a child's first words appear. At about 18 months, children's first attempts at syntactic construction can be observed. The first rule that young children are able to identify is that of the pertinence of word order. Most of the research into the emergence of syntax at age 2, calls attention to the importance of "pivot words", the category of words around which the others are or ganized. However, the function of these pivot words is comparable to the function of verbs in the sentences of adults. The study by G. Piéraut-Le Bonniec and J.-P. Pille shows that early on in the acquisition process, children do show an intuitive feeling for the verb's role as a sentence organizer. This ability is related to lan guage functioning and can thus be qualified as "epilinguistic". Young children, however, have difficulty considering verbs as words; for them, words only desig nate things. The ability to consider verbs as words that have specific characteristics that can be spoken about, i.e. metalinguistic competence, is only acquired through school learning (reading, writing, grammar). By 3 years old, children have tacit knowledge of a great number of rules that determine the grammaticality of sentences in their language. However, they have certain difficulties concerning the processing of complex sentences, both as to com prehension and production. These difficulties stem from the fact that while, theoretically, the linear organization of language is infinite, one's psychological ability to process it, is not. Thus, for example, the comprehension of complex sen tences often depends on the subject's ability to set up hypotheses and to "take risks" in developing an integrated meaning for a sentence. This is what is shown in J. Seguy and M. Kail's study concerning the processing of sentences containing ambiguous coreferential anaphoric pronouns. Subjects immediately assign reference to the pronoun in accordance with the interpretation bias induced by the semantic characteristics of the verb in the main clause of the sentence. It is not until after they realize that there is an incongruity that they reread it. The processing of complex sentences also includes the identification of a specific type of word, i.e. connectives, found in the speech chain. These words
6
G. PIÉRAUT-LE BONNIEC
have no referent other than the organizing function that is set up through their use and is attributed to them. M. Kail and J. Weissenborn have carried out a cross-lin guistic comparative study into the order of emergence of different semantic rela tions and their corresponding conjunctions in the spontaneous speech of young children. Discourse cohesion is an extremely important aspect of communication and its development is closely linked to cognitive development. Discourse cohesion comes from the use of linguistic elements that link the utterances that together make up a given discourse, and in so doing, they contextualize language. This topic is studied from two different points of view in two different papers in this volume. The first is J.-P. Bronckart and B. Schneuwly's and concerns how children between the ages of 9 and 12 use linguistic elements as textual organizers and what psychological operations underlie their production in relation to the conditions in which the texts are produced (situational discourse, theoretical discourse and conversational narra tive). The second paper is a cross-linguistic study by M. Hickmann concerning the ac quisition of a referential system. It shows that referential continuity is acquired later on in the learning process, especially when it is to be used in a context where the children must tell someone about what somebody else has said, i.e. indirect speech, where there are two levels of enunciation. Both these studies delve into both the comprehension and production of complex sentences and lead toward research that goes beyond sentential psycholinguistics into the experimental study of discourse.
Coherence Conditions (Organization of Knowledge in Discourse) The last five chapters of this volume treat the topic of textual psycholinguistics. The emphasis will thus be put on the functional aspects of language in that they are the most sophisticated way in which knowledge of the world is come by and its representations are set up. By putting subjects in the highly unusual situation where they are expected to assign meaning to an opaque text, M. Dolitsky and G. PiérautLe Bonniec present the subjects' discovery (in spite of their unease in the situation) of the interpretation process any reader or listener must go through, whatever the conditions. They must call forth what knowledge they have in order to set up hypotheses and to enrich meaning. Such a situation brings to light those strategies that are used to construct the meaning of a text under normal circumstances.
INTRODUCTION
7
Very roughly, it can be said that there are two types of discursive functions: nar rative and argumentative. Narrative, or descriptive, discourse is used to tell, while argumentative discourse is used to construct reasoning, with the view to persuade. Obviously an essentially narrative discourse might include explanations or justifica tion, and argumentative discourse might include some narration. What is important is that while in both cases the discourse must be cohesive, the discursive elements which ensure this cohesion are not the same. Two papers in this volume are about studies concerning children's acquisition of the narrative form. E. Esperet bases his approach to the subject on the idea that a narrative cannot be produced little by little, but must be planned. In other words, speaker/writers must plan what they will say and master the narrative schema. M. Fayol's study concerns the development of narrative skills in children. As in dicators of textual organization, he uses the child's choice of subject of the ut terance, verb forms, connectives and punctuation. Finally, the development of argumentative discourse in adolescence is studied by G. Piéraut-Le Bonniec and M. Valette. After a review of the specificity of ar gumentative discourse, they present the results of their experimental research which show that argumentative discourse presents real difficulties, even for adolescents who use their native language everyday and who can solve problems in various situations demanding a good deal of reasoning skill. However, they are not neces sarily capable of going beyond narration when setting up argumentative discourse. This is a problem that should be looked into by teachers and any others interested in pedagogy. The authors who have contributed to this volume are highly involved in the type of research they present here, offering both the difficulties inherent in work on their subjects and the results that have so far been obtained. Although they are all spe cialized in their fields, the attempt here has been to fit their work into a much broader frame so that it could be applied to the work of all those who might be in terested in a pragmatic view of language: linguists, teachers, psychologists and speech therapists.
Speech Bases
The forms of nursery rhymes that are developed and adopted by linguistic com munities clearly reflect the salient traits and tendencies found in children's lan guage development and show a certain universal homogeneity as well. In particular, the differences in phonemic variation found in parents' terms of endear ment have been shown to be "very limited". The principles underlying successive stages in children's acquisition of language have shed some light on and permitted the interpretation of "interlanguage parallels" found all over the world in the struc ture of such terms ... ... From working together and through their joint efforts, linguists, anthropologists and specialists in the development of mental and behavioral psychology have opened up a large area of research, which is certainly a very fruit ful one. R. Jakobson, 1941 Kindersprach, Aphasie und allgemeine Lautgesetze (our English translation from the French translation, 1969: Langage Enfantin et Aphasie, Paris: Editions de Minuit)
Phylogeny and Ontogeny of Languages F. Bresson Centre d'Études des Processus Cognitifs et du Langage École des Hauts Études en Sciences Sociales Centre National de Recherche Scientifique, Paris
In the positivist atmosphere of the nineteenth century, papers on the origins of language had a good chance of being rejected just as studies on perpetual motion or the quadrature of the circle were refused by the French Académie des Sciences, Today, the study of the phylogeny of language no longer seems absurd as it no longer needs to be based on arbitrary speculation. Furthermore, our attitude has been shaped by the acceptance of Homo sapien sapiens as a primate with living cousins, the anthropoid apes. However, investigation into the phylogeny of language is hampered by the fact that humans are the only living animals with a spoken language that is more than a mere extension of the vocal communication systems found in other species. No other known communication system includes both phonetic and syntactic com binatory systems, a referential organization, anaphoric linking markers and an in trinsically unlimited capacity to create statements. The unique characteristics of human language are such that it cannot be directly derived homologically from the communication systems found in other primates. Language, like all behavior, leaves no fossils, and without direct evidence of its origin, missing links must be inferred from indirect evidence, namely the conditions necessary for language development today. It is because these conditions were totally unknown that as sumptions concerning the origin of language remained mere speculations in the past. Furthermore, the unique quality of human speech led to refusal, on religious or ideological grounds, to approach the problem from the evolutionary angle, taking humans as a primate among primates and language as one behavior among others with roots further back in the past than the period covered by its known his tory. Worse yet, the history of language was limited to written records, i.e., to an extremely short period in the history of civilization and, of course, of hominization.
12
F. BRESSON
Even now, discussion of the phylogeny of language is of the realm of conjec ture, as it is based on factual convergences that can rarely be used to generate verifiable hypotheses. First, these conjectures concern those conditions in the or ganization of the central nervous system that are prerequisite to the production of language and the morphological consequences of this organization. However, these in turn can only be linked to the macroscopic structure of the brain. Other assump tions concern the conditions for the production and perception of the sounds of lan guage, the production of references and for predicate and syntax organization. If the development of language is an evolutionary process, it should be possible to derive an origin from each of its characteristics by homology from our knowledge of humankind and its closest living relatives, the apes. Other primates can also be used as a source of information on neural mechanisms. Ontogeny is another impor tant source of inference for phylogeny, not because it is felt that recapitulation is involved, but because the linguistic capacities apparent in the first days of life and the ability to apply linguistic rules without previous trial and error experimentation enable us to form hypotheses about the specific conditions for language production and perception. Data from these different sources cannot be fitted together to form a coherent whole before they are considered separately. Language appears as the result of the convergent evolution of different systems or as the product of changes in intrinsi cally related subsystems as they appear in normal language production today. Con sequently, we will discuss in turn morphology, handedness and cerebral lateralization, speech production and perception, referential behaviors and predicate and anaphoric organization.
Cerebral Morphology Hominization involved two transformations in cerebral morphology which can be seen today in fossil skulls and brain endocasts. The first was an increase in the volume of the skull, implying a concomitant increase in brain size, particularly ef fecting the frontal region (Falk 1983). The second concerns differences between the right and left endocasts of hemispheres. These differences are important as they are apparently related to language (Holloway and de Lacoste 1982). Geschwind found that the planum temporale is larger in the left than in the right hemisphere in humans (Geschwind and Levitsky 1968). The difference is already apparent in the foetus (Galaburda et al. 1978; Galaburda and Pandya 1982; Galaburda et al 1987). This enlargement, in a region which anatomical and clinical data have shown to be
PHYLOGENY AND ONTOGENY OF LANGUAGES
13
involved in language control, suggest that the correlation must be of functional sig nificance. The difference is not unique to modem humans. LeMay and Geschwind (1975) studied other primate brains, measuring the height of the posterior extremity of the sylvian sulcus, which was correlated with the size of the planum temporale. They found virtually no differences between the left and right hemisphere in the 69 right and left monkey brains examined (Cebus and Ateles) whereas a difference was present in apes (9 chimpanzees, 7 gorillas and 12 orang-utangs) although not to the extent observed in humans. Asymmetry was greatest in the orang-utang. The authors considered that in conjunction with previous observations, these results suggest that the study of this species could lead to important information on the evolution of language and manual dexterity (LeMay 1985). The finding is par ticularly intriguing as immunologically the orang-utang is the anthropoid ape fur thest removed from humans while the chimpanzee is the closest relative (Miyamoto et al. 1987). Following this up, Yeni-Komshian and Benson (1976) compared the brains of 25 rhesus monkeys, 25 chimpanzees and 25 humans. The left/right ratio was 1.15 in humans, 1.05 in apes and about 1 in the monkey. After examining all the published data on left-right hemisphere comparison, Witelson (1977) found that they generally confirm that asymmetry exists and in creases in anthropoids as we go up the evolutionary ladder. Although asymmetry may be related to language, since it is present in apes which do not speak, the dif ference is not a sufficient condition for language development, at least, not until asymmetry has reached a critical level. The same author considers that the data cur rently available on humans are not sufficient to conclude that the morphological differences are related to functional differences (Witelson 1977, 1982). The data may however be linked to differences in hormonal conditions affecting cerebral on togeny in the foetus and the differences in the rates of development of the two hemispheres (Galaburda et al. 1987; Geschwind and Behan 1982; Sherman and Galaburda 1982, 1984). The conditions responsible for developmental asymmetry linked to the hormonal regulation of ontogeny may produce the conditions needed for the development of a new behavior, language. Later development of the left hemisphere may explain its dominant role in language control.
Handedness and Language Right-handedness has been linked to language since the mid-nineteenth century when Broca (1861, 1865) showed that only lesions of the left hemisphere might cause aphasia or paralysis of the right arm and leg. It is now known that the left
14
F. BRESSON
hemisphere controls language in 96% of right-handed individuals. The same is true of 70% of left-handed people, while in 15% of these individuals language is con trolled by the right hemisphere and in the remaining 15% control is bilateral. Cur rent estimates suggest that at most 11% of the population is left-handed; however, this proportion varies because population is not typically divided by handedness. In fact, the choice of hand in some behaviors is socially determined. For instance, in our society, the right hand is always used to shake hands, even if more tolerance is shown today when it comes to writing or holding a knife. Dominance of the right hand is widespread and found in all cultures. Indeed, traces of rotation shown on paleolithic tools imply that it is a characteristic of Homo sapiens, although at least one implement has been found that was used in a counter-clockwise fashion. Nevertheless, handedness in one- or two-handed tasks is a matter of degree. In a large part of the population one hand is chosen consistently in a series of tasks. Distribution is bimodal with a first peak representing consistent right-handers and a second for left-handers; however, there is a whole series of intermediaries between the two peaks (Annett 1978, 1985). Some individuals are therefore less clearly lateralized and would consequently be more dependent on social constraints for at least some tasks. Handedness is subject to environmental constraints such as social rules and the fact that objects are designed for right-handed people. This explains why a choice of hand becomes more consistent with age, and suggests a developmental factor. However studies in infants disprove the belief. 80% of neonates turn their heads to the right when placed in a supine position whereas only 9% favour the left (Turkewitz and Birch 1971). This affects the mother's choice of side on which she carries the child (Ginsburg et al. 1979). Similarly, positioning of the foot is rightlateralized in the first days after birth and stabilized after one week (Melekian 1981; Peters and Petrie 1979). The 5-day-old infant exhibits left-hand superiority in open-loop visuo-manual reaching for an object (de Schonen and Bresson 1984). Stable bimanual coordination appears at 26 weeks. When a small object is presented on a tray, the left hand holds the edge of the tray and the right slides towards the object (Bresson et al. 1911; Bresson and de Schonen 1977; de Schonen 1977; de Schonen and Bresson 1984). All the evidence suggests that handedness and the complementary roles of the two hands are independent of environmental factors and related to language dominance (MacNeilage 1985). Bimanual coordina tion appears to be fundamental. The organization which will later be used for be haviors such as "peeling a potato" is already apparent in embryo. In this behavior the left hand assumes a localizing and propping function and represents the area to which the action of the right hand is applied. For this bimanual coordination the
PHYLOOENY AND ONTOGENY OF LANGUAGES
15
left hemisphere has a control function over both hands. This function is consistent with neurological observations of left hemisphere lesions which can affect both hands (Liepmann 1908; Wyke 1971; Kimura 1977). Many studies in monkeys (Lehman 1978, 1980; Vauclair and Fagot 1987; War ren 1977) and apes (Brésard and Bresson 1983) have found clear and relatively stable lateralization, independent of species. The ratio of right- to left-handedness was identical in all population studies. Handedness seemed to be subject to more situational and environmental constraints and age than other behaviors (Chorazina 1976; McGonigle and Flook 1978; Brésard and Bresson 1983). However, speciesrelated lateralization exists for handedness and other behaviors such as visuo-motor control (Brésard and Bresson 1987; MacNeilage et al. 1987). The right-handedness precision grip appears to be consistently lateralized for monkeys (Beck and Barton 1972; Preilowski 1979) and gorillas (Preilowski and Leder 1984) with left-handed ness for chimpanzees (Brésard and Bresson 1983, 1987). Species-related lateraliza tion seems to result from the developmental asymmetry of both hemispheres (Glick and Ross 1981; Sherman and Galaburda 1982; Geschwind and Galaburda 1985, 1987). As is the case with morphology, data on behavioral laterality do not reveal any clear filiation between apes and humans. Behavioral laterality and language laterality in humans has been established beyond doubt. The link between them does not seem to be fortuitous; it might stem from a shared relationship with the organization of the nervous system, consecutive to the hormonal regulation of on togeny (Geschwind and Galaburda 1987). It is important to determine whether similar relationships exist for left hemisphere dominance in other species (Brésard and Bresson 1987; MacNeilage et al. 1987). It would be reasonable to assume that, as with morphological differences, asymmetry might be related to different rates of development of the two hemispheres (de Schonen 1989). Later development of the left hemisphere may have effected the complex processing of several functions, such as bimanual coordination or language. However, a causal relationship between manual dominance and language would still have to be explained. Hominization has often been represented as the liberation of the position of the head and activity of the hands by assumption of an upright posture and bipedalism (Langdon 1985). Chipped tools, i.e. not mere choppers, were not used before the time that bimanual coordination was fully functional (Levalosian and Mousterian industries), i.e. where the left hand holds and presents the object to permit the work of the right, as in potato peeling. Such fine coordination seems to be rare or impos sible in apes. Wright (1972) tried to teach an orang-utang to make and use a flint flake. The ape was able to use the flake, but had difficulty learning to make it as it
16
F. BRESSON
could not hold the stone correctly while striking. Finally, a clamp was used to maintain the stone in place. However, there are many confirmed observations of wild chimpanzees breaking nuts (Boesch and Boesch 1981). The nut is placed on a small depression of a branch or a large stone. The animal selects a suitable stone to serve as a hammer and sets it aside for later use. This behavior, like screwing and unscrewing, shows a degree of coordination between the hands. It is therefore tempting to suppose that there is an evolutionary continuity starting with using a hand to maintain an object in place on an anvil followed by the ability to present the desired face of an object to the skilled and active hand for continual precise ad justment. Behavior of the skilled hand would also evolve, from simple striking (beating a nut) to the delicate craftsmanship of upper paleolithic flint scrapes. The increasing complexity of coordination during a 2-million-year period of evolution which is obviously correlated with the development of the cortex, does not imply a discontinuity that could be used to infer a relationship to language. Nevertheless, coordination may be linked to language in two mutually com patible ways, although these are both conjectural. The first theory, which is older and more frequently supported proposes that language was necessary for the learn ing and propagation of elaborate methods of tool production and for the collective activities, such as hunting large prey, that the new tools permitted. However, there is no evidence to suggest that language is the only way of transmitting and coor dinating tool production, an activity which remained stable and rudimentary for hundreds of thousands of years. Acquisition and transmission of behavior by imita tion of a visual model has been observed in Japanese macaques (Kawai 1965) and examples of acquisition of complex skills without the use of language in humans are also known (Bril 1986). Human's stone tools are considered to be different from the "primary" tools used by chimpanzees in their natural habitat, e.g. straws broken with the hands or teeth or stones that are picked up (Watanabe 1977, cited by Kitahara-Frisch 1978; Boesch and Boesch 1981). It is likely that, by 1.9 to 1.4 million years ago, hominids chipped tools and this implies the use of one stone (the primary tool) to fashion another (the secondary tool) (Toth 1985, 1986). The or ganization represented by the secondary tool, both in its elaborate shape and the purposefulness of its production, implies a cognitive complexity superior to that reflected in the simple food-gathering activities and primary-tool use of chimpan zees, even considering an intermediate process such as occasionally selecting chips for a breaking stone. However, this alone is not sufficient to state that language is a necessary condition of the behaviors, even if it is a sufficient condition. The second interpretation, which stresses the complementary role of the hands or their functional asymmetry, was proposed by Oakley (1961). According to this
PHYLOGENY AND ONTOGENY OF LANGUAGES
17
hypothesis, manual asymmetry is required to manufacture tools and would lead to cortical asymmetry (Frost 1980; Steklis and Hamad 1976). Frost considers one hand as skilled and the other as unskilled. The first would require more space in the cortex for its control, pushing concurrent functions into the other hemisphere (Kimura 1979). For Frost, the greater size of the planum temporale would explain why the space is occupied by the skilled hand which is the right hand. The skilled hand can perform the functions of the other hand (power grip, Connoly and Elliott 1972), but the reverse (precision grip) is not true. However, this argument is apparently incompatible with the early onset of manual complementarity (Bresson et al. 1977; de Schonen 1977; de Schonen and Bresson 1984). Bimanual coordination must have appeared far earlier than lan guage, as it is found in apes. Asymmetry "results from the greater pruning down of one of the sides during late foetal life and infancy, a process that implicates asym metry of developmental neuronal loss" (Galaburda et al. 1987: 867). The required conditions are certainly complex since some split-brain subjects can perform certain tasks requiring bimanual coordination. The duality of function involved in bimanual coordination, especially in its more elaborate forms, may appear to be parallel to the coordination necessary for language, which requires markers and ac tions related to the markers. In acalculia, the loss of algorithms is related to lesions of the left hemisphere and can be distinguished from loss of "addressing", which decides where algorithms should be placed and which is associated with lesions of right hemisphere (Bresson et al. 1972). However, if the analogy were tenable, some kinds of language disorders would be seen in subjects with lesions of the right hemisphere, but this is not the case as language is a left hemisphere function.
Speech Production and Perception Modem humans can produce from 100 to 200 different types of sounds. It is dif ficult to be more precise as there is no general agreement as to what constitutes a type of sound, since there are considerable variations which are perceptible be tween sounds identified as representing the same phoneme. Be that as it may, a human can produce 10 times the number of sounds a chimpanzee can. Humans can develop a modulation system to adjust the sounds they produce to the repertory of any language, and this repertory is almost complete before the age of three. This requires control of stable vocal productions in their successive occurrences (types), and in particular, the regulation of duration. An example is provided by the con trasting pair, voiced and unvoiced, with steady voice onset time. A further example
18
F. BRESSON
is provided by the "sounds of silence" or vocal prolongations with marked boun daries (Liberman 1982). Production must contain invariants which permit the identification of sounds of the same type, independent of voice or speaking rate. Visual perception constants provide an analogy. However speech constants pose very different problems from visual ones since objects have invariant physical dimensions independent of spatiotemporal transformations. Speech constancy implies invariance in the form of a gestural structure despite transformations due to different speakers and circumstan ces. Such relational invariance is known to exist. It has been established that lis teners take speaking rate (speed of articulation and flow) into account when assessing the duration of voice onset time (VOT); the same physical interval is per ceived as /p/ or /b/ depending on the speed of articulation (Crompton 1980; Miller and Grosjean 1981; Repp 1982; Rietveld 1980). Children therefore have to learn to produce the constants used in their language to mark different types of phonemes and boundaries, just as they must learn to recognize them (Chiat 1983; Macken and Barton 1979, 1980). The stable organization of the order of production of a series of movements implies a far more complex arrangement of the nervous system than that which provides for the invariance of the significant forms of vocalizations in other mammals. We will not discuss the system in birds because they belong to a more distant evolutionary branch, however a neural network model of the acquisi tion of songs by birds presents a general model (Dehaene et al. 1987). Perception and production of human speech reveal a capacity for control of the speech organs and hence the parallel development of the relevant areas of the cortex, which is en tirely different from the capacities related to the vocalizations of other mammals. This constitutes a discontinuity between apes and modern humans but reveals noth ing about the evolution of this skill. In a series of studies on the reconstruction of the supralaryngeal vocal tract in hominids, Liebrman (Lieberman 1973, 1975, 1979, 1984, 1985; Lieberman and Crelin 1971; Lieberman et al 1972) found that Neanderthals could not have been able to produce as many sounds as present-day humans. Their vocal system could not have accommodated the vowels /i/, /u/ or /a/. Lieberman considered that production limitations in apes and Neanderthals as well as in modem neonates might be due to peripheral obstacles, i.e. speech organ morphology. If this is so, limitation in cortical organization may also be involved although not necessarily. These results are still being disputed as the industry and symbolic behaviors of Neanderthal man, together with their skull volume, strongly suggest the possession of a language. The limitation in their repertory of contrastive traits probably ex cludes the production of a language with the same type of differentiation as found
PHYLOGENY AND ONTOGENY OF LANGUAGES
19
in languages in use today, but does not exclude some type of spoken language. In contrast, contemporary hominids such as Skuhl V, had a vocal tract that was similar to our own. It is now known that the Homo sapien sapiens "Proto-CroMagnon" were contemporaries of Neanderthals (Stringer and Andrews 1988; Val ladas et al. 1988). The speech insufficiency of Neanderthals possibly explains their disappearance in favor of Homo sapien sapiens. A second type of discontinuity has long been discussed, i.e., the inability to learn to produce sounds that are part of the repertory in response to arbitrary sig nals. The production of vocalizations would, if this were true, be restricted to situa tions which release them, as is the case for the production of emotional facial expressions. However, this is not so. Sutton et al. (1981) were able to teach rhesus monkeys two operant conditioning responses in a similar period of time. One was a classic behavior, pressing a lever; the second was a vocalization from their reper tory ("coo"). Aitken (Aitken 1981; Aitken and Wilson 1979) obtained similar responses in another species of macaque. Apparent stability in the use of the vocal repertory is not, therefore, due to the inability to control vocal emissions in new as sociations. The perception of language sounds presents a characteristic which has long been thought to be specific to language, i.e. their categorical nature. Artificial language sounds can be created to obtain a series of intermediates which are physically (acoustically) equidistant from each other and which constitute a graduated trans formation between, for instance, a syllable that everyone hears as /ba/ and another that everyone hears as /pa/. When subjects are presented with the intermediates and are asked to state whether they hear /ba/ or /pa/, there is an abrupt passage from one response to the other at a stable intermediate stimulus value. This shows that the stimuli are divided into 2 types (Liberman 1982; Liberman et al. 1961; Liberman et al. 1957; Liberman and Mattingly 1985). It has been proven that neonates discriminate categorically between language sounds at the same transition points as adults, even before they can produce speech (Eimas 1971; Bertoncini and Mehler 1981b; Jusczyk 1981; Mehler et al. 1988). It has also been demonstrated in both adults and infants that categorical perception is not unique to language sounds (Jusczyk et al. 1977). Finally, Kuhl has shown that categorical discrimination of artificially constructed linguistic sounds is possible in animals which do not possess language, such as chinchillas or macaques (Kuhl 1979b, 1981; Kuhl and Miller 1975). However there is a difference between the be havior of neonates and of animals in these experiments. The former learn to respond to the discrimination situation very rapidly, but the animals require long conditioning sessions. One reason for this difference is that the discrimination has
20
F. BRESSON
no functional value for the animals. Consequently, the number of trials must be in creased before the difference to be discriminated is perceived as a signal. In con trast, infants detect the sounds without difficulty as early as the first days after birth and even before. These experiments show that acoustic stimuli are processed identically in mam mals and humans at least until the processing level which brings the geniculate bodies into play (Hashimoto 1980). The evolutionary transformation that led to the development of language cannot therefore have been the categorical processing of linguistic sounds, which already exists at lower levels of the phylogenetic scale. Speech production, where there is discontinuity, and speech perception, where there is continuity, are completely different. This suggests that the processing of acoustic stimuli appeared first and shaped the form of production, rather than the reverse. Because these types of sounds were perceived categorically, they were used to produce language; they were not perceived categorically, because they were used in language. In 1957, Liberman proposed a motor theory of speech perception (Liberman et al. 1957, 1967). He considered that only invariant modes of articulation could ex plain categorical perception and the invariance of categorization in different con texts, despite acoustic variations. He suggested that humans, the language producers, had a speech mode of perception that was characteristic to them and that was different from the general mode of auditory perception. Studies with animals have shown that the theory is untenable in its original form. However, the findings do not exclude the possibility that the appearance of a special and differentiated speech mode of production has resulted in the reorganization of the auditory system and the formation of a specific speech mode of perception. The discriminations obtained in animals cannot be equated with the discrimina tions involved in speech perception. Experiments have concentrated on series of syllables with no more than vocal variations. These are not on a par with the rapid sequences of vocalic sounds segmented in distinguishable groups as in speech production. These are not even on the par of short adult vocalic sequences that must be parsed by infants. There is no evidence to suggest that animals, including apes, are capable of making a valid segmentation of the sequences, except possibly, for pygmy chimpanzees (Savage-Rumbaugh et al. 1986). Segmentation implies, among other factors, the discrimination between differences in duration, as is also required for the perception of the "sounds of silence" (Liberman 1982). In the latter case it has been found, for example, that if /s/ and /lit/ are separated by a short silence they are perceived as "slit" whereas if the silence is longer, about 100 ms, they are heard as "split". Although the nonvoiced occlusive is not physically
PHYLOGENY AND ONTOGENY OF LANGUAGES
21
produced in the transient forms that appear in the usual production of "split", "silence is bridged by, and participates in, integrative processes" (Repp 1987:11) and the two forms are perceived identically. This perceptive effect of sounds of silence disappears for Broca aphasics, while Wernicke aphasics show normal results (Python-Thuillard and Assal 1987). Morse et al. (1982) showed that 6month-old infants perceive the sounds of silence in the same way as adults. Segmentation appears at a very early stage as all children produce segmented forms, e.g. the noun without the article is produced before the noun accompanied by the article (Bresson 1974, 1977). Similarly, children produce the vocal elonga tions characteristic of word boundaries in French at a very early age (Bacri, this volume). This suggests that a speech mode of perception exists but does not dis prove the hypothesis that evolution consisted of adaptating production to a pre-ex isting form of acoustic signal processing where that acoustic signal is the signal used for speech. Linguistic sound perception is controlled by the left hemisphere. This was first demonstrated by Kimura (Kimura 1961, 1982) using a dichotic listening technique invented by Broadbent. The subjects are presented with two series of words simul taneously; one series is directed to the right ear and the other to the left. The sub jects are then asked to repeat what they have heard. The results are better for the series presented to the right ear, the one that is controlled by the left hemisphere. This advantage is not found when the stimuli are noises or musical notes. Other methods, such as an analysis of evoked potentials, which record the neuro-electrical correlates of perceptions give analogous results. A superiority of the right ear in language sound processing has also been found in neonates using the evoked potentials method (Molfese and Molfese 1979) as well as the dichotic listening method (Bertoncini et al. 1987, 1988; Entus 1977). Dominance of the left hemisphere is thus well established before speech production occurs and the left hemisphere is predisposed for the preferential treatment of speech signals. Sign-language as a system simultaneously qualifies as a language and as an ex tremely complex and precise set of gestures, executed in space and perceived visually. Left cerebral hemisphere subserves the expression and comprehension of sign language processing (Damasio et al. 1986; Kimura et al. 1982; Poizner et al. 1979; Poizner and Tallal 1987). The left hemisphere subserves language in a visuospatial as well as an auditory mode. The use of positron emission tomographic studies to measure regional changes in average blood flow while processing individual auditory or visual words, provides support for multiple, parallel routes between localized sensory-specific,
22
F. BRESSON
phonological, articulatory and semantic-coding areas (Petersen et al. 1988). A general network relating some of the areas could represent the different levels of lexical processing. For auditory processing, areas of activity were found in the primary auditory cortex, and left-lateralized in the temporo-parietal cortex, and in ferior anterior cingulate cortex. "The temporo-parietal region ... is a good candidate for a phonological coding region." (Petersen et al. 1988: 587). Areas related to motor output and articulatory coding that were activated included primary sensori motor cortex and a set of other structures near Broca's area related to general motor, rather than only language-specific, output. A series of studies obtained analogous results in monkeys (Beecher et al. 1979; Heffner and Heffher 1984; Petersen et al. 1978; Zoloth and Green 1979). Petersen et al. tested the capacity to discriminate between the two acoustic cues present in the "coo" sound of the Japanese macaque. The work was carried out on the above species and on two other species of macaque as well as on a cercopithecoid. The first acoustic cue has a significant value for the species while the second does not. The Japanese macaque learnt the task linked to the significant cue much more quickly than the one linked to the other cue. In the other species, there was no such difference, as neither of the cues have any significant value. The results are similar to the difference in the acquisition of categorical discriminations between the neonate and the chinchilla. The process is specific for cues that are significant in the vocalizations of a species, as has been demonstrated in the squirrel monkey by intracellular recordings (Wollberg and Newman 1972). In a second study, the same authors used a dichotic listening task and detected a superiority of the right ear over the left in response to significant vocal cues. This finding suggests that there is lateralized processing of the communication sounds of a species. If the results are confirmed in other species, they may offer a parallel which could cast light on the phylogeny of language. Patricia Kuhl (quoted by MacNeilage et al. 1987) tested 30 young macaques (of 3 species) on auditory dis crimination tasks involving speech stimuli and all favored the right hand for the re quired response. These findings should be considered in conjunction with Dewson's results (Dewson et al. 1970). In this experiment a macaque was taught to produce a pure tone by pressing the button on the left, and a white noise by pressing the button on the right. The animal was then presented with the 4 possible sequences of the two sounds to be reproduced by pressing the buttons in the right order. The task was then complicated by modifying the duration of the sounds and the intervals. When this task was mastered, a unilateral ablation of the superior temporal gyrus (the area that controls language in humans) was performed. After the ablation, a considerable
PHYLOGENY AND ONTOGENY OF LANGUAGES
23
and durable deficit was noted, whatever the side of the lesion and its relationship to the dominant hand. Dewson then modified the procedure (1977) so that the buttons were distinguished not by their position (left or right) but by color. With the nonspatial keyboard only left-sided lesions produced a deficit. These results, like those of Petersen et al (1978, 1984) and Heffner and Heffner (1984), suggest that the characteristics of the functions of the left hemisphere in humans and nonhuman primates are homologous. Not only is there lateral dominance in the control of temporo-auditory vocal information, but the left hemi sphere is dominant. As primates are phylogenetically close to man, their left-sided dominance is more striking than the analogous facts discovered in birds by Nottebohm(1971, 1972, 1980).
Referential Relationships The communication signals observed in animals, including apes, are devoid of the referential value of language. The signals consist of calls, threats and warnings. They denote emotional states and their referential value is limited to designating the state of the animal producing them. The chimpanzee can exhibit a spontaneous behavior which may be akin to pointing. The animal grasps the wrist of its human keeper and drags the hand towards a desired object that the animal cannot reach or is not allowed to have. This incitation to take and give is a very restricted form of designation. First, the two partners must move, and second, although limited, it is a sort of relay of a common action. The behavior can be compared to other ethological rituals that are performed in order to elicit the same or other appropriate response in another individual. Pointing is not observed, even in apes. Apes can learn this gesture, but do not seem to use it spontaneously. Menzel commented (1973) "one good reason why the chimpanzee rarely points is that it does not have to; as a quadruped, it points with its whole body." The author has also demonstrated that a chimpanzee's companion knows where it is going (Menzel and Halperin 1975). Information is only given in an immediate context and cannot be separated from the action. However, S. Savage-Rumbaugh's pygmy chimpanzees spontaneously used iconic gestures to communicate preferred directions and actions they wished to perform. The pygmy chimpanzees, Kanzi and Mulika, between 6 and 16 months of age, "used an out stretched arm and hand to point toward areas to which they wished to be carried; they made twisting motions toward containers when they needed help in opening twist-top lids... Gestures were often accompanied by vocalizations that served to
24
F. BRESSON
orient the listener's attention and to convey affect" (Savage-Rumbaugh et al. 1986:218). These behaviors are also observed in children (Clark 1978; Trevarthen and Hubley 1978). In humans, pointing is related to the ability to look in the same direction as another individual, i.e. to focus on the same object (Butterworth and Cochran 1980; Butterworth and Grover 1988; Leung and Rheingold 1981; Scaife and Bruner 1975). Focus agreement is at first limited to the visual field common to both partners. Pointing is also subject to limits, initially as it is only used to request or signal surprising objects or confirm a name. In sign language (ASL), personal pronouns are formed by pointing directly to the addressee or oneself. It was observed that deaf children initially (ages 10 to 12 months) pointed to persons, objects and locations. But these children manifested discontinuity in the transition from prelinguistic to linguistic communication (Pettito 1987). But they then exhibited a long avoidance period, during which orle func tion of the pointing gesture (pointing to self and others) dropped out completely. During this period they continued to use other types of pointing. When pointing to self and others, these children acquired knowledge of personal pronouns over a period of time, displaying errors similar to those of 2-year-old hearing children. The first linguistic shifters present similar restrictions. The reply "here" to the question "where is X?" can only be given if X is within the child's visual field and if s/he can move to the object to designate it (Cabrejo-Parra 1977). Complex referential capacity is not observed before predicative activity. Whether this is re lated to a conceptual activity or not, is another question (Savage-Rumbaugh et al. 1980).
Predication and Anaphoric Relationships The ability to produce predicative statements and the capacity to produce decla rative statements are probably related. They appear to be essential characteristics of linguistic representations. They undoubtedly constitute very evolved forms of lan guage and are not independent of social communication and cognitive conditions which lead to the adoption of a knowing and declarative mode as opposed to a practical procedural attitude which aims at an effect on the physical or social en vironment (and it may not be possible to know how it is learnt). Rules and strategies of these procedural abilities may not be conscious and reflect the unselective aggregation of frequency information (Berry and Broadbent 1984; Broadbent et al. 1986; Hayes and Broadbent 1988). Topics of children's early linguistic
PHYLOGENY AND ONTOGENY OF LANGUAGES
25
knowledge may appear to result from multiply-represented procedures (KarmiloffSmith 1986). The difference between these modes of knowledge is apparent in Sheeran's work (1973). As early as 18 months, children are capable of anticipating the con stancy of the weight of a cylinder they are presented several times. This can be seen from the invariance of the muscle tension of the arms; they still anticipate the invariance if the cylinder is deformed in front of them (Mounoud and Bower 1974). Sheeran has shown that 3-year-olds can acquire this sense of practical conservation even if they are shown to be nonconserving in Piaget and Inhelder's classical situa tion. Piaget predicted this when he states in The Construction of Reality (1937/1954) that, "without language the child manages to form a system of schemas capable of indefinite combinations which is the precursor of concepts and logical relationships ... these schemas are even suited to certain spontaneous group ings equivalent to mental deduction and construction" (317, my translation). He contrasted the products of these "quasi-operations" with propositions that can only be true or false, and for which he stated that language was a necessary condition. Work by Premack and the Rumbaughs has shown that chimpanzees seem to form concepts, subsume examples into a class, produce analogous relationships (i.e. relationships between relationships) and produce invariance responses in some con servation situations. These abilities are apparent in situations which involve the manipulation of arbitrary signs which leads to the construction of propositional or quasi-propositional structures. However, differentiation between items whose referents were similar, such as "juice" and "coke", are difficult for chimpanzees (Savage-Rumbaugh et al. 1986). By contrast, for young pygmy chimpanzees this type of differentiation waw made quite spontaneously and they did not use the symbolic items interchangeably. It is necessary to use the relation in both direc tions, from the categorical symbol to the specific one and from the specific to the categorical one. Transmission of tool-making techniques has been invoked as a sign of the exist ence of language, but is was seen that the argument was unconvincing; language is a sufficient but not a necessary condition. Coordination of collective activities could also be invoked such as hunting large prey. However the coordination may not require more than injunctive signals used in quite complex forms of nonverbal communication. If the orders are given in situ in a language, the injunctive forms do not require either a syntax including a predicate and anaphorics or semantics based on conceptual relationships such as those chimpanzees can be taught to produce. We have no proof that the earliest forms of language production took or may have taken a declarative form.
26
F. BRESSON
Chimpanzees never formed requests in which someone other than themselves was the beneficiary of the request. By contrast, for young pygmy chimpanzees the "ability-to conceptualize and then to symbolically initiate complex forms of interac tion between others would seem to be a precursor of syntactic structure, if not the basis itself for the occurrence of syntax" (Savage-Rumbaugh et al. 1986: 230). One may wonder whether the type of language learnt serves as a means of as king the questions that the investigator decides to ask, questions which are not part of the repertory of the species and are not everyday situations, or whether language is a necessary prerequisite for treating the problem. The latter interpretation can be understood in two ways. First, that language is needed to perform mental actions required to solve complex problems. Second, language permits one to abstract the properties of an object and relational properties. This is a necessary condition for making true/false statements. The ability to communicate about absent referents or events could be used by pygmy chimpanzees (Savage-Rumbaugh et al. 1986). Although chimpanzees may show a type of spontaneously developed social intelligence, this intelligence does not seem to imply a "knowledge set". It is more difficult to communicate about spatial or temporal order, as "x is before y" or "x is after y". Order relations, par ticularly in the past or the distant future, imply other forms of representations than those which are used for an absent event or referent by pygmy chimpanzees; they can communicate "I want to go out" or "I want a cake," (an object not present) but they can't communicate "I wished I had a cake," (but do not wish any longer) nor "I would like to have a cake tomorrow." This inability supports the claim that for the human species remembering cannot exist before the child is 3 years old, i.e. when minimal linguistic competence has been acquired. The inhumations by "Mousterian Proto-Cro-Magnon" (Vandermeersch 1981) is not a necessary or suffi cient condition of distant future representation, but a relationship between these representations and language is very probable. The first fire-places, of red ochre, laid down 500,000 years ago, might imply the communication of future events, but that is also an unsolved problem. There is reason to believe that apes do have the capacity to produce mental im ages and to use them (Brésard 1988), and to produce enactic representations (Bruner 1964) to solve the problems they encounter or those we set. Semantic rep resentation in language has a structure that is entirely different from that of mental images or pictures. For images, there are no markers of determination; inferences and abstractions are not included in figurative organization, which does not include syntactic markers controlling the nature and order of operations to be carried out. Picture comprehension proceeds by analysis whereas verbal statement proceeds by
PHYLOGENY AND ONTOGENY OF LANGUAGES
27
synthesis. The latter can be compared to the solution of a mathematical problem that produces a result, just as the solution of syntactic operations produces mean ing. Language must translate the forms of perceptively visual information, or infor mation coming from other modalities, into linguistic semantic forms (Jackendoff 1987). Representations of physical objects and locations must correspond to con ceptual and semantic representations. This implies more than immediate perception, because it is necessary to set up a correspondence both from categories to objects and from objects to categories. Language must also translate remembered iconic images and the ordering of absent referent rehearsal. Jackendoff maps the proper ties of the 3-D model of D. Marr (1982) which explicitly encodes an object's shape in long-term memory. For this model, objects are represented as (mentally) occupy ing volume in space, regardless of the viewer's position and in terms of a hierarchi cal decomposition of the object into parts and parts of parts. Jackendoff shows that the formation rules for semantic/conceptual structure include a vocabulary of primi tive conceptual categories or "semantic parts of speech"; these categories are such entities as Thing (or Object), Event, State, Action, Place, Path, Property. Past or fu ture must include order of temporal place. Thus, Jackendoff s model can be taken as a good example of the transition from perceiving thought without language to human language. The topic-comment structure in language (Gruber 1967) provides a model of the procedure of determining a reference and a determination. Another example is genitive structure. The structure is of the analytical type; the determinant "tells about" what is determined. Syntactic markers indicate the relationship whether it is stated by the speaker or to be inferred by the listener. The phrase, "the dog's paw", implies that "dog", the concept and paradigm image comprised of four paws, must be related to "dog-paw", which must be distinguished or detennined. However, forms of representation analogous to the analytical structure of topic-comment or genitive relationships have not been described in their behaviors. The speed of acquisition of syntactic rules has led many other authors to propose that humans have a universal grammar and a general disposition to detect the forms it takes in spoken languages. The rules of this grammar can be conceived metaphorically as schemas of rules, just as in logic one talks about schemas of axioms. Each language fixes the forms of production from a repertory of pos sibilities and compatibilities. This position corresponds, mutatis mutandis, to the one described above as being the probable one for the speech mode of perception (Liberman 1982). In one language word boundaries may be marked by a variation in the relative vocalic duration, as is the case with French, while another uses the
28
F. BRESSON
parameter of intensity. The same could then be true of determination markers or aspectual markers, for example. The existence of an innate predisposition as part of the inheritance of the species, would also explain that language acquisition does not require a highly developed intellectual capacity. Even children with very low intel ligence quotients learn to talk and do not take very much longer to do so than nor mal children. These arguments, although plausible, do not provide firm proof of these hypotheses and the alternatives cannot be rejected experimentally. Nevertheless, many convincing factors argue for an innate predisposition for the conditions necessary for easy language acquisition. We discussed above the positioning func tion of the left hand, complemented by the action of the right. This cooperation, where the right hand seems metaphorically to determine the object positioned by the left, is reminiscent of the topic-comment relationship. Similarly, focus of atten tion would appear to confer special qualities to shapes located at the focal point. Other activities which present analogies with linguistic determinations, particularly positioning operations, have also been described in this article. It is more difficult to find analogies with anaphoric operations, which consist of marking the identity of the determinate through several determinations like transforming a variable into a constant by its algebraic "binding". Rumelhart and McClelland have described a connectionist (distributed parallel processing) model of the acquisition of the past tense in English which successfully maps many stems onto their past forms (1986). Other connectionist models work mainly with "subsymbolic computation". The relationship between the "Neural net works that learn temporal sequences by selection" (Dehaene et al. 1987) and a model of 3-year-old children's language development which is not yet satisfyingly constructed, and the relationship between thought without language and the neurophysiological speech-brain, are the fundamental issues to be solved in the near fu ture.
Conclusion One of the properties of evolution is that it is possible to establish homologous links and to show from which organ a new organ has developed. It is the inability to establish a filiation for syntax which is the major difficulty in understanding the phylogeny of language today. The absence of traces of previous forms of language which were perhaps used by the common ancestor of apes and Homo sapien
PHYLOGENY AND ONTOGENY OF LANGUAGES
29
sapiens, does not allow for speculation at present. The initial forms of language ac tivity in human development are too specific. This review of the relevant points shows that language probably resulted from a convergent evolution of cerebral structures, behaviors and the methods of data processing that the structures permitted. A multicausal hypothesis is supported by the fact that lesions in various areas of the brain produce disturbances in language activities. The unity of language, as seen by linguistics, does not imply a unity of the structural support or its phylogenetic origins. Further progress in this field will certainly come about by increasing our knowledge of the cerebral mechanisms of language.
Prenatal Familiarization J.P. Lecanuet, . Granier-Deferre Laboratoire de Psycho-Biologie de l' Enfant EPHE (3ème section) CNRS, Paris and M.-C. Busnel Groupe Génétique, Neurogénétique et Comportements UER Biomedicale, Université René Descartes, Paris
Introduction In the last fifty years, prenatal and neonatal research has drastically modified the view of the neonate held by psychologists, physicians and parents. The newborn baby is no longer considered a passive subject requiring primarily to be fed, soothed and cleaned. S/he has become a true person equipped with many percep tive abilities, which s/he uses to actively seek out contacts. However, these abilities do not "blossom" suddenly at birth. Anatomo-pathological and physiological studies on fetuses and premature babies have demonstrated that all the sensory sys tems on which these abilities are based are functional at normal term. Moreover, some of these systems, such as the auditory system, are functional months before. Activation of these systems by intrauterine stimulation may have structural, func tional and, last but not least, behavioral consequences. In accordance with this hypothesis, this paper reviews the data on: (1) prenatal auditory system develop ment, (2) the nature of the sound environment inside the uterus, in particular, speech emergence and fetal reactions to this environment, and (3) the consequences of prenatal familiarization with these various sounds which may contribute to the development of the auditory discriminative capacities of the newborn and his/her perinatal voice and language perception.
32
IP. LECANUET, . GRANTER-DEFERRE AND M.-C. BUSNEL
Development of the Auditory System The human auditory system is very complex. We will only summarize the development of its two major elements: the middle and inner ear (a detailed review of this ontogeny can be found in Rubel 1978; Pujol and Uziel 1986). The middle ear transmits acoustic vibrations from the external ear to the inner ear. This is done through the tympanic membrane which begins its development around the 11th week of gestation and through the ossicles which develop from the 8th week. In the inner ear, the cochlea develops from the otocyst, a structure ap pearing around the 28th day. It reaches its full morphological development by the 10th week, and its final adult size by the 20th week. The organ of Corti, which bears the auditory receptors, develops within the cochlea from the 8th week on wards, and seems to be functional by the 18-20th week according to Pujol and Uziel (1986), who base their inference on the parallelism in chronology of cochlear development in humans and in every studied mammal. The first cochlear potentials can be recorded in all animal studies at the same developmental stage. Maturation of both inner and middle ear probably ends during the 8th month. At the onset of cochlear functioning, auditory competencies, which are well characterized in animal studies, are poor. Electrophysiological responses can only be recorded for medium frequencies (1000-2000 Hz, depending on the species). Auditory thresholds are high (around 100 dB). There is no frequency discrimina tion and no temporal coding. However, they improve rapidly and gradually: auditory thresholds decrease, temporal coding begins, frequency sensitivity widens, first in the low frequency range, then, in the high frequency range, and after that, unit frequency selectivity sharpens. The first cochlear potentials are evoked by midfrequencies, although the base of cochlea, which usually codes high frequencies when mature, is the first cochlear zone to develop. Rubel and Ryals (1983) showed that cochlear tonotopy (i.e. the spatial distribution of frequency sensitivity in the cochlea) changes during maturation. It has been thought that residual embryonic mesenchyma and amniotic fluid, still present in the external and middle ear at birth, may impair in utero middle ear functioning. However, the anatomical studies of Maclellan et al.(1964) and the tympanometric studies of Keith (1975) show that they do not induce ear drum and ossicle stiffness. Nevertheless, it can be assumed that the middle ear is not neces sary for fetal audition since it is adapted to the amplification of acoustic stimuli in aerial life. Without this amplification, clinical studies show that there is an average hearing loss of 35 dB from the aerial environment of the outer ear to the liquidian environment of the cochlea. In utero, since the outer and middle ears are filled with
PRENATAL FAMILIARIZATION
33
amniotic fluid, and since liquids, tissues and bones have close conducting proper ties, the acoustic energy inside the uterine cavity can reach the cochlear receptors with negligible energy loss; thus, there is no need for an amplifying system. How ever, middle ear transmission might still be possible, but as pointed out by Rubel (1985), its prenatal functioning is necessarily different from its postnatal function ing. Central processing of auditory information appears during fetal life, as shown by the development of auditory evoked potentials studied in premature babies. Brainstem responses, up to the inferior colliculi, are first recorded in premature babies by the 26th week; thalamic components are recorded at 35 weeks with only a 10 dB HL threshold (Schulman-Galambos and Galambos 1979; Starr et al. 1977). Anand and Hickey (1987), in their recent review, emphasized that electrophys iological and other convergent data relative to the maturation of the sleep states (Torres and Anderson 1985), brain metabolism (Chugani et al. 1986), and be havioral responses (Sammons 1980), show that cortical functioning, and thus some higher processing of sensory information, may begin around the 30th week. The kind of auditory information the fetus can process in a relatively rich acoustic en vironment will be described in the following section.
Intrauterine Acoustic Stimulation and Fetal Responses Intra-amniotic recordings have been performed during labor with either a hydrophone or a microphone inserted in utero after rupture of the amniotic membranes. They showed that the intrauterine noise is composed of (a) sounds of maternal, placental and fetal origins (endogenous background noise), (b) attenuated and filtered external sounds (Johansson et al. 1964; Bench 1965; Walker et al. 1971; Henschall 1972; Murooka et al. 1976; Querleu et al. 1981, 1986). The mother's voice has a particular status; it is an endogenous sound with some aerial components, also transmitted in utero.
Endogenous Background Noise It is composed of several vascular sounds, mainly from maternal heart, placental and umbilical cord origin, and of gastro-intestinal sounds. According to Querleu et al. (1981, 1986), its maximum energy is found in very low frequencies (below 70 Hz). The energy then decreases as the frequency increases, and reaches its lowest
34
J.P. LECANUET. . GRANTER-DEEERRE AND M.-C. BUSNEL
level at 700 Hz. Estimations of the mean acoustic level of the endogenous back ground noise range from 28 dB to 95 dB SPL depending on the recording devices and methods. The lowest values come from the most recent study (Querleu et al. 1986), in which the microphone was placed far away from the placenta.
External Sounds In hearing and in analyzing the acoustical structure of intrauterine recordings, it is clear that external sounds emerge from the background noise. These sounds are however attenuated in the high frequencies by the maternal abdominal tissues and partially masked by the low frequency background noise. Although the authors who performed in utero recordings measured different at tenuation curves, they all agree that attenuation increases with frequency. However, Querleu et al. (1986) pointed out recently that the attenuation of high frequencies has been overestimated in previous work. For example, the highest frequency com ponents of a musical triangle sound emitted at 70-80 dB ex utero can still be detected in intrauterine recordings with no more than 40 dB attenuation at 18 kHz. Since the mother's voice benefits from the additional means of transmission through the maternal body, it emerges from the background noise with a higher in tensity level than that of other female or male voices (24 dB vs. 8-12 dB) which are likewise emitted at the external level of 60 dB. In spite of a reduction of the frequency bandwidth and of the transients, in trauterine recordings of speaking adults indicate that a voice's melodic contours are not altered. A certain amount of phonetic features can be recognized by adult lis teners when emitted at 60 dB SPL (Querleu et al. 1986). For both maternal and ex ternal voices (Querleu et al. 1986) simple words, uttered at 90 dB, can be understood, especially if they include plosive sounds (Busnel 1979). It should be pointed out that the intensity level and frequency spectrum of in utero recordings might not exactly correspond to those which stimulate the fetal ear, since the recordings were made after the rupture of the amniotic membranes and leakage of some amniotic fluid. This methodological drawback was overcome in the ewe (Armitage et al. 1980), by permanently implanting a microphone and a radio emitting system inside the womb. Analysis of 24 hour recordings revealed many different noises: Meeting, chewing, breathing and gastro-intestinal noises but, curiously enough, only intermittent and very low level heart-beat sounds. The in-
PRENATAL FAMILIARIZATION
35
tensity levels ranged from 36 to 64 dB SPL with a mean of 60 dB. Attenuation of external sounds was similar to the lowest estimation of human studies.
Fetal Reactiveness to Sound Fetal reactiveness to acoustic stimuli was demonstrated in a series of early studies (Peiper 1925; Forbes and Forbes 1927; Ray 1932; Sontag and Wallace 1935, 1936), most of which investigated near term fetuses. Since then, research in this area has widely developed. Its aim has been mainly clinical, and has used responsiveness to acoustical stimuli to assess fetal well-being (Goodlin and Schmidt 1972; Jensen 1984; Leader et al. 1982a, b; Querleu et al. 1984a, Davey et al 1984; Divon et al 1985; Serafini et al 1984). Three major types of fetal responses can be recorded: (1) cardiac (accelerations, decelerations or biphasic responses), which can be accompanied by (2) motor reflex responses, and (3) brainstem evoked potentials (BER). A fourth type of response consisting of histochemical modifications has also been used in animal research (Horner et al.. 1987). Cardiotocography and real-time scanning systems are now generally used to record heart rate responses and to detect fetal movements. Fetal auditory evoked potentials (BER) have been recorded during labor after rupture of the amniotic membranes (Scibetta et al 1971; Barden et al 1978). In two studies (Sakabe and Arayama 1969; Schwartzman et al 1985), the authors claimed that they were able to record BER through the maternal abdominal wall but their origin is still uncer tain. In animals, cochlear and brainstem auditory potentials have been recorded with implanted electrodes inside the uterus in guinea-pigs (Scibetta and Rosen 1969) and sheep fetuses (Jost et al 1972; Woods et al. 1983; Armitage and Vince, personal communication). More than 25 recent studies (see Busnel and Granier-Deferre 1983) have demonstrated that, as in pre- and full-term neonates and young infants, fetal reac tiveness to acoustical stimulation depends on both stimulus characteristics and be havioral state. The stimulations classically used are very loud (over 105 dB, which can be es timated to be about 85 dB in utero) and of a relatively short duration (5 to 30 s). They may induce cardiac heart rate accelerations and one or several components of the startle motor responses such as: eye-blink, head retroflexion, limb extensionflexion or trunk rotation. These responses can be observed in some fetuses from
36
J.P. LECANUET, . GRANIER-DEFERRE AND M.-C. BUSNEL
22-24 weeks gestation (GA), while motor responses are evoked in all fetuses after 28-30 weeks (Birnholz and Benacerraf 1983; Leader et al. 1982). In the near-term fetus, the evocation of these responses clearly depends on stimulus characteristics and fetal state. At a lower level, accelerations are more reliably evoked than strong motor responses. When stimulus intensity and frequency rise, the number of stimu lated subjects performing one or both responses, as well as the mean amplitude of the accelerations (ranging from 7 to 45 bpm), increase (Dwornicka et al. 1964, Read and Miller 1977; Querleu and Renard 1981; Jensen 1984; Granier-Deferre et al. 1985; Lecanuet et al. 1986, 1988). Response ratios are higher when the stimula tions are delivered during a fetal state similar to the neonate's active sleep - when spontaneous heart rate variability is high - than during a fetal state of quiet sleep, when heart rate variability is low (Schmidt et al. 1985; Lecanuet et al. 1986). When emitting less intense stimuli (80 to 100 dB), 8-9 month-old fetuses display cardiac decelerations (mean amplitude 10 bpm). They are more often evoked by low frequency octave-band noise stimuli than by mid-high frequency stimuli and are never accompanied by a motor response. Cardiac decelerations are also reliably induced by speech sounds delivered during quiet sleep (Lecanuet et al. 1986, 1987, 1988). There is now important convergent data showing that fetuses, like neonates, per ceive acoustical stimuli. Near term fetuses can also discriminate between two com plex stimuli, each composed of a pair of syllables (emitted at 95 dB SPL), different in the order of the syllables (BABI versus BIBA) (Lecanuet et al. 1987). Such com petence may be partially a consequence of fetal familiarization to speech sounds.
Consequences of Fetal Familiarization to Speech Sounds The fetus is exposed to a large variety of simple and complex sounds. Music, song and speech sounds have been shown recently to produce behavioral changes in the fetus while they may also induce, as may other sounds, structural and func tional effects, some of which are required for the normal functioning of the auditory system. The data presented above have shown that during the last 10 weeks of in trauterine life the fetus, whose auditory system can accurately process acoustical stimulations, is exposed to maternal and, to a smaller extent, external speech sounds. Stimulation by speech sounds may exert general effects, i.e. effects which could be induced by any other type of acoustical stimulation, as well as specific ef fects. These effects have been described (A) at structural and functional levels
PRENATAL FAMILIARIZATION
37
where they can only be inferred from research on animals, and (B) at a behavioral level which can be directly assessed from human experiments.
Structural and Functional Effects Animal studies performed mainly on species in which the auditory function starts after birth (mouse, rat, gerbil, cat, guinea-pig, chick, owl) have shown that af ferent input is necessary to establish and maintain a correct functioning of the auditory system and to partially control sound integration in the brainstem. Bilateral auditory deprivation, starting before or during the period of appearance of cochlear potentials, causes the same anatomical alterations, but to a smaller extent, than a bilateral destruction of the cochleas: a reduction in the number and size of the neurons and dendritic arborization along the auditory relays of the brainstem acous tic pathway, higher electrophysiological auditory thresholds, and a reduction in the ability to discriminate complex rhythmic structures. Monaural deprivation causes neuronal and electrophysiological alterations affect ing mostly ipsilateral structures, dealing with binaural interactions. Some experi ments show that sound deprivation would mainly induce a delay in the development of auditory sensitivity. Impairment may be reversible if the depriva tion does not extend to a critical period (review in: Ruben and Rapin 1980; Conlee and Parks 1981; Moore 1985; Clopton 1986). Such deficits impair auditory spatial localization (Clements and Kelly 1978; Knudsen et al. 1982). On the other hand, prolonged exposure to selectively enriched sound environments - which do not in duce any acoustic trauma - stimulates local dendritic growth (Smith et al. 1983), modifies the reactivity of central acoustic units (Clopton and Winfield 1976; Sanes and Constantine-Paton 1983) and seems to facilitate some discriminative auditory tasks. Thus, it can be hypothesized that sounds reaching the fetal ear might con tribute to the structural and functional shaping of the auditory pathways (Wein berger and Diamond 1987).
Behavioral Effects Fetal learning abilities in the auditory mode (habituation, conditioning or im printing) have been demonstrated on animal and human subjects. For instance, human fetuses show a short-term habituation of the motor response to a loud sound (Leader et al. 1982a, b; Madison et al. 1985) and so do sheep (Leader et al. 1988).
38
J.P. LECANUET, . GRANTER-DEFERRE AND M.-C. BUSNEL
Concerning long-tenn habituation, human newborns (Ando and Hattori 1971, 1977), like neonate guinea pigs (Vince 1979), are less disturbed by a startling stimulus if they have been exposed to it for a long time during the prenatal period. Spelt, (1948) after Ray (1932), conditioned the motor response to a loud sound, a neutral vibro-acoustic stimulus. The authors claimed that this conditioning, which was subjected to extinction, could be reactivated later on. In a more recent study, a maternal relaxation state (US inducing fetal movements and supposedly positively perceived by the fetus) was coupled with a musical sequence (CS). On hearing the musical sequence, during a nonrelaxed maternal state, exposed fetuses started a motor activity sequence (Feijoo 1981). Because of the lack of control groups and of the difficulties in establishing classical conditioning in the neonate, the results of these attempts have been subjected to some controversy. Thus, one may consider that prenatal exposure to certain aspects of speech stimuli may promote various fetal acquisitions via associative learning or via a process similar to the chick's auditory imprinting in ovo which has been evidenced with artificial (Green and Adkins 1975; Lien 1976; Rajecki 1974) or natural stimuli, as with maternal vocalizations (Guyomarc'h 1974). Behavioral evidence of such acquisition can be found in: (i) the sensitivity newborns have for speech stimuli and (ii) their specific preferences for different aspects of prenatal speech stimulation. In the newborn, it has been shown that frequencies within the speech range are the most effective in inducing cardiac responses (Hutt and Hutt 1970). Synthetic syllables evoke cardiac decelerations (Eisenberg 1976). Voices are preferred to in strumental music by neonates and speech is preferred to nonspeech stimuli (Siperstein and Butterfield 1973; Eisenberg 1976). A female voice seems particular ly effective in catching babies' attention (Wolff 1969; Brazelton et al. 1974). The onset of a female voice elicits a decelerative cardiac response, whereas the onset of a male voice does not (Lipsitt 1979). Moreover, newborns are sensitive to the lin guistic structure of acoustic stimuli. It is now well documented that young infants possess very accurate discriminative capacities concerning various speech features (see Aslin et al. 1983 for review). Newborns appear to be able to discriminate the phonetic components of speech (Bertoncini and Mehler 1981a; Moon 1985). Three categories of speech preferences have been revealed in recent studies: (1) for a specific speaker: newborns prefer their mother's voice, (2) for a specific se quence: newborns prefer a speech sequence repeatedly spoken or sung during the last fetal weeks, (3) for a specific language: newborns prefer the maternal language.
PRENATAL FAMILIARIZATION
39
Maternal Voice. The maternal voice seems to be a particularly attractive stimulus to newborns. It is the most soothing voice at 15 days (Wolff 1969), it elicits preferential head orientations at 10 days (André-Thomas 1966) and even at 5 days (Hammond 1970). Babies less than 2 hours old tested in the delivery room, show more global reactiveness to the sound of their first name when spoken by their mother (36%) than by 4 different nurses (12%) (Querleu et al 1984b). The first comparison of maternal and other female voice attractiveness was con ducted by Mills and Melhuish (1974). They found that the mean nonnutritive suck ing score of 3-week-old babies was higher when the onset of a sucking burst started the play-back of the maternal voice than when it elicited the play-back of another female voice. Mehler et al. (1978) confirmed the higher attractiveness of the maternal voice with 1-month-old babies, this preference disappears if prosodic cues are removed. During the first few weeks of postnatal life, the maternal voice is often paired with various, highly reinforcing events (food, tactile and vestibular stimulations). Therefore, when experiments are conducted after a long period of postnatal ex perience with this voice, it is impossible to demonstrate that its positive (reinforc ing) value has developed during the prenatal period. This led DeCasper and Fifer (1980) to test 2- to 3-day-old babies, who had a very short postnatal experience with their mother (less than 12 h). In order to test direct preference, they designed a conditioned nonnutritive sucking procedure in which the baby controls which voice (maternal or other female) it may hear by modifying the duration of the pauses be tween bursts of sucks. Babies learned either to lengthen or to reduce their pauses in order to listen to a recording of their mother's voice. This preference was also demonstrated in an experiment using a modified procedure (Fifer 1981). In order to listen to their mother's voice, babies were to start sucking upon hearing a particular auditory signal emitted during the pauses. Sucking during the emission of another auditory signal triggered the play-back of an unknown female voice. In another study, DeCasper and Prescott (1984) found that male speech was not an effective reward for newborn learning and that three-day-old babies did not prefer their father's voice over another male voice. This absence of preference was not due to a lack of discrimination between different male voices, as shown by a control experiment, but might be related to a lack of prenatal experience of this voice, due either to the fact that external voices are less intense than the maternal one in utero, or to its less frequent occurrence in most pregnancies, or both. Moreover, DeCasper and Prescott's demonstration that an explicit postnatal ex-
40
J.P. LECANUET. . GRANTER-DEFERRE AND M.-C. BUSNEL
perience with the father's voice (an average duration of four hours within two days) was not sufficient to induce a preference for this voice, suggested that babies' preference for maternal voice was probably not a result of a postnatal experience with this voice. DeCasper concluded from the studies of this group and of the works by Wolff (1963) and Brazelton (1978) that babies show hierarchical preferences - first for the mother's voice, then for other female voices and last for male voices, including the paternal one. In search of more experimental evidence for a prenatal development of preference for maternal voice, Spence and DeCasper (1986) compared the reactiveness of babies to two types of recordings of their mother's and of another female voice either as the voice can be heard postnatally, or filtered in order to sound like an intrauterine recording (low pass filtered). Babies showed no preference between the filtered or the unfiltered recordings of their mother's voice, but, when tested with the other woman's voice they preferred the unfiltered version. For the authors, this difference between groups of subjects supports the hypothesis of the influence of prenatal experience with low-frequency characteristics of maternal voice on postnatal perception of her voice. Investigating the same question, Moon and Fifer (1986) found that babies preferred a filtered version of their mother's voice to a nonfiltered one. However, in their experiment, more mid-frequencies were filtered than in the Spence and DeCasper study, and the filtered voice was mixed with an intrauterine heart beat recording which may increase attractiveness for this stimula tion. Specific Speech Sequences. A second type of studies bears on the prenatal learn ing of prosodic sequences of maternal speech which might be similar to what has often been reported in individual or cultural stories and legends. Infants and some times even teenagers are particularly sensitive to musical sequences which have been repeatedly emitted during their fetal life. For example, Derlon (1981) reports that in XVIIth century France, Gipsy babies had been separated from their parents during a persecution period and placed in alien families. Several years later, when this particular persecution ended, Gipsy fathers travelled across the country playing their band's traditional songs on the violin, and when their children heard them, they came back, recognizing the tunes they had heard during their fetal and/or neonatal life. Postnatal recognition of specific prenatal speech sequences has also been inves tigated in two experimental set ups. DeCasper and Spence (1986) asked pregnant women to read out loud, twice a day, during the last 6 weeks of their pregnancy, a
PRENATAL FAMILIARIZATION
41
particular story containing highly recurrent vowels (A or for example). A second group of women read another story having different acoustic features. Babies born from both groups were tested on their third day with the nonnutritive sucking choice-test procedure (cited above) in order to assess their preference for the story their mother recited (target story) over the story recited by the other group of mothers (control story), both being read by an unfamiliar female voice. Half of the babies of each group had to shorten the pauses between sucking bursts in order to hear the target story, the other half had to lengthen these pauses. Results indicate that babies chose the target story more often than the control one, suggesting a prenatal learning of some acoustic features of the story. Using the same test procedure, Satt (1984) found that 3-day-old babies prefer a lullaby repeatedly emitted during the end of their fetal life to an unfamiliar one, both lullabies being recorded by the mother. Maternal Language. Results from Mehler et al. (1987) suggest that the maternal language may be more attractive than a foreign one to 4-day-old neonates. Babies born from French-speaking mothers not only discriminated between French and Russian spoken by a bilingual woman in a nonnutritive (noncontingent) sucking test, but they also displayed higher mean sucking rates during the habituation phase of the experiment when listening to their maternal language than to Russian. Moreover, babies born from mothers speaking a language other than French or Russian were unable to discriminate between them (see Bertoncini et al., this volume).
Conclusion The studies reviewed in this chapter show that (1) the human auditory system is functional 3-4 months before birth, (2) a large variety of acoustical stimulations, particularly maternal and external speech sounds, emerge from the intrauterine background noise and are loud enough to stimulate the fetal auditory system, (3) fetuses beyond 28-30 weeks of gestation reliably react to external sounds - includ ing speech later on - displaying either startle-type responses with heart rate ac celerations or heart rate decelerative responses as a function of the structure of the stimulation and of their behavioral state, (4) prenatal exposure to acoustical stimulations may exert structural and functional effects on the human auditory sys tem as it does in animals at the same stage of auditory development, (5) prenatal familiarization to speech sounds may contribute to the development of a specific
42
J.P. LECANUET, . GRANIER-DEFERRE AND M.-C. BUSNEL
sensitivity to these stimuli and to the formation of a preference for (i) a given speaker (mother), (ii) some particular prosodic sequence when read or sung repeatedly by the mother during the last weeks of her pregnancy and (iii) a given language (maternal). All these statements are not equally supported by the same amount of ex perimental data; they do bring, however, promising results which call for replica tions to be conducted with similar or other methodological procedures. Nevertheless, one may conclude that prenatal familiarization to speech sounds presumably bears both global and specific consequences. If, as stated by DeCasper and Fifer (1980), the less than 3-day-old newborns prefer their mother's voice because it is already the most familiar, and thus posi tively reinforcing and soothing, one is led to reconsider with Fifer (1981) the at tachment theory of Bowlby (1969) and Ainsworth (1979). The theory postulates that babies first go through a phase during which they do not discriminate their mother from other women (until 8 or 12 weeks); but experimental results described above show that this is not the case. It should be considered, in agreement with the interactionist position of Cairns (1979), that the very early recognition of the mother's voice organizes the discrimination and recognition of the other maternal characteristics that will gradually be associated with it. Several animal studies (Stickrod et al. 1982; Smotherman 1982a, b; Vince and Ward 1984; Hudson 1985) have demonstrated that there is also a continuity from fetal to postnatal life from the olfactory modality. These experiments suggest that human neonates may also be prenatally familiarized to their mother this way. Ele ments of recognition may come from the aromatic substances of maternal food which can be found in breast milk and colostrum as well as in amniotic fluid (Arena 1970; Vorherr 1975). Such physiological processes may underlie human babies' early preference for maternal odors compared to odors of other recently delivered women (Macfarlane 1975; Russel 1976; Schaal 1984). These experimental data indicate that, in spite of the relative immaturity of the fetal brain and in spite of the fact that most prenatal stimulations occur during sleep episodes, some learning takes place. As mentioned above, it has been frequently suggested that these acquisitions may develop via a perceptual imprinting-like learning. Could prenatal familiarization to sounds contribute to the development of the large and complex neonatal perceptual auditory abilities which have been reported in the last 20 years, in particular for speech sounds (see Bertoncini, this volume)? The yet little direct experimental results make it difficult to infer with certainty. One may consider that this familiarization may activate the human's specific innate
PRENATAL FAMILIARIZATION
43
abilities to process speech sounds, as postulated by Chomsky (1975). An alternative hypothesis is that there are innate potentialities (not specific to humans) that develop almost to their optimum functional state, very early in life, because of the relative "richness" of prenatal auditory activation. We have briefly mentioned the growing evidence from animal studies showing how critical they are during the development of the auditory system. Human clinical and psycho-acoustic studies (Rapin 1979; Ruben and Rapin 1980; Henry 1983) have shown that adequate stimulation of the auditory system is necessary to maintain its full functional properties, in particular for the processing of a large corpus of complex speech sounds, in the infant as well as in the adult. Lack of or reduced auditory stimulation, for example, after repetitive otitis media in the young infant greatly impairs infant language development (Needleman 1977). Speech production and processing are highly dependent on the auditory perceptual capacities which are also a function of the nature of the environment. Several adult studies (Mackain et al. 1980; Werker et al. 1981) have shown that the ability to distinguish phonetic categories is directly related to specific linguistic experience. Well known examples are given from adults' incapacities to produce, and beforehand to identify some language specific features (i.e. Japanese indiscrimina tion of (R) and (L) (Miyawaki et al. 1975). Werker and Tees (1984) have demonstrated that the ability to discriminate lin guistic contrasts declines within the first year of life and that its maintenance depends on the language experience during this period. This can explain earlier results obtained in cross-linguistic studies describing differential speech discrimina tive capacities in 2- to 6-month-old infants (Lasky et al. 1975; Streeter 1976; Eilers et al. 1979). The interpretation of the data was controversial (Mackain 1982); how ever, the experiments deserve to be extended. These results are usually interpretable either as a reduction as a function of age of our auditory discriminative capacities or as the result of a high specification or crystallization of the system for the specific sounds that have stimulated the auditory system. The effects of the acoustical environment on central speech sound processing have also been emphasized in neuro-physiological studies on interhemispheric functional differences. Tsunoda's first results (1971) have opened a new research area in which converging data (Ten Houten 1976; Scott et al. 1979; Tsunoda 1985; Nachsohn 1986) show that speech acoustic features are processed differently ac cording to the acoustic structure of the language, and thus, that exposal to a par ticular language sound environment might orient cortical speech processing. This particular effect probably involves other higher, nonauditory factors, like cognitive processes related to speech. However, all these human observations, very briefly
44
J.P. LECANUET, . GRANIER-DEFERRE AND M.-C. BUSNEL
mentioned here, and unfortunately mostly conducted on adults, are nevertheless in teresting to relate to the animal experiments cited above that have revealed the plastic properties of the auditory system, in particular during its development. It is now necessary to design experiments which would allow one to accurately assess the role of prenatal speech sound experience on the development of language perception competencies. Does this experience only contribute to trigger innate abilities, or does it also participate in a gradual perceptual acquisition process similar to the one hypothesized by Jacobson's model (1968) for the infancy period? Prenatal experience might functionally shape some discriminative potentialities of the auditory system to the characteristics of the sounds that will be relevant during postnatal life.
