The Crosslinguistic Study of Language Acquisition, Volume 1: The Data 0898593670, 9780898593679

Table of contents :
Contents
Format and Abbreviations for Glosses
Introduction: Why Study Acquisition Crosslinguistically? • Dan I. Slobin
SPOKEN LANGUAGES
1. The Acquisition of English • Jill G. de Villiers and Peter A. de Villiers
2. The Acquisition of German • Anne E. Mills
3. The Acquisition of Hebrew • Ruth A. Berman
4. The Acquisition of Japanese • Patricia M. Clancy
5. The Acquisition of Kaluli • Bambi B. Schieffelin
6. The Acquisition of Polish • Magdalena Smoczyhska
7. The Acquisition of Romance, with Special Reference to French • Eve V. Clark
8. Variation and Error: A Sociolinguistic Approach to Language Acquisition in Samoa • Elinor Ochs
9. The Acquisition of Turkish • Ayhan A. Aksu-Koc and Dan I. Slobin
SIGNED LANGUAGES
10. The Acquisition of American Sign Language • Elissa L Newport and Richard P. Meier
Subject Index
Author Index

Citation preview

T H E CROSSLINGUISTIC STUDY OF LANGUAGE ACQUISITION Volume 1: The Data

T H E CROSSLINGUISTIC STUDY OF LANGUAGE ACQUISITION Volume 1: The Data Edited by DAN ISAAC SLOBIN University of California, Berkeley

L A W R E N C E E R L B A U M ASSOCIATES, PUBLISHERS Hillsdale, N e w Jersey London

Copyright © 1985 by Lawrence Erlbaum Associates, Inc. All rights reserved. No part of this book may be reproduced in any form, by photostat, microform, retrieval system, or any other means, without the prior written permission of the publisher. Lawrence Erlbaum Associates, Inc., Publishers 365 Broadway Hillsdale, New Jersey 07642

L i b r a r y of Congress Cataloging-in-Publication Data Main entry under title: The crosslinguistic study of language acquisition. Includes bibliographies and indexes. Contents: v. 1. The data — v. 2. Theoretical issues. 1. Language acquisition. I. Slobin, Dan Isaac 1939P118.C69 1985 401'.9 85-27411 ISBN 0-89859-367-0 (set)

Contents

Volume

1: THE

DATA

Format a n d A b b r e v i a t i o n s f o r Glosses

Introduction: Why Study Acquisition Crosslinguistically? Dan I.

Slobin

Crosslinguistic Study as a Method in Developmental Psycholinguistics 4 Null Hypothesis: Developmental Universals Hypotheses of Specific Language Effects The Plan of the Book 18 References 22

SPOKEN L A N G U A G E S 1.

The A c q u i s i t i o n of English Jill G. de Villiers and

2.

Introduction The Data Conclusion

27 38 124

References

128

Peter A. de

The A c q u i s i t i o n of German Anne

E.

Mills

Introduction The Data References

141 158 249

Villiers

5 15

vi 3.

CONTENTS

The A c q u i s i t i o n of Hebrew Ruth A.

255

Berman

Introduction The Data

255 280

The Setting of Language Acquisition Conclusions 347 Postscript 363 References 364

329

(Hungarian Language A c q u i s i t i o n as an Exemplification of a General Model of Grammatical Development Brian MacWhinney—see Volume 2)

4.

The A c q u i s i t i o n of Japanese Patricia

M.

Introduction 373 The Data 383 The Setting of Language Acquisition Conclusions 502 List of Abbreviations 515 References 576

5.

470

The A c q u i s i t i o n of Kaluli Bambi

B.

Emphatic and Other Discourse Particles Conclusions 589 References 592 The A c q u i s i t i o n of Polish Magdalena

Introduction The Data Conclusions References

525

Schieffelin

Kaluli Language and Cultural Content Overall Sketch of Development 534 Word Order 543 Personal Pronouns 549 Nominal Casemarking 555 Verbs 566

6.

373

Clancy

Smoczyhska

595 620 667 681

525

586

595

CONTENTS

7.

The A c q u i s i t i o n of R o m a n c e , w i t h Special Reference t o French Eve V.

687

Clark

Introduction 688 The Data 702 The Setting of Language Acquisition Conclusions 756 Coda 762 References 767 8.

Vii

743

Variation a n d Error: A S o c i o l i n g u i s t i c A p p r o a c h to Language A c q u i s i t i o n in S a m o a Elinor

783

Ochs

783

What is an Error?

A Sociolinguistic Sketch of Samoan 788 Sources of Data 809 Overall Course of Development 810 Acquisition of Two Phonological Registers Acquisition of Word Order 821

812

Acquisition of Ergative Case Marking 826 Acquisition of Ergativity Through Word Order 831 Relating Sociolinguistics to Language Acquisition 833 References 836 9.

