The Child’s Communicative Competence: Language Capacity in Three Groups of Children from Different Social Classes 9783110802047, 9789027925954

Table of contents :
PREFACE
CONTENTS
1 THE PROBLEM
2. THE EXPERIMENT
3. RESULTS
4 DISCUSSION
5. SUMMARY
REFERENCES
APPENDICES

Citation preview

JANUA LINGUARUM STUDIA MEMORIAE N I C O L A I VAN WIJK D E D I C A T A edenda curai C. H. V A N S C H O O N E V E L D Indiana University

Series Minor,

202

THE CHILD'S COMMUNICATIVE COMPETENCE Language Capacity in Three Groups of Children from Different Social Classes

by

TON VAN DER GEEST, RUDIE GERSTEL, RENÉ APPEL, BERNARD TH. TERVOORT University of Amsterdam

1973 MOUTON THE HAGUE • PARIS

© Copyright 1973 in The Netherlands Mouton & Co. N.V., Publishers, The Hague No part of this book may be translated or reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publishers.

LIBRARY OF CONGRESS CATALOG CARD NUMBER: 73-80962

Printed in Belgium by NICI, Ghent

PREFACE

The research described in this monograph has been executed by a project-group of students and members of the teaching stafl of the Institute of General Linguistics of the University of Amsterdam. Initially the group intended to evaluate the effects of a language compensation program — a Dutch adaptation of the EngelmannBereiter program. Because no data of the children's language capacity at the time of the program's start were available, the group changed its goal to investigating the language capacity of three sociologically different groups of three- to four-year-old children. We hoped on the basis of our results to formulate some suggestions with respect to the language educational program. The results, however, were such that we restricted ourselves to discussing the problem of language compensation in general, and the differences in language behavior of the three experimental groups in particular. Both topics have been discussed in terms of two opposing sociolinguistic theories: a) the deficiency theory which presupposes that the language of the lower class children is retarded in comparison with the language of middle class children; b) the difference theory which presupposes that the language of both groups of children is different but essentially equivalent. We intend to make clear our position that the latter theory is the more reasonable one. A central part of this monograph is occupied with the problem

6

PREFACE

that recent linguistic findings cannot without adaptations measure syntactic maturity. Probably this is why syntactic complexity of language use is rarely dealt with systematically and exhaustively. For measuring syntactic complexity, this study proposes a method which implements a score list by which one is able to make a systematic and almost exhaustive inventory of the semantic and syntactic phenomena of the three to four-year-old child's language use. The communicative aspect of the competence — one's capacity to react appropriately by means of language within a certain context and situation — has also been systematically covered in the score list. A factor analysis on the scores of syntactic complexity resulted in 36 variables that are crucial for the determination of a child's communicative language capacity. Finally, the authors argue for a pure observational method of data sampling. The monograph is organized in the following way: — In chapter one the theoretically linguistic, sociolinguistic, and psycholinguistic assumptions are discussed in some detail. A number of conclusions that are of special interest for the actual project are presented at the end of each separate section. — In the second chapter the setup of the experiment is reported and the newly designed syntactic complexity score is discussed in detail. — In the third chapter the findings and conclusions are reported. Ample attention is paid to the linguistic interpretation of the results of two factor analyses. — In the fourth chapter the theoretical assumptions of the first chapter are compared with the findings of chapter three. Some tentative conclusions with respect to theoretical issues and to educational problems are drawn. It would be impossible for us at this point to acknowledge in detail the contribution that our students have made. We would like to thank especially Margreet Beunderman, Lilian Doorenbosch, Anita Koster, Dorine Plantenga, Max Verbeek, and Evie

PREFACE

7

Visch. We are sincerely grateful to Mr. J. de Leeuw, University of Leiden, for making available and adapting the factor analysis programs. We thank Dr. Catherine Snow for the many valuable criticisms and corrections of the preliminary version of this book, as well as Mrs. Anneke Vermeer for the typing of the text, and Mr. B. H. Kaart for statistical and administrative assistance. The authors Amsterdam, November 1972

CONTENTS

Preface

5

1. The problem 1.1. 1.2. 1.3. 1.4.

The notion of competence The measurement of children's language capacity . Message versus code Language capacity and social environment . . . .

2. The experiment 2.1. The composition of the groups 2.2. The collecting of the language samples 2.3. The language measures used 2.3.1. Traditional quantitative measures 2.3.2. Traditional qualitative measures 2.3.3. The new Syntactic Complexity Score . . . . 2.3.4. Some other measures 2.3.5. The sentence as basis of the analysis . . . . 3. Results 3.1. Factor analysis of the measures of syntactic complexity 3.2. Differences among the groups 3.2.1. Social class differences 3.2.2. Environmental differences on the traditional language measures 3.2.3. The relation between the measures of

11 11 17 22 24 33 33 35 37 37 39 40 49 51 53 53 62 62 64

10

CONTENTS

syntactic complexity and the different quantitative measures 3.2.4. Developmental differences between the social classes 3.2.5. Environment and play contacts 4. Discussion 4.1. Linguistic competence versus communicative competence 4.2. Qualitative vs. quantitative measures 4.3. The planning factor 4.4. The role of pragmatics in linguistics, sociolinguistics, and psycholinguistics 4.5. Some tentative conclusions regarding language compensation programs 4.6. The hereditary factor 5. Summary

66 67 75 79 79 86 91 93 100 103 105

5.1. The design of the project

105

5.2. The results

107

References

Ill

Appendices

115

1 THE PROBLEM

The purpose of this study is to describe the language capacity of three- to four-year-old children from three socio-economic backgrounds. Since LANGUAGE CAPACITY has been used with different meanings, we will first attempt to define the term as used in this study. By language capacity we mean not only verbal skills, such as those measured by quantitative measures of PERFORMANCE (e.g. mean length of utterance), but also the underlying COMPETENCE which must be analyzed qualitatively. This short characterization unfortunately remains inadequate because of imprécisions and inconsistencies in the definitions of competence and performance as given by linguists, sociologists, and psychologists over the last several years. We will, therefore, discuss the concepts of competence and performance in the next section. In the following chapters we will discuss some other basic problems in describing children's language capacity: (1) How does one describe language capacity in general and, more specifically, in children ? (2) What aspects of language capacity distinguish social class ? (3) The traditional measures of language skill are limited. What sort of test or measuring device is most appropriate as an index of language capacity ? 1.1. THE NOTION OF COMPETENCE

It is becoming more evident that there is a big gap between the standpoint of the linguist speaking about competence and per-

12

THE PROBLEM

formance and that of the psycholinguist and sociologist describing the puzzle called language as a phenomenon in reality, accounting for the proficiency in the processing of linguistic information, and (or) measuring the individual's or group's progress in the development of language. Chomsky 1 defines competence as the idealized knowledge of the language possessed by the "speaker-hearer", and performance as the actual usage of language in concrete situations. These definitions are too naive for any use beyond linguistcs, as is apparent from the fact that performance is composed of at least two facets: (a) the process of producing and understanding language, making use of the rules pertaining to the linguistic competence; and (b) the result of this process in actual language use — that which results from the process in actual uttering and understanding. Fodor and Garrett 2 point to another inconsistency in the competence-performance distinction: they argue that there are two (psycho)linguistic notions of competence, and therefore also two notions of performance. They posit in general that competence is the counterpart of behavior. Second, they argue that the notion is used in a rather limited sense, namely as "linguistic capacity independently of the other psychological mechanisms and competences with which linguistic capacity must be supposed to interact in the production of verbalizations". This means that there exists a distinction between competence and behavior and between the linguistic competence and such non-linguistic capacities as memory, perception, and the like. The latter distinction is taken as the starting point in modern transformational linguistic research, in the sense that linguists deal exclusively with the linguistic competence. How the influence of linguistic competence is brought to bear in the production and understanding of language is left to the psychologist to solve. 1

N. Chomsky, Aspects of the Theory of Syntax (Cambridge, 1965), p. 4. J. Fodor and M. Garrett, "Some Reflections on Competence and Performance", in Psycholinguistics Papers (Chicago, 1966).

2

THE PROBLEM

13

Psychological research on the latter question was carried out initially on the basis of the so-called correspondence hypothesis. According to this hypothesis, the linguistic rules proposed for the grammatical derivation of the sentence and the order in which they are applied is supposed to correspond step by step to the mental processes executed when somebody produces a sentence.3 This hypothesis is also present in language acquisition research, e.g. in the studies of McNeill,4 when he discusses the hierarchy of categories and the development of the basic grammatical relations, and in Brown and Hanlon's 5 article on derivational complexity in relation to the developmental sequence in language acquisition. In this article, Brown and Hanlon limit themselves to the cumulative derivational complexity according to the principle that "when the derivation of a sentence Y follows all the rules applied in the derivation of a sentence X plus at least one rule not applied in X then Y has greater cumulative derivational complexity than X". This correspondence hypothesis is not in accordance with Chomsky's views: "...it seems absurd to suppose that the speaker first forms a generalized Phrase marker by base rules and then tests it for well-formedness by applying transformational rules... . But this absurdity is simply a corollary to the deeper absurdity of regarding the system of generative rules as a point-by-point model for the actual construction of a sentence by a speaker." 6 It is, therefore, not so surprising that the correspondence hypothesis in its strict sense has been abandoned. Modern psycholinguistic research to a large extent focuses upon the discovery of sentence processing strategies. These strategies are shortcuts from surface 3

J. R. Hayes, "Introduction", in J. R. Hayes (ed.), Cognition and the Development of Language (New York, 1970). 4 "Developmental Psycholinguistics", in F. Smith and G. A. Miller (eds.), The Genesis of Language (Cambridge, Mass., 1966); and Language Acquisition, the Study of Developmental Psycholinguistics (New York, 1970). 5 R. Brown and C. Hanlon, "Derivational Complexity and Order of Acquisition in Child Speech", in J. R. Hayes (ed.), Cognition and the Development of Language (New York, 1970), p. 13. 6 N. Chomsky, Aspects of the Theory of Syntax, p. 139.

14

THE PROBLEM

structures to deep structure semantic relations which enable one to avoid some of the complexities of a complete derivational analysis. Two important problems arise in this type of research. The first originates from the fact that ambiguous sentences are usually used as testing material. If a subject is conscious of the possibility of double interpretation of a sentence, he finds himself in an experimental situation which cannot be compared with normal conversation. The second problem concerns the influence of context and situation in the processing of language. Given a normal conversational situation, the hearer can process sentences that might be only partially perceived and incorrectly or even ungrammatically realized. Moreover, in the conversational situation, sentences to which no linguist cares to pay attention, and for which he consequently does not give any rules, occur frequently; in the experimental situation these contextually-determined sentences would be uninterpretable and impossible to process. Examples are: (1)

Thank you, Mommy (I — Mommy — request of you that you say "Thank you, Mommy");

(2)

Coffee!¡.¡1

(Would you care for a cup of coffee?/ Coffee is the only thing I would like to have./ Look, there you can get coffee!/ etc.);

(3)

May I?

(May I take a cigarette?/ etc.).

Context and situation clearly should be taken into account as important factors in the processing of language. It would lead, however, to enormous complications in the setup of an experiment to incorporate these factors systematically. In the experimental situation, context and situation are missing or are inadequately presented by means of a picture or a story. The conclusions of such experiments are only valid for the processing of sentences which are well-formed and whose interpretation requires a minimum of context. Context, an extremely important variable, must

THE PROBLEM

15

be accounted for if one sets out to describe or measure language capacity both qualitatively and quantitatively. Supplementary support for this standpoint is provided from sociolinguistics by a recent development in the competence theory — COMMUNICATIVE COMPETENCE. This term was first used by Hymes and his group, and Chomsky's notion of "the creative aspect of language use" is central to it. In Aspects7 Chomsky states, "Thus an essential property of language is that it provides the means of expressing indefinitely many thoughts and for reacting appropriately (our italics) in an indefinite range of new situations." In other words, the focus of attention has to be shifted from the notion that: (a) a language is a set of sentences, and it belongs to the competence of the speaker/hearer to be able to understand and to produce all these sentences, at least in principle; to the idea that: (b) a language is a set of sentences. The speaker is capable of producing at any time whichever of these sentences he desires to produce. The hearer is capable of interpreting all sentences of that language given a certain context. The formulations (a) and (b) each have their consequences for the competence of the hearer and the speaker. For the hearer, it is stated in (a) that he understands the sentences: (4)

(5) (6)

I did it this way; Yes; Tomorrow;

etc. To a certain extent this is, of course, true. In (4) the hearer knows that I is the subject (the actor), it is the object, and the action is carried out in a certain way. In opposition to (6), for (4) it is possible to establish an interpretation of the sentence. The I is the 7

N. Chomsky, Aspects of the Theory of Syntax, p. 6.

16

THE PROBLEM

speaker in a conversational situation, the action and the object it apparently are given in the situation, and the way it was done may be expressed through gestures. Although the sentences (4), (5), and (6) can be generated in principle within both versions (a) and (b), the essential difference is that in case (a) it is impossible to formulate explicitly the conditions which context and situation have to meet. It is impossible, for example, to state explicitly that (4) has to be used and not its semantic equivalent (4a) I shaved it clean, which is possible if (b) is taken as point of departure. According to Hymes,8 an adequate theory of language usage has to make four types of judgements about the speaker: 1. the judgement of grammaticality and acceptability of occurring sentences (linguistic judgement); 2. the judgement as to whether a certain construction or rule can be realized (psychologic judgement); 3. the judgement of a sentence as to its adequacy within context and situation (communicative judgement); 4. the judgement as to what extent a linguistic phenomenon is used, and what this usage implies in terms of social valuation, intelligence, and development. These judgements are of essential importance in establishing someone's language capacity. The foregoing shows the central position of the speaker and thus of language production in the theory of communicative competence. Yet one should not forget that language capacity includes both comprehension and production, which are not necessarily each other's counterpart. This is what makes the traditional Transformational Generative model so useless in the psycholinguistic and sociolinguistic sense. Its insistence on the inseparable "speaker-hearer" unity makes it impossible to account for the 8

D. Hymes, "Competence and Performance in Linguistic Theory", Language Acquisition: Models and Methods (London, 1971), p. 12.