Initial Equipment for Speech Perception J. Bertoncini Laboratoire de Sciences Cognitives et Psycholinguistique École des Hautes Études en Sciences Sociales Centre National de Recherche Scientifique, Paris
Research into language acquisition has traditionally paid little attention to the problem of perception owing to a lack of appropriate techniques for demonstrating its functioning in infants. However the development of such techniques along with a new theoretical conception of language development (Chomsky 1975) have al lowed research in this area to progress considerably in recent years. After more than ten years of research into speech perception in young infants (from 0 to 6 months), it is now possible to give an initial account of the results. The nature of the mechanisms of perception that allow the infant to begin pertinent processing of the sounds of speech are not yet clearly known. However, the ex perimental data that have been collected bring out the enonnous variety of the newborn's abilities, particularly in sound discrimination. The different kinds of dis criminative abilities vary greatly, and, in general, are similar to those that have been observed in adult speakers. Thus, very early on, newborn infants can dis criminate most of the phonetic contrasts occurring in natural languages. Current re search is turning to more complex questions such as the perception of certain phonetic dimensions in a variable context. Again, young infants have shown a sen sitivity to parameters such as syllable length and structure comparable to that of adults. This suggests that there exist specialized mechanisms for processing speech sounds that are capable of functioning without prior exposure to the contrasts being studied. However hypotheses concerning the specific mode of perception depend ing on innate mechanisms (Eimas 1975a) has proved controversial. The existence of specific basic perceptive mechanisms remains questionable as the discriminative performance that has been observed in the newborn may depend on general proper ties of the auditory system which themselves function from birth (Jusczyk 1981).
46
J. BERTONCINI
The above controversy may be settled by the discovery of abilities which go beyond simple discrimination. In order to achieve this, however, it will be neces sary to develop experimental procedures that are better adapted to testing the exist ence of higher processing levels in infants less than six months old than those that exist at present. Some researchers are currently interested in the categorization of sounds. This process is just as necessary as discrimination, for both speech acquisi tion and phonological development. The determination of the nature of categories actually used by young infants would be a great step toward the explanation of lan guage processes, their initial-state characteristics and their sensitivity to environ mental influences.
Initial Perceptive Equipment Well before babies can produce certain sounds themselves, they exhibit percep tive abilities that already reflect speech-signal processing. These abilities have been considered to be innate in that they are not the result of some linguistic experience and are therefore of special interest. First of all, they are an essential component at the start of the acquisition process. Second, they may explain certain characteristics of perceptive mechanisms in the adult. Knowledge of these initial capacities might thus help in determining the intrinsic properties of the speech perception system. What we call initial equipment includes those abilities that function - and that can be exhibited - as of the first weeks of life, as well as other fundamental properties of the speech processing system that will emerge when the organism has reached a more advanced state of maturation. It is known that the auditory system is capable of functioning already several weeks before birth. It is possible that it begins functioning in utero. Research on prenatal audition in humans has developed recently with the help of modern techni ques which can record physiological parameters (Lecanuet, Granier-Deferre and Busnel, this volume). But given that the transmission of acoustic signals in utero is somewhat special in that it occurs in a liquid medium, it seems likely that certain properties in auditory functioning are not activated until birth. During a postnatal period of several months, the newborn's auditory thresholds are generally higher than those of adults, although it must be noted that they are difficult to determine precisely in very young infants (Trehub, Schneider and Endman 1980; Sinnot, Pisoni and Aslin 1983). However, they seem to be differen ces of degree rather than of nature (Schneider, Trehub and Bull 1979).
INITIAL EQUIPMENT FOR SPEECH PERCEPTION
47
Most of the research on the basic perceptive equipment for processing speech signals is concerned has been carried out on the newborn's and very young infant's ability to discriminate. This is mainly due to the available experimental techniques. Most of these are based on habituation, characterized by a decrease in response rate following the repetitive presentation of a stimulus. After a period of habituation with one stimulus, the presentation of a second stimulus elicits a recovery in response rate, which, according to the accepted interpretation, shows that the sub ject has perceived the difference between the two stimuli, i.e. that s/he has dis criminated between them. Many response types have been used: heart rate deceleration, visual fixation time and high amplitude sucking (Siqueland and DeLucia 1969). The latter has been used the most because of its ease and the stability of emission at birth. At first, research into the newborn's discrimination of speech sounds focused on voicing, distinguishing between voiced stops /b, d, g/ and voiceless stops /p, t, k/. This phonetic contrast was the focus of many studies on adults (Liberman, Cooper, Shankweiler and Studdert-Kennedy 1967), and appeared to be perceived by them categorically1. The study by Eimas, Siqueland, Jusczyk and Vigorito (1971) marked the beginning of research on infant speech perception to determine the origin of the perceptual mechanisms that had been suggested for adults. The results of this study show that 1- and 4-month-old infants discriminate two stimuli that vary on the VOT (voice onset time) only if they belong to two different categories, identified as /ba/ and /pa/ by adults. Other stimulus pairs, separated by the same distance on the VOT continuum, but from the same category, were not discriminated by the in fants. The authors concluded that the mechanisms underlying the perception of VOT were innate and that they were part of the basic biological equipment of the organism. Furthermore, the infants seemed to perceive this dimension in a way similar to adults since their discrimination was categorical. This particularity was interpreted as indicating the mode of perception that is specific to speech and func tional as of the first weeks of life. The same type of results was later replicated for other phonetic dimensions such as place-of-articulation that distinguishes sounds like /ba/ and /ga/ (Morse 1972; Eimas 1974). Once the ability of newborns to discriminate between variations in the initial segments of CV syllables was established, medial and final position variations in mono- or bisyllabic sequences were tested (Jusczyk 1977; Jusczyk and Thompson 1978). Many of the experiments were carried out on stop consonants as the acous tic correlates of voicing variations or place-of-articulation were better known. Other experiments have shown that newborns discriminate certain dimensions of frica-
48
J. BERTONCINI
tives (Eilers, Wilson and Moore 1977), and that they can distinguish between two sounds such as /ba/ and /ma/ that differ in their manner of articulation (Eiman and Miller 1980a). We now know that infants are capable of perceiving a large variety of phonetic contrasts long before they are able to produce them. No minimal period of ex posure to a given language seems to be necessary to bring discriminatory ability into play. The same discriminatory performance that has been found in infants of several months of age has been found in newborns (Bertoncini 1982). Moreover, newborns can discriminate certain phonetic contrasts which are not present in their linguistic environment (Eilers, Gavin and Wilson 1979; Lasky, Syrdal-Lasky and Klein 1975). This capacity to discriminate foreign language contrasts seems to be reduced by the end of the first year (Werker and Tees 1984). The mechanisms necessary for voice recognition, especially the mother's, must also be part of the baby's basic equipment for the perception of species-specific signals (Mills and Melhuish 1974; Mehler, Bertoncini, Barrière and Jassik-Gerschenfeld 1978); this ability has even been found in 3-day-old infants (DeCasper and Fifer 1980). More recently, some experiments suggest that discrimination of native language may also be part of the basic equipment, since not only 2-month-old infants but also 4-dayolds exhibited differential behavioral responses when listening to fragments of their own linguistic environment (Mehler, Lambertz, Jusczyk and Amiel-Tison 1986; Mehler, Jusczyk, Lambertz, Amiel-Tison and Bertoncini 1988). To summarize, initial equipment includes perceptual mechanisms necessary for the pertinent discrimination of most acoustic characteristics of natural languages. Moreover the discrimination of certain phonetic dimensions is categorical; all varia tions are not perceived in a unifonn manner. Thus, from birth, perceptual mechanisms show particular sensitivity to certain regions of the acoustic continua underlying phonetic variations. These perceptual mechanisms are part of the innate equipment that allows infants to begin processing the acoustic speech signal very early.
Specialized Basic Equipment? The fact that discrimination was categorical for certain phonetic dimensions was interpreted as the result of specialized mechanisms (Eimas et al. 1971; Eimas 1974). Following this finding researchers gave more direct attention to the specificity of perceptual mechanisms by comparing speech perception to nonspeech sound perception. For adult subjects, the categorical perception of consonantal
INITIAL EQUIPMENT FOR SPEECH PERCEPTION
49
phonemes was contrasted with the continuous perception of vowels and nonspeech sounds. In most experiments nonspeech sounds were created by presenting only the portions of speech patterns that distinguished two syllables (e.g., transitions of F2 and F3 in /ba, ga/, Morse 1972). These signals were perceived by adults as nonlinguistic, and the discrimination functions did not show discontinuities (Mattingly, Liberman, Syrdal and Halwes 1971). For children discrimination was not categori cal, either (Eimas 1974, 1975b). In addition, for infants as for adults, vowel percep tion appeared to be continuous (Swoboda, Morse and Leavitt 1976). These results suggest that there are two distinct modes of perception, one acoustic and the other using phonetic categorization reserved for speech. Two problems linked to the nature of the speech signal have led to the idea that speech perception requires specialized mechanisms. The first concerns the absence of any simple correspondence between acoustic and phonetic variations in the speech signal (Liberman et al. 1967). The second arises from the need to segment the spoken string into discrete units, which is indispensable for further operations in speech processing. Of the different models that have been proposed, the one based on the existence of specialized detectors for phonetic features (Cutting and Eimas 1975) has the advantage of accounting for both adult and infant performance by its use of common automatic processes. Categorical perception was used in experiments of selective adaptation (Eimas and Corbit 1973) and supported this type of model in that it seemed to be specific to speech perception. However, recent studies have shown that categorical percep tion does not reflect a specific process since it is found with complex nonspeech sounds presented to adults (Cutting and Rosner 1974; Miller, Weir, Pastore, Kelly and Dooling 1976; Pisoni 1977) and infants (Jusczyk, Rosner, Cutting, Foard and Smith 1977; Jusczyk, Pisoni, Walley and Murray 1980). Also some experiments on the perception of speech sounds in animals should be noted (Kuhl and Miller 1978). These have suggested that categorical discrimination does not necessarily imply that specialized perceptual processes for speech are brought into play. In any case, these results do not exclude the possible existence of a specialized mode of perception (Liberman 1982). Furthermore, if categorical perception does not apply exclusively to speech, it may, at least, be an efficient process for trans forming speech sounds into phonetically processable sequences. For authors like P. Jusczyk (1982), the findings in favour of a phonetic mode of processing that is functional at birth are not convincing. We believe that it would be necessary to take up more complex questions that go beyond simple discrimina tion in order to bring out the initial abilities more specifically linked to speech processing.
50
J. BERTONCINI
Accumulated findings on discrimination have clearly brought out the variety of a newborn's abilities, but they have not solved the problem of specificity. The task the subject is asked to carry out does not necessarily require processing that goes beyond the immediate detection of some difference. Many of the results can be ex plained by acoustic processing of the cues present in the signal (Stevens and Blumstein 1981) without it being necessary to postulate a distinct mode of percep tion. Furthermore, the available experimental procedures for testing subjects only a few weeks old provide results of an all-or-nothing type. The behavioral responses studies, e.g. dishabituation, do not directly reflect, for instance, the level of discriminability of different stimuli. It should be noted, however, that studies on discrimination have progressed greatly. For example, Miller and Liberman (1979) have shown that the location of the phonetic boundary between /ba/ and /wa/ depends on the duration of the syll able in relation to speaking rate. The effect is not the same if the syllable duration is modified by the lengthening of the vowel or by the addition of transitions of a final consonant. This same type of phenomenon is found in infants (Eimas and Miller 1980b); they are sensitive to contexts that combine parameters such as dura tion and syllable structure. In the same way, our own work is aimed at determining the relation between the initial perceptual system and the later perception of the integrated properties of speech sounds. What we are interested in is whether the perceptual system shows a special predisposition for processing the segmental units of the speech chain, such as syllables. We have therefore specifically studied the properties that may make the syllabic unit more directly accessible to processing by the infant. We have found that two CVC sequences are discriminated more correctly than two CCC se quences for equivalent variation in constituant order. Inverting initial and final con sonants in syllable strings (C1VC2 vs. C2VC1, e.g., /tap/ vs. /pat/) elicits a strong reaction in the majority of 3- to 8-week-old subjects, while a similar change in nonsyllabic strings (ClCxC2 vs. C2CxCl, e.g., /tsp/ vs. /pst/) does not elicit equivalent discriminatory reactions (Bertoncini and Mehler 1981b). In another experiment, carried out in collaboration with S. Blumstein (Blumstein et al. personal communication), we have shown that stimuli synthesized from the first 20 milliseconds of CV syllables (cf. Blumstein and Stevens 1980) were dis criminated by newborns when the place of articulation of the initial consonant (/ba/ vs. /ga/) or the quality of the following vowel (/ba/ vs. /bi/) was varied (Bertoncini, Bijeljac-Babic, Blumstein and Mehler 1987). This suggests that the acoustic cues contained in the beginning of the signal are sufficient to discriminate each of the two constituants of the syllable and that this integrated information specifies the
INITIAL EQUIPMENT FOR SPEECH PERCEPTION
51
CV sequence as a unit. Furthermore, the newborn's perceptual mechanisms may operate directly at this level of segmentation. Thus, the newborn seems to have at its disposal a natural system that is capable of processing certain types of verbal information answering the needs of speech analysis at the acoustic and phonetic levels. The newborn's perceptual system al ready includes certain properties that favor a schema of speech analysis. It can be considered that adult processing, too, shows some traces of these fundamental properties, as the syllabic unit plays an important role in phonetic analysis and lexi cal processing (Mehler 1981; Mehler, Dommergues, Frauenfelder and Seguy 1981). One must now question the specialization of perceptual mechanisms by studying the newborn's categorization capacities. Indeed speech perception does not only in volve discriminating phonetic contrasts. It also implies recognizing similarities among certain elements in the speech chain. These elements, described as invariant at the phonetic level, conceal potentially discriminable acoustic variations. Can in fants, then, ignore these variations when they are not pertinent to adequately decod ing phonetic speech? Perceptual constants in speech sounds have been studied very little. One must take special note of the work of P. Kuhl and her colleagues. Their results pertain mainly to vowel perception and show that infants less than 6 months old can main tain their discrimination of vocalic elements (/a/ vs. /if) despite variations in pitch contours (Kuhl and Miller 1982) and in the speaker's voice (Kuhl 1976). It seems, then, that infants can tolerate, to a certain extent, considerable physical variations in the cues of these phonetic categories. As concerns the nature of the categories employed by infants in processing, one can refer to the pioneer work carried out by Fodor, Garrett and Brill (1975) on 4month-old subjects. Their results on head-turning show that infants respond more easily to a pair of stimuli that share one phonetic element (e.g., /pi/ and /pu/), than to a pair based on acoustic similarity (initial burst frequency as in /pi/ and /ka/). However, mainly for methodological reasons, these results do not prove conclusive ly that infants group elements in terms of phonetic categories. This type of experiment helps in to judging the respective importance of the dif ferent elements of a sound sequence and their role in the categorization process. Thus young infants' categorization of various phonetic and/or acoustic elements can be compared. If infants categorize phonetically similar but acoustically dif ferent elements most easily, this may mean that specialized procedures have been employed. In a phonetic categorization task infants are presented with several different stimuli that share a given phonetic element during habituation. The test consists of
52
J. BERTONCINI
introducing a novel stimulus containing (or not) this same phonetic element. The hypothesis is that the introduction of a novel stimulus containing the given phonetic element does not elicit a dishabituation response, since the presence of the common element brings about categorization. Thus, in a highly variable acoustic context we can attempt to determine whether the infant first separates out those elements that have a linguistic (phonetic) value. This procedure can be used to study whether an infant is capable of maintaining discrimination of certain phonetic dimensions in spite of massive acoustic varia tions, like those due to the speaker's voice, speaking rate or pitch contour. The in fant would thus be confronted concurrently with several variations that s/he can potentially detect when any one of these variations is presented singly. The specialization of perceptual mechanisms can then manifest itself if different "weights" are attributed to the different variable cues of the signal. When a single variation is to be tested, within the limits of its perceptibility, a discrimination response is normally obtained. When several variables are associated in the set of stimuli that is presented, the system must make a choice. If no predisposition guides this choice, any detectable variation can trigger discrimination. If there is specialization, then only some variables will be "amplified" and treated at the ex pense of those that do not have linguistic value. More precise determination of the specialization of this system would help to overcome some of the obstacles in the studying the automatic recognition of speech. In recent experiments carried out in collaboration with P. Jusczyk, a modified version of the "High Amplitude Sucking" procedure was used in order to test 4day-old and 2-month-old infants' ability to categorize in phonetically varied con texts. The question was whether infants can extract and use the phonetic similarity between syllable components to categorize and represent spoken syllables. Infants were habituated with a randomized set of different CV syllables that shared a phonetic segment, either initial consonant (e.g., /bo, ba, bi, be/) or following vowel (e.g., /bi, si, li, mi/). Then, a new item was added to the initial set. If infants tend to represent syllables in ternis of common phonetic segments, the addition of a new item should be responded to differently depending on whether it contains the same phonetic segment as the already presented syllables. For instance, the addition of /di/ into the /bi, si, li, mi/ set may induce less reactivity (e.g., less dishabituation) than an entirely new syllable such as /da/. In fact, the results do not support the hypothesis above in terms of phonetic segments. In these conditions, 4-day-old neonates and 2-month-old infants did not react in the same way. While 2-month-
INITIAL EQUIPMENT FOR SPEECH PERCEPTION
53
old infants exhibited a significant response of dishabituation to the addition of any new item, 4-day-old neonates demonstrated sensitivity only to a change in vowel categories. Neonates did not react to the addition of /di/ in /bi, si li, mi/ nor to the addition of /da/ in /bo, ba, bi, be/. However, they did manifest dishabituation when the new item was /bu/ in /bo, ba, bi, be/ or /ma/ in /bi, si, li, mi/. In contrast, 2month-old infants detected any novelty, even a change of the consonant only or the vowel only in either context (for more details, see Bertoncini, Bijeljac-Babic, Jusczyk, Kennedy and Mehler 1988). Thus, there is no evidence from either age group that the representation of syllables was structured in terms of phonetic seg ments. Moreover, the results are compatible with a view that infants encode only global differences among the syllables. Thus, we have no evidence that infants' processing of spoken syllables involves some kind of phonetic analysis and, therefore, requires some "phonetically" special ized mechanisms. Before concluding this review of the questions concerning specificity, we must mention research being carried out in neuro-psychology. This focuses on interhemispheric asymmetry to determine whether or not the perception of linguistic material by babies is mainly sustained by the left hemisphere as in adults. The results, based on electrophysiological indicators (Molfese and Molfese 1979) or on behavioral responses (Entus 1977; Bertoncini 1982), seem to indicate greater invol vement of the left hemisphere. This may mean that the left hemisphere is predisposed to process certain analytic aspects of speech sounds.
Conclusion For adult speakers speech perception is most likely automatic and obligatory; thus under normal conditions early perceptual processing operations are still brought into play at the highest level of linguistic interpretation. For infants begin ning to apprehend speech signals, certain levels of processing are beyond their capacity. Thus, what must be determined are the limits of the first stages of percep tual processing, what type of information is processed and whether infant speech processing requires specialized mechanisms. We have seen that newborns respond to phonetic distinctions among syllabic elements. They can also process information about the structure and length of a syllable. These skills allow them to distinguish, very early, most of the charac-
54
J. BERTONCINI
teristic dimensions of the strings they hear. Moreover, it appears that newborns can recognize certain similarities between speech sounds while ignoring nonpertinent acoustic variations. Their perceptual capacities allow them to embark on complex speech sound processing from the first days of life. However, the question remains: Are these skills found in the perceptual system specific to speech processing or not? The search for specialized mechanisms must be based on comparisons between the perceptual performances of newborns and adults on specific aspects of the speech signal. Hitherto, the results obtained for categorical perception and for hemispheric asymmetry have proved inconclusive in determining the nature of the mechanisms that are brought into play. However, the systematic comparison of adult and infant performance has brought out certain similarities both in the nature of discriminatory skills, and the conditions in which they appear. It might then be that identical mechanisms are involved. For example, the initial perceptual system might have a tendency to process a particular type of verbal unit such as the syll able. The properties of the initial system would thus favour a speech analysis design that would continue to be the most efficient, even after more elaborate processing systems have developed. Research into the specificity of processing is now focusing on the skills that allow infants to extract pertinent information from a variable acoustic context. Ex perimentation should now be pursued on the discrimination of complex phonetic dimensions and categorization, as well as on the memorization of pertinent phonetic attributes. Research on the relations between perception and production in speech acquisi tion is much needed. Discriminatory ability is highly developed and functional long before the child begins to babble. However, it may be that perceiving certain phonetic parameters in a particular language becomes more refined as production ability develops. Sensitization enhances perceptual mechanisms, allowing them to respond to more precise values in a given acoustic dimension. Certain charac teristics of the language in the environment appear in the production of 6- or 8month-olds (cf. . de Boysson-Bardies, this volume). This data should provide supplementary indications as to the critical type of information that infants are able to extract from their linguistic environment. Studies on phonological development, data on categorization abilities and on the beginning of production for segmental aspects of speech, should provide some indications concerning the acquisition of phonological rules.
INITIAL EQUIPMENT FOR SPEECH PERCEPTION
55
NOTES 1. Perception is called categorical when discrimination is highly correlated to identification. Areas of the continuum where there is highest discriminability correspond to the boundaries, between categories defined by identification.
Target-Language Influences on Prespeech B. de Boysson-Bardies Laboratoire de Psychologie Expérimentale École des Hautes Études et Sciences Sociales Centre National de Recherche Scientifique, Paris
Diversity is one way of betting on possibilities. F. Jacob
Even if we accept that infants do have some linguistic competence, it must be recognized that speech is not their typical vocal performance. It is widely admitted that language learning is biologically determined. Thus lan guage should be a well-structured system with constraints on how it develops. Given the diversity of natural languages, the system has to be open, i.e. sensitive to the contingencies of the linguistic environment. For speech acquisition to take place, the developmental process must enable the infant to select and combine the units or segments that are necessary to process and produce a given language's relevant subset of units. When and how do the processes that give special importance to the relevant parameters determining the background language begin to influence infants' be havior? How and at what point in the developmental process is the phonetic struc ture of speech coded as language-specific information? Answering these questions is an important goal for studies on speech acquisition. More specifically, we must find out if the prespeech stage is involved in speech acquisition. Various and some times incompatible hypotheses have been offered to interpret the forms and func tions of so-called prelinguistic productions. From a phylogenetic point of view, it is generally admitted that perception and articulation develop in a compatible and complementary fashion. The very early ability of newborns to discriminate and then categorize speech sounds has now been well established (Eimas, Siqueland, Jusczyk and Vigorito 1971; Eimas 1974; Jusczyk 1985), although the underlying processes are still being disputed. At about
58
. DE BOYSSON-BARDIES
7 months, infants can disregard nonrelevant variations such as speaker, pitch or context to categorize segments phonetically (Kuhl 1979, 1980). These discrimina tions reside on preprogrammed functional bases since no differences have been found in the discrimination performances of 3- to 6-month-old infants from dif ferent linguistic communities, whatever the specificities of the phonological system of the different languages. However, this universal behavior is soon modulated by experience. Recent experiments show on the one hand, that 6-month-old infants are sensitive to the relevance of prosodic cues in their target-language and not in foreign languages (Jusczyk, personal communication), and on the other hand, that the general discriminative capacities of newborns are channelled at around 9-10 months of age when they become more specific, ans where only the contrasts that are phonologically relevant in the parental language are processed (Werker and Tees 1984). The capacity to process all phonetico-acoustic contrasts partly vanishes at around 12 months. During the first year of life linguistic experience permits in fants to select those parameters that are distinctive in their target language. Further more, word comprehension is evidenced (Lewis 1963) before the time children utter their first words. The process of setting relevant parameters is dependent on (among other things) perceivability, reliability, accessibility and the functional role of the different cues in the speech of the environment. The necessary flexibility of the system may allow for different paths in processing strategies to be followed. This explains the variability that has been found among infants of the same linguistic community (Ferguson 1979, 1986; Vihman, Ferguson and Elbert 1986). However, variability in prelinguistic productions among infants from the same linguistic background can be explained by endogenous factors. It could be ascribed to random variations depend ent on immanent factors and leading to individual differences. Therefore, individual variations only weakly challenge the position stating that babbling productions reflect general proclivities that are independent of perceptual input. One way to settle the debate concerning the relation between perception and production in speech development would be to show that characteristics of the lin guistic background may influence the structure of infants' productions during the babbling stage. If target-languages were found to influence prespeech productions, it would entail that auditory forms are already being selected and that auditory/articulatory forms are being mapped to motor command. Although the mechanisms and processes of speech perception are fundamentally different from those of production, it is obvious that humans are equipped with the ability to translate auditory information into information to be used by the motor system for articulation. There exist practically no experimental data on the relation
TARGET-LANGUAGE INFLUENCES ON PRESPEECH
59
between perception and production in infants. Experiments on adults emphasize the short amount of time required for "translation" to take place (Chistovich, Lissenko and Fedorova 1972; Chistovich 1983) suggesting that audio-motor bases for speech are innate. Informal observations on sound reproduction or speech gestures in new borns also suggest that the audio-motor bases of speech are at least partially innate. However, to say that a capacity is innate is not to say that it is not subject to a maturational or developmental process. Different theories have been proposed to account for the relation between speech perception and production in infancy. The "Independence hypothesis" postulates that motor and perceptual processes for speech develop independently during the first year and that therefore babbling1 productions reflect only the ease of articulation that has been channelled by matura tion and motor development (Studdert-Kennedy 1986). Consequently, these produc tions are not influenced by perception and reflect articulatory proclivities that are the same for all infants independent of the linguistic environment. This exercise al lows infants to master those articulatory patterns from which they will later choose their first words. The first-word stage is considered to indicate that perceptual and motor components are beginning to be linked into a perceptuo-motor system for speech. Another theory links the perceptuo-motor system for speech to the onset of com prehension processes. Simple experience with a target language is not considered to be sufficient to induce language specific phonemic representation; the convergence of sounds and meaning is necessary (MacKain 1982; Jusczyk 1985, 1986). In this case, the influence of linguistic background is considered to begin before the onset of first words since some word comprehension is already evident at this preword producing stage. The "Interactive approach" denies that there is discontinuity between prelinguistic and linguistic behavior and advances that "there is probably a predisposition towards speech-like characteristics on the production side... Such a predisposition would provide the infants with a basis for articulatory and auditory configurations" (Macken 1986). Both perceptual capacities and capacities to produce syllabic strings and intonational patterns are necessary prerequisites to speech mastery. During the first year of life infants already actively engage in important linguistic behavior (Ferguson 1979; Menyuk 1971; Menn 1976, 1983) that allows them to ac tively structure input and output (productions). The "Competition model" assumes that "articulatory structure is mapped right off the auditory structure and that articulatory learning focuses not on the acquisi tion of representations, but on a system for converting auditory/articulatory repre-
60
. DE BOYSSON-BARDIES
sentations to motor commands" (MacWhinney 1987). These positions could predict an early influence of perception on production. Our position is that speech development is partly dependent on a computational process which enables infants to perceptually select what is relevant from the input. Articulatory settings are influenced by this selection. These processes begin in per ception before the last quarter of the first year (Werker and Tees 1984) and thus can be reflected in production even before the first word is produced. In accordance with such a point of view, the study of speech development should use cross-linguistic studies to specify in prespeech: - which invariants define innate mechanisms or principles - when speech sounds begin to reflect the process of coding speech-specific in formation - how interaction between phonetic and intonational characteristics of speech can be exemplified in prespeech. This study we present (de Boysson-Bardies, Sagart and Durand 1984; de Boysson-bardies, Sagart, Hallé and Durand 1986; de Boysson-Bardies, Sagart and Hallé 1987; de Boysson-Bardies, Hallé, Sagart and Durand 1989) consists mainly of an attempt to determine through a comparative approach if, when and how the prosodic cues and/or the segmental structure of the target-language influence bab bling patterns.
Comparative Studies Phonetic and intonational structures as well as natural language timing can differ a great deal. Varying information in terms of different target languages could help to separate out, on the one hand, the general tendency governing maturational processes and, on the other hand, the variability resulting from specific phonetic and intonational characteristics of the target-language. However this approach poses a number of problems. There are few valid homogeneous phonetic descriptions of the prosodic aspects of different languages. Furthermore, because of the lack of unifonnity in the description of acoustical data and the lack of precise knowledge concerning the perceptual value of these nonsegmental parameters, we are only rarely able to make a direct comparison between adult speech and babbling. Thus, it is extremely difficult to set up comparison scales with which to assess the degree of confonnity of the nonsegmental patterns found in babbling with the nonsegmental patterns of the target language.
TARGET-LANGUAGE INFLUENCES ON PRESPEECH
61
On the other hand, although the study of segmental patters in babbling could ap pear as more accessible, due to the greater availability of studies on the sound sys tems of natural languages, the inherent difficulty in transcribing the vocal productions of babies (Stockman, Woods and Tishman 1981) is a serious obstacle to such an approach. In fact, there are no good experimental methodologies to study early productions. In order to evaluate the hypothesis that infant babbling reveals characteristics of the environmental language, we looked for systematic differences in the production of infants from different linguistic backgrounds in the light of the specific charac teristics of the target language. We carried out two types of studies: first we made experiments based on adult judgement in a babbling discrimination task and secondly, we made acoustic and perceptual analyses of prelinguistic productions and then compared them to hypotheses drawn from phonetic analyses of the targetlanguage.
Target-Language Influence on Prespeech: Adult Judgements This approach includes experiments on adults' perceptual judgments. Adult sub jects were asked to judge babbling as compatible or incompatible with the prosodic and phonological characteristics of their own language. The ability of adults to per form this discrimination task implies that some perceptual correlates of the speech dimensions of a given language can be found in the vocal productions of infants. Three factors were controlled in the application of this paradigm. The first point concerns the characteristics of the target language. We chose to maximize the number of dimensions on which the languages could differ. For this purpose French, Arabic (Maghreb dialects) and Cantonese were selected. These languages differ in the following ways, among others: a. Phonation type: tense voice is characteristic of Cantonese production whereas lax voice is common in French. b. Laryngeal and supralaryngeal settings: Arabic is characterized by a rather posterior tongue setting as opposed to fronted tongue setting in French. Prosody: languages with a stress system based on the phonological properties of words offer different rhythmical patterns compared to languages such as French which only have demarcative stress at the end of the sequence. In con trast, the Arabic stress system relies on syllabic weight (Hyman 1977). d. Tone system: Cantonese differs from French or Arabic in that it is a tone lan guage.
62
. DE BOYSSON-BARDIES
e. Phonetic repertory and syllabic structure: the phonetic repertory of Maghreb Arabic differs mainly from French and Cantonese by the larger proportion of consonantal sounds in comparison to vocalic sounds. f. Syllabic format and mode of onset and release of syllables. The second point relates to constraints on data collection. Given the high sen sitivity of infants to the sounds of language, all the data were collected in the infants' countries of origin. In our study the babies were audio recorded in Paris, Hong Kong and Tunis. The third point relates to the choice of age groups: 8- and 10-month-old infants were chosen as respectively exemplifying different stages in the control of intona tion and of CV productions (Oiler 1980).
Experiment 1 In the first experiment French adult judges were asked to discriminate between vocal productions of 8- and 10-month-old infants. The judges were presented with excerpts from babbling productions of 8 French children (four 8-month-old and four 10-month-old infants) and 8 non-French children (two 8-month-old Tunisians, two 8-month-old Cantonese, two 10-month-old Tunisians and two 10-month-old Cantonese). Samples were presented in two separate blocks depending on the age of the infants. Each block included 8 pairs of samples; each sample consisted of babbling by a French baby and by either an Arabic or Cantonese baby. Forty French adults (thirty-two women and eight men, among the women six teen had children) participated in the experiment. The judges heard 16 pairs of samples: eight pairs from the 8-month-block and eight from the 10-month block. Within a block, the order of presentation was counterbalanced across subjects. The adults were asked to judge which sample in a pair they thought came from a French infant. In this first experiment judges were not informed of the specific linguistic background of the non-French children. Table 1 indicates the percentage of correct identifications of French infants per age block and per language. An analysis of variance shows that the two main factors, linguistic community and age, are each significant. Interaction of the two variables is not significant. Computation for each pair shows that for 7 pairs out of 8 involving French/Tunisian babies, adults were able to correctly identify the French child. The same identification score is found for 3 out of 4 pairs in French/Cantonese com-
TARGET-LANGUAGE INFLUENCES ON PRESPEECH
63
Table 1 Child's age Comparison
8 months
10 months
French-Arabic
75.8%
74.4%
French-Cantonese
69.4%
31.9%
Percentages of correct identification of the French sample in the French-Arabic and French-Cantonese comparison pairs (N = 40).
parisons for 8-month-old infants. However 10-month-old French children were not discriminated from Cantonese babies. These results show that, under certain conditions, French judges are able to make a correct choice in discrimination tasks involving babbling productions of children belonging to their own linguistic community.
Experiment 2 We tried to test the value of prosodic versus segmental cues in the identification of babbling. In the first experiment, according to the judges, cues provided by intonational patterns and timing organization of pitch and intensity variations seemed to offer an easier means of discrimination than segmental cues. Our selection criteria favoured the selection of long vocalizations with pitch patterns spreading over whole breath-groups. Some of the 10-month-old samples which consisted of sequences of well-articulated, reduplicated syllables were poorly discriminated. This type of babbling, frequent for 8- to 10-month-old infants, is generally poor in intonational patterns. For the second experiment, the 10-month-old infants' samples consisted ex clusively of sequences of isolated or reduplicated syllables with few intonation pat terns. Samples of 8-month-old infants consisted mainly of melodic vocalizations. Because it was not possible to collect additional data from Cantonese infants, the comparisons were made only for French and Algerian infants. We used the same experimental method as in the first experiment.
64
. DE BOYSSON-BARDIES
The percentage of correct choice for the 8-month-old infants is close to the per centage of correct choice in the first experiment. The difference between correct and incorrect choice is significant. This difference for 10-month-old infants is not significant. With 8-month-old infants, French judges' choices are mainly based on prosodic features. These features, which are more prominent in long productions with a low rate of articulation, permit the judges either to recognize the babbling productions of infants belonging to their linguistic community or to eliminate as alien the production of non-French infants. The most useful cues are phonation types, rhythm and pitch patterns. When listening to 10-month-old reduplicated syllables, adults say they try to look for cues that are related to actual words in their lan guage. As reduplication of syllables such as /pa/, /ba/ or /ma/ is frequent in the bab bling of infants, independent of linguistic background, this strategy makes it difficult for judges to make relevant discriminations. We thus decided to analyze acoustically the 10-month-old infants' productions with a high rate of articulation, to see if evidence of early differentiation of laryn geal or supralaryngeal settings can be found in spite of the ambiguous results in the second experiment. For this purpose two studies were conducted.
Long-Term Spectra The first study concerns the general adjustments necessary to define the settings of the vocal tract that are required by different languages. Although descriptive data in this field are seldom available, it is known that particular settings of the larynx, pharynx, velum, tongue and lips underlie variations of voice quality and articulatory principles in the different languages. Acquiring such settings could be preliminary and prerequisite to learning the short term muscular settings which un derlie the segmental inventory of a language. The distribution of the energy/intensity spectra is dependent upon differences in the configuration of the vocal tract (Fant 1960). Consequently we thought it would be interesting to see if spectra run on long enough portions of speech could reveal specificities due to voice quality and articulatory settings of different languages. A comparison of long-term spectrum (LTS) analyses of adults and infants from three different linguistic communities was run to see of 10-month-old infants attune their vocal productions to certain long term characteristics of speech productions in
TARGET-LANGUAGE INFLUENCES ON PRESPEECH
65
their target language. The claim of an early attunement of laryngeal and supralaryngeal settings will be reinforced if LTS patterns are stable across each language group and if there is a general similarity of mean adult LTS patterns with mean in fant LTS patterns within each language group. Adult speech samples and babbling samples were recorded in Paris, Hong Kong and Algiers. Adults in each language group were asked to read a text. French and Algerian groups included 10 males and 10 females. In Cantonese, 9 males and 8 females contributed samples. Six infants' samples were contributed. Mean ages are 10; 13 for French infants, 10; 10 for Cantonese and 10;8 for Algerian infants. For each sample of babbling, well articulated syllables were extracted and spliced one after another in the order of occurrence in the recording. For adults as well as for infants, LTS were run on the first 40 seconds of each sample. A reference pattern was constructed by averaging the LTS of adult speakers of the same sex and from the same linguistic community. Six reference patterns were defined in this way (3 language groups x sex). The distances of individual adult LTS to each of the six reference patterns were then calculated.2 Individual LTS were classified in the sex and language group according to the best approximation to the reference pattern. Clear language-specific patterns are identifiable in the reference patterns, both regarding the first harmonic (HI) and the general spectrum shape. In the classifica tion test, 55 out of the 57 adults were classified in the correct language/sex group. Consequently, reference patterns were taken as representing language specific templates which may be considered as relevant "prototypes" in the study of infant LTS. The same procedure as with adults was followed with infants. It was found that in spite of the fact that infants produce babbling instead of speech, the reference LTS patterns for each language group present striking similarities to corresponding LTS patterns for adults. French and Cantonese infants show a one-peak pattern. This peak is centered at a higher frequency (1200 Hz) for Cantonese than for French infants (825 Hz). A similar discrepancy was found between French and Cantonese adult patterns for which main peaks are respectively at 400 Hz and 500 Hz. A double peak is found in Algerian infants' patterns as in adults'. The results of classification tests show that infants' patterns tend to be grouped according to the language of the community. By using the same distances, as with adults, 13 out of 17 infants were classified in the correct language group.
. DE BOYSSON-BARDIES
Table 2
Classification of Adult LTS Subjects Males French
Classes French Males
Algerian
Cantonese
French
1
2
9
Algerian Cantonese
| Cantonese |
10 1
8
French Females
Females Algerian
8
Algerian
10 8
Cantonese
Table 3
Classification of Infant LTS Subjects: 10-month-old infants Classes
French
French
5
Algerian
1
Cantonese Correct classification
5/6
Algerian
Cantonese
4
2
1
4
4/5
4/6
In the spectra of Cantonese and French infants, we fail to find a contrast on the first harmonic that would support the hypothesis that infants attune their phonation type early on. The interpretation of LTS curves is made difficult by the indirect relations be tween speech sounds and spectrum shapes and by our lack of knowledge about the long term articulatory characteristics of the different languages. Moreover the rela tion between babbling and adult speech has not proved to be the same for each lan guage independent of settings and segment characteristics in the different languages.
TARGET-LANGUAGE INFLUENCES ON PRESPEECH
Figure 1
Average (reference patterns) of the French 10month-olds' LTS (N - 6).
Average (reference patterns) of the Cantonese 10month-olds' LTS (N = 6).
67
68
. DE BOYSSON-BARDIES
Figure 2
Average (reference patterns) of the French female adults' LTS (N = 10).
Average (reference patterns) of the Cantonese female adults' LTS (N = 8).
TARGET-LANGUAGE INFLUENCES ON PRESPEECH
69
Which characteristics of the speech in their environment do 10-month-old in fants then attune their vocal productions to? The high ratio of vowel frames to con sonant frames makes it plausible that to a large extent language specific differences in LTS shapes relate to vowel production. For different languages, vowel systems differ not only in the number of vowel phonemes and the detail of oppositions among them, but also in the dimensions of the articulatory space in which the vowels are inscribed. Setting up loose limits to tongue and lip movements accord ing to the vowel system of a target language could be the first step in the environment's influence on articulation.
Acoustic Study of Vowels Consequently, the aim of the following study was to determine if systematic dif ferences can be found among vocalic productions among infants from different lin guistic communities and if so, whether these differences reflect adult vocalic patterns. This point was tested by making an acoustic analysis of vowels produced by five out of the six French and Cantonese, 10-month-old infants from the first study, five Algerian (three of them from the first study) infants and five London-English infants. The acoustic investigation of infants' vowels was limited to the examina tion of the first two formants. About 50 vowels per infant were selected. They were taken from the infant's babbling in order of appearance, provided that they were acoustically tractable. The total number of analyzed vowels reflects the distribution of the vowels in the infants' productions. For each vowel, Fl and F2 were es timated from a 20 Hz resolution short-time spectrum. The results show systematic differences among mean frequency formants for vowels produced by infants from different linguistic communities. Mean Fl and F2 values and standard deviations, together with F2/F1 ratio per linguistic community are given in Table 4. The F2/F1 ratio characterizes the vowels on the compact-diffuse dimension and has the ad vantage of being effected very little by variations in individual vocal tract length. Computation of statistical distances between infants' vowel sets indicate that in fants differ more among different linguistic communities than within any single lin guistic community. Main tendencies of infant vowels can also be visualized by means of "mean vowels" defined by infants (mean Fl and mean F2 for a given infant) or per lan guage community (mean Fl and mean F2 for the infants of the community).
70
. DE BOYSSON-BARDIES
Table 4 English
French
Moroccan
Cantonese
Fl
876 (207)
878 (239)
976 (282)
1047 (241)
EZ
2628 (452)
2456 (439)
2341 (459)
2343 (381)
3.00
2.80
2.40
2.24
F2/F1
Mean formant frequencies and standard deviations (Hz) of infant vowels and F2IF1 ratios by language community.
Figure 3 Frequency of F2 (Hz)
Frequency of Fl (Hz)
F1IF2 plot of the "mean vowels" of each infant. Infant's linguistic group is indicated: E (English), F (French), A (Algerian) or (Cantonese).
Table 5
English
French
Moroccan
Cantonese
Infants
3.00
2.80
2.40
2.24
Adults
3.68
3.28
3.03
2.71
F2IF1 ratios of infant and adult vowels by language community.
TARGET-LANGUAGE INFLUENCES ON PRESPEECH
71
We have attempted to give a similar account of mean vowels and F2/F1 ratios for adult speakers in the four languages under scrutiny. For each of these languages we used two kinds of data existing in the literature: frequency count of a vowel on corpora of running speech (Fry 1947; Wioland 1972; Fok 1979) and Fl and F2 values for individual vowels (Henton 1983; Lonchamp 1989; Lee 1985). Since data for the Algiers dialect could not be found, we used frequency counts and formant values from Moroccan Arabic (Benkirane 1982). For each language, Fl and F2 of "mean" vowels were computed as the frequency count weighted by Fl and by F2 values from the acoustico-phonetic study. A comparison was run between the adults' values of mean Fl and mean F2 and on F2/F1 ratios and the infants' data. The same trends are found for adult speech and for babbling; English and French have more diffuse vowels than Cantonese and Arabic. The similarity between infants' and adults' vowels across the four linguistic communities is well illustrated by the parallel patterns of F2/F1 ratios found in in fant and adult vowels.
Conclusion Babbling forms present general characteristics dependent on age (Cruttenden 1970; Stark 1980; Oiler 1980). These general characteristics have too often masked clear differences, some of which must be linked to the characteristics of the target language. Our cross-cultural attempts to study prespeech development give support to the claim of an early influence of environment on prelinguistic productions. Rhythm and intonation patterns of 8-month-old infants present differences that are consis tent with the rhythmic and intonational characteristics of the parental language. On the segmental level, setting up loose articulatory preferential position for tongue and lip movements in 10-month-old vowel production is the first indication that the vowel system is being shaped by the target-language. This implies that the speech sounds in the surrounding language already provide material for building up internal representations which infants use to set up ar ticulatory principles for their target-language. However, more data are needed to better understand the ways and the steps by which interaction with the environment allows for the selection and mapping of dif ferent parameters during the prespeech stage. The influence of linguistic back ground on intonational and segmental patterns does not yet permit any claim on "phonological" representation. However, an investigation of the links between
72
. DE BOYSSON-BARDIES
prespeech forms and first words may help us to understand some of the roles played in prespeech by the different use of segments or by intonation cues to con vey meaning contrasts.
NOTES 1. The babbling stage begins at around 7 - 8 months and is characterized by the emergence of syllabic productions that exhibit acoustic timing constraints which are close to mature speech (Oiler 1986). 2. The distance between individual LTS and each of the six reference patterns were calculated using a Euclidian distance between arrays of cepstral coefficients.
Prosodic Cues in Very Young Children's Speech N. Bacri Université René Descartes Laboratoire de Psychologie Expérimentale École des Hauts Études en Sciences Sociales Centre National de Recherche Scientifique, Paris
The language of children in the age group between 18 months and two years is characterized by the presence of productions that can be easily interpreted by the adult, of others whose meaning can only be guessed at, and still others that can only be perceived as incomprehensible babbling. Nonverbal information, i.e. mimicry, gestures and the contextual situation, facilitates comprehension, but is not sufficient in itself to render all of these children's productions intelligible. Underlying speech intelligibility are the complex processes involved in the iden tification of phonetic and syllabic units and in the grouping of these units into se quences that adults can recognize as belonging to their language. Generally speaking, the identification of speech sounds depends on a categorization that is set up by a bottom-up processing of the spectro-temporal information contained in the signal and a top-down processing of language-specific information. However this very general model leaves open the question as to what quantity and what type of information is necessary to enable the listener to understand the speech of a speaker who is very different from him or her, i.e., when this speaker is a child. Child speech can be considered as a case of a vocal signal of rather poor intel ligibility, but also as a case where the voice dimension has precedence over the syntactic dimension. The utterances of children from 22 months to 26 months old are best characterized by the simultaneous presence of both babbling and initial lin guistic forms. Babbling might be defined, depending on the point of view, either as a production of syllabic sequences forming auditory chunk (Bacri and de BoyssonBardies 1981), or as "chains of repetitive articulatory movements during continued or interrupted phonation of a breath unit" (Koopmans van Beinum and van der Stelt 1979). Initial linguistic forms are recognized by adults as belonging to their own
74
N. BACRI
language because they manifest a sufficient degree of approximation to adult lan guage to be understood, although they often present highly variable phones. Peter son and Barney (1952) demonstrated the intrinsic variety of phones, and particularly of vowels, by using perceptual and physical data. It seems to be quite possible for an adult listening to an adult speaker to abstract the invariants from the spectral structure of a speech signal (Stevens and Blumstein 1981). Template matching processes have been offered to account for the rapid identification of con sonants. But the variability of children's productions is such (Ferguson 1986) that only general patterns of phonemic distribution can be constructed, but not, strictly speaking, phone classes. At 18 months old this variability is mainly due to children's insufficient control of the articulators as they pass from one articulatory gesture to another. Moreover, during the transition period between babbling and thoroughly intelligible speech, phonological development follows a set of complex rules (Dore, Franklin, Miller and Ramer 1976; Oller 1980; Menyuk, Menn and Sil ber 1986). All children do not set up and follow the same cognitive strategies, and several patterns of continuity have been proposed (Oiler and Smith 1977; Ferguson and Macken 1982; Locke 1983; Ferguson 1986; Macken 1987; MacWhinney 1987). Thus it is not surprising that child language intelligibility varies from child to child, for the same perceiving adult, when the variability of child productions is taken into account. Variability in the realizations of a "same" segment has also supported models of phonetic intelligibility focusing on perceptual integration and segregation (for an overview, see Repp 1987). As for adults, these functions might rest on a reference of perception to production (Liberman and Mattingly 1985). One could infer from this assumption that when listening to child speech an adult can construct a repre sentation of the characteristics of the child's vocal tract. But this cannot be carried out without some difficulty. Whether the reference of perception to production is maintained or not, this theoretical position has a two-fold interest. First, it brings to the fore the importance of the perceptual integration processes which analyze in a parallel fashion the numerous cues produced by a single articulatory gesture that are dispersed along a time dimension. Second, context relations can be set up for cues over a stretch of time covering two or three syllables. It is within this theoretical framework that the perceptual processing of child speech by an adult might be considered, not only as concerns segments, but as con cerns word chains. As a matter of fact, the immaturity of the mechanisms that govern the child's co-articulation leads to the production of integrated units that are larger than the syllable and whose boundaries are seldom well delimited acoustical ly. In this way, processing cues that are spread out over a long time span, e.g. in-
PROSODIC CUES IN VERY YOUNG CHILDREN'S SPEECH
75
tonation cues, might be integrated into a perception model. If segmental informa is not regular enough to permit, by itself alone, segmentation and sequenciation, or to be the base of phonemic and syllabic sequence identification, the presence of timing and pitch information in speech signals could explain how a se quence may be represented as an intelligible production, according to a listener's normalization process (Bacri 1986). However, it should be noted that intonation has only been given a secondary role, both in production and in perception, until these last few years. Of course some authors had already noticed that intonation was one of the first aspects of children's vocal behavior that shows a "phonetic drift", i.e. language specific fea tures (Lewis 1951), but little empirical work was carried out in view of confirming this observation (Menyuk and Bernholtz 1969; Oiler, Weiman, Doyle and Ross 1976; de Boysson-Bardies, Bacri, Sagart and Poizat 1981). There were two main reasons for this. First, it had long been thought that first language differed from babbling not only in that it appears at a time when phonologically controlled forms emerge, but also when the rich intonative contours of babbling disappear. According to Jakobson (1941), the undivided attention of the child who is just beginning to speak is concentrated on segmental organization. As the child's articulation becomes more precise, the intonative contours become more monotone. Strong intonative movements were considered to be a sign of im maturity of the vocal apparatus. Intonation was said to perturb articulatory precision until the time when articulation was well-established and intonation could acquire a linguistic function linked to the development of the child's abilities as to syntactic and lexical organization. During the period of transition between babbling and first language, intonation would lose its expressive function without for as much acquiring a new functional value (Weir 1966). Second, the study of pitch, as linked to fundamental frequency, i.e. the voice en velope, was neglected due to a very narrow interpretation of the "articulatory model". Prosodic cues, which are dependent on pitch and on the temporal structure of speech, did not seem, in the case of children, to be processed by mechanisms analogous to those that are put into play in the perception of speech sounds. In fact, the study of speech perception in children gave rise to a great deal of research into infants' initial equipment and into their capacity to discriminate speech sounds categorically (Bertoncini, this volume). But variations in intonative contours are, in this theoretical framework, considered as "noise" which can perturb this discrimina tion and the infant's ability to set up perceptual constants (Kuhl 1980). Thus in tonation is not linguistically, i.e. phonetically, pertinent. It plays no role in the intelligibility of speech as it does not correspond to any precise cues.
76
N. BACRI
Already in the 1970's, and particularly during the last ten years, the hypothesis of discontinuity between babbling and first language, as well as between intonative and phonetic cues, has been questioned and a new approach to the relation between perception and production has appeared. For example, the belief in a discontinuity between strongly intonated babbling and a first monotone language was called into question by studies of corpora that were collected over a long period of time. While some children did in fact show development that accorded with Jakobson's model, others continued to produce intonative contours even as their speech became intel ligible (Nelson 1975). Thus, studies on the functions of intonation as the child sets up speech processing mechanisms has developed and a new field of study has opened which investigates children's ability to communicate with adults in their en vironment (Crystal 1986; Cruttenden 1986). New importance has been lent to the young child's prosodic cues; their presence in the child's vocal productions has been studied with a view to learning (1) what their specific characteristics and their developmental pattern are, (2) how children analyze prosodic cues when they listen to adult speech, and (3) how an adult listener may use them when assessing the in telligibility of young children's speech.
Intonational Organization of Young Children's Speech A prerequisite for intelligible speech in children is the control of the fine coor dination of lips, tongue, vocal cords and respiratory movements with each other, along with those of the vocal tract as a whole. Control of the mechanism involves two kinds of commands: a "sharp pulse" applied to the laryngeal structure which generates a long term component, and "a stepwise force applied to another part of the laryngeal structure", which generates local movements (Lehiste 1984). Coor dination between these commands develops progressively over time. As infants ma ture, they acquire better glottal control and thus pitch variations and speech timing become structured. Already during the first weeks of life the cry repertory of an infant includes about 4 to 6 gross patterns which are recombined to form intonational contours. In fant vocalizations have been classified according to their auditory and spectrographic features, and in particular voicing, breath direction, and the degree of constriction of the vocal tract above the glottis (Stark 1986). It should be noted that the acoustic and perceptual analyses that have been car ried out on vocalizations have lead to divergent theories on the development of in tonation. Some data have been reported with reference to a small number of
PROSODIC CUES IN VERY YOUNG CHILDREN'S SPEECH
77
"distinctive features" as, for example, melody type, continuity and lax/tense opposi tion, and have lead to a precise classification by adult subjects of those perceptions, i.e. pain, hunger and pleasure, that are associated with an infant's states (WaszHockert, Lind, Vuorenkoski, Partanen and Valanne 1968; Delack 1975). Other studies have shown that an infant's cries are continuous and do not form discrete cry-types (Muller, Hollien and Murry 1974; Brennan and Kirkland 1982). And still others have found that control of voice intensity seems to precede control of fun damental frequency which is still relatively undeveloped at 6 months old (Murry, Amundson and Hollien 1977). As for children between 8 to 10 months of age, intonation contours seem to be come more and more regular (Konopczynski 1984). As of 8 months the babbling of children whose native languages are different shows an adjustment to the linguistic environment which is specific to the mother tongue, mainly in its rhythmic and melodic properties. The metaphonological characteristics of the mother tongue in fluence babbling more than do segmental properties (de Boysson-Bardies, Sagart and Durand 1984). Even if clear contrasts among productions are not always recog nized by adults, acoustical analysis of the signal does show a great variety of con tours, some of which are in accordance with adult norms. At about 16 months intonation contrasts, a rise and fall succession, appear on a single vocal sequence (de Boysson-Bardies et al. 1981). The analysis of curves of fundamental frequency do not exhibit any discontinuity between the contours of late babbling and those of first language. Intonation contours can be used to class ify child utterances just as adult sentences are classified as questions, assertions, ap peals, demands and so on (Menyuk et al. 1986). If we look now at the temporal organization of child productions, the onset of canonical babbling between 6-8 months, which is characterized by sequences of consonant-vowel units (Oller et al. 1976; Oller 1986; Holmgren, Lindblom, Aurelius, Jailing and Zetterstrom 1986), is preceded by an early form of temporal regularity towards 4-5 months (Bickely, Lindblom and Roug 1986). For the sake of clarity, articulatory tempo will be separated from speech timing. Articulatory tempo depends on the maturation of sensori-motor coordinative structures. Consonant as well as vocalic durations tend progressively towards those of adult speech, even if they widely exceed the adult's for a rather long time (Bacri 1987b). At 10 months old, they present some characteristics of the language of the child's environment, but it is not before 18 months old that one can consider that the adult target is attained (Mack and Lieberman 1985). However it is worth noting that prosodic patterns can be acquired even before good control of phonemic dura tional values. There is a clear difference in the syllabic duration of short utterances
78
N. BACRI
and long ones compared to each other (de Boysson-Bardies et al. 1981) at a time when there is still no significant decrease in word duration (Mack et al. 1985). A lengthening of final syllables in relation to initial syllables cannot be observed at 12 months, but appears clearly towards 14 months (Oiler et al. 1977). At 20 months, lengthening is about 20% to 30% in French children's speech, and is found only in the utterances that adults can interpret as intelligible (Bacri 1984a, b). A more detailed account of these data will be given in the third section, Intonation Condi tions that Effect Intelligibility. The progressive mastery of prosodic patterns per taining either to intonative contours or to utterance timing characterizes the transition period between babbling and first language. These studies allow us to conclude that prosodic cues are present in young children's speech. The question now is whether, like adults, young children attend to prosodie cues and use them to process speech.
Young Children's Perception of Prosodic Cues Recent studies have shown that the processes of perceptual categorization are more complex than had originally been expected (Bertoncini, this volume). Varia tions in speech rhythm, an integral part of prosodie structure, must necessarily be taken into account by children as they process spontaneous speech. Studies on categorical perception have only been carried out on syllables or disyllables that were pronounced by a single speaker (for an overview see Miller and Eimas 1983). The problem is more complex when the syllables are in context and children are presented word sequences; in this case the critical cues vary with the co-articula tions and the intonation structure of the sentence. Can a young child extract these cues, which have trading relations and which are spread out over a given length of time? Infants from 2 to 4 months old are sensitive to both spectral characteristics of the signal and syllable durations (Eimas and Miller 1980a; Miller 1981). Relational processing of parallel cues is thus set up, and this enables the child to adjust for speaking rate without disturbing perceptual constancy. The discovery of a "set of configurational properties that typify a phonetic category may depend upon ex posure to a set of exemplars that differ with respect to phonetic context and to talkers" (Kuhl 1979a:385). This remark gives rise to two antinomic work hypotheses as to the use of prosodie cues for speech processing by the child:
PROvSODTC CUES IN VERY YOUNG CHILDREN'S SPEECH
79
a. "Fundamental frequency and timber differences between the voices of different talkers and pitch contours are prominent characteristics that a listener must ig nore, since they do not produce critical information concerning phonetic iden tification" (Kuhl 1979a). b. The properties of the signal that determine phonetic distinctions and those that specify pitch and speech rhythm can be methodologically distinguished, but they are carried by the same waveform; children cannot identify phonemes and syllables without, at the same time and interactively, processing the socalled "suprasegmental" cues. Their speech does not become intelligible unless they adjust it to the intonational properties of their mother tongue. Their ability to produce speech whose pitch and tempo are structured is dependent on their ability to analyze speech tempo and pitch perceptually and not on their ability to ignore them. It is this second hypothesis that our work is based on. The processing of speech timing brings into play mechanisms that are analogous to those used to extract temporal phonetic information, e.g. voicing, or other aspects of the temporal structure of speech. Allen and Hawkins (1980) have shown that the development of phonological rhythm corresponds to the development of the perception of speech rhythm. Children's two "clocks" that specify segmental durations and speech tempo also enable them to structure their speech according to a complex configuration of acoustic cues (Allen 1972). Production control goes beyond the realization of a single articulatory movement when the child is able to pick out relational cues that are spread out over several segments, at about 10 months of age. Little work has been done on the study of how children extract pitch values, dis criminate and categorize pitch changes, as pitch is the element, par excellence, of speech that seems to have no linguistic value. It is known that the ability to dis criminate pure tones is about 2% between 1000 Hz and 3000 Hz (Olsho, Schoon, Sakai, Turpin and Sperduto 1982). Terhard's (1974) model of development in cludes a peripheral analyzer that extracts the components of the signal, among which is fundamental frequency. The signal is then matched to a stored spectral pattern which appears when the child has some experience of the mother tongue. According to Fowler, Smith and Tassinary (1986), infants about 4 months of age "perceive stress beats and stress-beat timing of syllables as adults do" (814). As early as 5 months old, they are able to discriminate a change in peak fundamental frequency in the final syllable of multisyllabic speech-like stimuli (Bull, Eilers and Oiler 1985) as well as changes in peak intensity and vowel duration (Eilers, Bull,
80
N. BACRI
Oller and Lewis 1984). Discriminated values are, in these experiments, repre sentative of pitch and of temporal contrasts cueing linguistic stress. Does the child make use of this ability? Kuhl (1976) showed that 4- to 16-week-olds, as well as 6-month-olds (Kuhl 1980), may ignore variations in speakers and pitch contours when they have to make a phonetic categorization, for example for vowels, but cannot ignore a change in vowel color when they have to respond to a pitch contour change. Thus it is more difficult to categorize sounds that share a pitch contour than those that share a spectral quality. It would be more difficult, but still possible, for infants to categorize pitch contours (falling vs. rising), too. Miller, Younger and Morse (1982) showed that 7-month-olds can categorize speech according to the sex of the speaker, provided that fundamental frequency was linked to voice timbre. Clarkson and Clifton (1984) demonstrated that 7-month-olds can spectrally categorize dif ferent tonal complexes according to pitch, as well as harmonic complexes bearing the same pitch categories, which however, lack fundamental frequency. Finally, Mehler, Bertoncini, Barrière and Jassik-Gerschenfeld (1978) have shown by using a habituation technique that infants only recognize their mother's voice when she speaks naturally, but not when her intonation is altered. Even if infants show a poorer resolution of frequency components for complex stimuli than do adults (Aslin, Pisoni and Jusczyk 1983), it is likely that towards the end of the child's first year auditive mechanisms of peripheral analysis enable the child to recognize and identify different intonation contours. Speech perception as well as production rest on temporal patterns of control which are based on physiological rhythm. Variations in pitch and timing serve a contrastive function, as well as a linguistic function. Little is known, on the other hand, about central processing mechanisms, to which little study has been devoted. To summarize, our present knowledge of children's abilities to produce and per ceptually analyze the intonation contours related to speech tempo and pitch change bring out two main ideas: - that given a theoretical paradigm of continuity between babbling and first lan guage, it is plausible that voice modulation is concomitant with the maturation process of articulatory gestures; the mastery of intonation and of articulation, far from being antinomic, develop in a parallel fashion. - that the mechanisms that regulate co-articulation are not fully mastered until age 6 or 7. However, it would seem that, given the presence in young children's productions of contrasts in duration and pitch of the same type that will have a linguistic value in adult language - contrasts supported by the
PROSODIC CUES IN VERY YOUNG CHILDREN'S SPEECH
81
early processing of analogous contrasts of duration and pitch - adult com prehension of children's speech can be based on these contrasts. It remains to see how adults, when listening to children's speech, process these contrasts that bring two voice dimensions simultaneously into play, the first being a temporal dimension, and the other a spectral dimension that is evaluated by the fundamental frequency. A pluriparametric analysis of intonation that accounts for both duration and pitch will certainly carry with it a better understanding of children's voice percep tion. Between 22 and 26 months of age, the structuration of child speech does not have the same accuracy at articulatory, morpho-phonetic and intonation levels. However, when a child is about 18 months old, intonation contours of first lan guage are compatible with those of the target-language, particularly at the end of utterances (Allen et al. 1980; de Boysson-Bardies et al. 1981). In addition, the relationship between final rises or final syllabic duration, which was not found when children were 12 months old (Oiler et al. 1977), begins to be set up at 18 months old (de Boysson-Bardies et al. 1981; Bacri 1984b, 1987b). On the basis of these data, it is worth examining how pitch and duration cues might be matched to form an adult's representation of child language.
Intonation Conditions that Effect Intelligibility An experiment was carried out in order to get a more precise understanding of the effects of intonation on children's speech intelligibility (Bacri 1984a, b). Three adults (two females and one male), native speakers of Parisian French, were asked to sort spontaneous productions of two male children into three categories: intelligible, unintelligible and undecidable. Subjects listened only to children's utterances; the response category chosen was given orally. The first of the two children was 22 months old at the beginning of the study and 26 months old at the end. He was visited six times at home, every three weeks, and produced 1450 utterances. The second one, aged 23 months at the beginning of the study and 26 months at the end, produced 520 utterances, over three sessions. The voice char acteristics of both children were almost the same; their utterances were numerous, but rather difficult to understand; their intonation was very "expressive". For each utterance, whether intelligible or not, a timing analysis was performed on syllables and sequences. A pitch detector plotted variations of fundamental fre quency (F0) as a function of time. All the F 0 fluctuations exceeding 35 Hz during at least 50 ms, i.e. perceptually relevant for an adult listener, were analyzed.
82
N. BACRI
The first idea one may have when listening to a child is that child utterances are understood, or not, depending on how they measure up to adult language. So, first "words" would in fact be words and not babbling because they have the same phonemic and syllabic structure as adult words. However, "translations" made by the adult subjects of our experiment for 150 intelligible utterances did not show a high level of agreement, except for two kinds of words: "personal-social terms" (Nelson 1975), as, for example, names, "merci" (thank you), "au revoir" (good bye), "quoi" (what), and terms referring to objects of classes or objects, such as "dedans" (inside), "encore" (another/again). In other respects, syllabic sequence analysis, performed on 150 utterances produced by the first child and 50 by the second one, showed the high variability of syllabic patterns. Samples were indeed too small to permit generalization, but it is worth noting that similarity among children's productions was stronger than similarity to adult French. These preliminary data were not consistent with the assumption that a com prehension strategy is mainly founded on adult reference to lexical items. Neverthe less, syllable and utterance segmentation could be based on references to specific phonemes, which, being better articulated and occurring frequently at the beginning or at the end of sequences, could function as boundary markers. Maximum efficien cy would be attained at those points (initial or final) where contrasts are the most perceptually relevant. Moreover, at these points "words" and empty phonological forms could appear (Dore et al. 1976), and carry an iatelligibility judgement.
Influence of Speech Timing on Intelligibility Phonemic contrasts are carried by all the dimensions of speech, from its acousti cal segmental structure to pitch contours. Adult perceptual integration involved a continuous processing of spectral, temporal and pitch cues. Thus, in French as in other languages, vocalic lengthening is a cue for the distinction between postvocalic voiced or unvoiced consonants. Whether each feature is processed as an in dependent source of perceptual information or follows integrated processing, is still unknown (Bacri 1986). Nevertheless, speech timing appears to strongly influence its processing (Lehiste 1970; Klatt 1976; Nooteboom, Brokx and de Rooij 1978 among others). Two aspects of young children's temporal organization of intonation, local speech rate and speech tempo have been studied. They are relevant to segmental identification and intelligibility. Local speech rate is referred to as the timing of the
PROSODIC CUES IN VERY YOUNG CHILDREN'S SPEECH
83
syllabic segments along a stream of speech sounds, namely, an utterance. Swiftness of speech is related to the acoustic structure of segments (intrinsic variations), to the position of a segment in the utterance and to its context (co-intrinsic and extrin sic variations). These last variations are the most important for the comprehension of speech. Thus, in adult speech, initial syllable length varies in inverse ratio to ut terance length, and is still shorter if followed by an unvoiced stop (for an analysis of French see DiCristo 1978). Syllabic duration is most influenced by what is to follow. On the other hand, a syllable is lengthened when preceding a major syntac tic break or when it is in final position, in French like in English (Vaissière 1980; Grosjean and Deschamps 1972). The hypothesis was that temporal variations could be used by adults as anticipatory cues and boundary markers. They may also serve to trigger a retrospective perceptual normalization and so enable the adult listener to adjust to children's speech. As for tempo, speaking rate depends on the number and duration of pauses, and also on the articulation rate or time spent in articulating (Goldman-Eisler 1972; Grosjean et al. 1972; Miller 1981). The spectral quality of speech is affected by tempo variations. Thus Miller reported that listeners' judgements on phonetic struc ture and on tempo, interact and are interdependent. Adjustment by the listener to the speaker's tempo could be the basis for temporal information processing.
Figure 1
Mean differences and confidence intervals (in ms) between initial and final syllabic dura tions, in short and long sequences, for intel ligible (IU) and unintelligible (VU) utterances (Bacri 1984a, b).
84
N. BACRI
Our data showed that the intelligibility of children's speech was related to the presence of a strong contrast between initial and final syllabic durations. This con trast was absent in sequences that were judged as unintelligible (Figure 1). Moreover, changes in duration varied from 21% to 40% in intelligible utteran ces, and thus were higher than the differential threshold, according to Klatt (1976), whereas they varied from 1% to 8% in unintelligible utterances and it was not pos sible to discriminate among them. Speaking rate and articulation rate were quite similar for the two types of ut terances; it was almost constant throughout the whole corpus. Pause length, the second component of speaking rate, was about 20% to 27% of the total mean dura tion and had no effect on intelligibility. But the position of pauses in utterances clearly varied. Intelligible utterances were characterized by one pause in median position, just after the first syllable (short utterances) or after the second or third one (long utterances). In all cases, a pause was preceded by syllable lengthening. The temporal structure of initial syllables in intelligible utterances conformed to their most frequent phonetic structure, i.e. vowel or unvoiced consonant + vowel. In other respects, the main difference between intelligible and unintelligible utteran ces was that a strong contrast between initial and final syllabic durations was found only for intelligible utterances. The regularity of this contrast suggests the presence of a timing organization. Plausibly, this contrast corresponds to higher articulatory accuracy, which in turn is linked to a better approximation of adult language. Standard French is clearly marked by a strong contrast between the shortness of in itial syllables and the lengthening of final syllables. As a result, local speech rate variations can serve as perceptual cues, whether they are proceeded primarily by acoustic information or are specified as features in a phonetic representation on the basis of the acoustic properties of a speech signal. Timing infonnation may underlie the listener's adjustment to a signal according to a general normalization process. Such an effect created by the durational contrast between the beginning and the end of utterances would be consistent with two kinds of speech processing, namely phonemic identification and the perception of syllable and utterance boundaries.
Influence of Pitch on Intelligibility Intonation contours are usually analyzed on the basis of pitch movements, i.e. F 0 variations. F 0 configurations, as they are visualized by an instrumental analysis of intonation, depend on several factors, from segmental acoustic structure to the
PROSODIC CUES IN VERY YOUNG CHILDREN'S SPEECH
85
physiological and emotional states of speakers (Cruttenden 1986). Listeners must extract linguistically relevant information. Moreover, F 0 varies considerably with speaker's voice characteristics. Thus several models have been proposed to explain how linguistically significant cues are extracted from pitch variations, and how these cues can facilitate segmentation and identification of the speech waveform. As for children, some authors maintain that there is a link between pitch con tours and situations such as monologues, mother-child dialogues, etc. (Menyuk et al. 1986; Konopczynski 1984). However no recurrent contour was found in our data (Bacri 1984b). All possible contours occurred in all situations without any cor relation between pattern distribution and either situation typology or intelligibility categories. On the other hand, children's voice characteristics were strikingly con stant all through the corpus for every kind of contour (flat, rising and falling). However, it is worth noticing that the position of F 0 changes exceeding 100 Hz seemed to be related to intelligibility; changes occurred on the last syllable for in telligible utterances, or, in the case of long utterances, on a syllable preceding a pause. In the latter case, the utterance was identified by adults as a sequence of two constituents. In unintelligible, utterances, 100 Hz changes could occur at arbitrary locations. A second kind of analysis was carried out in order to determine the dynamic evolution of voice, since it might be that intelligibility depends on the auditory in tegration of pitch and thus on signal continuity. This analysis was born on the num ber of F 0 fluctuations per syllable and per second, and no longer on absolute values and mean amounts of F 0 changes (Bacri 1984a). In fact, it is not so easy for a lis tener to extract an intonation contour from a child's speech signal, given the steep jumps in children's voices. The average number of F 0 fluctuations per syllable was quasi-constant for the whole corpus. For each session and for the two children, mean value ranges were almost the same. But significant differences were found as a function of intelligibility (Figure 2). The positions of F 0 fluctuations in utterances differed between the two kinds of utterances, as did the main F 0 changes. Small fluctuations occurring in all syllabic positions perturbed intelligibility. Intelligibility depends on a small number of fluc tuations per second. A smooth F 0 slope increased intelligibility. The most salient result consisted of the close relationship between intelligibility judgment and the rate of F 0 change. This suggests a high correspondence between the temporal and pitch dimension of intonation. Intelligibility is impaired by either abrupt or frequent jumps occurring randomly, i.e. not preceding a pause; it decreased most when a jump occurred on a short syllable and when several small fluctuations intensified the more important F 0 changes. Thus, the data suggest that an intelligibility
86
N. BACRI
Mean number of F0
fluctuations
Figure 2
Mean number of F0 fluctuations per syllable and per second for inteligible (IU) and unin telligible (UU) utterances, with confidence in tervals (Bacri 1984a, b).
decision is supported by a listener's ability to process speech as a continuous signal and that regularity of F 0 evolution is a relevant cue for speech continuity.
Conclusions In spite of the great variability in intonation contours, 22- to 26-month-old children's intelligible utterances are characterized by a close correspondence be tween duration and pitch patterns. Final or prepausal syllabic lengthening and salient F 0 movements appeared on the same segments. Therefore, both can be in tegrated as boundary cues and facilitate either phonemic or syllabic segmentation. Intelligibility was increased still more by good correspondence between intonation and phonemic structure, such as shown by the data on unvoiced stop consonants, whereas unintelligibility was linked to the presence of a single kind of cue. For in stance, a large F 0 fluctuation corresponding to a short syllable regularly entailed an unintelligibility judgment.
PROSODIC CUES IN VERY YOUNG CHILDREN'S SPEECH
87
Our results are consistent with recent research in the pluriparametric structure of intonation (Rossi et al. 1981). As for adult speech, inappropriate pitch contours and temporal relations within utterances decrease intelligibility (Huggins 1978). In ex periments where the three "suprasegmental" variables (pitch, duration and amplitude) were interchanged, the magnitude of the pitch effect was about 15-20%, with regard to English sentences (Streeter 1978). It was about 10% when final words had to be identified in Dutch sentences (Nooteboom et al. 1978). As for French, intelligibility is mainly modified by syllabic lengthening, whereas boundary identification depends on pitch and duration interaction (Bacri 1987a). In fact, syl labic lengthening is an important determinant of speech rhythm and also cor responds to more acute articulatory movements. Final syllable lengthening in the children's corpus fits in with this feature of French. Joined with a smooth F 0 , it supports intelligibility. Similarly, the positions of pauses within intelligible utterances provided evidence of children's "microplanning" (Butterworth 1980) for phonetic sequence choice. Phonetic sequences could have a status analogous to adult words. For the same reason, listeners' se quence segmentation is facilitated by pauses. Moreover, pauses help to detect the boundaries of intonation units when following syllabic lengthening. Data also showed that the presence of a single cue, either relevant to timing or to F 0 , impaired intelligibility. The exact kind of interaction between temporal and pitch parameters is still a problem. Hierarchical models have been proposed, ac cording to which duration (either local speech rate or speech tempo) often dominates other dimensions; sometimes F 0 is defined as the most relevant parameter (Pierrehumbert 1981). The point that is important for us here is that studies bearing on the perception of stress within short segments gave evidence of trading relations among several cues (Thorsen 1980). What is perceptually relevant is not timing alone or F 0 change per se, but "F 0 timing", i.e. trading relations be tween both kinds of changes; speech continuity perception would thus be based on adjustments between parameters. Intonation contours could then be identified as to some abstract patterns which provide information for a meaningful representation. Although able to perceive all the signal fluctuations above threshold, an adult would process as intelligible only those fluctuations which correspond to a coherent signal and which bear information consistent with that usually found in adult speech. The relevance of intonation cues for an adult listener would be constructed by reference to an internal representation of what child language is and what speech production mechanisms are. Speech development presupposes very close coordination between a child's bab bling and a mother's words. Vocal exchanges between mother and child show that
88
N. BACRI
as of the first months of the child's life, a rhythmic pattern that derives from a mutual regulation of behavior is established. The existence of turn-taking in mother-child "dialogues" has often been observed, even if at times children's ut terances and mother's speech are super-imposed. But this superimposition of vocal behavior occurs only briefly and infrequendy, as in the case of a warning, distress signals, laughs and the mother's imitation of the child's vocal behavior (Schaffer, Collis and Parsons 1977; Trevarthen 1977). These exchanges are signs of a tem poral structure. Furthermore, mothers adjust their voices to the voice of their children as of the baby's birth, and continue this adjustment long after (Shockey and Bond 1980; Stem, Spieker and MacKain 1982; Stern, Spieker, Barnett and MacKain 1983; Bernstein-Ratner and Luberoff 1984). Gamica (1974) has shown that intonation patterns are better defined in the phrases addressed to children than in those addressed to adults. Parent-child communication is a particularly striking case of the adjustment that all listeners must make in order to understand their in terlocutors.