The A c q u i s i t i o n of T u r k i s h Ayhan

A. Aksu-Koc

and

Dan I.

839 Slobin

Introduction 839 The Data 847 The Setting of Language Acquisition Conclusions 869 References 876

861

SIGNED L A N G U A G E S 10.

The A c q u i s i t i o n of A m e r i c a n Sign Language Elissa L Newport

Introduction The Data

and

Richard

P.

Meier

881 907

The Setting of Language Acquisition Conclusions References

927 933

Subject Index 939 A u t h o r Index 947

975

881

Viii

CONTENTS

Volume 2: Theoretical

11.

Cognitive Prerequisites: The Evidence F r o m Children Learning English Judith

12.

R.

M.

MacWhinney

C r o s s l i n g u i s t i c Evidence f o r t h e Language-Making Capacity Dan I.

16.

Peters

Hungarian Language A c q u i s i t i o n as an Exemplification of a General Model of Grammatical Development Brian

15.

Givon

Language S e g m e n t a t i o n : Operating Principles For the Perception a n d A n a l y s i s of Language Ann

14.

Johnston

F u n c t i o n , Structure, a n d Language A c q u i s i t i o n Talmy

13.

Issues

Slobin

What Shapes Children's G r a m m a r s ? Melissa

Bowerman

Format and Abbreviations for Glosses*

A l l foreign language examples are given i n Italics. (Small caps are used for emphasis and other usual functions o f Italics.) I n running text, English glosses and grammatical codes are given i n single quotes, and optional free translations follow i n parentheses, indicated by an equal sign and single quotes. Grammatical codes are always given i n capital letters (see list, below). For example: gel-me-di-n 'come-NEG-PAST-2SG' (= 'you didn't come'). I n interlinear format, translation equivalents appear below each foreign element, and the free translation is placed below i n single quotes: gel

-me

-di

-n

come N E G PAST 2SG 'you didn't come' Hyphens i n a gloss always correspond to hyphens i n the foreign example. I f one foreign element corresponds to more than one English element and/or gram­ matical code, the collection o f meaning equivalents is joined by colons; e.g. gel-medin c o m e - N E G : P A S T : 2 S G \ or even gelmedin ' c o m e : N E G : P A S T : 2 S G \ 4

*The abbreviations are adapted from a list used by Bernard Comrie (The languages of the Soviet Union, Cambridge University Press, 1981, p. xv). The format is based on useful suggestions offered by Christian Lehmann in "Guidelines for interlinear morphemic translations: A proposal for a standardization" (Institut für Sprachwissenschaft, Universität Köln, Arbeitspapier Nr. 37, 1980). The system presented here is offered as a proposal for standardization in child language studies.

ix

X

FORMAT A N D ABBREVIATIONS

I f it is relevant to indicate the possibility o f segmentation, equal signs can be used in place o f colons. The preceding example consists o f segmentable morphemes, and could also be glossed, for example, as gel-medin ' c o m e - N E G + P A S T + 2 S G ' . Use o f colons is neutral i n regard to the possibility o f segmentation, and i n most instances either colons or hyphens are used. (The degree o f precision o f segmenta­ tion and glossing o f an example, o f course, depends on the role it plays i n the exposition.) I f a grammatical code consists o f t w o words or abbreviations they are j o i n e d by a period; e.g. D E F . A R T means ''definite a r t i c l e . " Combining the principles for use o f colons and periods, consider the gloss for the German definite article i n its masculine singular accusative form: den ' D E F . A R T : M A S C : S G : A C C \

LIST OF G R A M M A T I C A L CODES 1 First Person 2 Second Person 3 Third Person ABESS Abessive ('without X ' ) A B L Ablative ('from X ' ) ABS Absolutive ACC Accusative ACT Active ADESS Adessive ('toward X ' ) ADJ Adjective, Adjectival ADV Adverb(ial) AFFIRM Affirmative AGR Agreement AGENT Agent A L L A T Allative ('to(wards) X ' ) AN Animate AORIST Aorist ART Article ASP Aspect AUG Augmentative AUX Auxiliary BEN Benefactive CAUS Causative C L Clitic CMPLR Complementizer COMIT Comitative ('(together) with X') COMP AR Comparative COMPL Completive CONC Concessive COND Conditional CON J Conjunction