THE PROBLEM

17

irreducible differences between speaking and understanding. As an illustration of this, Hymes 9 reports that a certain group of lower class Negro children could understand but could not speak standard English. For the time being it is not evident, however, how this perceptual aspect can be sufficiently, effectively, and exhaustively dealt with in the theory. Precisely because context and situation are so crucial to the proper exercise of communicative competence, evaluations of communicative competence can only be made in observational, not experimental, situations. Observational methods limit one to studying production and not comprehension. We draw the following conclusions concerning the definition of competence: A. Competence has linguistic as well as psychological and sociological aspects, i.e. there are both the knowledge of rules and structures and the use of this knowledge in speaking and understanding; B. Within psychological competence the abilities for perceiving and producing have to be distinguished from each other, which is not true for linguistic competence; C. Psychologists focus mainly upon perception research in experimental situations and this has numerous drawbacks; D. Sociologically oriented linguists focus mainly upon production which is easier and safer to investigate than comprehension, especially where it is more a matter of measuring, comparing, judging, and evaluating the language capacity. Thus we think that Hymes' principles can be usefully applied in our attempt to assess language capacity on the basis of speech production. 1.2. THE MEASUREMENT OF CHILDREN'S LANGUAGE CAPACITY

Our observations concerning measurement of language capacity 9

D. Hymes, "Competence and Performance in Linguistic Theory", p. 8.

18

THE PROBLEM

in general hold true for the measurement of language capacity in children, with one major difference. In children, production has been intensively studied but little attempt has been made to study receptive abilities. In literature on the acquisition of language, it is generally accepted that children of 36 months and older have a command of language that can be described with the conceptual framework and the rule mechanism of adult language. This means in practice that the basic grammatical relations (see McNeill10) and transformations (see Bellugi 11 and Menyuk 12) are supposed to be present. Both generative rules and transformations are still developing, of course, but they are approaching the adult norms. Seen in the light of what has been worked out before, however, it should be stated that the description of an inventory of which rules are present in the competence of the child is limited to deductions made from his production. The communicative aspect of his competence is not accounted for. With regard to the establishment of the language capacity of the three- to four-year-old child, the following additional remarks should be made: A. Only Menyuk has tried to give a systematic and exhaustive analysis of the language of the three- to four-year-old child. For critical reviews of her proposals we refer to Bloom, 13 Van der Geest,14 and Shipley.15 From these it becomes evident that her descriptions are too sketchy, that they are based on a theory which is not fully correct linguistically, and that they do not penetrate deeply enough into the real process of language acquisition. An additional objection is that her analyses deal with isolated sentences instead of sentences which are functioning within context 10

See note 4. U. Bellugi, "The Emergence of Inflections and Negation Systems in the Speech of Two Children" (unpubl. Ph. D. thesis, 1964). 12 P. Menyuk, Sentences Children Use (Cambridge, Mass., 1969), Chapter 3. 13 L. Bloom, "Paula Menyuk's Sentences Children Use", Contemporary Psychology 15, pp. 182-84. 14 A. van der Geest, "Zinsstructuren die kinderen gebruiken", Ned. Ts. voor de psychologie (1971). 15 E. Shipley, "Review of Sentences Children Use", Language 46, pp. 931-38.

11

THE PROBLEM

19

and situation - a usual procedure in linguistic analyses of child language. The only exception to the latter condition is Bloom (see for discussion Van der Geest16). B. Brown, Cazden, and Bellugi 17 have pointed out that it is impossible to arrive at competence assessment on the basis of the frequency of occurrence of a certain rule in the production of the child. Their argument is based on differences in frequency of certain verbs being first used by the child all at the same time. This is not a matter of grammatical structures but of peculiarities in the use of the lexicon; words are more dependent on someone's communicative needs than the use of the subject-predicate distinction. In other words, both the usage of to bite or to analyze requires the realization of a subject, a verb, and an object. When it is purely a matter of making an inventory of the rules that the child has mastered, it can generally be maintained that frequency of occurrence does not matter much. The analysis in such cases is limited to the linguistic competence, i.e. to judgement (1) of Hymes. In terms of communicative competence, frequency really does tell us something about the communicative power of rules that are obviously known. As Hymes puts it in his discussion of judgement (4), probabilities in language usage are indeed indications of style, adequate response, and the like. Furthermore, "one must provide for the distinction between the marginally and merely possible and the actual normal: between what one will accept as a hearer and what one will produce as a speaker". The results of our experiment indicate that the frequency of production of certain structures in successive periods of child language are subject to systematic changes. These changes can be used as measuring sticks for the language development. C. A final point that is relevant for the stating of our problem deals with the relation and, eventually, the interrelation between 18

A. van der Geest, Evaluation of Theories on Child Grammars (Amsterdam, 1972), Chapter 3. 17 R. Brown, C. Cazden, and U. Bellugi, "The Child's Grammar from 1 t o 3", Minnesota Symposia, J. P. Hill (ed.) (1969), pp. 64-65.

20

THE PROBLEM

perception and production. The opinion that a child's comprehension is broader than his production is generally accepted. Experimental research, especially that by Fraser, Bellugi, and Brown,18 has reinforced this opinion. Their results support intuition: children understand sentences which they are not yet able to produce. This result, however, is not relevant to our understanding of the nature of the perception and the production. There are, moreover, a number of considerations that rather reduce the relevancy of such a statement. All parts of a sentence must be realized in production. This frequently is the big problem for the language-producing child, for only fragments of the intended sentence are realized. The absence of certain parts of the sentence does not necessarily indicate shortcomings in the underlying linguistic deep structure, but rather shortcomings of a more neuropsychological nature such as a language organizing capacity of the central nervous system which has not yet matured enough. Psychologically this is evidenced in a limited memory span. On the other hand, for the understanding of sentences it is not necessary that all parts of the sentence are perceived in order to arrive at the correct interpretation: morphological cues, function words, and the like are less relevant for the interpretation than content words. It is frequently sufficient that that part of the sentence that contains the most essential information (the comment) is picked up. It therefore makes sense to start from a memory span that has not been matured too much as yet when one is dealing with perception (for a tentative sketch on what kind of strategy the child uses for the perception of sentences see Tervoort, et al.19). This usually means that measuring the perceptive capacity of children is a risky affair: the intention of the sentence presented comes across most of the time, although only fragments are perceived. Perhaps this is one of the reasons why psycholinguists

18

C. Fraser, U. Bellugi, and R. Brown, "Control of Grammar, in Imitation, Comprehension, and Production", JVLVB (1963), pp. 121-35. 19 B. Tervoort, et al., Psycholinguistiek ( = Aula Pocket 481) (Utrecht, 1972), Chapter 2.

THE PROBLEM

21

usually tend to deal with the production side of language acquisition research. Another danger inherent in the evaluation of comprehension of speech is that often the perception abilities are overrated because context and situation are not taken into account. Clark 20 demonstrates that a child brought into a situation where the prepositions up and down could be used in a sentence did not appear to evaluate this difference in his behavior. This problem can be easily explained by taking context and situation into consideration. Standing at the bottom of the stairs a parent will not usually say go down the stairs, or we are going down the stairs. In other words, within the situation what the parent means is obvious. This contrasts with a situation in which a child can either go up or down the stairs. Our own observations indicate that the alternative put the papers on top of/under the table, when presented to a child who is accustomed to the papers always being under the (reading) table, invariably evokes the response that the child puts the newspapers under the table. In such cases the child often is more strongly influenced by context and situation than by the actual utterance. Thus it remains problematical whether the perception of the child surpasses his production. The question is whether it is feasible that these difficulties are accounted for in one's design. As an illustration of this question, an experiment of Shipley, Smith, and Gleitman 21 could serve well. They started from the position that a child had responded adequately if he just looked at the ball after the sentence Throw me the ball had been presented to him. Suppose, however, that the child had indeed thrown the ball, then we still would not know if the child had indeed perceived throw, for it is quite possible that the throwing of the ball is the normal response for the child to the stimulus ball. In summary, the following points are of essential importance for our investigation: 20

R. Clark, "Comprehension and Production in Language Acquisition", Work in Progress (Edinburgh University, 1972), pp. 91-106. 21 E. Shipley, et al., "A Study in the Acquisition of Language", Lg (1969), pp. 322-43.

22

THE PROBLEM

E. Unlike the researcher engaged in establishing the linguistic competence of the child, one cannot limit oneself to an inventory of linguistic structures when setting out to establish the communicative and psycholinguistic competence even when not yet taking context and situation into account. Frequencies of occurrence of these structures in the data are also of vital importance for the assessment of the child's language capacity. F. The analysis cannot be limited to isolated sentences: all utterances of the child will have to be evaluated for adequacy within the context. G. Describing the perception of the child more or less exhaustively in a systematic way is probably impossible for much of the perception will be beyond the reach of the investigator.

1.3.

MESSAGE VERSUS CODE

Bloom 22 demonstrates that when the child's sentence is evaluated within context and situation it becomes apparent that children know considerably more about their language than can be derived from their sentence realization. On this basis she makes a clear distinction between the SEMANTIC INTENT, i.e. the intention of the sentence, and the REALIZATION thereof in the surface structure. Intuitively, she comes surprisingly close to the recent developments in linguistics referred to as GENERATIVE SEMANTICS. Generative semantics developed in opposition to more traditional interpretative semantics as posited, for example, by Katz and Postal. Katz and Postal (1964) formulated the hypothesis that all semantic information is represented before the transformations start but for the interpretation that underlies the combination of sentences. This latter information is supplied by the "sentence combining transformations", and is regulated by the deep structure 22

L. Bloom, Language Development: Form and Function in Emerging Grammars (Cambridge, Mass., 1970).

THE PROBLEM

23

in Chomsky. 23 Various observations led to the later abandonment of the model as worked out by Chomsky. The point of departure then became the idea that even during the transformations all kinds of specifications of meaning were appearing. Generative semantics, on the contrary, has from the outset adhered strictly to the condition as put by Katz and Postal and Chomsky. They maintain a basic distinction between semantic information on the one hand and the sentence realization on the other. Thus Fillmore 24 argues that a deeper and more essential structure has to be postulated than the phrase structure accepted thus far. According to his proposals, such a structure has to be made up of a verb and a number of semantically marked constituents such as the actor, the dative, and the objective. McCawley's basic point of departure is that language essentially is a code through which semantic cognitive material (the message) is linked with encoded forms (the surface structure). One of the consequences of such a standpoint is that syntax ceases to exist as a branch in linguistics in its own right: the ways words can be linked together become purely the result of restrictions on the ways semantic material can be combined, and of transformations which are characteristic within a certain language for the conversion of semantic information into sentences.25 A bipartition between semantics - the linguistic equivalent of cognition - and the realization is also of interest for our problem. When the language capacities of two different individuals or groups are going to be compared, it is quite possible that eventually identical intentions will be uttered in different manners, and, specifically, in dependency of speech habits or evironment bound behavior. The differences that occur can be of a sociological nature; they can also be determined by the language development. Differences in the frequency of realization of certain structures may also 23

N. Chomsky, Aspects of the Theory of Syntax, p. 134. C. Fillmore, "The Case for Case", in E. Bach and R. J. Harms (eds.), Universals of Linguistic Theory (New York, 1968), pp. 1-88. 25 J. D. McCawley, "Where do Noun Phrases Come From?" (Chicago, 1967(7), unpubl.). 24

24

THE PROBLEM

go back to differences in conversational topics and the intention of the child. Precisely in a sociological viewpoint this kind of difference is interesting, as will become evident at a later stage. Such type of information gets lost if one takes an interpretative semantic standpoint as the starting point. Therefore, we have chosen the standpoint of generative semantics for our project.

1.4.

LANGUAGE CAPACITY A N D SOCIAL ENVIRONMENT

The difference in language use between children coming from diverse social environments has been subject to extensive investigations during the sixties. It is obviously an oversimplification to equate language use with language capacity. These investigations were inspired pragmatically by the generally poor school achievement of children from lower social classes (LC), which, in their turn, were blamed on the children's limited language capacity. It was hypothesized that these children could not express themselves adequately and could not understand expressions of subtle and abstract thought partly because of the limitations in their language capacity. Theoretical inspiration was drawn from the writings of Basil Bernstein, although he had been dealing more with adult language use.26 He distinguishes a RESTRICTED CODE and an ELABORATED CODE. Members of the LC have access to the restricted code only; those who have a middleclass (MC) background have both codes at their disposal. The codes are based upon the social relations within the group. Where there is a high group solidarity, as in LC and in MC in informal, intimate situations, the individual aspect in language use has to be restricted as much as possible: a restricted code thus originates. Specific intentions, opinions, and ideas need not to be expressed because they are shared by everyone in the group. In consequence, the restricted code is characterized by a low syntactic and lexical diversity, and by the 26

B. Bernstein, "A Sociolinguistic Approach to Socialization", in F. Williams (ed.), Language and Poverty (Chicago, 1970), pp. 25-61.