Basic Discourse Capacities
Structural preoccupations have been dominant in linguistics for some time; but the usefulness of a synthesis of structural and functional approaches has long been apparent from the work of the Prague linguists who developed Buhler's ideas, especially in the study of grammar. The particular form taken by the grammatical system of language is closely related to social and personal needs that language is required to serve. But in order to bring this out it is necessary to look at both the system of language and its functions at the same time; otherwise we will lack any theoretical basis for generalizations about how language is used. M.A.K. Halliday (1970)
Cohesion: Syntactic Organization Leading to Discourse
We use the term grammatical strategies, or sentence strategies, although it should be kept in mind that these are not strategies that are formulated by the grammar, but cognitive strategies that are used to produce or understand structures that are specified by the rules of the grammar. It is at this point that the difference between the notions of the rule and strategy become most obvious. Even if we accept the hypothesis that grammar is a theoreti cal - and hence general, abstract, and idealized - reconstruction of the language rules known by language users, there is still a crucial difference between grammati cal rules on the one hand and strategies on the other. Typically, rules apply to structures taken as complete entities, a posteriori. For both, synthetic and analytic types of rules in sentence grammars or cognitive models, the categories and units are typically structural... Strategies pertain to the latter kind of linear processing, that is to information that a language user processes fragmentarily. This does not mean that strategies do not also imply higher level structures, or that parallel processing would be excluded in a cognitive model of language use, but rather that the data are processed in real time. Once the relevant data are processed, a language user may use rules to check whether the strategies have been correctly applied. T.A. van Dijk and W. Kintsch (1983)
Verbs as Sentence Organizers G. Piéraut-Le Bonniec Laboratoire de Psychobiologie de l' Enfant, Paris and J.P. Pille Inspection Départementale de l' Education Nationale, Etampes
Whatever hypotheses might be proposed to account for early linguistic com petence, it is generally agreed that by the age of three years, and in some cases even younger, children have tacit knowledge of a great number of the rules which determine the grammaticality of sentences. Children make themselves understood and understand what is said to them, at least as far as the semantic content that is expressed, is within the grasp of their world knowledge. This early "awareness" of the rules underlying language functioning, does not of course imply any capacity to state them. Metalinguistic abilities evolve much later, very probably through the in fluence of scholastic schooling. The first grammatical rule that young children are able to isolate is the relevance of word order. From the start children's two-word utterances do not occur in hap hazard order. Braine (1963) found that as soon as the child has gone beyond simple holophrases, two classes of words can be distinguished: "open" and "pivot". The position of the words belonging to the former class is fixed in relation to the latter, at least for any given child although this might not be the same for all children. Braine called this "positional learning". According to McNeil (1970) the different grammatical categories, e.g., nouns, adjectives, verbs, adverbs, etc., will develop out of these two classes and be established. For H. Sinclair and J.-P. Bronckart (1972) the most important question is less how children set up these categories than how they set up the grammatical relations between the subject-noun phrase, the verb phrase and the object-noun phrase, al though the authors do believe that these two phenomena cannot be dissociated. They set up a schema of language development from the results of an experimental
94
G. PIÉRAUT-LE BONNIEC AND J.P. PILLE
study they carried out on children where the youngest were less than three years of age. According to this schema at the start there are two initial sentence patterns: 1. utterances where the strongest link is between the subject and the verb and where the object is only an extension of the verb; the two together make up the "predicate"; 2. utterances where a verb-object pair is set up and this in turn combines with the subject-verb to make up the /subject-verb-object/ (SVO or SOV) structure which seems to be universal. It can be seen that all these studies tend to bring out how important it is for lan guage development that a particular category of words (pivot- or linking-words) be set up whose function is similar to that of the verb in a standard sentence. The question, then, is to find out what processes lead to the differentiation between the categories of verbs and nouns and to study both its semantic and syntactic aspects. Authors who have followed in Piaget's line of thinking put greater emphasis on the semantic aspects. E. Ferreiro (1971), for example, like Sinclair and Bronckart (cf. Sinclair and Bronckart 1972; Sinclair 1987), have set up a parallel between the construction of a category of verbs and the construction of those schemas of action that enable the child to transform objects or learn about their properties. Babies are explorers and researchers (cf. Sinclair, Stambak and Lézine et al. 1982; Langer 1980, 1986) and, according to Sinclair and Bronckart, it might be that what children express by holophrases are schemas of possible action that are closely linked to what they want to obtain or what they are doing. While the actor, patient and action are an indivisible whole in holophrases, children very quickly begin to describe events in which they themselves are the actors (Sinclair and Bronckart 1972). Thus, words that express actions or transformations might be picked out by the child very early and would stand out distinctly from the others. A second point of view puts greater emphasis on syntactic aspects. It should be noted that while the category of nouns can fulfill different functions in the sentence (as subject or as object), verbs, on the other hand, have only one function, given the type of meaning they carry, which is to link noun phrases. Maratsos (1982) suggests that certain phenomena concerning the. sentence helps the child to dif ferentiate between noun phrases and verb phrases. Verbs are words that carry nega tion, and only verbs can change tense and aspect. De Boysson-Bardies (1976), for example, has shown that the French word, "pas" (not), plays the role of an operator of negation at a very early age (by 21 months in her corpora). Furthermore, a num ber of other studies have shown that, at a very early age children are able to apply to verbs, and only to members of the category of verbs, specific rules for mor phological transformation. Obviously) as Bronckart (1976) has shown, a tense-
VERBS AS SENTENCE ORGANIZERS
95
aspect system is not completely set up until the child reaches adolescence. Never theless, a number of studies have shown both for French and English that transfor mations carried out on verbs by very young children are definitely the product of internalized rules, and not merely imitations. The proof is that these rules, which can only be applied to regular verbs, are sometimes incorrectly applied to verbs whose conjugation is irregular, as in Brown's (1970) classic example of "he goed", where the child applied the regular past tense morpheme to a verb that cannot take it. How do children represent verbs, a distinctive category of words which, none theless, children are able to use correctly at a very young age? As has been shown by a number of authors (Hall 1976; Berthoud-Papandropoulou 1978, 1980; Evans, Taylor and Blum 1979; Huttenlocher 1964; McNinch 1974), children have a great deal of difficulty segmenting sentences into words. It also seems that until the age of seven, children do not differentiate the word from the thing it represents. When children below this age are asked to give a "long word" they give answers such as "train" or "snake" (Papandropoulou and Sinclair 1974; Berthoud-Papandropoulou 1980). And when Gérard (1981) asked children to shorten a sentence, they made it negative. The effect of this was to reduce the number of actions, as one of them was disaffirmed, but to increase the length of the sentence which is the opposite of what was asked to be done. Finally, a study by Costerman and Giurgea (1988) shows that until the age of 6 children have difficulties understanding that the parts making up a phonemic label do not necessarily correspond to aspects of meaning. For example, if a child is asked how the French word, "printemps" (spring), begins, the answer may be, "ça commence par les fleurs" (it begins with flowers), or the word "télévision" will be said to begin by "bouton" (button). Along the same lines can be found works carried out on the development of children's judgements of the grammaticality of sentences (Clark 1978; Gleitman and Shipley 1972; Evans and Tunmer 1980; Tunmer and Grieve 1984; Van Kleeck 1984). The results of these studies seem to show that children are unable to separate the form of the sentence from its semantic content before the age of seven or eight. Van Kleeck (1984) has attempted to fit her work on the development of metalinguistic abilities into the framework of Piaget's theory. Thus, according to her, a distinction must be made between the pre-operative stage where children cannot account for more than one aspect of language at a time, either the form or its content, and the operative state where they can manipulate both these aspects simultaneously. In fact, it would seem that first of all, learning to read, and then learning gram mar, both play a very important role in children's ability to consider language as a
96
G. PIÉRAUT-LE BONNIEC AND J.P. PILLE
thing which can be talked about. Thus, this study (cf. Pille 1986) is divided into two parts. The first set of experiments was carried out to find out how pre-grammar school children (age 5 and 6) represent the function of the verb. The second set of experiments studied the development of the notion of the verb for children between the ages of 7 and 11, i.e. children who have had a certain amount of grammar as part of their schooling.
Children's Representation of the Verb Before Reading Age Children from age 5 to 6 were put through two series of tests. The first set was run to learn how verbs were seen as words of the language. The second set studied verb recognition in their role as sentence organizers.
The Verb as a Word of the Language Twenty middle- to upper-class, kindergarten children (aged 5;4 to 6;4) who were born French and who had no apparent problems, were tested by the experimenter in a one-on-one situation. After giving a few examples, the experimenter asked the child to give him some words, "dis-moi un mot, celui que tu veux" (tell me a word, any word you want to). The test was considered to be over when the child had given 10 words. The result was that no child gave verbs spontaneously; all the words they gave were names of things, and in particular, things in the immediate surroundings. The experimenter then infonned the child that he would say something and the child was to say if it was a word or not ("maintenant, à mon tour je vais te dire quelque chose; tu me diras si c'est un mot ou si ce n'est pas un mot"). The ex perimenter then offered the child the following 10 words: CHAISE, TABLE, DORMIR, COUSSIN, SAUTER, DANSER, TÉLÉPHONER, MUSIQUE, SAGE, RÊVER (CHAIR, TABLE, TO SLEEP, CUSHION, TO JUMP, TO DANCE, TO PHONE, MUSIC, GOOD, TO DREAM).
(There is no ambiguity in French between the infinitive form of these verbs and their semantically corresponding noun from which they are clearly distinguished.) Only 7 out of 20 subjects recognized all the given words, including the 5 infinitive verbs, as such; 2 subjects eliminated all the verbs. The 11 others accepted some verbs and rejected others. The reasons the children gave to justify their answers show that children at this age do not yet distinguish verbs, as words, from their
VERBS AS SENTENCE ORGANIZERS
97
meanings and that acceptability criteria are affected by those meanings. One child, for example, thought that "jump" was not a word because it's not normal to jump ("quand on saute, c'est pas normal"). Another example is that "dance" is a word because when you dance well it's really a word ("oui quand on dance bien c'est vraiment un mof). Finally, the experimenter asked the child to give him a word that resembled each of the 5 previously mentioned verbs ("Donnes-moi des mots qui ressemblent à ..."). It was found that when a subject did give verbs (only one subject gave 5; 5 sub jects gave a few), they were always verbs that belonged to the same semantic space. For example, SAUTER (TO JUMP) as a prompt, gave SAUTILLER (TO SKIP); DORMIR (TO SLEEP) gave "faire une sieste" (to take a nap). 14 subjects did not produce a single verb, but gave names of objects semantically related to the prompt. Therefore, it would seem that at this age, words are mainly names of things. Verbs are problematical for children insofar as they want to differentiate them from names of things and thus they hesitate to consider them words. It is interesting to note, oh the other hand, that children at this age who have no difficulty using verbs correctly in the correct context, cannot produce this category of words when they are asked to do so in a context-free situation.
The Verb as a Sentence Organizer In this test the children were presented with 5 sets of 4 sentences, half of which were deviant. Each set included: (1) a basic sentence which was correctly struc tured and made up of a subject noun phrase, a verb phrase in the present tense and an object noun phrase, (2) a sentence constructed out of a temporal transformation of the first, where the verb phrase was in the past or future tense, (3) a sentence where the verb phrase was missing, and (4) a sentence made up of the same words as the basic sentence but where the word order was changed around. For example: (1) (2) (3) (4)
Agnès brosse ses dents. Agnès a brossé ses dents. *Agnès ses dents. *Agnès dents ses brosse.
(Agnes brushes her teeth) (Agnes brushed her teeth) (*Agnes her teeth) (*Agnes teeth her brushes)
The order of presentation of these sentences was, of course, changed around for each set.
G. PIÉRAUT-LE BONNIEC AND J.P. PILLE
Table 1
Given Sentences Incorrect
Correa Present tense
Other tenses
Without a verb
Incorrect order
Acceptable
75%
77%
70%
7%
Judged unacceptable
25%
23%
30%
93%
The percentages of sentences judged to be acceptable or unacceptable as a function of the type of utterance presented (N = 100 for each type; 5 utterances of each type were pre sented to 20 subjects).
Table 1 gives the percentage of sentences judged acceptable for each of these four types across children. As can be seen, a majority of children judged the nondeviant sentences as ac ceptable, and judged the incorrectly ordered sentence as unacceptable. Curiously, however, 70% of the children judged the sentences with missing verbs as accept able. Another experiment with children of the same age, gives results that point to the same type of conclusion. In this case the experimenter told the children he would ask them to do something. They were to do what he said and could ask questions if they wanted to ("Je vais te demander de faire quelque chose. Tu feras ce que je te demande de faire. Tu as le droit de me poser des questions si tu veux). He then read out the following noun phrases: LA PORTE (THE DOOR), LE CRAYON (THE PEN CIL), LE PAPIER (THE PAPER), LE BONBON (THE CANDY), LE DÉJEUNER (THE LUNCH), LA POUPÉE (THE DOLL), LE RIRE (THE LAUGH). T h e objects w e r e present
in the immediate environment (except for the lunch and the laugh!). It is remark able that very few of the subjects had any trouble in the given situation and that only one child asked, and for the first item only, what she was expected to do ("Quand tu dis LA PORTE, il faut faire quoiV (When you say DOOR what must I do)). For all the other subjects, the instructions posed no difficulty at all and they carried out an action which they spontaneously commented on. For example, for "LA PORTE", came comments of the sort: "je l'ouvre" (I open it), "on l'ouvre" (you open it), "ça s'ouvre" (it opens); for "LE BONBON" there was "ça se mange" (you can eat it); for "LE CRAYON" there was "j'écris" (I write), etc. 75% of the responses
VERBS AS SENTENCE ORGANIZERS
99
were of this type. But when subjects were asked who told them to do it ("Qui t'a dit de le faire?"), or how they knew what to do ("Comment tu sais ce qu'il faut faire?"), the answers were of the type "LA PORTE, c'est pour rentrer et sortir" (THE DOOR is for going in and out), or "parce qu'il y a une feuille et UN CRAYON" (be cause there is some paper and A PENCIL). The other children responded by pointing at the object and saying, for example "le voilà LE CRAYON" (there's THE PENCIL) or by mentioning its absence, as in "il n'y a pas de DÉJEUNER" (there is no LUNCH). Thus, for children under age 7, verbs seem to have a status that is different from other words of the language. In fact, their status would almost be extralinguistic since the child has difficulties including verbs in the category of words. A verb seems to function like a sort of appendix to names of things, and if it is missing, the child will construct meaning out of the known properties of the nouns that are part of the utterance.
The Development of the Notion of a Verb Between Ages 7 and 13 The French school curriculum plans for children to learn to read between age 6 7. It would seem that learning to read would help children to separate the word as a signifier, from its meaning, and help them to develop a rudimentary form of metalinguistic activity, so that they could begin to "stand back from" and "see" their language. At 8 years old children begin to learn about grammar, morphology (in the study of conjugations) and syntax, e.g., the expression of temporal relations. This learning should help the child to differentiate verbs from other words in the language and to establish the verb's role as sentence organizer. However learning to read and thinking about grammar are completely different activities. That is why the experiments as discussed below, were carried out on three different groups of subjects, representing three different levels of schooling. The first group (which will be referred to as NS.l) was made up of 21 second graders, aged 6,10 - 8,9. Normally by this time children have learned to read, but learning about grammar has just begun. The second group (which will be referred to as NS.2) was made up of 22 third graders, aged 7,11 and 10,5. As far as grammar is concerned, these children have mainly studied morphology and in particular, the verb conjugations. Finally, the third group (which will be referred to as NS.3) was made up of 28 fifth graders. These children have been taught to recognize the main grammatical func tions.
100
G. PIÉRAUT-LE BONNIEC AND J.P. PILLE
The Importance of Knowing the Meaning of a Verb in order to Understand the Meaning of a Sentence During their time with the experimenters, the children were presented with some sentences containing neologisms; the same neologism could occur in three different places: in the subject noun phrase (NP1), the verb phrase (VP) or the object noun phrase (NP2). Thus in the following example, the neologism BREUDE occurs: - as a subject: "BREUDE m'a offert un joli livre" (BREUDE gave me a nice book) - as a verb: "Aurélien BREUDE un petit ordinateur" (Aurélien BREUDEs a little computer) - as an object: "Jean s'est acheté un BREUDE" (Jean bought a BREUDE) The three sentences were printed on a card and the children were asked to read them. They were then asked to say which sentence they understood the best, and the one they understood the least, i.e. the one that seemed the most difficult to un derstand. Four series of sentences were constructed and presented in the same way but in varying orders. As can be seen in Table 2, a definite change appears between NS.l and NS.2. Only 50% of the younger children considered that the sentence with the neologism in the NP1 position was the easiest to understand and only a little more than 50% considered that the sentence with the neologism in the VP position was the most difficult. On the other hand, as of age 8 (NS.2 and NS.3), the neologism in the VP situation is considered understandable only in a few cases (9% and 8%). The verb is thus seen as the key to the sentence; and while it can help in assigning meaning to an unknown subject, the contrary is not true. It is interesting to note that the greatest behavioral change in the subjects occurs between the first and second level, between the ages when learning to read is the main subject of study and when learning about grammar begins.
Preeminence of Regular Verbs While it can be said that the 8-year-old has developed a verbal phrase function, does this mean that all the verbs of the language have been identified as words that
101
VERBS AS SENTENCE ORGANIZERS
Table 2
Easiest utterances to understand
Most difficult utterances to understand
Position of the neologism
NP1
VP
NP2
NP1
VP
NP2
School level 1 (NS.l) N = 21
52%
15%
32%
25%
55%
20%
School level 2 (NS.2) N = 22
71%
9%
21%
13%
75%
12%
School level 3 (NS.3) N = 28
81%
8%
11%
5%
81%
14%
Percentages of utterances which were judged easy or difficult to understand given the position of the neologism in the sentence and changes in choices as the school level goes up.
are able to fulfill this function? And then, if the grammatical category of verbs is identified as such, is the source of the recognition process rooted in semantics or morphology? In an attempt to answer these questions, two experimental situations were presented to the children of NS.2 (age 8 - 9 ) and NS.3 (age 10 - 13). At first, the children were asked to classify a set of mixed-up words which in cluded names of things, verbs and adjectives. These words had been chosen from a development scale of words defining lexical knowledge in French which had been set up by Ehrlich, Bramaud du Boucheron and Florin (1978). The children were tested in a group. The words were written one after another in random order on sheets of paper that were handed out to the children. They were to put the words from this list in order, as if they were putting them into boxes. They were to decide on the number of boxes that were needed and to put names on them. The results are as follows: a. Coherent classification, i.e. where a unique criterion is used, does not appear until NS.3, be it classification by grammatical categories or by first letters. This is true for 13 out of 28 subjects. b. For some of the youngest children the data was impossible to interpret, e.g. Sophie put dictée (dictation) with entier (entire). At NS.3 there are no longer uninterpretable assignments.
102
G. PIÉRAUT-LE BONNIEC AND J.P. PILLE
c. Three types of criteria were found: i) semantic, e.g., one child put together dictée (dictation), attentif (attentive) and punir (punish), while another put together diable (devil) and noir (black); ii) phonetic, classification by shared first letters or endings; and iii) grammatical, classification by grammatical categories. For a number of subjects these criteria could be used simultaneous ly, e.g., Karine who put together words containing -oir. boire (drink), noir (black), voir (see), couloir (hallway), mouchoir (handkerchief), croire (believe) and also put together words seeming to have something to do with school: dictée (dictation), punir (punish), absent (absent), attentif (attentive), répondre (answer), papier (paper). Table 3 gives the types of classification found according to age. Classification by semantic space was most often found for children between 8 - 9. On the con trary, for 10- to 13-year-olds, meaning was ignored and phonetic categorization was most often used, far more than grammatical categorization. However, it must be noted that if the subjects used grammatical categories, it is the category of verbs that was the most often used, e.g., a subject put together all the verbs but classed the nouns alphabetically. The question then comes up whether there wouldn't be more grammatical clas sification, and less semantic classification, if the children were presented with neologisms having correct morphological features, as found in French nouns and conjugated verbs. Thus, the next experiment was carried out. It was divided into two parts. For the first part, 22 NS.2 and 28 NS.3 children were handed out lists on which were presented 24 neologisms in random order. There were 24 neologisms altogether, in cluding 12 with possible French verb endings, e.g., parlIR, fauchOlR, roumER, 6 with possible noun endings, e.g., beurANCE, laitEUR, tabEAU; and finally 6 with en dings not found in French, e.g., voulAB, sardOF. The children were asked to class-
Table 3
Types of Classification
NS.2 NS.3
Uninterpretable
Semantic
Phonetic
16%
61%
24%
5%
72%
Grammatical
23%
Percentages of the types of classification made by the children as a function of school level.
103
VERBS AS SENTENCE ORGANIZERS
ify these unfamiliar words into two categories: those that could be considered like verbs, and those they could not be considered verbs. The second part of the experi ment consisted of a one-on-one discussion with the children. Table 4 gives the percentage of neologisms that were identified as verbs given their endings. It can be seen that as of NS.2 the verbs have been clearly differentiated from other words in the language in more than half the cases, and at NS.3 in more than 80% of the cases. However a distinction should be made among the verb endings that were attached to the neologisms as can be seen in Table 5. Here it can be seen that the neologisms that are identified as verbs by children at NS.2 are those that have regular verbal conjugations, i.e. verbs whose infinitive ends in -er or -ir\ these are identified as verbs 72% of the time. On the other hand,
Table 4 Identified as verbs by the children at level: NS.2
NS.3
French verb
58%
89%
French noun
10%
2%
7%
2%
Neologisms whose ending mimics a:
Neologisms whose ending does not conform to French endings
Percentages of neologisms that were identified as verbs as a function of the sub jects' school level and of its type of ending, neologisms that mimic French verb or noun endings and neologisms that do not conform to French endings.
Table 5
Endings of the Neologisms Conjugation type
NS.2
NS.3
regular (ending in -er or -ir)
72%
97%
irregular (ending in -oir or -re)
58%
88%
Percentages of neologisms that were identified as verbs as a function of their verbal ending.
104
G. PIÉRAUT-LE BONNIEC AND J.P. PILLE
those neologisms whose endings are similar to more irregular French verbs are treated as verbs only 58% of the time. The discussions with the children that followed this experiment helped to bring out which procedures were the most often used to identify the neoverbs. NS.2 children explained that they tried to conjugate the neologism: Example 1 - Sandra: E: Essaie de m'expliquer comment tu as fait pour savoir si tel mot était un verbe ou n'était pas un verbe. (Try to tell me how you knew if a word was a verb or not.) S: Bah, je l' ai conjugué. (Well, I conjugated it.) E: CONVOIRE, par exemple tu m'as dit que c'était un verbe. Pourquoi? (CONVOIRE, for example, you said that it was a verb. Why?) S: Parce qu'on peut dire: Je convois, tu convois, il convoit. (Because you can say, I convois, you convois, he convois.) Example 2 - Frédéric: E: Tu me dis que CONVOIRE est un verbe. Comment as-tu trouvé? (You told me that CONVOIRE was a verb. How did you know?) F: J'ai dit "convoirer". (I said "[to] convoir".) E: Et BAUM? (And BALIM?) F: Parce que "balimer". (Because [to] balim.) E: Et COSTID? (And COSTID?) F: C'est un verbe parce qu'on peut dire "costider". (It's a verb because you can say [to] costid.) E: Comment sais-tu que tel mot est un verbe et que tel autre n'est pas un verbe? (How do you know that one word is a verb and another isn't a verb?) F: Il faut que ça fasse bien à l'oreille. (It has to sound good.) E: Et si tu hésites? (And if you're not sure?)
VERBS AS SENTENCE ORGANIZERS F:
105
Alors je mets "je", "tu", "il". (Well, I use "I", "you", "he".)
It can be seen that the words that are classified as verbs generally undergo a transformation in accordance with what is referred to as the first group of verbs, those ending in -er, which make up the majority of verbs in French and which appear to be the most regular. Moreover, new verbs in the language invariably take on the conjugations of this group. Thus, in the one-on-one discussions it was found that PARLIR, which should normally be identified as a verb belonging to the second group, those ending in -ir, were not identified as a verb unless they could be assimilated to the first group verbs: "je peux le transformer en 'parle' et alors je peux le conjuguer" (I can change it into 'parle' [a verb belonging to the first group meaning 'speak'] and so I can conjugate it), said one subject. In the same way CHINDRE and FAUCHOIR, which should be put into the third group of verbs, i.e. those ending in -re, were considered to be verbs be cause one could say "chindrer", while FAUCHOIR (normally a 3rd group form) permitted "nous fauchoirons", as if it were the verb, fauchoirer. Thus, at this grade children are able to identify certain inflexions as characteristic of verbs, and in particular forms ending in -ir or -oir. However, on the other hand, it would seem that for them these are not really verbs unless they can be assimi lated to the most prototypical form, i.e. verbs whose infinitives end in -er. The results are different for NS.3 children, who recognize several types of con jugations: Example 1 - Grégory: E: Comment as-tu fait pour trouver les verbes? (How did you find the verbs?) G: Je savais que les verbes que je connaissais se terminaient par IR, OIR, les verbes du 1er groupe en ER; et les autres mots ne se terminaient ni en IR, ni en ER, ni en INDRE. (I knew that the verbs I know end in IR or OIR; the verbs of the 1st group end in ER; and the other words didn't end in IR, or ER, or INDRE.) Example 2 - Alain: E: Pourquoi m!as-tu dit que CONVOIRE était un verbe? (Why did you tell me that CONVOIRE was a verb?) A: Parce que c'est un verbe en OIRE. (Because it's a verb in OIRE.)
106
G. PIÉRAUT-LE BONNIEC AND J.P. PILLE
E:
Et BAUM? (And BALIM?)
A:
Ce n'est pas un verbe; ça ne se termine pas comme les verbes du 1er, du 2ème ou du Sème groupe. (It's not a verb; it doesn't end like the verbs in the 1st, 2nd or 3rd group.)
During the discussion the children were also asked what a verb was and what it did. It is only at NS.3 that they respond in such a way that they show they have un derstood the function of verbs, as can be seen in the examples below: Example 1 - Grégory: E: À quoi sert un verbe? (What does a verb do?) G: 'est pour faire des phrases; il en faut un, sinon cela ne voudrait rien dire. (It's to make sentences; you need one; otherwise it doesn't mean anything.) Example 2 - Alain: A: Ça sert à décrire quelque chose, à préciser la fonction du sujet. Il y en a besoin dans la phrase. (It's to describe something, to give the function of the subject. You need one in the sentence.) Example 3 - Valérie: V: On peut le conjuguer. Je m'en sers pour faire une phrase, s'il n'y a pas de verbe, cela n'irait pas. (You can conjugate it. I use it to make a sentence; if there is no verb, it won't work.)
Conclusion All the difference between intuition and the thought-out comprehension of how language functions can be found in the difference of the results between children of 5 - 6 years old who spontaneously reconstitute a verb when they hear a series of words that does not include one, and the child of 10 - 13 years old who can explain that a sentence must necessarily contain one. This tacit intuition of the set of rules that determines the grammaticality of sentences that speakers possess concerning their language, can be called, using Culioli's (1968) expression, epilinguistic com-
VERBS AS SENTENCE ORGANIZERS
107
petence (cf. Gombert 1986). This competence does not imply that subjects are able to state the rules they use when they speak or when they listen to others. On the other hand, when children become capable of speaking about their language, it is then possible to speak of metalinguistic abilities. It is obvious that schooling cer tainly plays an important role in the development of metalinguistic competence. Does, then, the passage from epilinguistic to metalinguistic ability present any difficulties as to the verb phrase? Some of the results presented here would seem to confirm this. Difficulties may be both semantic and syntactic. It was seen that children age 5 - 6 had difficulties coming up with context-free verbs; when they were asked to give a word, they only named objects. For children of this age a word cannot be used without its meaning. Stating a verb implies a transformation in the state of the world or it means taking a relation into consideration, which is very difficult for these children (cf. Sternberg 1984). A transformation process can have an effect; pouring a liquid from vase A into vase means that vase is full while vase A is empty. But the pouring itself has left no trace; it has only left a result which can be expressed in the form of properties of objects. Vase has only become full and vase A has become empty through the comparison of the previous state and the present state. Specifying a transformation, e.g., pouring the contents of vase A into vase B, or specifying a relation, e.g., X is supporting Y, independent of the objects implied in the process, presupposes a high level of abstraction. It is thus more difficult to give the meaning of a verb than the meaning of an object that is in front of one's eyes. And even if it cannot be seen, it is still easier to give, especial ly if it is a familiar one. All the same, these children had no difficulty making sen tences or restoring sentences correctly with the verbs that were used. What then, is the linguistic status of these verbs that were used but never thought up, and which, as it was seen, children often even refuse to consider as words altogether? It might be that for children age 5 - 6, a verb would only be some sort of appendage that is necessary to speak about things and would only be a way to qualify them, e.g., a door opens and closes just like it can be made of wood, red or blue, or for the bedroom or the kitchen. The verb would be some part of the name of the object; as one child said, "une porte ça s'ouvre" (a door opens). What then, are the conditions that permit children to get over these difficulties? First of all, learning to read teaches the child to segment written utterances into words. Secondly, the recognition of a word as a specific entity should help children to differentiate a word from its meaning. Finally, what children first learn about grammar teaches them to consider verbs as words that can be transformed more than any other words in the language. The consistency of a verb's function in spite of morphological transformations which express changes in aspect or temporal rela-
108
G. PIÉRAUT-LE BONNIEC AND J.P. PILLE
tions, may help children to discover what is particular to verbs as compared to other words. The importance of the role played by morphological learning in the development of the notion of a verb explains why verbs are considered first of all to be a word that takes on certain prototypical transformations, particularly those that correspond to the most regular, and most frequent type of conjugation. Once verbs have become objects that have specific forms and that can be dis cussed, their function in the sentence can be recognized and the verb can appear as the word that is necessary to give meaning to an utterance. As concerns verbs, metalinguistic competence develops out of epilinguistic competence.
Pronoun Assignment in the Processing of' Locally Ambiguous Sentences J. Seguy and M. Kail Laboratoire de Psychologie Expérimentale, Université René Descartes Centre National de Recherche Scientifique École Pratique des Hauts Études (3ème Section), Paris
The coexistence of the sequential nature of language and the limitations of psychological processing capacities constitutes one of the major problems that must be solved by every theory of language comprehension. Indeed, the linear organization of language frequently forces the subject to form hypotheses concerning the structure of the utterance being processed without suffi cient information to guarantee their correctness. In other words, in order to elaborate an on-line integrated representation of an utterance, the subject must take risks concerning its possible development. Recent proposals concerning different parsing strategies in language comprehen sion (Kimball 1973; Frazier 1978; Frazier and Fodor 1978; Frazier and Rayner 1982; Rayner, Carlson and Frazier 1983) have been advanced to tackle these problems. In the experiments presented here we studied the comprehension strategies which operate when processing sentences containing a pronoun whose coreferent is ambiguous. In these sentences the ambiguous coreference of the pronoun is resolved by the properties of the final adjective. Thus, for example, the difference in the interpretation of sentences (1) and (2) is determined by the fact that in (1) the final adjective assigns the subject of the main clause (the robber) as a pronominal referent; whereas in (2) it is the object of this clause (the jeweler) that is assigned. (1) (2)
The robber frightens the jeweler because he is armed. The robber frightens the jeweler because he is helpless.
110
J. SEGUY AND M. KAIL
Although the first part of the sentence which is shared by (1) and (2) contains the ambiguous coreference just mentioned, a great number of experiments have shown that a preferential interpretation does exist. Indeed, when subjects are asked to complete this part of the sentence with an adjective, most of them use an adjective which is intended to modify the subject of the main clause. Such a response bias seems to be directly related to certain seman tic properties of the verb of the main clause, such as the "causal valence" studied by Caramazza and his colleagues (Garvey and Caramazza 1974; Garvey, Caramazza and Yates 1974; Grober, Beardsley and Caramazza 1978; Caramazza and Gupta 1979). According to these authors, "this feature selects one noun or the other as the probable instigator or causal source for a series of events" (Caramazza, Grober and Garvey 1977:601). For certain verbs (VN1 verbs such as frighten, impress, consult...) the cause of the action is ascribed to the subject of the main clause. For other verbs (VN2 verbs such as: admire, congratulate, blame...) the cause of the action is ascribed to the ob ject of the main clause. Finally, there exists a third category of verbs (such as: wait, protect, question...) which do not produce any systematic bias towards either the subject or the object of the main clause. Kail (1979b) obtained results similar to those of Caramazza et al. with tasks involving the completion of utterances and coreferent assignment in French. The fact that the verb of the main clause directs the assignment of the pronominal referent should have implications for the perceptual analysis of this kind of utterance. In a study dealing with language comprehension, Caramazza et al. (1977) presented their subjects with sentences containing verbs of type VN1 and VN2. The sentences were built so that the subordinate clause could or could not be compatible with the semantic bias produced by the verb of the main clause. The authors used two types of sentences which we shall call: - first type - compatible sentences - second type - incompatible sentences Thus in the sentence: (3) John telephoned Bill because he wanted some information. the interpretation of the subordinate clause is compatible with that of the main clause, i.e. with the bias produced by the verb VN1. On the other hand, in sen tence (4),
ASSIGNMENT IN PROCESSING OF AMBIGUOUS SENTENCES
111
(4) John telephoned Bill because he withheld some information. the interpretation of the subordinate clause opposes the semantic bias produced by the verb of the main clause. In this experiment the subjects had to decide as quickly as possible whether the pronoun that followed "because" referred to the first or the second person men tioned in the main clause. The decision time was smaller with compatible sentences (3) as opposed to incompatible sentences (4). This result holds even when the presence of gender markers permit an unambiguous assignment of pronominal referent. Sentence (5) gives an example of such a case: (5)
Sue telephoned Bill because he withheld some information.
These results, which have been confirmed by Grober and Kellar (1981), suggest that there exists a difference in the global complexity of sentence analysis as a function of the compatibility between the sentence and the preferential reading strategy. All the same, this difference in complexity might be due to the type of ex perimental task used by these authors, in which the subject is explicitly asked to find the pronominal referent. Indeed, this could have introduced some ad hoc strategies which would not necessarily be those used in more natural comprehen sion tasks. The criticism could equally address Hirst and Brill's (1980) experiments on the role of contextual constraints in pronoun assignment. Moreover, their type of experimental procedure gives very delayed responses which reflect the result of reading strategies rather than their on-line processing. We need information about this on-line processing in order to answer the main question to be dealt with in the article, i.e., at what moment in the processing of these sentences does the subject resolve the local ambiguity, or in other words, assign the pronominal referent? Three hypotheses may be formulated concerning this question:
Hypothesis 1 Due to a lack of information necessary to definitively and unambiguously assign the pronominal referent when the pronoun was perceived, the subject waits for the end of the sentence before doing so.
112
J. SEGUY AND M. KAIL
Hypothesis 2 The subject calculates the two possible interpretations of the coreferences parallelly and decides which is correct when s/he arrives at the end of the sentence.
Hypothesis 3 As soon as the pronoun is encountered, the subject assigns the pronominal referent according to the interpretation derived from the preferential reading strategy induced by the causal valence of the verb of the main clause. In the following experiments we tried to test these three hypotheses, using diverse experimental paradigms in order to avoid the interpretation biases due to particular experimental tasks.
Experiment 1 In the first experiment we studied how sentences with local ambiguities are processed by using an acceptability judgment task. The sentences contained anaphoric pronouns which may be categorized as either "compatible" or "incom patible" with the preferential reading strategy determined by the semantic properties of the verb of the main clause. The use of a preferential reading strategy for the comprehension of this type of sentence should lead to shorter decision times for the "compatible" sentences com pared to the "incompatible" ones. The advantage of the acceptability judgement paradigm lies in that it doesn't attract the subject's attention to the variable being manipulated. Indeed, the judgement asked of the subject concerns the global mean ing of the sentence and not certain particular properties (such as the referential am biguity of the pronoun). The relevance of this type of procedure in subjects' utterance interpretations was confirmed by Holmes (1979) and Marquer, Le Nes tour, Lévenlé, and Welitz (1982).
Method Sixteen experimental sentences with the same syntactic structure were used; all of them were complex sentences composed of two clauses related by the conjunc-
ASSIGNMENT IN PROCESSING OF AMBIGUOUS SENTENCES
113
tion "because". Eight of the sentences contained a VN1 type verb, in other words a verb that biases the choice of the pronominal referent towards the subject of the main clause. These verbs were chosen from a group of 72 verbs whose interpreta tion biases where established using a completion task carried out by 120 subjects (Kail 1979b). The experimental verbs were the following: VN1:
choquer (shock) déranger (bother)
impressioner (impress) convaincre (convince)
agacer (irritate) émouvoir (touch emotionally)
effrayer (scare) influencer (influence)
VN2:
féliciter (congratulate) soigner (care for)
récompenser (reward) gronder (scold)
envier (envy) reconnaître (recognize)
punir (punish) admirer (admire)
Each of these verbs appears in two complementary sentences that differ only with respect to the final adjective. Depending on the properties of this adjective, the acceptable pragmatic interpretation of the sentence will or will not agree with the interpretation imposed by the verb of the main clause (compatible and incompatible sentences respectively). Sentences (1) and (2) are an example of a pair of sentences with a VN1 type verb. Sentences (6) and (7) presented below are an example of a VN2 type pair. (6) The usherette envies the actress because she is elegant. (compatible sentence) (7) The usherette envies the actress because she is jealous. (incompatible sentence) Besides the 16 experimental sentences, 12 unacceptable sentences with the same syntactic structure as the experimental sentences were used. These sentences were unacceptable in the sense that the final adjective could not be related to either of the noun phrases of the sentence (e.g. "The girl looked at the woman because she was liquid."). These sentences are semantically anomalous. Two experimental lists were used in order to counterbalance the presentation of the sentence pairs so that each subject was only given one version (either incom patible or compatible) of each pair. Each list was composed of 16 experimental
114
J. SEGUY AND M. KAIL
sentences, 8 of which were VN1 type and the other 8 were VN2 type. Half of each type were compatible and the remaining half were incompatible. These sentences, plus six trial sentences used to familiarize the subject with the task, were the same for the two lists.
Procedure The sentences were presented visually on a Hewlett-Packard screen controlled by a computer. Each sentence remained on the screen until the subject responded by pressing either the "yes" or "no" key depending on whether he considered the sentence to be acceptable or not. The "yes" response was associated with the subject's dominant hand. Response and reaction times were recorded by the computer (reaction times in clude the reading time). The 28 sentences were presented in a random order which was different for each subject. The subjects were instructed to decide as quickly as possible whether the sentences were acceptable or not.
Subjects Two independent groups of 20 subjects were used for the two experimental lists, L1 and L2. The subjects were all psychology students at the University of Paris V.
Results The results of four subjects (two from each list) were excluded from the analysis, as these subjects had considered all the incompatible sentences as unac ceptable. The results presented in Table 1 (reaction time in cs. and error percent age) are those of the 18 subjects who gave a minimum of 2 correct answers (out of the four possible) for each sentence category. As we can see in Table 1, the compatible sentences produced less errors and shorter reactions times than the incompatible sentences. There was a significantly greater number of errors recorded for the incompatible sentences as compared to the compatible sentences (F(l - 16) - 21.8 ; p < .001). An analysis of variance of the reaction times revealed no effect of the intergroup factor (experimental list) and no interaction between this and the other experimental factors. A comparison be-
ASSIGNMENT IN PROCESSING OF AMBIGUOUS SENTENCES
115
Table 1
Verb Type
Sentence Type
Compatible Incompatible
VN1
VN2
352
341
(9.7)
(3.7)
441
421
(23.5)
(20.1)
Reaction times for acceptability judgements (in cs.) as a function of sentence type (compatible vs. incompatible) and verb type (VNl vs. VN2). The number in parentheses indicates the per centage of errors.
tween compatible and incompatible sentences revealed a significant difference (F(l - 16) = 40.6 ; p < .0005). On the other hand, the type of verb (VNl, VN2) had no significant effect (F < 1), and the interaction of these two factors was not sig nificant.
Discussion The results of this experiment support the third hypothesis which predicted that attribution of the pronominal referent takes place according to the preferential inter pretation indicated by the semantic properties of the verb of the main clause as soon as the pronoun is processed. In fact the large number of errors (greater than 20%) observed with the incompatible sentences and the fact that these were judged as unacceptable by four of the subjects (excluded from the analysis, as indicated above), reveals the difficulty involved in adopting a nonpreferential reading strategy. Also, the significantly greater reaction times for the incompatible sentences sug gest that the subjects have reprocessed these sentences in order to give them an adequate pragmatic interpretation. Hypothesis 3 therefore seems to explain the dif ference in the processing complexity of compatible and incompatible sentences. This difference does not appear to be the result of ad hoc strategies induced by the experimental task, but of more general strategies used in sentence comprehension.
116
J. SEGUY AND M. KAIL
All the same, the experimental conditions adopted here, in particular the fact that the subject must process the sentences as quickly as possible, might have privileged a simple preferential reading rather than the parallel use of two reading strategies as postulated by hypothesis 2. Experiment 2 was set up to test this possibility by using an experimental situa tion that encourages comprehensive processing of the different interpretations of the sentence.
Experiment 2
Method The material used in this experiment consisted of the same two lists, L1 and L2, as in experiment 1.
Procedure The sentences were presented visually on a Hewlett-Packard screen controlled by a computer. Each sentence was presented in two steps: first the whole sentence minus the final adjective was presented. This remained on the screen for three seconds, after which the final adjective was presented. The adjective remained on the screen until the subject responded. The subject had to judge whether the adjec tive could or could not represent the final element of the sentence by pushing on the "yes" or the "no" key. The "yes" response was associated with the dominant hand of the subject. A warning signal was displayed for half a second before the presentation of each sentence. A gap of three seconds was left between the subject's response and the warning signal for the next sentence.
Subjects Two independent groups of twelve subjects were given the two experimental lists L1 and L2. All the subjects were psychology students at the University of Paris V.
ASSIGNMENT IN PROCESSING OF AMBIGUOUS SENTENCES
117
Results Table 2 shows the results of this experiment which appear to agree with those of experiment I. There were more errors with the incompatible sentences than with the compatible sentences (F(l - 22) = 19.8 ; p < .001). An analysis of variance of the reaction times revealed a significant effect of sen tence type: (compatible/incompatible) (F(l - 22) = 15.33 ; p. < .001). On the other hand, there is no significant effect for the type of verb (F(l - 22) = 3.6 ; p > .10), and the interaction between these two factors is not significant.
Table 2 Verb Type
Sentence Type
Compatible Incompatible
VN1
VN2
147
143
(7.2)
(4.2)
198
177
(24)
(16.6)
Reaction times (in cs.) for the final adjective as a function of sentence type (compatible vs. incompatible) and verb type (VN1 vs. VN2). The numbers in parentheses indicate the per centage of errors.
Discussion These results indicate that even when the processing time available is long enough (theoretically) to allow the subject to work out two interpretations of the sentence, only the preferential interpretation is actually calculated. A complemen tary analysis is carried out only when the final adjective calls this interpretation into question. In this experiment we asked the subjects to judge whether the test-word could or could not constitute the last word of the previously presented sentence fragment. It
118
J. SEGUY AND M. KAIL
is clear that in order to make such a judgement the subject must relate the proper ties of the test-word to the global interpretation of to the remainder of the sentence. The test-word will or will not be able to adequately integrate the rest of the sen tence depending on how the subject reads the context. In the third experiment presented below we examined this last point by using a task where the subject need not consider compatibility between the context and the test-word.
Experiment 3 In this experiment we studied the effect of context on the lexical decision times for the final adjective of the sentences. As already mentioned, deciding whether the last element of the sentence is or is not a word does not a priori necessitate the conscious construction of relations between the properties of the test-item and those of the context. All the same, a large number of studies have shown that sentence context does influence the recognition of the test-item in this type of experimental situation. Tulving and Gold (1963) and Morton (1969) observed that the tachistoscopic recognition threshold of a word is much lower if this word is semantically con gruent with its context than if it is not. Also, the recognition thresholds of high transition probability words are lower than those for low transition probability words. Later on, the influence of sentence context on the recognition of a test-word was demonstrated using a variety of experimental procedures such as the reading latency of the word (Stanovich and West 1981a, b; West and Stanovich 1982, 1986; Forster 1981), lexical decision times (Schubert and Eimas 1977; Fischler and Bloom 1979, 1980; Kleiman 1980), and phoneme detection (Morton and Long 1976; Foss and Blank 1980). However, while the role of sentence context in the recognition of a test word ap pears to be well established, the exact nature of its effects and its interpretation are still controversial issues. In particular, the facilitation and inhibition effects of con text on the recognition of a test-word appear to be directly dependent on the type of experimental procedure used. Indeed, facilitatory effects are greater in reading tasks than in lexical decision tasks, whereas the inverse is true for inhibition effects (on this particular point see Forster 1981; and West and Stanovich 1982). This last point suggests that the lexical decision concerning a test-word may be influenced
ASSIGNMENT IN PROCESSING OF AMBIGUOUS SENTENCES
119
by processes which would appear to occur after lexical access. In particular, longer response times to a test-word in the present case would result form the difficulty to integrate it into the remainder of the sentence (Cairns, Cowart and Jablon 1981). Considering our experimental material, the presence of an inhibition effect on the lexical decision times for the test words which oppose the preferential reading of the context sentence, would constitute important empirical support for hypothesis 3. In fact, this result would indicate that during the first stage of processing, subjects set up a representation of the sentence into which the incompatible test-word can not be integrated.
Method Sixteen sentence-test word pairs were used which were the same as the ex perimental sentences in our two previous experiments. However, we changed the final adjective in some of the sentences in order to match the adjective pairs as sociated to the same context-sentence in terms of length and frequency of use. As before, two lists of 16 items were each constructed in order to balance the presenta tion of the final adjectives associated with each sentence context among the sub jects. Along with the 16 experimental items, a supplementary list of 16 context-sentence and test-item (nonsense words that obeyed French graphophonological rules) couples were constructed. The context-sentences were of the same type as those of the experimental items and they were the same for the two experimental lists. Each experimental list contained 40 items including 8 trial items which were of the same type as the context-sentences used for the experimental items and which were associated with a (French) word for 4 of the items and a nonsense word for the 4 others. The 32 experimental items were composed of 8 compatible items and 8 incompatible items (4 VN1 and 4 VN2 in each case) and 16 items with a nonsense word as the final element. Two control lists containing 60 words and 60 nonsense words plus 10 trial items (5 words and 5 nonsense words) were also constructed. Among the 60 words of each of the control lists were the 16 final adjectives (test-words) of the L1 and L2 experimental sentences. These con trol lists were introduced in order to check that in a neutral situation, i.e. without a context-sentence, the lexical decision times for the final adjectives in the com patible and incompatible experimental items were comparable. Such a control is necessary if we are to be able to attribute any differences observed among the testwords to the effects of the context-sentence.
120
J. SEGUY AND M. KAIL
Procedure The experiment was divided into two blocks of trials separated by a 15-minute interval. Each block corresponded to the presentation of one of the experimental lists (L1 or L2) and the corresponding control list (LC1 or LC2). Half of the sub jects were presented with the experimental list first followed by the control list, and the other subjects were presented the lists in reverse order. The experimental items were presented visually on a screen linked to a computer. The context-sentence was presented for two seconds and the test item (word or nonword) was presented after a 300 ms. interval. The subject had to decide whether the sequence of letters that followed the context-sentence represented a word or not. The subject responded by pressing one of two buttons associated with "yes" and "no" answers. As for the other experiments, the "yes" response was associated with the dominant hand of the subject.
Subjects 48 psychology students at the University of Paris V participated in the experi ment. 4 groups of 12 subjects were each assigned to 4 conditions resulting from the different combinations of the list factor and the order of presentation (experimental list - control list).
Results Table 3 shows the mean response times (in ms.) which correspond to the 4 situations which result from crossing the two main factors: verb type and com patibility. The control list reaction times are indicated in parentheses. The error percentages obtained were very low: 1.6% for the compatible testwords and 1.3% for the incompatible test-words. As is indicated in Table 3, the compatible test-words gave lower reaction times than the incompatible test-words. This difference was larger for items with VN1 type verbs than for those with VN2 type verbs. An analysis of variance indicated a significant effect of the main factors: 1) verb type (VN1 vs. VN2, F(l - 46) = 4.96 ; p < .005) and 2) compatibility between word and context (compatibility vs. incompatibility, (F(l - 46) = 10.86 ; p < .005).
ASSIGNMENT IN PROCESSING OF AMBIGUOUS SENTENCES
121
Table 3 Verb Type VN1 Sentence Type
Compatible Incompatible
VN2
651
629
(651)
(627)
752
688
(652)
(629)
Lexical decision times (in ms.) for thefinalad jective as a function of sentence type (compati ble vs. incompatible) and verb type (VN1 vs. VN2). The numbers in parentheses indicate lexical decision times for the experimental ad jectives of the control lists.
There was no significant interaction between these factors (F < 1). Specific com parisons indicated that the effect of the compatibility factor was significant for both types of verbs (F(l - 46) = 11.18 ; p. < .005 for type VN1, and F(l - 46) = 4.09 ; p < .05 for type VN2). It must be emphasized that the differences obtained for the various types of con texts cannot be assigned to the properties inherent in the test-words which are as sociated with them. In fact, when the same test-words are presented in isolation (control list), no difference is observed among them (651 ms. vs. 652 ms. for type VN1 and 627 ms. vs. 629 ms. for type VN2).
Discussion The results obtained in this experiment clearly suggest that the subjects read the context sentence in one manner only, conforming to the direction induced by the verb of the main clause. When the test-word is incompatible with this preferential analysis of the context-sentence, its recognition is inhibited. The inhibition seems to be due to the difficulty the subject encounters while trying to integrate the testword into the remainder of the sentence. According to this interpretation, context does not play a role in the actual access of the test word, but rather in its integra tion into the global meaning of the context.
122
J. SEGUY AND M. KAIL
In fact, the difference in lexical decision times for compatible and incompatible test-words is analogous to that observed by other authors for sentences containing a congruent test-word and for sentences containing a test-word that renders them anomalous. For example, Fischler and Bloom (1979) who used an experimental procedure identical to ours, observed (in their experiments 1 and 4) differences varying from 138 ms. to 73 ms. between the congruent and the anomalous test-words, depending on the transition probabilities of the congruent test-words. We did not directly manipulate this last parameter in the present study, but it is quite plausible that the compatible final adjectives have a transition probability that lies between the high and low transition probabilities for Fischler and Bloom's congruent words. Differences of the same order between items that are congruent or incongruent with the context-sentence were also obtained by Forster (1981) and West and Stanovich (1982). Even if the control situation in our experiment was essentially aimed at linking the differences obtained between the compatible and incompatible words to the ef fect of the context and not to the properties inherent in these items, we can never theless mention here a number of points where this control situation could be considered to be a neutral context. From this point of view, it should be noted that the lexical decision times for the words presented in isolation in the control list are very similar to those observed where these words are compatible with the context-sentence, with this being so for the two types of context (VN1 and VN2). The context-sentence does not seem therefore to have facilitated the recognition of the compatible test-word. On the other hand, the same comparison made with the incompatible test-words clearly in dicates that the context produces a strong inhibition effect. These results agree with those obtained by Fischler and Bloom (1979) and in particular Forster (1981). Ex actly as in our experiment these authors observed very strong inhibition effects whereas the facilitation effects only occurred for words with a very high transition probability.
General Discussion The principle aim of the experiments presented in this article was to find out how subjects process locally ambiguous sentences containing an anaphoric pronoun. Theoretically three different strategies are possible: 1) subjects do not definitively assign pronominal referents until they possess all the information con-
ASSIGNMENT IN PROCESSING OF AMBIGUOUS SENTENCES
123
veyed by the sentence; 2) subjects assign two pronominal referents during sentence processing (hypotheses 1 and 2, above); 3) subjects immediately assign the pronominal referent in accordance with the interpretation bias induced by the semantic properties of the verb of the main clause. This third possibility is supported by the results of an experiment conducted by Dell, McKoon and Ratcliff (1983). Using a priming technique, these authors managed to show that the presentation of the pronoun activated the representation of its referent within a time lapse of 250 ms. Moreover, the activation was not limited to the referent but extended to the rest of the clause in which it was situated. This is also supported by our own work as presented here. Experiment 1 showed that sentences which are compatible with the preferential pronoun assignment strategy are judged acceptable more often and more quickly than the sentences in compatible with this strategy. Experiment 2 suggested that even when subjects have enough time to carry out the two possible readings of the first part of the sentence, only the preferential reading is actually accomplished. Finally, experiment 3 con firms this interpretation by showing that lexical decision times are longer when the test-word contradicts a subject's interpretation of the preceding context. The lexical decision time for a test-word presented after an unfinished sentence seems therefore to indicate the difficulty of integrating this test-word into the sentence. Automatic calculation of the compatibility between the "candidate" words and the preceding context takes place at this level of processing and inhibits identification of the testword. The results as a whole support the hypothesis that a unique referent is immedi ately assigned to a potentially ambiguous pronoun. Pronoun assignment, therefore, depends directly on the semantic properties of the clause the referent belongs to.
Conjunctions: Developmental Issues M. Kail Laboratoire de Psychologie Expérimentale, Université René Descartes Centre National de Recherche Scientifique École Pratique des Hauts Études (3ème Section), Paris
and J. Weissenborn Max-Planck Institut für Psycholinguistik, Nijmegen
Preliminary Considerations One of the fundamental characteristics of human language is its recursiveness. Recursiveness allows for the combination of sentences or parts of sentences and is thus the basis of the syntactic complexity of an utterance. This recursiveness is made possible, in part, by certain lexical elements, connectives, of which conjunc tions constitute the most important class. The ontogenesis of connectives is thus in separable from that of complex sentences and so the majority of studies on the development of connectives have been carried out on conjunctions. Logical reason ing is another perspective in which certain connectives such as AND, OR and IF have been the object of developmental study. In this article we do not intend to review the studies done on connectives in these two perspectives. Partial syntheses of psycholinguistic studies can be found in Limber (1973), Bowennan (1979), and Bloom, Lahey, Hood, Lifter and Fiess (1980), French (1981), and French and Nelson (1985) and a general survey of studies using a logical perspective appears in Falmagne (1975) and Paris (1975). What we would specifically like to do here is isolate several aspects which have been neglected in previous studies and which nevertheless are indispensable for a better understanding of the ontogenesis of these elements. Our study will focus on conjunctions (e.g., BEFORE, AFTER, BECAUSE, BUT...) with no mention of syntactic connectives (e.g., THAT). Every study of language acquisition, it seems, needs to answer at least two ques tions: what is acquired by the child and how is it acquired? With the understanding that the second question presupposes the first. However, to get valid answers to
126
M. KAIL AND J. WEISSENBORN
these two questions it is first necessary to adequately analyze the linguistic proper ties of the phenomena under discussion. Very often the linguistic analyses underlying developmental studies of connec tives are much too cursory to advance the understanding of their acquisition. This has been a problem with analyses of the French causal conjunctions CAR (FOR, AS), PUISQUE (SINCE, SEEING THAT), and PARCE QUE (BECAUSE). Apparently the obser vations that causal conjunctions are mastered after additive coordination conjunc tions (AND, OR) and temporal conjunctions (and, then) (Clancy, Jacobson and Silva 1976; Bowerman 1979; Bloom et al. 1980) masks the reality of their differences. Apart from the common expression of a causal relationship, the syntactic and semantic-pragmatic use conditions for each of them are so distinct that the question of their differential acquisition is unavoidable. Thus, as has been noted by Barbault, Ducrot, Dufour, Espagnon, Israel and Manesse (1975) and Ducrot (1972), at the syntactic level an utterance "PARCE QUE (BECAUSE) Q" can among other things be questioned, denied or embedded, whereas the same operations are excluded in the case of utterances with CAR and PUISQUE. Thus, note the ungrammatically of (2) and (3): (1) Est-ce que la chaudière a éclaté PARCE Qu'il a gelé? (Did the boiler explode BECAUSE it froze?) (2) * Est-ce que la chaudière a éclaté PUlSQU'il a gelé? (Did the boiler explode SINCE it froze?) (3) *Est-ce que la chaudière a éclaté CAR il a gelél (Did the boiler explode FOR it froze?) These syntactic use conditions depend on the semantic-pragmatic functions of PARCE QUE, on the one hand, and CAR and PUISQUE on the other. Whereas PARCE QUE can be treated as an operator analogous to AND and OR in propositional logic, adding to a basic "connection between two ideas P and Q, a new idea of causal relationship between P and Q", CAR and PUISQUE serve "to accomplish an act of speech (justification, inference). They mark that the speaker accomplishes a par ticular speech activity at a given moment by means of the ideas" (Barbault et al. 1975:254). From this analysis results a number of questions for a developmental study of these elements: - How does a native French-speaking child arrive at a differentiation of these functions?
CONJUNCTIONS: DEVELOPMENTAL ISSUES
127
- Can a hierarchical relationship be determined among these functions such that one would be more fundamental than the others? If so, what are the implica tions for the order of acquisition of these functions? - What is the relationship between the acquisition of these discourse functors and of other functors, which, like MÊME (EVEN), SEULEMENT (ONLY), and AUSSI
(ALSO), also fulfill a discourse function? However, there are other consequences of such an analysis which go beyond the context of the acquisition of a particular language. A correspondence can be noted between the system of causal conjunctions of French and that of German. The syn tactic and semantic-pragmatic use conditions of the causal conjunction WEIL (BE CAUSE), DA (SINCE) and DENN (FOR) are apparently identical in the two languages (Pasch 1982). Thus sentences (5) and (6) are ungrammatical: (4) Ist der Heizkessel geplatzt, WEIL es gefroren hat? (Did the boiler explode BECAUSE it froze?) (5) *Ist der Heizkessel geplatzt, DA es gefroren hat? (Did the boiler explode SINCE it froze?) (6) *Ist der Heizkessel geplatzt, DENN es hat gefroren? (Did the boiler explode FOR it froze?) This comparison with German corroborates the observation of the close relation ship between certain syntactic use conditions and the discourse functions of these conjunctions. One is thus led to ask whether this parallelism is or is not present from the first appearance of causal connectives in the child's language. The response to this question may help in understanding the interaction of syntactic and semantic factors in acquisition differently from how it has traditionally be under stood. Furthermore, a comparative study of developmental hypotheses formulated with regard to French causal conjunctions would help to differentiate between more general processes of acquisition, based on either cognitive or linguistic constraints. Similarly, this comparative method allows for the evolution of the influence on the acquisition process of syntactic properties which are independent of semantic fac tors. For example, German DA (SINCE) and WEIL (BECAUSE) are associated with an SOV structure, whereas DENN (FOR) is associated with an SVO structure. There is no apparent semantic reason for this. Such an analysis also has repercussions for the study of the acquisition of causal conjunctions in languages such as English and Italian, where these connectives are distributed differently (see, for example, Clancy et al. 1976). The question concerns
128
M. KAIL AND J. WEISSENBORN
what influence these linguistic differences may have on the course of acquisition. We examine this question in the second part of this paper, where we present an analysis of the connective MAIS (BUT). In the following section, we will examine the explanations proposed to account for the order of emergence of different conjunctions, the interconnection between the mastery of conjunctions and that of the utterances in which they occur and some general models proposed to explain their acquisition.
Specific Problems in Developmental Studies of Conjunctions
Order of Acquisition It is possible to study the order in which conjunctions are acquired in terms of either the first appearance of a specific conceptual relation or the complete under standing of all the conjunctions that express this relation. This order may not coin cide. For example, BECAUSE is the first of the conjunctions expressing causality to emerge, and it does so quite early. Other members of the class, e.g. SINCE, THERE FORE, etc., do not emerge until much later. Taking the first exemplar in spon taneous production as evidence for some basic notion of the relation it expresses, we find, with some variation the following order of appearance: Type of Relation: additive coordination (AND) temporal sequence (THEN) causal (BECAUSE) adversative (BUT) conditional (IF) 5. concessive (ALTHOUGH)
1. 2. 3. 4.
This order remains essentially stable across languages (Stern and Stern 1928; Bloom et al. 1980; Clark 1985), beginning at about age 2 and finishing at about 5. Three different types of explanations of acquisition order have been proposed: linguistic, cognitive and "discursive" (discourse-oriented). They have in common a formulation in terms of the complexity of processing operations. These explana tions will be examined in the light of some specific cases.
CONJUNCTIONS: DEVELOPMENTAL ISSUES
129
With regard to explanation in terms of linguistic complexity, it should be remembered that the two dimensions most frequently evoked to account for the order of emergence are the greater or lesser semantic complexity of the conjunc tions in question or the greater or lesser complexity of the syntactic structures in which they occur. Semantic complexity is often described in terms of features, for which we will later present an example concerning the acquisition of BEFORE and AFTER. The syntactic complexity of a structure has been most often defined in terms of its derivational history, i.e., the complexity of a sentence is considered to be a func tion of the number of grammatical rules implied in its generation. This suggests that the grammatical model is considered to be isomorphic with a model of lan guage production and comprehension. The weak point in this purely quantitative (in terms of numbers of rules) perspective in that it presupposes an equivalence of all these rules when processing. To remedy this, Brown and Hanlon (1970) have proposed that only structures generated by the same rule (with the exception of one or two) be compared. It is thus, for example, that the interrogative (a) "Is he com ing?" is considered to be more complex than the corresponding affirmative (b) "He is coming", because (a) would be derived from (b) by the application of a sup plementary rule of auxiliary displacement. This means that for a model of language acquisition, the most grammatically complex structures will be mastered after the simpler corresponding structures. It is thus, on the basis of a grammatical model deriving coordinated structures of the type, "Pierre and Michèle are dancing" from a coordinated sentence structure, "Pierre is dancing and Michèle is dancing" (by application of a rule which deletes identical elements), that the hypothesis can be formulated that AND as a conjunc tion of sentences will be mastered before AND as a conjunction of sentential ele ments. In a series of longitudinal as well as experimental studies, Lust and her col leagues have tried to verify this hypothesis (Lust and Mercis 1980; Lust 1981). These results have been questioned by other studies (Ardery 1979, 1980; TagerFlusberg, DeVillersa and Hakuta 1982) which have found that the coordination of sentence-internal constituents precedes rather than follows the coordination of com plete sentences. These conflicting results suggest a processing model in which coordinated constituents (e.g., NPs, VPs) are generated directly and are not the result of the deletion rules applied to sentences containing them. Evaluating such contradictory results as these is difficult on the basis of the data alone, because the models assume different principles of syntactic complexity. To
130
M. KAIL AND J. WEISSENBORN
decide among them, one would first have to agree on which assumptions are cor rect. Whether they attribute greater importance to one or the other dimension, the majority of developmental studies are sensitive to the interaction of syntax and semantics (including the meaning of propositions and conjunctions). However, this interaction is usually considered to be secondary to conceptual determinants which underlie the acquisition process. After analyzing productions of children 1:7-3 years old according to a typology based on their fundamental semantic relations (additive, temporal, causal, adversa tive, epistemic, object specification or notification relational), Bloom et al. (1980) invoke a general principle of functional implication to explain the observed developmental sequence: additive < temporal < causal < adversative. It should be noted that the later acquired conjunctions possess the properties of earlier ones: temporal, causal and adversative conjunctions are additive (i.e., they refer to the connection of two events); causals are at the same time additive and temporal. While certain adversatives are additive, temporal and quasi-causal, adversatives also include a dimension of opposition or contrast. According to Bloom et al., this principle of semantic plurality has a corresponding parallel in the child's conceptual development (cf. Brainerd 1978, for a discussion of the principle of cumulative complexity in cognitive development). These authors recall that, as Sinclair (1970) has shown, the child is capable of grouping objects together before being able to organize these objects into a relation of seriation. They find an analogy in their own study, where children begin to speak of things or events which go together (additive conjunction) before speaking of things or events which go together in an ordered relationship. However, it must be said that the explanations advanced remain rather tentative and interpretive. Moreover, it seems that depending on the method of data analysis, this developmental sequence presents variation. Thus Clancy et al. (1976) have proposed a somewhat different order from that of Bloom et al. after analyzing children's productions in four languages (English, Italian, Turkish and German). These analyses include all sentences expressing "complex semantic contents" whether or not conjunctions are present. In the four languages the order of emer gence of semantic notions is identical (findings in agreement with Slobin's (1973) claim that in spite of formal variations among languages, the order of emergence of semantic notions is constant). The order found by Clancy et al. (1976) is that the first juxtaposed sentences express notions of symmetric coordination, followed by antithesis (adversatives), and then sequentially (temporal conjunctions) and causality. After these appear the first notions of conditionality emerge. The authors
CONJUNCTIONS: DEVELOPMENTAL ISSUES
131
say that an underlying cognitive basis is the source of this cross-linguistic regularity; however this is insufficiently supported. Because the study of cognitive determinants of language functioning have been somewhat disappointing, it has been proposed to analyze the development of con junctions in terms of discourse. Halliday and Hasan (1976) have studied conjunc tions in these terms in written texts, claiming that five principles assure discourse cohesion: a) reference, b) substitution, c) ellipsis, d) lexical cohesion, and e) con junction. In this framework, Eisenberg (1980) found the classical order AND, BE CAUSE, BUT, THEN, and SO. The analysis which she presents is very different from that of Bloom et al. Ac cording to these authors, in 90% of additive and temporal conjunctions and 80% of causal and adversative conjunctions, children attempt to link together their own productions. (This is contrary to Greenfield and Smith (1976), and Bates (1976) who suggest that children begin to produce complex sentences to connect their dis course to that of the adult. Bates notes that children's first uses of BUT and BE CAUSE are anaphoric and constitute commentaries on adult productions.) Eisenberg's data (and ours as well, cf. p. 000) confirm the idea that children first use these forms to make their own discourse coherent. According to Eisenberg, the majority of productions containing AND, THEN, and SO involve the child's own dis course cohesion, and less than 15% are connected with the discourse of the adult. Similarly, BECAUSE, although used at an early age in responses to questions, is mainly produced within an utterance of the child's. In contrast, BUT is the first con junction to initially fulfil an interactive function before fulfilling an autoregulative function, mainly to mark disappointment in relation to an expectation, or a "contrastive opposition" with respect to an interlocutor. While Bloom et al.'s findings are compatible with Eisenberg's, their conclusion, which claims that the emergence of conjunctions is determined by constraints rela tive to the form of the message rather than those introduced by the discourse situa tion, must be re-examined. We return to this issue below. Eisenberg's analysis highlights the necessity to stop viewing conjunctions on a unidimensional continuum of complexity, one whose contents vary with the dif ferent authors who do this. (This point is also made by Trosborg (1982) and KellerCohen (1987)). The emergence of the various conjunctions does not really seem to have a common origin, since each one involves different factors in the domains of linguistic knowledge, knowledge of the world, and discourse abilities that the others do not. One further demonstration of this point is made by Wing and Scholnick (1981) who show that difficulty in mastering certain conjunctions (BECAUSE,
132
M. KAIL AND J. WEISSENBORN
IF, ALTHOUGH, and UNLESS) is a function of the evolution of metadiscourse abilities, and in particular, the child's ability to judge the beliefs of a speaker rela tive to both the propositional content of the utterance and the nature of the interpropositional connections. Conjunctions such as BECAUSE, which presuppose uniquely positive beliefs, are easier for 6- to 11-year-olds than conjunctions such as UNLESS, which presuppose beliefs which are not positive. These pragmatic dimensions also interact with uncertainty and familiarity, as Scholnick and Wing (1982) have shown. Whether the material is familiar or not ef fects the evaluation of propositional truth only when the conjunction conveys un certainty. Wing and Scholnick (1986) have also found considerable linguistic growth during adolescence, as regards material involving factuals and counterfac tuals as well as uncertainty. They explained this in terms of two factors that under lie sentence interpretation: information-processing load (counterfactuals requiring juggling several levels of meaning at once) and the nature of the inference required (it may be easier to infer the external referent for uncertainty sentences than for counterfactuals). An example illustrating the difficulty of separating out the child's lexical ability from syntactic knowledge about complex sentences is provided by the controversy in the literature on BEFORE and AFTER. The point of departure, supported by her empirical research, is the theory proposed by E. Clark (1971) concerning the exist ence of different levels of semantic knowledge. According to this theory, two main factors can account for the order of acquisition of word meanings. - a principle of semantic complexity, which claims that in a given semantic field the child acquires the meanings of simple words before complex words. Simplicity is defined by Clark's Semantic Feature Hypothesis. For example BEFORE has a positive value for the feature [prior] whereas AFTER has a nega tive one. The hypothesis is that the positive features are simpler and, thus, ac quired earlier. - nonlinguistic strategies by which children identify the order of events with the order of enunciation. Having learned that BEFORE and AFTER refer to time, both are assumed to have the same meaning, that determined by the order of enunciation. This makes the right predictions for cases like "Mary ate breakfast BEFORE she went to school." but not for "Mary ate breakfast AFTER she went to school." Nonreplication of the findings of the conflation of enunciation and event orders have led many investigators to claim that children's difficulties are more syntactic than semantic in nature. Amidon and Carey (1972) and Barrie-Blackley (1973), finding errors committed by 5- and 6-year-olds in an act-out task involving events
CONJUNCTIONS: DEVELOPMENTAL ISSUES
133
mentioned in a subordinate clause, suggest that the event described in the main clause is more accessible. Other studies (Harner 1976; Coker 1978; Hood and Bloom 1979) insist on syntactic determinants in the comprehension of temporal connectives. Moreover, many studies mention a bias introduced by the nature of Clark's task (Johnson 1975; Amidon 1976; Hood 1977; French and Brown 1977; Flores d'Arcais 1978, 1981; Silva 1984). For example, the classical act-out tasks are inadequate if one is interested in the meaning of the connectives and the dif ferences between them, insofar as they constitute a specific domain of the lexicon of the child. Using different experimental methods, lexical decision tasks, with 7and 12-year-old Italian and Dutch children Flores d'Arcais (1978, 1981) showed that BECAUSE is the first conjunction to be clearly differentiated, followed by BEFORE and AFTER. SO THAT and IN ORDER TO are progressively differentiated from those with which they are often confounded. Lastly come UNLESS and AL THOUGH.