CONN Connective CONSEC Consecutive CONT Continuous, Continuative COP Copula DAT Dative D E C L Declarative D E F Definite DEICT Deictic DEM Demonstrative DESID Desiderative DIM Diminutive DIREC Directional DO Direct Object DYN Dynamic (Nonstative) E L A T Elative ('out of X ' ) EMPH Emphatic E R G Ergative ESS Essive ('as X ' ) EVID Evidential E X C L Exclusive EXIS Existential FACT Factive FEM Feminine FIN Finite FOC Focus FUT Future GEN Genitive HAB Habitual HON Honorific HUM Human I L L Illative ('into X ' )

FORMAT A N D

IMP Imperative INAN Inanimate INCH Inchoative INCL Inclusive INDEF Indefinite INESS Inessive ('in X ' ) INF Infinitive INFER Inferential INSTR Instrumental INT Interrogative INTENT Intentive INTERJ Interjection INTRANS Intransitive 10 Indirect Object IPFV Imperfective IRR Irrealis ITER Iterative LOC Locative MASC Masculine MOD Modal N Noun NEG Negative NEUT Neuter NEUTRAL Neutral NOM Nominative NOML Nominal NONPAST Non-past NONVIR Non-virile NUM Numeral, Numeric OBJ Object OBLIG Obligatory OPT Optative PART Participle PARTIT Partitive PASS Passive PAST Past PAT Patient PERF Perfect PFV Perfective PL Plural POL Polite POSS Possessive

ABBREVIATIONS

POST Postposition POT Potential PP Past Participle PRE Prefix PREP Preposition PRES Present PRESUM Presumptive PRO Pronoun PROG Progressive PROL Prolative ('along X ' ) PTL Particle PURP Purposive Q Question QUANT Quantifier QUOT Quotative RECENT Recent RECIP Reciprocal R E F L Reflexive R E L Relative REM Remote REPET Repetition REPORT Reportative RES Resultative SG Singular SIMUL Simultaneous STAT Stative SUBJ Subject SUBJV Subjunctive SUBL Sublative ('onto X ' ) SUFF Suffix SUPER Superessive ('on X ' ) SUPERL Superlative TEMP Temporal TNS Tense TOP Topic TRANS Transitive TRANSL Translative ('becoming X ' ) V Verb VIR Virile VN Verbal noun VOC Vocative

T H E CROSSLINGUISTIC STUDY OF LANGUAGE ACQUISITION Volume 1: The Data

Volume 1 THE DATA

Introduction: Why Study Acquisition Crosslinguistically?

Dan

I. S l o b i n

University

of California,

Berkeley

Contents Crosslinguistic Study as a Method in Developmental Psycholinguistics 4 Null Hypothesis: Developmental Universals 5 Conceptual Development and the Construction of Grammar 6 Formal Pattern Preferences and Grammar 10 Concepts are Combined in Grammatical Morphemes According to Semantic Affinities 10 Grammatical Morphemes are Positioned According to Their Scope of Operation 11 Grammatical Markers are placed According to Principles of Semantic Relevance 12 Morphological Systems are Constructed According to Formal (Non-semantic) Criteria 13 Hypotheses of Specific Language Effects 15 Form-Function Interaction Influences Rate or Sequence of Development 15 Particular Linguistic Forms are Relatively More Accessible, Holding Content Constant 16 Separate Marking of Notions in Particular Languages Reveals a Conceptual Substratum 16 Language-specific Co-occurrences Reveal Patterns of Conceptual Relavance 17 Crosslinguistic Differences in Degree of Coding an Area Reveal Linguistic Capacities 18 The Plan of the Book 18

Child language study can be more than the analysis of individual instances of language development, because it is possible to for­ mulate laws of formation that are operative in every child lan­ guage. This commonality is even international; therefore we,will not need to limit our evidence solely to German children.

1

—Clara and Wilhelm Stern, Die Kindersprache (1907)

' U n d dennoch kann die Kindersprachkunde mehr als die Analyse individueller Sprachent­ wicklungen sein; denn sie vermag Bildungsgesetze zu formulieren, die in jeder Kindersprache wirksam sind. Diese Gemeinsamkeit ist sogar international; wir werden unsere Belege daher nicht ausschließlich auf deutsche Kinder zu beschränken brauchen." l4

3

4

Slobin

C R O S S L I N G U I S T I C S T U D Y A S A M E T H O D IN DEVELOPMENTAL PSYCHOLINGUISTICS I n 1907, when the Sterns published the first edition o f their great diary study o f German child language, they were able to point to a significant body o f nine­ teenth century research on the acquisition o f a number o f European languages. They, and their contemporaries, were aware that there were clear crosslinguistic parallels i n development, based on common principles o f child psychology. The goals o f investigators were directed beyond individual languages to the discovery of general principles. For example, the great Russian diarist, A . N . Gvozdev, publishing an article i n 1928 under the title, ' T h e significance o f the study o f child language for l i n g u i s t i c s , " noted that (1961, p . 9): The acquisition of the native language follows strict regularities and is charac­ terized by the same features in different children. This supports the idea that native language acquisition is determined by general psychophysiological conditions which function uniformly in all people, thus leaving their mark on the strcuture of language. 2