THE PROBLEM

25

implicit, particularistic, and concrete features of the meanings that are transferred. Where in a culture the individual aspect gains significance over the social (for example, in the MC culture), the elaborated code comes into being: speakers have to utter feelings and opinions specifically because it can no longer be assumed as a matter of course that these are shared by other members of that culture. The elaborated code can be characterized as a flexible type of language use with a great number of syntactic and lexical possibilities suited to carry across abstract, universalistic, and explicit meanings. Bernstein emphasises that the restricted code should not be evaluated as qualitatively inferior to the elaborated code: "Lest the restricted code be misinterpreted as simply poor language, we must be aware that it contains a vast potential of meanings. It is a form of speech which symbolizes a communally based culture. It carries its own aesthetic. It should not be disvalued."27 The latter, however, was indeed done by the investigators dealing mainly with the language of the child as related to school achievements. They stated that the language of the LC children was "deficient" or "deprived". This was maintained in its most extreme form by Bereiter and Engelmann 28 who considered the young LC child as practically non-verbal. This type of child, according to Bereiter and Engelmann, hardly knows single words and treats groups of words as one word ("giant words"), whereby the number of possible syntactic combinations is radically reduced. These children lag in language development, according to Bereiter and Engelmann, because of environmental deficiencies. Children from a LC environment experience far less verbal interaction with their mothers at an early age, whereas it is just this interaction which is so important for the acquisition of language and cognitive development (see for example Hess and Shipman29). Apart from this, the world of experience of these children is 27

B. Bernstein, "A Sociolinguistic Approach to Socialization", p. 37. C. Bereiter and S. Engelmann, Teaching Disadvantaged Children (1966). 29 R. Hess and V. Shipman, "Early Experience and the socialization of Cognitive Modes in Children", Child Development (1965), pp. 869-86. 28

26

THE PROBLEM

limited; the number of stimuli in their environment is low and unorganized to the extent that the development is minimally activated. 30 Moreover, since the home environment is usually busy, noisy, and overcrowded, presumably language acquisition slows down. Against this, Schultz and Aurbach 31 argued that the noisy, busy environment may well provide valuable stimulation for the LC child's development. A completely opposite opinion on the differences in language capacity between LC and MC children can be found among those who are sometimes termed the DIFFERENCE THEORISTS,32 among whom are Labov, Baratz, and Stewart. They claim that the language of the LC child does not lag behind, as compared to that of the MC child, but that it is only different. They are of the opinion, first, that this specific "being-different" has not been well enough acknowledged and identified by investigators such as Deutsch, and, second, that it is incorrect to demand of the LC child that suddenly at school he has another language capacity at his disposal, namely, that of the MC child. Although Bernstein's theories have been used extensively ("abused", Bernstein himself puts it) by adherents of the deficiency theory, they actually come much closer to the difference theory. Bernstein himself 33 states that the language capacity of LC children is potentially equal to that of MC children, but they put it to use in a different way. He, moreover, is of the opinion that children cannot be expected to drop the restricted code, based upon their own culture, at the school door and use an elaborated code, based upon another culture, once they are within the school. Houston 34 rejects the deficiency theory. She argues that it is 30

See M. Deutsch, et al., The Disadvantaged Child (New York, 1967). C. Schultz and H. Aurbach, "The Usefulness of Cumulative Deprivation as an Explanation of Educational Differences", Merrill-Palmer Quarterly 17 (1971), pp. 27-39. 32 See M. Bruck and G. Tucker, "Lower Class Language" (unpubl., McGill University, 1972). 33 B. Bernstein, "Education Cannot Compensate for Society", New Society (1971), pp. 344-47. 34 S. Houston, "A Reexamination of Some Assumptions about the Language of the Disadvantaged Child", Child Development (1970), pp. 947-63. 31

THE PROBLEM

27

based on results that were obtained with the help of research methods that are inappropriate and unfair to the LC child. Labov, too, is convinced that the language usage as elicited by researchers is a long way from being exemplary for the language potential of LC children. Labov 35 gives a striking example of the possibility of the child "clamming up" when the investigator wants to elicit language. Labov also shows how such a boy, almost completely reticent at first, suddenly starts talking unrestrictedly once the investigator commits himself through words and gestures as being culturally one with that boy (cf. our way of collecting data, pages 35-7). According to Houston, 36 LC children have two registers at their disposal ("A register consists of a range of styles which have in common their appropriateness to a given situation or environment"): a SCHOOL REGISTER and a NONSCHOOL REGISTER. Of these, it is the school register which can be characterized as a "deficient language". The context expressed in the school register is limited and reveals little about the attitudes, opinions, or ideas of the LC children. As an authority, the investigator usually takes the place of the teacher, and will have access only to the school register; he will not analyse the far richer and more varied nonschool register that the child also has at his disposal. The investigator, therefore, incorrectly concludes that the child lags considerably behind the MC child as far as language capacity is concerned. Cazden 37 rejects both the difference theory and the deficiency theory in an article that carries the telling title "The Neglected Situation in Child Language Research and Education". Cazden follows Hymes 38 when she states that we should not evaluate language by itself, but according to the ways it is used in certain situations. A corollary to this is that the "language of the child 35

W. Labov, "The Logic of Nonstandard English", in F. Williams (ed.), Language and Poverty (Chicago, 1970). 36 S. Houston, "A Reexamination of Some Assumptions", p. 95?. 37 C. Cazden, "The Neglected Situation in Child Language Research and Education", in F. Williams (ed.), Language and Poverty (Chicago, 1970), pp. 81-101. 38 D. Hymes, "Competence and Performance in Linguistic Theory".

28

THE PROBLEM

is not a static phenomenon, but a flexible capacity either to react or not to react adequately upon a verbal context". Investigations not incorporating this context into its analysis will therefore draw incorrect conclusions. As has been remarked above, much research into language capacity has been directly inspired by the poor school achievement of children from lower social classes.39 When they enter school they already lag considerably behind pupils coming from higher social classes. This gap increases in the course of the following years, practically preventing their transition to higher education. These children are forced into what Williams 40 called the "poverty cycle" (see figure 1), which the child cannot break out of even as an adult, and which will trap his offspring as well. socioeconomic phase

sociocultural phase

Economic disadvantage/

v

Employment disadvantage

\

/

Developmental disadvantage

Educational disadvantage

Fig. 1. Poverty cycle (Williams, 1970).

To decrease the developmental disadvantage one must dissolve it as much as possible before the child goes to school. The solution which has been offered therefore is compensatory preschool education. The accent is being put more and more on the young

39

See, for example, D . Lawton, Social Class, Language and Education (London, 1968). 40 F. Williams, "Some Preliminaries and Prospects", in F. Williams (ed.), Language and Poverty (Chicago, 1970).

THE PROBLEM

29

child. 41 Some compensators give the impression that they would like to snatch the baby from his mother's arms right after birth because it is beyond her to give him a decent education. Of course, the call for compensatory education came especially from adherents of the deficiency theory: what the LC child misses must be provided; where he lags behind, compensation is in order. As a euphemism for compensation the term enrichment is sometimes also used. Enrichment does not suggest that LC children are missing something, so it is more acceptable to the LC parents. In fact, however, both flags cover the same cargo. Roughly speaking, three types of compensation programs can be distinguished. First, there are programs aiming at supplying the LC child in general with those MC experiences which he has not enjoyed - school trips, small excursions, nicely structured toys, and the like. Second, there are programs which try to teach all kinds of cognitive skills (thus, among others, language capacity). Third, there are the language compensation programs which start from the supposition that it is the deficient language capacity which causes the poor achievement at the cognitive level; a child lacking abstractions in his language use, according to these compensators, is also incapable of abstract thinking. The school is normative in all this: what the school does, is well done, and the LC child has to adapt himself at all times to the school which is the representative of the MC culture. Bereiter and Engelmann 42 put it this way: it is fully correct to term the LC culture impoverished because its children cannot comply with the requirements of the schools. Naturally, the adherents of the difference-theory offer strong resistance to these ideas. They do not believe in a monocultural society with a monocultural school in which the only culture has to be that of the higher social classes. Precisely because teachers too belong to a higher social class, real communication with LC children is missing, and these children are further frustrated 41

See, for example, Bereiter and Engelmann, Teaching Disadvantaged Children. 42 Bereiter and Engelmann, Teaching Disadvantaged Children.

30

THE PROBLEM

in their development. As soon as the norms and values of the LC culture are appreciated in the school, the educational achievements of the LC children will improve automatically. In this way of thinking compensation programs not only are to be rejected theoretically (see for example Brands, 1972), but they are also doomed to failure because they are based upon a culture - they even idealize and propagate a culture - that conflicts with the indigenous LC culture of the children. Bernstein 43 says that it is nonsense to talk about compensatory education as long as regular education is so poor - too many students in one classroom, poor teaching material, and the like. Nelissen 44 lists a number of critical points against compensation programs; he joins this idea with the following point: "From educational theory the criticism reads that with the introduction of c.p.'s (compensation programs) none of the essential educational problems have been solved. C.p.'s, especially language c.p.'s, are of little use, as long as so much has to be improved in education itself. A c.p. then is a mock solution, it can even block educational improvements" (page 303 - our translation). This argument, however, does not seem to us altogether relevant for rejecting compensation programs. We think they can be rejected on internal shortcomings only. We shall come back to this point in chapter 4 Discussion.

Practice seems to support the difference theorists. In 1967, after an investigation into the results of the work with different compensation programs, the United States Commission on Civil Rights came to the following conclusion: "The Commission's analysis does not suggest that compensatory education is incapable of remedying the effects of poverty on the academic achievement of individual children. There is little question that school programs involving expenditures for cultural enrichment, better teaching, and other educational services can be helpful to disadvantaged children. The fact remains, however, that none of the programs 43

B. Bernstein, "Education Cannot Compensate for Society". Nelissen, "Aantekeningen bij Kompensatieprogramma's", Pedagogische studien (1972), p. 303-18. 44

THE PROBLEM

31

appear to have raised significantly the achievement of participating pupils, as a group, within the period evaluated by the Commission." 45 The Head Start program is the best known of the various unsuccessful compensation programs. It should be noted, however, that according to other analyses some programs did have a positive influence upon the development of disadvantaged children.46 The fact that most programs failed to a considerable extent has led to the origin of theories, termed dangerous and racist, about the intelligence of different environmental and racial groups. According to Jensen 47 LC children and Negro children are hereditarily less intelligent than children from a high social (white) background. Jensen's conclusions are strongly assailed by Labov 48 and others: it is Labov's conviction that Jensen could only come to this conclusion because he adheres to the deficiency theory. Kagan 49 states that compensation programs have nowhere been completely developed or correctly evaluated as yet; thus conclusions from poor results of these programs are premature. We shall not dwell upon this matter any further because to a large extent it has to do with the status of the intelligence tests, a field in which we feel we are not enough at home to be able to judge Jensen's ideas. For a review of Jensen's ideas and a critical evaluation we refer to Nelissen 50 in addition to the authors mentioned. Undoubtedly both heredity and environment play an important role in intelligence. We maintain that one must search for explanations which can be controlled methodically. For our project this implies that we want to try to reduce differences between the environmental groups investigated to, among others, differential educational patterns, speech habits of the environment, and 45

Cited from A. Jensen, "How Much Can We Boost I. Q. and Scholastic Achievement?", Harv. Educ. Rev. (1969), p. 3. 46 See, for example, C. Bereiter, "An Academically Oriented Preschool", in Brison and Hill (eds.), Psychology and Early Education (1968). 47 A. Jensen, "How Much Can We Boost I. Q. and Scholastic Achievement ?". 48 W. Labov, "The Logic of Nonstandard English". 49 J. Kagan, "Inadequate Evidence and Illogical Conclusions", Harv. Educ. Rev. (1969). 50 J. Nelissen, "Aantekeningen bij Kompensatieprogramma's".

32

THE PROBLEM

communicative contacts. Only if these explanations turn out to be unsatisfactory would we feel that as a last resort one could turn to the Great Unknown - the hereditary factor. The insight in this whole problem can, moreover, be deepened by studying the development of language capacity and intelligence in the different environmental groups. We should, however, note here that our investigation has not been longitudinal. The results of such a cross-sectional project have to be interpreted carefully. In summary we can conclude that the field of sociolinguistics of child language as a whole is difficult to survey, but three points clearly present themselves: 1) The most important question is whether the language of LC children ought to be termed different or deficient. 2) This question can only be answered through experimental research in which the language of the child is analyzed in the whole of the communicative situation; and 3) Are there developmental differences to be identified between children from diverse environments, and can they be explained in terms of environmental variables ?

2 THE EXPERIMENT

2.1.

THE COMPOSITION OF THE GROUPS

The children from the three groups investigated came from different social environments; they all spent about four hours a day in a nursery school playroom. Group one, to be further identified as the ISOLATED LC group or I-LC group, consisted of fourteen boys, nine of whom came from unskilled working class families and five from skilled working class families. All fourteen lived in an isolated ghetto-like quarter of Haarlem, a town of about 200,000 inhabitants in North-West Holland, where the male population consists primarily of unskilled laborers. The playroom frequented by these children was situated in their neighborhood in a building that functioned as a discussion and action center open to all those of the area; selection factors, therefore, play hardly any role in the participation in the nursery school. During a period of half a year, this group had been exposed daily for about 10 to 15 minutes to a Dutch adaptation 51 of the Bereiter and Engelmann language compensation program. 52 The investigation has been limited to children who had participated in the program for at least three months. At the time of the project the playgroup consisted almost exclusively of boys; therefore, the few girls present were not tested. It appeared to be impossible to compose a group comparable in age and distribution of parental occupation from one other nursery school since participating in nursery school playroom activities still is more or less an affair of the privileged. Group two therefore 51 52

R. Gerstel, Een taalprogramma voor peuters en kleuters (Groningen, 1970). Bereiter and Engelmann, Teaching Disadvantaged Children.