Task Biases and Interpreting Results While such research sheds new light on a domain in which children's knowledge of conjunctions is too often inferred from their comprehension of sen tences in which they are inserted, the problem remains that the highly metalinguis tic nature of the task does not allow us to get at the implicit knowledge at work in the linguistic differentiation of conjunctions. What is particularly striking in the results of the research devoted to conjunctions is the significant time interval be tween the appearance of diverse types of relations and forms in the child's spon taneous production as compared to their mastery in experimental production tasks (Gallivan 1982; French 1986; French and Nelson 1985; McCabe, Evely, Abramovitch, Corter and Pepler 1983; McCabe and Peterson 1985; Peterson and McCabe 1987). In this regard, Hood's (1977) study of the production of the con nectives BECAUSE and SO is very instructive and confirms the necessity of compar ing comprehension and production data. Hood shows, in fact, that children who tend to produce the order cause-effect in juxtaposed sentences, acquire SO first, whereas children who produce the order effect-cause use BECAUSE first. The two connectives appear simultaneously in the speech of children who show no order preference. Similarly, in oral descriptions of events required from children (4 to 8), Fayol and Mouchon (1988) showed that from 4 years old, children are able to produce conjunctions such as MAIS (BECAUSE). These productions seemed to depend on the specificities of the sequence of events and not on linguistic
134
M. KAIL AND J. WEISSENBORN
knowledge. Inasmuch as similar discrepancies between comprehension and produc tion arise in other areas of language development (such as the acquisition of spatial prepositions, Weissenborn 1981), they present a more general problem which goes beyond the scope of this study. In general, experimental tasks seem to tap the child's metalinguistic abilities, which develop much later. As French (1981) noticed in evaluating comprehension studies, an important issue is the problem interpreting the child's answer. While a correct answer generally can be taken as indicating that the child knows the connective, an incor rect answer does not necessarily mean that the child lacks this knowledge. It may simply be the result of performance factors; e.g. insufficient attention, misunderstanding of the task, etc. It also seems that in these studies the existence of systematic comprehension strategies has been overemphasized. When analyses of individual responses are carried out, they provide evidence of great variability (e.g., French and Brown 1977). It should be noted, moreover, that, as Cromer (1976) remarks, the concept of "strategy", fruitful as it may be, runs the risk of eluding the question of how strategies are abandoned once the child has mastered the necessary conceptual and linguistic knowledge. And as Clark (1983) points out, it is difficult to determine what a child's spontaneous productions mean, for example, whether they correspond to adult usage or whether they are unanalyzed formulas. In order to overcome some of the difficulties of interpreting comprehension and production results, French and Nelson (1985) invented a method where they ask children aged 2:11 to 5:6 to describe activities they have participated in. The outof-context descriptions of birthday parties, getting dressed, etc., they claim, directly tap the children's mental representations ("General Event Representations" or "Scripts") of these familiar events. A[???] to these anything the early emergence of relational terms like BEFORE, AFTER, BECAUSE, SO and BUT is related to the facilitating role of this kind of memory representation of familiar events. That is, the relations expressed by these connectives are already analyzed in the underlying event representations they refer to. French and Nelson's model assumes the steps of acquisition of relational terms to be as follows: initially, children can use and understand relational terms only in descriptions of familiar events for which they already have mental representations. Gradually their knowledge of relational terms becomes decontextualized, until eventually, they can use them to express and comprehend new relationships. This model does not specify the mechanisms that are responsible for the early mapping of conjunctions to familiar contexts. It also suggests that conjunctions are correctly used from the beginning. In addition, the model does not explain how relations not yet known by the child are derived from those that are known. The in-
CONJUNCTIONS: DEVELOPMENTAL ISSUES
135
terest of this study, in contrast to componential semantic models, which do not ac count for the child's precocious productions, lies in its use of cognitive repre sentations.
The Acquisition of BUT: A Cross-linguistic Analysis Suggesting that developmental studies devoted to the mastery of subordinate clauses group together some very heterogeneous psycholinguistic facts, Kail (1983) has insisted on the necessity of going beyond the study of isolated sentences to take into account utterances and their contents (e.g., Jakubowicz 1978; Piéraut-Le Bonniec 1980b), the speech situation (e.g., Champaud and Jakubowicz 1979; Champaud 1983), conditions of use and presuppositions (e.g., Berthoud and Othenin-Girard 1983). Similarly, it is advisable to integrate contextual factors into analyses of conjunctions. Conjunctions are not just logical operators organizing interpropositional links, but rather linguistic elements which also play a role in a dis course situation. Such an approach has been adopted by only a few authors up to now. There are, thus, practically no longitudinal analyses of conjunctions from this perspective, except that of Bates (1976) on two Italian children. While confirming the chronological precedence of AND (a classic observation), Bates related the ap pearance of BUT, BECAUSE, and EXCEPT to the child's production of utterances about utterances, "metapragmatic" forms which assure a certain auto-regulation of discourse. In experimental studies on the pragmatic function of conjunctions, Caron (1979, 1983) investigated the semantic space of the French connective SI (IF) basing his linguistic analysis of "if p then q" statements on Ducrot's (1972) theory. Ducrot shows the relationship between supposition and presupposition by opposing sup position to implication. Presupposition, like supposition, creates a universe of dis course. "Supposition and presupposition have in common the property of transforming, one explicitly and the other implicitly, the universe of discourse, and of transporting the listener into a situation chosen by the speaker." Utilizing an utterance classification procedure inspired by Miller (1967, 1969), with adult and 13 to 15 year-old adolescent subjects, Caron carried out a set of studies which indicate an organization of utterances based on their "illocutionary" characteristics. He considers the types of relations that utterances establish with and between interlocutors, for example, the acceptance of the responsibility for the statement introduced by SI, the "given" or "new" character of this proposition, and
136
M. KAIL AND J. WEISSENBORN
whether or not the relation between interlocutors is cooperative. Such research, mainly carried out on adults until now, should be carried out in a developmental perspective. Similar considerations underlie the work of Bronckart and Schneuwly (1984) who view the analysis of conjunctions in terms of textual organizers. In their work, however, the emphasis is less on the functional analysis of one conjunction or another than on their differential distribution according to the nature of a particular type of text (situational discourse, theoretical discourse, conversational narration). In the following, we will discuss studies on the child's growing understanding of the presuppositional and argumentative values attached to BUT (French: MAIS) in discourse. Linguistic analyses of this conjunction claim that for an utterance of the type "P but Q" to be used correctly, Q must contradict a proposition presupposed by the lis tener and inferred from P in a specific context (Pusch 1975; Anscombre and Ducrot 1977; Lang 1977; Asbach-Schnitker 1978). In other words, if P can serve as an ar gument for R, Q is an argument which cancels out R. These analyses call into question the idea of a symmetry between P and Q with distinct argumentative values. The conclusion is supported with less force for P than for Q. In an initial study, Kail (1980) showed that learning to understand the argumen tative value of MAIS is a gradual process whose result can be observed equally in production and acceptability judgements of utterances of the type P MAIS Q. Up to 6 years of age, MAIS is considered a coordination operator. Children thus rely on pragmatic regularity in accepting utterances of the type, "Le camion est chargé MAIS il roule lentement" (The truck is loaded BUT it is going slowly). It is actually only at around age 10 that children understand the argumentative value of MAIS, and reject utterances which contradict the presuppositions that they imply. This finding is confirmed by Bassano and Champaud's work (1987) on the French con junctions MAIS, BIEN QUE, POURTANT, QUAND MÊME and MÊME SI that relate, ac
cording to an analysis by Anscombre and Ducrot (1983), two propositions expressing opposite conclusions. However, questions about these results arise from two distinct points of view. Firstly, as mentioned in a critical review of the psycholinguistic research on presup positions (Kail and Plas 1979), the ecological validity of this kind of research is relative. In fact, in the experimental situations, especially those relating to com prehension, the child is required to make metalinguistic decisions, probably of a different nature than decisions s/he normally makes in natural communication situa tions.
CONJUNCTIONS: DEVELOPMENTAL ISSUES
137
The second question concerns the fact that the area of conjunctions is particular ly sensitive to cross-linguistic variations. Clearly in the case of MAIS, certain lan guages make different lexico-semantic distinctions not found in French. German (as well as Spanish) has at its disposal two distinct morphemes, SONDERN and ABER (Spanish: SINO and PERO), which have different use conditions (cf. the analysis by Anscombre and Ducrot (1977) on the existence of two MAIS in French). More precisely "P SONDERN Q" implies that the proposition P contains an ex plicit (i.e., syntactically marked) negation; its use indicates that the speaker is refut ing an actual or virtual assessment. This is expressed by MAIS in its contrastive value, "Il n'est pas jaune MAIS bleu" (It is not yellow BUT blue). In contrast "P ABER Q" does not imply negation of P (the argumentative value of MAIS, "Il est in telligent MAIS bavard" (He is intelligent BUT talkative)). The problem is to deter mine the extent to which the development of the interpretation of the presuppositional and argumentative components of the meaning of conjunctions like BUT by children is influenced by the structural properties of a particular lan guage. The syntactic-lexical organization of a language can favor or not access to pertinent indices, i.e. "local cues" (Slobin 1982), which the child uses in mastering the presupposed dimension. The general hypothesis of our research (Kail and Weissenborn 1984a) predicts that since German possesses two conjunctions (ABER and SONDERN) while French MAIS is plurifunctional, access to the presuppositional and argumentative com ponents of ABER will be facilitated by the existence of SONDERN. Linked to this hypothesis is the prediction that German children will distinguish SONDERN from the most primitive conjunction, UND (AND) at a younger age than French children acquire MAIS. In one study children were presented a short story which preceded the utterance containing MAIS or ABER/SONDERN which the children were to complete or judge. The results show the child is able to infer the presupposition by virtue of the lin guistic context. These results confirm previous results showing that in processing utterances of the type "P MAIS (BUT) Q" most young children (7 years old) base their responses on the contents of P and Q, the conjunction seeming to play only a very limited role. Utterances with MAIS (ABER) are mastered later (9:11) than ut terances with MAIS (SONDERN) (8:8), a difficulty explained by the greater cognitive complexity of the ABER contrast. Indeed, in the case of MAIS (SONDERN), the ex plicit negation of P supplies the necessary information permitting the child to deter mine the class of predicates within which it will be necessary to construct P's
138
M. KAIL AND J. WEISSENBORN
opposite, Q. In the case of MAIS (ABER), the child must construct a proposition P' presupposed by P on the basis of the context. These first results suggest that mastery of the semantics of ABER is accessed through that of SONDERN. That is, in order to comprehend the presupposition at tached to utterances with ABER, it is necessary first to master oppositions of the ad versative type. The prediction according to which the lexical differentiation present in German allows children to distinguish between UND and ABER/SONDERN is con firmed. We have indeed shown that in judgement tasks of "P MAIS (SONDERN) Q" utterances German children show a better performance than do French children; the plurifunctionality of the French MAIS accounts for this difference. Finally, it is in teresting that for both the French and the German children, utterances with MAIS (ABER/SONDERN) whose topic is the theme of the preceding context, are more easily dealt with, a result which brings out the functional connection between presupposi tions and given information. To evaluate more systematically the acquisition rates of linguistic devices used to express "contrast" and "negation" in different languages, we have compared our experimental work with naturalistic data. The study of the emergence of BUT in spontaneous production implies the analysis of more primitive syntactico-semantic organizational systems (e.g., juxtaposition of a negative proposition and a positive proposition), or the analysis of comparison procedures (similarities, opposition) which pave the way for its appearance and which vary across languages. Languages were chosen as a function of relevant contrasts (lexical differentia tion, degree of flexibility of the order of surface structure elements, phonological opposition, etc.) which they manifest. Thus we can compare languages such as German, Spanish and Hebrew which are characterized by great lexical differentia tion of conjunctions used for denial. The data in the different languages consist of corpora (longitudinal and cross-sectional data) collected in various communicative situations (exchanges with an adult, games, narratives, argumentations...), and in volve children from 1:6 to 3:6. The analysis of these corpora is based on a typol ogy of utterances with MAIS. This typology, inspired by Bruxelles et al. (1976), has been developed on the basis of the analysis of the corpus of Philippe from 2:3 to 3:3 (CHILDES data base; cf. Suppes, Léveillé and Smith 1974). Three classes of utterances have been distinguished.
CONJUNCTIONS: DEVELOPMENTAL ISSUES
139
Utterances in which the Child Says "P MAIS Q" In this class P and Q are propositions produced by the child; the connective MAIS may have a contrastive value (generally one of the two propositions is negative) or an argumentative one. In Philippe's corpus, we did not find this type of complete utterance.
Utterances in which X (Interlocutor) Says P and the Child Says "MAIS Q" These are the utterances in which the contrastive value of MAIS fulfills a dis course function of picking up on the discourse of others. a. The child controls the content of P (contestation by global refutation or con trast with expression of an alternative). Example
Exp: Phil (3:0):
On peut pas manger. MAIS si on peut manger. (We can't eat.) (BUT yes we can eat.)
Example
Ben, je croyais que c'était pour voler. Exp: Phil (2:11): MAIS non, c'est pour conduire. (Uh, I thought it was for flying.) (BUT no, it's for driving.)
b. The child contests an inference that he has drawn from P. Example
Exp: Phil (3:1):
Tu sais pas à qui il est ce jouet? MAIS tu peux bien prendre celle-là aussi. (You don't know whose toy this is?) (BUT you can take that one too.)
Whereas global refutations (MAIS SI... MAIS NON) (BUT YES... BUT NO) appear around 2:3, inference-based refutations do not appear before 3:1.
140
M. KAIL AND J. WEISSENBORN
Utterances in which the Child Says "MAIS Q" in Response to Extralinguistic Context. These are utterances in which MAIS introduces an opposition by the child to an element of situational context (e.g., behavior of the interlocutor, aspects of the speech situation). Example Exp: Qu' est-ce que je fais? Phil (3:3): Qu'est-ce que tufais? MAIS non le fais pas. (What am I doing?) (What are you doing? BUT no, don't do it.) Example
Exp: Phil (3:0):
Allez, le petit coup pour Clara. Vas-y. MAIS je vais boire avec la cuillère. (Come on, a little bit for Clara. Go ahead.) (BUT I am going to drink with the spoon.)
The "MAIS Q" Situation (Linguistic and Nonlinguistic) This category includes productions which have been called "denial of expecta tion" by Lakoff (1971) and Tannen (1977). As in the preceding category, the ut terances of this category express an evaluative opposition (transition between contrastive MAIS and argumentative MAIS). MAIS signals that the child expected dif ferent verbal or nonverbal behavior from the interlocutor, keeping in mind his or her beliefs regarding what the interlocutor knows. Example Exp: Phil (2:11): Exp: Phil:
Philippe returns Qu'est-ce que tu as trouve] Un petit jouet pour mon petit mouton. Quel petit jouetl MAIS le camion. (What have you found?) (A little toy for my little sheep.) (Which little toy?) (BUT the truck.)
The analysis of the first occurrences of MAIS in Philippe's corpus show that beginning at 2:3, this conjunction fulfills differentiated discourse functions. These
CONJUNCTIONS: DEVELOPMENTAL ISSUES
141
results accord with those of French (1981) and emphasize the fact that the first in stances of MAIS have a true opposition function. The naturalistic data indicate that very early children are sensitive to the implications of their own productions, and to the constraints of discourse. A very similar conclusion can be found in Peterson's study (1987) on the semantic and pragmatic uses óf BUT in children's narratives (3:6 to 9:6). children at all ages use BUT for primarily pragmatic func tions, to interrupt the flow of their narrative in order to insert relevant comments, monitor the listener's attention or change topic. In contrast to what comprehension research has indicated, i.e., that MAIS is ini tially understood as confirming or adding to a previous assertion, our analysis of spontaneous production data indicates that the child's initial refutations constitute the basis for the acquisition of explicit contrastive forms of MAIS (corresponding to SONDERN in German and SINO in Spanish). It is clear that the first instances of MAIS fulfill a discourse function whose most complex form is argumentative MAIS (ABER, PERO), which implies previous mastery of contrastive MAIS. From the perspective of a comparative approach to the acquisition of conjunc tions, the first question to be asked is whether the evolution of different functions evidenced in French is associated with linguistic procedures which vary among lan guages (presence or absence of lexical differentiation of conjunctions, order of propositions, ordering of affirmative and negative propositions, etc.) and to deter mine what modifications these variations introduce into the construction of the semantic domain we have discussed here. German data (from 2:1 to 3:4) confirm the initial emergence of "ABER Q" structures (where ABER has a discourse value), followed by "P SONDERN Q" structures appearing from 3:0, which appears earlier than in the French data. For English, the picture is more variable. While Eisenberg (1980) found a preponderance of "BUT Q" structures for the youngest children (2 years), Bowerman (personal communication) observed in the speech of one of her daughters a majority of complete structures of the type "P BUT Q". It is clear, how ever, that these observations are based on very fragmentary data from which it is as still premature to draw definite conclusions.
Final Remarks In concluding her critical review of the acquisition of complex sentences, Bowerman (1979) emphasizes that although our picture of the ways in which sen tences emerge in children's language has progressed, there still remain many areas to be explored. Thus it would be informative to study how complex sentences of
142
M. KAIL AND J. WEISSENBORN
different types are acquired by a single child in order to understand the links exist ing among artificially separated areas in experimental studies (temporals, causals and conditionals). The analysis of common conceptual contents, the study of the development of discourse skills and the analysis of constraints on the processing system, in addition to an investigation of possible semantic transfer from one domain to another, need further development. Conversely, detailed research into the relationship between form and function with regard to particular conjunctions must be carried out, keeping in mind that such studies will have to include detailed cross-linguistic analyses, and more detailed examination of conjunctions which have closely related meanings.
Children's Production of Textual Organizers J.-P. Bronckart and B. Schneuwly Section des Sciences de l' Education Université de Genève
Conjunctions are words with no declension expressing various operations of our mind and helping to relate members or parts of discourse. (Restaut, 17321)
Although this classical definition applies to conjunctions only, it satisfactorily specifies the aim of the present study, namely the study of those units of language that function as textual organizers (or connectives) and the psychological operations underlying their production. The experimental studies that we shall report here are characterized by an interactionist and differential approach aimed at revealing the relationships of interdependence between certain configurations of linguistic units and certain configurations of extralinguistic (or contextual) parameters. As will be seen later, such an approach implies the description of types of texts, including specific linguistic characteristics and the formulation of linguistic operations "repre senting" the different forms that the relationships of interdependence between text and context may take. Except for Antoine's monograph devoted to La coordination en français (1958 and 1962), very few linguists belonging to the structuralist trend dealt with textual organizers2. Indeed, Brunot (1926), Sechehaye (1950) and Bally (1932) inves tigated textual organizers, but their main concern was to introduce some order and consistency in their rather vague classification of this category of units. On the other hand, for the proponents of generative grammar, connectives were used main ly as "privileged criteria" for validating some hypotheses on syntactic description (cf. in particular the synthesis proposed by Grunig 1977).
144
J. P. BRONCKART AND B. SCHNEUWLY
For about the last fifteen years, however, as an extension of the studies per formed by logicians and philosophers focusing on the characteristics of "natural language" (Stawson 1950; Searle 1969), connectives and organizers have been analyzed in numerous, often stimulating studies. Dealing with the "logic of lan guage", from a pragmatic perspective, Ducrot and his collaborators (Ducrot 1973; Ducrot et al. 1980), analyzed the conditions of use in French of such units as MAIS (BUT), EH BIEN (WELL), DÉCIDEMENT (UNDOUBTEDLY), D'AILLEURS (BE SIDES), etc.; they demonstrated that the functioning of these connectives was high ly dependant on the characteristics of the situation of enunciation. In an analogous approach inspired by the theory of speech acts, Roulet and his collaborators studied the marks of structuration of genuine dialogues and proposed hypotheses specifying the level of functioning of a substantial subgroup of organizers (cf. L'analyse des conversations authentiques, ELA 44, 1981). Finally, within the framework of tex tual linguistics, Van Dijk (1979) and above all Biasci (1982) proposed to distin guish the "semantic" functioning (chaining of clauses) from the "pragmatic" functioning (chaining of speech acts) of various connectives in English, German and Italian. The most exhaustive study of the functioning of textual organizers was proposed however, by Gulich (1970). We shall examine it in more detail before analyzing its extensions in Quasthoff's (1979, 1980) and Auchlin's (1981) work. Gulich considers that every text has a structure which is marked at the surface level by a specific class of linguistic units, the "Gleiderungssignale" (segmentation signals); these are invariant words which have lost their proper lexical meaning and are henceforth capable of indicating and/or specifying the structural relationships in a text. On the basis of the analysis of a corpus of dialogues and conversational nar ratives in French, from a typological perspective, Gulich demonstrated that signals of segmentation in dialogues are essentially used to mark a change in topic or a change of speaker, whereas in narratives the same units contribute mostly to the structuration of episodic units. Gulich also demonstrated that the distribution of units is dependent on the type of text, where PUIS (THEN) and ALORS (SO) are produced with a high frequency in narratives, and where MAIS (BUT) and EH BIEN (WELL) occur mostly in dialogues. However, the author notes that it is often dif ficult to distinguish dialogue from narrative and that there is often a transition from one type to the other within the same text. Quasthoff advanced the hypothesis that segmentation signals would have two distinct functions: on one hand, they would constitute the involuntary trace of an "intense activity of cognitive planning" (to fill in pauses, to signal that one is going to continue to speak...); on the other hand, they would mark the different phases in text organization (or semantic macrostructure) and would serve to orient the
CHILDREN'S PRODUCTION OF TEXTUAL ORGANIZERS
145
listener's attention. The results obtained by Quasthoff seem to show that there is no one-to-one correspondence between either of these two functions and a subclass of organizers. She proposes however to distinguish marks of hesitation (repetitions, hum, prolonged phonemes) which would mostly constitute the trace of cognitive planning, from marks of segmentation (metanarrative sentences, abstraction, orien tation, coda, phrases such as "Wie dem auch sei", etc.) which primarily indicate macromantic caesuras, and finally connection marks which we are particularly inter ested in and which would have a twofold status. They would be the trace of general planning as well as the trace of means used for connecting events within a temporal or causal sequence in order to orient the listener's attention. Within the theoretical framework proposed by Roulet (1981), Auchlin studied markers of conversation (MCS) in French, hypothesizing that the function of MCS's is to indicate the "level of textualization" of the segments (or constituents) of a text. He proposed a classification which takes into account "the behavior of MCS's relative to the levels of textualization of the constituents they articulate" (linear chaining, etc.) on the one hand, and the type of constituent articulated on the other hand by the MCS's (constituent of the same speaker, constituent of the lis tener, nonverbal constituent). Like Quasthoff, Auchlin emphasizes the plurifunctionality of many textual organizers. After its introduction into linguistics, the pragmatic perspective was adopted by a number of psycholinguists who investigated the problems posed by the function ing of textual organizers. Using metalanguage tasks, Fillenbaum (1974, 1975, 1977) revealed the "referential" and "contextual" use of AND, OR and IF in adults. Using analogous tasks, Caron (1983) and Fayol (1981) described the status given by French speakers to such units as AINSI (SO), COMME (LIKE), SI (IF), ET (AND), OU (OR), APRÈS (AFTER) and ALORS (THEN). In children, we should first mention many studies focusing on the comprehension of connectives (cf. especially Katz and Brent 1968; Kail 1979a; and Hamer 1980) which go beyond the logical perspective initiated by Piaget as early as 1924. Concerning production - which we are more particularly interested in - Bates (1976) and Bloom et al. (1980) have shown that at an early age Italian-speaking children like English-speaking children, produce equivalents of AND and a little later equivalents of BUT, BECAUSE and IF NOT. Other connectives occur still later and are probably used as temporal or ganizers before indicating causality relationships. In an investigation of children's narratives, Kernan (1977) advanced a distinction analogous to Quasthoff's proposal and demonstrated that AND, THEN and AND THEN organizers are not only used to express relationships of succession, but could also relate semantically independent events. Fayol's (1981) very detailed study of children's narratives in French con-
146
J. P. BRONCKART AND B. SCHNEUWLY
firms the precociousness and plurifunctionality of ET (AND) demonstrates that or ganizers, interacting with punctuation marks, function as traces of the enunciative structuration of narratives. Except for Schneuwly's work (1981) on the production of dialogues, most of these studies are limited to the study of one particular type of text, children's narratives. Other types of textual production should be investigated if a real description of the ontogenesis of organizers is to be achieved. With this purpose in mind, we will summarize here a broad study which has been presented elsewhere in detail (cf. Bain et al. 1982; Bronckart 1983; Schneuwly and Bronckart 1983). 1. The textual corpora that has been collected constitute a first level of data the linguistic units - that can be analyzed using distributional procedures and statistical methods of quantification. 2. These corpora are dependent on a second category of parameters, namely contextual parameters in a broad sense. In order to define the problem of text functioning in a real interactionist perspective, we believe it indispensable to develop a model of context. Contextual constitutes a theoretically infinite col lection from which a finite set should be extracted, namely the set of parameters exerting an observable influence on text production. We propose to distinguish three subgroups of parameters: first, those which refer to social interaction, goals of the linguistic activity, institutional locus of this activity, listener and speaker; second, those which refer to the material act of enuncia tion (speech), speaker, possible listeners, time and physical location of enun ciation; third, those which concern the referential content, notions and cognitive relationships which are expressed. 3. The production of a text is the result of psychological operations on contex tual parameters. These operations are materialized in the syntactic and seman tic categories of a natural language. Each unit of a textual corpus then constitutes the trace of one or several operations and, for this reason, may be submitted to a second functional analysis. 4. The text unit is the part of a textual corpus produced in interaction with a stable configuration of the parameters of social interaction and the act of production. Types of texts are defined on one hand by the particular con figuration of linguistic parameters determining them and, on the other hand, by the configuration of linguistic units at the surface level. 5. General operations of contextualization should be distinguished from textualization operations. The former aim at the "representation" of contextual parameters, i.e., at the selection of one value for each parameter, while the
CHILDREN'S PRODUCTION OF TEXTUAL ORGANIZERS
147
latter contribute to the construction of the verbal chain. The latter category of operations may be described in terms of computations on contextual variables and may be divided into three subgroups: a) operations of textual anchoring (which determines the type of text); b) operations of text organization (semantic macrostructure); and c) operations of discursive strategies (con tributing to marking the phases of text composition, cohesion and modalization). On the basis of these theoretical proposals, an experimental method was derived which aimed at bringing out the relationships of dependence existing between cer tain types of context and certain configurations of surface units in a corpus. More precisely, the types of text production conditions which constitute the independent variables in the experiments were defined by controlling the parameters of social interaction and by their relations with the parameters of the act of production. On the other hand, the referential content was not controlled. The texts produced in these conditions were collected and analyzed and the ratios of dependance between contextual parameters and surface units in the texts were then computed. First of all, contemporary texts produced by adults were studied. The analysis resulted in a typology in which "polar" categories were differentiated from an "intermediary" one (cf. Bain et al. 1982). The present phase of study focuses on children's texts, using the same procedure. The aim is not to delineate how operations of textual production develop - such an aim is too premature - but rather to improve our knowledge of the relevant extralinguistic parameters and the linguistic operations underlying text production. It is our belief that for children the paradigms of lin guistic units corresponding to a determined linguistic operation are narrower than for adults and that this limitation should help us refine our present model of opera tions. In the present study, 140 texts were selected from the general corpus of texts produced by children. 70 of them were produced by children (9 and 10 years old) attending school (4th grade, this group will be referred to as 4P) in the suburbs of Geneva and the other 70 were produced by 11- and 12-year-olds attending the same school in another grade (6th grade, 6P). Texts were produced orally in three controlled conditions. a) Situational discourse (SD): The children were taught how to build a device with the help of cardboard, wheels, straw and nails. One of them was asked to explain verbally to a classmate (who was not trained to build the device) how to proceed with the construction. The necessary materials were available to the speaker, but he or she was required not to use them for his/her verbal ex planations.
148
J. P. BRONCKART AND B. SCHNEUWLY
b) Theoretical discourse (TD): The children were required to describe familiar objects; the experimenter asked one child to describe a carriage to children who were supposedly unfamiliar with this vehicle. c) Conversational narrative (CN): In the course of a conversation the ex perimenter asked the child to tell him what s/he did the day before. The texts produced by the children were recorded and transcribed. Subsequently, textual organizers were extracted from the transcriptions. These units constitute what is known as a semi-open paradigmatic class consisting of conjunctions, phrases, adverbs, and various "expressions". Due to their morpho-lexematic charac ter (cf. Bronckart 1983; Schneuwly and Bronckart 1983), they cannot be identified by rigid distributional procedures. For this reason all morpho-lexemes that are given in Grevisse's Bon usage (1936/1959) were noted. Expressions of oral punctuation, BON (WELL), BEN (SO), as well as prepositional phrases with a tern-
Table 1
Group 4P (9 to 10 Year-Olds)
Number of texts Number of words Number of organizers Ratio of organizers to number of words
SD
TD
CN
30
20
20
12.072
1.408
1.594
1.103
122
144
1/10.94
1/11.54
1/11.07
Group 6P (10 to 11 Year-Olds)
Number of texts Number of words Number of organizers Ratio of organizers to number of words
SD
TD
CN
30
20
20
12.137
838
3.317
1.130
46
257
1/10.75
1/18.20
1/12.90
General characteristics of the texts produced. SD: situational discourse. TD: theoretical discourse. CN: conversational narrative.
CHILDREN'S PRODUCTION OF TEXTUAL ORGANIZERS
149
poral value, APRÈS TROIS JOURS (THREE DAYS LATER), DEUX JOURS PLUS TARD
(TWO DAYS LATER), etc. were also entered in this category of organizers. Table 1 presents a summary of the general quantitative characteristics of the texts produced. First of all, it can be seen that texts produced in the SD condition are much longer than those in the TD or CN conditions. In spite of this difference in length, probably due to the real verbal exchanges in SD, the relative frequency of organizers is stable in the 4P group; the ratio of the number of words to the number of organizers ranged from 10.94 to 11.54. In the 6P group, while the rela tive frequency of organizers is analogous in SD and CN (10.74 to 11.07), it is much lower in texts produced in the TD condition (one organizer for 18.2 words). In a previous study, Schneuwly (1981) had already observed such an impoverish ment; at the end of primary school, children produce very short theoretical texts (on the average 45 words per text in the present study) with very few and poorly dif ferentiated organizers. Table 2 presents the different organizers found in the texts. For lack of a better solution, we maintain the differentiation between coordinate conjunctions and sub ordinate conjunctions which is disputed by various linguistics but which can be ap plied to our texts without any major problem. Also with all the necessary reservations that should be made, temporal values are distinguished from logical values. The fourth class (adverbs and adverbial phrases) groups together units that probably ought to have been differentiated had their frequency in the present texts been higher. Finally, it should be noted that the fifth class does not include VOILA (THERE), which occurs very frequently in SD (256 in the 4P group and 204 in the 6P group), nor OUI (YES) nor NON (NO), also very frequent in this type of text. Table 3 shows the distribution of organizers for each class. The digits indicate the relative frequency of production, i.e. the percentage of occurrence of the units of a class relative to the total number of organizers produced by each age group and for each type of text. On the whole, two thirds of the organizers consist of coordinate conjunctions. But at variance with Kernan's results, in the 4P as well as in the 6P group, connec tives with a logical value (ET (AND), MAIS (BUT), CAR (BECAUSE or FOR), DONC (THEREFORE)) are clearly more frequent in conversation narratives than in the two types of discourse. In SD and TD, most connectives are temporal {PUIS (THEN), ALORS (SO), ENSUITE (AND THEN), APRÈS (AFTER, THEN)). This paradoxical result can be better understood by comparing the theoretical discourse and the conversa tional narrative produced by two 4P students (cf. Table 4). At the first level of analysis, it seems that the order of the succession of events (semantic macro-struc ture) in CN is so salient that the order of the succession of clauses and the presence
150
J. P. BRONCKART AND B. SCHNEUWLY
Table 2
Classification of the Main Organizers Occuring in the Texts Classes
Main Occurences
la.
"Logical" coordinate conjunctions
ET (AND), MAIS (BUT), DONC (THEREFORE), CAR
lb.
"Temporal" coordinate conjunctions
PUIS ALORS (THEN), APRÈS (AFTER), ENSUITE
(FOR).
(AFTERWARDS), MAINTENANT (NOW). 2a.
"Neutral" subordinate conjunctions
QUE ( T H A T ) followed by a subordinate clause, SI (IF) followed by a subordinate clause.
2b.
"Logical" subordinate conjunctions
PARCE QUE (BECAUSE), PUISQUE (SINCE), SI (IF), POUR ( F O R ) followed by the infinitive, POUR QUE (SO THAT), SINON (IF NOT).
2c.
"Temporal" subordinate conjunctions
QUAND (WHEN), LORSQUE (WHILE), PENDANT (DU RING/ WHILE), CHAQUE FOIS QUE (EVERY TIME THAT), COMME (AS), EN MÊME TEMPS QUE TAN DIS QUE (AT THE SAME TIME THAT/WHILE), APRÈS (AFTER) followed b y t h e infinitive (after + verb + -ing), AVANT QUE (BEFORE).
3.
Prepositional phrases
QUELQUES JOURS PLUS TARD (A FEW DAYS LA TER), PENDANT CE TEMPS (MEANWHILE).
4.
Adverbs and adverbial phrases
AUTREMENT (OTHERWISE), AUSSI (ALSO), AINSI (SO), DES FOIS, QUELQUEFOIS PARFOIS (SOME TIMES), NORMALEMENT (NORMALLY).
5.
Oral punctuation
BON, BEN, EH BEN (WELL, O.K.).
of the neutral connective ET suffices to mark it at the surface level. On the contrary, in TD the elements to be described are not "referentially" ordered and the subjects seem to feel the need to mark more strongly the succession of phases in the description activity. In this type of text ET PUIS is probably also a mark of hesita tion, in Quasthoff's sense, and should be studied in more detail. When considering subordinate conjunctions, one can see that there are very few "neutral" units in the 4P group, whereas their number increases in the 6P group, especially in narrative texts. The role of these conjunctions seems to be highly de-
CHILDREN'S PRODUCTION OF TEXTUAL ORGANIZERS
151
Table 3 SD 4P
TD 6P
4P
CN 6P
4P
6P
la.
Logical coord. conj.
14.3
13.3
0.4
15.1
40.8
36.9
lb.
Temporal coord. conj.
63.6
57.6
56.6
52.2
25.6
22.8
1.
TOTAL
77.9
70.9
60.6
67.3
66.4
59.7 !
2a.
Neutral subord. conj.
1.8
2.6
3.2
4.3
3.5
8.9
2b.
Logical subord. conj.
9.4
10.3
12.3
4.4
7.5
10.1
2c.
Temporal subord. conj.
1.5
0.4
4.8
0.0
18.0
12.9
2.
TOTAL
12.7
13.3
20.3
8.7
29.0
31.9
3.
Prepositional phrases
0.0
0.0
0.0
0.0
4.2
5.8
4.
Adverbs and Adverb. phrases
3.4
6.4
18.7
23.8
0.0
1.6
5.
Oral punctuation
5.8
8.4
0.0
0.0
0.0
0.0
Relative frequency of organizers in 4P and in 6P. SD: situational discourse. TD: theoretical discourse. CN: conversational narrative.
pendent on the verbal forms (DEMANDER SI (ASK IF), DIRE QUE (SAY THAT), CROIRE QUE (BELIEVE THAT)) and one may wonder whether they are real textual organizers. As to the other two classes (2b and 2c), in both types of discourse, logi cal subordinate conjunctions predominate while temporal subordinate conjunctions are rare. In contrast, in narrative texts the latter predominate. It should be em phasized that in this type of text the development from 4P to 6P is mainly charac terized by diversification in the marks of subordination, be they logical or temporal. The distribution of the last three classes of organizers is very clear. Prepositional phrases occur in narratives only, and marks of oral punctuation are characteristic of SD. In both cases, the occurrence frequency is very low. Adverbs and adverbial phrases are produced mainly in TD and in most cases have the value of a frequen tative phrase (SOUVENT (OFTEN), QUELQUEFOISorDES FOIS (SOMETIMES), etc.). One might hypothesize that this type of unit (which amounts to a quarter of the or ganizers in TD for the 6P group) constitutes the trace of an operation of aspectual determination of the speaker's perceptual activity, whereas for adults one would ex pect that the determiners quantify the topic. For example, instead of the expected utterance "certains carosses ont de jolies lanternes" (some carnages have pretty
152
J. P. BRONCKART AND B. SCHNEUWLY
Table 4
Examples of Typical Texts Produced by 9- to 10-Year-Olds
Theoretical Discourse: Sandro (4P) Un carosse - ALORS - il y a quatre roues - il a quatre chevals (sic) QUELQUEFOIS - PIS en haut il y a des sièges pour celui-là qui conduit le carosse - ET PIS derrière il y a une grande planche ET PIS il y a des bancs dedans - des sièges pour s'asseoir - PIS c'est QUELQUEFOIS pour les dames qui viennent au bal PIS ils vont - PIS le haut, il y a des fenêtres - PIS des portes - QUELQUEFOIS on peut mettre des bagages en haut ET PIS il y a des phares. Translation: A carriage - THEN - it has four wheels - it has four horses SOMETIMES THEN on top there are seats for the one who drives the carriage - AND THEN behind there is a huge board AND THEN there are benches in it - seats for sitting on - THEN it is SOMETIMES for ladies who go to a ball AND THEN they go - THEN the top, there are windows - THEN doors - SOMETIMES you can put luggage on top AND THEN there are lights. Conversational Narrative: Bruno (4P) Dimanche - QUAND f ai été au golf - j'ai tiré de belles balles - ET au quatrième trou, j' ai tiré une balle très belle ET je suis tombé - ET c'est très bien le golf de Divonne - MAIS on a dû partir CAR j'étais tout mouillé. Translation: Last Sunday - WHEN I went to the golf course - I made some beautiful shots AND at the fourth hold I made a very beautiful shot AND I fell down AND the golf course at Divonne is very nice - BUT we had to leave BE CAUSE I was all wet.
CHILDREN'S PRODUCTION OF TEXTUAL ORGANIZERS
153
lights), children produce sentences of the following type: "DES FOIS, ils ont de jolies lanternes" (SOMETIMES, they have pretty lights). Table 5 presents a more detailed analysis of the distribution of coordination con junctions. It appears that for "logical" organizers the difference between these two types of discourse (SD and TD) on the one hand, and narratives on the other hand, is mainly due to the massive use of ET in narratives; other types of logical connec tives are equally frequent in SD and in CN. Temporal conjunctions are distributed over four subclasses. PUIS (THEN) is nearly the only unit produced in TD; it also occurs more frequently in SD, which confirms its status as an organizer linked to the continuity of discursive activity (according to Quasthoff). MAINTENANT (NOW) is found almost exclusively in SD, together with VOlLÀ (THERE), OUI (YES), NON (NO); its functioning seems to be related to controlling the listeners' activity, and it could be considered as a deictic. APRÈS (AFTER) and ENSUITE (AFTERWARDS) predominate in CN, but also appear in SD; they may be considered as serving to mark the succession of events. The status of ALORS (THEN) is polyvalent, but the increase in its frequency of occurrence in CN for the 6P group seems to reveal that later on it becomes specialized in marking temporal succession. To what extent, then, do these data permit us to increase our knowledge of the operations underlying the production of textual organizers? Within the framework of the general model that we are attempting to develop, textual organizers might constitute the trace of four separate operations of textualization:
Table 5 SD 4P ET (and)
TD 6P
4P
CN 6P
4P
6P
6.6
6.8
5.3
16.0
48.2
45.6
MAIS (but), CAR (for)
11.7
12.0
1.3
6.4
13.2
16.2
PUIS (then)
31.4
35.1
92.1
75.3
5.3
4.5
MAINTENANT (now)
18.8
16.8
0.0
3.3
0.0
0.7
APRÈS (after), ENSUITE (afterwards)
13.1
11.0
0.0
0.0
23.9
12.9
ALORS (then), other conjunctions
18.5
18.3
1.3
0.0
9.3
20.1
Relative frequency of occurence of the various types of conjunctions of coordination in the two age groups. SD: situational discourse. TD: theoretical discourse. CN: conversational nar rative.
154
J. P. BRONCKART AND B. SCHNEUWLY
T1: which would mark textual anchoring {UN JOUR (ONE DAY), HIER (YESTER DAY)). T2: which would mark the phases of textual structuring, or to put it differently, the semantic macrostructure {PUIS (AFTER), C'EST ALORS QUE (SO THEN)). T3: which would contribute to grammatical cohesion, in Halliday's (1964) sense {TANDIS QUE (WHILE), PUISQUE (SINCE)). T4: which would mark modalizations, i.e. the speaker's explicit modal interven tions {MALHEUREUSEMENT (UNFORTUNATELY), ON POURRAIT CROIRE QUE (WE MIGHT BELIEVE THAT)). These hypotheses which are derived from our model, integrate and organize most of the functions that up to now linguists have attributed to connectives, with the exception, however, of the function of "marking the cognitive planning" proposed by Quasthoff (1979). In agreement with this author, we propose a fifth operation which expresses the very essence of discursive activity: TO: which would mark the continuity of the textual chain {EH! (UH), BON (WELL)). In accordance with the theoretical perspective presented above, it would seem that organizers which "reflect" the same operation are organized in different subsets depending on the type of anchorage and, as a result, on the type of text. Differen ces between types are especially visible at the level of macrostructure marks. Thematic units in SD are very highly dependent on the listeners (marks of interac tion), while in TD they are more dependent on the social pair listener-speaker (marks of abstraction) and in CN they are closer to the events reported (episodic marks). An attempt will now be made to identify the operations each unit of the eight surface classes of organizers are the trace of. Let us consider class la ("logical" coordinate conjunctions). As mentioned above, the functioning of the ET unit is paradoxical; it occurs more frequently in CN than in the two types of discourse. Generally considered as a precocious and polyvalent connective (cf. Bates and Fayol especially), in the groups of subjects studied here (whose ages vary between 9 and 12 years) it clearly functions as an organizer that is specific to narrative. We think that this use constitutes an inter mediary stage, probably a necessary one, which follows the production of general marks of interaction and which precedes the use of marks specific to the macrostructure of narratives. Compared to the other organizers of this class, MAIS is the
CHILDREN'S PRODUCTION OF TEXTUAL ORGANIZERS
155
only one to have a high frequency of occurrence. It seems to play two different functions: in SD its function is frequently to manage and control interaction and is analogous to what Bruxelles et al. (1976) described in adult discourse, while in CN and TD (and for the remaining occurrences in SD) it clearly marks "logical" relationships between events. MAIS thus functions in our corpora as an index of the semantic macrostructure; in SD, it marks interaction and in the other two types it marks episodes. As mentioned above, MAINTENANT has a particular status in class lb (temporal coordinate conjunctions) and indisputably functions as a mark of interaction. The other units of this class are clearly marks of connection (in Quasthoff's sense); they mark both the continuity of discursive activity and temporal caesuras in the macrostructure. However three subsets of this class can be distinguished: PUIS, particular ly in TD, is related to discursive activity and its temporal character is less accentuated; APRÈS and ENSUITE seem to be more closely related to the macrostruc ture, while ALORS occupies an intermediary position. Things are much clearer concerning subordinate conjunctions. "Neutral" units (2a) do not function as textual organizers, and units with a logical (2b) or temporal (2c) value have a "transphrastic" role; they are hence marks of cohesion, in Halliday's sense. Predictably, units of class 3 (prepositional phrases ) function exclusively in nar rative. They mark textual anchoring and its relays in the text macrostructure (a clearly episodic function). Class 4 includes rather diverse units (adverbs and adver bial phrases) whose main function seems to be to characterize the "modus" of the relationships between speaker and listener (NORMALEMENT (NORMALLY), QUAND MÊME (ANYWAY), DES FOIS (SOMETIMES); these units undoubtedly express opera tions of modalization. It should be emphasized that insofar as connectives of class 3 and class 4 organize the text in its relationships with the act of production on the one hand (anchoring) and with the parameters of social interaction on the other (modalizations), they may be considered as marks of "coherence" in Slakta's (1975) sense. Finally, units of class 5 clearly function as marks of continuity in dis cursive activity as well as marks of interaction. The interpretation that has just been proposed obviously necessitates solid ex perimental confirmation. Nevertheless the data demonstrate the heuristic validity of the operative model as well as the relevance of differentiation among several types of text. Situational discourse is characterized by the massive presence of marks of discursive continuity (TO) and marks of interaction (T2), by the scarcity of units referring to cohesion and modalization and by the absence of units expressing anchorage. In this type of text the macrostructure is clearly dependent on the act of
156
J. P. BRONCKART AND B. SCHNEUWLY
production and on the control exerted by the listeners. Hence it is not necessary to make "coherence" explicit. Conversational narrative generally contains marks of anchoring (Tl) that are reinforced in the macrostructure by episodic marks (T2) and marks of consistency (T3, mostly in the 6P group). As a result, units express ing discursive continuity are very rare in this type of text. Finally, very poor struc turation of theoretical discourse is due to a lack of macrostructure organizers. The units that are found in this type of discourse contribute to the continuity of speech (TO) on the one hand and "reassure" the speaker in his/her exchanges with the lis tener on the other hand (T4, modalizations).
NOTES 1. Mentioned by Biasci, 1982. 2. To simplify we consider as structuralist the linguistic schools that existed from Saussure to Chomsky. 3. This type of operation is often difficult to differentiate from operations of contextualization concerning the relations of referential content.
The Development of Discourse Cohesion: Some Functional and Cross-Linguistic Issues M. Hickmann -Planck Institut für Psycholinguistik, Nijmegen
Introduction Discourse cohesion is a fundamental aspect of communication. Roughly, cohesion is created by the conjoined uses of linguistic devices which in various ways link together utterances in discourse and thereby create a "co-text" necessary for the unfolding of discourse. The ontogenesis of discourse cohesion corresponds to the development of children's ability to "anchor" speech in co-text and thus to use language as its own context. It is only fairly recently that the study of child language has been directly con cerned with the uses of linguistic devices in discourse, rather than in isolated sen tences. This interest for discourse partakes of a more general tendency in child language research to focus on the speech context in order to explain the functions of linguistic devices used by children (e.g., Bates 1976; Karmiloff-Smith 1979; Ochs and Schieffelin 1979; Bruner 1983). The systematic study of children's dis course skills, as can be found in studies of conversations, narratives, argumentation, game explanations, route directions, etc., has thus shown there is progression, sometimes at a late age, in how children organize personal, spatial, and temporal reference across utterances in discourse (cf. among others, Karmiloff-Smith 1981; Weissenborn 1981; Hickmann 1982; Schneuwly 1984; Bamberg 1985, 1986; Bronckart 1985; Fayol 1983a, 1985a). In this paper we discuss the development of cohesion by examining children's uses of referring expressions in discourse. We will first briefly present a few prin ciples governing these uses, as well as some methodological problems involved in the study of children's referential system. For example, children's mastery of the referential system seems to be a much later development if one includes in thenpragmatic repertoire the intralinguistic uses of referring devices linking together ut terances within discourse. We illustrate these points on the basis of how children
158
M. HICKMANN
introduce referents and maintain reference to them across utterances in narrative discourse. On the basis of these aspects of reference in children's narratives, we charac terize the ontogenesis of discourse cohesion by a gradual development of a linguis tic repertoire that is necessary to develop the capacity to anchor speech in speech itself. This capacity is an essential part of what all children have to acquire in order to become "competent" native speakers and it may have consequences not only for their communicative skills, but also for their cognitive skills. However, the par ticular linguistic devices that must be acquired for this development vary from lan guage to language. From a cross-linguistic perspective, we will briefly illustrate some of the different problems that must be confronted by children acquiring dif ferent languages in organizing reference in discourse. These comparisons suggest there are both universal principles in the development of discourse cohesion and some different developmental courses depending on the language being acquired.
Deixis, Coreference, and Anaphora As shown in previous reviews (Hickmann 1982, 1984), studies that have focused on the development of reference (e.g., on the acquisition of definite versus in definite detenniners and of pronouns) present divergent conclusions. Some con clude that children master the referential system as early as two or three years old, while others conclude that they do not do so until 10 or even 12 years. A detailed comparison of these studies shows that these divergences can be resolved if one takes into account the relation between children's utterances and different aspects of the speech situation from a functional point of view. This review agrees with studies that conclude that children's mastery of the referential system is a late development, showing that their referential expressions have, in part, different func tions in comparison to those of the adult's. One of the conclusions of this synthesis that concerns us particularly here is that, despite some elaborate distinctions in young children's referential expressions they have, in part, different functions in comparison to those of the adult's and it is precisely the systematic uses of these elements for discourse cohesion that are acquired late. One can summarize the general principle that leads to this conclusion as follows: in order to determine if children "master" the referential system, and particularly whether they can use referring expressions in discourse like adults, one must iden tify what we will call here "strict intralinguistic" uses of referring expressions, namely those that necessarily and maximally depend on the co-text (for more
THE DEVELOPMENT OF DISCOURSE COHESION
159
precise definitions, see Halliday and Hasan's notion of the "textual" function of language, 1976; also Hickmann 1982). For example, consider some of the different types of referring expressions that could be used by a speaker of English to mention a referent in discourse. The list in (1) shows four types of expressions, ordered in terms of the extent to which they presuppose the existence and identity of the referent. In this list, the expression a N is the least presupposing and the zero anaphor (e.g., "John left and 0 went home") the most presupposing. In contrast to a N, the definite determiner in the expression the N indicates that the existence and identity of the referent is mutually presup posed. However, it also contains relatively precise lexical information, in com parison to explicit pronouns, where this information is minimal. (1)
a N < the N < it < 0
Now, compare two different situations in which a speaker may want to "intro duce" a referent in discourse and say something about it. In one situation the denoted referent is present and both interlocutors are paying attention to it. In this case a pronoun (personal or demonstrative), possibly accompanied by a gesture, could in principle suffice to initiate discourse about the referent, as well as to main tain reference to it thereafter from utterance to utterance. In contrast, consider another situation which would be characterized by the following three conditions: (1) the referent is not present in the speech situation, (2) it has not yet been men tioned in previous discourse (or even in another speech situation involving the two interlocutors), (3) its existence and identity cannot be presupposed a priori (e.g., the referent is not unique and known to all, as would be the case for the expression the sun). In this case, the locutor depends necessarily and maximally on discourse to introduce the referent, as well as to maintain reference to it in order to communi cate efficiently. S/he will use one of the least presupposing expressions available in his/her language (typically a nominal with an indefinite determiner in a language such as English or French) in order to introduce the referent, sometimes in "presentative" utterances (e.g., "there was an N"). Once s/he has introduced the referent, s/he will use presupposing devices (definite nominals or pronouns, including zero anaphors) in order to go on denoting the referent across utterances. Speakers' choices among more or less presupposing devices in reference maintenance depend at this point on a number of factors, some of which will be mentioned below. As can be seen from the above, the same devices can be used in relation to dif ferent aspects of the context and thereby have different "functions" (e.g., in the sense of Halliday and Hasan, 1976). For example, even though all pronouns are
160
M. HICKMANN
deictic elements, since they depend on some aspect of the speech situation, one can distinguish the intralinguistic uses of these devices as a particular type of deixis which refers to an entity by "pointing" to some coreferential element in the co-text, rather than to the actual referent in nonlinguistic context. This distinction between deictic and strict intralinguistic uses has several im plications. First, it implies that one cannot equate coreference and anaphora. For example, consider the sequences of utterances uttered by a speaker A to inter locutor in (2) to (4); (2)
(A has just received a present from B) It's beautiful. Where did you get it? (3) (A and are in an art museum, looking at different paintings) Look at this one. It's beautiful. (4) John bought a painting at the auction yesterday. You should see it. It's beautiful.
In (2) the presence of the two coreferential pronouns in adjacent utterances does not guarantee that these uses are strictly intralinguistic. In particular, unlike (4) where the referent is absent, both pronouns in (2) are deictic, so that in this respect the sequence it-it is akin to a sequence involving (one or more) demonstrative pronouns as in (3). Thus, although coreference is necessary for an expression to be anaphoric, it is not a sufficient criterion to distinguish deictic from anaphoric uses of the same forms. A second implication of the distinction between deictic and strictly intralinguis tic uses, directly related to the first one, is the fact that an adequate functional inter pretation of how pronouns are used requires that we also consider how the denoted referents are introduced in discourse. Thus, it follows from the discourse situation illustrated above that the distinction between anaphora in (4) and mere coreference in (2) and (3) correlates with different devices used to first mention the referent. As we have seen, if the referent is mutually known by virtue of the nonlinguistic con text, the speaker does not have to introduce it strictly in discourse. In fact, the use of some appropriate linguistic device (e.g., a nominal with an indefinite determiner) to introduce a referent that is not mutually known in discourse, becomes a neces sary criterion to determine whether subsequent coreferential expressions maintain ing reference to this entity are intralinguistic or not. The absence of such an appropriate referent introduction in discourse often makes it difficult to determine unambiguously what function(s) pronouns have in discourse.
THE DEVELOPMENT OF DISCOURSE COHESION
161
However, the presence of an appropriate referent introduction, although neces sary, is not sufficient to determine whether children master the discourse-internal functions of referring expressions. There is yet a third implication of the distinction between deictic and intralinguistic uses. Necessary and maximal reliance on dis course also implies that speakers must handle referential continuity and discon tinuity in reference maintenance. There are numerous factors affecting speakers' choices among available presupposing devices within discourse. They include structural, semantic, and pragmatic considerations, some of which are languagespecific3 (cf. Li 1976; Silverstein 1976; Bolinger 1977; Givon 1983; Klein and Stutterheim 1985; Tomlin 1987; etc.). For example, consider the uses of referencemaintaining devices in examples (5) and (6). In (5) explicit or zero pronouns suf fice to refer to the policeman unambiguously, so that definite nominals are actually inappropriate after it has been made clear who the speaker is talking about. (5)
John came into the room and there was a policeman there. The policeman (he, the man) was sitting at the desk and he (0, *the policeman) was wear ing a hat and coat. Suddenly, he (*the policeman) took a gun out of the desk drawer and he (0, *the policeman) pointed it at John.
In (6) more than one referent is centrally involved. All other things being equal, the zero anaphor in (6a) can only be coreferential with the NP in the preceding utterance which is in the same (agent) subject role {Harry). Although the pronoun he in (6b) can, in principle, refer to either Harry or to John, the most natural interpretation is that it refers to Harry. By virtue of these constraints, the repetition of the full NP in (6c) is inappropriate, since it is clear what referent this expression denotes. Inversely, if the speaker wants to refer unambiguously to John as the one who left in (6d), s/he would typically have to use a full nominal. (6)
Harry (h) greeted John (j) and... a. 0(h, *j) left b. he(h, ?j) left *Harry left d. John left
Although the distinction between deictic and strict intralinguistic functions is theoretically clear, it is difficult to apply when analyzing spontaneous productions, particularly those of young children in natural situations. For example, it is often
162
M. HICKMANN
difficult to determine whether young children presuppose the existence and identity of referents on the basis of the nonlinguistic context or of the co-text. In many cases, although there is indeed a coreferential relation in the co-text, it must be in ferred by the adult who interprets the utterances of the child. One can illustrate some aspects of this problem from studies that have claimed that children master these uses very early (between 2 and 3 years). Many of these studies have concluded that young children master both the uses of anaphora and the uses of devices to introduce referents in discourse on the basis of examples such as (7) and (8) (from Huxley 1970 and Brown 1973, respectively; for a full discussion of these and other examples, see Hickmann 1984): (7) (8)
He's a clever pilot. He can fly upside down. *That a jeep. I put some in the jeep.
Example (7) is interpreted as showing the development of the anaphoric function of pronouns. Similarly, example (8) is interpreted as showing both a "correct" use of the indefinite detemiiner to introduce a referent and a "correct" use of the definite determiner to maintain reference to it in discourse. As noted above, given that the denoted referents were (presumably) present, such conclusions should be qualified in at least two ways: (1) the expressions that maintain refer ence, although coreferential with a preceding expression, are not necessarily (strictly) intralinguistic, and it is, in fact, quite probable that they are deictic; (2) referent introductions such as the one in (8) are part of predicating constructions that "label" the referents and denote them deictically (N.B. the demonstrative that) or even presuppose them entirely (e.g., "A jeep", "jeep"), all of which are quite common in early child speech (cf. Keenan and Klein 1975; Bruner 1983). From such examples, therefore, one cannot conclude that referential devices have a (strict) intralinguistic function in young children's repertoire. Even if it can be shown that children have learned some formal opposition between indefinite and definite determiners at an early age, a careful and qualitative analysis of how this opposition is used in their utterances and in the context is necessary to determine whether these forms have the same functions as in an adult's repertoire. A number of studies on children's discourse organization have shown that the strictly intralinguistic uses of referring expressions are in fact a late development. For example, Karmiloff-Smith (1981) concludes that children begin to use referring expressions deictically and only later as intralinguistic devices which establish links across utterances in discourse, for example reserving pronominal anaphora to main tain reference systematically to the "thematic subject" of discourse. It should be
THE DEVELOPMENT OF DISCOURSE COHESION
163
noted again that in this study children were producing narratives in a situation where they and their adult interlocutor were leafing through a picture book together. Given that joint attention to the referents was assumed in this situation, children did not have to introduce referents or to maintain reference to them unam biguously with presupposing referents. Thus this situation might in fact "invite" them to depend on the nonlinguistic context for communication. The importance of being able to determine whether strictly intralinguistic uses are part of the child's repertoire can be highlighted if we consider the typical in ferences that have been made about children's cognitive and interpersonal develop ment on the basis of their referential system. For example, children's uses of devices that are "too presupposing" for the introduction of referents (i.e., am biguous for the listener) have often been interpreted as showing some degree of "egocentricity" at an earlier stage which prevents them from "taking their listeners' perspective". However, children's frequent uses of labellings have also been inter preted as showing that they are concerned with establishing joint attention to referents in discourse and therefore that they are not egocentric. Given the general principles discussed above, there is yet another interpretation of children's uses that would resolve this question. Young children are able to "introduce" referents which they want to talk about, but only in speech situations that allow them to use referential devices deictically and that do not require them to depend on strict in tralinguistic relations in discourse. In other words, children would only seem "egocentric" in situations where their pragmatic repertoire does not allow them to depend maximally on discourse to communicate (also see Hickmann 1986, 1987a). In order to test this hypothesis, however, it is necessary to study children's dis course in situations that differentiate as much as possible deictic from (strict) in tralinguistic uses of the same forms. For example, some studies show that if referents are not present, or if they are present but not visible to the child's inter locutor, referents are not introduced by means of the strictly intralinguistic devices that would be necessary in these cases until a relatively late age (Warden 1976, 1981; Hickmann 1982, 1984, 1986, 1987a; Kail, Hickmann and Emmenecker 1987). The results summarized below illustrate this phenomenon for both referentintroductions and for reference-maintenance.
Cohesion in Children's Narratives The results presented below are part of a larger research project on children's narrative discourse between the ages of 4 and 10 years old. In a first series of
164
M. HICKMANN
studies (Hickmann 1982), various situations and materials were used to elicit narra tives from English-speaking adults and children of three age groups: 4, 7, and 10 years (40 children in each age group). This research was then extended within a cross-linguistic approach comparing elicited narratives in several language groups (see the section entitled Cross-Linguistic Perspective, below). The data summarized here were collected in the following situation: each child was asked to narrate the two picture sequences shown in Figures 1 and 2 for a naïve (blindfolded) inter locutor, who then had to tell the stories back. Note that in sequence A (but not in B), one of the characters (the horse) was clearly more central than the others.
Introduction of Referents The first mentions of the (animate) referents fell into four types: (1) appropriate expressions, namely those that did not presuppose the existence and identity of the referents, mostly NP's containing indefinite detemiiners, with or without presenta tional constructions ("there is/was a horse", "a horse was running"); (2) inap propriate expressions, namely those that clearly presupposed in different degrees the existence and identity of the referent on first mention, mostly definite nominals and pronouns ("The horse/he is running"); (3) ambiguous expressions5; (4) expres sions (regardless of presence and type of determiner) that were used in explicit predicating constructions ("this/that is a/the horse") or in "elliptical" utterances that could be akin to such labellings (including utterances such as "A horse", "the horse", "horse", and "a horse running"). Tables 1 and 2 display the proportions of the different types of first mentions that were found for picture sequences A and B, respectively. These tables show a clear progression, particularly an increase in the uses of appropriate introductions between 4 and 7 years, then between 7 and 10 years, accompanied by a decrease in the uses of inappropriate expressions and in the uses of expressions with explicit or potential labelling constructions. Explicit and potential labellings were found mostly at 4 years and they have practically disappeared by 7 years. A qualitative analysis of these uses at 4 years shows that the large majority contained some determiner (54% indefinite, 37% definite) and consisted either of explicit predicating constructions (39%) or expres sions used alone to label the referents (46%). Examples (9) and (10) illustrate the uses of explicit and potential labellings by the 4-year-old children (sequences A and B, respectively). Note that the 4-year-olds' referring expressions were often accompanied by deictic elements (such as the demonstratives here and there), as
THE DEVELOPMENT OF DISCOURSE COHESION
Figure 1
165
Figure 2
166
M. HICKMANN
THE DEVELOPMENT OF DISCOURSE COHESION
167
Table 1
First Mentions of the Animate Referents in Sequence A Types of Expressions
4 Years
7 Years
10 Years
Appropriate
19%
59%
82%
Inappropriate
48%
35%
15%
Ambiguous
5%
5%
2%
Explicit or potential labellings
28%
1%
1%
Total
(91)
(109)
(113)
Table 2
First Mentions of the Animate Referents in Sequence Types of Expressions
4 Years
7 Years
10 Years
Appropriate
25%
55%
87%
Inappropriate
48%
38%
12%
Ambiguous
5%
5%
1%
Explicit or potential labellings
22%
2%
-
Total
(131)
(148)
(158)
well as by nonverbal pointing to the pictures. In addition, these labellings were often used not only to first mention the characters, but also to denote them there after in subsequent discourse, especially as children went from one picture to the next (also see the section entitled Reference Maintenance, below). For example, in (9) the label "horse" (first mention, picture 1) is followed by another potential label "and a horse and a cow" (picture 2). In (10) the first mention of the dog "with a dog there" (picture 4) is followed by the labellings "here's a dog biting the kitty cat's tail" (picture 5) and "here's a dog who's chasing a cat" (picture 6).
168
M. HICKMANN
(9)
Horse. A horse is running. [...] And a ... horse and a cow. [...] And a horse is ... still running. [...] Horse fell down. [...] And a bird's ... ca-and the bird came. [...] Cow ... helped him. (4 years) (10) First a duck she's in her nest ... [...] here's duck she's out of her nest ... with a cat there ... and here's a cat ... climbing up the tree with a dog there ... [...] and here's a dog biting the ... kitty cat's tail ... and here's ... a dog who's chasing a cat and ... and that thing is getting back into her nest. (4 years)
There are two ways to interpret the young children's first mentions of referents. On the one hand, the majority of their expressions are deictic. This is clearly the case for explicit labelling and for definite nominals, and it is quite likely to be the case for potential labelling. According to the distinctions we made above, such uses do not contribute directly to the cohesion of the narrative, if one defines this notion as a system of strict intralinguistic relations. On the other hand, one could interpret at least some of these first mentions as "primitive" referent introductions. Although their main function is to label the referents, they are also indirectly deictic means of introducing referents in discourse (and of reintroducing them from picture to picture) for children who have not yet mastered strict intralinguistic uses of these devices and that still depend on nonlinguistic context to structure their discourse.
Reference Maintenance The analyses of reference-maintenance involve all the referring expressions that were used in the narratives to denote the characters after they had been first men tioned. These expressions fell into five types: (1) nominals with indefinite deter miners; (2) definite nominals; (3) pronouns; (4) zero anaphors;8 (5) nominals (regardless of presence and type of determiners) used in explicit or potential labell ing. Overall, relatively few indefinite determiners were used outside of the label in reference-maintenance (e.g., "a horse is still running" in (9)). These uses were found mostly at 4 years and to a lesser degree at 7 years (sequence A: 3% at 4 years, 2% at 7 years, and 0.4% at 10 years; sequence B: 3% at 4 years, 1% at 7 years, 0.1% at 10 years). As noted above, some explicit and potential labels were used not only for first mentions, but also for subsequent ones, especially at 4 years
THE DEVELOPMENT OF DISCOURSE COHESION
169
(23% for sequence A, 15% for sequence B). They were rare at 7 years (2% for A, none for B), and there were no occurrences at all at 10 years. The two picture sequences differed with respect to the uses of definite nominals versus highly presupposing expressions (pronouns and zero anaphors), especially at 7 and 10 years. These uses were much more frequent than all other (nominal) ex pressions in sequence A (51% at 4 years, 73% at 7 years, 80% at 10 years), but not in sequence (44% at 4 years, 47% at 7 years, 54% at 10 years). This result is due in great part to the fact that sequence A had a "main" character (horse), whereas se quence did not. Thus, in sequence A, the great majority of expressions denoting the horse were highly presupposing (pronouns and zero anaphors), whereas expres sions of both types were equally frequent for the other referents. In sequence B, no referent was more presupposed than the others.9 Examples (11) and (12) illustrate this first result. Both narratives were produced by the same 7-year-old child for se quences A and B, respectively. In (11) the child uses a whole series of pronouns to maintain reference to the horse, whereas in (12) he uses practically only definite nominals in reference maintenance. (11) Once there was a horse running and ... he saw a cow and then he stopped and then he jumped over the fence and he broke the fence and got hurt. And then ... a bird came along and ... with a ... doctor's kit and ... bandaged the horse up. (7 years) (12) Once there was a bird in her nest and a cat came along and the bird was flying away and the cat was looking up at the nest and the cat - and a dog came along and the cat was trying to climb up the tree and he was almost there and the dog - bit the dog - bit the cat's tail ... and the cat ... ran away. (7 years) Each referring expression maintaining reference to the characters was also analyzed in terms of its relation to the co-text that preceded the utterance. This analysis can be simplified for our purposes here as follows (for more details, see Hickmann 1982). First, a distinction was made between expressions that were used in agent and/or subject role (A/S) and those that were used in other roles (not-A/S). From a semantic point of view, the A/S expressions were accompanied by predi cates representing nonverbal or verbal actions performed with volition by an agent (hit, go, say), as well as internal processes or states (know, think), processes affect ing the referent (fall, die) and various properties, including location (e.g., be big, beautiful, on the fence). Second, each expression was characterized as having one
170
M. HICKMANN
of the following three types of contexts: (1) the expression was preceded by a coreferential expression and both were used in A/S role within their utterance (coreference A/S); (2) the expression was preceded by a coreferential expression, but they were not both in A/S role (non-A/S coreference); (3) the expression was not preceded by a coreferential expression (noncoreferential context). Tables 3 and 4 summarize the results concerning the co-text of different refer ring expressions.11 These tables show that the great majority of pronouns (explicit and zero) were preceded by coreferential expressions and that this coreference rela tion was of the A/S type. Inversely, nominals (regardless of determiners) were more frequently used in noncoreferential contexts. This phenomenon holds for both picture sequences and for all ages, although it is most marked for sequence A and at 10 years. Examples (11) and (12) above illustrate this aspect of the narratives. In (11) the child stops using pronouns to denote the main characters (horse) at the point where one of the secondary characters (bird) becomes active in the plot in a sequence of A/S roles ("and then a bird came along and bandaged the horse up"). In (12) the se quence involves frequent changes in which referents occur in A/S role across suc cessive utterances. No pronoun is used except when the same referent is mentioned in A/S role across two adjacent utterances ("and the cat was trying to climb up the tree and he was almost there").
Table 3
Relation between Expressions and Coreference in the Co-Text (Sequence A)
Age 4 Yrs 7 Yrs 10 Yrs
Coreferential A/S
Non-A/S
Non coreferential
Zero/Pro
62%
21%
18%
Def/Ind
33%
17%
50%
(76)
Zero/Pro
74%
18%
9%
(245)
Def/Ind
30%
19%
51%
(95)
Zero/Pro
76%
14%
10%
(394)
Def/Ind
14%
7%
78%
(110)
Expression
Total (141)
171
THE DEVELOPMENT OF DISCOURSE COHESION
Table 4
Relation between Expressions and Coreference in the Co-Text (Sequence B)
Age 4 Yrs 7 Yrs 10 Yrs
A/S
Non-A/S
Noncoreferential
Zero/Pro
56%
27%
17%
(187)
DeF/Ind
24%
19%
57%
(187)
Zero/Pro
62%
26%
11%
(223)
Def/Ind
13%
19%
68%
(285)
Zero/Pro
76%
14%
10%
(394)
Def/Ind
16%
15%
70%
(349)
Expression
Coreferential
Total
At all ages, then, children presupposed the referents in their narratives when the co-text had the following properties: reference maintenance was "dense" across suc cessive utterance and coreference was of the A/S type. However, the data also show a developmental progression in children's uses of nominals. Whereas the 10year-olds and some 7-year-olds used these expressions mostly in noncoreferential contexts, the younger children often used them in coreferential contexts of the A/S type. Examples (13) and (14), produced by two 7-year-old children illustrate this point. In (13) the first mention of the horse, which is explicit labelling ("it's a pony running"), is followed by the frequent repetition of a definite nominal ("the pony") from utterance to utterance, even when there are A/S coreferential relations. Similarly, in (14) the child uses nominals ("a horse", "the horse") to maintain refer ence to the main character in contexts where pronouns would have been clearly possible. (13) It's a pony running and then the pony sees a pal, the pony jumps over the ... the fence and then [...] the pony falls and hurts his leg so the cow bandages the pony's leg. (7 years) (14) A horse is running ... [...] a horse sees a fence ... a horse is jumping, the horse broke the fence and hurt his leg and now they're bandaging the horse legs - the horse leg up. (7 years)
172
M. HICKMANN
The repeated uses of indefinite determiners in example (14) can be compared to the repeated uses of these devices and/or of labelling in examples (9) and (10) above. The children "re-introduce" the character from one utterance to the next, as they go from one picture to the next (at least at the beginning of the narrative in (14)). From a functional point of view, one can therefore see a progression in how indefinite determiners are used in discourse. They are used at first in deictic labellings, then in other utterances, to introduce and re-introduce referents from one pic ture to the next. The repeated uses of definite nominals in contexts where pronouns would have been possible is also in part determined by the change from one picture to the next. Each utterance is independent from the preceding one and each is de pendent on the nonlinguistic context. In contrast, children later use oppositions in the referential system in order to link utterances together within discourse, resulting in the following properties of their narratives: (1) they introduce referents in ut terances where the expressions are not deictic; (2) they do not re-introduce them from one picture to the next; (3) they presuppose them when it is possible to do so (A/S coreference across adjacent utterances); (4) they do not presuppose them when it is not possible to do so (for example, when there is a change of A/S across ad jacent utterances). Finally, note the child's use of the predicate got hurt in example (11) above ("he broke the fence and 0 got hurt"). This type of utterance presents a referent which had an active role earlier in the narrative as a patient undergoing a process (here it undergoes the consequences of its own actions), while still presenting this referent as "subject" (zero anaphor). A related but more complex case is shown in an ex cerpt from a 10-year-old's narrative in example (15): (15) [...] So ... urn ... she started to climb the tree. But unfortunately she didn't get ... to the first branch ... before the dog came. And the poor little cat got pulled down ... by the dog [...] (10 years) Here the child maintains reference to the cat with a series of pronouns in A/S role, then with a definite nominal (after a mention of the dog in the role A/S) in a full ger-passive that presents the cat in the role of patient/subject. Such uses show the further development of discourse cohesion, involving complex clause structures to ensure referential continuity.
THE DEVELOPMENT OF DISCOURSE COHESION
173
Cross-Linguistic Perspective Acquiring the discourse-internal functions of linguistic devices, and thereby the ability to use language as its own context, is an essential part of becoming a mature native speaker, regardless of the particular language which is being acquired. In this respect, the developmental progression which was found with the referring ex pressions of English-speaking children is compatible with the general hypothesis that anaphora derives from deixis both in diachronic and in ontogenetic language development (e.g., Lyons 1975, 1977). More generally, universal functional prin ciples play a role in how all children learn to use referential devices in discourse. Thus, cross-linguistic analyses (e.g., Li 1976; Silverstein 1976, 1987; Givon 1983; Tomlin 1987) have shown the existence of universal principles of discourse or ganization which involve interaction among intrasentential properties of referring expressions (e.g., their referential content, their propositional role within the clause) and intersentential properties, particularly the degree to which referents are presup posed across clauses (what is the "topic" at a given point of discourse). However, although all languages provide some means to solve discourse problems that must be confronted by all children, these means differ a great deal from language to language, making any given discourse problem more or less com plex for them. It is therefore necessary to study children's uses of referential devices in discourse within a cross-linguistic perspective in order to determine the extent to which the specificities of different languages might affect the develop mental course of discourse cohesion. We can illustrate some of the issues that arise in such a cross-linguistic approach by briefly comparing the typical formal means available for referent introductions and reference maintenance in English with those available in other Indo-European languages (e.g., French), as well as in a typologically different language (e.g., Mandarin Chinese). The ways in which children use these formal means are illustrated below on the basis of a presently ongoing com parative study for which narratives were collected in these languages with some of the same materials (the picture sequences shown in Figure 1) and the same proce dures as those discussed above (for more details, see Hickmann 1987b).