The emphasis was on the universal, rather than the particular. The value o f data from various languages was the same as that o f data from various children: to demonstrate commonalities. A n d w i t h the rise o f an insular American psychol­ ogy, ' 'language development" became a summary o f the facts o f the acquisition of English, taken as representative o f general patterns. Thus Dorothea McCarthy could summarize a large body o f systematic observational studies and conclude, in 1950, that such studies: have yielded considerable uniformity of results, and a fairly accurate description can now be given of linguistic development in the age range of two to five years (p. 165). It is the burden o f the present collection o f studies to demonstrate that crosslinguistic study does more than reveal uniformities o f development, because properties o f individual languages influence the course o f development. Begin­ ning w i t h Melissa Bowerman's study o f the acquisition o f Finnish i n 1965 (Bowerman, 1973), followed by a series o f crosslinguistic studies organized by John Gumperz, Susan E r v i n - T r i p p , and Dan Slobin at Berkeley (Slobin, 1967), it has become clear that different types o f languages pose different types o f acquisition p r o b l e m s . One cannot study universals without exploring particulars 3

"Usvoenie rodnogo jazyka po otnoseniju ko mnogim gruppam jazykovyx javlenij proxodit so strogoj zakonomernost'ju i xarakterizuetsja u raznyx detej odnimi i temi ze certami; i eto podtverzdaet mysl' o torn, cto usvoenie rodnogo jazyka opredeljaetsja takimi obscimi psixofiziologiceskimi uslovijami, kotorye dejstvujut edinoobrazno u vsex ljudej i kotorye poetomu kladut svoj otpecatok i na strukturu jazyka." The following Berkeley dissertations on the acquisition of various native languages emerged 2

3

Introduction

5

(Slobin, 1982). The w o r l d provides us w i t h a marvelous set o f "natural experi­ ments," i n w h i c h children w i t h similar endowments master languages o f varying forms. Gvozdev was right that psychophysiological commonalities "leave their mark on the structure o f language," but that mark is on a more abstract l e v e l — the level o f language universals. B y combining attention to universals and partic­ ulars, we are beginning to discern a more differentiated picture o f child lan­ guage—one i n w h i c h we can see w h y patterns o f acquisition o f specific proper­ ties V A R Y from language to language, while they are determined by common principles o f a higher order. Such principles are summarized i n m y concluding chapter (Slobin, 1985). The task o f this Introduction is to spell out the ways i n which crosslinguistic study constitutes a M E T H O D for the discovery o f general principles o f a c q u i s i t i o n . 4

There are t w o major pacesetters to language development, involved w i t h the poles o f function and o f form (Slobin, 1973): (1) on the functional level, devel­ opment is paced by the growth o f conceptual and communicative capacities, operating i n conjunction w i t h innate schemas o f cognition; and (2) on the formal level, development is paced b y the growth o f perceptual and informationprocessing capacities, operating i n conjunction w i t h innate schemas o f grammar. The course o f acquisition o f any particular linguistic form reflects an interaction of the child's abilities to decipher and cognize both structure and content. B y examining the meanings o f children's grammatical forms crosslinguistically, we can determine conceptual starting points for linguistic form; and by examining the forms o f child grammar we can determine children's strategies for construct­ ing morphosyntactic systems. The crosslinguistic method can be used to reveal both developmental univer­ sals and language-specific developmental patterns i n the interaction o f form and content. L e t us first examine evidence for the N U L L HYPOTHESIS that language development is everywhere the same, and then turn to HYPOTHESES O F SPECIFIC L A N G U A G E EFFECTS upon the course o f development.

NULL HYPOTHESIS: DEVELOPMENTAL UNIVERSALS The null hypothesis, as discussed earler, guided early work i n child language development, and i t still provides an important part o f the picture. Wherever we from A field manual for cross-cultural study of the acquisition of communicative competence (Slobin, 1967): Arabic (Badry, 1983), Black English (Mitchell-Kernan, 1969), Finnish (Argoff, 1976), Hungarian (MacWhinney, 1973), Japanese (Clancy, 1980), Luo (Blount, 1969), Mandarin (Erbaugh, 1982), Samoan (Kernan, 1969) Serbo-Croatian (Radulovic, 1975), Spanish (Eisenberg, 1982), Turkish (Aksu, 1978), Tzeltal (Stross, 1969). I n this chapter, various pieces of data from the volume are presented as examples of the use of the crosslinguistic method. Many of the same facts are also discussed in my final chapter, "Crosslinguistic evidence for the Language-Making Capacity," where they are used to support the formulation of general "operating principles" for acquisition. (See volume 2.) 4