34

THE EXPERIMENT

is composed of boys visiting different Haarlem nursery school groups. Of these, eight came from unskilled working class families and six from skilled. The difference between the children of groups one and two lies in the fact that the daily interaction for children of group one remains limited to their own social class, whereas the children of group two have possibilities of broadening their horizon through social interaction with children from different social environments, both on the street and in the nursery school playgroup. Moreover, this group possibly has profited from the fact that such playgroups attract children from certain working class families which enjoy privileged situations. Because of the composite character of the playgroups visited by our investigators, group two from now on is termed the MIXED LC group, or M-LC group. This group received no language compensation programs. The third group (MC group) is composed of children from middle class families. This group included only boys from families whose head had an academic degree or held an executive position in business, and who lived in the well to do Haarlem suburbs Heemstede and Aerdenhout. These boys visited different nursery school playrooms in that area. None of the children in the three groups suffered any hearing loss. Two boys living in Holland since their birth had Italian fathers. The nursery school's teacher's first observation - that their foreign extraction did not influence the boys' language capacity unfavorably - was found to be correct. After the analysis was completed, one of the boys even appeared to be among the best of his group - the isolated L C group. To exclude eventual contaminating effects from the most important covariate - age - we tested whether the composition of the groups was matched for age. Table 1 gives the average age expressed in months and its range within each group. An analysis of variance gave an F-ratio of .057 with 2.39 degrees of freedom. There is therefore no reason to suppose that the groups differ from each other as far as age is concerned. For financial reasons we could not measure the IQ of all children. The IQ scores of 11 boys of the M C group and of all boys of the isolated L C group were put at our disposal. The results were obtained with

35

THE EXPERIMENT TABLE 1 Average age expressed in months and its range within each social class group i-LC

m-LC

MC

Total

Average

41.71

41.78

42.14

41.88

Range

36-48

36-47

36-46

36-48

the Terman intelligence test. 53 The averaged Terman IQ of the MC group was 110.27, and of the LC group for the testing before the language program 94.58, and after the language program, 101.70. The increase in scores of the isolated LC group can perhaps be explained in terms of test experience. Apart from that, the question can be raised whether traditional intelligence tests such as the Terman do not derive their content too much from middle class culture, and thus favor children from middle and high social classes over children from lower social classes. Moreover, children of higher social classes have more toys at their disposal which in shape and in the potential for handling are similar to the objects used in intelligence tests for that age.

2.2.

THE COLLECTING OF THE LANGUAGE SAMPLES

In order to collect language samples, two students followed a child around in the playroom as much as possible and wrote down the utterances of the child to a total of 150 utterances per child. All the non-verbal aspects of the communication which were deemed necessary for the understanding of the utterances were also recorded. A verbal context is needed, for example, for the 53

Stanford-Binet, third revision, 1964.

36

THE EXPERIMENT

understanding of the utterance Home (in this case, as an answer to the question Where is Mama!); without a non-verbal context, the utterance all gone is utterly uninterpretable - one needs the information that the child while saying this looked into a drawer that appeared to be empty. From the 150 utterances collected, the pure repetitions and those that were going to prove uninterpretable were eliminated by the two students immediately after they collected the utterances, with the help of the protocols made by them both. More than 100 utterances were left per child; the first 100 were always used for the actual analysis. We have used 100 rather than the usual 50 utterances because we wanted to be maximally sure that the samples collected were representative for the normal language use of the child, and because our groups consisted of relatively few subjects. According to Darley and Moll, 54 a sample of 50 utterances suffices to make reliable statements on the mean length of sentence of the language of the child, but it is insufficient for statements on the syntactic complexity by means of a traditional complexity measure such as Templin's Syntactic Complexity Score.55 Darley and Moll suggest that the syntactic complexity score from utterance to utterance is perhaps a more variable phenomenon than sentence length. They also do not exclude the possibility that the arbitrary way of scoring within Templin's SCS (see also chapter 4: Discussion) could strongly determine the cause of the varying character of the complexity score. The latter seems highly probable to us: omitting the article, for example, can topple a sentence in Templin's score system from 4 to 0 points. For a number of reasons we have recorded the language used by the children in written form rather than in tape recordings. A microphone attached to the child can easily be a disturbing factor. Children who know something about the working of a tape recorder tend to produce different or affected speech. Precisely because of 54

F. Darley and K. Moll, "Reliability of Language Measures", JSHR (1960),

pp. 166-73. 55

See M. Templin, Certain Language Skills (Minneapolis, 1957).

37

THE EXPERIMENT

our basic tenets sketched in the introduction, it was, moreover, necessary that the child would be enabled to behave and move around normally in the playroom; running around or sliding on a slide, etc., is not very well possible with an attached microphone. Another possible solution would have been to work with hidden microphones; however, financial means for such a technique were lacking, and we also felt corroborated by the experiences reported by Labov 56 in our judgement that it is nearly impossible to isolate the voices of the subjects among those of all others, especially in the nearly continuous bustle and noise of a playroom. We therefore made a choice in favor of the normal play situation in order to be able to record the spontaneous and informal language of the child. According to the nursery school supervisors, the verbal and non-verbal behavior of the children in the presence of the two students did not deviate from their customary behavior. For most of the boys it was necessary to observe an initial short habituation period in which no utterances were noted and the child could become familiar with the students. The normal course of action in the playroom was changed in only one respect. Normally, the children were not supposed to talk while they were eating the food they had brought with them. For the benefit of the language recording this was permitted.

2.3. THE LANGUAGE MEASURES USED

Roughly speaking, language measures can be divided into two kinds: quantitative and qualitative ones. 2.3.1. Traditional quantitative

measures

The quantitative measures attempt to say something about the complexity of the language use of a speaker by giving average scores. Best known among the quantitative language measures is the mean length of sentence. Length never precisely measures 56

W. Labov, "The Logic of Nonstandard English", pp. 153-87.

38

THE EXPERIMENT

the syntactic complexity, of course, for it cannot give an analysis of what contributes intrinsically to such complexity but it is presumed to correlate highly with complexity.57 Next to the mean length of sentence there are a great number of so-called traditional quantitative measures, the best known of which are the number of different words (NDW), the type token ratio (TTR) which is computed by dividing the number of different words (types) by the total number of words (tokens), the number of one-word sentences (N1WS), the standard-deviation of the mean length of sentence (SD/MLS), and the mean length of the five longest sentences (ML5LS). Although all five correlate more or less with the real syntactic complexity score as does the mean length of sentence (MLS), it will be evident that the function of NDW and TTR in first instance is different from that of the other measures. The NDW is primarily a measure of one's vocabulary; as such, it can serve to indicate how many topics of conversation someone a child - has at his disposal. The TTR is a measure for the diversity of the language used. Carroll 58 had already noted that the TTR is not reliable for increased or differently sized samples, as is the case with our material. Carroll proposed an alternative TTR, to be termed TTR(C) here, which is computed by dividing the number of different words by the square root of twice the total number of words. Precisely because these measures have been applied so often in child language studies, we have applied them also to our project. The MLS can be considered the most important one, both because of its frequent use and because of its reliability computed by Minifie, et al., 59 A disadvantage of the MLS, according to Cowan, et al., 60 is that the scores on this measure depend on a number of situational variables in the investigation, mostly on the stimulus 57

See, for example, D. Sharf, "Some Relationships between Measures of Early Language Development", JSHD (1972), pp. 64-74. 58 J. Carroll, Language and Thought (Englewood Cliffs, 1964). 59 F. Minifie, et al., "Temporal Reliability of Seven Language Measures", JSHR (1963), pp. 139-48. 60 P. Cowan, et al., "Mean Length of Spoken Responses as a Function of Stimulus, Experimenter and Subject", Child Development (1967), pp. 191-203.

THE EXPERIMENT

39

intended to elicit the utterances to be collected and on the presence of the investigator. These intervening variables, however, do not occur in our project since we have used spontaneous, nonelicited utterances. As a matter of course, we did include answers to questions in our language samples because they occur normally in the daily communicative situation. 2.3.2. Traditional qualitative measures Two types of qualitative measures can be distinguished: quantifying qualitative measures and the descriptive measures. In the former method, sentences are credited with points on the basis of a superficial difference in degree of complexity - the more points, the higher the complexity of the language. The simplest example of such a qualitative measure is one in which a simple sentence would get one point, and a complex sentence two points. Most qualitative measures have a more involved scoring system, but according to Sharf 6 1 none measures better than the MLS. If this is true, we feel it is useless to apply them because the time-consuming work of computing the complexity score results in no information at all about what causes or prevents the attainment of a high complexity score. Just as with quantitative measures, the final result is a number which says nothing about the syntactic quality of the language use. An assessment of syntactic quality of language use can be provided only by a descriptive measure. A descriptive complexity scale is based on an inventory of all factors which might contribute to the complexity of the sentences produced. Structures, elements, and constructions found in the child's language can be scored on such an inventory. Children can then be compared to see which items each of them produces and how often. Thus it is possible not only to establish that the language of the one child is more complex than that of the other, but also what exactly contributes to that greater complexity. 61

D. Sharf, "Some Relationships between Measures of Early Language Development".

40

THE EXPERIMENT

Because the existing descriptive measures, for example Muma, 62 looked incomplete to us, we have constructed a new Syntactic Complexity Score which meets our conditions, specifically as far as the marking of deviant language use is concerned. The application of this descriptive measure to the scoring of 4200 utterances on syntactic complexity will be explained in the next section. 2.3.3. The new Syntactic Complexity Score In order to measure effectively someone's language capacity, a number of difficulties in the analysis of spontaneous language production have to be overcome. First, it is is necessary to handle all kinds of performance disorders such as false starts, resumptions, hesitations, and the like. Second, language use is influenced by various factors with which a linguist merely describing the knowledge of his language intuitively is not confronted, i.e. the factors of the speech event. 63 Two factors of the speech event are of prime importance: i. Every sentence is embedded in a piece of information already known - the context. This information is relevant for the semantic intent described on the level of deep structure. Moreover, the context results in certain peculiarities appearing in the surface structure. ii. Because both speaker and hearer know the situation, much information is transferred covertly, referringly, nonverbally (for example pointing behavior), or even not transferred at all. This information is presupposed to be common knowledge in the conversation. Because of these two factors, deletions and substitutions (for example pronouns and the verb phrase do so) occur in the actual use of the language. For example, the information John hit Pete 68 See J. Muma, "Syntax of Preschool Fluent and Disfluent Speech: A Transformational Analysis", JSHR (1971), pp. 428-41. 63 See D. Hymes, Language in Culture and Society (New York, 1964).

THE EXPERIMENT

41

with a ruler can be rendered in dependence on context as John!, Pete!, With a ruler!, or He did it with a ruler, or even with just He! or Him! These factors are even stronger in the analysis of child language because a child's speech is more of a reaction to the immediate environment than is an adult's. For these reasons, Bloom 64 came to the conclusion that we must not exclude the analysis of context, situation, and nonverbal behavior if we want to penetrate into the knowledge the child has of his language. The best way to do this is to distinguish systematically between the semantic intent or message - the information the child intends to give, as determined from context, situation, and nonverbal behavior - and the realization or code which is realized on the verbal level. The scoring form of the syntactic complexity had two separate parts: SEMANTIC ASSESSMENT and REALIZATION ASSESSMENT. In the Semantic Assessment the paraphrase is scored and in the Realization Assessment the actual utterance with all its surface peculiarities. The paraphrases were made by two persons according to the following guidelines: (1) paraphrases ought to be as simple as possible; (2) paraphrases ought to be well-formed adult sentences; (3) paraphrases ought to be contextually independent. In case of uncertainty the decision lay with the total group of experimenters (about ten people). The following example from Bloom can serve as an illustration of this procedure: (1) Looking at a picture of a baby eating cereal Mother: What is the baby doing? Child: Eating cereal. Clearly, the paraphrase has to be: (2) 64

The baby/He is eating cereal.

L. Bloom, Language Development : Form and Function in Emerging Grammars, Chapters 1 and 2.

42

THE EXPERIMENT

In the Semantic Assessment both the subject and the time auxiliary have to be scored. In the Realization Assessment, on the other hand, both deletions have to be accounted for. Van der Geest 6 5 argues more extensively that present-day linguistics has no focal point as it did a number of years ago with Chomsky's Aspects.66 The most complete theoretical linguistic elaboration is undoubtedly Seuren's.67 His theory is the more interesting for our purposes because it is based on a semantic standpoint that does not deviate fundamentally from that of generative semantics. His study has therefore been taken as the theoretical point of departure for the scoring form. Some expansions and modifications, accounted for by Van der Geest, appeared to be necessary.68 The basic terms and distinctions used in Seuren's theory are presented in the tree diagram in figure 2:

NUCLEUS

T

(Modal)

[Pres 1 "— Past J (NEG)

subject, object, verb, etc.

\ tense

Noun

i

M SQL QL QN ASS IMP NEG QU

= = = = = = =

sentence qualifier qualifier quantifier assertion imperative negation question

n; QN + n

I

["possibility permission Lnecessity J

Fig. 2. Seuren's (1969) theory in a simplified tree diagram. 65

A. van der Geest, "Naar een syntaktische kompleksiteitsskore", wetenschap in Nederland (1972, in press). 66 N. Chomsky, Aspects of the Theory of Syntax. 67 P. Seuren, Operators and Nucleus (Cambridge, 1969). 68 A. van der Geest, "Naar een syntaktische kompleksiteitsskore".