Referent-Introductions and Reference-Maintenance in Chinese While the use of determiners is obligatory with all nominals (at least singular ones) in Indo-European languages, determiners are not formally obligatory in Chinese. The uses as opposed to nonuses of these devices with nominals, however,
174
M. HICKMANN
do typically mark both the semantic distinction between specific versus nonspecific reference and the pragmatic distinction between given versus new information in discourse. Chinese determiners consist of numerals or demonstratives used in conjunction with a classifier (either specific classifiers or the general classifier -ge) and fol lowed by a nominal.12 The uses of the following most frequent types of expressions can be summarized briefly as follows (for more details, see, for example, Li and Thompson 1981): (1) nominals with numeral and classifier can only be used when reference is specific and when the existence/identity of the referent is not mutually known (e.g., "yil-pi2 ": "one-CL-horse"); (2) nominals with demonstrative and classifier, as well as pronouns (also see note 2), can only be used when reference is specific and when the existence/identity of the referent is mutually known (e.g., "nei4-ge ": "that horse"; "tal": "he/him/she/her/it/them"); (3) bare nominals must be used when reference is nonspecific and they can be used when reference is specific, regardless of whether the existence/identity of the referent is mutually known (e.g., "": "horse").13 In addition, as shown in examples (16) to (19) (from Li and Thompson 1981), these contrastive forms interact with topicality and clause structure.14 Thus, in (16) and (17) indefinite forms in principle cannot be used in (preverbal) sentence-initial position, a position reserved for necessarily definite topical NP's (see below). In addition, as shown in (18) and (19), bare nominals in preverbal position are definite (the referent must be mutually known). When they are in postverbal position, they are indefinite (the referent is not mutually known): (16) lail-le yil-ge renl come-PCL one-CL person (A person came) (17) *Yil-ge renl lai2-le one-CL person come-PCL (18) renl lail-le person come-PCL (The person(s) has/have come or People came) (19) lail-le renl le come-PCL person PCL (A/some person(s) has/have come) This interaction between form and clause-structure must be seen in the light of the typological properties of Chinese as a so-called "topic-prominent" language, in
THE DEVELOPMENT OF DISCOURSE COHESION
175
contrast to languages that have been characterized as "subject-prominent" languages (e.g., English). Utterances in Chinese are better characterized by a topic-comment structure than by a subject-predicate structure. Thus, Chinese topics contrast with English subjects in many ways that can be roughly summarized here as follows (for more details, see Keenan 1976; Li and Thompson 1976): topics need not be in tegrated within the semantic/syntactic network of the clause. For example no NP agrees with the verb formally (Chinese verbs are uninflected) and any NP can be a topic, independently of the selectional restrictions or case frame of the verb, as long as the NP is either generic or definite (i.e., either reference is nonspecific or it is specific and the referent must be mutually known). Analyses of the Chinese narratives produced by adults show that they used both oppositions among NP types and word-order to differentiate the first mentions of animate referents as opposed to subsequent ones. Thus, as shown in examples (20) to (22), when introducing referents, they used nominals with numeral determiners (most often accompanied by specific classifiers) in postverbal position in various types of constructions. In (20) a nominal with numeral is used to introduce a referent in an existential construction. In (21) and (22) the same NP types are used for referent introductions and placed after the verb. (20) You2 yil-zhil niao3 malma have one-CL bird mother (there is/was a mother bird) (21) pao3-lai2-le yil-zhil maol run-come-PCL one-CL cat ((there) comes/came a cat running) (22) zhei4-ge lil bal shang4-mian4 zhan4-zhe this-CL fence top-surface stand-PCL (On this fence is/was standing a bird)
yil-zhil one-CL
niaor3 bird
In contrast, (23) and (24) illustrate how referring expressions were used in refer ence maintenance. They consisted of pronouns, bare nominals, or nominals with demonstrative determiners (in conjunctions with either specific classifiers or with the general classifier -ge), most often placed in preverbal position (depending to some extent on their semantic roles, e.g., preverbal agents, postverbal patients). (23) tal lai2-le 3p come-PCL (He came) (24) (nei4-ge)
pao3-le
176
M. HICKMANN this-CL horse gallop-PCL ((This) The horse galloped)
The narratives produced by the Chinese children indicate a developmental progression both in the uses of local markings in the NP's and in the uses of wordorder variation for referent-introductions. Example (25) shows the beginning of a 4year-old's narrative where neither type of marking is used to introduce referents. First mentions (in bold) consist of bare or demonstrative nominals, used either in labellings or in preverbal positions, and they are therefore inappropriate. (25) 1.
2.
3.
4.
5.
6:
Gel... zhe4-ge gelzi he2 xiao3 gelzi pigeon this CL pigeon and little pigeon (The pigeon... this pigeon and the little pigeons,) ran2hou4 ne tal... zou3le after PCL 3p leave PCL (and after he... left) Ran2hou4 xiao3 gelzi maol kan4-jian4 le after little pigeon cat look-see PCL (and after the little pigeons the cat saw (them)) Tal yao4... chil 3p want eat (He wanted to... eat) Tal jiu4 pa2 shu4 3p then climb tree (So he climbed up the tree) Ran2hou4 gou3 ran2hou4 gou3 diaol-zhe maol after dog after dog bite PCL cat de yi3ba /.../ POS tail /.../ (afterwards the dog bit in the tail of the cat /.../)
In contrast, example (26) shows excerpts from a 7-year-old's narrative. The child uses an inappropriate first mention in (26.1) (preverbal bare nominal: niao3) and some appropriate ones in (26.2) and (26.5) that combine both numeral deter miners and postverbal positions ("lia2le yil-zhil hu2U2" and "jiu2 lai2le yil-zhil gou3").
THE DEVELOPMENT OF DISCOURSE COHESION
26. 1.
2.
3.
4.
5.
6.
Niao3 zai4 wol bird be-at nest in (A bird was in the nest) lai2-le yil-zhil hu2U2 come-PCL one-CL fox (there came a fox) i feil-zo /.../ bird fly-leave /.../ (the bird flew away /.../) hu2li2 pa2-shang4 shu4 qu4 fox climb-ascend tree go (the fox climbed up the tree) jiu4 Iai2-le yil-zhil gou3 then come-PCL one-CL dog (but then a dog came) gou3 yao3-zhu4-le hu2li2 de POS dog bite-ZHU-PCL fox (the dog bit the fox's tail /.../)
wei3ba tail
177
/.../ /.../
On the whole, preliminary analyses indicate that Chinese children learn local markings on NP forms before they learn word-order variation for referent intro ductions. The data also indicate, however, that Chinese children may acquire these local markings later than English-speaking children. This result may be be cause local NP markings are not obligatory and interact with word-order in adult discourse to a much greater degree than in English.
Referent-Introductions and Reference-Maintenance in French Although French has often been described as a subject-prominent language, some analyses have shown that spoken French is in fact similar to a topicprominent language such as Chinese, or at least that both subject-predicate and topic-comment structures coexist (cf. especially Givon 1976; Lambrecht 1981, 1987b).16 For example, it has been argued that the existence of "dislocated" struc tures such as (27a) and (27b), in contrast to (27c), are best described as topic-com ment structures, where the full dislocated NP is the topic (27a) or antitopic (27b) and the pronoun is a marker of topic-verb agreement. Indeed, the dislocated topic
178
M. HICKMANN
NP has many of the properties of Chinese topics, e.g., it need not be integrated in the clause as subjects must be, it is always sentence-initial, and, as shown in (27d) and (27e), it must be definite if reference is specific (with nonspecific reference, a different pronoun must be used). 27. a. b. d. e.
Le chien il court (The dog he runs) Il court le chien (He runs the dog) Un/le chien court (A/the dog runs) *Un chien il court (*A dog he runs) Un chien ça court (Dogs run) (generic)
Analysis of some 4- to 10-year-olds ' narratives show a developmental progres sion in the uses of such dislocated structures. We illustrate one aspect of this progression here (for more details, see Hickmann 1987b, 1988). Children under 10 years frequently used left dislocations, not only to maintain reference to entities, but also to first mention them in discourse. In a few cases (mostly found among the younger children) these first mentions consisted of a dislocated indefinite form. Ex ample (28) illustrates such uses (in bold) in an excerpt from a 4-year-old's narrative of sequence ("un cheval i court" (a horse 'e's running), "une bamby elle court" (a bamby, she's running)). Although indefinite forms as such are appropriate for referent-introductions, they are actually inappropriate in these constructions since the topic NP must be definite when reference is specific. (28) Un cheval i court. Euh une bamby elle court... elle regarde le cheval. Mum... le cheval i court et puis la ch- le cheval... i regarde la bamby Euh... le cheval est mort... i ferme les yeux, il est par terre. /.../ (4 years) (A horse 'e's running. A bamby she's running... she's looking at the horse. Mm... The horse 'e's running and then the h- the horse 'e's looking at the bamby Uh... The horse is dead... 'e closes his eyes, he's on the ground. /.../) In the great majority of the cases, however, dislocations that were used (for both referent-introductions and reference-maintenance) consisted of dislocated definite
THE DEVELOPMENT OF DISCOURSE COHESION
179
NP's. Example (29) illustrates such uses in a narrative produced by a 7-year-old for sequence (first mentions are in bold). (29) Un arbre avec un nid et des petits... trois petits et avec la maman. Après il y a encore un arbre et trois petits, la maman qui s'évole et le chat il vient. Y a un autre arbre, les trois... ils y sont encore les trois et le chat qui es saie (décide?) qu'il va monter. Après il monte et puis le chien il vient et après sur la cinquième le chat il - il est... jusqu'au nid et le chien il lui mord la queue. Et après la maman elle revient, elle apporte du fil pour les petits canards et puis... elle leur donne et puis après, le chat il est - il... il boude. (7 years) (A tree with a nest and some little ones... three little ones and with the mother. Then there is still a tree and three little ones, the mother who flies away and the cat he comes. There's another tree, the three... they're still there the three and the cat who tries (decides?) to climb up. Then he climbs up and then the dog he comes and then on the fifth one the cat he - he is... up to the nest and the dog he bites his tail. And then the mother she comes back, she brings string for the little ducks and then... she gives [it] to them and then after, the cat he is - he... he sulks.) Dislocations are relatively frequent in (29) for both first and subsequent men tions of referents. With respect to first mentions, the narrative opens with a series of labellings that constitute deictic introductions of the mother and baby birds. Fur ther into the narrative, both the cat and the dog are introduced with left-disloca tions. Note that the uses of left-dislocations are highly systematic. They all occur when there is a "topic-shift" across adjacent clauses, regardless of whether the dis located NP's constitute first of subsequent mentions. When no "topic-shift" occurs, the referent is denoted with a pronoun. Such examples show that, although some children under 10 years have learned the "topic-shift" function of left-dislocations, they have not yet learned to reserve such constructions for reference-maintenance only, i.e., children have to learn to exclude them when introducing referents that are not mutually known. Similarly, example (30) illustrates the uses of left-dislocations from a 4-yearold's narrative of sequence (first mentions are in bold type). (30) L'oiseau il-il-il est avec des bébés, le chat il regarde les oiseaux, la maman elle part et le chat il essaye de monter à l'arbre. Il grimpe. Et après il
180
M. HICKMANN
regarde encore les oiseaux et il va remonter à\arbre. Regarde, il essaye là encore de monter à l' arbre. Et après il grimpe et le chien il (inaudible)... il-il crie une fois et... et le chien va lui mordre la queue. Il recommence à mordre la queue. Et après le chat il court. Et la maman elle revient et le chien il revient là après et le-le chien il court après lui. (4 years) (The bird he-he-he is with some babies, the cat he looks at the birds, the mother she leaves and the cat he's trying to climb up the tree. He climbs. And then he's still looking at the birds and he's going to climb up the tree again. Look, he's still trying here to climb up the tree. And then he's climbing and the dog he (inaudible)... he-he screams once and... and the dog is going to bite his tail. He starts to bite his tail again. And after the cat he runs. And the mother she comes back and the dog he comes back there then and the-the dog he runs after him.) In this case, the narrative opens with a left-dislocation, which constitutes the first mention of the mother bird, and where the baby birds are simultaneously intro duced ("l' oiseau il est avec des bébés" (the bird it's with the babies)). As the narrative unfolds both the cat and the dog are first mentioned in left-dislocations. Note the use of the left-dislocation which closes the narrative ("et le chien il court après lui" (and the dog he runs after him)). The child uses a full NP and pronoun to refer to the dog, although it is topical at this point in the discourse, and he uses a pronoun to refer to the cat, although it is less topical. On the whole, preliminary analyses indicate developmental progression in how French children learn to differentiate referent introductions from reference maintenance. Some of these progression types are specific to these children and they occur be tween 4- and 7-years (learning to use determiners), as well as between 7- and 10-years of age (learning to restrict dislocated structures to reference main tenance).
Conclusion The progression that was found in children's narrative organization shows the gradual development of the capacity to use language as its own context. In the par ticular narrative situation that was discussed here, the referents were present, but they were visible only to the children and not to the interlocutor. As we saw from the English data, younger children do not use referring expressions in the same way as older ones. In particular, they use the referential system at first with deictic tunc-
THE DEVELOPMENT OF DISCOURSE COHESION
181
tions, rather than with strict intralinguistic ones. With respect to referent introduc tions, they use devices that presuppose the referents on first mention (e.g., definite nominals), as well as other devices (e.g., nominals with indefinite determiners) in deictic labelling, sometimes repeated from one picture to the next. The latter uses could constitute appropriate introductions in some situations, in the sense that they draw the interlocutor's attention to referents (e.g., see Keenan and Klein 1975; Keenan and Schieffelin 1976; Ninio and Bruner 1978; Atkinson 1979; etc.). How ever, they are still directly bound to the nonlinguistic context rather than to the un folding co-text, and therefore not as "appropriate" in this situation, where maximal reliance on discourse is required. Young children, then, either do not introduce referents in discourse appropriately, or they introduce them with primitive deictic means that do not contribute to discourse-internal cohesion as such. Analyses of reference maintenance show that children's (and adults') use of referring expressions are highly constrained by the nature of the coreferential rela tions across utterances. In particular, they use relatively presupposing devices (ex plicit and zero pronouns) when they maintain reference to the same character across adjacent utterances and when these coreferential relations involve expres sions in agent and/or subject role. Inversely, they use less presupposing devices (nominals) to denote a referent if it has not been mentioned in the immediately preceding clause at all or if it has not been mentioned in agent/subject role in both clauses. In contrast to other studies (e.g., Karmiloff-Smith 1981), it seems, then, that from 4-years on, reference maintenance follows some of the rules of the adult system which involve interactions among intrasentential properties (e.g., role of NP within the clause) and intersentential properties (e.g., pragmatic role of NP as more or less topical in discourse). However, as was found in other studies, reference maintenance in young children's narratives also shows that in some respects it is deictic. For example, these children often mentioned the same referent with repeated deictic labellings (at 4-years) and with nominals in agent/subject coreferential relations (at 7-years) across adjacent utterances. Thus, it is not until after seven years that children stop re-introducing referents across utterances as they go from picture to picture and that they rely maximally on presuppositions established within discourse. These properties of children's narratives, then, give the impression that until at least 7 years old, children presuppose referents either too much or too little. These aspects of children's repertoire show that their mastery of the referential system is a late development, at least if one includes as part of the native speakers' competence uses of language that are strictly intralinguistic. When the speech situation is con trolled in such a way as to "invite" children to rely necessarily and maximally on
182
M. HICKMANN
discourse to communicate, their uses of devices both to introduce referents and to maintain reference, indicate a dependence on nonlinguistic context. Similar con clusions can be drawn from narratives elicited in other situations, some of which show types of progression that occur yet later in the development of discourse cohesion (for results based on other situations, see Hickmann 1982, 1984, 1987a). These and other related results based on English can be summarized in terms of a developmental progression whereby children learn to "anchor" linguistic devices within discourse, using discourse-internal function forms which were at first only deictic in their repertoire. Whatever cognitive skills might be involved in this development, acquiring the discourse-internal functions of referring expressions has important implications for children's communicative skills, because it provides them with the necessary tools to rely maximally on discourse when reliance on nonlinguistic context is not possible in the speech situation. Thus, what has been typically described as cognitive capacity for "displaced reference" or "decontextualized communication", becomes possible when children learn to use language as "its own context". As we have seen, the particular ways in which children learn how to use lan guage as its own context may vary depending on the language they are acquiring. Although all children must learn some set of discourse-internal constraints on the uses of linguistic devices in order to communicate efficiently, and although these constraints involve universal functional/pragmatic principles of discourse organiza tion, there are wide cross-linguistic variations in the particular means available for discourse cohesion. Little is still known about cross-linguistic similarities and dif ferences in the ontogenesis of reference in discourse from a functional/pragmatic point of view. As we have illustrated above on the basis of a brief comparison of children's narratives in English, French and Chinese, the particular developmental course of discourse cohesion may vary across languages. For example, given the properties of Chinese, Chinese children may learn the uses of NP variation for referent introductions later than English-speaking children, although they may learn other means earlier, e.g., word-order variation. Similarly, given the properties of French, French children's acquisition of NP variation for referent introductions re quires that they learn to reserve some types of clause structures (e.g., dislocations) for reference maintenance. Such language-specific aspects of the development of discourse cohesion clearly must be taken into account when making claims about children's language and cognitive development on the basis of their discourse or ganization. In this respect, a final general remark can be mentioned here concerning some further implications of discourse cohesion in child development. It is highly prob-
THE DEVELOPMENT OF DISCOURSE COHESION
183
able that the development of discourse cohesion implies not only more efficient uses of signs in communicative situations involving other people but also more ef ficient uses of signs in other situations, e.g., those involving internal reasoning. Such a hypothesis would be more compatible with theories of ontogenesis that as sume mutual influences between language and cognitive development (e.g., Vygotsky 1962) than with those that assume that language development is a by product of independent underlying cognitive processes (e.g., see discussions in Ochs 1987; Karmiloff-Smith 1987; Hickmann 1986, 1987a). A full discussion of the issue is beyond the scope of this paper. It should be pointed out, however, that it has implications for the kinds of inferences we can make from children's language use. Thus, as we have mentioned, the notion of cognitive "egocentricity" has often been invoked to explain children's referential system, independently of the functional/pragmatic properties of their linguistic repertoire. Such a notion should be at least qualified in such a way as to take into account the fact that language is a semiotic reality in itself that must be confronted by all children and may be a dynamic mechanism in development. The semiotic properties of language are partially the same for all children and partially different across linguistic systems. Further research is necessary to disentangle these two aspects of language development in order to determine how all children learn to use language as its own context.
NOTES 1. The editor of this volume invited me to contribute this synthetic paper which includes both re search I have discussed elsewhere and ongoing research not published previously. 2. There is a great deal of variation in how lexical information is encoded in pronouns across languages. For example, the pronouns of Indo-European languages typically encode person, number, gender, and case distinctions. In contrast, Chinese pronouns do not distinguish gender and case (e.g., tal "he/him/she/her/it") and, although some plural forms are available, they are not always used in the third person (e.g., talmen "they/them" can sometimes be replaced by tal). 3. The factors effecting the extent to which more or less presupposing forms can be used in dis course to ensure referential continuity are too numerous and varied to be discussed in detail here. They include structural constraints, register, irony, formality of the communicative situa tion, episodic structure, the extent to which one among several referents is more "central" than the others, etc.
184
M. HICKMANN
4. It is difficult to evaluate the conclusions of some previous studies when insufficient contextual information is provided. For example, Huxley (1970) concludes from example (7) that the pronoun he in "he can fly upside down" is anaphoric (rather than deictic) without specifying: (1) whether the referent was present and (2) whether some preceding utterance contained an introduction of this referent. 5. For example, the distinction between the determiners a and the is sometimes difficult to hear in some types of phonological environments. 6. Utterances such as "a horse running" have at least two possible interpretations. The most like ly interpretation is that they are deictic labellings (akin to "It/this is a horse (who is) run ning"). Another interpretation is that the NP is the subject in a subject-predicate construction without an auxiliary (akin to "a horse is running"). 7. In all the examples cited, the symbol [...] indicates that the child's interlocutor gives feedback, e.g., in the form of encouragement, repetition, or punctuation (uh-hum). The symbol /.../ indi cates that part of the child's narrative was left out. 8. Most zero anaphors were "free", namely they were in free variation with other NP's in coor dinated or juxtaposed utterances. Some occurrences were used in restricted contexts where syntactic constraints played a role. 9. More precisely, the only difference among the referents in sequence was that the baby birds were the least presupposed. 10. Cases of coreference within utterances (e.g., the reflexive and possessive pronouns in "he hurts himself', "he falls on his back"), are not included in these analyses. 11. For these analyses of the co-text, all expressions used in explicit or potential labelling are ex cluded, since they are either the predicate or ambiguous (also see note 6). 12. Chinese speakers can use either the general classifier (-ge) or specific classifiers. Different specific classifiers must be used as a function of various dimensions characterizing classes of referents (features such as animacy, shape, size, function, material, etc.). For example, in the stories elicited for this study, if children used specific classifiers, they had to use different ones for the various animate referents (e.g., pi3 for the horse, zhil for the cat, the dog, and for the birds, tou2 for the cow). If adjectives and other modifiers are used, they are inserted before the nominal (e.g., "yil-pi3 da4 ma3"\ "one+classifier big horse"). 13. The constraints on the uses of zero elements in Chinese are also different from those dis cussed above for English. For example, while subjects are obligatory in most declarative ut terances of English (e.g., in contexts other than juxtaposition and coordination), zero anaphors can be used in Chinese in any structural context, as long as the referent can be presupposed from the discourse situation.
THE DEVELOPMENT OF DISCOURSE COHESION
185
14. In all Chinese examples below, both morphemic and free translations are provided. The fol lowing symbols are used: in the Chinese text, numbers correspond to tones; in the morphemic translation, 3p corresponds to third person pronouns, POS to the possessive marker, PCL to various particles, CL to classifiers. Since Chinese verbs are not inflected, they can be trans lated in several ways (e.g., singular/plural, past/nonpast tense, etc.). Some verbs consist of complex verbal constructions and some forms can be translated either as verbs or as preposi tions). 15. Note that such a clausal structure is quite common for referent introductions in Chinese and is not "stylistically" marked as it is in other languages (e.g., English). 16. Written (standard) French does fit the subject-predicate structure of English. Spoken (nonstan dard) French either does not fit this structure or only fits it partially, although it is not clear which variable or combinations of variables determine which structure is used (e.g., including socio-linguistic variables such as social class, formality or informality of the communicative situation, spoken versus written channel, etc.). 17. Although French has verbal suffixes, the distinctions they encode are much less specific than those of other languages (e.g., German), particularly in the spoken channel (see Givon 1976; Lambrecht 1981, 1987).
Coherence: Language as it Underlies and Organizes Knowledge
Effect-events are certainly causally related to their physical causes. However, this is not a necessary relation, but the one that is expressed. On the other hand, be tween these effect-events and their physical causes, or between the former and their ideal cause, there exists a relation which is no longer even a causal one, but simply one of expression. The question thus becomes: What is the expressed relation among events? Among events there seem to form extrinsic relations of silent compatibility and in compatibility, of conjunction and disjunction, which are very difficult to evaluate. What makes an event compatible or incompatible with another? G. Deleuze, 1969, La Logique du Sens. Paris: Editions de Minuit (our translation).
An Opaque Text as a Mobilizer of Knowledge M. Dolitsky Etudes Interculturelles de Langues Appliquées Université Paris VII
and G. Piéraut-Le Bonniec Laboratoire de Psychobiologie de l' Enfant, Paris
"The meaning of utterances is not part of the structure of the utterance, but as signed ... by the language user" (Van Dijk 1977:2) who, far from being dumb, has his/her own inner language. An Other's discourse passes through the filter that this inner language constitutes (Bakhtine 1929 (1977)). This mediatization creates a cer tain amount of uncertainty in the correspondence between speaker and listener meaning. However, this uncertainty that accompanies all speech acts normally remains unnoticed by interlocutors and the comprehension process is generally un conscious (Hakes 1980). It is only when the obscurity in the Other's words goes beyond a certain limit that the reader or listener becomes conscious of the com prehension process. And in this situation the comprehension strategies that are brought into play to automatically clear up unintelligible forms might be observed. Obviously, a study concerning the interpretation itself of nonsense texts would be of little value as most interpretation is idiosyncratic. However, the study of in terpretations as an indication of the strategies available to a listener/reader, which themselves are limited in number, should offer important enlightenment in the study of comprehension and meaning. Nonsense is a literary genre that puts a reader into this type of situation. When faced with a text based on lexical nonsense, i.e. a text that contains a high number of neologisms for which no commonly accepted, or coded, meaning has yet ac crued, a reader must consciously turn to meaning-finding strategies that can be ap plied to the new elements. Nonsense does not mean no sense; however the relations between words and meaning, words and world, remain opaque. These texts are
190
M. DOLITSKY AND G. PIÈRAUT-LE BONNIEC
carefully structured and it is precisely by attending to the structure that meaning will appear. It is the generally accepted meanings and uses of the coded morphemes that serve as a base from which the meanings of the uncoded ones can be established. Warrin's translation (1931; cf. texts below) of Lewis Carroll's (1871) "Jabberwocky" is, like the original, a perfectly balanced combination of precoded gram matical and lexical morphemes and of uncoded ones, i.e. neologisms. Thus it was chosen for this study of comprehension. Since Ebbinghaus (1885) first worked with neologisms, many experiments using them have been carried out in the field of learning. Berko (1958) was the first to use them to test children's acquisition of syntax, and this technique was picked up by countless others. The difference between the work done here and other work with neologisms is that they are not being presented in isolation or as part of a sen tence out of context, but, in number, as integral elements of a complete and coherent text. The importance of their use here is that the readers find themselves in a situation where they must develop meanings from the different neologisms in relation to their role as part of a larger text and context, where the known and un known interact. On the one hand, each neologism contributes to the meaning of the text as a whole in a down-up process: on the other, the context will suggest the meanings that the neologisms can take on in a top-down process. Finally, the same sort of effects will exist on a microcontextual level, where the neologism is part of a phrase and relates syntagmatically to the other words surrounding it, be they part of the language or other neologisms. Moreover, we feel that 11-year-old children offer a particularly interesting sub ject of study. Their linguistic capacity is fairly comparable to the normal adult speaker's, although their skill in constructing texts is not quite on the par of an adult's. Results of their work should then offer indications to general comprehen sion abilities while at the same time offering spontaneous samples of productions as witness to children's still developing skills in the area of text construction and comprehension. It also seemed that children would be amenable to working on a nonsense text. The interpretation of nonsense was studied from two main perspectives: first, the effect the opaque text had on the children's language as they attempted to explicate it and second, the way in which word meaning could be gleaned from the text. Thus it is an enquiry into both comprehension and production.
AN OPAQUE TEXT AS A MOBILIZER OF KNOWLEDGE
Jabberwocky
191
Le Jaseroque
'Twas brillig, and the slithy toves Did gyre and gimble in the wabe: All mimsy were the borogoves, And the mome raths outgrabe.
II brilgue: les tôves lubricilleux Se gyrent en vrillant dans le guave, Enmimés sont les gougebosqueux, Et le mômerade horsgrave.
['Beware the Jabberwock, my son! The jaws that bite, the claws that catch! Beware the JubJub bird, and shun The frimious Bandersnatch!"
Garde-toi du Jaseroque, mon fils! La gueule qui mord; le griffe qui prend! Garde-toi de l'oiseau Jube, évite Le frumieux Band-à-prend!
He took his vorpal sword in hand: Long time the manxome foe he sought So rested he by the Tumtum tree, And stood awhile in thought.
Son glaive vorpal en main il vaT-à la recherche du fauve manscant; Puis arrivé à l'arbre Té-Té, Il y reste, réflichissant.
And, as in uffish thought he stood, The Jabberwock, with eyes of flame, Came whiffling through the tulgey wood, And burbled as it came!
Pendant qu'il pense, tout uffusé Le Jaseroque, à l'oeil flambant, Vient siblant par le bois tullegeasis, Et burbule en venant.
bne, two! One, two! And through and through The vorpal blade went snicker-snack! He left it dead, and with its head He went galumphing back.
Un deux, un deux, par le milieu, Le glaive vorpal fait pat-à-pan! La bête défaite, avec sa tête, Il rentre gallomphant.
['And hast thou slain the Jabberwock? Come to my arms, my beamish boy! O frabjous day! Callon! Callay!" He chortled in his joy.
As-tu tué le Jaseroque? Viens à mon cœur, fils rayonnais! jour frabbejeais! Calleau! Callai! Il cortule dans sa joie.
'Twas brillig, and the slithy toves Did gyre and gimble in the wabe: All mimsy were the borogoves, And the mome raths outgrabe.
Il briligue: les tôves lubricilleux Se gyrent en vrillant dans le guave, Enmimés sont les gougebosqueux, Et le mômerade horsgrave.
192
M. DOLITSKY AND G. PIÉRAUT-LE BONNIEC
Method "Le Jaseroque" was handed out to 29 9-year-old French children in two separate (arbitrarily divided) groups during their regular Friday afternoon class . They were seated on the floor with their teacher and asked to read it to themselves. Many in terrupted their reading to exclaim that the text did not mean anything, but they were told to please finish the text which they would discuss when everybody had finished. The teacher asked for volunteers to read the text out loud and opened the discussion of the meaning of the poem. This discussion was taped (which the children knew) and later transcribed. At the end of the discussion, the neologisms were written on the board, and the children were asked to choose among them those that expressed the most for them and then write what the neologisms meant. Each class had between one hour and one and a half hours of discussion and about a half-hour writing time, although one group had a ten-minute recess in the middle of their discussion.
Results In this section we will first take into account the children's reactions to this type of text and second, we will study the strategies brought into play by the children as they attempt to make sense of the text.
The Children's Reaction to the Text The children reacted first by an avalanche of metadiscourse (cf. Kintsch and Van Dijk, 1978) expressing their rejection of the text. For example they said: - "ça ne veut rien dire" (it doesn't mean anything) - "ce sont des mots farfelus" (they're crazy words) - "c'est des mots l'un à coté de Vautre ... il y a des mots qui veulent rien dire" (they're words one after another ... there are words that don't mean anything) - "c'est bizarre" (it's strange)
AN OPAQUE TEXT AS A MOBILIZER OF KNOWLEDGE
193
- "c'est quand même pas un mot correct" (it's still not a correct word) At first, the teacher had to push the children to read the text to the end and to get them to try to make some sense of it. Then, after the discussion was underway, there was a change in their attitude. They no longer said that the text did not mean anything, but expressed, instead, their own incapacity to understand the text: - "moi, je ne comprend aucun mot" (I don't understand a word) - "je ne comprend rien" (I don't understand anything) - "ça, je ne sais rien" (I don't know anything about it) - "ça me trouble un petit peu" (It upsets me a little bit) - "ça me gène" (It bothers me) The teacher tried to get them to use their imaginations to figure out the mean ings of the words, but they found this inacceptable in that, for them, as they ex plained, words have meaning and the author wrote what he did to say something meaningful, even if the reader might have trouble understanding what it is. They refused to invent a meaning because they felt that their role was to find out what the author meant and not to make it up, themselves: - "/7 a voulu dire quelque chose le monsieur en écrivant ces mots, ça veut dire quelque chose" (the man meant something by writing these words; it means something). - "il n'a pas fait des mots n'importe comment" (he didn't make up these words any old way) - "même qu'il dit des mots qu'on ne comprend pas, ça va encore; il a raconté une histoire, cela veut dire quelque chose, il n'a pas écrit comme ça" (even if he says words we can't understand, that's still OK; he told a story; that means something; he didn't write it just like that) The children found it completely unacceptable that meaning could be seen from the recipient's point of view as well as from the producer's point of view and that
194
M. DOLITSKY AND G. PIÉRAUT-LE BONNIEC
the two viewpoints could be relatively independent. If they rejected what they were being asked to do, it was because, for them, their could be no representation without the intention to represent (Pateman 1982). The children refused the right to decide on a word's meaning. Words exist before they are used by a speaker and it is not for the listener to supply their meaning after the fact. Words are used correct ly by speakers to say what they mean without any ambiguity. Of course, there are times one must guess the meaning of a word that isn't known, but there is a "right answer" which is unique; it is in agreement with the speaker's intentions and all guesses are not equal. Given the difficulties involved in the comprehension of the text, some of the children rejected it by investing it with a negative affect, as in the following ex amples: - "c'est horrible" (it's horrible) - "elle faisait peur cette histoire" (this story is scary) - "c'est une histoire où tout est sombre" (it's a story where everything is dark) - "il y a quand même quelque chose de mystérieux pour nous tous" (there's still something mysterious for all of us) - "il y a des herbes où il y a plein de mots étranges" (there is grass where there are lots of strange words) - "c'est la jungle que j'ai impression que ça se passe autrefois" (it's a jungle and I have the impression it's happening in the past) Sometimes the children tried to imagine the author: - "j'imaginerais quelqu'un d'ancien comme le grand-père ou Varrière grandpère qui raconte ça à son fils comme s'il était passé une catastrophe dans le pays où il était" (I would imagine someone very old like a grandfather or a great-grandfather who tells this to his son as if there had been a catastrophe in the country where he was) - "j'ai I'impression que c'est quelque chose qui dit mais pas quelqu'un, que c'est le langage du JASEROQUE" (I have the impression that it is something that speaks but not someone, that it's the JABBERWOCK's language)
AN OPAQUE TEXT AS A MOBILIZER OF KNOWLEDGE
195
- "j'avais pensé à un monsieur qui rève" (I thought of a man who is dreaming) The children had no confidence in their intuition to interpret the text. Not only did they often say, 'I don't know", but they used expressions like: "a little" or "al most" to introduce the idea of approximation, or they used expressions like "for ex ample", "perhaps", "it's possible that" to introduce ideas of possibility. They also used conditionals and alternatives: "ou" (or), "soit que ...soit que": - "c'est peut-être un peu un être humain" (perhaps it's a human being) - "/7 prend son..., je ne sais pas, son glaive" (he takes his..., I don't know, his sword) - "peut-être aussi par exemple, 'sais pas, il y a une forêt à traverser et il y a le JASEROQUE dedans, et peut-être pour aller soit à un autre village ou je ne sais pas quoi" (perhaps also for example, don't know, there's a forest to go through and the JABBERWOCK is in it, and perhaps to go either to another town or I don't know what) This absence of confidence in their ability to interpret what the text meant con tributed to the "movement" of the discussion, i.e. to modifications in opinions and ideas. Their own uncertainty helped the children to accept each other's ideas, to further develop them, even to the point of forgetting their own ideas: - "les autres, ils disent leur pensées, et puis après on en sait plus laquelle c'est la notre" (the others tell their ideas and then afterwards we don't know anymore which one is ours) It was found that the two groups focussed their attention differently. In one, the children were most interested in the JABBERWOCK itself, and it was described by using a number of strategies. There are descriptions by analogy or descriptions of the different parts of its body: - "un animal à quatre pattes, qui a une paire d'ailes et une gueule d'animal féroce avec des cornes et des griffes et puis un glaive qui lançait par lui-même en ouvrant la bouche"
196
M. DOLITSKY AND G. PIÉRAUT-LE BONNIEC
(an animal with four feet, that has a pair of wings and the head of a wild animal with horns and claws and then a sword that threw itself when its mouth opened) and descriptions of its personality: - "il n'est pas vraiment méchant, il a peur, 'est pour cela qu'il attaque" (it's not really mean; it's afraid; that's why it attacks) It is also described through its actions: - "il vient dans la jungle la nuit, il enlève tout sur son passage, et puis il fait ce qu'il veut avec tout le monde" (it goes through the jungle at night and takes away everything along its way and then it does whatever it wants to with everyone) On the other hand, the second group was most interested in reconstructing the story line: - "C'est une mère et un fils qui s'en vont dans les bois et ils voient il y a une espèce de crocodile volant et ils viennent le voir et un jour ils voient quelqu'un qui se fait attaquer par le JASEROQUE et la mère dit 'garde-toi du JASEROQUE' là-bas, et le petit il prend son glaive, il va pour le tuer. Ensuite il lui coupe les veines et le JASEROQUE tombe et ensuite il est tué" (There's a mother and a son who go into the woods and they see a sort of lying crocodile and they come to see it and one day they see someone getting attacked by the JABBERWOCKY and the mother says 'beware the JABBERWOCK' over there and the little one takes his sword; he goes to kill it. Then he cuts its veins and the JABBERWOCKY falls and then it's killed) - "Le JASEROQUE avance, il sort de la truc du bois et puis il le voit, il le regarde puis dans ses yeux, puis il prend son glaive et puis il coupe la tête et puis il s'en va à toute allure" (The JABBERWOCK advances, it takes out its wooden thing and then he sees it; he looks at it, then in its eyes; then he takes his sword and then he cuts off its head and runs off as fast as he can) This does not mean that there was no description at all, but that the children showed a preference to describe the story's movement which gives meaning to the
AN OPAQUE TEXT AS A MOBILIZER OF KNOWLEDGE
197
poem. In this group there were more discussions and differences of opinion. For example, the children threw themselves into a discussion of the number of animals in the poem: - "Le JASEROQUE, on le nomme par son nom, et puis il le traite par un autre nom d'un animal aussi féroce que lui et puis i le traite encore par un autre nom" (The JABBERWOCK is named by its own name, and then they call it by another name of an animal as ferocious as it is! and then it's called by still another name) - "L'oiseau JUBE et le JASEROQUE c'est deux animaux; celui qui parle, il dit, "garde-toi du JASEROQUE de l' oiseau JUBE" (The JUBJUB bird and the JABBERWOCK are two animals; the one who speaks says, "beware the JABBERWOCK, the JUBJUB bird) - "A mon avis, il y a trois animaux parce que il y a le JASEROQUE, l'oiseau JUBE et les MÔMERADES HORSGRAVES" (In my opinion there are three animals because there is the JABBERWOCK, the JUBJUB bird and the MOME RATHS OUTGRABE)
In this group some of the children endowed the characters with thoughts, as if they were trying to get into their psyche: - "pendant qu'il réfléchit le JASEROQUE de son côté près du bois, il le regarde et lui aussi il réfléchit s'il va sauter sur lui" (while he was thinking, the JABBERWOCKY on its own side near the woods was looking at him and he, too, was thinking whether to jump on him) - "il réfléchit, parce que, il sait pas si c'est mieux pour lui de le tuer ou pas le tuer, de l'autre côté le JASEROQUE se dit la même chose" (he thinks because he doesn't know if it's better for him to kill it or not to kill it; on the other side the JABBERWOCKY is wondering the same thing) A child could even seem to identify with the hero: Child: "il rencontre le JASEROQUE, et puis il réfléchit s'il doit le tuer ou pas, j'ai l' impression qu'il ose pas, alors il prend son glaive et puis, il le lance sans regarder et puis il part parce que c'est là qu'il se dit 'est-ce-que je l'ai tué?"
198
M. DOLITSKY AND G. PIÉRAUT-LE BONNIEC
(he meets the JABBERWOCK and then he thinks if he should kill it or not; I have the impression that he doesn't dare; so he takes his sword and then he throws it without looking and then leaves because that's where he wonders, 'have I killed it?') Teacher: "Et là, finalement, il a atteint le JASEROQUE ou pas?" (And there, finally, has he hurt the JABBERWOCKY or not?) Child: "mais je ne sais pas! il ne sait pas; il n'a pas regardé" (but I don't know! he doesn't know; he didn't look) However, the interest of this study was not how the subjects expressed their fan tasies which spring from their own personal experiences and feelings. The impor tant point was to see if it was possible to find specific strategies that were used to assign meaning to what had been presented, to see if there existed any cognitivelinguistic processes that were shared by all the children regardless of their idiosyncratic interpretations. In order to bring out these strategies, the children's in terpretations were studied from two different perspectives: 1. the way in which they assigned meaning to the poem as a whole, and 2. the way in which they assigned meaning to the neologisms.
General Meaning and Paraphrase Our interest here, lies in how paraphrase is used to establish and communicate the meaning of an opaque text. Typically, paraphrase can be considered as the reformulation of a statement in equivalent terms to express what seems to be obscure in a given text in order to render it more precise or clear or to give it more strength. Thus, paraphrase has the paradoxical characteristic of being the same, yet different, from the original (Fuchs 1982). The perspective of paraphrase taken here is both psychological and pragmatic. From a pragmatic point of view, the speaker and listener, as subjects in a given context, must be taken into consideration. From a psychological point of view, we would like to delve into those operations of the thought process whose existence depends on symbolic language, and in particular, natural language. Using Fuchs (1982) model of paraphrase, we have distinguished two different types of paraphrastic situations. The first is where the subject paraphrases the opa que text and thus confers meaning to obscure terms that are found therein; this
AN OPAQUE TEXT AS A MOBILIZER OF KNOWLEDGE
199
would correspond to what Fuchs refers to as paraphrasic recognition, which can be schematized as follows: X "means/signifies" Y. The other situation corresponds to what Fuchs calls paraphrasic reformulation, which can be schematized as: "what I mean by saying X, is Y" or "what S meant by X, is Y". Corresponding to these two situations are two processes by which the meaning of an opaque text will be estab lished. Paraphrasic Recognition. In this situation the subject is in search of a known word to be substituted for the unknown one. However, the data would seem to show that there are two different types of substitutions depending on the way in which subjects see their role as interpreters of the meaning the author intended for the text. On the one hand, a subject can consider his or her interpretation as the in terpretation of an unknown term (X = Y). In this case the subject seems quite con vinced of having correctly interpreted what the author intended. On the other hand, the subject may put a certain distance between him- or herself and the interpreta tion (I guess that X could be Y; X might be Y), as if the author's intention is out of reach. This difference in paraphrasing would thus correspond to two distinct types of relations. In the first case, it is a sign/sign relation of the type "X means Y". Ex amples of this are: - "VRILLER c'est creuser" (GYRE is to dig) - "MANSCANT, quelque chose qui se voit pas" (MANXOME, something that can't be seen) - "BURBULE c'est des sables mouvants qui commencent à faire des énormes bul les" (BURBLE is quicksand that is starting to make enormous bubbles) - "garde-toi ça veut dire méfie-toi" (beware means be careful) In the second case, we have a sign/object relation of the type, "I imagine X as a type of Y"; "X is like a Y". Examples of this are: - "Moi, j'imagine ça comme un espèce de venin" (I imagine it like a kind of venom) - "j'imagine un peu comme un vampire" (I imagine [it] a little like a vampire)
200
M. DOLITSKY AND G. PIERAUT-LE BONNIEC
- "moi, je pense de plus en plus c'est,comme un,comment je pourrais dire, vam pire, aigle" (I think more and more that it's like a, how can I say [it], vampire, eagle) - "c'est peut-être un peu un être humain" (perhaps it's a human being in a way) - "pour moi, le JASEROQUE c'est pas un animal, ce serait un chemin, une route" (for me, the JABBERWOCK is not an animal; it would be a path, a road) In other words, two types of operations must be distinguished here: operations of denotation where the subject working within the linguistic code seeks a synonym for the neologism, and operations of reference where the subject attempts to con struct objects of discourse. Paraphrasic Reformulation. In situations of this type subjects use paraphrases to explain or justify their interpretation of the poem or of parts of it. A child may pick up on what another child has just said and paraphrase that; more often, however, children paraphrase their own productions. It seems, in any case, that the three paraphrasic functions that have been found: explicative, summarizing, and expan sive, may appear either as children's paraphrases of their own productions or of the productions of others. 1. Explicative paraphrase increases the quantity of information that was first given. For example: "Moi j'imagine ça comme une espèce de venin, quelque chose qui quand il arrive et, il laisse rien derriere lui. Il arrive puis quand il prend quelqu'un et les griffes qui sortent et une gueule et puis qui mange quelqu'un" (I imagine it like a kind of venom, something that when it arrives and it leaves nothing behind it. It arrives and when it takes someone and the claws come out and a head and then who eats someone) Thus the subject speaks about how the JABBERWOCKY left nothing behind it, that it eats its victim; besides this, the subject speaks about its weapons, its claws and its head. Samuel says, "c'est peut-être quelqu'un qui creuse, plusieurs gens qui creusent" (perhaps it's someone digging, several people digging)
AN OPAQUE TEXT AS A MOBILIZER OF KNOWLEDGE
201
Charlotte continues with, "quelqu'un qui creuse une tombe, ou quelque chose comme ça" (someone digging a tomb, or something like that) 2. Summarizing paraphrase synthesizes what has gone before. For example: "un espèce, je ne sais pas, c'est comme une espèce d'esprit, d'ouragan qui ne se voit pas, qui vient dans la jungle, la nuit, il enlève tout sur son pas sage, et puis il fait ce qu'il veut de tout le monde. Vraiment, c'est un ton nerre, c'est horrible" (a sort of, I don't know, it's like a kind of spirit, a hurricane that you can't see, that comes through the jungle at night; it takes away everything in its way and it does what it wants to with everyone. Really, it's thunder; it's horrible) 3. Expansive paraphrase justifies the subject's interpretation or gives reasons for what happened according to the subject. For example: "non il le tue pas, parce que après, il s'enfuit; puisqu'il s'enfuit après, tu vois, il le tue pas" (No, he doesn't kill him, because then he runs away; since he runs away then, you see, he doesn't kill him) "Le JASEROQUE, il doit garder quelque chose, garder une grotte secrète avec quelque chose à l' intérieur, et alors je pense que le JASEROQUE, il doit pas attaquer les villages, puisqu'il reste toujours proteger cette grotte" (The JABBERWOCK, he has to guard something, guard a secret cave with something inside, and so I think the JABBERWOCK mustn't attack villages, since he always stays to protect the cave)
Discussion These results show how important paraphrasing is when approaching the mean ing of a text, especially when readers have the feeling that they can't quite put their finger on the meaning and pin it down.
202
M. DOLITSKY AND G. PIÉRAUT-LE BONNIEC
Paraphrase permits listeners to clarify, summarize or cut up what they hear, and if necessary, to reorganize it; it is thus, a dynamic process in establishing meaning. In any case, it is necessary to distinguish two paraphrasic processes. One would be a way of first getting an anchor in the linguistic code. The subject works at find ing a synonym by retaining a certain amount of homonymy (see Dolitsky 1984) and syntactic markers. This type of metalinguistic activity is a way of mastering the text that has been presented. Another process would be to construct or reconstruct referents in some way, which would be a type of conceptualization, thus mastering the meaning of the text.
Lexical Meaning Below is a list of the neologisms that occurred in the French text and as they ap peared on the blackboard. In front of each word is the number of times a given word was chosen to be defined: 17 JASEROQUE, JUBE 15 ENMIMÉS, FRUMIEUX, BURBULE, CORTULE 14 TÉTÉ 11 GOUGEBOSQUEUX, VORPAL, SIBLANT 10 TOVES, GALLOMPHANT 9
GYRENT, VRILLANT MANSCANT, RAYONNAIS, CALLAU! CALLAI!
8
LUBRICILLEUX, GUAVE, UFFUSÉ, FRABBEJEAIS
7
MOMERADES, HORSGRAVE, BAND-À-PREND
5
TULLEGEAIS
2
BRILGUE
The definitions were studied with respect to whether or not the child had taken the macro- and/or microcontext into account, and the effect of a neologism's phonetic form on its definition.
Macrocontextual level: macrostrategy Did the definition take into account the overall story and atmosphere of the poem?
AN OPAQUE TEXT AS A MOBILIZER OF KNOWLEDGE
203
Definitions that could be linked to a world of dangerous dragons or mysterious forests were considered to be macrocontextual positive (+). Examples of this type include: MANSCANT = méchant (mean), JUBE = un chemin en forêt (a path in the forest), il BRILGUE = il l' attaque (he attacks it), FRABBEJEAIS = frappé (hit). Those that could not be so linked were considered macrocontextual negative (-). Examples of this type are: GYRENT = bouillent (boil), TÉTÉ = télé (TV), UFFUSÉ = amusé (amused), BURBULE = botte de savon (soapbox).
Microcontextual level: microstrategy Does the child define the neologism is accordance with the syntagmatic role it plays in the sentence where it appears? A definition was considered microcontex tual positive (+) if it could replace the neologism in the chain where it appears in the poem; specifically, it had to be of the same part of speech. For example: - Le fauve MANSQUANT - méchant (the MANXOME beast - mean) - les TÔVES se GYRENT - bouillent (the TOVES did GYRE - boil) - Il pense, tout UFFUSÉ - amusé (in UFF1SH thought - amused) Following are examples of microcontextual negative (-) definitions: - l' oiseau JUBE - un chemin (the JUBJUB bird - a path) - l' arbre TÉTÉ - télé (theTOMTOMtree - TV) - il BURBULE - boite à savon (he BURBLED - soap box) - O jour FRABBEJEAIS - frappé (Oh, FRABJOUS day - hit)
204
M. DOLITSKY AND G. PIÉRAUT-LE BONNIEC
Morphemic level: phonetic strategy Did the child pay attention to the phonetic form of the neologism? A definition was considered to be phonetically positive (+) if it was subjectively judged as similar-sounding in respect to the neologism. For example: MANSQUANT / méchant (+), TÉTÉ / tele (+), FRABBEJEAIS / frappé (+) JUBE I un chemin(-), GYRENT / bouillent (-), BRILGUE I attaque(-) As the different levels are not exclusive but could be attended to simultaneously, there are eight possible combinations of (+)'s and (-)'s corresponding to eight dif ferent global strategies. These strategies, however, can be divided into four groups according to the number of levels that are taken into account by the child: 1. None of the aspects that have been defined above are taken into account, or at least the experimenter has found no reason to consider a given response as an instance of one of the strategies. In such a case, it will be labeled SO, to mark an absence of the set strategies. 2. The response shows focus on one of the 3 above-mentioned aspects, either macrocontextual (Sl), microcontextual (S2) or phonetic (S3). 3. The response shows that two aspects have been taken into account, either macrocontextual + microcontextual (Sl-2), macrocontextual + phonetic (Sl-3) or microcontextual + phonetic (S2-3). 4. The response shows that all three aspects have been taken into account (Sl2-3). As can be seen on Table 1, only 13% of the responses belong to the SO category; 198 out of 230 responses can be classed in one of the Sn categories. Two interpretive strategies stand out in our results: macrocontextual and phonetic. When only one aspect is attended to, it is either macrocontextual (30% of 1-aspect strategies) or phonetic (55%); a microcontextual strategy is least used (7%). Microcontext is attended to in only 7% of these cases, and when the data are taken on the whole, such a strategy, even when combined with another, is used the least.
AN OPAQUE TEXT AS A MOBILIZER OF KNOWLEDGE
205
Table 1
The Different Strategies used by the Subjects Number of As pects Accounted for
Strategies
0
SO
1
Sn
Macro
Micro
Morpho
N
-
-
-
N = 230 32
S1
31
+ + +
S3
3%
43
21%
55
27%
11
5%
+
+
22
11%
+
+
+
30
15%
N = 127
113
106
+
S1-3
+
S1-2-3
+
15%
6
+
S1-2 S2-3
3
13% |
N=198
S2
12
%
Attention to the microcontext is almost always in relation with attention to either the macrocontext or the phonetic form of the word (94% of the times this aspect is attended to in a 1- or 2-aspect strategy), as if taking into account the microcontextual level were, in a way, dependent on one of the other strategies, where the other two are dominant, but not particularly compatible. This can be seen by the fact that macrocontextual and phonetic strategies are taken into account together in only 11 cases (=5% overall and 12% of 2-aspect strategies).
Discussion The results show that there exist two main strategies of interpretation: semantic and phonetic. If they show up here together, it would seem to be due to the par ticular nature of the text that was used in that it could be treated like a poem and we would expect phonetic strategies to be used most often in texts where words
206
M. DOLITSKY AND G. PIERAUT-LE BONNIEC
can take on a context-free value, where language would be considered to be like music. Moreover, words carry an affective charge; it is true for everybody that some words are endowed with a particularly rich meaning due to life's experiences and words that are similar to these might also carry a similar affective charge. However a semantic strategy is probably the most usual and the most natural. It should be noted that already at 11 years of age, subjects have sufficient linguistic awareness that when assigning meanings to neologisms, they will in many cases at tempt to maintain the categories of speech these neologisms seem to belong to given their position in the sentence or morphological marking, no matter how strange the neologism might seem to be. But, most often, as can also be seen by these results, subjects are guided by the meaning of the poem as a whole when assigning meaning to the neologisms. The meaning of a text seems to be set up dialectic ally through the constant interaction between the macrocontext and the neologism. On the one hand the objects of the text are not fully defined at its onset, and significations are constantly being added to it, enriching the text throughout the reading (cf. Grize 1982). On the other hand, the syntagmatic environment of a given word can only give it a probable value; it is the context that confers meaning to it. This meaning is worked out as the reader goes along, 1) reading the text, and 2) interpreting it through paraphrase. This is why "until we understand the meaning, we do not know what has been said" (Wig gins 1971).
Conclusion In this part of the study Lewis Carroll's poem was used as a type of magnifying glass in order to try to bring out what kind of strategies are used to assign meaning to words in a text. Only poetic texts, of course, leave the reader so much freedom; in most texts, the words used impose a lot more constraints on the reader's inter pretation. Meaning-finding strategies are thus relatively sound mechanisms that can function even in unusual situations such as the one described here and in such a way that interpretation is never simply arbitrary. This study has shown that children are perfectly capable of assigning meaning to an opaque text by using a limited number of meaning-finding strategies that can be enumerated. The subjects' initial difficulty in finding meaning in the given opaque text was less due to a lack of available meaning-finding strategies than to their feel ings that the usual communicative "contract" had been broken, that, in a way, this text went against their "linguistic theory" where words had only one meaning, the
AN OPAQUE TEXT AS A MOBILIZER OF KNOWLEDGE
207
author's; the author wants to say something and he does so with unambiguous words. The comprehension of a text should not pose a problem; what we under stand is what the author meant. Such confidence in verbal communication is ob viously an illusion, even if it is true that language is often used to communicate information and that we do not speak to say nothing, even if at times we might seem to say very little. Such confidence in the good functioning of verbal com munication does present some problems in that subjects seldom participate, in a communicative situation with a metalinguistic frame of mind. They rarely question whether they have said what they meant to, or if the interlocutor has done so; and it is even more rare that they question their interpretation of what has been said. And this is what misunderstandings are made of!
NOTES 1. This study was carried out at the École Alsacienne in Paris, France. The authors would like to thank the principal of this school as well as Anne-Marie Béasse-Arnaud, the teacher of the class, for letting them carry out this research.
The Development and Role of Narrative Schema Storytelling E. Esperet Laboratoire de Psychologie du Langage Université de Poitiers, CNRS
For about ten years the number of psychological studies concerning language comprehension has increased, particularly in the field of discourse processing. This growing interest in the cognitive processes which intervene in the comprehension, memorization or recall of complex verbal material has not been accompanied by a similar attention to production processes. Several reasons may explain this. First, from a historical point of view, a great part of the studies related to lan guage phenomena beyond the word or utterance level, were carried out after numerous studies about semantic memory had been. Thus, they have followed the experimental paradigms conventionally used in this framework, i.e. presentation of verbal stimuli whose characteristics, as handled by the experimenter, are related to the subject's verbal responses. The hypothesis is that these responses have their source in the cognitive processing applied by the subject to stimuli. On the other hand, in production, only the characteristics of the situation can be connected to those of the collected utterances. However, these characteristics are not necessarily linguistic, which renders the inferences concerning the subject's cognitive processes more indirect. A second reason, apparently methodological, has theoretical implications. It is assumed, rightly or wrongly, that the characteristics of pre-built verbal material can be described and consequently controlled. Models and their objectives are available and provide a grid for the lexical, syntactic or semantic analyses of a stimulus. On the contrary, situations of social interlocution can be easily characterized psychologically. It is extremely difficult to define which aspects of a situation have a real influence on the discourse uttered, and to propose a plausible model of psychological processes which would connect these aspects in the speaker with some specific linguistic behavior.
210
E. ESPERET
The third reason is due to the social representation sometimes connected with the study of text production. For some researchers, these studies have their specificity and their interest, but a strict linguistic approach with a corpus descrip tion would be preferred. For others, if these studies are in the field of psychology, they remain at a descriptive level and the highly speculative interpretations that come out of them raise theoretical problems. Thus the study of the ontogenesis of discourse production has progressed less rapidly than the study of speech comprehension processes. Recently, however, re search devoted to the former has been carried out by a team in Poitiers working on the development of narrative behavior (Genèse de la conduite de récit), and the results will be reported here in this paper. First, we will deal with the conception of language behavior on which the studies rely. Then we will examine some data relative to the role of narrative schema in story production. Finally, we will tackle some of the difficulties and the direction we intend to take in further research.
Language Behavior – Storytelling Our starting hypothesis is that children first learn to handle more and more dis course situations, and gradually the functional requirements of these situations bring them either to build new means of linguistic expression, or to modify their use of the means they already possess. Consequently, rather than speaking of lan guage development, it might be better to speak of language behavior, which avoids the view of language as a formal and polyvalent tool independent of tasks and con tests. This study centers on the analysis of production processes at a cognitive and metacognitive level rather than on production itself. Obviously, this does not ex clude a study of syntactic, lexical or semantic acquisition, but this does imply put ting psycholinguistic processes back into the framework in which they take place. For example, we shall attempt to specify the functional characteristics of the narra tive production situation which might explain the appearance in French of the passé composé, as opposed to the imparfait (two past verbal tenses). This is in contrast to the view that mastering these verbal tenses explains the advent of some discourse categories such as the narrative. The ontogenesis of language thus corresponds to a gradual appropriation of a more or less extensive range of language behaviors in children; dialogue, argumen tation and narration are the most frequent ones in the first years. These behaviors are gradually organized into a system where they are coordinated. The description
DEVELOPMENT AND ROLE OF NARRATIVE SCHEMA STORYTELLING
211
of a child's language at a given age would be made up of the combined descriptive characteristics of the whole system, which would include the lexicon, syntactic structures, etc. and each characteristic would first be interpreted in the framework of the behavior in which it appeared.
The Notion of Language Behavior Language behavior (Esperet 1982; Esperet and Gaonac'h 1981, 1986) will be considered as the product of a hierarchical sequence of cognitive processes leading to the production of discourse in a situation. From a functional and sequential point of view, we hypothesize that there exist two levels in the organization of the processes involved. - The first level includes mechanisms which assure monitoring the behavior as a whole. These would first be activated by the interlocutionary situation. They would call upon specific representations of the behavior concerned: repre sentations of structural characteristics of the discourse category under con sideration, socio-linguistic rules defining appropriate contents, linguistic forms, etc. In relation to these, there would occur operations of selection and content organization, which would be taken from the speaker's semantic memory. This level would also monitor, more or less completely depending on the production difficulties encountered, planning the psycholinguistic operations which assure the construction of discourse from selected and structured contents. - The operations of implementation and relatively punctual semantic choices take place at a second level where the processes of the gradual construction of dis course come together, as do inter- and intrasentence operations (Noizet 1980). Also at this level "textuality" (referentiation and coherence) are constructed and the paralinguistic and nonverbal aspects which contribute to the "setting" of discourse are governed. This schema requires three explanations: - This conception of discourse production considers that the second level is functionally dependent on the first, which implies the existence of a metacognitive process. But while it is still difficult to define precisely the nature and sequencing of these processes (in terms of procedure, according to Brown (1975) or in terms of knowledge, according to Flavell (1981)), it seems never theless necessary to assume their existence in discourse production. From this point of view pragmatic and semantic aspects of discourse are given priority over syntactic functions which would be dependent on them.
212
E. ESPERET
- For reasons of clarity, this paper separates the two levels artificially. Obvious ly, in reality, discourse is constructed through a permanent back and forth movement from one level to the other. We do not assume, for example, that in a situation eliciting argumentative behavior, selection-combination operations would define immediately and precisely, the whole discursive content, since it would be "worded" afterwards. The activated representation of this behavior would rather define general semantic frames and rules which gradually monitor a content. This first content would be "translated" into linguistic fonn before selecting of the following content, and could modulate this selection in return. Thus a difficulty in linguistic "translation" could indeed bring the speaker to choose another expression, and thereby change the orientation of the discourse within limits consistent with the representation that is activated. - Finally, the process described at levels 1 and 2 must each have an ontogenesis which could account for the ontogenesis of the whole behavior. Therefore, it is necessary to study the stages of construction of level 1 representations, or psycholinguistic level 2 operations, and show how the stage of development of these two levels can explain the discourse produced at a given age. In this way, it should be possible to bring out the main, constant features, as well as the genesis of the invariants, for the types of language behavior already mentioned. We have chosen to analyze some aspects of one of them, the narrative.
Narrative Behavior To be more precise, this research centers on narrative behavior in a school situa tion. Thus it refers to a limited category of narrative productions whose charac teristics are not automatically transposable to other types of narratives, e.g. conversational narratives. However, it would seem that narrative behavior of all types share the same invariants. Role of the Narrative Schema. Work on the school narrative was chosen because of its institutional importance. It is the subject of many exercises in school and is used to assess pupils, even if its educational principles often remain inadequate (cf. Bain, Bronckart, Davaud et al.'s analyses 1982; Feagans 1982). On the other hand, in the last ten years, psychologists working on discourse processing have been par ticularly interested in the narrative. They have been able to bring to light several apparently stable procedures, mainly in comprehension, recall or recognition. Studies of production have been carried out, mainly in their symmetrical relation to
DEVELOPMENT AND ROLE OF NARRATIVE SCHEMA STORYTELLING
213
comprehension (cf. Freedle and Duran 1979; Kintsch and Van Dijk 1978; Van Dijk 1980; Van Dijk and Kintsch 1983). It would seem that knowledge acquired by re searchers through studies of production processes could provide some elements for a model of narrative production. The numerous studies, whose analyses can be found in Denhière (1981-1982, 1984), Fayol (1985b) or Mandler (1982b) all em phasize the role semantic structures, which are common to the members of the same culture play in the processing of narrative discourse. However, while some research is based on a theory of narrative grammar (Mandler and Johnson 1977; Mandler 1982; Stein and Glenn 1979; Thorndyke 1977; etc.), where the existence of a narrative schema may be considered as the result of the internalization of cer tain elements described by these narrative grammars (Mandler 1982b), other theories hold that culturally acquired superstructures play a role of guidance and monitoring in the construction of semantic representations of discourse (Kintsch and Van Dijk 1978; Van Dijk 1980). What seems relevant in these two approaches is that they both hypothesize that a speaker activates a representation of a narrative and thus "knows" what it is. The narrative schema would thus guide the construction of discourse while respecting certain constraints, i.e. something must be judged interesting, or something unusual must happen; the framework, the characters and the situation must be specified; an outcome, consequences or a final assessment are expected. Hypothesizing such a narrative schema seems necessary to understand narrative production, even if this notion has been submitted to justifiable criticism as to its role in memory processes in children (Thorndyke and Yekovitch 1980; Denhière 1984). This monitoring, situated at the first level of our model, obviously does not ac count for all discourse operations. Speakers' psycholinguistics abilities and the or ganized whole of their knowledge about the world also intervene. Thus, if the speaker situates the beginning of a story in Africa, the actors will probably be lions, giraffes or elephants rather than cats or cows. Furthermore, and this is where the schema really plays a role, the scene will probably present a threatening lion, rather than one that is taking a nap after a meal. The intervention of the schema accounts for feedback, and modifications are fre quently observed in oral narratives, e.g. supplementary information concerning the situation, considered necessary by the speaker in order for the interlocutor to com prehend the event being told about, changes in the verb tense in order to maintain the temporal coherence of the narrative, etc. Elements for a Model of Narrative Behavior. Figure 1 summarizes the main stages postulated for the construction of a narrative by a speaker who has the re-
214
E. ESPERET
Figure 1
Sequential schema of narrative beihavior (inspired by Esperet and Fayol 1982).
DEVELOPMENT AND ROLE OF NARRATIVE SCHEMA STORYTELLING
215
quired ability. The interlocution situation activates a representation of narrative be havior, whose major component is the narrative schema. This schema will play an essential role in the different operations in selecting/combining the content ele ments that will be actualized as constituents. The latter includes the notion of a Narrative Event, which corresponds to the core of the story (e.g., "a wolf catches a little boy") and which is thought to be acquired before the other constituents. These elements give the typical narrative macrostructure, i.e. Framing or thematic type ("It is the story of..."), Initial Situation ("one day, ..."), narrative Event ("then, the wolf caught the little boy"), Attempt(s) ("his father took a gun"), Result ("he killed the wolf') and possible, Finality ("he never went back there"). This schema which only offers empty "spaces" semantically will gradually allow for the selection and organization of a content that corresponds to the different components, and will, at the same time help maintain coherence with respect to the event ("narrative dilata tion" occurs through macrostructures similar to the ones described by Kintsch and Van Dijk 1978). Each macrostructure thus obtained will or will not be the object of further development (micro-operations). It should be remembered that our model does not imply that all of the schema's components are semantically specified before the beginning of the other operations. The hypothesis is that, once the schema is activated, the first component (framing) will guide the search for a content that is consistent with the target event, and that this content will possibly be subjected to operations of narrative expansion (a story can be presented simply by a title or by a long discourse), and will finally be for mulated. Then, a similar sequence, involving retroactions, will begin for the second component. At each stage, monitoring by the narrative schema is obviously not suf ficient to entirely determine the selected semantic content. The speaker's repre sentations of the world also play a part, e.g. the scripts described by Schank and Abelson (1977) which constitute the conventional structure of a sequence of actions (Going to a restaurant: entering choosing a table, asking for the menu, e t c . ) . The speaker's lexical knowledge also intervenes (cf. Ehrlich and Florin 1981) as do socio-cultural representations, e.g.: What content is admissible or desired according to the speaker and what frame? What type of language must be used? etc. Another type of determination operates on macrostructural and microsutructural operations. This determination results from many retroactions executed at all levels, in par ticular in oral production. A lexical obstacle, for example, will entail another choice at a target action level, and this modification in turn will entail corrections or addi tions at the already produced beginning of the narrative. Finally the speaker's reac tions will influence the way discourse proceeds. However, this last point will not be studied here; the school narrative is not a communicative situation, per se.
216
E. ESPERET
The Construction and the Role of Narrative Schema In order to set up and validate our model, we have carried out research focused on the following three questions: 1. Do children construct a narrative schema, and if they do, in what form'? 2. Does the genetic evolution observable in narratives correspond to the instoration of a schema organizing these narratives? 3. Is there, for any given child, a relation between the type of cognitive schema s/he possesses and the type of narrative produced? This third question is the most delicate. The results which will be presented come mainly from a longitudinal study car ried out on 34 children who were followed for 4 years (from the age of41/2Vi to 7 Vi years old). Some cross-sectional studies will also be briefly referred to in order to shed light on the evolution of a particular aspect between 5 and 11 years.
Study of Narrative Schema Construction Considering the age of the children, it was difficult to use a definition task or a choice of criteria corresponding to a narrative notion or not. So a judgment task was prepared which would allow the narrative schema to emerge, if it existed. The hypothesis was that schema construction develops in three main stages be tween 4 and 11 years. First, a coherence condition which differentiates a set of un connected utterances from organized discourse would be set up. This first component, which is not always easily perceived in narrative content, remains, however, nonspecific of the narrative type. It is, in a way, a minimal condition which reveals children's metalinguistic awareness of a textual reality. Second, the criterion event appears. In order for a discourse to be judged "a story", it is neces sary to have a semantic inflexion, i.e. a complication, to set it apart from a simple temporal sequence of actions or a description (this corresponds to the notion of "opening" used by Fayol 1981, or to that of the first unit of the narrative dyad "complication, resolution", in Botvin and Sutton-Smith 1977). Finally, the sup plementary criterion of a reaction to the event would appear ("closing" in Fayol; second part of the dyad in Botvin et al.). Then the notion of minimal episode is set up, i.e. initial situation —» complication → resolution, the structure of which will enriched to include Spatio-temporal framing, Attempt, Evaluation, etc. Cross-Sectional Research on the Notion of Story. This study was carried out on 60 children who were divided into 4 age groups: 51/2 years, 71/2 years, 91/2 years, 111/2 years (N = 15 per group). The material and the procedure used were the fol-
DEVELOPMENT AND ROLE OF NARRATIVE SCHEMA STORYTELLING
217
lowing: four types of verbal sequences with roughly the same length and the same syntactic structure were constructed: 1. Sequences called nontexts (NT). Each of these consisted of a sequence of 5 statements, whose lexicon was taken from dif ferent semantic fields, so that there would be no connection between them, and in which there were neither anaphorical pronouns nor indefinite-definite relations for determinants. 2. Sequences of the same structure which can be called texts (T), as they possess all the indices of coherence, but only make up a typical script. 3. Se quences called "openings" (O) in which an event occurs without specifying either the character's reactions or the consequences. 4. Stories (S) which are made up of the minimal episode already mentioned. The different sequences were presented randomly, and the subjects were asked if the discourse offered was a story or not and to justify their responses. For each child, a score was calculated per type of sequence (NT, T, O, S) according to the following principles: a "yes" response to a sequence was scored as +1; a "no response, -1; "I don't know" was given the mark 0. As each type had 4 sequences, the scores could range from -4 to +4. The results obtained support the hypothesis of a three-stage evolution in the construction of the narrative schema. It was found that: - Firstly, at 5 to 6 years of age, there is no differentiation of the 3 types O, T and S, but a beginning of a rejection of the NT type. Nondifferentiation be tween O, T ans S persists up to the age of 71/2 years but the nontexts are well distinguished from the other sequences and are only rarely accepted. Secondly, the S sequences are distinguished from the T and sequences which start to be rejected at only 91/2 years and 111/2 years. - The presence of surface marks, which usually accompany the dyad "complica tion-resolution", of the type "SUDDENLY", and "AT LAST", "LUCKILY", that were inserted voluntarily in half of the sequences of each type, slows down the rejection of NT, T or productions, but only for the two youngest groups. Then, it is the semantic criterion, i.e. the presence or absence of the dyad, that becomes crucial. - The analysis of the arguments given by the children supports the interpretation. Up to 71/2 years old, the only categorical justifications (except "Because it's nice", "because it's true", etc.) are references to text coherence. At 91/2 years some arguments of the structural type ("there's a beginning") appear; these be come dominant and better worded at 111/2 years ("It's quiet and then, some thing happens and then, it becomes quiet again," for example). An evolution really does then occur in the representation children have of a story, an evolution which tends towards the typical schema described by many
218
E. ESPERET
Figure 2
Story
Opening
Text
Nontext 11-12
Mean scores of narrative judgments according to age group (N = 15 per group) and sequence type being judged. The scores -4 and +4, respectively, correspond to refusal or acceptance of the 4 sequences of a given type.
authors. Obviously, this is not at all "natural" and reflects the various cultural in fluences received by the child mainly from school. In other respects, this evolution appears slower than would be imagined, at least as far as setting up metacognitive schema is concerned. Finally, the group data conceal that within a strong genetic tendency there exist interindividual variations probably implying differences be tween children in their contact with narrative discourse and with judgment tasks. Thus a study carried out by Ballaire (1982) shows that some 7- to 8-year-old children possess a representation much closer to the typical schema than other 10to 11-year-old children (cf. also Denhière 1982). Longitudinal Research. This allowed us to verify the reliability of a portion of the results obtained in the previous research, through a somewhat different ex perimental situation. 34 children were individually questioned once a year for four
DEVELOPMENT AND ROLE OF NARRATIVE SCHEMA STORYTELLING
219
years between the ages of 41/2 to 71/2 years. They came from a sample which at first, numbered about 70. During the first year (4Vi years), only the sequences NT and S were presented orally. Then at 51/2, 6V2 and 71/2 years, for the 4 types of se quences (cf. Cross-Sectional Research on the Notion of Story, above), comic strips, which were semantically equivalent to the sequences (1 picture per utterance) were presented. It was thought there was no need to present the 4 types as differentiated (NT, T, and S) to 41/2-year-old children, and it was not sure that they could read a comic strip. The different sequences (oral or comic strips) were presented in ran dom order, and the children were asked if the pictures or the discourse made up a story; the subjects justified their responses afterwards. Few of them used the responses "I don't know", or "a little", which were possible. Table 1 presents the percentages of "yes, it makes a story" responses, for the dif ferent situations. It can be seen that the only clear evolution, between the ages of 41/2 and 7Vi, concerns the sequences NT (Nontexts) which are rejected more and more as stories. The S sequences are virtually accepted from the age of 6 years onward, but the two other categories (T and O) are not clearly differentiated. The analysis of the argu ments given by the children fully confirm this assumption. For the youngest, every thing is a story so long as there is a character, (even in only one of 5 utterances). Then, the category NT is refused with the justifications "it does not go together", "it does not follow", "it isn't about the same thing", etc. When the child gives a positive answer to the three other categories, it is justified by building up a "narra tive" (even if it is a script), e.g. "it's a story, because the man is going for a walk, and then...". This result confirms and specifies the result obtained in the cross-sec-
Table 1 Sequence type
Verbal form
Comic-strip form
41/2 years
5 Vi years
61/2 years
Nontext
74%
71/2 years
35%
21%
0%
Text
63%
88%
88%
Opening
78%
86%
89%
69%
93%
96%
Story
76%
Percentage of yes responses to the narration judgement according to the type of sequences presented. (Number of subjects followed longitudinally f or 4 years: 34.)
220
E. ESPERET
tional research. Only the necessity of semantic coherence is gradually perceived be tween the ages of 41/2 and 71/2,which corresponds to the first stage of schema con struction, the stage we have hypothesized. Therefore, we can conclude that a narrative schema, even if it refers to a reality that is supposed to be familiar to the children, does not involve a metacognitive task, like a judgment. This does not imply that a causal schema may not intervene in memory and reconstruction tasks (cf. Born 1975, 1986).
The Evolution of Produced Narratives If it is true that narrative production responds to socially admitted standards and implies the use of a narrative schema, then the structure of the stories built up by children must, with age, become closer to this schema. An evolution of production must correspond to the gradual construction observed in the previous paragraphs. So, the children were asked to tell a story in various situations. The analysis of this type of material raises a number of problems, e.g. there exists no text to serve as a reference; semantic ambiguities linked to oral communication are numerous in young children's, productions, etc. The few studies already carried out cannot be exploited to solve this problem. Either they deal with another aspect of the narra tive, or the way the narratives were obtained conceals the real performance (Botvin and Sutton-Smith 1977), i.e. the experimenters took the narratives "in hand", which partially explains their textual quality. The corpus used by Applebee (1978) was obtained through numerous interventions by the interviewer: "and now, what is he doing?", and "what happens after that?" etc.). We have therefore been forced to make some arbitrary decisions and to limit our ambitions. The Analysis of Narratives by Children between 41/2 and 71/2 Years Old. Once a year, the 34 children of the previous study (cf. Longitudinal Research) were also asked to tell a story, "to think of it inside their head", "in order for it to be interest ing", and to begin whenever they wanted to. If the instructions didn't result in a story, they were asked for a story they might have seen, or heard or "read". If a poem or a song was given, they were again asked for a story. The story was taperecorded and then fully transcribed. This task, which was carried out in a very school-like situation, without any ex plicit motivation for the child, seemed difficult for the youngest children. The examination of Table 2 shows that for half of the children, at 41/2 years, narrative production, when asked for, is apparently impossible (47% of nonrespon-
DEVELOPMENT AND ROLE OF NARRATIVE SCHEMA STORYTELLING
221
Table 2
Children's age Responses
41/2 years
51/2 years • 61/2 years
71/2years
% of nonresponses (including song, sto ry, ...)