6

Slobin

find similar patterns o f development crosslinguistically, i n function or i n f o r m , we see evidence for strong developmental universals which may operate across all settings. O n the functional level, there is evidence for the primacy o f concep­ tual development i n providing the first meanings for grammatical forms and i n pacing the course o f development o f certain forms. A n d on the formal level, there is evidence for general language acquisition strategies that take precedence over the constraints o f particular linguistic forms i n individual languages.

Conceptual Development and the Construction of Grammar The acquisition studies o f numerous languages reported i n this volume reveal conceptual underpinnings o f child grammar that could not be determined by study o f any one language i n isolation. This is because the crosslinguistic method allows us to track the course o f acquisition o f particular semantic notions across a range o f differing surface expressions. Children acquire word-order patterns and morphological markers, such as case inflections, adpositions, and verbal inflec­ tions, before they master the full range o f uses o f these grammatical devices i n the adult language. Across languages, common meanings are assigned to diverse forms. The point can be made by examining the f o l l o w i n g four propositions about the ways i n w h i c h children use grammatical marking to express l i n ­ guistically relevant notions. 1. Conceptual for

Development

Grammatical

Provides

Starting

Points

Marking.

One o f the clearest examples o f this proposition is offered by comparison o f the notions expressed by children i n their early marking o f transitivity. I n child speech i n various languages early grammatical marking o f agent-patient relations (accusative or ergative inflections, word-order patterns) focuses on basic causal events i n w h i c h an agent carries out a physical and perceptible change o f state i n a patient by means o f direct body contact or w i t h an instrument under the agent's control (Slobin, 1981). Crosslinguistic comparison shows a striking U N D E R extension o f both accusative and ergative inflections to such " h i g h l y transitive" (Hopper & Thompson, 1980) events. For example, Gvozdev (1949) noted that i n his son's acquisition o f Russian, the accusative inflection was apparently first limited to the direct objects o f verbs i n v o l v i n g manipulative physical action on things—such as ' g i v e ' , 'carry', ' p u t ' , and ' t h r o w ' , while uninflected nouns served as objects o f verbs like 'see' and 'read'. Investigating the acquisition o f K a l u l i , an ergative language o f N e w Guinea, Schieffelin (1985) found that the ergative inflection first appears only on the subjects o f verbs such as ' g i v e ' , 'grab', 'take', and ' h i t ' , and that it tends to be omitted i n sentences w i t h verbs such as 'say', 'call-out', and 'see'. She has found i n addition, that the ergative is used earlier, and w i t h greater consistency, i n utterances w i t h past-tense verbs, as opposed to present or future; and that negated verbs tend to be accompanied by

Introduction

7

grammatically unmarked agents. I n all o f these instances an event has to be overtly manipulative and actually realized i n order to receive ergative marking. Furthermore, i n t w o quite different ergative languages, K a l u l i and Quiche (Pye, 1979), children do not extend ergative markers to intransitive constructions, indicating that they are specifically grammaticizing only agents o f highly tran­ sitive, manipulative activities. These examples suggest that such activities con­ stitute a central semantic organizing point for grammatical m a r k i n g — a starting point w i t h a particular salient conceptual basis from the point o f view o f child development. Crosslinguistic study is necessary to discover and define such conceptual bases o f grammatical marking. 2. Preferred Event for Grammatical

Perspectives Marking.

Provide

Starting

Points

W e have noted that children focus on events that have been realized and actually completed. T a k i n g a perspective on immediate results has consequences for the meanings underlying the first uses o f forms used for marking tense, aspect, and voice crosslinguistically. For example, the first uses o f past-tense, perfect, or perfective verb inflections seem to comment on an immediately completed event that results i n a visible change o f state or location o f some object, w i t h later development into a more general past tense. The past tense or perfect appears first on verbs like ' f a l l ' , ' d r o p ' , 'break', and ' s p i l l ' across a very wide range o f languages and language types. Focus on results often precludes or excludes focus on agency. There are examples i n child speech i n w h i c h a past-tense form is reinterpreted as a descrip­ tion o f an affected object, such as attempts by Italian children to make the past participle o f transitive verbs agree i n number and gender w i t h the direct object, though the participle does not agree w i t h the object i n the input language ( A n t i nucci & M i l l e r , 1976). I n Hebrew (Berman, 1985), children appear to simplify the system o f verb patterns into an opposition between a middle-voice " r e s u l t " perspective and a transitive "agentive perspective." A n d i n Turkish (Savasir & Gee, 1982) the first passive verb forms are used i n situations i n w h i c h a c h i l d , having failed to bring about a desired result, focuses on the object o f manipula­ tion w i t h a negative, third-person passive. For example, a child tries to open a door, fails, and says the equivalent o f ' i t isn't being opened', thereby shifting attention from her o w n action to the resisting object. I n instances such as these, children narrow the meaning o f a particular grammatical m a r k i n g — a verb par­ ticiple, a derivational verb pattern, a passive marker—to express a conceptually salient perspective on events o f a particular type. 3. Universal Language