Taal-

THE EXPERIMENT

43

A short description of this tree-diagram, specifically in its adaptability to the communicative competence, will be given in the next section. A. Nucleus-Operators Seuren argues that a distinction has to be made between the nucleus and the operators of a sentence. Under the nucleus he includes the verb and those constituents that are internally connected with the verb. The total notion of, for example, to convince implies that there is someone who convinces (actor or subject), someone who is being convinced (object), and something of which he is being convinced (prepositional object). One could say - though Seuren does not - that the nucleus gives information on entities in reality (persons, actions, objects, and the like) and their relation to each other. Seuren does not define the operators. Each of them is empirically motivated. The operators can be taken as performed by the speaker on the reality expressed in the nucleus. Essentially, the operators give information about the attitude of the speaker towards that reality, so the sentence qualifier expresses the intention of the speaker. The need for information, for example, is expressed by the question (QU). The qualifier usually expresses the standpoint taken: the information in the nucleus is true or not true, possible or not possible, necessary or not, and (or) its setting is in the present or the past. In fact, in the operators the speaker has a number of sets of options at his disposal which enable him to express nuances about the information in the nucleus. The nucleus-operator distinction was maintained systematically in the scoring form. The nucleus has been subdivided into nominal and verbal predicates; furthermore, all possible relations to the verb have been accounted for. As a background for this inventory we have taken the study on Dutch syntax by Klooster, Verkuyl, and Luif.69 In the operator part, a systematic distinction has been 69

W. Klooster, et al., Inleiding tot de syntaxis (Culemborg, 1969).

44

THE EXPERIMENT

realized between the sentence qualifier and the qualifier. Within these subdivisions the set of options that has been incorporated has been taken as completely as possible. The distinction between the nucleus and the operators as well as that between SQL and QL had to be assumed for both the Semantic Assessment and the Realization Assessment. B. Expansion It is an important developmental feature of language acquisition that the child learns how to include more and more information within a sentence. All traditional syntactic complexity scores are based upon this developmental trend. This applies strictly speaking also to the T-unit as developed by Hunt. It is important, therefore, to keep the basic information of the sentence strictly separate from those constituents that can be added to a sentence ad libitum. From a linguistic standpoint, all parts added to a sentence are based upon a sentence in its own right, for example: (3)

Sentence 1. The dog is my aunt's. Sentence 2. The dog is named Ajax.

become, depending on what the speaker is attending to: (4) My aunt's dog is named Ajax. or: (5)

The dog named Ajax is my aunt's.

The same reasoning holds true for adverbials, modals, auxiliaries, and all sorts of adverbial clauses. The total inventory of expansions, as incorporated in the Semantic Assessment and the Realization Assessment, consists of: adjective, relative, coordination, topicalization, adverbs and adverbial clauses, and modals and auxiliaries. Those clauses that have a function within the nucleus of the main sentence, such as subject-clause, object-clause, and the like, have been treated as a separate group. According to recent linguistic research, such sentences do not have an underlying conjunction.

THE EXPERIMENT

45

Several of the expansions just enumerated have been described as operators by Seuren, for example, the modals and adverbials, and the modal clauses; thus we have treated all expansions as operations. Seuren does not deal with complex sentences, but the conjunction of two or more sentences also comes under the characterization that we have given of operators: (6) It is raining. We stay home. (7) We stay home because it is raining. (8) It is raining so we stay home. (9) We stay home because of the rain. The complex sentence makes a statement about realities with more subtle shades of expression than can be given in successive simple sentences. C. Realization In Chomsky's theory of 1965, as well as in generative semantics and in Seuren's model, the differences between the deep structure of a sentence - doing as much justice to its meaning as possible and its surface structure are fully accounted for by means of transformations. Of these, four types can be distinguished - deletion, permutation, addition, and substitution. If we may consider these to be universal, the conclusion is warranted that the difference between semantic intent and realization in child language can be described in terms of these four transformations. In practice this has been done already by Tervoort 70 and Menyuk. 71 Both describe deviances of their performance data from the adult model language in terms of these four types of transformations, although using a different terminology to indicate the deviance from normal transformations. Linguistic theorists, furthermore, have made complete lists of transformations and their order of application in adult language; 70

B. Tervoort, Final Report (...) Analysis of Communicative Structure Patterns in Deaf Children (Groningen, 1967). 71 P. Menyuk, Sentences Children Use, Chapter 2.2.

46

THE EXPERIMENT

eventually, they could be applied to the scoring form under discussion. It would be in order to actually do so if the order of acquiring these transformations is identical with the order of the transformations as listed in linguistic theory. Two hypotheses concerning the relationship between the order of transformations and the order of acquisition have been formulated: (i) The correspondence hypothesis (discussed in Section 1.1.). The correspondence hypothesis must be rejected on the basis of data showing that V.O. constructions initially occur much more frequently than S-V-(O) constructions. Also modals, adverbs and adjective constructions, such as also chair, Mommy Sandra's doll, nice doll, occur quite early.

kitchen,

Evidence that the transformational development of a child's language can take a different course from the order of transformations in linguistic theory appears in this example: (10)

Hester goes => I goes => I go

The intrinsic development of the child's language can take its own course. (ii) The other hypothesis states that the child learns how to break down surface structures into deep structures.72 These two opposing hypotheses make it clear that caution is necessary when linguistic theory is applied to the language development of a child. This is why we have given the Realization Assessment the character of a mere inventory. The system followed in this matter has been that we have kept correct realizations strictly separated from incorrect ones. The following points still have to be dealt with: 72

See J. Fodor, "How to Learn to Talk", The Genesis of Language (Cambridge, Mass.); I. Schlesinger, "A Note on the Relationship between Psychological and Linguistic Theories", FoL (1967), pp. 397-402; and E. Ingram, "The Requirements of Model Users", Language Acquisition: Models and Methods (London, 1971).

THE EXPERIMENT

47

A. Deletion In language usage, a number of deletions can be pointed out which linguistics does not take into account because they are structural independent, i.e. they are realized on the basis of context and situation. Added to these in child language are the deletions the child cannot avoid because some rules that are needed to produce grammatical sentences are still lacking in his competence. Only this latter type of deletion has been scored in the section approximation. As has been stated before, linguistics does not supply us with indications of how to systematize approximation deletions; an error analysis afterwards is the only way to reveal developmental tendencies. B. Order, redundancy, and substitution For order, redundancy, and substitution, the same holds true. Samples of language use which cannot be incorporated into a linguistic analysis are common, such as : (11) Said Kissinger: We had a plain-spoken conversation without results. (12) Nemen we toch lekker de trein. 'Take we the train anyhow' (lit.) In child language the following types are to be expected: (13) I goes. (14) I does hammer. (15) Goes train choo choo. To merely describe this sort of constructions as deviant, as Menyuk does, is not sufficient. It is more reasonable to consider and evaluate these phenomena as milestones in the child's development in so far as is possible. The peculiarities of such sentences are scored in the approximation section; this does not imply so much that they are incorrect as that they are typically childish idiosyncrasies in the use of language.

48

THE EXPERIMENT

C. Realization as worked out in the SCS As has been demonstrated in the foregoing, the order of acquisition of the transformations does not necessarily run parallel to the transformational order in the linguistic derivation. No one has studied the order of acquisition of transformations in Dutch. For English the studies are limited to the order of acquisition of linguistic adult transformations. Only Bellugi 73 is an exception to this. She goes more extensively into the matter of the linguistic development of specific transformations by the child, namely negation, auxiliary-inversion in questions, negation -f- indefinites, and case marking in personal pronouns (I versus me). Altogether this gives scant information, but we have incorporated it into the scoring form in as far as it can be applied to Dutch. Instead of elaborating upon all kinds of possibly hypothetical transformations, we have incorporated into the scoring form the end point and the different intermediate stages of a number of transformations. Illustrative examples of these are: (16) Hester gaat => Ik gaat => Ik ga 'Hester goes' => 'I goes' => 'I go' (17) Ik => mij •I' => 'me' (18)

Omdat ik kwam gisteren => Omdat ik gisteren kwam 'Because I yesterday came' => 'Because I came yesterday'

As far as deletions are concerned, it looks as if developmental transformations are hard to describe systematically. A number of different statements seem possible, though none is precise: 1. words carrying main accent in the sentence cannot be deleted. 2. content words are deleted less easily than function words. It is quite difficult to translate this kind of remark into an explicit and formalized rule. Even if statements on deletions in terms of 73

See, for example, U. Bellugi, "Simplification in Children's Language", Language Acquisition: Models and Methods (London, 1971).

THE EXPERIMENT

49

constituents or their functions would be possible in general, it is difficult to trace them back from the heading approximation deletion in the score list. It should be stressed that deletions cannot be described in terms of what is higher or lower in the tree, as both higher and lower constituents have to be considered as given in context and situation which means that both can drop out in normal usage. In the scoring form one can trace how the different items of the Semantic Assessment are realized in actual language use. Next, a column has been put into the Realization Assessment containing contextual (correct) deletions, and a column of developmental (wrong) deletions. This latter column gives information on formal regularities in the development of the deletion transformation. For a good survey of the Syntactic Complexity Score as it has been worked out, table 2 presents a short scheme. For the worked out SCS, one is referred to Appendix I, where the total scores per group are also given. 2.3.4. Some other measures

Apart from this new SCS and the six traditional measures, the three groups have also been compared on the following measures: - The relation function words/total number of words. It is known 74 that function words such as prepositions, articles, and auxiliaries are acquired later than content words such as nouns and adjectives. It seems worthwhile to check whether there is a difference between the three groups on this point. - The number of sentences not dealing with the here and now. This measure could be an indication for the number of possible topics the child can talk about. If one is not bound to one place (here) and one time (now), the amount of communicative possibilities is greatly enlarged. According to sociolinguistic literature summarized above, we can expect that children from the middle 74 See, for example, R. Brown and U. Bellugi, "Three Processes in the Child's Acquisition of Syntax", Harv. Ed. Rev. (1964), pp. 133-51.

50

THE EXPERIMENT

TABLE 2 Framework of the Syntactic Complexity Score

Semantic Assessment Nucleus

Operator

subject verb object etc. Expansion: subordination coordination Adverbials etc. Qualifiers tense modality passive etc. Sentence qualifiers declarative question etc.

Realization Assessment Adult norm

Approximation

Nucleus

(deletion)

deletion

Operator: Expansion

(deletion)

deletion/ substitution

Qualif.

(deletion)

deletion/substitution redundancy

SQL

(deletion)

deletion

morphol. word order

deletion/substitution redundancy inversion

THE EXPERIMENT

51

class will more often use sentences whose meaning is not related to the here and now, specifically because they have learned to abstract. This type of reasoning is also valid for the next measure. - The number of abstract nouns. If we may believe Bernstein 75 as to the concrete character of the restricted code, we can expect few abstract nouns in the language of LC children. 2.3.5. The sentence as basis of the analysis Presently it is customary not to analyze child language on the basis of sentences, but to divide the sentences into T-units and to use the latter for the analysis. A T-unit is made up of one main clause with its subordinate constructions, so that for example a coordinate sentence consisting of two main clauses is divided into two T-units. This method of analysis has been developed by Hunt 7 6 because children have a strong tendency, both in oral and written language use, to string together unrelated sentences use run-on sentences - or to coordinate them continuously, as in the well-known "and then ... and then" constructions. Hunt thinks that we get a wrong picture of children's language capacity if we apply all kinds of qualitative measures such as the MLS to language with those types of constructions. Through his investigations, Hunt, who investigated written language of children of six years and older only, came to the conclusion that the quantitative measure number of words per T-unit ought to be preferred over the MLS. We are of the opinion that, first, it is incorrect to split up all coordinated sentences, especially where the functional coordinations are concerned. Wijnstra 77 has already pointed out that working with a T-unit for Dutch data gives the following rather strange result. Comparing sentences coordinated by the causal 75

B. Bernstein, "A Sociological Approach to Socialization", pp. 25-61. K. Hunt, Grammatical Structures Written at Three Grade Levels (= NCTE 3) (Champaign, 111., 1965). 77 J. Wijnstra, "Syntaktische Kompleksiteit in schriftelijk taalgebruik", Ned. ts. v. d. Psych. (1972), pp. 29-55. 76

52

THE EXPERIMENT

coordinating conjunction want 'for', which does not cause inversion of words - the only indication in Dutch for the distinction between coordination and subordination - with sentences where a subordinate clause beginning with the causal conjunction omdat 'because', which causes inversion and is connected with the main clause, leads to the noting of two T-units in the former and one T-unit in the latter case. This is obviously an absurd distinction. Second, a T-unit seems more suited for written than for oral language because in the latter case one has phonological information which makes clear what should be considered a sentence and a communicative unit. Third, according to Hunt a T-unit should contain at least one subject and one predicate. Obviously, this could not be a starting point for our project, as our samples included many sentences with correct or incorrect deletions of the subject; a sentence like playing with the toy cars as an answer to the question What are you doing ? with correct deletion of the subject and the auxiliary is functional within the communicative situation but does not qualify under the T-unit definition. Finally, in the stage of language acquisition investigated (i.e. of three- to four-year-olds) the capacity to coordinate sentences has to be evaluated positively, not negatively as in the T-unit analysis where the average number of T-units per sentence is considered to be a negative measure for language capacity. Negative means here that the higher the score is the lower the language capacity is reckoned to be. A higher score is reached when the child coordinates many sentences, and therefore has many T-units per sentence.