47%
32%
29%
3%
% of narratives inspired by books, TV, films
41%
41%
26%
29%
% of real-life narratives
9%
9%
9%
9%
% of imaginary narratives
3%
18%
36%
59%
Percentage of nonresponses and types of narratives obtained. (Number of subjects followed longi tudinally for 4 years: 34.)
ses). This is not a classical problem of blocking. Most of the children, during the same session, performed other tasks with material support, i.e. they spoke, they ex plained, etc. Some even made their nonresponse explicit, "I cannot tell stories", "I am too average", "my older sister tells them". For the other half of the children, most of the stories that were told were inspired by narratives they had already heard (41% of the total answers + the nonanswers, i.e. 78% of the narratives produced). Finally, developmental evolution shows a progression towards a more assured mastery of the task. The nonanswers finally disappear; they decrease from 47% to 3%; the imagined narratives gradually win out over the other two types of production (from 3% to 59%). This is probably because the narrative schema, or at least some of its elements, begin to control the observed production. To check the evolution of the narrative structure, all the productions were analyses with a grid to code each narrative utterance. This grid, which is based on narrative grammars, in cluded about ten categories, e.g. Action, Attempt, Event, Outcome, Obstacle, etc. The presence or absence of some marks of a setting in narrative discourse were also coded: T: title given, I: theme introducer ("once upon a time there was a little girl who...", "this is the story of a ... who ...", "it tells about a ... who ..."), F: spatio-temporal framing ("one morning, in the woods...", "one day, as she was going to the market...") and E: end mark ("here it is", "that's finished", "that's all"...). This code brought out patterns in production, where six typical narrative structures stand out. These correspond to hierarchical levels: at level 0, there is no discourse per se, but a series of nominal groups, or unconnected parts of sentences. Level 1 is made up of temporal sequences of actions: "we went home, we ate, then,
222
E. ESPERET
we went to bed", etc. At level 2, the wording of an event is eventually accom panied by a wording of actions, but the episode remains very incomplete (no at tempts, reactions or outcomes) or there is a worded sequence of actions with a marked causality, without an event. Level 3 is made up of "opening/closing" worded sequences which already foreshadow the full narrative; for example, a cer tain character's need creates an action, which may have consequences. This will result in the disappearance of the need at level 4. Here the unitary narrative episode appears (initial situation, event, reaction or attempt, outcome). This sometimes takes the form of an "opening/closing" sequence which involves complications; an obstacle must be removed, or cleared, in order to reach the end. Finally, level 5 is characterized by the presence of several interconnected, or concatenated, narrative episodes; it represents the most complete form of the narrative. Only the productions of levels 4 and 5 correspond to full implementation of the narrative schema. Table 3 presents the proportions of each of the six levels ob served in the productions. It can be seen that, with age, levels 0 and 1 productions clearly decrease while generally level 3 and 4 productions increase. Level 5 narratives were almost all (5 out of 6) inspired by tales the children knew very well, and there should be no il lusion about the children's abilities. Overall improvement occurs over a greater proportion of imaginary narratives with age (see Table 3). Consequently, the per centage of imaginary narratives increases with the structural level of the narratives, from level 0 to level 4: 26%; 37%; 65% and 60%. The real-life narratives are with one exception, of types 1 or 2 (temporal sequence or script according to Schank). Therefore children tend to produce more and more discourse structures similar to the schema, Initial situation - Complication - Attempt - Resolution, and this hap pens without a model that was either heard or read. This evolution is accompanied by a more frequent use of narrative discourse markers. Table 4 shows that the higher the structural level, the more markers there are in the produced narrative; these are mainly introducers and the spatio-temporal frames. Furthermore, no narrative from level 4 involves direct presentation of the event or complication without a prior setting. Moreover, this type of narrative always involves an end mark, whether it is the presence of an ending (E) in discourse, or an explicit intonation. Thus, the progres sive construction of the schema does not only appear in the discourse structure, but it also implies a certain type of wording (second level of the model). The overall tendency appears to be clear and consistent with our hypothesis. However it must be said that less than a third of the subjects were able to build up a full narrative at the age of 71/2. Thus development was far from finished.
DEVELOPMENT AND ROLE OF NARRATIVE SCHEMA STORYTELLING
223
Table 3
Narrative level
Children's age 411 years
51/2 years
61/2 years
71/2 years
0
17%
5%
0%
3%
1
45%
39%
21%
12%
2
28%
13%
25%
27%
3
5%
22%
14%
27%
4
5%
9%
27%
27%
5
0%
13%
8%
3%
Percentage of narratives of each level produced by children, as a function of age. (Number of subjects followed longitudinally for 4 years: 34.)
Cross-Sectional Research on Produced Narratives. The 60 children of the crosssectional research already mentioned also produced narratives. They were asked to tell a story and the beginning was provided for them; "The clock struck 8 o'clock; all was quiet ..." This beginning, unfortunately, brought about a high proportion of scripts (typical actions following one another according to a time sequence and fol lowing a familiar script, e.g. the departure to school). There seems to be a strong effect of the instructions on these production tasks which could also be seen in other respects. The evolution observed here, between 51/2and 111/2 years, remains however convergent. The percentage of narratives cor responding to levels 4 and 5 are for each age group respectively: 11%, 5%, 46% and 71%; the percentages of levels 2 and 3 narratives range from 90% to 24% be tween 51/2 and111/2years of age. Mastery of the narrative schema for production is generally considered to be ac quired at about the age of 11. Our results show that this type of exercise still causes difficulties for of these children. Metacognitive Schema and the Structure of Produced Narratives. If a narrative schema is considered to develop with metacognitive tasks, and if produced narra tives develop parallel to constructions of this same schema, there is still not suffi cient proof of a causal relation between the two phenomena. At best, this parallel provides a favorable assumption. But the results indicate a temporal discrepancy between the two types of development. Children can produce a full narrative some-
224
E. ESPERET
Table 4 % narration marks Narrative level
I
F
Theme introducer
Spatio-temporal framing
End
1
11%
11%
59%
2
29%
21%
46%
3
63%
25%
63%
4
80%
25%
55%
5
100%
66%
50%
Percentage of narrarivesof each level which present narative marks I, F, E.
what earlier than they can refuse the quality of being a narrative to a sequence without a narrative schema. On the contrary, those who produce a full narrative recognize this category when the corresponding sequence is presented to them. Therefore, it can be supposed from these tasks and situations that a narrative schema (or these elements that have already been developed in the child) can be more or less easily brought into play. It was tempting to connect metacognitive schema and schema produced in the narrative (correlational approach), but the correlation, significant from a statistical point of view, did not seem to reach a value which really supported our hypothesis (0.51). Furthermore, the longitudinal research is still too "short". Over the first 4 years, the patterns of children's answers remain very similar. The absence of suffi cient differentiation makes it difficult to find a relation between metacognitive schema and produced narratives. Therefore, there is still no direct data concerning such a relation at present.
Problems and Directions of Research Our approach is based on the idea that discourse, in this case narratives, is not produced utterance after utterance, but requires planning and monitoring by meta cognitive processes of all the operations involved. The fundamental implication of such control is the mastery of the narrative schema by the speaker. This hypothesis
DEVELOPMENT AND ROLE OF NARRATIVE SCHEMA STORYTELLING
225
has grown out of the studies carried out on the comprehension and memorization of narratives. It is in agreement with Clark and Hecht's analysis (1983), according to which speakers might use representations they have built into their memory during comprehension as criteria to guide their productions. Thus, we have in fact worked in a direction others have taken, even if the term "metacognition" is not always used (Chafe 1980; de Beaugrande 1980; Quasthoff and Nikolaus 1982). While at this general level, the different conceptions of narrative production that have been put forward are quite convergent, any attempt to clarify the mechanisms that are implied in the process raises many problems.
Narrative Schema, Metacognition and Cognitive Monitoring The first question concerns the content, and particularly the status of narrative schema. The narrative schema has been presented as an organized representation of the canonical structure of a narrative; whether it is made up of elements of narra tive grammar (Mandler type), rather than narrative categories forming a superstruc ture (Kintsch and Van Dijk type) is secondary here. On the other hand, the idea that there exists a functional and genetic hierarchy of schema components seems important. A child rarely utters, for instance, attempts or resolutions without eventcomplication being built up beforehand. Moreover, the assumption that the core of a story is selected first, explains the youngest children's answers, which only pro vide "a news announcement" (Sacks 1972), such as "I have fallen" without the "narrative dilatation". Secondly, is a narrative schema a relatively static representation, activated by a metacognitive process, which is used in checking content units coming from semantic memory? or is it, itself, an operative structure which does this checking? The terms used by different authors are not very precise on this point. They say that the schema intervenes, or plays an important role, in the control of operations. At present, no analysis seems to clearly answer this question. Some authors (e.g. Mandler 1982b) even think that the functional relationships between the schema and other basic operations, in production of judgment, are so complex that it is still too delicate to approach them. This question is linked to one concerning the degree of metacognition implied in the guiding processes. This term covers many assumptions that are not always very clear. Fisher and Mandler (1982b) have recently brought out the theoretical "fuzziness" and the difficulty of empirical validation linked to terms such as procedures, strategies or metacognitive competence in Flavell and Brown's models.
226
E. ESPERET
In our research, the fundamental intervention of a narrative schema is similar to cognitive monitoring such as Flavell conceives it (1981). The narrative schema would stem from a skill which would not necessarily imply metacognitive function ing. Metacognitive functioning would only be required in difficult situations, error correction and tasks of narrative judgment (for more recent studies, cf. Esperet 1987; Esperet and Piolat 1988). This would mean a child possesses the schema at the level of production monitoring without necessarily possessing metacognitive understanding. This would explain the observed discrepancies between judgments and productions. It appears that the question of "monitor" identity cannot be posed very precisely yet.
Narrative Schema and Surface Marks One should not forget that the influence of the schema is not limited to setting up a semantic macrostructure, but that it may even intervene in psycholinguistic operations at the second level of the model. If the listener correctly recognizes a narrative in a given discourse, it is not only because s/he perceives the initial situa tion, the complication, etc., but also because s/he can find some specific markers in the text: the oppositive function of verb tenses, some expressions of coherence through lexical and anaphorical markers of cohesion, the spatio-temporal framing mentioned above, etc. To design a production model, it is therefore necessary to build up a theory of textual operations which are not only those used in the construction of autonomous utterances. But as Fayol has pointed out (1981), psychological models of narrative production and comprehension often leave out this phase (the phase of translation into writing, according to Hayes and Flower 1980). Thus there is important work to be done here. This implies that it will be possible to describe the mechanisms that match content organization which include pragmatic components, to performance linguistic. The work done by Bronckart's team (Bronckart 1985; Bronckart and Schneuwly 1981, 1984) based on enunciation theory (discourse and argumentation) studies the surface markers specific to some textual categories (theoretical discourse, storytell ing), which are the result of underlying textual operations that are gradually mastered by the child. In our opinion, there remains the problem of combining the two approaches: one which tends to focus on semantic structuration (this being our approach up to now) and the second one which focuses more on the expression of
DEVELOPMENT AND ROLE OF NARRATIVE SCHEMA STORYTELLING
227
the characteristics of the production situation at the textual level. Work in this second direction (Gaonac'h and Esperet 1985) has already been undertaken. These are not the only problems; the study of production or more exactly of production processes in young children requires, for instance, methodological in ventiveness. The frequent use in studies on metacognition of speakers' accounts of their own processes is very difficult to use with young children. But these problems must not prevent us from considering this field of research as fundamental, for casting it aside might bias the conception psychology has of language construction.
Stories A Psycholinguistic and Ontogenetic Approach to the Acquisition of Narrative Abilities M. Fayol Laboratoire de Psychologie Université de Dijon
After the explosion of psycholinguistics studies concerning sentences, interest shifted little by little at the end of the seventies toward research into the memoriza tion and comprehension of texts and, to a lesser degree, into text production. While the term "text" covers a great variety of forms of discourse, it is the story that inter ests us most here. This is both because of the facility with which it can be isolated, and because of its observable regularities. Thus, it is easy to parse a story and to set up a schema. Most of the studies on stories are attempts to prove the "psychological validity" (Mandler 1982a) of subjacent organization that is supposed to serve as a guide to memorization, comprehension and production of stories. It seems to be well estab lished today that stories are organized in episodes composed of a small number of narrative categories (setting, initiating event, goal, attempts, consequence...) which are presented one after another according to relatively strict rules and which effect rememorization and on-line processing time (Cirilo and Foss 1980; Haberlandt, Berian and Sandson 1980; Kintsch and Van Dijk 1978; Mandler and Goodman 1982; Mandler and Johnson 1977; Rumelhart 1977; Thorndyke 1977). However, it would seem that given the schemas thus set up, this sort of prototypic narrative organization, would be rooted in the prelinguistic cognitive representations of actions and chains of actions. But since the work of these authors was carried out, the near total of all the studies done on the subject of stories, in contrast with all expectations, ignores psycholinguistics (or, if one prefers, the psychology of language). They almost never take into account how the text is set up or what relationship this might have with subjacent categories. And thus re search into story production is all but inexistent.
230
M. FAYOL
But as soon as such a perspective is taken up, the data offered by a theory of schema are found to be insufficient, even if they do, for the time being, provide a conceptual framework. In the pages that follow, we are going to report on the most important findings that have come out of a study on the acquisition of the ability to write stories about one's own personal experience, and attempt to define the most important questions that must be addressed on this subject.
A Psycholinguistic Approach to Stories According to Labov (1978), a story about personal experience can be defined as a way to recapitulate past experience in such a way as to set up a correspondence between a (supposedly) real series of events and an identical series of verbal propositions. This implies that: a. the author has selected an event worthy of being related - the focus of infor mation - that, in a dialogical type of exchange, can be put into the form of a "news announcement", e.g. "I fell yesterday" (Fillol and Mouchon 1977). b. this event will be expanded upon (Bronckart 1977; Fayol 1981), analyzed into constituent actions and states that are dependent on the subject's world knowledge and evaluated from the subject's point of view. This fragmentation might be seen as making up a type of framework that would correspond to macropropositions (Kintsch and Van Dijk, ibid.) or to categories of a conven tional narrative schema (Esperet and Fayol 1982). the narrator will set up a series of micropropositions that will end up becom ing a text after a certain number of transformations will have been applied to them in such a way that the story will not simply be made up of a juxtaposi tion of independent utterances. It is at this level, and only at this level, that psycholinguistic problems come into the picture. Thus, the core of the problem resides in the passage from a cognitive-semantic organization, which is most likely nonsequential (Ehrlich 1982, 1984), to one that is linear, linearity being, of course, inherent in speech which unfolds along the dimension of time. The narrative schema can be considered as an intermediary step (Esperet 1984). Because setting up a text means concatenating propositions where each refers to an action or a state, the narrator is forced to carry out a certain number of opera tions. As Bronckart, Gennari and De Weck (1981) have noted, utterances made up according to a certain number of autonomous rules regulating how they function, get organized into texts which in tum are regulated by the context. The importance
STORIES
231
this has for stories is that, we will find, on the one hand, a certain number of lin guistic phenomena functioning intratextually, and, on the other hand, elements that serve as traces that anchor the text into the context (time, place, addressee ...) where the story is being told. And while the data from psycholinguistic studies of the sentence are compatible with the former, this is not so for the latter. In fact, the manner in which utterances must succeed one another in a story for ces the narrator to use index systems that help the reader-listener find the objects of the predication(s). These systems concur topical unity to the text and assure its cohesion; they appear in the surface structure through systematically used marks. This applies in the way the topic is maintained as the syntactic subject (KarmiloffSmith 1981). It also applies to the way in which determiners and pronouns are used, ways that cannot be accounted for by findings from work on isolated senten ces or sentence pairs . However, topical centration and cohesion, although indispensable to insure that the sequences of propositions combine to form a whole, are not sufficient in them selves. They must be related to new information. This will culminate into "the point of the story" (Schank 1975) which can be considered equivalent to Labov and Waletsky's (1975) complexity and outcome. What this means is that, as has been pointed out by Applebee (1978) and Slakta (1975), the narrator must continually negotiate a compromise between cohesion and progression. As a general rule, it can be considered that cohesion is maintained mostly through reference to what has gone before (Halliday and Hasan 1979). The same is true for maintaining a topic. The author must in some way set up from the begin ning an "origin" for the story (Fayol 1981). This can be considered more or less similar to the setting, which gives the topic and the point of view from which the events will be seen (Black, Turner and Bower 1979) as well as the time and place. Later actions and states will then be plotted on the sort of coordinate system (Culioli 1976) set up by this origin. This dependence will show itself on the surface by a series of linguistic marks of the type: - "UN homme entra ...IL portait un chapeau ..." CET homme (A man entered ... HE was wearing a hat ... THIS man) - "UN JOUR je devais aller ...LE matin ... DEUX HEURES PLUS TARD ... TROIS JOUR APRÈS ..." (ONE DAY I had to go ... IN THE morning ... TWO HOURS LATER ... THREE DAYS LATER...)
232
M. FAYOL
Progress toward the climax of the story is signposted in a number of ways. The chronological succession of propositions referring back to earlier events is one of them, but not the only one unless the story is about ordinary events that can be "scripted" (Schank and Abelson 1977). If the story includes an unanticipated ele ment whose consequences are unforeseen, certain grammatical connectives will ap pear ("suddenly", "all of a sudden") to signal the rupture from the expected continuum of events (Bronckart and Schneuwly 1984). There are, associated with this, changes in verb forms (Schiffrin 1981; Weinrich 1973) which put the setting in contraposition with the focus of information as we have shown elsewhere (Fayol 1985a). In this case, verb tenses lose the usual values that are linked to the parameters of the context where the utterance occurs, and, instead, take on an intracontextual function. The imperfect and pluperfect (Fayol 1982b) in French, for example, will be applied to background states and actions, while the simple past or the present will mark the events that are being focused on. In sum, the system of references here will be different from the one used to insure cohesion. For while in the latter case, reference is made to what has gone before, in this case references are to what will come and thus create expectations. This is what Larochette (1945) called a "schema of incidence", e.g. "je dormAIS (I WAS sleepING); il entrA" (He enterED). It should be noted that the accumulation of French imperfects at the beginning of the narrative will lead the reader-listener to wonder "What is he get ting at?" the "main point" will only appear with a simple past or a present. This system of marks that works intratextually which has just been discussed ob viously only concerns average adults. It can in no way be applied to children (Fayol 1983a). The question then becomes, how do children acquire such a system?
The Acquisition of the Ability to Write a Story from One's Own Experience Two hundred and sixty-five children, from six to ten years of age, from various socio-cultural groups were asked to, "Write a story about something that happened to you." The textual organization of the resulting stories was analyzed in detail with focus on the author-subjects, verb forms, punctuation and connectives. We will not discuss here the conclusions concerning these elements as they have been reported elsewhere (cf. Fayol 1981, 1982). What concerns this paper is the general tenden cies that were brought out by the study and these will be illustrated by a number of examples. Our analyses have shown that the acquisition under discussion can be divided and schematized in five stages. These would correspond approximately to age or
STORIES
233
grade level in school. However, we cannot be sure that all children pass through all these stages or that if the instructions were changed there would be no change in the following chronology that was obtained. Type I. This is the most basic type or organization; it is found mainly in the work of children of six and seven years old. Either the author restricts the story to a single "news announcement" or there is a whole series of them. Each event is told in a perfect form (French passé composé) and related directly to the context of the speaker at the time of the writing {HIER (YESTERDAY), IL Y A ... JOURS (... DAYS AGO), AUJOURD'HUI (TODAY) ...)2. There is no intratextual marking system. The sentences are in no way integrated to create a whole, but are instead an assembly of independent actions and states, as in the following examples (the age in years and code name of the author are given in parenthesis): (1)
"Le père-noël m'a apporté une poupée et des cuillers" (Santa Clause brought me a doll and some spoons) (Lam, 6) (2) "LE JOUR DE LA PENTECÔTEj'ai été dans la forêt LES JEUDIS je vais à la piscine LE DIMANCHE je vais à la forêt. MERCREDIj'ai été à la peché des truites MERCREDI J' ai été à la piscine LUNDI j'ai ceuilli des fleurs pour maman" (On WHITSUN I went to the forest on THURSDAYS I go to the pool on SUN DAY I go to the forest. WEDNESDAY I went fishing for trout
WEDNESDAY I went to the pool MONDAY I picked flowers for mommy) (Aff, 6) (3) "Un dimanche j' ai été chez ma tata manger. Un jour ma meme est venue repasser chez moi. Je joue tout le temps au ballon. Un jour, j' ai joué à la corde. Un jour, j'ai été au zoo" (One Sunday I went to eat at my auntie's. One day my grandma came to iron at my house. I always play with a ball. One day I played with a rope. One day I went to the zoo) (Lav, 7) It is as if the children knew (at least for (2) and (3)) that a story is made up of several successive sentences, but have not understood that they are supposed to be
234
M. FAYOL
expansions of the same event. That being the case, their productions take on the aspect of successive, juxtaposed utterance turns. This is borne out by the reiteration of the text's "origins" - "LE JOUR DE LA PENTECÔTE" (WHITSUN), "LES JEUDIS" (ON
THURSDAYS), "LE DIMANCHE" (ON SUNDAYS) ... - which are directly related to the author's context at the time of the writing. It might be considered here, as Bereiter and Scardamalia (1982) have done, that these children situate their writing in a speech situation rather than in the framework of a story-telling situation. Type II. productions seem to be the continuation of Type I productions. Clearly, there is no longer simply a listing of completely independent "news an nouncements", however there is still no integration of the utterances into a whole. The child immediately defines a spatio-temporal frame - a sort of "container" - and "pours into it" a series of events which is generally in chronological order. For the most part, however, the successive facts, except for their order of occurrence and the given time and space in which they occur, appear quite independent, as can be seen in the following examples: (4)
"Dimanche j'ai été à la pêche avec mon papa et ma maman et ma soeur j' ai pris un poisson c'est un arc-en-ciel mon papa a pris un poisson ma maman a pris un arc-en-ciel" (Sunday I went fishing with my daddy and my mommy and my sister I caught a fish it is a rainbow my daddy caught a fish my mommy caught a rainbow) (Aug, 7) (5) "Un dimanche. Je suis allée à la forêt. Je me suis bien amusée avec mes trois cousins. J'ai vu une vipère au bord de la rive. J'ai été cueillir du muguet pour ma marraine, j'en ai trouvé beaucoup. J'ai couru jusqu'à la voiture on faisait une course et c'est moi qu'a gagné. C'est le soir. Il faut s'en aller. Tout le monde monte dans la voiturre" (One Sunday, I went to the forest. I had a lot of fun with my three cousins. I saw a snake on the riverside. I picked lily of the valley for my godmother, I found a lot. I ran all the way to the car we had a race and I won. It's evening.
STORIES
235
We must go. Everybody gets into the ) (, 7) (6) "UN JOUR en arrivant en Hollande un samedi, nous avons été dans un magasin où il y avait beaucoup de gens. DEUX MINUTES APRÈS je me suis perdu, heureusement j' ai retrouvé mes parents. LE LENDEMAIN MATIN on avait été à A ... Le LUNDI on a visité un musée ..." (ONE DAY when we arrived in Holland one Saturday, we were in a store where there were a lot of people. TWO MINUTES LATER I got lost, luckily I found my parents again. THE NEXT MORNING we were at A ... MONDAY we visited a museum ...) (Boy, 9) These types of stories are similar to scripts in that they almost always relate usual, everyday events that are inserted between two "borders" of world knowledge (go to/come back from; begin/finish, cf. Fayol 1980). The proceedings are related in the passé composé while only the states are given in the imperfect tense. The pronoun "I" remains dominant as in the preceding stage. On the other hand, the first marking system, one that insures cohesion, appears. Now the children no longer set up an origin on which the coordinates of the events will be plotted (cf (6) with its sequence "ONE DAY ... TWO MINUTES LATER ... THE NEXT MORNING ... MONDAY ...). The punctuation repeats these marks; it co-occurs with them and disappears when several facts are joined to one of them (cf (5) from "I ran ..." to "I won"). Although a story centered around a single event does not yet appear at this age, at least some definite progress concerning topical unity and cohesion can be found. In any case, the proceedings continue to be told in the passé-composé, a tense which relates to a finished action, or perfect form in the present. Thus, children favor as the story's anchor, the context they find themselves in, at the moment of the telling. Thus, simultaneously, there is both initial marking in relation to this context as well as positioning of other interprocedural marks, although these latter are almost always limited to chronological relations. The only frequently used con nectives are "AND" and the paradigm, "APRÈS/PUIS/ENSUITE' (AFTER THAT, THEN, THEN)3. It might be said then, that the story is still situated in the instance of its telling and develops step by step without the child anticipating an integrating event. Type HI. This third type is a turning point in that the productions that appear are quite heterogeneous. In the same texts can be found linking and marking strategies similar to those that are described for Type II, and other, more sophisticated ones, characteristic of Type IV, as in the following examples:
236
(7)
M. FAYOL
"Un dimanche mon papa m'a emmenée chez ma meme avec ma sœur et ma maman. On avait emmené ma chienne Miss. Nous sommes arrivés, dit papa. On entre à la maison, on emmène aussi Miss dit papa, bien sûr on ne veut pas la laisser dans la voiture. On s'installe à table et on mange. Je mange en cinq minutes. Ma chienne venait tout le temps près de moi. Après je vais m'amuser avec la chienne. Je vais par terre, ma chienne croyait que f avais un morceau de pain alors elle m'a fait glisser sur le carreau et les autres rigolaient tant qu'ils pouvaient" (One Sunday my daddy took me to my grandma's with my sister and my mommy. We took my dog Miss. We're here says daddy. We go into the house, we'll take Miss, too says daddy, of course we don't want to leave her in the car. We sit down at the table and eat. I eat in five minutes. My dog came to me all the time. Then I go AND play with the dog. I go on the floor, my dog thought that I had a piece of bread so she made me fall on the floor and the others laughed as hard as they could) (Sal, 7) (8) "Un jour chez mon grand-père, je donnais du grain aux poules et aux lapins. Ensuite j'ai ramassé des noisettes ... Puis j'ai fait quatre heures. Et après avec mon grand-père nous avons attelé le poney ... J'ai fait le tour du champ de mon grand-père tout allait bien ensuite, je l'ai promené et il m'a mordu. Alors mon grand-père s'est précipité sur moi et pompom s'est en allé dans son champ. Et j'avais un gros bleu dans le dos ..." (One day at my grandfather's, I gave some grain to the chickens and rab bits. Then I gathered nuts ... Then I had a snack. And then my grandfather and I saddled the pony ... I made a tour of my grandfather's field every thing went well after that, I walked him and he bit me. So my grandfather hurried over to me and pompom went off in the field. And I had a big black and blue mark on my back ...) (Sey, 8)
STORIES
237
At this stage, children, about 8 years old, write very long texts. This is a result of the fact that their productions are made up of one or several scripts which start off and/or finish the story (cf. both examples (7) and (8)), and, at the same time, in clude an episode which generally corresponds to a complication followed by a resolution. However, this episode is often "drowned" in a succession of ordinary events. Children at this stage do not yet center on a single event. At the same time, some linguistic marks, which have been heretofore absent, can now be observed. They appear mainly in episodes which, unlike scripts, are unpunctuated. This same is true for certain connectives - "SOUDAIN" (SUDDENLY), "TOUT À COUP" (ALL OF A SUDDEN), "MAIS" (BUT), "ALORS" (SO) ... - and verb
forms like the (narrative) present and the passé simple. While these are still used infrequently, the children use them "locally" and strategically when they recount complications or resolutions, as in the following examples: (9)
"TOUT À COUP le bouchon s'enfonce mon papa tire le poisson s'en va" ... (This episode follows five passé-composé's) (all of a sudden the cork fell in my daddy pulled the fish fell off) (Mar, 8) (10) "... et TOUT À COUP il se met à sonner" (This comes after a very long se quence of passé-composé''s, imperfects and pluperfects) (...and all of a sudden it started to ring) (11) "... je pris un verre ... le verre glissa et le verre tomba ... et se cassa. Maman me fâcha" (This is part of a story twenty-one propositions long) (...I picked up a glass ... the glass slipped and the glass fell ... and broke. Mommy got mad at me) (Dur, 9) But the most important phenomenon to be found in this type of production is certainly the setting up of what Culioli (ibid.) call "translated origins", i.e. the use of sequences of events related in the imperfect or pluperfect in order to start off the story (cf. Franckel 1976; Fuchs and Leonard 1979). Obviously, the use of such forms is hesitant and not always correct, however, there is certainly an attempt here to go from a system characterized by speech marked by the passé-composé to one where there is greater "distance" (Fayol 1984, 1985b), thus pennitting it to become somewhat autonomous. In this latter case, events are "plotted" into a system of spatio-temporal coordinates that are relatively independent of the context in which the utterance occurs. Type 1 productions are made up of elements belonging to speech and others belonging to stories. This can be seen in the "content" - scripts describing usual, chronologically-ordered facts that occur along with episodes made up of complica-
238
M. FAYOL
tions and solutions - as well as in the presence and function of certain surface marks. The passé-composé, the connectives "AND" and "THEN", and a greater use of punctuation is associated with scripts. The occurrence of episodes brings with it corresponding changes in verb tenses (the use of the present and the passé-simple), the use of "SUDDENLY" and "ALL OF A SUDDEN", "BUT" and "SO", and a decrease of punctuation. Marks relative to the situation of the child as s/he tells the story dominate in scripts; marks relative to the climax of the story appear with episodes. Type IV. Henceforth, the child situates him/herself straightaway in the story. Narratives about usual, everyday happenings disappear and the story centers on one or more episodes. These productions feel intuitively "interesting" as they relate un expected facts. The surface marks of these texts are clearly different from those found in speech. The initial utterance sets up a system of spatio-temporal coordinates separated from the child's real and immediate situation; "translated origins" are used more widely. Schemas based on incidents are used regularly; the child begins the narration by a sequence, which can sometimes be very long, in the imparfait and/or the pluperfect and reserves the present or the passé-simple for unexpected events, or those that are felt by the child to be unexpected. The connectives SOUDAIN (SUDDENLY), ALORS (SO) and MAIS (BUT) are used regularly. Punctuation becomes more varied but becomes more scarce when the story describes complica tions or attempts to resolve them. For example: (12) "Nous attendions depuis deux jours que la neige tombe, mais aucun espoir. A chaque instant nous regardions à la fenêtre. Rien! encore rien! toujours rien! disait-on. Le jour suivant une neige dure, épaisse et sale faisait son plein aucune trace de verdure la fontaine était gelée je m'habillais d'un gros manteau de fourrure et de mes grosses bottes vertes. Je sortais dehors sans prévoir que mon voisin allait me faire une farce. Bien caché derrière une haie envahie de neige il avait préparé les dix boules il en tira neuf arrivé à la dixième il s'arrêta deux secondes croyant qu'il n'en avait plus je m'étais montré et piaf il m'avait touché, hélas c'était déjà l'heure de manger, le repas terminé nous recommencions à jouer" (We had waited two days for snow to fall, but no hope. We constantly looked out the window. Nothing! still nothing! always nothing! we said. The next day a hard, thick and dirty snow was everywhere there was not a trace of green the fountain was frozen I put on a big fur coat and my big
STORIES
239
green boots. I went out without knowing my neighbor was going to play a trick on me. Well hidden behind a hedge covered with snow he had prepared the ten balls he threw nine and at the tenth stopped for two seconds thinking he didn't have anymore I showed myself and "splat" he hit me unfortunately it was already time to eat, when the meal was finished we began to play again) (Ner, 10) Story (12) clearly illustrates the way children proceed when they are at this stage in story-writing. The story is told in a past that is relatively independent from where the child is situated. The very long sequence of imperfect verbs does not stop until the story gets to the complication; the string of imperfects is picked up immediately after this. This differentiation of verb tenses seems to be the child's way of focusing on the climax of the story. Verbs now function intratextually and no longer relative to the child's real situation. Type V. This type was very rare in our sample. It differs from Type IV by its "form", e.g. surface marks, only; its episodic organization remains similar. Children who write narratives of this type relate them in the present rather than in the past that, in fact, characterizes the events and states that are told about. They use "today" as their origin, thus "dedeixifying" and placing the story in the realm of "always" (Benveniste 1965), as in the following examples: (13) "Ma chambre est tout déballée AUJOURD'HUI ..." (My room is a complete mess TODAY ...) (Leb, 9) (14) "AUJOURD'HUI c'est mercredi ..." (TODAY is Wednesday ...) (Lag, 11) (15) "AUJOURD'HUI c'est dimanche ..." (TODAY is Sunday ...) (Reb, 10) (16) [Extracts] "Aujourd'hui nous devons emmener le troupeau dans le champ pour le surveiller. A quatre nous partons avec le troupeau et le chien ...Il est dix heures, nous devons rentrer le troupeau. 'Youpi appporte-les!' Le jeune chien qui vient just de commencer son métier de chien de berger part chercher les brebis. Le troupeau remonte en courant et vient vers nous sans le chien. Nous comptons le troupeau et il y en a que quatre cent quarante-neuf il en manque une. nous rentrons le troupeau à la bergerie et nous repartons dans le champ. Nous en faisons le tour et nous trouvons Youpi avec la brebis qui a le ventre ouvert. Nous sommes furieux ...
240
M. FAYOL
Depuis ce jour-là nous n'emmenons plus le chien avec nous pour garder le troupeau" (Today we must take the herd to the field to guard them. Four of us leave with the heard and the dog ... It is ten o'clock, we must bring the herd back. 'Youpi get them! ' The young dog who had just begun its career as a sheep dog leaves to get the ewes. The herd runs up toward us without the dog. We count the sheep and there are only four hundred and forty-nine one is missing. We take the to the shed and go back to the field. We make a tour of it and find Youpi with the ewe whose belly is wide open. We are furious... From that day on we don't take the dog with us to guard the herd) (Lec, 10) All references to the speaking situation have disappeared leaving nothing but in terprocess relations. It is not until the final phrase, the "coda" as Labov and Waletsky (ibid) call it, that any traces anchoring the utterance, appear. Except for this, the story has become autonomous; it unfolds by itself according to an intratextual organization that is independent of the parameters of the reader-listener's situa tion. The author seems to have selected one and only one event and cut it up into constitutive components organized causally (in the largest sense of the word). In relating this event, s/he has anticipated the climax and put a distance between him/herself and the facts as well as of the story itself, so that there remains no "trace" of the speaker.
Conclusion The analysis that we have carried out on texts written by children in elementary school shows the very close correspondence that exists between two large categories of phenomena. First, the "contents" of what is told about and their or ganization clearly change. At best, the youngest children string independent events together, each of these being a "news announcement". The oldest children for the most part, however, center their story around a single fact that they expand upon and which they relate through elements organized into an episode. This seems to correspond to the "schema" that has been isolated by a number of researchers (Glenn 1978; Stein and Glenn 1979, 1982; Stein and Trabasso 1982).
STORIES
241
Second, there can be seen a gradual change from speech forms to story forms. At six or seven children write as if they were in the framework of a conversation. The facts that they tell about each relate to a "speech turn" showing signs of a strong anchor in the author's immediate situation (predominance of "I", the passécomposé, elements marked as "near" the author's spatio-temporal situation). On the other hand, subjects of ten to eleven years of age regularly use linguistic markers that belong specifically to stories; they more or less completely get rid of all traces of an anchor in their immediate situation (cf. Types IV and V). They regularly make use of contrasting verb forms, specific connectives etc. At an intermediate stage, at around eight to nine years old, productions show both speech and narra tive forms; cohesion and topical continuity alone can be observed; there is no focus on an episode. This concurrent development of the organization of content and the implementa tion of a system of linguistic markers is extremely important and deserves com ment. One might ask why this is so and what factors contribute to it. Our hypothesis is that the child can understand how intratextual markers function only if they can be related to a cognitive-semantic structure representing series of actions and their connections. Concentration on a single event along with the expansion of that event into com ponents that are organized according to rules governing the succession of narrative categories as modeled by a schema (Mandler and Johnson, ibid.; Stein and Glenn, ibid.) seems to be prerequisite to the inductive understanding of surface markers. This understanding, however, does not come about all at once. Topical continuity (Karmiloff-Smith, ibid.) and cohesion come first. First, the child seems to perceive that a story enlarges upon a series of events relative to a single topic and would then infer from examples what linguistic means contribute to this. Following this, the child sets up a system of contrasting, intratextual verb forms. This comes much later and is slow to develop; in fact, it is not yet fully mastered at age ten or eleven (Fayol 1985a). Contrasting forms only appear "in places". The present or passésimple would be applied to a single happening (most often an unexpected fact) before its use is extended to the full complication-resolution component of the story. Finally, anchorage of the story in the child's immediate situation as s/he writes, using markers that are characteristic of speech forms (where all past events are treated as finished) disappear as "distance" is created between the story and the child's situation. Use of "translated origins" (the imparfait) plus the regular use of the present indicates the story's progress toward autonomy from speech parameters; information is given as to who is telling the story, when, where and to whom.
242
M. FAYOL
It must be admitted that proof of this general conception demands a series of studies that combine the analyses of corpora of different types (we have limited ourselves here to stories concerning personal experience) and experimental studies with better controlled variables. This is, in fact, what we intend to begin doing now. It still remains to determine what factors contribute to this development. The ab sence of research into this problem does not permit us to determine exactly what they are. However, it might be hypothesized that they fall into three main categories (Fayol 1985b): First, the child certainly reconstructs a representation of events and rules for linking these events. This results in cognitive-semantic representation modeled in terms of scripts or plans (Rumelhart, ibid.; Schank, ibid.; Schank and Abelson, ibid.), i.e. conventional narrative schemas (Esperet 1984). Second, in their daily interactions, whenever children "announce news", they find themselves being questioned by adults or peers for more information concern ing the antecedents and the consequences of the event (and then? why? ...) Here too, it might be that as the subject gains more and more experience of this sort, s/he internalizes this type of reaction. Going so far as to anticipate potential ques tions. This would lead to his/her answering them in advance, thus including right in the story those elements that are essential to its understanding. Unfortunately, there has still been very little work carried out on this problem (cf. Magee and SuttonSmith 1983) and this hypothesis has yet to be confirmed. Third, the child is bombarded with texts daily, stories told or read by adults at school. The child, thus, has a great number of more or less literary examples from which to make inferences about linguistic markers and their intratextual function ing. It is most likely that the mastery of (written or oral) story-telling depends on the interaction of all three of these factors and not just on one of them, e.g. the cogni tive-semantic schema on which almost all the research has been carried out on. Determining the share in the process of each one, their pertinence and the moment that they come into play is a problem that has not yet been attacked.
NOTES 1. There have been a number of studies carried out on isolated sentences, sentence pairs (cf. for example the recent work by Bronckart, Kail and Noizet 1983) and coreference phenomena. However given the way in which the data are presented, they are cannot be applied to issues concerning texts.
STORIES
243
2. Although a large amount of research has been carried out on the acquisition of verb tenses in recent years, a certain number of problems still remain. An attempt at a synthesis of the theoretical and methodological problems concerning the acquisition of verb tenses as concerns texts can be found in Fayol (1985a, b). 3. Studies on connectives have been carried out mainly by using sentence pairs. (cf. Kail and Weissenborn 1984b for a synthesis of these studies). But here, too, the results cannot neces sarily be directly applied to texts which pose problems that are specific to texts (cf. Haberlandt 1982 for an approach to the issue from the point of view of texts).
The Development of Argumentative Discourse G. Piéraut-Le Bonniec Laboratoire de Psychobiologie de l'Enfant, Paris
and M. Valette Collège Saint Simon, Jouars-Pontchartrain
In general, it can be said that in the case of narrative or descriptive discourse, speakers tell something to others or to themselves, while in the case of argumenta tive discourse, the speaker reasons things out in order to be more convincing. The study of argumentation is situated at the intersection of two important issues in lin guistics, one that concerns language processes (linguistics and communication) and the other which concerns what Piaget calls the "mental operations" that are carried out by a subject. Thus, before any study can be made on how argumentative ability develops in adolescents. The properties that make up this type of discourse must be studied and the following questions must be addressed. 1. If argumentative discourse is characterized by reasoning, what differentiates this type of reasoning from other forms of reasoning, e.g. formal logical reasoning or mathematical reasoning? 2. In what ways does the study of argumentation differ from those studies that have been carried out on reasoning in other situations, e.g. problem-solving situations? 3. If argumentation is a specific type of discourse, what linguistic markers dif ferentiate it from other forms of discourse, e.g. from narrative discourse?
246
G. PIERAUT-LE BONNIEC AND M. VALETTE
Argumentative Discourse
Argumentation and Propositional Logic Philosophers have long questioned how logic is expressed in language. Thinking on this subject goes back as far as Aristotle in Ancient Greece, the Scholastics (cf. Moody 1958) in the Middle Ages and the Port-Royal grammarians of the 17th cen tury (Arnauld and Nicole 1662). Over the years logicians have drawn attention to the "weaknesses" in natural language where most words are polysemous and propositions are rarely exempt from ambiguity. Logical relations between state ments vary in rigor because of the ambiguity and polysemy of logical connectives. For this to become clear we need only think of all the meanings the coordinating conjunction "and" and the subordinating conjunction "if'"can take on. A number of studies (Wason 1966, 1968, 1983; Johnson-Laird and Tagart 1969; Johnson-Laird 1983) have shown that although nothing prevents us from using the "if ... then" material implication in speech, it is rarely assigned this value. This is why, since the time of the Ancient Greeks, logicians and mathematicians have attempted to devise symbolic languages free of ambiguity. Thus, Leibnitz' dream has, to a cer tain extent at least, been fulfilled by the formal logic of the 20th century, which has become the science of formal systems. Does this mean that the reasoning carried out in natural language is not rational since it cannot be symbolized mathematically or expressed in logical propositions? But logic employs only a restricted number of logical connectives among those available in natural languages. It ignores connectives indicating cause or concession as well as markers of adversative opposition such as BUT, HOWEVER, ON THE ONE HAND... ON THE OTHER HAND, etc. Thus it is, in fact, the richness of natural lan guages that permit us to reason in a way that is both productive and adapted to the problem situations that have to be dealt with. Furthermore, formal logical deduction, which is true for all times, all places and all speakers, is totally unsuited as a method of solving the type of problems we en counter daily, be they identifying the cause when a system breaks down, making a decision, or convincing an allocutor that a solution is adequate. On the other hand, argumentative reasoning never takes a form that results in necessary truth as is found in mathematical statements. Solutions in natural language differ from solu tions in formal logic in that they are based in the realm of the possible or the prob able and never in the realm of truth functions. Argumentative reasoning is based on
THE DEVELOPMENT OF ARGUMENTATIVE DISCOURSE
247
a kind of "everyday logic" (Grize 1982) which differs from, but is not incompatible with, classical prepositional logic.
Argumentation and Reasoning In problem-solving situations, subjects may use argumentation to offer evidence as to their reasoning, so that it can be understood. This is particularly true of the experiments that Piaget and his colleagues carried out using methods taken from clinical psychology. This type of experiment included a dialogical exchange be tween the experimenter and the subject, thus allowing subjects to explain their ar guments. And most often the type of argumentation used by his subjects was what Piaget used to define the different operational stages of development. The impor tance Piaget and Inhelder (1941/1962) lent to their subjects' argumentation in their studies concerning the conservation of physical quantities and which they lent to the development of hypothetico-deductive reasoning in adolescents (Inhelder and Piaget 1955), is well known. However, many other nondiscursive elements could be taken into account: the way in which subjects explore the situation or their trials and errors as they look for a solution. The subjects in Piagetian-type experiments are always placed in situations which resemble those the physicist deals with. The subject has direct access to reality and can test, for example, whether the flexibility of a rod is a function of its length, thickness, shape (round/square), or of the material it is made of, etc. The question asked here, however, is that while subjects are able to answer questions that are asked in situations that they control, how well can they construct through language and ex situ a global representation of that same situation by logically organizing the reasons that validate their solution? A Piagetian perspective of language does not answer such a question, nor can it be answered within the perspective taken by authors interested in problem-solving. But the study of argumentation is specifical ly aimed at learning how discourse functions as an organizer of thought and reasoning. Argumentation is always supported by reasoning, but the study of ar gumentation is the study of the discourse in which this reasoning is expressed, of how the evidence that is extracted from the situation takes on meaning and values that render it pertinent for the argument the locutor wants to defend, of how knowledge is articulated so that a possible world is created for the allocutor, of how the locutor expresses degrees of conviction in relation to what s/he says. In short word, studying argumentation means studying how language resources are
248
G. PIÉRAUT-LE BONNIEC AND M. VALETTE
used to create that representation of the world, which Grize (1982) calls a schema, that the locutor hopes to share with the allocutor. It has been observed that very early on children are able to use certain foms of argumentation as can be seen in the following as publlished by Caron (1985): "ƒ/ pleure Jérome. Si Jérome, il pleure, alors c'est un bébé. C'est un bébé Jérome". (Jerome is crying. If Jerome is crying, then he's a baby. Jerome's a baby (Julie, age 4.1). It is obvious here that the distinction must be made between reasoning and ar gumentation. Literally, Julie seems to want to "show" that Jerome is a baby, since he's crying. But Julie's aim can clearly be seen to be much more subtle. She al ready knows how to use language resources to create a universe in which her little brother cuts a sorry figure, and where in contrast, her own worth as a sensible little girl is increased in her father's eyes.
Argumentation and Narration The problem, then, is to characterize argumentation and narration and to dif ferentiate them. Both types of discourse include properties and constraints that are found in everyday language. Thus, narration like argumentation, as opposed to for mal logic, are both found in real interpersonal speech situations. Arguments are presented to someone; someone is told something, even if that someone is one's self as, for example, in a diary. Discourse in natural language is never neutral; speakers choose to say what they do in accordance with their representation of the person in front of them. Furthermore, the referents that occur in natural language discourse belong to the domain of cultural knowledge known as cultural pre constructions. For example, although it is never stated explicitly, the word "re volver" evokes, among other things, the ideas of murder, blood, danger, etc. Also, in natural language, both in the case of argumentation and narration, discourse referents are never defined once and for all as they are in formal logic. Thus, in a detective novel, for example, the revolver can either be the offensive weapon used by the criminal or the defensive weapon used by the guard. Finally, for both narra tion and argumentation, even if speakers do not want to demonstrate a necessary truth, their discourse must be both plausible and cohesive. Even in the most out landish fiction, once the rules setting the story outside the domain of reality are es-
THE DEVELOPMENT OF ARGUMENTATIVE DISCOURSE
249
tablished, some sort of coherence must be maintained (cf. in this volume: Kail and Weissenborn; Bronckart and Schneuwly; Hickmann). Therefore, in order to distin guish argumentative discourse from narrative discourse the following four points must be considered: a. Argumentative discourse corresponds to a specific interactive situation where there is, or could be, disagreement between the interlocutors. The speaker wants to and tries to convince the listener that s/he is right, be it to arrive at a shared opinion, a decision or a justification. Argumentative discourse is, thus,linked to a specific context. b. The objects of argumentative discourse are chosen in function with the place they can occupy within the general economy of the argumentation. This in cludes the meaning they may take on, the positive or negative value they may be attributed. The relations that will be established among them depend on the argument to be supported. Furthermore, if a speaker wants to convince a lis tener of something, the meaning s/he assigns to the discourse referents must be explained. Finally, the meaning of referents can change during the course of the discourse since given features of referents are not equally relevant during the course of the argumentation. c. Argumentative discourse is strongly structured. In the case of narration it is often sufficient to mark temporal relations, anteriority, simultaneity or posteriority. Argumentative discourse often expresses many more causal rela tions which can be either explicative (e.g. It's raining because the atmospheric pressure is low) or justificatory (e.g. It's raining because you can see people walking down the street with their umbrellas open). It expresses more relations of consequence or of opposition (cf. Grize and Piéraut-le Bonniec 1983), and sometimes even hypothetical deductive relations. To create certain speech ef fects, the conjunction that should be expressed might remain implicit (e.g. He has blood on his hands; he's the murderer). But in any case, it is only with great difficulty that reasoning can be presented in speech without logical links, something which is perfectly possible when narrating a story. d. The speaker is strongly implicated in argumentative discourse and in contrast to narration, argumentation rarely takes a declarative form. This is a form of discourse where opinions are put forward and listeners are expected to indicate their attitude toward speakers and their speech. Ideas may be presented from one point of view or another at the beginning of a discussion, but through ar gumentation one view will finally become believable or it will no longer be believable. Because of this, argumentative discourse must include many more modal expressions than narrative discourse. The following are not the same: "I
250
G. PEÉRAUT-LE BONNIEC AND M. VALETTE
believe Peter is guilty" / "Maybe Peter is guilty" / "It is likely that Peter is guilty" / "I am sure that Peter is guilty" / "People say that Peter is guilty". In these utterances, the speaker is not implicated in the same way as in the state ment, "Peter is guilty". Thus an argumentative text can be seen as a repository of those linguistic markers of the reasoning process that the subject masters.
The Development of Argumentative Discourse in Adolescents: Experimental Studies A lot of work from a number of perspectives has been carried out on adult ar gumentation in French (Perelman and Olbrechts-Tyteca 1970; Grize 1982; Borel, Grize and Mieville 1983; Grize and Piéraut-Le Bonniec 1983; Meyer 1982; Anscombre and Ducrot 1983). While these authors have been mainly interested in adult argumentation others have worked on acquisition. Several studies have been made concerning the acquisition of logical connectives both in French and English (cf. Kail and Weissenbom in this volume). A few authors like Caron (1987), Eisen berg and Garvey (Eisenberg and Garvey 1981; Garvey and Eisenberg 1986; Eisen berg 1987) have studied the emergence of persuasive strategies in the spontaneous language of very young children (age: 2 - 6 years). On the other hand, specific aspects of the development of argumentative skills in older children (age 7 - 1 4 years) have been studied in experimental situations. Esperet et al. worked on how implication markers were used by children in different types of argumentative dis course. Brandt (1988) has done some very important experimental work on the development of argumentation in children aged 7 to 12 years. Placed in a game situation, two children make up a team that play against a single adult. They were then observed as to how they used argumentation to convince their teammate of the validity of the strategy they proposed. The study centered on the use of negation. As for adolescents, a good amount of work exists on their hypothetico-deductive reasoning (Inhelder and Piaget 1955; Matalon 1962; Grize and Matalon 1962; Overton et al. 1987; Johnson-Laird 1983; Wason 1966, 1968, 1983; Cox and Griggs 1982), however nothing has been carried out on their ability to organize their argumentation. But adolescents certainly have a problem. To be convinced of this, one need only listen to their teachers' complaints about their great difficulty in formulating and clearly transmitting their ideas and reasoning them out. This would seem to be paradoxical in that first of all, children master their native language
THE DEVELOPMENT OF ARGUMENTATIVE DISCOURSE
251
very early on (cf. Part 3 of this volume) and secondly, studies have shown (PiérautLe Bonniec 1980a, 1990,) that at the time when children enter adolescence period, they dispose of sufficient cognitive equipment to use all kinds of reasoning. Their acquisition of the notions of possibility and necessity, on the one hand, and of decidability and indecidability on the other hand, brings the complex reasoning needed for mathematics and physics within their reach. This, then, brings out the importance of studying the development during adolescence of the ability to use the resources of natural language to reason in situations which are different from the forms of reasoning used in solving physical or mathematical problems.
Method The source material for the experiment was a short thriller by F. Brown (1981) entitled "Une histoire de fou" ("The Story of a Madman", our trnslation). The sub jects were given the main points of the mystery: a crime was committed in an in sane asylum located in a small town; there is every reason to think that the wanted criminal has got away by taking a train from the station in town; the police detec tive has just questioned the station-master who has given him a description of the three men who had been in the station's waiting room before the train stopped. The police detective thinks he can identify one of these three men as the escapee and writes a report for his sergeant. These three suspects were described only as to their physical appearance or behavior, as they would have been seen by someone who did not know the men and who was not aware that a crime had just been com mitted in the area. Fourteen features were provided: 5 for the first two characters and 4 for the third. The subjects' task consisted of writing the report the investigat ing officer had to send to the police chief, i.e. the subjects were asked to use these descriptions to decide, like the policeman, who was the most likely culprit. Table 1 presents the text in the form it was given to each subject. The subjects were told to consider the task as a written assignment and they could have all the time they needed, although, in general, less than an hour was sufficient. The subjects found the task interesting and enjoyable. The population consisted of thirty adolescents: ten between 11 and 12 years (School level 1 = SL.l), ten between 14 and 15 (SL.2), and ten between 16 and 17 (SL.3). They attended two schools in neighborhoods of relatively high socio economic standing in the Paris area.
252
G. PIÉRAUT-LE BONNIEC AND M. VALETTE
Table 1
Instructions Text
Police Enquiry into the Crime Committed at Saint-Guy sur Bièvre A madman escaped yesterday from the insane asylum at Saint-Guy sur Bièvre lafter killing a guard. A police detective has come to talk to the station-master of this small town. Here is the description the station-master gave him of the people] [who were in the waiting room last night between 6 p.m. and 8 p.m.: First man:
was out of breath was slightly bald was wearing tortoise-shell glasses had a suitcase said he had just come from the barber's
Second man:
was tall and thin was wearing good-quality clothes but his pants were too short had red eyes constantly played with the poker for the stove
Third man:
was wearing something that looked like a uniform was tall and heavy was cross-eyed was carrying a revolver
The police detective gave these facts to his chief and explained why he thought] that one of these men was the madman. Write the detective's report.
THE DEVELOPMENT OF ARGUMENTATIVE DISCOURSE
253
Results The Criteria of Argumentative Discourse. First of all the children's texts were studied from the point of view three of the traits that have already been mentioned above (points to b, and d, page 247), as most pertinent in characterizing an ar gumentative text. It was deemed unnecessary to study the setting of the discourse, as the situation was defined in advance by the experiment. The subjects had been instructed to "put themselves in the shoes" of the police detective and to convince their chief that they had identified the murderer. The issue then was to see how the subjects adapted to this situation. The Objects of Discourse: How the Given Clues Were Used. Normally in the given situation, the subjects would be expected to consider the clues and assign a value (positive, negative or neutral) to each one according to the argument they wanted to defend. However, as has already been mentioned, in an argumentative situation, the significance of a discourse object is not necessarily assigned once and for all. Consequently, it was interesting to see how well the subjects could discuss the importance of the clues. Furthermore, in an argumentative situation the specific significance that is conferred on objects of discourse is dependent on the way in which they relate to each other, e.g. an adolescent's wearing pants that are too short does not have the same significance as an adult's doing so. Number of Clues Used. The mean number of clues used by the 11 to 12-year-old group is 3.7 (sd=2.34), by the 14 to 15-year olds it is 8.1 (sd=1.61) and by the 16 to 17-year-olds it is 7.3 (sd=3.54). These results cohere with the fact that in the youngest group the clues are used to narrate the investigation and are enhanced by invented details. The 11 to 12-year-olds use less given information than clues they invent. As of 14 to 15-years old, 50% or more of the clues are used. All the clues are not used to the same degree: the revolver and the uniform appear in 22 texts whereas the clue "glasses" appears in only 8 out of the 3 0 texts. However, what is more important is that whereas in the 11 to 12-year-olds' texts the revolver and the uniform alone account for 32% of the clues mentioned, in the 16 to 17-year-olds', they only count for 16% as other clues, such as the suitcase or red eyes, have been taken into account at least as often as the revolver and the uniform. This shows that as of 13 to 14-years old the subjects were able to focus on clues other than the most salient ones.
254
G. PIERAUT-LE BONNIEC AND M. VALETTE
Use of Clues. A distinction must be made between cases where the clue is only cited and those cases where it is discussed: Example 1: "He is big and heavy-set; he is wearing a uniform and is armed with a revolver" Example 2: "The third man was carrying a revolver on him that he could have used to kill the guard after stealing his uniform. But I don't see how this man could have obtained a revolver if he was locked into an in sane asylum." In the 11 to 12-year-old group, 80% of the clues are cited without explicitation whereas this figure drops to 24% in the 14 to 15-year-old group and further falls to 4% in the 16 to 17-year-old group. One of the major differences between age groups is the way in which subjects establish connections between clues. In the youngest group of children, only one such relationship was observed ("expensive clothes, but pants too short") whereas there were 13 in the middle group and 21 in the oldest group. Example:
"This man stated he had just come from the barber shop which is surprising since he is somewhat bald. In addition, you rarely see people going to get their hair cut with a suitcase just before they decide to go on a trip."
Lastly, whereas the 11 to 12-year-olds' clues are only used in an incriminating fashion, the 14 to 15-year olds', and even more so the 16 to 17-year-olds', used them either in an incriminating or a nonincriminating fashion. Clues which are first placed in an incriminating light may be presented in a nonincriminating fashion later on in a text. Example:
"One of the three men was wearing something that resembled a uniform. I suspected him first. But I decided later he was innocent on the basis of the other clues. He was big and heavy-set, whereas the fugitive should be thin and agile since he demonstrated his agility by escaping via the rooftops."
Use of logical connectives. Words that were used to express logical relations were singled out and listed. They were then classified in accordance with the rules of French grammar as published by Grévisse (1959) and Wagner and Pinchon
THE DEVELOPMENT OF ARGUMENTATIVE DISCOURSE
255
(1962) who are both linguists which are considered authorities on the subject. Seven logical relations were considered to be expressible either by conjunctions of coordination or of subordination. - Cause can be expressed by CAR (FOR/AS), EN EFFET (INDEED) or by PARCE QUE (BECAUSE), PUISQUE (SINCE), VU QUE (IN VIEW OF THE FACT THAT), ÉTANT DONNÉ QUE (BEING THAT), SURTOUT QUE (ESPECIALLY SINCE), DE PEUR QUE (OUT OF FEAR THAT) Example 1: "Nous pouvons aussi penser qu'il s'est pressé CAR son retard lui aurait fait manqué son train" (We may also think that he was in a hurry AS his tardiness would have made him miss his train) Example 2: "Ceci prouve qu'il avait couru, non pas par peur de manquer son train, PUlSQU'il attendait avec les autres et qu'il avait une cer taine avance". (This shows that he had been running, not out of fear of missing his train, SINCE he was waiting with the others and he had been somewhat early.) - Consecution can be expressed through the use of the conjunction DONC (THEREFORE) or ALORS (THUS) or e x p r e s s i o n s s u c h as EN SORTE QUE, SI BIEN QUE, AU POINT QUE, (SO THAT, SUCH THAT).
Example 1: "Le fou échappé de l'asile aurait DONC attaqué le gardien par derrière" (The madman who escaped from the asylum would THUS have at tacked the guard from behind) Example 2: "Je m'empresse DONC de prendre le train vers la direction indiqué" (THUS I hurry to take the train in the direction shown.) It should be noticed that the value of DONC is not the same in the two ex amples. In the first one, it marks the conclusion of the reasoning while in the second it weakly links the two events. This distinction will be made when oc currences of DONC are counted. - Concession is expressed by QUOIQUE, BIEN QUE (ALTHOUGH) or MALGRÉ (EVEN THOUGH).
256
Example:
G. PIERAUT-LE BONNIEC AND M. VALETTE
"Il a dit qu'il revenait de chez le coiffeur BIEN Qu'il soit prati quement chauve". (He said he was coming back from the barber's EVEN THOUGH he was almost bald.)
- Goals are expressed by AFIN QUE (SO THAT) or POUR (TO). Example:
"Il portait des lunettes d'écaillés AFIN DE passer inaperçu". (He was wearing tortoise-shell glasses IN ORDER TO go un noticed)
- Conditionals are expressed by SI, (IF) À SUPPOSER QUE, (LETS SAY THAT) AU CAS OÙ (IN CASE). Example:
"ET MÊME S'IL était réellement allé chez le coiffeur besoin de l'annoncer à tout le monde!".
aurait-il
eu
(And EVEN IF he had really gone to the barber's did he have to announce it to everybody?) - Opposition can be expressed by MAIS (BUT), PAR CONTRE, (ON THE OTHER HAND) POURTANT or TANDIS QUE (WHILE/WHEREAS). Example:
"Il reste le troisième suspect. Ses yeux louchaient, je ne pus porter un jugement sur un si petit point MAIS on ne peut pas nier que si nous représentons souvent un fou en louchant..."'. (There remains the third suspect. He was cross-eyed. I cannot judge him on such a small point BUT you cannot deny that if we often represent a madman as cross-eyed ...)
- Co-occurrence is expressed by OR (WELL, INDEED) which punctuates two ele ments of the reasoning. Example:
"Il a aussi déclaré sortir de chez le coiffeur et le chef de gare m'a affirmé qu'il portait une valise. OR qu'est-ce qu'un fou aurait-il pu faire avec une valise!"
THE DEVELOPMENT OF ARGUMENTATIVE DISCOURSE
257
(He also declared that he had come from the barber's and the sta tion-master told me that he was carrying a suitcase. WELL, what could a madman be doing with a suitcase?) Table 2 shows the distribution of occurrences of these different markers of logi cal relations for each of the three age groups. It should be noted that SL.2 has the greatest number of total occurrences. The drop between SL.2 and SL.3 might be explained by the fact that the texts of the older children are generally shorter and the argumentation is tighter; but it will be seen that this group's argumentation is not weakened. However the comparison of marker occurrences between SL. 1 and SL.2 is the most interesting. First, there is an increase in their number of occurren ces, from 53 to 85. But even more important is the evolution that can be seen in the distribution of the different types of relations that are expressed. In particular, there is a reversal in the distribution of cause and consecution: 28% and 41% respective ly for SL.l, and 39% and 23% for SL.2. It should also be mentioned that for the SL.l group DONC is used 8 out of 17 times to link events (and only 9 times to mark the conclusion of reasoning) while for the SL.2 group, it is used 18 out of 19 times to introduce the conclusion of the reasoning. The proportions remain the
Table 2 SL.l
SL.2
SL.3
N = 53
N = 85
N = 661
Cause
28%
39%
45%
Consecution
41%
23%
18%
11%
10%
12%
9%
3%
13%
19%
Logical relations
1%
Concession Goals Conditionals Opposition Co-occurrences
18%
5%
Occurrences of expressions marking logical relations in percentages (TV = number of occurrences). 1. The texts of the older children are shorter and the ar gumentation is tighter.
258
G. PIÉRAUT-LE BONNIEC AND M. VALETTE
same for the SL.3 group. Finally, it should be noted that conditional relations are not expressed until SL.2.
Modalities There are a number of ways in which speakers can express how far they can be held for what they say, particularly when they want to limit their responsibility. In French, the most frequently used strategies involve the use of - a modal auxiliary, e.g. POUVOIR (MIGHT, MAY) to indicate a possibility: Example 1: "Le troisième homme portait un revolver avec lequel il AURAIT PU tuer le gardien". (The third man carried a revolver which he COULD HAVE killed the guard with.) Example 2: "Il portait des lunettes d'écaillés, ce qui POURRAIT être pour lui un moyen de ne pas être reconnu." (He was wearing tortoise-shell glasses which MIGHT have been a way for him not to be recognized.) - or a conditional form: Example:
"Voilà comment se SERAIT passé le crime suivant mes hypothèses". (That's how the crime WOULD HAVE occurred according to my hypotheses.)
- But other means also exist: - an adverb expressing possibility or probability can be used: Example 1: "Après réflexion je pensais que le coiffeur l'avait PROBABLEMENT retardé". (After some consideration I thought the barber had PROBABLY held him up) Example 2: "...cet homme portait une valise. Personne ne sait ce qui se trouvait à l'intérieur: PEUT-ÊTRE une arme à feu".
THE DEVELOPMENT OF ARGUMENTATIVE DISCOURSE
259
(...this man was carrying a suitcase. Nobody knows what was in side: MAYBE a firearm) - verbs such as SEMBLER or PARAÎTRE (seem) can be used: Example 1: "Les deux hommes ne me SEMBLENT pas suspects" (The two men don't SEEM suspicious to me) Example 2: "Ce comportement PARAIT normal chez un fou". (This behavior SEEMS normal for a madman.) - verbs or expressions with epistemic value, e.g. SUPPOSER QUE, (SUPPOSE, GUESS) AVOIR L'IMPRESSION QUE (IMAGINE) can be used:
Example 1: "J'AI PLUTÔT L'IMPRESSION QUE c'est le second". (I RATHER IMAGINE it's the second.) Example 2: "Cet homme était légèrement essouflé ce qui TENTERAIT à faire croire qu'il craignait d'être poursuivi". (That man was slightly out of breath which WOULD TEMPT us to think that he was afraid of being followed.") - certainty is generally expressed through the use of an adverb: Example 1: "Cet essoufflement était CERTAINEMENT dû au nombre de kilomètres qu'il devait avoir parcouru pour arriver à la gare" (His being out of breath was CERTAINLY due to the number of kilometers he had to go to get to the station.) Example 2: "Il portait un uniforme, SÛREMENT celui du gardien" (He wore a uniform, SURELY the guard's) - or through the use of an epistemic verb or expression: Example 1: "Je pouvais ÊTRE certain qu'il n'était pas le coupable". (I could BE CERTAIN he wasn't the guilty one.) Example 2: "Toutes les preuves nous DÉMONTRENT bien qu'il est coupable". (All the evidence clearly PROVES that he is guilty.) Table 3 gives the percentage of the types of modalization that occurred in each group.
260
G. PIÉRAUT-LE BONNIEC AND M. VALETTE
Table 3 SL.l
SL.2
SL.3
N=17
N = 45
N = 351
Veibal transformations
53%
40%
46%
Expressions of possibility
18%
29%
34%
Expressions of probability
11%
22%
14%
Expressions of certainty
17%
8%
5%
Percentage of types of modalisation (Number of occurrences). 1. Cf. Table 2.
It should be noticed that here too, it is in the SL.2 group that the most modal ex pressions are found. But we have already seen that this could be explained by the fact that it is the age group where the subjects are the most explicit in their ar gumentation. The most interesting fact to note here is the considerable increase in the use of modal expressions between SL.l and SL.2, and the increase in expres sions of uncertainty as compared to expressions of certainty.
Argumentation and Narration The three points that we have just discussed, i.e., ways the clues could be used, the expression of logical relations and modalization, are the criteria that were used to evaluate each of the texts as to the roles played by narration (narrative mode) and argumentation (argumentative mode). The Narrative Mode. Narration can take on a number of forms: - The subject repeats the information that was given in the instructions. Example 1: "Parmi les 3 hommes j'en ai soupçonné un: il est grand et mince, ses vêtements sont de bonne qualité mais son pantalon est court, ses yeux sont rouges, et il tripotait le tisonnier du poële."
THE DEVELOPMENT OF ARGUMENTATIVE DISCOURSE
261
(Among the 3 men I suspected one: he is tall and thin, his clothes are of good quality but his pants are short, his eyes are red, and he played around with the poker.) - The subject tells about the enquiry. Instead of working on the clues that have been given, the subject gives a detailed account of an imaginary enquiry as carried out by the policeman and from that the subject invents a quantity of other clues that will be used or not in the argumentation. Example 2: "Je me rendis à l' asile et je demandai une autosie (sic) du gar dien tué. Je l'obtins et j'appris qu'il avait été tué par balle. J'interrogeai les autres gardiens le directeur et les fous. Les gar diens et le directeur n'avaient rien entendu pendant la nuit mais leur réponse était malheureusement négative. J'interrogeai dans les fous et l'un d'eux qui était sans doute le plus lucide m'affirma avoir entendu un coup de feu mais il se refusa à d'autres com mentaires". (I went to the asylum and asked for the autopsy of the guard that was killed. I got it and I learn that he had been killed by a bullet. I question the other guards the director and the mad people. The guards and the director hadn't heard anything during the night but their response, unfortunately, was negative. I questioned the mad people and one of them who was probably the most lucid told me that he had heard a gun go off but he refused to say anything else.) - The subject reconstructs the crime. Example 3: "Voilà ce que je pense des circonstances du crime. Le fou attend son gardien derrière la porte. Dès que celui-ci entre dans la cel lule du fou il l'attaque par derrière. Il veut juste l'assommer mais il est très fort et ne voit pas très bien (il louche) il lui casse les vertèbres du cou; puis il se sauve va à la gare et prend le train de 17h30 pour Rouen ou celui de 17h55 pour Paris." (This is what I think of the circumstances of the crime. The mad man waits for his guard behind the door. As soon as he enters the madman's cell he attacks him from behind. He just wants to
262
G. PIÉRAUT-LE BONNIEC AND M. VALETTE
knock him out but he is very strong and can't see very well (he's cross-eyed) he breaks two of the vertebrae in his neck; then he runs off to the station and takes the 5:30 train to Rouen or the 5:55 to Paris.) Example 4: Il s'était échappé de sa cellule après avoir neutralisé son gar dien; il s'empara de son pistolet et le tua d'une balle dans le coeur puis il escalada le mur, s'échappa de l'asile se rendit à la gare sans avoir pu retiré son uniforme et sans s'être débarrassé de son pistolet" (He escaped from his cell after neutralizing his guard; he took his pistol and killed him with a bullet in his heart and then climbed over the wall, escaped from the asylum went to the station without being able to take off his uniform and without getting rid of his pistol) It should be noted that all of these examples are in an assertive mode, i.e. the events are reported like facts whose truth is generally accepted. It should also be noted, on the other hand, that while there are numerous temporal relations (DÈS QUE (AS SOON AS), PUIS (THEN), APRÈS QUE (AFTER), ET (AND) + TIME VALUE), there are no words that introduce logical articulations with the exception of MAIS (BUT). However, this conjunction in an adversative sense is not used here with an argumentative value; it is only part of the story. Thus in the third example it intro duces an explanation of how the murder took place, but it is not used to justify the subject's choice of suspect. The clues that are used (the man is strong and he is cross-eyed (example 3), the uniform, the pistol (example 4)) are integrated into the telling of the crime and do not serve to prove the guilt of the 3rd man who is not the object of discussion. The 3rd man has been 'chosen' as the guilty one; the jus tification of this is simply to tell the facts. The Argumentative Mode. Let's consider the following example: Example 1: "Je pense que le premier homme est le fou que la police recher che. EN EFFET, certains détails portent à croire que c'est lui. Il était le seul à être essoufflé c'est certainement DÛ AU FAIT QUE cet homme a dû courir pour échapper à toute poursuite. Il portait des lunettes d'écaille AFIN DE passer inaperçu, POUR QUE per sonne ne puisse le reconnaître. Seule les lunettes d'écaille ne sont pas très discrètes et sont très rares. Il est le seul à s'être justifié
THE DEVELOPMENT OF ARGUMENTATIVE DISCOURSE
263
EN EFFET il a déclaré qu'il sortait de chez le coiffeur; MAIS pour quoi s'est-il embarassé d'une valise pour y aller. Il aurait très bien pu aller chez le coiffeur la veille, dans sä fuite il s'est embarrase d'une valise, il ne faut pas oublier qu'il s'agit d'un fou. ET, pensant passer inaperçu avec une valise, des lunettes il s'est trahi". (I think the first man is the madman that the police are looking for. Indeed certain details lead us to believe that it is him. He was the only one out of breath; this is certainly DUE TO THE FACT THAT this man had to run to escape from everyone chasing him. He wore tortoise-shell glasses so that nobody would notice him. Except that tortoise-shell glasses are not very discreet and are very rare, he was the only one with an excuse; Indeed he stated that he had come from the barber's. BUT why was he car rying a suitcase to go there. He could have gone to the barber's the night before, and he took the suitcase along with him to es cape; we mustn't forget that we're talking about a madman. AND, expecting not to be noticed with a suitcase and glasses he betrayed himself.) First of all it can be seen that the meaning lent to the clues are the object of dis cussion: - out of breath: the man is the only one to be out of breath; explanation: he had to run if he didn't want to be caught. - tortoise-shell glasses: 1st possible meaning: try to pass unnoticed, but tortoiseshell glasses are noticeable, especially when they do not go with the rest of the outfit. Thus there is incoherence, which can, however, be found acceptable in the case of a madman. - say that he has just come from the barber's: he is the only one to explain where he had been before coming to the station. This might mean that the man is trying to set up an alibi, especially since it would have been more usual to go to the barber's the day before a trip. - the suitcase: it is burdensome for someone who is trying to flee, but this is the case of a madman, so the contradiction disappears. Secondly, the text includes a good number of logical articulations: EN EFFET (IN DEED), DÛ AU FAIT QUE (DUE TO THE FACT THAT), POUR (TO), AFIN DE (IN ORDER TO), MAIS (BUT), ET (AND) which serves to introduce the conclusion.