Conceptual Patterns.

Schemas

can Override

Input

The aforementioned examples deal w i t h children's narrowing o f the semantic content o f a grammatical category i n the first phases o f its acquisition, apparently making distinctions that are not grammatically marked i n the input language. I n

Slobin

8

other instances, children's definitions o f grammatical categories may be too broad, i n that they ignore or neutralize distinctions that are grammatically marked by mature speakers. The resulting mapping o f form and content reflects basic conceptual organization w h i c h may w e l l be universal. For example, early uses o f grammatical markers o f location, movement, and possession (case inflec­ tions, pre- and postpositions) focus simply on the relation o f a figure to a ground, whether or not the figure is static or moving w i t h respect to the ground, and whether or not the ground is a physical location or an animate being (possessor or recipient o f the figure). For example, i n languages which provide distinct case inflections for locative S T A T E (e.g., ' i n ' , ' o n ' ) and locative G O A L ( ' i n t o ' , ' o n t o ' ) , a typical child error consists i n confusion o f the t w o forms, generally w i t h the stative f o r m used for both functions. Such errors are found i n German ( M i l l s , 1985) i n regard to dative (static) and accusative (directional) casemarking on articles; i n Slavic languages (Gvozdev, 1949; Radulovic, 1975; Smoczyhska, 1985) i n regard to nominal suffixes o f the accusative and other cases; and i n Turkish (Aksu-Koc & Slobin, 1985) i n regard to the locative and dative-direc­ tional suffixes on nouns and deictics. These patterns suggest a conceptual sche­ ma oriented simply to the figure-ground relation, w i t h later development o f

TABLE 1.1 Order o f A c q u i s i t i o n of Locative Expressions in Four Languages and Percentage o f Subjects P r o d u c i n g E a c h a

Scale Point

English

Italian

Serbo-Croatian

Turkish

1 2 3 4

IN ON UNDER BESIDE

90 83 81 74

IN ON UNDER BESIDE

91 88 84 77

ON IN BESIDE UNDER

88 84 82 72

IN ON UNDER BESIDE

90 80 79 79

5 6 7

BETWEEN FRONT/ BACK/

49 30 21

BETWEEN BACK/ FRONT/

57 42 41

BACK/ BETWEEN FRONT/

31 26 19

BACK/ FRONT/ BETWEEN

71 53 50

8 9

BACK FRONT

14 3

BACK FRONT

23 18

BACK FRONT

16 12

BACK FRONT

7 4

Gutman coefficient of reproducibility Number of subjects

0.93

0.89

0.86

0.91

86

74

90

70

°Data come from an elicitation task carried out with subjects between the ages of 2;0 and 4;8. The subscript / on B A C K and FRONT denotes location with regard to reference objects that have an inherent front-back orientation (e.g., cars, houses), while B A C K and FRONT without the subscript denote non-oriented reference objects (e.g., plates, blocks). This table appears as Table 5 in Johnston and Slobin (1979, p. 537).

Introduction

9

orientation to the means i n w h i c h the relationship is temporally manifested (enduring versus coming-into-being). I n addition, we find that children are not particularly concerned to indicate grammatically whether the locative reference point is animate or inanimate. For example, i n languages like Hungarian (MacWhinney, 1985), which distinguish an allative (directed movement) from a dative (beneficiary) case, children often use the dative for both meanings. Bowerman (1981) gives related examples from English, where the data consist o f confusion o f forms that mark a distinction that is not marked i n some other languages: English-speaking children confuse give and put, as i n languages that do not distinguish animate from inanimate goals, saying things like " p u t one to m e " and " g i v e some i n here." Similar inattention to animacy as a locative feature is seen i n expressions o f location and possession. Broadly conceived, possession is a locative state i n which an object is located i n relation to a person i n an enduring or sociallysanctioned manner. I n languages i n w h i c h location and possession are marked by distinct means, we sometimes find that children use a single means o f expression for both notions, as i n German children's use o f the locative preposition zu ' t o ' to indicate possession as w e l l ( M i l l s , 1985). Such examples indicate that a basic conceptual schema can override distinctions that are grammatically marked i n the input language. I n the instances noted above, a schema o f object location is recruited to grammatical use o f case inflections, adpositions, or verbs, overriding distinctions o f static location versus direction, or distinctions o f animate versus inanimate reference points for object location or object transfer. 4. Conceptual Emergence

Development Determines of Grammatical Forms.