3 RESULTS

3.1.

FACTOR ANALYSIS OF THE MEASURES OF SYNTACTIC COMPLEXITY

In order to reduce the total of 109 measures 78 of structural complexity to a limited number of factors that could be interpreted, a factor analysis was carried out. This was done in two phases. Preliminary factor analysis A first preliminary factor analysis (see Appendix II) resulted in three factors that could be interpreted reasonably well.79 The factorial variances were respectively 19.87, 13.16, and 8.02 percent. Factor 1 The following linguistic features load highly on factor 1: a) the composition of the nucleus with regard to verbs and their NP's; b) the adverbial clauses of time and degree; c) the adjective+noun construction; d) the complex sentence; e) tense phenomena; f ) the declarative and the comparative; 78

See Appendix II; the total of 109 variables resulted from a reduction of the variables mentioned in Appendix I. The reduction was based on a minimal cell frequency (25 per cell). 79 This F. A. should be considered tentative because of the large number of variables in relation to the limited number of subjects. We will discuss this analysis only because of its linguistically attractive results.

54

RESULTS

g) irregularities of the verb forms of present tense, past tense, and of the participle; h) the congruence between subject and verb, between adjective and noun, and between article and noun; i) prepositions with a function within the nucleus; j) morphological phenomena within the NP - singular and plural, pronominalization, and the like; k) the conjunctions; and 1) the word order in the sentences containing expansions. This whole series of phenomena can be reduced to the following two general points: 1. Phenomena that have to do with the development of the nucleus, both with regard to the semantic structure and to the explication thereof in the surface structure. See for this aspect the listings a), e), f), g), h), i), and j). High scores on these indicate use of the capacity to organize the nucleus; apparently the latter is in full development. 2. Phenomena that have to do with the organization under one sentence heading of two nuclei that are independent of each other in principle by means of linguistic composition. These phenomena, too, relate both to the Semantic and the Realization Assessments. The following listings are important for this expansion of the sentence: b), c), d), h), k), and 1). High scores on these characteristics imply that the child is already capable of combining two independent realities on the interpretative level and of realizing them on the level of morphology and syntax. As far as the negative scores are concerned, we limit ourselves to those below -.20. On factor 1 the following points load negatively to that extent: 1) The frequent use of deixis (for example, that, there, etc.), must, will, and may, the different types of question, and the imperative. In various publications such as those by Bernstein these linguistic phenomena have been shown to have sociological

RESULTS

55

implications, specifically in the sense that they occur frequently within the class of unskilled laborers, i.e. in the so-called restricted code. That the modal verbs must, will, and may and the imperative load negatively on this factor, in opposition to can and the declarative sentence, points to the fact that language is employed for hierarchical purposes more often by the lower class than by the higher class: the imperative from higher placed persons to lower placed persons and the so-called desiderative question from lower to higher placed persons. The children of the higher class, on the other hand, employ more frequently the declarative sentence and the modal auxiliary can. Both groups know and use modal operations. 2) Another negative phenomenon is the lack of, or at least the less than optimal use of, syntactic means to realize on surface structure level those structures for which the organizing competence seems to be already present, specifically with regard to nucleus and modal operator. In other words, although the nucleus has come to full development already in semantic sense, it is not fully correctly organized in using them, pronominalization rules are applied the wrong way, and all kinds of constituents are deleted that should not be. As a whole, factor 1 can be seen as indicating the typical developmental language problems for the three- to four-year-old child both in Semantics and in Realization. We will, therefore, refer to it with the generic term the DEVELOPMENTAL LANGUAGE FACTOR. Factor 2

Factor 2 is to be considered as a supplement to factor 1 with somewhat different emphasis. Generally speaking, the relevant positive loadings of factor 1 are considerably lower and can even become negative in factor 2. High loadings were found in factor 1 for the declarative sentence and in factor 2 for the question (in factor 3 the imperative takes a central position). In factor 2 the highest positive scores and lowest negative loadings are related to the realization of the sentence. This is why it seems best to speak

56

RESULTS

of it as the REALIZATION FACTOR. The high loadings concentrate around the verb form, the pronominalization, and the correct word order in the sentence. The negative loadings are to be found especially in the approximation section. Because approximation deletions are important in this factor - especially those of the tense aspect, the copula, the subject first and third person, the deixis, the article, and the modal auxiliary - it is also in order to speak of this as a DELETION FACTOR. Factor 3

Factor 3 measures differences in the instrumentality of the language use in the different subjects. Under instrumentality of language and language use is understood here that language is employed to cause a certain effect on the hearer: one may want to inform the hearer (the declarative sentence), or receive some information (the question), or desire that the hearer performs a certain action (the imperative sentence). More obviously than in factors 1 and 2, the focal point here is the instrumentality of the language act. Of these three instrumental aspects, only the loading on the imperative sentence is very low (-.76), which indicates that this factor concentrates on the transfer of information. This is again strongly emphasized by the fact that can loads positively, and the desiderative aspect, as expressed by may, very negatively. In the transfer of information it is efficiency that counts; therefore, in this factor the correct deletions, i.e. those deletions that are effectuated within context and situation, have a high load. The same reasoning is valid for the occurrence of the deixis instead of constituents that are fully explicitated: there and this (one) are permissable simplifications

of, for example, on the green chair and the big

truck.

Connected with this is the preference for the definite article. The use of the latter is limited to those cases in which the person or object named by the constituent has been introduced already in the foregoing part of the conversation, or can be supposed to be known to the hearer by the speaker on other grounds. This factor is especially concerned with the here and now; therefore, the

57

RESULTS

imperfect present tense loads highly and the loads on the imperfect past tense and the perfect tense, present and past, are negative to very negative, respectively. In summary, for this factor the efficiency in the concrete transferring of the information takes a central position. The second and definite factor

analysis

Considering the large number of linguistic variables in relation to the limited number of subjects, it was decided to keep for the analysis only those linguistic variables whose communality - the percentage variance explained by the factors - came to 50 % or higher. After having thus eliminated the noise, a new factor analysis was carried out on the remaining 36 variables. The three factors that resulted could be interpreted easily; they explain, respectively, 47.7%, 15.5%, and 8.8% of the total variance. The meaning of these factors - which, by the way, were non-rotated and orthogonal - had to be established again on the basis of an analysis of items loading high and low respectively on the pertinent factor. Factor 1

Factor 1 can best be considered as a general planning factor concerning that mental activity which is needed to bring into agreement two or more parts of the sentence as somehow related to each other. An illustrative example is the congruence between subject and predicate. After the decision as to person and number of the subject, the form of the verb is fixed in principle. The information on person and number therefore has to be kept in mind until the verb is realized. Concerning factor 1, of the twenty-four items with a load over +.70 or below —.47, twenty-three could be described as planning items (see table 3).80 On the level of linguistic interpretation of the planning factor we propose three types of planning (see table 4). 80

See also Appendix III.

58

RESULTS TABLE 3 Items with the highest loadings on the factors 1, 2 and 3 item

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 13. 14. 15. 16. 17. 18. 19. 21. 22. 25. 26. 28. 29. 30. 31. 32. 33. 34. 35. 36.

Prepositional phr. Determiner, sing. Determiner, indef. Adverb, tense Adverb, degree Subordination Complements Coordination present tense past tense perfect tense imperative morpheme tense auxiliaries participle morpheme agreement subj.-verb agreem. det. - noun agreem. adj. - noun prep, of the verb noun sing. 1st pers. pronoun subject pronoun conjunctions order: S.V. Adv. O order of sentences tense approximation copula deletion (appr.) subject del. appr. 1 subject del. appr. 2 deixis del. appr. article del. appr. Modal del. appr.

FI

83 76 71 74 73 88 88 88 61 83 61 -15 67 55 92 84 77 84 75 81 85 88 73 91 -42 -47 -53 -50 -45 -47 -50

F2

F3

-61 20 62 58 62 67

60 62 61 80 69 67 67

Comm. V /o 81 69 75 55 54 82 83 87 84 76 81 36 94 80 93 76 61 81 65 68 84 81 58 87 65 63 78 90 76 83 82

1. The planning of the constituents which are inherently connected with the verb (the constituents in the nucleus) after this verb has been chosen. The planning concentrates here around the

RESULTS

59

TABLE 4 Interpretation of factor 1 : the planning factor Type of planning

Numbers R/S

Characterization

planning 1

Is, 19»

prepositional relation between verb and related constituents

planning 2

2s, 3s, 16B, 17B, 18B, 21B, 22R, 25K, 31R-, 32R-, 33B-, 35R-

agreement/congruency a) article + noun b) subject + verb c) adjective + noun

planning 3

4s, 5s, 6s, 7s, 8s, 26R, 28B, 29B, 36B-

Interpretational relations between independent kernels: adverbials coordination, subordination, topicalization, complement sentences, and their realization in the surface structure

R and S mean that the items belong to the realization or semantic assessment of the score list resp. means that the item has a negative load.

optional choice of the preposition as the following synonymous sentences illustrate: (1) (2) (3) (4)

Mary gives Johnny a kiss. Mary gives a kiss to Johnny. Mary kisses Johnny. Johnny gets a kiss from Mary.

2. The planning that is necessary for the application of agreement rules, i.e. the choice of the article dependent on the noun (nouns are marked for gender in Dutch), the morphological marking of the adjective dependent on both the article and the noun (adjectives are inflected in Dutch), and congruence. Examples: (5)

Een mooi schip loopt van stapel. 'A beautiful ship is launched.'

60 (6)

RESULTS

De mooie schepen lopen van stapel. 'The beautiful ships are launched.'

3. The planning that is necessary to bring possibly independent kernel sentences into relation to each other, i.e. sentence combining transformations. Only this third type of planning is met with in literature as an important variable in the development of language capacity. 81 It has consequences for both the Semantic and the Realization Assessments. In the Semantic Assessment the interpretation of the relation between the two sentences plays an important role; in other words, the complex sentences that eventually result can differ on the semantic level, for example: (7a) The man became seasick. (7b) The man was the steersman. (8) The man who was the steersman became seasick. (9) The man who became seasick was the steersman. In (8) it is communicated that somebody became seasick, in (9) that someone was the steersman. On the realization level, all kinds of phenomena occur that have to be planned beforehand, for example, the relative, the conjunction, and (in Dutch!) the word order of the subordinate clause. In the following examples: (10) Het regent te hard. We blijven thuis. 'It is raining too hard. We stay home.' the causal relation is not explicated, as it is in the following sentences: (11) Het regent te hard; daarom blijven we thuis. 'It is raining too hard, therefore we stay home.' 81

See, for example, K. Hunt, "Syntactic Maturity in School Children and Adults" ( = Monogr. Soc. Res. Child Dev. 134) Vol. 35 (1970); R. O'Donnell et al., Syntax of Kindergarten and Elementary School Children (Champaign, 111., 1967).

RESULTS

61

(12)

We blijven thuis want het regent te hard. 'We stay home because (Dutch want) it is raining too hard.'

(13)

We blijven thuis omdat het te hard regent. 'We stay home because (Dutch omdat, causing inversion) it is raining too hard.'

(14) Het regent te hard zodat we thuis blijven. 'It is raining too hard so that we stay home.' (lit.) (15)

We blijven thuis vanwege de regen. 'We stay home because of the rain.'

In the semantic and syntactic descriptions we can account for these linguistic differences between sentences; however, it remains unexplained why a given speaker prefers one of these over the others. Therefore an analysis and description is needed in terms of communicative competence. Such an analysis ought to have both socio- and psycholinguistic validity. Arguments for the planning interpretation can also be derived from the negative loads. It appears then that approximation deletions are specifically bound to get negative loadings. As has been exposed in the positing of the problem, deletion is caused by shortcomings of a neuropsychological type such as a language capacity that has not been organized completely; in other words, it is caused by shortcomings in planning capacity. Finally, it should be noted that factor 1 is applicable to both the semantic intent and to the realization. Factors 2 and 3 The two other factors can be considered as complementary to factor 1 with regard to their interpretation. In factor 2 only those items which are explained negatively by factor 1 load higher than +.60. The items that load highly on this factor all deal with incorrect deletions - a high score on factor 2 is a negative indication for the language capacity. This factor is relevant to the realization of the utterance. Factor 3 loads very highly on grammatical tense phenomena,

62

RESULTS

both regarding semantics and realization. Comparing factor 3 with factor 1, we see that, except for the universal or present tense, both the first and the third factor have high loadings on the tense aspects and the other morphological aspects of the verb. Those items which score high on factor 1 score low on factor 3 and vice versa. That the universal or present tense loads very negatively (—.61) on this tense factor cannot be understood from a linguistic point of view but must be explained from the standpoint of developmental psychology. Children almost always begin their linguistic career with sentences that are all related to the here and now (see for example Brown & Bellugi).82 In this phase, hardly any tense aspect is realized; the verb forms used are almost exclusively infinitives. Even when the tense aspect is discovered in a following phase, it is still limited to the semantic intent. The realization of the grammatical tense then is generally indicated by tense adverbs; for example, yesterday can indicate in the past, whether this is in fact either two hours ago or a whole week ago. The verb form is often still limited to the infinitive. The negative load on the present or universal tense therefore indicates that use of this tense is a negative developmental feature.

3.2.