264
G. PIÉRAUT-LE BONNIEC AND M. VALETTE
Finally, there are a number of modal forms: JE PENSE QUE (I THINK), PORTER À CROIRE (WOULD MAKE US BELIEVE), CERTAINEMENT (CERTAINLY), IL A DÛ COURIR (HE MUST HAVE RUN), IL AURAIT TRÈS BIEN PU (HE COULD VERY WELL HAVE), IL NE FAUT PAS OUBLIER QUE (IT MUSTN'T BE FORGOTTEN THAT). For these reasons it would seem justified to consider this text to be in an ar gumentative mode. But there is argumentation that is supported, not by the information that was fur nished in the instructions, but by imaginary information, especially some concern ing the invented police inquiry. Example 2: "Voilà pourquoi je l'ai inculpé. Parce que un homme ni grand ni petit était venu rendre visite à son frère qui est dans l' asile de St Guy sur Bièvre et cet homme que j'inculpe avait déjà été dévêtu. Alors j'ai pensé que l'homme que j'inculpe avait déjà préparé sa fuite et qu'il avait déshabillé l'homme et ça explique pourquoi son pantalon est trop petit, et aussi ça explique pourquoi il tripote le tisonnier (il l'a tué avec le tisonnier)". (This is why I accused him. Because a man who was neither tall nor short came to visit his brother when he was in the St. Guy Asylum in Bièvre and the man that I accuse was already un dressed. So I thought that the man that I arrest had already prepared his escape and had undressed the man and that explains why his pants are too small, and it also explains why he played around with the poker (he killed him with the poker)). Argumentation may include narration as in the above example, which offers a hypothetical reconstruction of the crime. This is completely different from the case where subjects tell about the crime as if they were an eye-witness. Example 3. "Je vais d'un pas à l'asile. Je cherche sa cellule et je lui demande de m'expliquer son plan. (J'AVAIS DES HYPOTHÈSES mais il faut les VÉRIFIER) "Je me suis taillé un pistolet dans du bois et j'ai assommé le gardien mais il a riposté et il m'a blessé au bras. Après je suis allé à la gare mais je me suis rappelé les copains de l'asile et j'ai voulu y retourner. Voilà l'histoire". (I go quickly to the asylum. I look for his cell and I tell him to explain his plan to me. (I HAD SOME IDEAS but I must CHECK) "I
THE DEVELOPMENT OF ARGUMENTATIVE DISCOURSE
265
carved a piece of wood into a gun and I knocked out the guard, but he hit back and hurt my arm. Then I went to the station but I remembered my mates in the asylum and wanted to go back. That's the story.) Example 4. "Le troisième n'a aucun alibi. Il portait un pantalon court. Ce fait qui POURRAIT SEMBLER inintéressant, est très important; dans sa fuite, il AURA CHERCHER des vêtements et ne trouvant rien à sa taille, il AURA PRIS les premiers venus. J'accuse donc le deuxième suspect pour le meurtre du gardien." (The third man has no alibi. He was wearing short pants. This fact, which MIGHT SEEM unimportant, is very important in his es cape. He WILL HAVE LOOKED for some clothes and finding noth ing his size, he WILL HAVE TAKEN the first ones that came along. I THEREFORE accuse the second suspect of the murder of the guard.) Most often then narratives can be differentiated from argumentative texts through the presence or absence of modalities. All the texts included a certain amount of these two textual modes and only a very small proportion of any text is not in either one of these modes. Examples of this are polite forms which the policeman uses when addressing the inspector, or some of the advice the inspector gives in order to catch the murderer. The children's texts were all re-transcribed onto measured graph paper to study the number of half-lines devoted to textual modes. This, in turn, allowed for a cal culation of the proportion of each of these modes in each text. Table 4 gives the mean percentage for the textual modes for each school level. It can be seen that more than half (58%) of a SL.l text is made up of telling how the police enquiry took place, i.e. in adding a number of imaginary clues to those that were given in the instructions. On the average of 11% of the text is simply a repeti tion of the clues that were already given and 12% is used to recount the crime, while only 18% of the text is used to defend the writer's choice of the culprit. On the other hand, the proportion of the average SL.2 text devoted to argumentation for or against one of the suspects is 63%. It should be noted, too, that this ar gumentation is based mainly on the clues given in the instructions and not on clues invented by the subject. Three-quarters of the average SL.3 text is made up of ar gumentation.
266
G. PIÉRAUT-LE BONNIEC AND M. VALETTE
Conclusion This study has shown that organizing argumentative discourse presents real dif ficulties. Adolescents who use their native language everyday and who can solve problems in various situations demanding a lot of reasoning skill, are not necessari ly capable of going beyond narration when setting up argumentative discourse. There are several reasons for this. First of all is the fact that we believe more easily those things we see, or what is said by someone who has supposedly seen them, than what is supposedly demonstrated verbally. A story is convincing because, normally, it does not leave room for listeners to question its validity. In many courtrooms, only the witnessing of a crime and its telling or of its admission are proof of a defendant's guilt. Secondly, narratives seem to be the most spontaneous type of linguistic activity and an assertive mode of language is easier to use than modalities and logical con nectives. Even if argumentative texts always include some anecdotes and narration
Table 4
Mean Proportion of Each of the Textual Modes for the Three Groups Groups
Textual Modes Repetition of information Story of the enquiry Story of the crime
SL.l
SL.2
SL.3
11%
13%
5%
s = 5.29
s = 7.9
s = 6.9
58%
17%
19%
s = 17.5
s = 15.2
s = 10.3
12%
6%
2%
s = 12
s = 7.9
s = 3.4
Argumentation
18%
63%
73%
• given clues
2%
54%
58%
s = 4.2
s = 25.3
s = 20
16.5%
9%
15%
s = 15.6
s = 5.6
s = 19
• invented clues
THE DEVELOPMENT OF ARGUMENTATIVE DISCOURSE
267
which are used to illustrate one's reasoning, this in itself is not sufficient to set up a sound argumentation which is a form of discourse that presents more constraints than storytelling. Judging from our subjects' behavior, the main difficulty from which stem all the others, resides in the way information is used. It has already been mentioned that one of the differences between argumentative reasoning and formal reasoning is that the objects of the former are never defined once and for all and that in this case there is always a place for new information. Perhaps this freedom is difficult to handle, as argumentation means knowing just when to stop giving information. On the one hand, too much information unfocuses interest and might seem to be a digression and is not convincing. On the other hand argumenta tive reasoning demands a lot of leeway as concerns the interpretation and sig nificance of the objects of discourse, even if it is only to avoid the alocutor's objections. Being able to anticipate these objections is to be able to refute them in advance, e.g. being out of breath might mean that the man was afraid of being fol lowed by the police, but this can be countered as it can also mean that he is afraid of missing his train. Thus, argumentation implies being able to think in both a metacognitive and a metalinguistic framework. It means accepting that no given is obvious and that it is not sufficient to say something, for it to be true for an inter locutor. It is thus necessary to be able to situate ideas in the realms of the possible, the probable and the disputable and to find ways to present them as plausible and convincing. And if the representation of the world that the speaker constructs is to be convincing, the objects of discourse must not only be carefully chosen, but the relations that are set up among them must be stronger and "more logical" than simple juxtaposition (which corresponds perfectly well to the narrative form) would be. The construction of argumentative discourse is probably one of the most subtle and most elaborate ways to use language.
NOTES 1. The texts that are quoted here are exactly as they appeared in the children's work and no cor rection whatever has been made of any grammatical, spelling, punctuation, etc. mistakes.
References
Ainsworth, M.D.S. 1979 Attachment as Related to Mother-Infant Interaction. In J.S. Rosenblatt, R.A. Hinde, CG. Beer and M.-C. Busnel (eds.), Advances in the Study of Behavior, vol. 9. 2-52. New York: Academic Press. Aitken, P.G. 1981
Cortical Control of Conditioned and Spontaneous Vocal Behavior in Rhesus Monkeys. Brain and Language 13. 171-184.
Aitken, P.G. and W.A. Wilson, Jr. 1979 Discriminative Vocal Conditioning in Rhesus Monkeys: Evidence for Volitional Control? Brain and Language 8. 227-240. Allen, G.D. 1972
Timing Control in Speech Production: Some Theoretical and Methodological Is sues. Language Centre's Occasional Papers 13. 170-201.
Allen, G.D. and S. Hawkins 1980 Phonological Rhythm: Definition and Development. In G.H. Yeni-Komshian, S.F. Kavanagh and C.A. Ferguson (eds.), Child Phonology, vol. 1: Production. 227-256. New York: Academic Press. Amidon, A.O. 1976
Children's Understanding of Sentences with Contingent Relations: Why are Temporal and Conditional Connectives so Difficult? Journal of Experimental Child Psychology 4. 22, 423-437.
Amidon, A.O. and P. Carey 1972 Why Five-Year-Olds Cannot Understand Before and After? Journal of Verbal Learning and Verbal Behavior 11. 417-423. Anand, K.J.S. and P.R. Hickey 1987 Pain and its Effects in the Human Neonate and Fetus. New England Journal of Medicine 317. 1321-1329.
270 Ando, Y. and H. Hattori 1971 Effects of Intense Noise During Fetal Life upon Postnatal Adaptability (Statisti cal Study of the Reactions of Babies to Aircraft Noise). Journal of the Acousti cal Society of America 47. 1128-1130. Ando, Y. and H. Hattori 1977 Effects of Noise on Sleep of Babies. Journal of the Acoustical Society of America 62. 1. André-Thomas, A.S. Locomotion from Prenatal Life. London: Spastic Society, Heineman. 1966 Annett, M. 1978
Annett, M. 1985
Antoine, G. 1958/1962
A Single-Gene Explanation of Right- and Left-Handedness and Brainedness. Coventry: Lanchester Polytechnic.
Left, Right, Hand and Brain: The Rightshift Theory. Hillsdale, NJ: Lawrence Erlbaum Associates.
La coordination en français, 1 and 2. Paris: d'Artrey.
Anscombre, J.C. and O. Ducrot 1977 Deux "mais" en français? Lingua 43. 24-40. Anscombre, J.C. and O. Ducrot 1983 L'argumentation dans la langue. Brussels: Mardaga. Applebee, A.N. 1978 The Child's Concept of Story. Chicago: University of Chicago Press. Ardery, G. 1979
The Development of Coordinates in Child Language. Journal of Verbal Learning and Verbal Behavior 18. 745-756.
Ardery, G. 1980
On Coordination in Child Language. Journal of Child Language 1. 305-320.
Arena, J.M. 1970
Contamination of Ideal Food. Nutrition Today 5. 2-8.
Aristotle 1938/1983
The Categories on Interpretation. H.P. Cooke (ed.). Cambridge, MA: Harvard University Press.
REFERENCES Aristotle 1938/1983 Aristotle 1960/1976
Aristotle 1960/1976
271
Prior Analytics. H. Tredennick (ed.). Cambridge, MA: Harvard University Press.
Posterior Analytics. H. Tredennick (ed.). Cambridge, MA: Harvard University Press.
Topica. E.S. Forster (ed.). Cambridge, MA: Harvard University Press.
Armitage, S.E., B.A. Baldwin and M.A. Vince 1980 The Fetal Sound Environment of the Sheep. Science 208. 1173-1174. Arnauld, A. and P. Nicole 1662/1965 La logique ou l'art de penser. P. Clair and F. Girbal (eds.). Paris: PUF. Asbach-Schnitker, B. 1978 Die adversativen Konnektoren aber, sondern und but nach negierten Sätzen. In H. Weydt (ed.), Die Partikeln der deutschen Sprache. 457-468. Berlin: De Gruyter. Aslin, R.N., D.B. Pisoni and P.W. Jusczyk 1983 Auditory Development and Speech Perception in Infancy. In M.M. Haith and J.J. Campos (eds.), Handbook of Child Psychology, vol. II: Infancy and Developmen tal Psychobiology. 573-687. New York: Wiley and Sons. Atkinson, M. 1979
Auchlin, A. 1981
Bacri, N. 1984a
Bacri, N. 1984b
Prerequisites for Reference. In E. Ochs and B.B. Schieffelin (eds.), Developmen tal Pragmatics. New York: Academic Press.
Réflexions sur les marqueurs de structuration de la conversation. Études de lin guistique appliquée 44. 88-103.
Pitch and Timing Cues in Speech Intelligibility: The Cause of Child Language. In A. Cohen and M.P.R. vd. Broecke (eds.), Proceedings of the Tenth Interna tional Congress of Phonetic Sciences. 589-594. Dordrecht: Foros Publications.
L'intelligibilité du langage enfantin: intonation et compréhension de la parole. In M. Moscato and G. Piéraut-Le Bonniec (eds.), Le langage: construction et ac tualisation. 81-99. Rouen: PUR.
272 Bacri, N. 1986
Bacri, N. 1987a
Bacri, N. 1987b
Fonctions de l' intonation dans V organisation perceptive de la parole. Doctoral Thesis. Université de Paris VIII. Miméo.
Perceptual Spaces and the Identification of Natural and Synthetic Sentences. Proceedings of the XIth International Congress of Phonetic Sciences, vol. 2. 219-222. Estonia, USSR: Tallinn.
Contrôle de la temporalité dans la production du langage chez le jeune enfant entendant et malentendant. Bulletin d'audiophonologie (6). 639-658.
Bacri, N. and B. de Boysson-Bardies 1981 Babillage ou prélangage? Bulletin de l'association internationale de linguistique appliquée 2. Bain, D., J.-P. Bronckart, Davaud, A. Pasquier and B. Schneuwly 1982 Classification de textes pour la rédaction et l'analyse. Pratiques et théorie 28. 74. Bakhtine, M. 1929/1977
Le Marxisme et la philosophie du langage. Paris: Ed. de Minuit.
Ballaire, M.-F. 1982
The Narrative Event. Roneotyped Paper. Laboratoire de Psychologie de Poitiers.
Bally, 1932
Linguistique générale et linguistique française. Berne: Francke.
Bamberg, M. 1985
Bamberg, M. 1986
Form and Function in the Construction of Narratives: Developmental Perspec tives. PhD. dissertation. University of California, Berkeley.
A Functional Approach to the Acquisition of Anaphoric Relationships. Linguis tics 24. 227-284.
Barbault, M.C., O. Ducrot, J. Dufour, J. Espagnon, Israel and D. Manesse 1975 Car, parce que, puisque. Revue romane X-2. 248-280. Barden, T.P., P. Peltzman and J.T. Graham 1968 Human Fetal Electroencephalographic Response to Intrauterine Acoustic Signals. American Journal of Obstetrical Gynecology. 100. 1128-1134.
REFERENCES
273
Barrie-Blackley, S. 1973 Six-Year-Old Children's Understanding of Sentences Adjoined with Time Ad verbs. Journal of Psycholinguistic Research 2. 153-165. Bassano, D. and C. Champaud 1987 La fonction argumentative des marques de la langue. Argumentation 1. 175-201. Bates, E. 1976
Language and Context. The Acquisition of Pragmatics. New York: Academic Press.
Beaugrande, R. de 1980 Text, Discourse and Process. Norwood: Ablex Publishing Co.. Beck, C.H.M. and R.L. Barton 1972 Deviation and Laterality of Hand Preference in Monkeys. Cortex 8. 339-363. Beecher, M.D., M.R. Petersen, S.T. Zoloth, D.B. Moody and Stebbins 1979 Perception of Conspecific Vocalizations by Japanese Macaques. Brain, Behavior and Evolution 16. 443-460. Bellugi, U. 1967
Bench, R.J. 1968
Benkirane, T. 1982
Benveniste, E. 1965
A Transformational Analysis of the Development of Negation. Doctoral disserta tion. Cambridge, MA: Harvard University.
Sound Transmission to the Human Fetus through the Maternal Abdominal Wall. Journal of Genetic Psychology. 113. 1172-1174.
Étude phonétique et fonctions de la syllabe en arabe marocain. Doctoral Thesis. Université de Aix en Provence.
Le langage et l'expérience humaine. Diogène 51. 3-13.
Bereiter, . and M. Scardamalia 1982 From Conversation to Composition: The Role of Instruction in a Developmental Process. In R. Glaser (ed.), Advances in Instructional Psychology, vol. 2. Hillsdale, NJ: Lawrence Erlbaum Associates. Berko, J. 1958
The Child's Learning of English Morphology. Word 14. 150-177.
274 Bernstein-Ratner, N. and A. Luberoff 1984 Cues to Post-Vocalic Voicing in Mother-Child Speech. Journal of Phonetics 12. 285-289. Berry, D.C. and D.E. Broadbent 1984 On the Relationship between Task Performance and Associated Verbalisable Knowledge. Quarterly Journal of Experimental Psychology 36A. 209-231, Berthoud-Papandropoulou, I. 1978 An Experimental Study of Children's Ideas about Language. In A. Sinclair, R.J. Jarvella and W.J.M. Levelt (eds.), The Child Conception of Language. Berlin: Springer Verlag. Berthoud-Papandropoulou, I. 1980 La réflexion métalinguistique chez l'enfant. Geneva: Imprimerie Nationale. Berthoud, I. and Othenin-Girard 1983 L'étude génétique des subordonnées propositions méthologiques. In J.-P. Bronckart, M. Kail and G. Noizet (eds.), Psycholinguistique de l''enfant 189190. Neuchâtel and Paris: Delachaux et Niestlé. Bertoncini, J. 1982
Quelques aspects de la perception de la parole chez V enfant de 0 à 2 mois. Doctoral dissertation. Paris: École des Hautes Etudes en Sciences Sociales.
Bertoncini, J., R. Bijeljac-Babic, S.E. Blumstein and J. Mehler 1987 Discrimination in Neonates of Very Short CVs. Journal of the Acoustical Society of America 82. 31-37. Bertoncini, J., R. Bijeljac-Babic, P.W. Jusczyk, LJ. Kennedy and J. Mehler 1988 An Investigation of Young Infants' Perceptual Representation of Speech Sounds. Journal of Experimental Psychology: General 117. 21-33. Bertoncini, J. and J. Mehler 1981a Quelques remarques à propos de la perception chez le nouveau-né. Foetologie 31. 387-396. Bertoncini, J. and J. Mehler 1981b Syllables as Units in Infant Speech Perception. Infant, Behavior and Develop ment 4. 271-284. Biasci, C. 1982
Konnektive in Sätzen und Texten. Hamburg: Buske.
REFERENCES
275
Bickely, , . Lindblom and L. Roug 1986 Acoustic Measures of Rhythm in Infant's Babbling or "All God's Children Got Rhythm". Toronto: International Congress on Acoustics. Birnholz, J.C. and B.B. Benacerraf 1983 The Development of the Human Fetal Hearing. Science 222. 516-518. Black, J.B., J.T. Turner and G.H. Bower 1979 Point of View in Narrative Comprehension, Memory, and Production. Journal of Verbal Learning and Verbal Behavior 18. 187-198. Bloom, L., M. Lahey, L. Hood, K. Lifter and K. Fiess 1980 Complex Sentences: Acquisition of Syntactic Connectives and the Semantic Relations they Encode. Journal of Child Language 7. 235-261. Blumstein, S.E. and K.N. Stevens 1980 Perceptual Invariance and Onset Spectra for Stop Consonants in Different Vowel Environments. Journal of the Acoustical Society of America 67. 648-662. Boesch, C. and H.Boesch 1981 Sex Differences in the Use of Natural Hammers by Wild Chimpanzees: A Preliminary Report. Journal of Human Evolution 10. 585-593. Bolinger, D. 1977
Pronouns and Repeated Nouns. Indiana University Linguistic Club.
Borel, M J., J.B. Grize and D. Mieville 1983 Essai de logique naturelle. Berne: Peter Lang. Botvin, G.J. and B. Sutton-Smith 1977 The Development of Structural Complexity in Children's Fantasy Narratives. Developmental Psychology 13. 377-388. Bowerman, M. 1979
The Acquisition of Complex Sentences. In P. Fletcher and M. Garman (eds.), Language Acquisition. 285-305. Cambridge: Cambridge University Press.
Bowlby, T.G.R. 1969 Attachment and Loss, vol. 1: Attachment. London: Hogarth. Boysson-Bardies, . de 1976 Négation et performance linguistique. Paris and La Haye: Mouton. Boysson-Bardies, B. de, N. Bacri, L. Sagart and M. Poizat 1981 Timing in Late Babbling. Journal of Child Language 8. 525-539.
276 Boysson-Bardies, . de, L. Sagart and C. Durand 1984 Discernible Differences in the Babbling of Infants According to Target-Lan guage. Journal of Child Language 11. 1-15. Boysson-Bardies, B. de, L. Sagart, P. Hallé and C. Durand 1986 Acoustic Investigation of Cross-Linguistic Variability in Babbling. In B. Lindblom and R. Zetterstrom (eds.), Precursors of Early Speech. New York: Stockton Press. Boysson-Bardies, B. de, L. Sagart and P. Hallé 1987 The Influence of Target-Language on Vocalic Space in Ten-Month-Old Infants. Tallinn: Proceedings of 11th International Congress of Phonetic Science. August 1-7. Boysson-Bardies, B. de, P. Hallé, L. Sagart and C. Durand 1989 A Cross-Linguistic Investigation of Vowel Formants in Babbling. Journal of Child Language 16. 1-17. Braine, M.D.S. 1963
Brainerd, 1978
Brandt, P.Y. 1988
On Learning the Grammatical Order of Words. Psychological Review 70. 323348.
The Stage Acquisition in Cognitive Developmental Theory. The Behavioral and Brain Sciences 2. 173-213.
Contribution à l'étude de l'argumentation négative chez l'enfant de 7 à 12 ans. Doctoral dissertation. University of Geneva.
Brazelton, T.B. 1978 The Remarquable Talents of the Newborn. Birth and the Family Journal 5. 4-10. Brazelton, T.B., B. Koslowski and M. Main 1974 The Origins of Reciprocity in Mother-Infant Interaction. In M. Lewis and A. Rosenbloom (eds.), The Effect of the Infant on its Caregiver. 49-76. New York: Wiley and Sons. Brennan, M. and J. Kirkland 1982 Classification of Infant Cries Using Descriptive Scales. Infant Behavior Development 5. 341-346. Brésard, B. 1988
and
Primates Cognition of Space and Shapes. In L. Weiskrantz (ed.), Thought without Language. 356-415. Oxford: Oxford University Press.
REFERENCES
277
Brésard, . and F. Bresson 1983 Handedness in Pongo pygmæus and Pan troglodytes. Journal of Human Evolu tion 12. 659-666. Brésard, . and F. Bresson 1987 Reaching or Manipulation: Left or Right? Open Peer Commentary of MacNeilage, Studdert-Kennedy and Lindblom: Primates Handedness Reconsidered. Behavioral and Brain Sciences 10. 265-266. Bresson, F. 1974
Bresson, F. 1977
Remarks on Genetic Psycholinguistics: The Acquisition of the Article System in French. In F. Bresson and J. Mehler (eds.), Current Problems in Psycholinguis tics. 67-72. Paris: CNRS.
Semantics, Syntax and Utterance: Determining a Referent. International of Psycholinguistics 4.31-41.
Journal
Bresson, F., L. Maury, G. Piéraut-Le Bonniec and S. de Schonen 1977 Organization and Lateralization of Reaching in Infants: An Instance of Dissym metric Functions in Hand Collaboration. Neuro-Psychologia 15. 311-320. Bresson, F. and S. de Schonen 1977 À propos de la construction de l'espace et de l'objet: la prise d'un objet sur un support. Bulletin de psychologie 30. 151-158. Bresson, F., S. de Schonen and Tzortzis 1972 Étude des perturbations dans des performances logico-arithmétiques chez des sujets atteints de diverses lésions cérabrales. Langage 7. 108-132. Bril, B. 1986
The Acquisition of an Everyday Technical Motor Skill: The Pounding of Cereals in Mali (Africa). In M.G. Wade and H.T.A. Whiting (eds.), Themes in Motor Development. 315-326. Dordrecht: Martinus Nijhoff.
Broadbent, D.E., P. FitzGerald and M.H.P Broadbent 1986 Implicit and Explicit Knowledge in the Control of Complex Systems. British Journal of Psychology 11. 33-50. Broca, P. 1861
Remarques sur le siège de la faculté du langage articulé, suivies d'une observa tion d'aphémie. Bulletin de la société d'anatomie de Paris 6, 2ème série. 398407.
278 Broca, P. 1865
Sur le siège de la faculté du langage articulé. Bulletin de la société d'anthropologie de Paris 6. 337-393.
Bronckart, J.-P. 1976 Genèse et organisation des formes verbales chez l'enfant. Brussels: Dessart et Mardaga. Bronckart, J.-P. 1977 Théories du langage. Brussels: Mardaga. Bronckart, J.-P. 1983 Cas et prédicats. Une étude de psychologie du langage. La Linguistique 19. 2953. Bronckart, J.-P. 1985 Les opérations temporelles dans deux types de textes d'enfant. In J. Wittwer (eds.). Bulletin de psychologie, Numéro spécial "La psycholinguistique textuelle". Bronckart, J.-P., M. Kail and G. Noizet 1983 Psycholinguistique de l'enfant. Neuchâtel and Paris: Delachaux et Niestlé. Bronckart, J.-P., M. Gennari and G. de Weck 1981 La compréhension des phrases simples. La perspective représentative et la perspective communicative. International Journal of Psycholinguistics 8 (2). 529. Bronckart, J.-P. and B. Schneuwly 1981 Une approche totalitaire de la langue. Paris: Colloque "Le langage de l'enfant en milieu scolaire; problématiques et méthodes". December. Bronckart, J.-P. and B. Schneuwly 1984 La production des organisateurs textuels chez l'enfant. In M. Moscato and G. Piéraut-Le Bonniec (eds.), Le langage, construction et actualisation. Rouen: Presses Universitaires. Bronckart, J.-P. et al. 1985 Le fonctionnement des discours. Neuchâtel: Delachaux et Niestlé. Brown, A.L. 1975
Recognition, Reconstruction and Recall of Narrative Sequences by Preoperational Children. Child Development 46. 156-166.
Brown, A.L. and L.A. French 1976 Construction and Regeneration of Logical Sequences Using Causes or Conse quences as the Point of Departure. Child Development 47. 930-940.
REFERENCES
279
Brown, F. 1981
À chacun son meurtre. Paris: Clancier-Guénaud.
Brown, R.W. 1973
A First Language: The Early Stages. Cambridge, MA: Harvard University Press.
Brown, R. and C. Hanlon 1970 Derivational Complexity and Order of Acquisition in Child Speech. In J. Hayes (ed.), Cognition and Development of Language. 11-53. New York: Wiley and Sons. Bruner, J.S. 1964
The Course of Cognitive Growth. American Psychologist 19. 1-15.
Bruner, J.S. 1983
Children's Talk: Learning How to Use Language. New York: Norton.
Bruxelles, S., . Ducrot, E. Fouquier, J. Gouazé, G. Dos Reis Nunes and A. Remis 1976 "Mais" occupe-toi d'Amélie. Actes de la recherche en sciences sociales 6. 4762. Brunot, F. 1926
La pensée et la langue. Paris: Masson.
Bull, D., R.E. Eilers and D.K. Oller 1985 Infant's Discrimination of Final Syllable Fundamental Frequency in Multisyllabic Stimuli. Journal of the Acoustical Society of America 11 (1). 289-295. Busnel, M.-C. 1979
Mesures intravaginales du niveau et des distorsions acoustiques de bruits mater nels. Électrodiagnostic therapie 16. 142.
Busnel, M.-C. and C. Granier-Deferre 1983 And What of Fetal Audition? In A. Oliveirio and M. Zapella (eds.), The Behavior of Human Infants. New York: Life Science Coll., Plenum Press. Butterworth, B. 1980 Evidence from Pauses in Speech. In B. Butterworth (ed.), Language Production, I: Speech and Talk. London: Academic Press. Butterworth, G. and E. Cochran 1980 Towards a Mechanism of Joint Visual Attention in Human Infancy. International Journal of Behavioral Development 3. 253-272.
280 Butterworth, G. and L. Grover 1988 The Origins of Referential Communication. In L. Weiskrantz (ed.), Thought without Language. Oxford: Oxford University Press. Cabrejo-Parra, E. 1977 Quelques aspects comparatifs de l'acquisition du langage. In J.-P. Bronckart et al., La genèse de la parole. 259-267. Paris: P.U.F. Cairns, H.S., W. Cowart and A.D. Jablon 1981 Effects of Prior Context upon the Integration of Lexical Information During Sen tence Processing. Journal of Verbal Learning and Verbal Behavior 20. 445-453. Cairns, R.B. 1979
Social Development: The Origins and Plasticity of Interchanges. San Francisco: Freeman.
Caramazza, A., E. Grober and Garvey 1977 Comprehension of Anaphoric Pronouns. Journal of Verbal Learning and Verbal Behavior 16. 601-609. Caramazza, A. and S. Gupta 1979 The Roles of Topicalization, Parallel Function and Verb Semantics in the Inter pretation of Pronouns. Linguistics 17. 497-519. Caron, J. 1979
Caron, J. 1983 Caron, J. 1987
Carroll, L. 1871
Chafe, W.L. 1980
La compréhension d'un connecteur polysémique: la conjonction "si". Bulletin de psychologie 32. 791-801.
Les régulations du discours. Paris: PUF.
Pour une approche psycho-linguistique de l'argumentation. In G. Piéraut-Le Bonniec (ed.), Connaître et le dire. Brussels: Mardaga.
Through the Looking-Glass and What Alice Found There. London: Tenniel. In M. Gardner (1960), The Annotated Alice. New York: Bramball House.
The Deployment of Consciousness in the Production of a Narrative. In W.L. Chafe (ed.), The Pear Stories. 9-50. Norwood: Ablex Publishing Co..
REFERENCES Champaud, . 1983
281
Les énoncés conditionnels: une approche diachronique. In J.-P. Bronckart, M. Kail and G. Noizet (eds.), Psycholinguistique de l'enfant. 221-243. Neuchâtel and Paris: Delachaux et Niestlé.
Champaud, C. and C. Jakubowicz 1979 Situation hypothétique et conditions de production des énoncés avec "si": étude génétique. Bulletin de psychologie 32. 773-790. Charney, R. 1978
The Development of Personal Pronouns. Doctoral dissertation. University of Chicago.
Chistovich, L.A. 1983 Relation between Speech Production and Speech Perception. Utrecht: Abstracts of the Tenth International Congress of Phonetic Science. 15-19. 1-6 August. Chistovich, L.A., D.M. Lissenko and N.A. Fedorova 1972 Speech Production Control under Syncronisation with Periodically Presented Speech Stimulus. Sensornye Systemy 2. 56-85. Chiat, S. 1983
Chomsky, N. 1968
Why Mikey's Right and My Key's Wrong: The Significance of Stress and Word Boundaries in a Child's Output System. Cognition 14. 275-300.
Language and Mind. New York: Harcourt, Brace and World, Inc. Chicago: Harcourt.
Chomsky, N. 1975
Reflections on Language. New York: Pantheon Books.
Chorazina, H. 1976
Shrifts in Laterality in a Baby Chimpanzee. Nueropsychologia 14. 381-384.
Chugani, H.T. and M.E. Phelps 1986 Maturational Changes in Cerebral Function in Infants Determined by 18FDG Positron Emission Tomography. Science 231. 840-843. Cirilo, R.K. and D.J. Foss 1980 Text Structure and Reading Time for Sentences. Journal of Verbal Learning and Verbal Behavior 19. 96-109. Clancy, P., T. Jacobson and J. Silva 1976 The Acquisition of Conjunction: A Cross-Linguistic Study. Papers and Reports on Child Language Development Stanford 12. 71-80.
, R.A. 1978
Clark, E.V. 1971
Clark, E.V. 1972
Clark, E.V. 1973
Clark, E.V. 1978
Clark, E.V. 1982
Clark, E.V. 1983
Clark, E.V. 1985
The Transition from Action to Gesture. In A. Lock (ed.), Action Gesture and Symbol. The Emergence of Language. 231-257. London, New York and San Francisco: Academic Press.
On the Acquisition of the Meaning of Before and After. Journal of Verbal Learning and Verbal Befiavior 10. 266-275.
On the Child's Acquisition of Antonyms into Semantic Fields. Journal of Verbal Learning and Verbal Beliavior 11. 750-758.
Non-Linguistic Strategies and the Acquisition of Word Meanings. Cognition 2. 161-182.
Awareness of Language: Some Evidence from What Children Say and Do. In A. Sinclair, R.J. Jarvella and WJ.M. Levelt (eds.), The Child's Conception of Lan guage. Berlin: Springer Verlag.
The Young Word Maker: A Case Study of Innovation in the Child's Lexicon. In E. Wanner and L.R. Gleitman (eds.), Language Acquisition. Cambridge: Cambridge University Press.
Meaning and Concepts. In P.H. Mussen (ed.), Carmichael's Manual of Child Psychology, vol. 3: Cognitive Development (edited by J.H. Flavell and E.M. Markman). 787-840. New York: Wiley and Sons.
Acquisition of Romance, with Special Reference to French. In D. Slobin (ed.), The Cross-linguistic Study of Language Acquisition. 687-782. Hillsdale, NJ: Lawrence Erlbaum Associates.
Clark, E.V. and B.F. Hecht 1983 Comprehension, Production and Language Acquisition. Annual Review of Psychology 34. 325-349. Clarkson, M.G. and R.K. Clifton 1984 Pitch Extraction and Categorization by 7 Month Olds. Paper presented at the Fourth Biennial International Conference on Infant Studies, New York City,
REFERENCES
283
Clements, M. and J.B. Kelly 1978 Auditory Spatial Responses of Young Guinea Pigs (Cavia Procellus) During and After Ear Blocking. Journal of Comparative Physiological Psychology, 92. 3444. Clopton, B.M. 1986
Neural Correlates of Development and Plasticity. In W.T. Greenough and J.M. Juraska (eds.), Developmental Neuropsychology. Academic Press.
Clopton, B.M. and J.A. Winfield 1976 Effects of Early Exposure to Patterned Sound on Unit Activity in Rat Inferior Colliculus. /. Neurophysiol. 39. 1081-1089. Coker, P.L. 1978
Semantic and Syntactic Factors in the Acquisition of Before and After. Journal of Child Language 5. 261-277.
Conlee, J.W. and T.N. Parks 1981 Age and Position Dependent Effects of Monaural Acoustic Deprivation in Nucleus Magnocellularis of the Chicken. Journal of Comparative Neurology. 202. 373-374. Connoly, K. and J. Elliott 1972 The Evolution and Ontogeny of Hand Function. In N. Blurton Jones (ed.), Etiological Studies of Child Behaviour. 329-383. Cambridge: Cambridge Univer sity Press. Costerman, J. and D. Giurgea 1988 L'influence du sens sur la segmentation syllabique chez des enfants de trois à sept ans. Archives de psychologie 56 (217). 137-149. Cox, J.R. and R.A. Griggs 1982 The Effects of Experience on Performance in Wason's Selection Task. Memory and Cognition 10. 496-502. Cromer, R.F. 1976
Crompton, A. 1980 Cruttenden, A. 1970
Developmental Strategies for Learning. In V. Hamilton and M.D. Vernon (eds.), The Development of Cognitive Processes. London and New York: Academic Press.
Timing Patterns in French. Phonetica 37. 205-234.
A Phonetic Study of Late Babbling. British Journal of Disorders of Communica tion 5. 110-118.
284 Cruttenden, A. 1986 Crystal, D. 1986
Culioli, A. 1968 Culioli, A. 1976
Intonation. Cambridge: Cambridge University Press.
Prosodic Development. In P. Reicher and M. Garman (eds.), Language Acquisi tion. 174-197. Cambridge: Cambridge University Press.
La formalisation en linguistique. Cahiers pour l'analyse 9. 106-117.
Recherche en linguistique. Théorie des opérations énonciatives. Doctoral Semi nar (roneo.). Paris: Université de Paris VII.
Cutting, J.E. and P.D. Eimas 1975 Phonetic Feature Analyzers and the Processing of Speech in Infants. In J.F. Kavanagh and J.E. Cutting (eds.), The Role of Speech in Language. 127-148. MIT Press. Cutting, J.E. and B.S. Rosner 1974 Categories and Boundaries in Speech and Music. Perception and 16. 564-570.
Psychophysics
Damasio, A., U. Bellugi, H. Damasio, H. Poizner and J. Van Gilder 1986 Sign Language Aphasia During Left-Hemisphere Amytal Injection. Nature 322. 363-365. Davey, D.A., J. Dommisse, M. MacNab and D. Dacre 1984 The Value of an Auditory Stimulatory Test in Antenatal Fetal Cardiotocography. European Journal of Obstetrical Gynecology Reproductive Biology. 18. 273-277. DeCasper, A.J. and W.P. Fifer 1980 Of Human Bonding: Newborns Prefer their Mother's Voice. Science 208. 11741176. DeCasper, A.J. and P.A. Prescott 1984 Human Newborns' Perception of Male Voices: Preference Discrimination, and Reinforcing Value. Developmental Psychobiology. 17. 481-491. DeCasper, A.J. and M.J. Spence 1986 Prenatal Maternal Speech Influences Newborn's Perception of Speech Sounds. Infant Behavioral Development. 9. 133-150. Dehaene, S., J.-P. Changeux and J.-P. Nadal 1987 Neural Networks that Learn Temporal Sequences by Selection. Proceedings of the National Academy of Sciences, USA 84. 2727-2733.
REFERENCES
285
Delack, J.B. 1975 Prosodic Features of Infant Speech: The First Year of Life. Leeds: Proceedings of the Eighth International Congress of Phonetic Sciences. Dell, G.S., G. McKoon, R. Ratcliff 1983 The Activation of Antecedent Information During the Processing of Anaphoric Reference in Reading. Journal of Verbal Learning and Verbal Behavior 22. 121132. Denhière, G. 1982
Denhière, G. 1984 Derlon, P. 1981
Dewson, J.H. 1977
Schémas? Vous avez dit... schémas? Bulletin de psychologie "Langage et compréhension". 717-731.
35, Special Issue
// était une fois... Lille: PUL
Enfantement... enchantements (chez les gitans). In E. Herbinet and M.-C. Busnel (eds.), L'aube des sens, cahiers du nouveau-né. 381-384. Paris: Stock.
Preliminary Evidence of Hemispheric Asymmetry of Auditory Function in Monkeys. In S. Harnad, R.W. Doty, L. Goldstein, J. Jaynes and G. Krauthamer (eds.), Lateralization in the Nervous System. 63-71. New York: Academic Press.
Dewson, J.H., A. Cowey and L. Weiskrantz 1970 Disruptions of Auditory Sequences Discrimination by Unilateral and Bilateral Ablations of Superior Gyrus in the Monkey. Experimental Neurology 28. 529548. DiCristo, A. 1978
De la microprosodie à l'intonosyntaxe. Doctoral thesis. Université de Provence.
Divon, M.Y., L.D. Platt, C.J. Cantrell, C.V. Smith, S.Y. Yeh and R.H. Paul 1985 Evoked Fetal Startle Response: A Possible Intrauterine-Uterine Neurological Ex amination. American Journal of Obstetrical Gynecology. 153. 454-456. Dolitsky, M. 1984
Under the Tumtumtree: from Non-Sense to Sense. A Study in Non-Automatic Comprehension. Amsterdam: John Benjamins Publishing Company.
Dore, J., M.B. Franklin, R.T. Miller and A.L.H. Ramer 1976 Transitional Phenomena in Early Language Acquisition. Journal of Child Lan guage 3. 13-28.
286 Ducrot, O. 1972
Dire et ne pas dire. Paris: Herman.
Ducrot, O. 1973
Le dit et le dire. Paris: Minuit.
Ducrot, O. et al. 1980 Les mots du discours. Paris: Minuit. Dwornicka, B. A. Jasienska, W. Smolarz and R. Wawryk 1964 Attempt of Determining the Fetal Reaction to Acoustic Stimulation. Acta-OtoLaryngol (Stock.) 57. 571-574. Ebbinghaus, H. 1885 Über das Gedächtnis. Leipzig: Von Veit. Ehrlich, S. 1982
Ehrlich, S. 1984
Construction d'une représentation de texte et fonctionnement de la mémoire sémantique. Bulletin de psychologie 35 (356). 659-671.
Construction des représentations sémantiques: le fonctionnement du sujet. In M. Moscato and G. Piéraut-Le Bonniec, Langage: construction et actualisation. Rouen: PU Rouen.
Ehrlich, S., G. Bramaud du Boucheron and A. Florin 1978 Le développement des connaissances lexicales à l'école élémentaire. Paris: PUF. Ehrlich, S. and A. Florin 1981 Niveaux de compréhension et production d'un récit par des enfants de 3 à 11 ans. Bordeaux: Éditions du CNRS. Eilers, R.E., D.H. Bull, D.K. Oiler and D.C. Lewis 1984 The Discrimination of Vowel Duration by Infants. Journal of the Acoustical Society of America 75. 213-218. Eilers, R.E., W. Gavin and W.R. Wilson 1979 Linguistic Experience and Phonemic Perception in Infancy: A Cross-Linguistic Study. Child Development 50. 14-18. Eilers, R.E., W.R. Wilson and J.M. Moore 1977 Developmental Changes in Speech Discrimination in Infants. Journal of Speech and Hearing Research 20. 766-780.
REFERENCES Eimas, P.D. 1974
Eimas, P.D. 1975a
Eimas, P.D. 1975b
287
Auditory and Linguistic Processing of Cues for Place of Articulation by Infants. Perception and Psychophysics 16. 513-521.
Speech Perception in Early Infancy. In L.B. Cohen and P. Salapatek (eds.), In fant Perception. 193-231. New York: Academic Press.
Auditory and Phonetic Coding of the Cues for Speech: Discrimination of the R1 Distinction by Young Infants. Perception and Psychophysics 18. 341-347.
Eimas, P.D. and J.D. Corbit 1973 Selective Adaptation of Linguistic Feature Detectors. Cognitive Psychology 4. 99-109. Eimas, P.D. and J.L. Miller 1980a Discrimination of Information for Manner of Articulation. Infant Behavior and Development 3. 367-375. Eimas, P.D. and J.L. Miller 1980b Contextual Effects in Infant Speech Perception. Science 209. 1140-1141. Eimas, P.D., E.R. Siqueland, P.W. Jusczyk and J. Vigorito 1971 Speech Perception in Infants. Science 171. 303-306. Eisenberg, A.R. 1980 A Syntactic, Semantic, Pragmatic Analysis of Conjunctions. Papers and Reports on Child Language Development 19. 70-78. Eisenberg, A.R. 1987 Learning to Argue with Parents and Peers. Argumentation 1(2). 113-125. Eisenberg, A.R. and Garvey 1981 Children's Use of Verbal Strategies in Resolving Conflicts. Discourse Processes 4. 149-170. Eisenberg, R.B. 1976 Auditory Competence in Early Life: The Roots of Communicative Behavior. Bal timore: University Park Press. Entus, A.K. 1977
Hemispheric Asymmetry in Processing of Dichotically Presented Speech and Nonspeech Stimuli by Infants. In S.J. Segalowitz and F.A. Gruber (eds.), Language Development and Neurological Theory. 63-73. New York: Academic Press.
288 Esperet, E. 1981
Esperet, E. 1982
Esperet, E. 1984
Esperet, E. 1987
Genèse des conduites langagières: propositions d'analyse pour la conduite de récit. In Équipe récit, Technical Report no. 1: Genèse de la conduite de récit. 125. Poitiers: Laboratoire de Psychologie. Roneotyped paper.
Genesis of Language Behaviours and Acquisition Situations: Application to Storytelling. In F. Lowenthal, F. Vandamme and J. Cordier (eds.), Language and Language Acquisition. 309-313. New York: Plenum Press.
Processus de production: genèse et rôle du schéma narratif dans la conduite de récit. In M. Moscato and G. Piéraut-Le Bonniec, Langage: construction et ac tualisation. Rouen: PU Rouen.
Micro and Macrostructural Planning and Control in Production: Approaches in Storytelling Situation. Paper delivered at the 2nd European Conference for Re search on Learning and Instruction. Tübingen, September.
Esperet, E., P. Coirier, D. Coquin and J.M. Passerault 1987 L'implication du locuteur dans son discours: discours argumentatifs formel et naturel. Argumentation 1(2). 155-174. Esperet, E. and M. Fayol 1982 Éléments pour un modèle génétique de la conduite de récit. Poitiers: Laboratoire de psychologie. Dijon: Laboratoire de psychologie. Ronéotyped paper. Esperet, E. and D. Gaonac'h 1981 What does "Storytelling" Mean for Children: Narrative Schema Representation and Storytelling at Different Ages. Paper presented at the International Sym posium on Text Processing. Fribourg, September. Esperet, E. and D. Gaonac'h 1986 The Role of Narrative Schema on Story Production and Recall: A Longitudinal Study. Paper presented at the 2nd European Conference on Developmental Psychology (ISSBD). Rome, September. Esperet, E. and A. Piolat 1988 Production: Planning and Control. In G. Denhière and J.-P. Rossi (eds.), Text and Text Processing. Amsterdam: North Holland (in press). Evans, M., N. Taylor and I. Blum 1979 Children's Written Language Awareness and its Relation to Reading Acquisition. Journal of Reading Behavior 11. 7-19.
REFERENCES
Falk, D. 1983
Cerebral Cortices of East African Early Hominids. Science 221. 1072-1074.
Falmagne, R J . (ed.) 1975 Reasoning: Representation Wiley and Sons. Fant, G. 1960 Fayol, M. 1980
Fayol, M. 1981
Fayol, M. 1982a
Fayol, M. 1982b
Fayol, M. 1983a
Fayol, M. 1983b
Fayol, M. 1984
Fayol, M. 1985a
Fayol, M. 1985b
289
and Process in Children and Adults. New york:
Acoustic Theory of Speech Production. La Haye: Mouton.
À propos de la clôture de récits. Éléments pour une approche expérimentale du problème. Psychologie française 25. 139-147.
L'organisation du récit écrit chez l'enfant. Son évolution de 6 à 10 ans. Doc toral Thesis. Université de Bordeaux II.
Le plus-que-parfait. Étude génétique en compréhension et production chez l'enfant de 4 à 10 ans. Archives de psychologie 50. 261-283.
Le récit écrit d'expérience personnelle, son évolution chez l'enfant de 6 à 10 ans. Bulletin de psychologie 35. 261-265.
L'acquisition du récit: un bilan des recherches. Revue française de pédagogie 62. 65-82.
L'emploi des temps verbaux chez l'enfant, l'adolescent et l'adulte. Problèmes théoriques et méthodologiques. Paris: Colloque International de Didactique et Pédagogie du Français (Langue Maternelle). December.
La distanciation dans le langage. L'exemple du calcul de l'origine dans le récit écrit d'expérience personnelle chez l'enfant de 6 à 10 ans. Enfance 1. 5-19.
L'emploi des temps verbaux dans les récits écrits. Études chez l'enfant, l'adulte et l'adolescent. In J. Wittwer (ed.), Bulletin de psychologie, Numéro spécial, "La psycholinguistique textuelle".
Le récit et sa construction. Neuchâtel: Delachaux et Niestlé.
290 Fayol, M. and S. Mouchon 1988 L'emploi des connecteurs dans les descriptions orales d'événements; étude expérimentale chez l'enfant de 4 à 8 ans. Paper presented at the meeting Ac quisition du Langage et Développement Cognitif. Rouen, March. Feagans, L. 1982
Feijoo, J. 1981
The Development and Importance of Narratives for School Adaptation. In L. Feagans and D.E. Farran (eds.), The Language of Children Reared in Poverty. 95-116. New York: Academic Press.
Le Fetus, Pierre et le loup. In E. Herbinet and M.-C. Busnel (eds.), L'aube des sens. Cahiers du Nouveau-né 5. 192-209. Paris: Stock.
Ferguson, CA. 1979 Phonology as an Individual Access System: Some Data from Language Acquisi tion. In C.I. Fillmore, D. Kempler and W.S. Wang (eds.), Individual Differences in Language Ability and Language Behavior. New York: Academic Press. Ferguson, CA. 1986 Discovering Sound Units and Constructing Sound Systems: It's Child's Play. In J.S. Perkell and D.H. Klatt (eds.), Invariance and Variability in Speech Proces ses. 36-51. Hillsdale, NJ: Lawrence Erlbaum Associates. Ferguson, CA. and M.A. Macken 1982 Phonological Development in Children. In K.E. Nelson (ed.), Language Develop ment TV. New York: Gardner Press. Ferreiro, E. 1971 Fifer, W.P. 1981
Les relations temporelles dans le langage de l'enfant. Paris: Droz.
Early Attachment: Maternal Voice Preference in One- and Three-Day-Old In fants. PhD. dissertation. North Carolina: University of Greensboro.
Fillenbaum, S. 1974
Or: Some Uses. Journal of Experimental Psychology 103. 913-921.
Fillenbaum, S. 1975
If: Some Uses. Psychological Research 37. 245-260.
Fillenbaum, S. 1977
Mind Your P's and Q's: The Role of Content and Context in Some Uses of "and", "or" and "if. In G.H. Bower (ed.), The Psychology of Learning and Motivation. 41-98. New York: Academic Press.
REFERENCES
291
Fillol, F. and J. Mouchon 1977 Les éléments organisateurs du récit oral. Pratiques 17. 100-127. Fischler, I. and P.A. Bloom 1979 Automatic and Attentional Processes in the Effects of Sentence Contexts on Word Recognition. Journal of Verbal Learning and Verbal Behavior 18. 1-20. Fischler, I. and P.A. Bloom 1980 Rapid Processing of the Meaning of Sentences. Memory and Cognition 8. 216225. Fisher, P.M. and H. Mandl 1982 Metacognitive Regulation of Text Processing: Aspects and Problems Concerning the Relation between self-Statements and Actual Performance. In A. Rammer and W. Kintsch (eds.), Discourse Processing. 339-351. New York: North Hol land Publishing Co. Flavell, J.H. 1981
Cognitive Monitoring. In W.P. Dickson (ed.), Children's Oral Communication Skills. New York: Academic Press.
Flores d'Arcáis, G.B. 1978 Levels of Semantic Knowledge in Children's Use of Connectives. In A. Sinclair, R.J. Jarvella and W.J.M. Levelt (eds.), The Child's Conception of Language. 133-153. Berlin, Heidelberg and New York: Springer Verlag. Flores d'Arcáis, G.B. 1981 The Acquisition of Meaning of the Connectives. In W. Deutsch (ed.), The Child's Construction of Language. New York: Academic Press. Fodor, J.A., M.F. Garrett and S.L. Brill 1975 PI/KA/PU: The Perception of Speech Sounds by Prelinguistic Infants. Perception and Psychophysics 18. 74-78. Fok, A. 1979
The Freqnency of Occurrence of Speech Sounds and Tones in Cantonese. In R. Lord (ed.), Hong Kong Language Papers. Hong Kong: Hong Kong University Press.
Forbes, H.S. and H.B. Forbes 1927 Fetal Sense Reaction: Hearing. Journal of Comparative Physiological Psychol ogy. 7. 353-355.
292 Forster, K.I. 1981
Priming and the Effects of Sentence and Lexical Contexts on Naming Time: Evidence for Autonomous Lexical Processing. Quarterly Journal of Experimental Psychology 33A. 465-495.
Foss, D.J. and M.A. Blank 1980 Identifying the Speech Code. Cognitive Psychology 12. 1-31. Fowler, CA., M.R. Smith and L.G. Tassinary 1986 Perception of Syllable Timing by Prebabbling Infants. Journal of the Acoustical Society of America 79(3). 814-825. Franckel, JJ. 1976
Frazier, L. 1978
Introduction à l'étude de l'organisation et de la genèse du système aspectuel en français. Paris: EHESS.
On Comprehending Sentences: Syntactic Parsing Strategies. Unpublished doc toral dissertation. University of Connecticut.
Frazier, L. and J.D. Fodor 1978 The Sausage Machine: A New Two-Stage Parsing Model. Cognition 6. 291-325. Frazier, L. and K. Rayner 1982 Making and Correcting Errors During Sentence Comprehension: Eye Movements in the Analysis of Structurally Ambiguous Sentences. Cognitive Psychology 14. 178-120. Freedle, R. and R.P. Duran 1979 Sociolinguistic Approaches to Dialogue with suggested Applications to Cognitive Science. In R.D. Freedle (ed.), New Directions in Discourse Processing, vol. II. Norwood: Ablex Publishing Co. French, L.A. 1981
French, L.A. 1986
But of Course Preschoolers Understand the Meaning of but! Paper presented at the 6th Annual Boston University Conference on Language Development. Oc tober.
Acquiring and Using Words to Express Logical Relationships. In S.A. Kuczaj, II and M.D. Barrett (eds.), The Development of Word Meaning. Berlin, Heidelberg and New York: Springer Verlag.
French, L.A. and A.L. Brown 1977 Comprehension of Before and After in Logical and Arbitrary Sequences. Journal of Child Language 4. 247-256.
REFERENCES
293
French, L.A. and K. Nelson 1985 Young Children s Knowledge of Relational Terms. New York: Springer Verlag. Frost, G.T. 1980
Fry, D.B. 1947
Fuchs, . 1982
Tool Behavior and the Origins of Laterality. Journal of Human Evolution 9. 447-459.
The Frequency of Occurrence of Speech Sounds in Southern English. Archives néerlandaises de phonetique expérimentale. 103-106.
La paraphrase. Paris: PUF.
Fuchs, . and A.M. Leonard 1979 Vers une théorie des aspects. Les systèmes du français et de l' anglais. Paris and The Hague: Mouton and EHESS. Galaburda, A.M., J. Corsiglia, G.D. Rosen and G.F. Sherman 1987 Planum Temporale Asymmetry, Reappraisal since Geschwind and Levitsky. Neuropsychologia 25. 853-868. Galaburda, A.M., M. LeMay, T.L. Kemper and N. Teschwind 1978 Right-Left Asymmetries in the Brain. Science 199. 852-856. Galaburda, A.M. and D.N. Pandya 1982 Role of Architectronics and Connections in the Study of Primate Brain Evolu tion. In E. Armstrong and D. Falk (eds.), Primate Brain Evolution: Methods and Concepts. 203-216. New York: Plenum Press. Gallivan, J. 1982
Children's Understanding of BUT. Evidence from Spontaneous Production. Paper presented at the Annual Meeting of the Canadian Psychological Association. Montreal.
Gaonac'h, D. and E. Esperet 1985 Fonctions des temps verbaux dans la production de récits libres: évolution génétique entre 4 et 8 ans. Bulletin de psychologie 38, Special Issue "La psycholinguistique textuelle". 705-716.
Gardner, M. 1960 Gamica, O.K. 1974
The Annotated Alice. New York: Bramball House
Acoustical Characteristics of Speech to Young Children. Saint Louis, MO: Proceedings of the 88th Meeting of the Acoustical Society of America.
294 Garvey, G. and A. Caramazza 1974 Implicit Causality in Verbs. Linguistic Inquiry 5. 459-464. Garvey, G., A. Caramazza and J. Yates 1974 Factors Influencing Assignment of Pronoun Antecedents. Cognition 3. 227-243. Garvey, C. and A.R. Eisenberg 1986 Logical Patterns in Child-Mother Conversation. Toronto: Biennial meeting of the Society for Research in Child Development. Gérard, J. 1981
La maîtrise du principe de récursivité langagière chez l'enfant. Archives de psychologie 49. 95-114.
Geschwind, N. and P. Behan 1982 Left-Handedness: Association with Immune Disease, Migraine and Developmen tal Learning Disorders. Proceeding of the National Academy of Sciences 79. 5097-510. Geschwind, N. and A.M. Galaburda 1985/1987 Cerebral Lateralization. Biological Mecfianisms, Associations, and Pathology. Cambridge, MA: MIT Press. Geschwind, N. and W. Levitsky 1968 Human Brain: Left-Right Asymmetries in Temporal Speech Region. Science 161. 186-189. Ginsburg, HJ., S. Fling, M.L. Hope, D. Musgrove and Andrews 1979 Maternal Holding Preferences: A Consequence of Newborn Head-Turning Response. Child Development 50. 280-281. Givon, T. 1976
Topic, Pronoun, and Grammatical Agreement. In C.N. Li (ed.), Subject and Topic. Academic Press.
Givon, T. (ed.) 1983 Topic Continuity in Discourse. Amsterdam: John Benjamins Publishing Co. Glenn, 1978
The Role of Episodic Structure and of Story Length in Children's Recall of Simple Stories. Journal of Verbal Learning and Verbal Behavior 17. 224-247.
Gleitman, L.R., H. Gleitman and E.F. Shipley 1972 The Emergence of Child as Grammarian. Cognition 1. 137-164.
295
REFERENCES
Glick, S.D. and D.A. Ross 1981 Right-Sided Population Bias and Lateralization of Activity in Normal Rats. Brain Research 205. 222-225. Goodlin, R.C. and W. Schmidt 1972 Human Fetal Arousal Levels as Indicated by Heart Rate Recording. American Journal of Obstetrical Gynecology. 114. 613-621. Goldman-Eisler, F. 1972 Pauses, Clauses, Sentences. Language and Speech 15. 103-113. Gombert, J.E. 1986
Le développement des activités métalinguistiques chez l'enfant: le point de la recherche. Études de linguistique appliquée 62.
Granier-Deferre, G, J.-P. Lecanuet, H. Cohen and M.-G Busnel 1985 Feasibility of Prenatal Hearing Test. Acta-Oto-Laryngol 93-101.
(Stock.), suppl., 421.
Green, J.A. and E.K. Adkins 1975 The Effect of Prenatal Auditory Stimulation on Early Vocalization and Approach Behavior in the Japanese Quail (Coturnix Coturnix Japonica). Behaviour L.II, 2. 145-155. Greenfield, P. and J. Smith 1976 The Structure of Communication Academic Press.
in Early Language Development. New York:
Grevisse, M. 1936/1959
Le bon usage. Seventh edition. Gembloux: Duculot.
Grize, J.B. 1982
De la logique à l'argumentation. Geneva: Droz
Grize, J.B. and B. Matalon 1962 Introduction à une étude expérimentale et formelle du raisonnement naturelle. In J. Piaget (ed.), Études d'épistémologie génétique, XVI. Paris: PUF. Grize, J.B. and G. Piéraut-Le Bonniec 1983 La contradiction. Essai sur les opérations de la pensée. Paris: PUF. Grober, E., W. Beardsley and A. Caramazza 1978 Parallel Function Strategy in Pronoun Assignment. Cognition 6. 117-133.
296 Grober, E. and L. Kellar 1981 Semantic Influences on Pronoun Assignment in Aphasia. Applied tics 2. 253-268.
Psycholinguis-
Grosjean, F. and A. Deschamps 1972 Analyse des variables temporelles du français spontané. Phonetica 26. 129-157. Gruber, J. 1967
Topicalization in Child Language. Foundations of Language 3. 37-65.
Grunig, B.N. 1977
Bilan sur le statut de la coordination. DRLAV 15. 46-77.
Gulich, E. 1970
Makrosyntax Fink.
der Gliederungssignale
im Gesprochenen
Französisch.
Munich:
Guyomarc'h, J.-C. 1974 L'empreinte auditive prénatale chez le poussin domestique. Revue of Comparative Animalry. 8. 3-6. Haberlandt, K. 1982
Les expectations du lecteur dans la compréhension du texte. Bulletin psychologie 35 (356). 733-740.
de
Haberlandt, ., Berian and J. Sandson 1980 The Episode Schema in Story Processing. Journal of Verbal Learning and Ver bal Behavior 19. 635-650. Hakes, D. 1980
The Development Verlag.
of Metalinguistic Abilities in Children. New York: Springer
Hakes, D.T., J.S. Evans and W.E. Tunmer 1980 The Development of Metalinguistic Abilities in Children. Berlin: Springer Verlag. Hall, N.A. 1976
Children's Awareness of Segmentation in Speech and Print. Reading 10. 11-19.
Halliday, M.A.K. 1964 The Linguistic Study of Literary Texts. Proceedings Congress of the Linguists.
of the IXth International
Halliday, M.A.K. 1970 Language Structure and Language Function. In J. Lyons (ed.), New Horizons in Linguistics. Harmondsworth: Penguin Books.
REFERENCES
297
Halliday, M.A.K. and R. Hasan 1976 Cohesion in English. London: Longmans Group Limited. Hammond, J. 1970
Hamer, L. 1976
Hamer, L. 1980
Hashimoto, T. 1980
Hearing and Response in the Newborn. Developmental Medicine in Child Neurology. 12. 3.
Children's Understanding of Linguistic Reference to Past and Future. Journal of Psycholinguistic Research 5. 65-84.
Comprehension of Past and Future Revisited. Journal of Experimental Child Psychology 29. 170-182.
Information Processing of Speech Sounds in the Medical Geniculate and the In ferior Colliculus. Proceedings of the Japanese Academy 36 B. 294-299.
Hayes, J.R. and L.S. Flower 1980 Identifying the Organization of Writing Processes. In L.W. Gredd and E.R. Steinberg (eds.), Cognitive Processes in Writing. 3-30. Hillsdale, NJ: Lawrence Erlbaum Associates. Hayes, N.A. and D.E. Broadbent 1988 Two Modes of Learning for Interactive Tasks. Cognition 28. 249-276. Heffher, H.E. and R.S. Heffner 1984 Temporal Lobe Lesions and Perception of Species-Specific Vocalizations by Macaques. Science 226. 75-76. Henry, K.R. 1983
Abnormal Auditory Development Resulting from Exposure to Autotoxic, Chemi cal, Noise and Auditory Restriction. In R. Romand (ed.), Development of Auditory and Vestibular Systems. Academic Press.
Henschall, W.R. 1972 Intrauterine Sound Levels. Journal of Obstetrical Gynecology. 112. 577. Henton, C.C. 1983
Hickmann, M. 1982
Changes in Vowels of Received Pronunciation. Journal of Phonetics 11. 353371.
The Development of Narrative Skills: Pragmatic and Metapragmatic Aspects of Discourse Cohesion. Doctoral dissertation. University of Chicago.
298 Hickmann, M. 1984
Hickmann, M. 1986
Hickmann, M. 1987a
Hickmann, M. 1987b
Hickmann, M. 1988
Contexte et fonction dans le développement du langage. In M. Deleau (ed.), Langage et communication à l'âge préscolaire. Presses Universitaires de Rennes 2.
Psychosocial Aspects of Language Acquisition. In Fletcher and Garman (eds.), Language Acquisition (2nd edition). Cambridge: Cambridge University Press.
Pragmatics in Child Language: Some Issues in Developmental Theory. In M. Hickmann (ed.), Social and Functional Approaches to Language and Thought. Orlando, FL: Academic Press.
Clause Structure Variation in Narrative Discourse: A Functional and Cross-Lin guistic Approach. Paper presented at the Conference on Clause Structure in Language Acquisition. Nijmegen: -Planck Institute for Psycholinguistics. November.
Topics and Subjects in Discourse: An Analysis of French Children's Narratives. Paper presented at the Third European Conference on Developmental Psychology. Budapest, 15-19 June.
Hirst, W. and G.A. Brill 1980 Contextual Aspects of Pronoun Assignment. Journal of Verbal Learning and Verbal Behavior 19. 168-175. Holloway, R.J. and de Lacoste 1982 Brain Endocast Asymmetry in Pongids and Hominids: The Paleontology of Cerebral Hemisphere Dominance. American Journal of Physical Anthropology 58. 101-110. Holmgren, K., B. Lindblom, G. Aurelius, B. Jailing and R. Zetterstrom 1986 On the Phonetics of Infant Vocalizations. In B. Lindblom and R. Zetterstrom (eds.), Precursors of Early Speech. Basingstoke: MacMillan. Holmes, V.M. 1979
Hood, L. 1977
Accessing Ambiguous Words During Sentence Comprehension. Quarterly Jour nal of Experimental Psychology 31. 569-589.
A Longitudinal Study of the Development of the Expression of Causal Relation in Complex Sentences. Doctoral dissertation. Columbia University.
299
REFERENCES
Hood, L. and L. Bloom 1979 What, When and How about Why: A Longitudinal Study of Early Expressions of Causality. Monographs of the Society of Research in Child Development 44. Homer, K., J. Servières and Granier-Deferre 1987 Deoxyglucose Demonstration of in utero Hearing in the Guinea-Pig Fetus. Hear ing Research. 26. 327-333. Hudson, R. 1985
Do Newborn Rabbits Learn the Odor Stimuli Releasing Nipple-Search Be haviour? Developmental Psychobiology. 18. 575-585.
Huggins, A.W.F. 1978 Speech Timing and Intelligibility. In J. Requ (ed.), Attention and VIL Hillsdale: Lawrence Erlbaum Associates. Hurt, S.J. and C. Hurt 1970 Direct Observation Thomas.
and Measurement
of Behaviour.
Performance,
Springfield, IL: C.C.
Huttenlocher, J. 1964 Children's Language: Word-Phrase Relationship. Science 143. 264-265. Huxley, R. 1970
Hyman, L. 1977
The Development of the Correct Use of Subject Personal Pronouns in Two Children. In G.B. Flores d'Arcáis and W.J.M. Levelt (eds.), Advances in Psycholinguistics. North Holland Publishing Co,
On the Nature of Linguistic Stress. In L. Hyman (ed.), Southern California Oc casional Papers in Linguistics.
Inhelder, B. and J. Piaget 1955 De la logique de l'enfant à la logique de l'adolescent. Paris: PUF. Jackendoff, R. 1987
Jakobson, R. 1941
Jakobson, R. 1968
On Beyond Zebra: The Relation of Linguistic and Visual Information. Cognition 26. 89-114.
Kindersprache, Wiksell.
Aphasie,
une allgemeine
Lautgesetze.
Uppsals: Almquist
Child Language, Aphasia, and Phonological Universals. The Hague: Morton.
and
300 Jakubowicz, . 1978 Fait actuel ou fait virtuel? La compréhension d'énoncés conditionnels chez l'enfant. L'année psychologique 78. 105-128. Jensen, O.H. 1984
Fetal Heart Rate Response to a Controlled Sound Stimulus as a Measure of Fetal Weil-Being. Acta Obstetrical Gynecology Scandanavia. 64. 97-101.
Johansson, ., E. Wedenberg and B. Westin 1964 Measurement of Tone Response by the Human Fetus. A Preliminary Report. Acta Otolaryngology, 57. 188-192. Johnson, H.L. 1975
The Meaning of Before and After for Preschool Children. Journal of Experimen tal Child Psychology 19. 88-99.
Johnson-Laird, P.N. 1983 Mental Models. Cambridge: Cambridge University Press. Jost, R.G., E.J. Quilligan, A. Yeh and G.G. Anderson 1972 Intrauterine Electroencephalogram of the Sheep Fetus. American Journal of Obstetrical Gynecology. 114. 535-539. Jusczyk, P.W. 1977
Jusczyk, P.W. 1981
Jusczyk, P.W. 1982
Jusczyk, P.W. 1985
Perception of Syllable-Final Stop Consonants by 2-Month-Old Infants. Percep tion and Psychophysics 21. 450-454.
Infant Speech Perception: A Critical Appraisal. In P.D. Eimas and J.L. Miller (eds.), Perspectives on the Study of Speech. Hillsdale, NJ: Lawrence Erlbaum Associates.
The Processing of Speech and Non Speech Sounds by Infants: Some Implica tions. In R.B. Aslin, J.R. Alberts and M.R. Peterson (eds.), Sensory and Percep tual Development: Influences of Genetic and Experiential Factors. New York: Academic Press.
On Characterizing the Development of Speech Perception. In J. Mehler and R. Fox (eds.), Neonate Cognition: Beyond the Blooming, Buzzing Confusion. Hillsdale, NJ: Lawrence Erlbaum Associates.
REFERENCES Jusczyk, P.W. 1986
301
Some Further Reflections on How Speech Perception Develops. In J. Perkell and D.H. Klatt (eds.), Invariance and Variability of Speech Processes. Hillsdale, NJ: Lawrence Erlbaum Associates.
Jusczyk, P.W., D.B. Pisoni, A. Walley and J. Murray 1980 Discrimination of Relative Onset Time of Two-Component Tones by Infants. Journal of the Acoustical Society of America 67. 262-270. Jusczyk, P.W., B. S. Rosner, J.E. Cutting, CF. Foard and L.B. Smith 1977 Categorical Perception of Nonspeech Sounds by 2-Month-Old Infants. Perception and Psychophysics 21. 50-54. Jusczyk, P.W. and E. Thompson 1978 Perception of a Phonetic Contrast in Multisyllabic Utterances by 2-Month-Old Infants. Perception and Psychophysics 32. 105-109. Kail, M. 1979a
Kail, M. 1979b Kail, M. 1980
Kail, M. 1983
Compréhension de "seul", "même" et "aussi" chez l'enfant. Bulletin de psychologie 32. 763-771.
Coréférence et thématisation. Année psychologique 79. 411-427.
Étude génétique des présupposés de certains morphèmes grammaticaux. Un ex emple: "mais". 55-64. Approches du langage. Paris: Éditions de la Sorbonne.
Le traitement des phrases complexes. In J.-P. Bronckart, M. Kail and G. Noizet (eds.), Psycholinguistique de l'enfant. 249-259. Neuchâtel and Paris: Delachaux et Niestlé.
Kail, M., M. Hickmann and N. Emmenecker 1987 Introduction des referents dans le récit: étude développementale des contraintes contextuelles. Paper presented at the conference Acquisition des Langues. Aixen-Provence, December. Kail, M. and R. Plas 1979 Psycholinguistique des présuppositions. Éléments pour une critique. Semantikos 3. 1-26. Kail, M. and J. Weissenborn 1984a A Developmental Cross-Linguistic Study of Adversative Connectives: French "mais" and German "aber"/"sondern". Journal of Child Language 11. 143-158.
302 Kail, M. and J. Weissenborn 1984b L'acquisition des connecteurs: critiques et perspectives. In M. Moscato and G. Piéraut-Le Bonniec, Le langage: construction et actualisation. Rouen: PU Rouen. Karmiloff-Smith, A. 1979 A Functional Approach to Child Language: A Study of Determiners and Refer ence. Cambridge: Cambridge University Press. Karmiloff-Smith, A. 1981 The Grammatical Marking of Thematic Structure in the Development of Lan guage Production. In W. Deutsch (ed.), The Child's Construction of Language. Academic Press. Karmiloff-Smith, A. 1986 From Meta-Processes to Conscious Access: Evidence from Children's Metalin guistic and Repair Data. Cognition 23. 95-147. Karmiloff-Smith, A. 1987 Function and Process in Comparing Language and Cognition. In M. Hickmann (ed.), Social and Functional Approaches to Language and Thought. Academic Press. Katz, E.W. and S.B. Brent 1968 Understand Connectives. Journal of Verbal Learning and Verbal Behavior 501-509 Kawai, M. 1965
Keenan, E. 1976
1.
Newly Acquired Precultural Behavior of Natural Troop of Japanese Monkeys on Koshima Islet. Primates 6. 1-30.
Towards a Universal Definition of "Subject". In C.N. Li (ed.), Subject and Topic. New York: Academic Press.
Keenan, E.O. and E, Klein 1975 Coherency in Children's Discourse. Journal of Psycholinguistic Research 4. Keenan, E.O. and B.B. Schieffelin 1976 Topic as a Discourse Notion: A Study of Topic in the Conversations of Children and Adults. In C.N. Li (ed.), Subject and Topic. New York: Academic Press. Keith, R.W. 1975
Middle Ear Function in Neonates. Archives of Oto-Laryngology.
101. 376-379.
REFERENCES
303
Keller-Cohen, D. 1987 Context and Strategy in Acquiring Temporal Connectives. Journal of Psycholinguistic Research 16. 165-183. Kernan, K.T. 1977
Kimball, J.P 1973
Kimura, D. 1961
Kimura, D. 1977
Kimura, D. 1979
Kimura, D. 1982
Semantic and Expressive Elaboraiton in Children's Narratives. In S. Ervin Tripp and J. Michell Kernan (eds.), Child Discourse. 91-104. New York: Academic Press.
Seven Principles of Surface Structure Parsing in Natural Language. Cognition 2. 15-47.
Cerebral Dominance and the Perception of Verbal Stimuli. Canadian Journal of Psychology 15. 166-171.
Acquisition of a Motor Skill After Left Hemisphere Damage. Brain 100. 527542.
Neuromotor Mechanisms in the Evolution of Human Communication. In H.D. Steklis and M.J. Raleigh, Neurobiology of Social Communication in Primates. New York: Academic Press.
Left-Hemisphere Control of Oral and Brachial Movements and their Relation to Communication. Philosophical Transaction of the Royal Society of London B298. 135-149.
Kintsch, W. and T.A. van Dijk 1978 Toward a Model of Text Comprehension and Production. Psychological 85. 363-394.
Review
Kitahara-Frisch, J. 1978 Stone Tools as Indicators of Linguistic Abilities in Early Man. The Annals of the Japan Association for Philosophy of Science 5. 101-109. Klatt, D.H. 1976
Kleiman, G.M. 1980
Linguistic Uses of Segmental Durations in English: Acoustic and Perceptual Evidence. Journal of the Acoustical Society of America 59. 1208-1221.
Sentence Frame Contexts and Lexical Decisions: Sentence-Acceptability and Word Relatedness Effects. Memory and Cognition 8. 336-344.
304 Klein, W. and C. Stutterheim 1985 Referenttal Movement. Unpublished manuscript. Knudsen, E.I., P.F. Knudsen and S.D. Esterly 1982 Early Auditory Experience Modifies Sound Localization in Barn Owls. Nature 295. 238-240. Konopczynski, G. 1984 Allongement final: contrainte physiologique ou comportement acquis? Travaux de l'Institut de Phonétique de Strasbourg 16. 143-156. Koopmans-van Beinurn, F J . and J.M. van der Stelt 1979 Early Stages in Infant Speech Development. Proceedings Phonetic Sciences. University of Amsterdam. Kuhl, P.K. 1976
Kuhl, P.K. 1979a
Kuhl, P.K. 1979b
Kuhl, P.K. 1979c
Kuhl, P.K. 1980
Kuhl, P.K. 1981
of the Institute of
Speech Perception in Early Infancy: The Acquisition of Speech-Sound Categories. In S.K. Hirsch, D.H. Eldredge, I.J. Hirsh and S.R. Silverman (eds.), Hearing and Davis: Essays Honoring Hollowell Davis. 265-280. Saint Louis: Washington University Press.