Order

of

W i t h i n a domain o f form-function mapping, we typically find orders o f emergence based on conceptual development. For example, when children begin to acquire locative expressions (pre- or postpositions, locative case inflections), they do not acquire the means for expressing all locative relations at once; rather, there is a common order o f development across languages. Johnston and Slobin (1979) found similar sequences o f development i n English, Italian, Serbo-Croa­ tian, and Turkish, as shown i n Table 1.1. The means o f expression are different across these languages, i n v o l v i n g prepositions (English, Italian), prepositions and case inflections (Serbo-Croatian), and postpositions and case inflections (Turkish). The absolute ages o f acquisition o f the various terms also differ between these languages. However, the common order o f development o f major conceptual categories is determined by children's ability to conceptualize spatial relations, m o v i n g from simple topological relations to more complex projective relations (Parisi & A n t i n u c c i , 1970). (Differences i n order o f development i n the middle section o f the table are due to language-specific effects, as discussed below. Tables such as these reveal both common orders o f development on conceptual grounds and the influence o f factors o f linguistic complexity on

10

Slobin

particular points i n the sequence.) Similarly, i n regard to other conceptuallypaced domains, various investigators have found common crosslinguistic orders of development o f question words and connectives based on the meanings o f these terms (e.g. B l o o m et a l . , 1980; Clancy et a l . , 1976). Data such as these support the claim "that there is a fairly atuonomous development o f intentions to express various semantic n o t i o n s " (Slobin, 1973, p. 183). I n addition, on the level o f linguistic S T R U C T U R E , the null hypothesis predicts instances i n w h i c h general language acquisition strategies override F O R M A L characteristics o f the input language.

Formal Pattern Preferences a n d G r a m m a r A l o n g w i t h universal preferences for the semantic content o f grammatical mark­ ers i n early child speech, crosslinguistic study also reveals preferences for the placement o f such markers and for the construction o f morphological paradigms and word-order patterns o f particular types. Some formal pattern preferences are tied to the underlying meanings o f grammatical expressions, while others seem to be relatively content-free. 1. C o n c e p t s a r e C o m b i n e d i n G r a m m a t i c a l M o r p h e m e s According to Semantic Affinities W e have noted that children make use o f a particular set o f concepts as the semantic bases for early grammatical marking crosslinguistically. W e can now make an additional observation about basic semantic categories—namely: they exist i n a sort o f S I M I L A R I T Y SPACE i n w h i c h categories are arrayed according to certain N A T U R A L AFFINITIES o f varying degree. W e can begin to map out this similarity space by examining the acquisition o f grammatical forms for which two or more semantic notions combine or interact i n determining the choice o f form. As an example, let us examine the ways i n w h i c h verb forms can be conditioned by various semantic distinctions. I n the Slavic languages, the verb has different forms for perfective and imperfective aspect. For example, i n Polish one uses the verb otworzyc 'open' to refer to a single, completed act o f opening, and the verb otwierac to refer to ongoing or repeated acts o f opening. Children acquiring such languages have no difficulty i n learning pairs o f perfec­ tive and imperfective verbs, and even create their o w n neologisms for new perfective or imperfective forms before age 3 (Gvozdev, 1949; Smoczyhska, 1985). Notions o f verbal aspect are not only highly accessible to the c h i l d , but they are also so close to the meaning o f the verb itself that children quickly learn to combine verb meaning and verb aspect i n a single f o r m , easily learning separate forms for separate aspects. However, other notions are not so easily combinable w i t h verb meaning. Wherever we encounter verbs w h i c h change