DIFFERENCES AMONG THE GROUPS

Social class differences in the three factors found within the measures of the structural complexity and on the traditional language measures have been evaluated by means of an analysis of variance. 3.2.1. Social class differences on the factors found in the investigation of the syntactic complexity In order to compare the groups on the factors found from the syntactic complexity scores, it was necessary to compute factor scores per child. At the moment of processing no program for 82

R. Brown and U. Bellugi, "Three Processes in the Child's Acquisition of Syntax", pp. 133-51.

63

RESULTS

this was available. We have therefore attempted to confirm the results of the factor analysis by the application of a principal component analysis. The principal component analysis on the 36 language phenomena of the definite factor analysis produced components which were almost exactly equal to the factors found earlier. The components found explain 44.7 %, 14.6 %, and 7.6 % of the total variance. For their meaning, reference can be made to the interpretation of the second factor analysis. (For the component scores see Appendix IV.) Since the difference from this second factor analysis is so small, we will not discuss the result of the principal component analysis separately. We were able to compute the three components per child and per social class. We shall now present first the means of the components per environmental group in a table (see table 5). The scores are standard scores with a mean of 0 and a standard deviation of 1 (z-scores). In the same table the F-ratio (a measure, computed in the analysis of variance, indicating the size of the difference between the groups) is given, followed by the level of significance. TABLE 5 Social class differences on the linguistic components in z-scores

Component I Component II Component III

i-LC

m-LC

MC

—.54

—.20

.74

8.05

p < .01

.20

—.29

.09

.93

non-sign.

—.52

.10

.42

3.15

p < .05

F-ratio (2,39)

Level of sign.

The non-significant difference on component II is related to the interaction between age and social class which wipes out the environmental effect (see 3.2.4.: Developmental differences between the social classes). The table demonstrates that on both components I and III there is an increase in the scores in the order i-LC, m-LC,

64

RESULTS

MC. In the case of existing significant environmental differences (p < .05), it seemed to make sense to check which intergroup difference contributed especially to this (i-LC vs. m-LC, i-LC vs. MC, or m-LC vs. MC). The intergroup distance could possibly have to be ascribed to the position of one of the environmental groups. A priori it seems likely that the MC group will differ most from both the i-LC and the m-LC groups, for the lack of any group from the transitory class. For the analysis, a procedure indicated by Tukey was used. 83 For component I, the environmental difference is mainly due to the MC group showing a significant difference (p < .01) from i-LC and m-LC. The difference between the i-LC and the m-LC groups is due to chance. On component III the MC group is significantly different from the i-LC only (p < .01), whereas the difference between the i-LC and the m-LC groups is barely significant at the 5 % level. We note hereby that Tukey's test is rather conservative. 3.2.2. Environmental

differences on the traditional measures

language

The environments are further compared on six traditional language measures, i.e. the number of different words, the mean length of sentence, the type token ratio calculated using Carroll's technique, the number of one word sentences, the mean length of the five longest sentences, and the ratio of the number of function words to the total number of words. Table 6 indicates the means and the significant differences found. With the exception of the ratio function words/total number of words, where the F-ratio does not exceed the critical value, significant social class differences are found on all measures. Further individual group comparisons using Tukey's method indicate that the MC group is different on all five measures at the 1 % level from the i-LC group, and from the m-LC group on the measures NDW, MLS, TTR(C) and ML5LS. The difference between the 83

See B. Winer, Statistical Principles in Experimental Design (New York, 1962).

65

RESULTS TABLE 6 Means for different traditional language measures per social class i-LC

m-LC

MC

F-ratio

120.79

127.71

163.50

11.14

p < .01

MLS

3.32

3.90

5.07

12.44

p < .01

TTR(c)

4.71

4.61

5.13

4.75

p < .05

N1WS

12.30

4.93

.86

9.76

p < .01

ML5LS

6.20

7.06

10.14

12.19

p < .01

Ratio FW/TNW

.318

.348

.361

2.45

non-sign.

NDW

Level of sign.

MC and the m-LC groups on the N1WS - a negative indication of the level of language development - also shows a trend in the direction expected (p difference MC - m-LC < .05). The i-LC group differs in a negative way from the m-LC group, on MLS (p difference < .05) and on N1WS (p difference < .01). At the same time, two language measures based on Bernstein's theory were applied: the ratio of the number of abstract nouns to the total number of nouns, and the number of sentences not related to the here and now. Both can give an indication of the abstracting capacity of the child. The scores of the groups on these two measures are given in table 7. TABLE 7 The group means on two measures of abstraction Social group Ratio abstract nouns/ total number of nouns sentences not about the here and now

i-LC

m-LC

MC

F-ratio

.0802

.0592

.1001

3.12

p < .05

9.14

4.46

20.50

5.77

p < .05

66

RESULTS

In both instances the social class differences are significant. Further analysis indicates that regarding the ratio abstract nouns/all nouns only the difference between the MC and the m-LC groups is significant, and regarding the sentences not about the here and now the MC group is positively different from both LC groups, whereas the difference between the LC groups is not significant at the 5 % level. In both cases the i-LC group is different from the m-LC group in a positive rather than in a negative way, whereas the difference from the MC group in the first case (ratio abstract to total number of nouns) is small. 3.2.3. The relation between the measures of syntactic and the different quantitative measures

complexity

Comparing two traditional language measures to two quantifying qualitative ones, Sharf 8 4 came to the conclusion that the MLS is satisfactory as a measure of syntactic complexity. We will now compare our newly developed non-quantifying but qualitative measure of syntactic complexity to the traditional and otherwise applied qualitative language measures in order to be able to establish whether the latter, which can more easily be calculated, can be as usefully applied in further research. Table 8 gives the product-moment correlation between the linguistic components found, the language measures under scrutiny, and the IQ. The scores on component I appear to relate closely to the scores obtained on the traditional measures and to the IQ. In three instances - the NDW, the MLS, and the ML5LS - the correlation with the first component of the syntactic complexity test is nearly perfect. Components II and III appear to be rather independent of the traditional measures. The negative correlation between component II and the IQ, the ratio between function words and all words, and the positive correlation with the N1WS, a negative measure, look acceptable. The rather high positive correlation of component II with the TTR (Carroll's adaptation) seems quite 84

D. Sharf, "Some Relationships between Measures of Early Language Development", pp. 64-74.

67

RESULTS TABLE 8

The relation between the structural complexity, the pertinent qualitive measures, and the IQ{*p< .05; ** p < .01) Comp. I

Comp. II

Comp. III

NDW

.92**

.06

.00

MLS

.96**

—.09

.13

TTR(C)

.58**

N1WS

—.70**

.76**

—.17

.35*

—.28

ML5LS

.93**

.17

.09

ratio function w.

.34*

—.34*

.31*

ratio abstract nouns

.41**

—.25

.00

sentences not on the here and now

.74**

.25

.16

IQ

.59**

—.32*

.25

surprising, but can be explained. A child who scores highly on component II deletes much. These deletions are for a major part function words, a class limited in number and closed to expansion, in opposition to the class of content words. The chopping of a limited number of frequently occurring words adds to the verbal diversity as measured by the TTR(C). Thus there is a negative correlation (—.34) between the ratio function words/total number of words and component II. 3.2.4. Developmental

differences

between

the social

classes

The relationship between chronological age and both the linguistic components identified and the traditional measures is presented for each group in graphical form in graphs A to E. The scores on the three components have been presented as z-scores to facilitate

68

RESULTS

comparison. The regression lines have been computed on the basis of a statistical analysis. The results presented in the graphs have to be interpreted carefully since we are dealing here with a cioss-sectional investigation; moreover, the regression lines per group are based on fourteen cases only. The interpretation of the results is more reliable on the basis of the exact testing of a possible differential developmental aspect between the groups, carried out with the help of a test for the homogeneity of the regression. Each of the components will be discussed with regard to chronological age. Component I (see graph A) Differences between the environments on component I were small at age 3.0 and increase around age 4. Precise testing for homogeneity of regression results in a perceivable difference with an F-ratioof2.20. Graph A:

Correlation between age and component I per social class

MC m-LC .+2

.+1

0

.-1

3,0 3,1 3,2 3,3 3,4 3,5 3,6 3,7 3,8 3,9 3,10 3,11 4,0

RESULTS

69

Since the critical value for F in testing at 10 % level is 2.48, only a trend can be established in the direction of an increase of the differences between the environments for component I. Component II (see graph B) A higher score on component II implies more approximation deletions and is thus a negative indicator for language skill. Graph B shows us that there is a tendency toward interaction between environment and age. The test for homogeneity of regression results in an F-ratio of 2.69 (p < .10). We feel that the following interpretion can be proposed: Graph B:

Correlation between Component II and age per social class

MC m-LC

3,0 3,1 3,2 3,3 3,4 3,5 3,6 3,7 3,8 3,9 3,10 3,11 4,0

Among children from a lower social class the deletion scores decrease, whereas in the language of children from a MC group more and more approximation deletions occur. For an explanation of this curious phenomenon we have to go back to component I. There it appeared that at age 3.0 there are only minor differences in the planning factor. At that age fewer approximation deletions appear to occur in the language of MC children, and the semantic

70

RESULTS

input is identical. That these children start applying more approximation deletions at a later age can be logically explained by the fact that their semantic input per sentence is more extensive than is the case in the other groups. For the children of the LC groups this means, conversely, that parallel to a lesser increase of the semantic input the number of approximation deletions decreases. At this point, we would like to come back for a moment to the result of the analysis of variance for the deletion component which resulted in no significant differences between the social classes on component II. This suggests that the notion deletion as a negative measure for language development is ready for revision. So far, the approximation deletion ("omission", cf. for example Menyuk 85 ) has been considered to be a negative indicator for language capacity. The result as shown in graph B in relation to that shown in graph A indicates that evaluation of omissions can only be accepted if they are weighed against the complexity of the semantic structure and the realization thereof (component I). Follow-up research could produce more decisive answers on the later relationship, as deletions will naturally decline in frequency. Component III (see graph C) Regarding component III the seeming divergence of the three groups cannot be taken as important as this difference is not significant (F-ratio of .51). More extensive research is needed regarding the difference in grammatical tense both for the Semantic and the Realization (morphology, and the like) level of language capacity. NDW and MLS Graphs D and E show the relation between chronological age and the Number of Different Words and the Mean Length of Sentence respectively. 85

P. Menyuk, "Comparison of Grammar of Children with Functionally Deviant and Normal Speech", JSHR (1964), pp. 109-21.

100 3,0 3,1 3,2 3,3 3,4 3,5 3,6 3,7 3,8 3,9 3,10 3,11 4,0

72

RESULTS

Graph E: Correlation between Mean Length of Sentence (MLS) and age per social class MC

7

6

5

MLS

4

3

3,0 3,1 3,2 3,3 3,4 3,5' 3,6 3,7 3,8 3,9 3,10 3,11 4,0

The non-normalized raw scores are marked on the Y-axis. For both the NDW and the MLS, the regression coefficients are given in tables 9 and 10. The regression of the pertinent measure on the age means that an estimated increase per month for the different groups is given. TABLE 9 Estimated increase per month on the NDW (Significance of developmental difference p < .10) Social groups Number of different words

i-LC

m-LC

MC

1.80

3.01

7.78

TABLE 10 Estimated increase per month on the MLS CSignificance of developmental difference p < .01) Social group Mean length of sentence

i-LC

m-LC

MC

.035

.089

.26

73

RESULTS

Regarding the MLS, not only do significant environmental differences appear to exist (see the analysis of variance, mentioned before in 3.2.2.: Environmental differences on the traditional language measures), but there are also reasons to assume that the social class groups, starting with minimal differences at age 3.0, diverge further toward age four. In terms of development with regard to the MLS measure, at the age of 4.0 the children of the LC groups are reaching the level that is comparable to the level of ten months younger children of the MC group. Some correlations between the other measures and chronological age have been computed. Table 11 gives these correlations for each social class group. TABLE 11 Product-moment correlations of some language measures with chronological age

Social groups

i-LC

m-LC

MC

totals

Ratio abstract nouns

.48

.41

.24

.38

Sentences not on the here and now

.34

.40

.58

.30

N1WS

.002

—.32

—.82

—.12

TTR(C)

.42

.44

.56

.43

The column totals show the correlation of the pertinent measure with chronological age computed over the total number of 42 subjects. These correlations indicate that the relation between measure and age is rather remote. Regarding the number of one word sentences, the social classes appear to differ strongly (significance of difference p < .01). In the MC group the N1WS decreases strongly, almost to 0 by age 4.0, whereas the decrease in the m-LC group is very small and in the i-LC group nonexistent.

74

RESULTS

Mental age The tendency towards diversification between the groups on the language measures is not reflected in the intellectual development. The developmental lines for the two groups investigated run nearly parallel (see graph F indicating the relation between chronological and mental age for the i-LC and the MC groups). Graph F: Correlation between Mental Age (Terman) Age (CA) for the MC and i-LC

(MA) and

Chronological

groups.

4,10 4,8 4,6 4,4 4,2 4,0 MA 3,10 3,8 3,6 3,4 3,2

3,0

3,2

3,4

3,6

3,8

3,10

4,0

CA

Testing homogeneity of regression did not reveal any significant difference in the developmental progress of the three groups. With identical chronological age, the difference in mental age between

RESULTS

75

the i-LC (n=14) and the m-LC ( n = l l ) groups is significant at the 1 % level. At the same time, the result indicates that children of the i-LC group at a given chronological age reach a nearly corresponding mental age. In other words, these children are of normal intelligence even though an extensive investigation of the school career of children from this isolated environment indicated that by age eleven and twelve one out of five of these children was attending a school for the mentally retarded. 3.2.5.