The Perception of Speech in Early Infancy. In N.J. Lass (ed.), Speech and Lan guage: Advances in Basic Research and Practice. New York: Academic Press.
Models and Mechanisms in Speech Perception. Species Comparisons Provide Further Contributions. Brain, Behavior and Evolution 16. 374-408.
Speech Perception in Early Infance: Perceptual Constance for Spectrally Dis similar Vowel Categories. Joural of the Acoustical Society of America 66. 16681679.
Perceptual Constance for Speech-Sound Categories in Early Infancy. In G.M. Yeni-Komshian, J.F. Kavanagh and C.A. Ferguson (eds.), Child Phonology, II: Perception. 41-66. New York: Academic Press.
Discrimination of Speech by Nonhuman Animals: Basic Auditory Sensitivities Conducive to the Perception of Speech-Sound Categories. Journal of the Acous tical Society of America 70. 340-348.
REFERENCES
305
Kuhl, P. and J.D. Miller 1975 Speech Perception by the Chinchilla: Voiced-Voiceless Distinction in Alveolar Plosive Consonants. Science 190. 69-72. Kuhl, P.K. and J.D. Miller 1978 Speech Perception by the Chinchilla: Identification Functions for Synthetic VOT Stimuli. Journal of the Acoustical Society of America 63. 905-917. Kuhl, P.K. and J.D. Miller 1982 Discrimination of Auditory Target Dimensions in the Presence or Absence of Variation in a Second Dimension by Infants. Perception and Psychophysics 31. 279-292. Labov, W. 1978
Le parler ordinaire, vol. 1. Paris: Minuit.
Labov, W. and J. Waletsky 1967 Narrative Analysis: Oral Versions of Personal Experiences. In D. Helm (ed.), Es says on the Verbal and the Visual Arts. Seattle: University of Washington Press. Lakoff, R. 1971
Lambrecht, K. 1981
Lambrecht, K. 1987
If's, and's and but's about Conjunction. In Fillmore and T. Langendoen (eds.), Studies in Linguistic Semantics. New York: Holt Rinehard, Winston.
Topic, Antitopic and Verb Agreement in Non-standard French. Series Prag matics and Beyond. Amsterdam: John Benjamins Publishing Co.
On the Status of SVO Sentences in French Discourse. In R.S. Tomlin (ed.), Coherence and Grounding in Discourse. 217-261. Amsterdam and Philadelphia: John Benjamins Publishing Co.
Lang, E. 1977
Analyse sémantique de la connexion coordinative. DRLAV 15. 168-187.
Langdon, J.H. 1985
Fossils and the Origin of Bipedalism. Journal of Human Evolution 14. 615-635.
Langer, J. 1980
The Origins of Logic: Six to Twelve Months. New York: Academic Press.
Langer, J. 1986
The Origins of Logic: One to Two Years. New York: Academic Press.
306 Larochette, J. 1945
L'imparfait et le passé-simple. Les études classiques 13. 55-87.
Lasky, R., A. Syrdal-Lasky and R. Klein 1975 VOT Discrimination by 4- to 6-and-a-Half-Month-Old Infants from Spanish En vironments. Journal of Experimental Child Psychology 20. 215-225. Leader, L.R., P. Baillie, B. Martin and E. Vermeulen 1982a The Assessment and Significance of Habituation to a Repeated Stimulus by the Human Fetus. Early Human Development. 7. 211-283. Leader, L.R., P. Baillie, B. Martin and E. Vermeulen 1982b Fetal Habituation in High-Risk Pregnancies. British Journal of Obstetrical Gynecology. 89. 441-446. Leader, L.R., A.D. Stevens and E.R. Lumbers 1988 Measurement of Fetal Responses to Vibroacoustic Stimuli. Habituation in Fetal Sheep. Biology of the Neonate 53. 73-85. Lecanuet, J.-P., C. Granier-Deferre, H. Cohen, R. Le Houezec and M.-C. Busnel 1986 Fetal Responses to Acoustic Stimulation Depend on Heart Rate Variability Pat tern, Stimulus Intensity and Repetition. Early Human Development. 13. 269-283. Lecanuet, J.-P., C. Granier-Deferre and M.-C. Busnel 1988 Fetal Cardiac and Motor Responses to Octave-Band Noises as a Function of Central Frequency, Intensity and Heart Rate Variability. Early Human Develop ment. 18. 81-93. Lee, T. 1985
Lehiste, I. 1970 Lehiste, I. 1984
The Phonetic Quality and Long/Short Distinction of Cantonese Vowels. Fangyan (Dialect) 1. 28-38.
Suprasegmentale. Cambridge, MA: MIT Press.
Semantics, Syntax and Prosody. In M.P.R. van den Broecke and A. Cohen (eds.), Proceedings of the Tenth International Congress of Phonetic Sciences. 143-150. Dordrecht: Foris Publ.
Lehman, R.A.W. 1978 The Handedness of Rhesus Monkeys, I: Distribution. Neuropsychologia 16. 3342.
REFERENCES
307
Lehman, R.A.W. 1980 The Handedness of Rhesus Monkeys, : Consistency within and across Ac tivities. Cortex 16. 197-204. LeMay, M. 1985
Asymmetries of the Brain and Skulls of Nonhuman Primates. In S.D. Glick (ed.), Cerebral Lateralization in Nonhuman Species. New York: Academic Press.
LeMay, M. and N. Geschwind 1975 Hemispheric Differences in the Brains of Great Apes. Brain, Behavior and Evolution 11. 48-52. Leung, E.H.L. and H.L. Rheingold 1981 Development of Pointing as a Social Gesture. Developmental Psychology 17. 215-220. Lewis, M.M. 1951
Lewis, N.N. 1963
Li, C.N. (ed.) 1976
Infant Speech: A Study of the Beginnings of Language. Second edition. New York: Humanities Press.
Language, Thought and Personality in Infanee and Childhood. London and New York: Harrap and Basic Books.
Subject and Topic. New York: Academic Press.
Li, C.N. and Thompson 1976 Subject and Topic: A New Typology of Language. In C.N. Li (ed.), Subject and Topic. New York: Academic Press. Li, C.N. and Thompson 1981 Manadrin Chinese: A Functional Reference Grammar. University of California Press. Liberman, A.M. 1982 On Finding that Speech is Special. American Psychologist 37. 148-167. Liberman, A.M., F.S. Cooper, D.P. Shankweiler and M. Studdert-Kennedy 1967 Perception of the Speech Code. Psychological Review 74. 431-461. Liberman, A.M., K.S. Harris, H.S. Hoffman and B.C. Friggith 1957 The Discrimination of Speech Sounds within and across Phoneme Boundaries. Journal of Experimental Psychology 34. 358-368.
308 Liberman, A.M. and I.G. Mattingly 1985 The Motor Theory of Speech Perception Revised. Cognition 21. 1-36. Lieberman, P. 1973 Lieberman, P. 1975
Lieberman, P. 1979
Lieberman, P. 1984
Lieberman, P. 1985
On the Evolution of Human Language: An Unified View. Cognition 2. 59-94.
On the Origins of Language: An Introduction to the Evolution of Human Speech. New York: MacMillan.
Hominid Evolution, Supralaryngal Vocal Tract Physiology and the Fossil Evidence for Reconstructions. Brain and Language 7. 101-126.
The Biology and Evolution of Language. Cambridge, MA: Harvard University Press.
On the Evolution of Human Syntactic Ability. Its Pre-Adaptative Bases - Motor Control and Speech. Journal of Human Evolution 14. 657-668.
Lieberman, P. and E.S. Crelin 1971 On the Speech of Neanderthal Man. Linguistic Inquiry 2. 203-222. Lieberman, P., E.S. Crelin and D.H. Klatt 1972 Phonetic Ability and Related Anatomy of the Newborn and Adult Human, Neanderthal Man, and the Chimpanzee. American Anthropologist 74. 287-307. Liepmann, H. 1908 Limber, J. 1973
Lien, J. 1976
Lipsitt, L.P. 1979
Drei Aufsätze aus dem Apraxiegebiet. Basel: Karger.
The Genesis of Complex Sentences. In .. Moore (ed.), Cognitive Development and the Acquisition of Language. 171-185. New York: Academic Press.
Auditory Stimulation of Coturnix Embryo (Coturnix Coturnix Japonica) and its Later Effect on Auditory Preferences. Behav. Biol. il. 231-235.
The Pleasures and Annoyances of Infants: Approach and Avoidance Behavior. In E.B. Thoman (ed.), Origins of the Infant's Social Responsiveness. The Johnson and Johnson Baby Products Co., Pediatric Round Table, II. Hillsdale, NJ: Lawrence Erlbaum Associates.
REFERENCES
Locke, J.L. 1983 Lonchamp, F. 1989
Lust, B. 1981
309
Phonological Acquisition and Change. New York: Academic Press.
Acoustique descriptive des sons du français. In J.P. Tubach (ed.), La parole et son traitement automatique. 79-120. Paris: Masson.
Coordinating Studies of Coordination: to Reply to Ardery. Journal of Child Lan guage 8. 457-470.
Lust, B. and G.A. Mercis 1980 Development of Coordination in the Natural Speech of Young Children. Journal of Child Language 7. 279-304. Lyons, J. 1975
Lyons, J. 1977
Deixis as the Source of Reference. In E.L. Keenan (ed.), Formal Semantics of Natural Language. Cambridge: Cambridge University Press.
Semantics. Cambridge: Cambridge University Press.
MacFarlane, J.A. 1975 Olfaction in the Development of Social Preferences in the Neonate. iba Found. Symp. 33. 103-117. Mack, M. and P. Lieberman 1985 Acoustic Analysis of Words Produced by a Child from 46 to 149 Weeks. Jour nal of Child Language 12. 527-550. MacKain, K.S. 1982
Assessing the Role of Experience on Infant's Speech Discrimination. Journal of Child Language 9. 527-542.
MacKain, K.S., C.Y. Best and W. Strange 1980 Native Language Effects on the Perception of Liquids. Journal of Analyses of the Acoustical Society of America. 27. 527-533. Macken, M.A. 1986
Phonological Development: A Cross-Linguistic Perspective. In P. Fletcher and M. Garman (eds.), Language Acquisition. Second edition. Cambridge: Cambridge University Press.
310 Macken, M.A. 1987
Representation, Rules and Overgeneralization in Phonology. In B. MacWhinney (ed.), Mechanisms of Language Acquisition. 367-397. Hillsdale, NJ: Lawrence Erlbaum Associates.
MacKen, M.A. and D. Barton 1979 The Acquisition of Voicing Contrast in English: A Study of Voice Onset Time in Word-Initial Stop-Consonants. Journal of Child Language 7. 41-74. MacKen, M.A. and D. Barton 1980 The Acquisition of the Voicing Contrast in Spanish: A Phonetic and Phonologi cal Study of Word-Initial Stop-Consonants. Journal of Child Language 7. 433458. MacLellan, M.S., J.R. Brown, H. Rondeau, E. Shoughro, R.A. Johnson and A.R. Hale 1964 Embryonal Connective Tissue and Exudate in Ear. American Journal of Dis turbed Children 108. 164-170. MacNeilage, P.F. 1985 Bimanual Coordination and the Beginnings of Speech. In B. Lindblom and R. Zetterström (eds.), Precursors to Infant Speech. London: MacMillan. MacNeilage, P.F., M.G. Studdert-Kennedy and B. Lindblom 1987 Primate Handedness Reconsidered. Behavioral and Brain Sciences 10. 247-303. MacWhinney, B. 1987 The Competition Model. In B. MacWhinney (ed.), Mechanisms of Language Ac quisition. 249-308. Hillsdale, NJ: Lawrence Erlbaum Associates. Madison, L.S., S.A. Adubato, J.K. Madison, R.B. Nelson, J.C. Anderson, J. Erickson, L. Kuss and R.C. Goodlin 1986 Fetal Response Decrement: True Habituation? Journal of Developmental Be havioral Pediatrics. 7. 14-20. Magee, M.A. and B. Sutton-Smith 1983 The Art of Story Telling: How do Children Learn it. Young Children, May. 412. Mandier, J.M. 1982a
Mandler, J.M. 1982b
Recent Research in Story Grammars. In J.F. Le Ny and W. Kintsch (eds.), Lan guage and Comprehension. Amsterdam: North Holland Publishing Co.
Recherches récentes sur les grammaires de récit. Bulletin de psychologie 35, Special Issue "Language et compréhension". 705-715.
311
REFERENCES
Mandier, J.M. and M.S. Goodman 1982 On the Psychological Validity of Story Structure. Journal of Verbal Learning and Verbal Behavior 21. 507-523. Mandler, J.M. and N.S. Johnson 1977 Remembrance of Things Parsed: Story Structure and Recall. Cognitive ogy 9. 111-151. Maratsos, M. 1982
Psychol
The Child's Construction of Grammatical Categories. In E. Warmer and L. Rigleitman (eds.), Language Acquisition. Cambridge: Cambridge University Press.
Marquer, P., J. Le Nestour, M. Léveillé and J. Welitz 1982 Rôle de la fréquence relative des différentes acceptions d'homographes lors de l'accès au lexique. L'année psychologique 82. 353-377. Marr. D. 1982 Matalon, . 1962
Vision. San Francisco: Freeman.
Étude génétique de l'implication. In J. Piaget (ed.), Études génétique, XVI. Paris: PUF.
d'epistemologie
Mattingly, LG., A.M. Liberman, A.K. Syrdal and T. Halwes 1971 Discrimination in Speech and Nonspeech Modes. Cognitive Psychology 2. 131157. McCabe, A. and . Peterson 1985 A Naturalistic Study of the Production of Causal Connectives by Children. Jour nal of Child Language 12. 145-159. McCabe, A., S. Evely, R. Abramovitch, C.M. Corter and D.J. Pepler 1983 Conditional Statements in Young Children's Spontaneous Speech. Journal of Child Language 10. 253-258. McGonigle, B.O. and J. Flook 1978 The Learning of Hand Preferences by Squirrel Monkey. Psychological 40. 93-98. McNeil, D. 1970
Research
The Acquisition of Language. New York: Harper and Row.
McNinch, G.H. 1974 Awareness of Aural and Visual Word Boundary within a Sample of First Graders. Perceptual and Motor Skills 38. 1127-1134.
312 McWhinney, B. and E. Bates 1978 Sentential Devices for Conveying Giveness and Newness: A Cross-Cultural Developmental Study. Journal of Verbal Learning and Verbal Behavior 17. 539558. Mehler, J. 1981
The Role of Syllables in Speech Processing: Infant and Adult Data. Philosophi cal Transactions of the Royal Society London B295. 333-352.
Mehler, J., J. Bertoncini, M. Barrière and D. Jassik-Gerschenfeld 1978 Infant Recognition of Mother's Voice. Perception 7. 491-497. Mehler, J. J.Y. Dommergues, U. Frauenfelder and J. Seguy 1981 The Syllable's Role in Speech Segmentation. Journal of Verbal Learning and Verbal Behavior 20. 298-305. Mehler, J., P. Jusczyk, G. Lambertz, N. Halsted, J. Bertoncini and Amiel-Tison 1988 A Precursor of Language Acquisition in Young Infants. Cognition 29. 143-178. Mehler, J., G. Lambertz, P.W. Jusczyk and Amiel-Tison 1986 Discrimination de la langue maternelle par le nouveau-né. Comptes rendus de l' Académie des Sciences de Paris 303, Série III. 637-640. Melekian, B. 1981
Menn, L. 1976
Menn, L. 1983
Menyuk, P. 1971
Lateralization in the Human Newborn and Birth-Asymmetry of the Stepping Reflex. Neuropsychologia 19. 707-711.
Evidence for an Interactionist Discovery Theory of Child Phonology. Stanford Papers and Reports on Child Language Development 12. 169-177.
Development of Articulatory, Phonetic and Phonological Capabilities. In B. Butterworth (ed.), Language Production, vol. II. London: Academic Press.
The Acquisition and Development of Language. Englewood Cliffs, NJ: PrenticeHall.
Menyuk, P. and N. Bernholtz 1969 Prosodic Features and Children's Language Production. (QPR no. 93). Boston: Research Laboratory of Electronics, MIT.
REFERENCES
313
Menyuk, P., L. Menn and R. Silber 1986 Early Strategies for the Perception and Production of Words and Sounds. In P. Fletcher and M. Garman (eds.), Language Acquisition. Second edition. Cambridge: Cambridge University Press. Menzel, E.W. 1973
Leadership and Communication in Young Chimpanzees. In E.W. Menzel (ed.), Precultural Primate Behaviour. Basel: Karger.
Menzel, E.W. and S. Halperin 1975 Purposive Behavior as a Basis for Objective Communication between Chimpan zees. Science 189. 652-654. Meyer, M. 1982
Logique, language, argumentation. Paris: Hachette.
Miller, CL., B.A. Younger and P.A. Morse 1982 The Categorization of Male and Female Voices in Infancy. Infant Behaviour and Development 5. 143-159. Miller, G.A. 1967
Miller, G.A. 1969
Empirical Methods in the Study of Semantics. In D.D. Steinberg and L.A. Jakobowits (eds.), Semantics: An Interdisciplinary Reader in Philosophy, Lin guistics and Psychology. Cambridge: Cambridge University Press.
A Psychological Method to Investigate Verbal Concepts. Journal of Mathemati cal Psychology 6. 169-191.
Miller, J.D., C.C. Weir, R.E., Pastore, W.J. Kelly and R.J. Dooling 1976 Discrimination and Labelling of Noise-Buzz Sequences with Varying Noise-lead Times: An Example of Categorical Perception. Journal of the Acoustical Society of America 60. 410-417. Miller, J.L. 1981
Effects of Speaking Rate on Segmental Distinctions. In P.D. Eimas and J.L. Miller (eds.), Perspectives on the Study of Speech. Hillsdale, NJ: Lawrence Erlbaum Associates.
Miller, J.L. and P.D. Eimas 1983 Studies on the Categorization of Speech by Infants. Cognition 13. 135-165. Miller, J.L. and F. Grosjean 1981 How the Components of Speaking Rate Influence Perception of Phonetic Seg ments. Journal of Experimental Psychology: Human Perception and Performance 7. 208-215.
314 Miller, J.L. and A. Liberman 1979 Some Effects of Later-Occurring Information on the Perception of Stop Con sonant and Semivowel. Perception and Psychophysics 25. 457-465. Mills, M. and E. Melhuish 1974 Recognition of Mother's Voice in Early Infancy. Nature 252. 123-124. Miyamoto, M.M., J.L. Slightom and M. Goodman 1987 Phylogenetic Relations of Humans and African Apes from DNA Sequences in the ψη-Globin Region. Science 238. 369-373. Miyawaki, K., W. Strange, R.R. Verbrugge, A.M. Liberman, J.Y. Jenkins and O. Fujimura 1975 An Effect of Linguistic Experience: The Discrimination of (R) and (L) by Na tive Speakers of Japanese and English. Perception and Psychophysics 18. 331340. Molfese, D.L. and V.J. Molfese 1979 Hemisphere and Stimulus Differences as Reflected in the Cortical Responses of Newborn Infants to Speech Stimuli. Developmental Psychology 15. 505-511. Moody, E.A. 1953
Moon, C. 1985
Truth and Consequence in Medieval Logic. Amsterdam: North Holland Publish ing Co.
Phonological Universals and Syllable Discrimination in One- to Three-Day-Old Infants. Unpublished Doctoral dissertation. Columbia University.
Moon, C. and W.P. Fifer 1986 Newborns Prefer the Sound of Mother's Voice as Experienced in the Womb. Paper presented at the ICIS meeting. Annapolis. Moore, D.R. 1985
Morse, P.A. 1972
Postnatal Development of the Mam alian Central Auditory System and the Neural Consequences of Auditory Deprivation. Acta Otolaryngol, (Stockh.), Suppl. 421. 19-30.
The Discrimination of Speech and Nonspeech Stimuli in Early Infance. Journal of Experimental Child Psychology 14. 477-492.
Morse, P.A., R.E. Eilers and W.J. Gavin 1982 The Perception of the Sound of Silence in Early Infancy. Child Development 53. 189-195.
REFERENCES
Morton, J. 1969
315
The Interaction of Information in Word Recognition. Psychological Review 76. 165-178.
Morton, J. and Long 1976 Effect of Word Transition Probability on Phoneme Identification. Journal of Ver bal Learning and Verbal Behavior 15. 43-51. Mounoud, P. and T.G.R. Bower 1975 Conservation of Weight in Infants. Cognition 3. 29-40. Muller, E., H. Hollien and T. Murray 1974 Perceptual Response to Infant Crying: Identification of Cry Types. Journal of Child Language 1. 89-95. Murooka, H., Y. Koie and D. Dusa 1976 Analyse des sons intrautérins et de leurs effets tranquilisants sur le nouveau-né. Journal of Gynecological Obstetrical Biological Reproduction. 5. 367-376. Murry, T., P. Amundson and H. Hollien 1977 Acoustical Characteristics of Infant Cries: Fundamental Frequency. Journal of Child Language 4. 321-328. Nachsohn, I. 1986
Needleman, H. 1977
Nelson, K.E. 1975
Cross-Cultural Differences in Laterality. International Journal of 30. 291-295.
Neuroscience.
Effects of Hearing Loss from Early Recurrent Otitis Media on Speech and Lan guage Development. In B. Jaffe (ed.), Hearing Loss in Children. 640-649. Bal timore: University Park Press.
The Nominal Shift in Semantic-Syntactic Development. Cognitive Psychology 1. 461-479.
Ninio, A. and J. Bruner 1978 The Achievement and Antecedents of Labelling. Journal of Child Language 5. 1-15. Noizet, G. 1980 Nottebohm, F. 1971
De la perception à la compréhension du language. Paris: PUF.
Neural Lateralization of Vocal Control in a Passerine Bird: 1 - Song. Journal of Experimental Zoology 111. 229-262.
316 Nottebohm, F. 1972
Nottebohm, F. 1980
Neural Lateralization of Vocal Control in a Passerine Bird: 2 - Subsong, Calls and a Theory of Vocal Learning. Journal of Experimental Zoology 179. 35-50.
Brain Pathways for Vocal Learning in Birds: A Review of the First 10 Years. In J.M.S. Sprage and A.N.E. Epstein (eds.), Progress in Psychobiology and Physiological Psychology 9. 85-124. New York: Academic Press.
Nooteboom, S.G., P.L. Brokx and J.J. de Rooij 1978 Contributions of Prosody to Speech Perception. In W. Level and G. Flores d'Arcais (eds.), Studies in the Perception of Language. New York: Wiley and Sons. Oakley, K. 1961 Ochs, E. 1987
Man the Tool-Maker. Chicago: Chicago University Press.
Input: A Socio-Cultural Perspective. In M. Hickmann (ed.), Social and Function al Approaches to Language and Thought. Academic Press.
Ochs, E. and B.B. Schieffelin (eds.) Developmental Pragmatics. New York: Academic Press. 1979 Oller, D.K. 1980
Oller, D.K. 1986
The Emergence of the Sounds of Speech in Infancy. In G. Yeni-Komshian, C. Kavanagh and Ferguson (eds.), Child Phonology: Perception and Production. London: Academic Press.
Metaphonology and Infant Vocalizations. In B. Lindblom and R. Zetterstrom (eds.), Precursors of Early Speech. Basingstoke: MacMillan.
Oiler, D.K. and B.L. Smith 1977 The Effect of Final-Syllable Position on Vowel Duration in Infant Babbling. Journal of the Acoustical Society of America 62(4). 994-997. Oiler, D.K., L.A. Weiman, W.J. Doyle and Ross 1976 Infant Babbling and Speech. Journal of Child Language 3. 1-11. Olsho, L.W., C. Schoon, R. Sakai, R. Turpin and V. Sperduto 1982 Auditory Frequency Discrimination in Infancy. Developmental Psychology 18. 721-726.
REFERENCES
317
Overton, W.F., L.W. Shawn, I.A. Novec, J. Black and D.P. O'Brien 1987 Form and Content in the Development of Deductive Reasoning. Developmental Psychology 23. 22-30. Papandropoulou, J. and H. Sinclair 1974 What is a Word? Experimental Study of Children's Ideas on Grammar. Human Development 17. 241-258. Paris, S.G. 1975
Pasch, R. 1982
Propositional Logical Thinking and Comprehension of Language Connectives. The Hague: Mouton.
Untersuchungen zu den Gebrauchsbedingungen der deutschen Kausalkon junktionen da, denn und weil. Linguistische Studien. Teihe A: Arbeitsberichte 104. 41-243.
Pateman, T. 1982
Communicating with Computer Programs. Journal of Pragmatics 6. 225-240.
Peiper, A. 1925
Simmesempfíndugen des Kinder vor Seiner Geburt. Msr., Kinderhuj 29. 236.
Perelman, and L. Olbrechts-Tyteca 1970 Traité de l'argumentation. Brussels: Editions de l'Université de Bruxelles. Peters, M. and B.F. Petrie 1979 Functional Asymmetries in the Stepping Reflex of Human Neonates. Canadian Journal of Psychology 33. 198-200. Petersen, M.R., M.D. Beecher, S.R. Zoloth, D.R. Moody and W.C. Stebbins 1978 Neural Lateralization of Species-Specific Vocalizations by Japanese Macaques (Macaca Fuscata). Science 207. 324-326. Petersen, M.R., M.D. Beecher, S.R. Zoloth, S. Green, P.R. Marker, D.B. Moody and W.C. Steb bins 1984 Neural Lateralization of Vocalization by Japanese Macaques: Communicative Significance is more Important than Acoustic Structure. Behavioral Neuroscience 98. 779-790. Petersen, S.E., P.T. Fox, M.I. Posner, M. Mintum and M.E. Raichle 1988 Positron Emission Tomographic Studies of the Cortical Anatomy of Single-Word Processing. Nature 331. 585-589. Peterson, 1986
Semantic and Pragmatic Uses of "but". Journal of Child Language 13. 583-590.
318 Peterson, C. and A. McCabe 1985 Understanding "because". How Important is the Task? Journal of Psycholinguis tic Research 14. 199-218. Peterson, and A. McCabe 1987 The Structure of AND Conditions in Psycholinguistic Research 16. 467-490.
Children's
Narratives. Journal
of
Peterson, G.E. and J.L. Barney 1952 Control Methods Used in a Study of the Vowels. Journal of the Acoustical Society of America 24. 175-184. Pettito, L.A. 1987
Piaget, J. 1924/1926
Piaget, J. 1937/1954
On the Autonomy of Language and Gesture: Evidence from the Acquisition of Personal Pronouns in American Sign Language. Cognition 27. 1-52.
Le jugement et le raisonnement. Neuchâtel: Delachaux et Niestlé. {Judgement and Reasoning in the Child. New York: Harcourt and Brace).
La construction du réel chez l'enfant. Neuchâtel and Paris: Delachaux et Niestlé. (The Construction of Reality in the Child. New York: Basic Books).
Piaget, J. and B. Inhelder 1941/1962 Le développement des quantités physiques chez l' enfant: conservation atomisme. Second edition. Neuchâtel and Paris: Delachaux et Niestlé.
et
Piéraut-Le Bonniec, G. 1980a The Development of Modal Reasoning. New York: Academic Press. Piéraut-Le Bonniec, G. 1980b Développement de la compréhension des phrases conditionnelles. psychologique 80. 65-79.
L'année
Piéraut-Le Bonniec, G. 1990 The Logic of Meaning and Meaningful Implication. In W.F. Overton, Reasoning, Necessity and Logic: Developmental Perspectives. 67-995. Hillsdale, NJ: Lawrence Erlbaum Associates. Pierrehumbert, J. 1981 Synthesizing Intonation. Journal of the Acoustical Society of America 70 (4). 985-995.
REFERENCES
Pille, J.P. 1986
Pisoni, D.B. 1977
319
La notion de verbe: un exemple du développement de l'activité métalinguistique de l'enfant. Doctoral dissertation. Paris: Université René Descartes.
Identification and Discrimination of the Relative Onset Time of Two Component Tones: Implications for Voicing Perception in Stops. Journal of the Acoustical Society of America 61. 1352-1361.
Poizner, H., R. Battison and H. Lane 1979 Cerebral Asymmetry for American Sign Language: The Effects of Moving Stimuli. Brain and Language 7. 351-362. Poizner, H. and P. Tallal 1987 Temporal Processing in Deaf Signers. Brain and Language 30. 52-63. Preilowski, B. 1979
Performance Differences between Hands and Lack of Transfer of Finger Posture and Sensory-Motor Skill in Intact Rhesus Monkeys: Possible Model of the Origin of Cerebral Asymmetry. Neuroscience Letters (supplement) 3. 89.
Preilowski, B. and F. Leder 1984 Comparative Studies of Laterality: Hand Use of a Captive Group of Lowland Gorillas. In Proceeding of the 34th Congress of the German Society for Psychol ogy. Vienna: A. Albert. Pujol, R. and A. Uziel 1986 Auditory Development: Peripheral Aspects. In P. Timiras and E. Meisami (eds.), Handbook of Human Biologic Development. Boca Raton, FL: CRC Press. Pusch, L. 1975
Über den Unterschied zwischen aber und sondern, oder die Kunst des Widersprechens. In I. Batori et al. (eds.), Syntaktische und semantische Studien zur Koordination. 42-62. Tübingen: G. Narr.
Python-Thuillard, F. and G. Assal 1987 Troubles de la perception du language chez les aphasiques: étude d'une illusion Audio-verbale. Archives de psychologie 55. 305-313. Quasthoff, U.M. 1979 Verzögerungsphänomene, Verknùpsfungs une Gliederungssignale in Alltagsar gumentationen und Alltagserzählungen. In H. Weydt (ed.), Die Partikeln der deutsche Sprache. 39-57. Berlin: De Gruyter. Quasthoff, U.M. 1980 Erzählen in Gesprächen. Tübingen: G. Narr.
320 Quasthoff, U.M. and . Nikolaus 1982 What Makes a Good Story? Towards the Production of Conversational Narra tives. In A. Rammer and W. Kintsch (eds.), Discourse Processing. 16-28. New York: North Holland Publishing Co. Querleu, D., Boutteville, X. Renard and G. Crépin 1984a Évaluation diagnostique de la souffrance foetale pendant la grossesse au moyen d'un test de stimulation sonore. Journal of Gynecological Obstetrical Biological Reproduction. 13. 125-134. Querleu, D., C. Lefebvre, M. Titran, X. Renard, M. Morillon and G. Crépin 1984b Réactivité du nouveau-né de moins de deux heures de vie à la voix maternelle. Journal of Gynecological Obstetrical Biological Reproduction. 13. 125-134. Querleu, D. and X. Renard 1981 Les perceptions auditives du foetus humain. Medecine et gene. 39. 2101-2110. Querleu, D., X. Renard and F. Versyp 1986 Vie sensorielle du foetus. In M. Toumaire and G. Levy (eds.), L'environnement de la naissance. 15-41, Paris: Vigot. Rajecki, D.W. 1974
Rapin, I. 1979
Ray, W.S. 1932 Rayner, . , 1983
Effects of Prenatal Exposure to Auditory or Visual Stimulation on Post Natal Distress Vocalisations in Chicks. Behavioral Biology. 11. 525-536.
Conductive Hearing Loss Effects on Children's Language and Scholastic Skills. In D. Hanson and R.F. Ulvestatd (eds.), Ototos Media and Child Development: Speech Language and Education. Annals of Otol. Rhinol. Laryngol, (supplement) 88. 3-12.
A Preliminary Study of Fetal Conditioning. Child Development. 3. 173-177. M. Carlson and L. Frazier The Interaction of Syntax and Semantics During Sentence Processing: Eye Movements in the Analysis of Semanatically Biased Sentences. Journal of Ver bal Learning and Verbal Behavior 22. 358-374.
Read, J.A. and F.C. Miller 1977 Fetal Heart Rate Acceleration in Response to Acoustic Stimulation as Measure of Fetal Well-Being. American Journal of Obstetrical Gynecology. 129. 512-517. Repp, B.H. 1982
Phonetic Trading Relations and Context Effects: New Experimental Evidence for a Speech Mode of Perception. Psychological Bulletin 92. 81-110.
REFERENCES
Repp, B.H. 1987
Integration and Segregation in Speech Perception. Ha skins Laboratories Report on Speech Research SR-89/90. 1-32.
321
Status
Rietveld, A.C.M. 1980 Word Boundaries in the French Language. Language and Speech 23. 289-296. Rosch, E. 1873
Natural Categories. Cognitive Psychology 4. 328-350.
Rossi, M., A. DiCristo, D. Hirst, P. Martin and Y. Nishinuma 1981 L'intonation: de l' acoustique à la sémantique. Paris: Klincksieck. Roulet, E. 1981
Rubel, E.S. 1978
Rubel, E.W. 1985
L'analyse des conversations authentiques (coordination). Études de linguistique appliquée 44.
Ontogeny of Structure and Function in the Vertebrate Auditory System. In M. Jacobson (ed.), Handbook of Sensory Physiology, vol. 9: Development of Sensory Systems. Berlin: Springer Verlag.
Auditory System Development. In G. Gottlieb and N. Krasnegor (eds.), Meas urement of Audition and Vision in the First Year of Postnatal Life, a Methodological Overview. 53-90. Norwood: Ablex Publishing Co..
Rubel, E.W. and B.M. Ryals 1983 Development of the Place Principle: Acoustical Trauma. Science 219. 512-514. Ruben, R J . and I Rapin 1980 Plasticity of the Developing Auditory System. Ann. tol. Rhinol. Laryngol. 89. 303-311. Rumelhart, D.E. 1977 Understanding and Summarizing Brief Stories. In D. Laberge and S.J. Samuels (eds.), Basic Processes in Reading: Perception and Comprehension. Hillsdale, NJ: Lawrence Erlbaum Associates. Rumelhart, D.E. and J.L. McClelland 1986 On Learning the Past Tenses of English Verbs. In J.L. McClelland, D.E. Rumel hart and the DPD Research Group, Parallel Distributed Processing: Explorations in the Microstructure of Cognition, vol. 2: Psychological and Biological Models. Cambridge, MA: Bradford Books/MIT Press.
322 Russel, M.J. 1976 Sacks, H. 1972
Human Olfactory Communication. Nature 260. 520-522.
On the Analyzability of Stories by Children. In J.J. Gumperz and D. Hymes (eds.), Directions in Sociolinguistics. 325-345. New York: Rinehart and Winston
Sakabe, N., T. Arayama and T. Suzuki 1969 Human Fetal Response to Acoustic Stimulation. Acta Oto-Laryngology. (supple ment) 252. 29-36. Sammons, W.A.H. 1980 Premature Behavior and the Neonatal Intensive Care Unit Environment. In J.P. Cloherty and A.R. Stark (eds.), Manual of Neonatal Care. 359-363. Boston: Lit tle and Brown. Sanes, D.H. and M. Constantine-Paton 1983 Altered Activity Patterns During Development Reduce Neural Tuning. Science 111. 1183-1185. Satt, B.J. 1984
An Investigation into the Acoustical Induction of Intrauterine-Uterine Learning. PhD. dissertation. Cal. School Prof. Psychol.
Savage-Rumbaugh, S., K. McDonald, R.A. Sevcik, W.D. Hopkins and E. Rubert 1986 Spontaneous Symbol Acquisition and Communicative Use by Pygmy Chimpan zees {Pan Paniscus). Journal of Experimental Psychology: General 115. 211-235 Savage-Rumbaugh, E.S., D.M. Rumbaugh, S.T. Smith and J. Lawson 1980 Reference: The Linguistic Essential. Science 210. 922-925. Scaife, M. and J.S. Bruner 1975 The Capacity for Joint Visual Attention in the Infant. Nature 253. 265-266. Schaal, B. 1984
La compétence olfactive du nouveau-né humain: Aspects fonctionnels. Psychol. Med. 16. 305-314.
Schaal, ., H. Montagner, E. Herling, D. Bolzoni, E. Moyse and R. Quichon 1980 Les stimulations olfactives dans les relations entre l'enfant et la mère. Reproduc tive Nutritional Development. 20. 843-858. Schaffer, H.R., G.M. Collis and G. Parsons 1977 Vocal Interchange and Visual Regard in Verbal and Pre-Verbal Children. In H.R. Schaffer (ed.), Studies in Mother-Infant Interaction. 291-324. London: Academic Press.
323
REFERENCES
Schank, R.D. 1975
The Structure of Episodes in Memory. In D.G. Bubrow and A. Collins (eds.), Representation and Understanding. New York: Academic Press.
Schank, R.C. and R.P. Anderson 1977 Scripts, Plans, Goals and Understanding. Hillsdale, NJ: Lawrence Erlbaum As sociates. Schiffrin, D. 1981
Tense Variation in Narrative. Language 57. 45-62.
Schmidt, W., R. Boos, J. Gniers, L. Auer and S. Schulze 1985 Fetal Behavioural States and Controlled Sound Stimulation. Early Development. 12. 145-153.
Human
Schneider, B.A., S.E. Trehub and D. Bull 1979 The Development of Basic Auditory Processes in Infants. Canadian Journal of Psychology 33. 306-319. Schneuwly, B. 1981
Schneuwly, B. 1984
Étude approfondie des textes discursifs. Geneva: Rapport à l'intention du Fonds National Suisse de la Recherche Scientifique.
Le texte discursif écrit à l'école. Doctoral dissertation. Université de Génève.
Schneuwly, B. and J.-P. Bronckart 1983 Pour une psychologie du language. Archives de psychologie 51. 155-160. Scholnick, E.K. and C.S. Wing 1982 The Pragmatics of Subordinating Conjuctions: A Second Look. Journal of Child Language 9. 461-479. Schonen, S. de 1977
Schonen, S. de 1989
Functional Asymmetries in the Development of Bimanual Coordinations in Human Infants. Journal of Human Movements Studies 3. 144-156.
First Come, First Served: A Scenario about the Last Stages in the Development of the Hemispheric Specialization in Face Processing. European Bulletin of Cog nitive Psychology / Cahiers de psychologie Cognitive 9. 3-44.
Schonen, S. de and F. Bresson 1984 Développement de l'atteinte manuelle d'un object chez l'enfant. 1. 99-114.
Comportements
324 Schubert, R.E. and P.D. Eiman 1977 Effects of Context on the Classification of Words and Nonwords. Journal of Ex perimental Psychology: Human Perception and Performance 3. 27-36. Schulman-Galambos, and R. Galambos 1979 Brainstem Evoked Response Audiometry on Newborn Hearing Screening. Ar chives of Otolaryngology. 105. 86-90. Schwartzman, J., B. Wilkonson and P. Bertola 1985 Prenatal Testing of Hearing through ERA, Preliminary Communication. Santa Barbara Meeting. Scibetta, J.J. and M.G. Rosen 1969 Response Evoked by Sound in the Fetal Guinea-Pig. Obstetrical Gynecology. 33. 830-836. Scibetta, J.J., M.G. Rosen, C.J. Hochberg and L. Chick 1971 Human Fetal Brain Response to Sound During Labor. American Journal of Obstetrical Gynecology. 109. 82-85. Scott, S., G.W. Hynd, L. Hunk and W. Need 1979 Cerebral Speech Lateralization in the Native American Navajo. sychologia 17. 503-506. Searle, J.R. 1969
Speech Acts. Cambridge: Cambridge University Press.
Sechehaye, A. 1950
Essai sur la structure logique de la phrase. Paris: Champion.
Neurop-
Serafini, P., M.B.J.N. Lindsay, D.A. Nagey, M.J. Pupkin, P. Tseng and Crenshaw 1984 Anterpartum Fetal Heart Rate Response to Sound Stimulation: the Acoustic Stimulation Test. American Journal of Obstetical Gynecology. 148. 41-45. Sheeran, L. 1973
Vertical Decalage in Weight Conservation between Sensori-Motor and Concep tual Levels. Doctoral dissertation. Edinburgh University.
Sherman, G.F. and A.M. Galaburda 1982 Cortical Volume Asymmetry and Behavior in the Albino Rat. Abstracts 8. 627.
Neuroscience
Sherman, G.F. and A.M. Galaburda 1984 Neocortical Asymmetry and Open-Field Behavior in the Rat. Neurology 86. 473-482.
Experimental
REFERENCES
325
Shockey, L. and Z.S. Bond 1980 Phonological Processes in Speech Addressed to Children. Phone tica 37. 267-274. Silva, M.N. 1984
Silverstein, M. 1976
Silverstein, M. 1987
Sinclair, H. 1970
Sinclair, H. 1971
Sinclair, H. 1987
Developmental Issues in the Acquisition of Conjunction. Papers and Reports on Child Language Development 23. 106-114.
Hierarchy of Features and Ergativity. In R.M.W. Dixon (ed.), Grammatical Categories in Australian Languages. New Jersey: Humanities Press.
Cognitive Implications of a Referential Hierarchy. In M. Hickmann (ed.), Social and Functional Approaches to Language and Thought. Academic Press.
The Transition from Sensori-Motor Behavior to Symbolic Activity. Interchange 1. 119-126.
Sensori-Motor Section Patterns as a Condition for the Acquisition of Syntax. In R. Huxley and E. Ingram (eds.), Language Acquisition: Models and Methods. 121-130. London: Academic Press.
Activités organisatrices du jeune enfant et début du language. In G. Piéraut-Le Bonniec (ed.), Connaître et le dire. Brussels: Mardaga.
Sinclair, H. and J.-P. Bronckart 1972 SVO, a Linguistic Universal? A Study in Developmental Psycholinguistics. Jour nal of Child Experimental Psycliology 14. 329-348. Sinclair, H., M. Stambak, I. Lezine, et al. 1982 Les bébés et les choses, ou la créativité du développement cognitif Paris: PUF. Sinnot, J.M., D.B. Pisoni and R.N. Aslin 1983 A Comparison of Pure Tone Auditory Thresholds in Human Infants and Adults. Infant Behavior and Development 6. 3-17. Siperstein, G.N. and E. Butterfield 1973 Neonates Prefer Vocal-Instrumental Music to Noise and Vocal Music to In strumental Music. Unprepared manuscript. University of Kansas Medical Center. Siqueland, E.R. and CA. DeLucia 1969 Visual Reinforcement of Nonnutritive Sucking by Human Infants. Science 165. 1144-1146.
326 Slakta, D. 1975 Slobin, D.I. 1973
Slobin, D.I. 1982
L'ordre du texte. Études de linguistique appliquée 19. 30-42.
Cognitive Prerequisites for the Development of Grammar. In C.A. Ferguson and D.I. Slobin (eds.), Studies of Child Language Development. 175-208. New York: Holt, Rinehart and Winston.
Universal and Particular in the Acquisition of Language. In L.R. Gleitman and E. Wanner (eds.), Language Acquisition: State of the Art. Cambridge: Cambridge University Press.
Smith, Z.D.J., L. Gray and E.W. Rubel 1983 Afferent Influences on Brain Stem Auditory Nuclei of the Chicken: Nucleus Laminairs Dendritic Length Following Monaural Acoustic Deprivation. Journal of Comparative Neurology. 220. 199-205. Smotherman, W.P. 1982a In utero Chemosensory Experience Alters Taste Preference and Corticosteroid Responsiveness. Behavioral Neurological Biology. 36. 61-68. Smotherman, W.P. 1982b Odor Aversion Learning by the Rat Fetus. Physiological Behavior. 28. 5-7. Sontag, L.W. and R.F. Wallace 1935 The Movement Response of the Human Fetus to Sound Stimuli. Child Develop ment. 6. 253-258. Sontag, L.W. and R.F. Wallace 1936 Changes in the Rate of the Human Fetal Heart in Response to Vibratory Stimuli. American Journal of the Disturbed Child. 51. 583-589. Spelt, D.K. 1948
The Conditioning of the Human Fetus in utero. Journal of Experimental Psychology. 38. 338-346.
Spence, M.J. and A.J. DeCasper 1986 Prenatal Experience with Low Frequency Maternal Voice Sounds Influences the Newborn's Perception of Maternal Voice. Infant Behavior Development. 10. 133142. Stanovich, K.E. and R.F. West 1981a The Stability and Generality of Sentence Context Effects on Word Recognition Speech. Paper presented at the Annual meeting of the Psychonomic Society. Philadelphia, November.
REFERENCES
327
Stanovich, K.E. and R.F West 1981b The Effect of Sentence Context on Ongoing Word Recognition: Tests of a TwoProcess Theory. Journal of Experimental Psychology: Human Perception and Performance 1. 658-672. Stark, R.E. 1980
Stark, R.E. 1986
Stages of Speech Development in the First Year of Life. In G. Yeni-Komshian, J. Kavanagh and C. Ferguson (eds.), Child Phonology, vol. 1: Production. New York: Academic Press.
Prespeech Segmental Feature Development. In P. Fletcher and M. Garman (eds.), Language Acquisition. 149-173. Cambridge: Cambridge University Press.
Starr, A., R.N. Amlie, W.H. Martin and S. Sanders 1977 Development of Auditory Function in Newborn Infants Revealed by Auditory Brainstem Potentials. Pediatrics 60. 831-839. Stein, N.L. and C G . Glenn 1979 An Analysis of Story Comprehension in Elementary School Children. In R. Freedle (ed.), New Directions in Discourse Processing, vol. II. Norwood: Ablex Publishing Co. Stein, N.L. and C G . Glenn 1982 Children's Concept of Time: The Development of a Story Schema. In W J . Friedman (ed.), The Developmental Psychology of Time. New York: Academic Press. Stein, N.L. and T. Trabasso 1982 What's in a Story: An Approach to Comprehension and Instruction. In R. Glaser (ed.), Advances in Instructional Psychology, vol. 2. Hillsdale, NJ: Lawrence Erlbaum Associates. Steklis, H.D. and S.R. Hamad 1976 From Hand to Mouth: Some Critical Stages in the Evolution of Language. In S.R. Harnad, H.D. Steklis and J. Lancaster (eds.), Origins and Evolution of Lan guage and Speech. Annals of the New York Academy of Sciences 280. 445-455. Stern, and W. Stern 1928 Die Kindersprache. Leipzig: Barth. Stern, D.N. S. Spieker and K. MacKain 1982 Intonation Contours as Signals in Maternal Speech to Prelinguistic Infants. Developmental Psychology 18. 727-735.
328 Stern, D.N., S. Spieker, R.K. Barnett and . MacKain 1983 The Prosody of Maternal Speech: Infant Age and Context Related Changes. Journal of Child Language 10. 1-15. Sternberg, R J . 1984
A Theory of Knowledge Acquisition in the Development of Verbal Concept. Developmental Review 4(2). 113-138.
Stevens, K.N. and S.E. Blumstein 1981 The Search for Invariant Acoustic Correlates of Phonetic Features. In P.D. Eimas and J.L. Miller (eds.), Perspectives on the Study of Speech. 1-38. Hillsdale, NJ: Lawrence Erlbaum Associates. Stickrod, G., D.P. Kimble and W.P. Smotherman 1982 In utero Taste/Odor Aversion Conditioning in the Rat. Physiological 29. 769-771.
Behavior.
Stockman, I., D.R. Woods and A. Tishman 1981 Listener Agreement on Phonetic Segments in Early Infant Vocalizations. Journal of Psycholinguistic Research 10. 593-617.
Stawson, P.F. 1950 Streeter, L.A. 1976
Streeter, L.A. 1978
On Referring. Mind 59. 320-344.
Language Perception of 2-Month-Old Infants Show Effects of both Innate Mechanisms and Experience. Nature 259. 39-41.
Acoustic Determinants of Phrase Boundary Perception. Journal of the Acoustical Society of America 64 (6). 1582-1592.
Stringer, C.B. and P. Andrews 1988 Genetic and Fossil Evidence for the Origin of Modern Humans. Science 239. 1263-1268. Studdert-Kennedy, M. 1986 Sources of Variability in Early Speech Development. In J.S. Perkell and D.H. Klatt (eds.), Invariance and Variability in Speech Processes. Hillsdale, NJ: Lawrence Erlbaum Associates. Suppes, P., M. Léveillé and R.C. Smith 1974 Developmental Models of a Child's French Syntax. Technical Report Number 243. December 4. Psychology and Education Series. Institude for Mathematical Studies in the Social Sciences, Stanford University.
329
REFERENCES
Sutton, D., R.E. Trachy and R.C. Lindeman 1981 Vocal and Non-Vocal Discriminative Performance in Monkeys. Brain and Lan guage 14. 93-105. Swoboda, P., P. Morse and L. Leavitt 1976 Continuous Vowel Discrimination in Normal and at-Risk Infants. Child Develop ment 47. 459-465. Tager-Flusberg, H., J. DeVillersa and K. Hakuta 1982 The Development of Sentence Coordination. In S.A. Kuczaj II (ed.), Language Development, vol. I: Syntax and Semantics. 201-243. Hillsdale, NJ: Lawrence Erlbaum Associates. Tannen, D. 1977
Well What Did You Expect? Proceeding Berkeley Linguistic Society. 506-515.
of the 3rd Annual Meeting of the
Ten Houten, N. 1976 More on Split-Brain Research, Culture and Cognition. Current Anthropology 503-506. Terhard, E. 1974
17.
Pitch, Consonance, and Harmony. Journal of the Acoustical Society of America 55. 1061-1069.
Thomdyke, P.W. 1977 Cognitive Structures in Comprehension and Memory of Narrative Discourse. Cognitive Psychology 9. 77-110. Thomdyke, P.W. and F.R. Yekovitch 1980 A Critique of Schemata as a Theory of Human Story Memory. Poetics 9. 32-50. Thorsen, N. 1980
A Study of the Perception of Sentence Intonation - Evidence from Danish. Jour nal of the Acoustical Society of America 67. 1014-1030.
Tomlin, R.W. (ed.) 1987 Coherence and Founding jamins Publishing Co.
in Discourse.
Amsterdam, Philadelphia: John Ben
Torres, F. and Anderson 1985 The Normal EEG of the Human Newborn. Journal of Clinical 2. 89-130.
Neurophysiology.
330 Toth, N. 1985
Archaeological Evidence for Preferential Right-Handedness in the Lower and Middle Pleistocene, and its Possible Implications. Journal of Human Evolution 14. 607-614.
Toth, N. and K.D. Schick 1986 The First Million Years: The Archaeology of Protohuman Culture. Advances in Archaeological Method and Theory 9. 1-96. Tsunoda, T. 1971
Tsunoda, T. 1985
The Differentiation of the Cerebral Dominance of Vowel Sounds among Dif ferent Languages. Journal of Auditory Research. 11. 305-314.
The Japanese Brain. Tokyo: Taishukan.
Tulving, E. and Gold 1963 Stimulus Information and Contextual Information as Determinants of Tachistrscopic Recognition of Words. Journal of Experimental Psychology 66. 319-327. Tunmer, W.E. and R. Grieve 1984 Syntactic Awareness in Children. In W.E. Tunmer, Prat and M.L. Herriman (eds.), Metalinguistic Awareness in Children. Berlin: Springer Verlag. Trehub, S.E., B.A. Schneider and M. Endman 1980 Developmental Changes in Infants' Sensitivity to Octave Band Noises. Journal of Experimental Child Psychology 29. 282-293. Trevarthen, C. 1977
Descriptive Analyses of Infant Communicative Behaviour. In H.R. Schaffer (ed.), Studies in Mother-Infant Interaction. 227-270. London and New York: Academic Press.
Trevarthen, C. and P. Hubley 1978 Secondary Intersubjectivity: Confidence, Confiding and Acts of Meaning in the First Year. In A. Lock (ed.), Action, Gesture and Symbol. The Emergence of Language. 183-229. Trosborg, A. 1982
Children's Comprehension of "Before" and "After" Reinvestigated. Journal of Child Language 9. 381-402.
Turkewitz, G. and G.H. Birch 1971 Neurobehavioral Organization of the Human Newborn. In J. Hellmuth (ed.), Ex ceptional Infant. 2: Studies in Abnormalities. New York: Bruner and Marel.
REFERENCES
Vaissière, J. 1980
331
Structuration prosodique de la phrase française. Annali delia Scula Normale Superiore di Pisa, seria III, X, 2. 529-560.
Valladas, H., J.L. Reyss, J.L. Joron, G. Valladas, O. Bar-Yosef and B. Vandermeersch 1988 Thermoluminescence Dating of Mousterian "Proto-Cro-Magnon: Remains from Israel and the Origin of Modem Man. Nature 331. 614-616. Vandermeersch, B. 1981 Les hommes fossiles de Qafzeh, Israel. Paris: CNRS. Van Dijk, T.A. 1977 Text and Context: Explorations in the Semantics and Pragmatics of Discourse. London and New York: Longman. Van Dijk, T.A. 1979 Pragmatic Connectives. Journal of Pragmatics 3. 447-456. Van Dijk, T.A. 1980 Semantic Macrostructures. New York: Lawrence Erlbaum Associates. Van Dijk, T.A. and W. Kintsch 1983 Strategies of Discourse Comprehension. New York: Academic Press. Van Kleeck, A. 1984 Metalinguistic Skills: Cutting across Spoken and Written Language and ProblemSolving Abilities. In G.P. Wallach and K.G. Butler (eds.), Language Learning Disabilities in School-Age Children. Baltimore, MD: Williams and Wilkins. Vauclair, J. and J. Fagot 1987 Spontaneous Hand Usage and Handedness in a Troop of Baboons. Cortex 23. 265-274. Vihman, M.M., C.A. Ferguson and M. Elbert 1986 Phonological Development from Babbling to Speech: Common Tendencies and Individual Differences. Applied Psycholinguistics. 3-40. Vince, M.A. 1979
Postnatal Consequences of Prenatal Sound Stimulation in the Guinea-Pig. Animal Behavior. 27. 908-918.
Vince, M.A. and T.M. Ward 1984 The Responsiveness of the Newly Born Clun-Forest Lambs to Odor Sources in the Ewe. Behaviour 89. 117-127.
332 Vorherr, H. 1975
The Breast: Morphology, Physiology and Lactation. New York: Academic Press.
Vygotsky, L.S. 1962
Thought and Language. Cambridge, MA: MIT Press.
Wagner, R.L. and J. Pinchon 1962/1966 Grammaire du français. Paris: Hachette Walker, D.W., J.C. Grimwade and . Wood 1971 Intrauterine Noise: A Component of the Fetal Environment. American Journal of Obstetrical Gynecology. 109. 91-95.
Warden, D. 1976
Warden, D. 1981 Warren, J.M. 1977
Warrins, F. 1931/1960 Wason, P.C. 1966
Wason, P.C. 1968
Wason, P.C. 1983
The Influence of Context on Children's Uses of Identifying Expressions and References. British Journal of Psychology 67. 101-112.
Learning to Identify Referents. British Journal of Psychology 72. 93-99.
Handedness and Cerebral Dominance in Monkeys. In S. Harnad, R.W. Doty, J. Jaynes, L. Goldstein and G. Krauthamer (eds.), Lateralization in the Nervous System. New York: Academic Press.
Le jaseroque. In M. Gardner, The Annotated Alice. New York: Bramball House.
Reasoning. In B. Foss (ed.), New Horizons in Psychology. Harmondsworth: Pen guin Books.
Reasoning about a Rule. Quarterly Journal of Experimental Psychology 20. 273281.
Realism and Rationality in the Selection Task. In J.S.B.T. Evans (ed.), Thinking and Reasoning: Psychological Approaches. London: Routledge and Kegan Paul.
Wasz-Hockert, O., J. Lind, V. Vuorenkoski, T. Partanen and E. Valanne 1968 The Infant Cry: A Spectrographic and Auditory Analysis. Clinics in Develop mental Medicine 29. London: Heinemann.
REFERENCES
333
Weinberger and Diamond 1987 Physiological Plasticity in Auditory Cortex: Rapid Induction by Learning. Progress in Neurobiol. 29. 1. Weinrich, H. 1973 Weisseborn, J. 1981
Weir, R. 1966
Le temps. Paris: Seuil.
L'acquisition des prépositions spatiales: problèmes cognitifs et linguistiques. In C. Schwarze (ed.), Analyse des prépositions. 251-285. Tübingen: Niemeyer.
Some Questions on the Child's Learning of Phonology. In F. Smith and G. Miller (eds.), The Genesis of Language. Cambridge, MA: MIT Press.
Werker, J. and R.C. Tees 1984 Cross-Language Speech Perception: Evidence for Perceptual Reorganization During the First Year of Life. Infant Behavior and Development. 7. 49-63. Werker, J.F., J.H.V. Gilbert, K. Humphrey and R.C. Tees 1981 Developmental Aspects of Cross Language Speech Perception. Child Develop ment. 52. 349-355. West, R.F. and K.E. Stanovich 1982 The Source of Inhibition in Experiments on the Effect of Sentence Context and Word Recognition. Journal of Experimental Psychology: Learning, Memory and Cognition 8. 385-399. West, R.F. and K.E. Stanovich 1986 Robust Effects of Syntactic Structure on Visual Word Recognition. Memory and Cognition 14. 104-112. Wiggins, D. 1971
On Sentence-Sense, Word-Sense and Difference of Word-Sense. In D.D. Stein berg and L.A. Jocobovits (eds.), Semantics. Cambridge: Cambridge University Press.
Wing, G.S. and E.K. Scholnick 1981 Children's Comprehension of Pragmatic Concepts Expressed in "because", "al though", " i f and "unless". Journal of Child Language 8. 347-365. Wing, G.S. and E.K. Scholnick 1986 Understanding the Language of Reasoning: Cognitive, Linguistic and Develop mental Influences. Journal of Psycholinguistic Research 15. 383-401.
334 Wioland, F. 1972
Witelson, S.F. 1977
Witelson, S.F. 1982
Estimation de la fréquence des phonèmes en français parlé. Travaux de l'Institut de Phonétique 4. 177-204.
Anatomic Asymmetry in the Temporal Lobe: Its Documentation, Phylogenesis and Relationship to Function Asymmetry. Annals of the New York Academy of Sciences 299. 328-354.
Bumps of the Brain: Right-Left Anatomic Asymmetry as a Key to Functional Lateralization. In SJ. Segalowicz (ed.), Language Functions and Brain Or ganization. 117-114. New York and London: Academic Press.
Wollberg, Z. and J.D. Newman 1972 Auditory Cortex of Squirrel Monkey: Response Patterns of Single Cells to Species-Specific Vocalization. Science 175. 212-214. Wolff, P.H. 1963
Wolff, P.H. 1969
Observations on the Early Development of Smiling. In B.M. Foss (ed.), Deter minants of Infant Behavior. New York: Wiley and Sons.
The Natural History of Crying and other Vocalizations in Early Infance. In B.M. Foss (ed.), Determinants of Infant Behavior. London: Methuen.
Woods, J.R., M.A. Plessinger and E.C. Mack 1983 Fetal Auditory Brainstem Evoked Response (ABR). Pediatric Research. 18. 8385. Wright, R.V.S. 1972 Imitative Learning of a Flaked Stone Technology. The Case of an Orang-Utang. Mankind 8. 296-306. Wyke, M. 1971
The Effects of Brain Lesions on the Learning Performance of a Bimanual Coor dination Task. Cortex 1. 59-71.
Yeni-Komshian, G.H. and D.A. Benson 1976 Anatomical Study of Cerebral Asymmetry in the Temporal Lobe of Humans, Chimpanzees and Rhesus Monkeys. Science 192. 387-389. Zoloth, S. and S. Green 1979 Monkey Vocalizations and Human Speech: Parallels in Perception? Brain, Be havior and Evolution 16. 430-442.
Index A ability cognitive 2 metalinguistic 93, 95, 107, 134 perceptive 31 perceptual 2 absent referent 26-27 acalculia 17 acoustic cue 22, 50, 79 processing 50 stimulus 4, 20, 32-33, 35, 38 variation 20,51-52, 54 acquisition 1-7, 9, 16, 18, 22, 24-28, 38, 4446, 54, 57, 59, 64, 75-77, 125-130, 132-135, 137-138, 141-142, 158, 173, 177, 182, 190, 210, 213, 215, 223, 229230,232,243,250-251 anaphora 5, 11, 28, 131, 158, 160, 162, 173, 184 anaphoric organization 12 pronoun 5, 112, 122 relationship 24 anchor 147, 154-158, 182, 202, 231, 235, 240-241 antithesis 130 aphasia 13 approach interactive 59 metalinguistic 2 argumentative discourse 7, 245-246, 249-250, 253, 266-267 articulation 18, 20, 22, 47-48, 50, 57-60, 6365, 69, 71, 73-75, 78-84, 87, 145, 247, 262-263 rate 83-84 articulator 74 articulatory characteristic 66
gesture 74, 80 tempo 77 aspectual determination 151 asymmetry cortical 17 functional 16 manual 17 auditory functioning 37, 46 perception 20 system 20, 31-32, 36-37, 41, 43-46 babbling 2-4, 54, 58-66, 69, 71-78, 80, 82, 87 basic perceptive equipment 47 behavioral laterality 15 bimanual coordination 14-15, 17 bipedalism 15 boundary cue 86 cardiac response 38 categorical discrimination 19, 49 perception 19-20, 48-49, 54, 78 processing 20 symbol 25 category 5, 27, 38, 43, 46-47, 51, 53, 55, 78, 80-81, 85, 91, 93-95, 97, 99, 101-103, 110, 114, 140, 143, 146-147, 149, 204, 206, 210-212, 219, 221, 224-226, 229230, 240-242 causal conjunction 126-127 valence 110, 112 causality 13, 15, 110, 126-128, 130-131, 133, 145, 187, 220, 222-223, 246, 249, 255, 257 central nervous system 12
336
INDEX
cerebral lateralization 12 morphology 12 climax 232, 238-240 co-occurrence 256 co-text 157-158, 160, 162, 169-171, 181, 184 coda 145,240 cognition 1 cognitive ability 2 planning 144-145, 154 coherence condition 4, 6, 216 temporal 213 cohesion condition 4-5 discursive 6, 131, 157-158, 172-173, 182183 lexical 131 communication decontextualized 182 parent-child 88 competence epilinguistic 106, 108 metalinguistic 5, 107-108 competition model 59 complication-resolution 217, 241 concession 246, 255 condition coherence 4, 6, 216 cohesion 4-5 production 147 conditional 128, 258 conditioning 19, 37-38 consecution 15, 255, 257 conservation 25, 247 constant perceptual 75 speech 18 context-free situation 2, 97 verb 107 contour intonation 77, 80-81, 84-87 pitch 79-80, 82, 85, 87
conversational narrative 6, 144, 148-149, 156, 212 coreference 109-110, 112, 158, 160, 170-172, 184, 242 cortex primary auditory 22 temporo-parietal 22 cortical asymmetry 17 criterion event 216 cry repertory 76 cue acoustic 22, 50, 79 boundary 86 prosodic 39, 58, 60, 73, 75-76, 78 segmental 63 vocal 22 D decidability 251 declarative mode 24 statement 24 decontextualized communication 182 deixis 153, 158, 160-164, 168, 172-173, 179182, 184 designation 5, 23-24 development linguistic 9, 11, 13, 28, 43, 93-94, 134, 173, 183,210 semantic 1 dialogue 85, 88, 144, 146, 210 dichotic listening technique 21 discourse argumentative 7, 245-246, 249-250, 253, 266-267 cohesion 6, 131, 157-158, 172-173, 182183 function 127, 139, 141 situational 6, 136, 147, 155 strategy 147 theoretical 6, 136, 148-149, 156, 226 discriminability 19, 22, 36-38, 41-43, 57, 62, 75, 79, 84 discrimination categorical 19, 49 task 22, 61, 63
INDEX dislocated topic 177 displaced reference 182 E egocentricity 163, 183 ellipsis 131 enunciation theory 226 epilinguistic competence 106, 108 epilinguistics 5, 107 episode 144, 154-156, 183, 216-217, 221-222, 237, 239-241 equipment basic perceptive 47 initial 1, 75 perceptual-motor 3 F familiarization 31, 36, 41-42, 107, 114, 132, 134, 148, 220, 223 female voice 38-41 first mention 145, 160, 164-167, 171, 176, 178, 180-181 words 3 , 5 , 5 8 - 5 9 , 7 2 function discursive 127, 139, 141 opposition 141 verbal phrase 100 functional asymmetry 16 G genitive structure 27 goal 57, 146, 229, 256 grammatically 5, 89, 91, 93, 95, 99, 101-102, 106, 129, 154, 190, 232, 267 grip 15, 17 H habituation 37-38, 41, 80 handedness 12-15 hemisphere left 13, 15, 17, 21, 23 right 12-14, 17 hesitation 97, 145, 150 holophrase 93-94 hominization 11-12, 15
337
Homo sapiens 14 homology 12 hypothesis Independence 59 Semantic Feature 132 I image 26-27 implication 17, 130, 135, 160-161, 224, 246, 249-250 indecidability 251 Independence Hypothesis 59 inhibition effect 119, 122 initial equipment 1, 75 innate predisposition 28 intensity spectra 64 interactive approach 59 interlocutionary situation 211 intonation 4, 62-63, 71-72, 74-78, 80-82, 8488, 222 contour 77, 80-81, 84-87 pattern 59, 63 intratextual marking 233, 241 intrauterine noise 33 invariant 18, 20, 25, 60, 74, 144, 212 L label 95, 162, 164-168, 204 language development 9, 11, 13, 28, 43, 93-94, 134, 173, 183, 210 laterality 15 perception 31, 44 subject-prominent 177 target 3, 58-61, 65, 69, 71 topic-prominent 177 lateral dominance 23 laterality behavioral 15 linguistic 15 lax voice 61 learning morphological 108 positional 93 left hemisphere 13, 15, 17, 21, 23
338
INDEX
length syllabic 83 utterance 83 lexical access 119 cohesion 131 processing 22 linguistic development 9, 11, 13, 28, 43, 93-94, 134, 173, 183,210 environment 4, 57, 59, 77 long-term spectra 64 M macrostructure 144, 147, 154-156, 215, 226 male voice 34, 38-40 manual asymmetry 17 marker surface 226,241 syntactic 26-27, 202 maternal voice 4, 33-34, 38-40, 42, 80 memory 27, 134, 209, 211, 213, 220, 225 metalanguage 145 metalinguistic ability 93, 95, 107, 134 activity 99,202 approach 2 competence 5, 107-108 modalization 147, 155, 259-260 morphological learning 108 motor theory 20 N narrative conversational 6, 144, 148-149, 156, 212 dilatation 215,225 schema 7, 209-210, 212-217, 220-226, 230 Neanderthal 18 necessity 1-2, 11, 16-17, 25-28, 32, 37, 43-44, 57-59, 64, 73, 89, 107-108, 111, 118119, 126, 131, 133-135, 137-138, 147, 149, 154-158, 160-163, 173, 182-183, 187, 202, 211-213, 216, 220, 226, 246, 248, 251, 267
neologism 100, 102-104, 189-190, 192, 198, 200, 202-204, 206 neonate 3-4, 14, 18-19, 21-22, 31, 35-36, 3839, 41-42 news announcement 225, 230, 233-234, 240 nonsense 119, 189-190 normalization process 75, 84 ontogeny 11-13, 15, 32, 173, 229 opposition 24, 61, 77, 111, 119, 130-131, 135, 138, 140, 162, 173, 175, 210, 246, 248-249, 256 function 141 organization anaphoric 12 referential 11 syntactic 4-5, 12, 91 textual 7, 144, 147, 232 timing 63, 84 P paralinguistics 211 parent-child communication 88 parsing 20, 109, 229 pattern intonational 59, 63 prosodic 77-78 perception auditory 20 categorical 19-20, 48-49, 54, 78 linguistic 31, 44 sound 21 speech 20, 58-59, 75, 80 visual 18 perceptive ability 31 perceptual ability 2 constant 75 integration 74, 82 processing 4, 74 perceptual-motor equipment 3 phone class 74 phonetic categorization 4, 80, 102 categorization skills 4
INDEX
intelligibility 74 repertory 62 phonological representation 71 rhythm 79 phylogeny 11-12,20,22,28-29,57 pitch 58, 63-64, 75-76, 79-82, 84-87 contour 79-80, 82, 85, 87 value 79 pivot word 5 planum temporale 12-13, 17 pointing 23-24, 99, 160, 165-167 positional learning 93 possibility 2, 12, 19-20, 22, 24, 28, 33, 63, 74, 80, 84-85, 94, 102, 107, 109, 111112, 114, 116, 122-123, 125, 128, 142, 146, 163, 171-172, 182, 184, 195, 198, 204, 212, 215, 219, 226, 246, 249, 251, 258, 263, 267 possible world 247 potato peeling 15 predication 12, 24-25, 94, 137, 162, 164, 169, 172, 175, 177, 184-185,231 predicative statement 24 preferential reading strategy 111-112 prenatal experience 39-40, 44 prespeech 57-58,60-61,71-72 presupposition 1, 87, 107, 125, 129, 132, 135138, 159, 161-164, 169, 171-173, 181, 183-184 primary auditory cortex 22 processing acoustic 50 categorical 20 level 20 lexical 22 perceptual 4, 74 speech 43, 76, 78, 84 speech sound 43 temporal information 83 production condition 147 speech 12, 17,20-21,43,87 textual 4, 146-147, 210, 229 pronoun anaphoric 5, 112, 122
339
referent 109-113,115,122-123 prosodic cue 39, 58, 60, 73, 75-76, 78 pattern 77-78 prosody 40, 42, 58, 60-61, 63-64, 78 punctuation 7, 146, 148, 151, 184, 232, 235, 238, 256, 267 R recursiveness 125 referent absent 26-27 pronominal 109-113,115,122-123 referential organization 11 system 6, 157-158, 163, 172, 180-181, 183 value 23 referentiation 211 relationship anaphoric 24 topic-comment 28 repertory cry 76 phonetic 62 rhythm phonological 79 speech 78-79, 87 right hemisphere 12-14, 17 S schema construction 216, 220 narrative 7, 209-210, 212-217, 220-226, 230 script 134, 215, 217, 219, 222-223, 235, 237238, 242 segment 57-58, 66, 72, 74, 79, 83, 86-87, 107, 145 segmental cue 63 segmentation 20-21, 75, 82, 85-87, 95, 144145 semantic bias 110-111 development 1 Semantic Feature Hypothesis 132
340
INDEX
sentence context 118-119 organizer 5, 93, 96-97, 99 sequenciation 75 sequentiality 109,130,211 setting 7, 58, 61, 69, 71, 211, 218, 221-222, 226, 229-232, 237, 248, 253, 266 shifter 24 sign language 21, 24 silence 18, 20-21, 187 situation context-free 2 interlocutionary 211 situational discourse 6, 136, 147, 155 social representation 210 sound perception 21 speech 3-4, 36, 41-44, 57, 60, 66, 71, 73, 75,83 sounds of silence 18, 20-21 spatio-temporal framing 216, 221, 226 speaking rate 18, 78, 83-84 spectra intensity 64 long-term 64 speech constant 18 perception 20, 58-59, 75, 80 processing 43, 76, 78, 84 production 12, 17, 20-21, 43, 87 rhythm 78-79, 87 signal 4, 21, 74-75, 84-85 sound 3-4, 36, 41-44, 57, 60, 66, 71, 73, 75,83 sound processing 43 timing 76-77, 79, 82 statement declarative 24 predicative 24 story 40-41, 137, 164, 184, 193, 196, 202, 210, 213, 215-217, 219-221, 223, 225, 229-235, 237-242, 248-249, 251, 262, 265-266 storytelling 209-210, 226, 267 stress beat 79
system 61 structure genitive 27 SVO 2, 127 syllabic 62, 82 topic-comment 27, 175 subject-prominent language 177 subordinate clause 110-111, 133, 135 substitution 131, 199 superstructure 213, 225 supralaryngeal vocal tract 18 surface marker 226, 241 SVO structure 2, 127 syllable 19-20, 36, 38, 62-65, 74, 78-79, 8186 duration 77, 81, 83 format 62 length 83 lengthening 84, 87 structure 62, 82 sylvian sulcus 13 syntax 1, 5, 25-26, 28, 99, 130, 190 marker 26-27, 202 organization 4-5, 12, 91 system auditory 20,31-32,36-37,41,43-46 referential 6, 157-158, 163, 172, 180-181, 183 stress 61 tone 61 vocalic 18 T target language 3, 58-61, 65, 69, 71 temporal coherence 213 connective 133 information processing 83 phonetic information 79 structure of speech 75, 79 variation 83 temporo-parietal cortex 22 tense voice 61 text linguistics 3, 144 organization 7, 144, 147, 232
INDEX
production 4, 146-147, 210, 229 textualization 145-146, 153 theoretical discourse 6, 136, 148-149, 156, 226 theory enunciation 226 motor 20 timing organization 63, 84 speech 76-77, 79, 82 tone system 61 tools 14-17, 25, 182, 210 topic-comment relationship 28 structure 27, 175 topic-prominent language 177 topic-shift 179 topical unity 231,235 translation 25, 27, 58-59, 185, 190, 212, 226, 237-238, 241 U universal grammar 27 utterance length 83 V value pitch 79
referential 23 variation acoustic 20, 51-52, 54 temporal 83 verbal phrase function 100 visual perception 18 vocal cue 22 system 18 vocalization 19 voice envelope 75 female 38-41 lax 61 male 34,38-40 maternal 4, 33-34, 38-40, 42, 80 onset time 17-18 tense 61 W word first 3, 5, 58-59, 72 meaning 1, 190 order 5, 93, 97 pivot 5
341