Environment and play contacts

Analysis of the differences between the two LC groups has demonstrated that the i-LC group was inferior to the m-LC group on a number of language producing measures, especially on the N1WS, the MLS, the relative number of function words, and the grammatical tense as measured according to component III, and less so on the planning factor as measured in component I. Generally speaking, the developmental perspective on these measures is less favorable for the i-LC group. As this i-LC group comes out reasonably on both measures for abstraction and also in the IQ test, it seems to be in order to investigate further what kind of influences are working specifically on the use of language. Such post hoc hypotheses can give indications for further research. For the differentiation of both LC groups, the composition of the play groups in which the children were playing in the morning and the limitation of play contacts with children from different environments imposed by the housing situation have been decisive factors. Within the i-LC group children, the play contact is limited to their own environment; the part of town in which they grow up seems to be designed, so to say, to isolate the families that live there. This isolation is further reinforced by the geographical situation: the district in question is cut off from the rest of town by main traffic routes. Children of the m-LC group play in nursery schools also visited by middle and higher class children. The living situation of this m-LC group is, moreover, far more heterogeneous than that of the i-LC group.

76

RESULTS

Part of the differences between the groups may be accounted for through a selection factor. This threshold would hardly play a role for the i-LC group's choice of play facilities as the latter are offered nearly for free and are situated within their own district. Children of the m-LC group frequent play facilities that ask a fairly high financial contribution. That this leads to a threshold of admission is clear from the fact that children from lower social classes visit these nursery schools considerably less frequently than could be expected from a nonselective sample. The i-LC group is composed of five children from skilled and nine children from unskilled working class families. The m-LC group is represented by six children of the skilled working and eight of the unskilled working class. It was decided to split up both groups into skilled vs. unskilled in order to investigate the effect of the play contacts, the effect of the schooling of the father, and the interaction of these two factors separately. The i-LC group has almost exclusively homogeneous play contact because of its isolation. The m-LC group, on the other hand, has heterogeneous play contacts, i.e. with children from the middle class environment. For this investigation we used as dependent variables the NDW, the MLS, and the planning factor of the syntactic complexity score. Tables 12, 13, and 14 show the mean values. TABLE 12 Number of different words (for explanation see text)

skilled

unskilled

homogeneous play contacts

115.6

123.7

120.8

heterogeneous play contacts

135.8

121.6

127.0

126.6

122.7

Social background

An analysis of variance was carried out per measure in order to find out the respective influence of schooling, play contacts

RESULTS

77

TABLE 13 Mean length of sentence (for explanation see text) Social background

skilled

unskilled

Homogeneous play contacts

3.12

3.43

3.32

Heterogeneous play contacts

4.47

3.48

3.90

3.86

3.45

TABLE 14 Component I (for explanation see text) Social background

skilled

unskilled

Homogeneous play contacts

17.6

24.0

21.7

Heterogeneous play contacts

33.2

22.4

27.0

26.1

23.2

(i-LC vs. m-LC), and possible interaction. The results of these analyses are given in table 15. The results for the M L S measure TABLE 15 Summary of the results of the analysis of variance on schooling and play contacts Measures

Schooling

Play contacts

Interaction

NDW

not sign.

not sign.

not sign.

MLS

pC.05

not sign.

p

> co O £

> -i D, 13

a£>

cs

vo

cS u 3 C3 "3 8

>

+ +

J2 0 1p.

(3 •a a

I

o

u CO

t/5 Ô •

I I ih

Io

S M

w .2

.o

o

aj

.S "S

VI

us

g

S 2

a

£ 0

1 u Q

D 13

ao •a JD 3 w

®

,22

'-3

•o' o 13 S

X>' o

i-l

s

O

123

APPENDICES

o S I S f t

APPENDIX II

PRELIMINARY FACTOR ANALYSIS ROTATED FACTOR MATRJX (3 factors) (The numbers between parentheses correspond with those of the syntactic complexity score in Appendix I.) variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

20 21 22 23 24 25 26

(l,a) (l,b) (l,c) (3) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (17) (18) (19) (20) (21) (22) (23) (24) (25) (26) (27) (28)

factor 1

factor 2

factor 3

0.53 -0.07 0.18 0.52 0.66 0.79 -0.10 -0.08 -0.03 0.42 0.70 0.65 -0.11 -0.29 0.79 0.40 -0.01 0.64 0.66 0.28 0.30 0.21 -0.17 -0.20 0.31 -0.21

-0.03 0.55 -0.12 0.31 0.12 0.08 -0.09 -0.05 -0.12 -0.23 0.12 0.03 0.05 -0.06 -0.15 0.03 0.36 0.29 0.27 0.51 -0.02 0.39 0.06 -0.13 0.31 -0.48

-0.20 -0.22 0.54 -0.25 -0.18 0.01 -0.38 0.20 0.41 0.01 0.23 0.25 0.48 0.35 0.16 0.31 0.53 0.18 -0.03 -0.07 -0.28 0.08 -0.49 0.17 0.33 -0.09

communality %*

63

56

67 53 50

* The percentages of communality are only mentioned when they are higher than 50.

125

APPENDICES

variables

factor 1

factor 2

factor 3

27 28 29 30 31 32 33 34 35 36 37 38 39 40

(30) (32) (33) (34) (35) (38) (39) (40-41) (42) (43) (44) (45) (46) (47)

-0.12 0.58 0.84 0.82 0.87 0.51 0.75 0.36 0.31 0.81 -0.34 -0.24 -0.08 0.49

0.09 0.16 0.16 0.18 0.13 -0.14 0.18 0.47 0.30 -0.29 0.58 -0.11 0.19 0.14

-0.26 0.13 -0.05 -0.03 0.05 0.66 -0.05 -0.41 0.04 0.17 -0.03 -0.76 0.29 -0.01

41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71

(54) (55) (56-57) (58) (59) (59) (61) (62) (63) (64) (65) (66) (67, 69) (68) (70) (71) (72) (73) (74) (75) (76) (78) (79, 80, 81) (82-88) (86,a) (89, 90, 91) (89, 90) (92, 93) (94) (95) (97)

0.57 0.42 0.36 0.33 0.63 -0.16 -0.14 0.73 0.66 0.21 0.33 0.28 0.77 -0.05 0.69 0.68 0.64 -0.01 0.46 0.46 0.59 0.53 0.58 0.81 0.29 0.51 0.03 0.32 -0.12 -0.18 -0.19

0.29 0.52 0.55 0.38 0.66 0.03 -0.11 0.29 0.33 0.21 0.29 0.56 0.11 0.04 0.11 0.14 0.44 0.68 0.17 0.32 0.62 0.16 0.33 0.17 0.25 0.24 -0.06 0.23 -0.06 -0.02 0.62

0.11 -0.10 -0.40 -0.55 0.03 -0.63 -0.63 0.25 0.16 -0.09 0.29 0.16 0.09 -0.42 0.24 0.22 -0.04 -0.34 -0.03 -0.45 -0.19 0.19 0.03 -0.01 0.22 -0.01 0.48 0.09 -0.61 -0.60 -0.01

communality %

73 71 78 71 61 52 77 65

60 56 83 69 57

62 55 53 61 58 52 78 68

126

APPENDICES

variables

CD+ CD+ CD+ CD+ CD+ CD+ CD+ CD+ A+ A+ S+ S+ A+ A+ AD+ AO+ AO+ AD+ AD+ AD+ AD+ AD+ AD+ AD+ AD+ AD+ AD+ AD+ AD+

72 98) 73 99) 74 100) 75 101) 76 102) 77 103, 105, 106) 78 104) 79 107) 80 108, 110) 81 111) 82 112) 83 113) 84 115) 85 121, 122) 86 123, 124) 87 126, 127) 88 128) 89 52) 90 53, 54, 55) 91 59) 92 62, 63) 93 66-69) 94 70-76) 95 89) 96 92) 97 94-109) 98 111,a) 99 l l l . b ) 100 111.C) 101 112) 102 113) 103 115) 104 121-122) 105 123) 106 124) 107 126) 108 127) 109 128) Factor variance

factor 1

factor 2

factor 3

communality °/0

-0.30 0.08 -0.11 0.63 0.71 0.23 0.00 0.28 0.84 -0.04 -0.20 -0.35 0.07 0.27 0.00 -0.15 -0.07 -0.31 -0.05 -0.25 -0.10 0.01 -0.32 -0.32 0.00 -0.09 -0.37 -0.25 -0.27 -0.15 -0.09 -0.33 -0.29 -0.28 -0.11 -0.28 -0.26 -0.36 19,87

0.65 0.38 0.32 0.22 0.08 -0.03 0.57 0.32 0.18 0.14 -0.01 -0.10 0.16 0.15 -0.18 0.09 -0.00 -0.66 -0.15 0.27 0.26 -0.49 -0.04 -0.69 -0.25 -0.18 -0.71 -0.29 -0.82 -0.49 -0.35 -0.54 -0.46 -0.68 -0.36 -0.71 -0.53 -0.74 13,16

0.01 -0.04 -0.24 0.09 -0.03 -0.22 0.28 -0.02 0.01 0.46 0.16 0.19 -0.21 0.06 0.32 0.43 0.20 -0.02 -0.28 -0.25 0.12 -0.10 0.16 0.08 -0.29 -0.09 -0.14 0.10 0.12 0.11 0.09 -0.18 0.04 -0.00 0.03 -0.13 -0.06 -0.07 8.02

51

52

74

54

59 65 75

54 60 68

+ CD = correct deletion; AD = approximation deletion; A = approximation; S = substitution; O = order.

APPENDIX 111

DEFINITE FACTOR ANALYSIS

FI

F II

F III

Comm V /o

1= 2= 3= 4= 5= 6= 7=

6 11 15 18 19 29 30

8= 9= 10= 11 = 12= 13 = 14= 15 = 16= 17 = 18 = 19 = 20= 21 = 22 = 23 = 24= 25= 26= 27= 28=

31 32 33 34 36 38 43 44 45 48 49 53 55 56 57 58 60 61 64 72 76

Prepositional object Singular noun phrase Indefinite determiner Tense adverb Degree adverb Adverbial clause Subordination, nucleus function Coordination Present tense Past tense Perfect tense Negation Imperative Tense auxiliaries Participle morpheme Agreement subject verb Agreement determiner noun Agreement adjective noun Prepositions within nucleus Plural noun Singular noun Pronom. 1st pers. Pronom. 2nd person Marked pronoun Unmarked pronoun Subordinating conjunction Word order wh-question Word order S V Adv O

83 76 71 74 73 88

33 33 48 -03 08 21

09 02 -14 02 -06 01

81 69 75 55 54 82

88 88 45 83 61 63 -15 67 55 92 84 77 84 69 75 81 37 62 85 88 15 73

22 29 52 18 -24 55 -02 -31, -22 -29 14 06 32 27 31 -12 -47 -10 -33 20 -41 20

05 -06 -61 20 62 -26 58 62 67 -07 -19 -08 06 -14 00 04 26 40 06 07 39 -02

83 87 84 76 81 77 36 94 80 93 76 61 81 57 65 68 42 56 84 81 35 58

128

APPENDICES F I

29= 80 Word order Main sentence after subord. 30 = 89 Wrong deletion of tense 31 = 95 Wrong deletion copula 32 = 98 Wrong deletion subject 1st pers. 33 = 100 Wrong deletion subject 3rd pers. 34= 105 Wrong deletion deictic "this", etc. 35 = 107 Wrong deletion indef. determiner 36 = 109 Wrong deletion modal auxiliary

Explained variance:

FII

Fill

Comm

%

91 -42 -47

23 60 62

00 33 17

87 65 63

-53

61

37

78

-50

80

14

90

-45

69

28

76

-47

67

39

83

-50

67

34

82

47.69

15.51

8.81

Unrotated factor-analysis on 36 variables.

APPENDIX IV

PRINCIPAL COMPONENT ANALYSIS (for the meaning of the items see Appendix III)

item 1= 2= 3= 4= 5= 6= 7= 8= 9= 10= 11 = 12= 13 = 14= 15 = 16= 17= 18 = 19 = 20= 21 = 22 = 23 = 24= 25 = 26= 27 = 28 = 29 =

6 11 15 18 19 29 30 31 32 33 34 36 38 43 44 45 48 49 53 55 56 57 58 60 61 64 72 76 80

component I

component II

component III

-0.191 -0.173 -0.157 -0.174 -0.170 -0.205 -0.204 -0.203 -0.092 -0.192 -0.137 -0.138 0.066 -0.153 -0.121 -0.219 -0.197 -0.182 -0.192 -0.158 -0.171 -0.193 -0.080 -0.140 -0.204 -0.203 -0.022 -0.168 -0.210

-0.145 -0.154 -0.230 0.026 -0.040 -0.096 -0.100 -0.143 -0.269 -0.068 0.167 -0.272 0.127 0.196 0.169 0.112 -0.088 -0.040 -0.145 -0.141 -0.144 0.054 0.285 0.105 0.153 -0.090 0.283 -0.093 -0.105

0.152 0.111 0.009 -0.060 -0.061 0.034 0.091 0.031 -0.363 0.187 0.350 -0.059 0.293 0.310 0.367 -0.113 -0.087 -0.051 0.128 -0.043 0.093 -0.026 -0.051 0.188 -0.086 0.079 0.067 0.004 0.043

130

APPENDICES item

3 0 = 89 3 1 = 95 3 2 = 98 33 = 100 34=105 35 = 107 36=109

component I

0.141 0.153 0.169 0.160 0.149 0.154 0.162

component II

-0.171 -0.198 -0.166 -0.267 -0.216 -0.194 -0.192

component III

0.182 0.068 0.212 0.095 0.187 0.258 0.188