Introduction

11

their form on the basis o f present versus past tense, or affirmative versus nega­ tive, or distinctions o f person, we find that children prefer to use a single form o f the verb for all tenses, or for both affirmative and negative, or for all persons. Tense, negation, and person are apparently not inherently part o f verb meaning for children, as we find i n early errors i n Japanese, Slavic, and Romance lan­ guages. I n Spanish (Clark, 1985), for example, children w i l l use a single verb stem for all persons or tenses, even i f the input language provides systematic stem variations for particular verbs. I n Japanese (Clancy, 1985), where the form of negation is different for past and non-past tense verbs, we find children using a single negative form regardless o f tense. Apparently children operate w i t h builtin preferences for the combinability o f notions i n morphemes. I n these exam­ p l e s — w h i c h could not be determined by the study o f any individual language alone—it seems that children recognize that person does not change the meaning of the verb i n the way that aspect does, and that notions like tense and negation affect the meaining o f the entire clause, and not just the meaning o f the verb. 2. G r a m m a t i c a l M o r p h e m e s a r e P o s i t i o n e d A c c o r d i n g to their Scope o f O p e r a t i o n These examples move us from the morphological to the syntactic level o f early child grammar. Children are sensitive to the SCOPE or R A N G E o f operation o f grammatical elements. For example, they seem to recognize that i f an element operates on the meaning o f the clause, it should, ideally, be placed outside o f the clause, and should not alter the internal form o f the clause. Negation provides a prime example o f this principle. Wherever possible, children w i l l move negative elements, leaving verb forms and w o r d order intact. For example, i n Turkish (Aksu-Kog & Slobin, 1985) and Japanese (Clancy, 1985), where negative parti­ cles are sometimes placed inside the verb, children tend to move these particles to the end o f the clause, f o l l o w i n g the standard verb-final order o f these lan­ guages. Similar examples could be offered from other languages i n regard to the placement o f negation as w e l l as other forms that operate on clauses as a whole, such as markers o f questions and conditional forms. For example, i n both Polish (Smoczyhska, 1985) and Hungarian ( M a c W h i n n e y , 1985) a conditional marker is placed immediately after the verb stem, preceding person-marking. However, children tend to reverse the order, marking person close to the verb stem and placing the conditional marker after markers o f person and tense. A l l o f these examples reflect a general language acquisition strategy to extrapose operators whose scope is the clause: Japanese: * V E R B 4- P A S T + N E G A T I V E (=VERB+NEGATIVE+PAST)

12

Slobin

Turkish: *VERB+TENSE+PERSON NEGATIVE (= V E R B + N E G A T I V E + T E N S E + P E R S O N ) Polish: * VERB+PERSON+CONDITIONAL ( - VERB+CONDITIONAL+PERSON) Hungarian: *VERB +PERSON CONDITIONAL ( - V E R B + CONDITIONAL+PERSON) A g a i n , invididual instances o f such phenomena could be interpreted i n terms o f language-specific features, while the crosslinguistic commonality o f such pat­ terns, across differing morphosyntactic forms, suggests general formal pattern principles.

3. G r a m m a t i c a l M a r k e r s a r e P l a c e d A c c o r d i n g t o Principles of Semantic Relevance The examples we have just considered also suggest another type o f syntactic principle i n regard to the placement o f inflectional morphemes i n relation to the words that they operate on. I n these examples we note preferences to keep grammatical markers o f aspect, tense, and person close to the verb, while keep­ ing negation and conditionality peripheral. M o r e generally, we can speak o f the degree o f "relevance" (Bybee, 1985) o f the meaning o f a grammatical marker to the meaning o f the stem to w h i c h i t is affixed. I n Polish, for example, personmarking is attached to conditional particles and conditional connectives, but children avoid applying a verbal suffix to such non-verbal elements as condi­ tionals and connectives, preferring to mark person and number on the verb. I n Hungarian, the verb receives a suffix indicating the definiteness o f the direct object, and children have great difficulty i n acquiring this form. O n the other hand, where definiteness is indicated by a marker on the nounphrase as i n English and German definite and indefinite articles, or Bulgarian (Gheorgov, 1908) noun suffixes for definiteness, children have little difficulty i n learning to mark these notions grammatically. Apparently the noun is a more " r e l e v a n t " locus o f definiteness than is the verb. Findings such as these suggest that children follow a general principle, across an array o f language-specific instances: Morphemes that go together semantically should be placed together syntactically. T o briefly summarize, many o f the crosslinguistic child language patterns we have examined suggest that c h i l ­ dren operate w i t h a hierarchy o f relevance o f grammatical markers i n relation to the part-of-speech they modify. For example, tense and person are more inher­ ently part o f the meaning o f a verb than are negation and conditionality, w h i c h operate on the meaning o f an entire clause. Accordingly, it is easier for children

Introduction

13

to learn to affix tense and person markers to the verb than to affix negation and conditional markers to the verb, and it is difficult for them to learn to affix verbal notions like tense and person to non-verbal elements. A hierarchy o f relevance also affects the ease w i t h w h i c h children can acquire stem changes o f the verb. Whereas children can easily acquire stem changes for perfective and imperfective aspect, they find i t difficult to vary the form o f the verb for person or tense. This whole array o f data suggests a hierarchy o f relevance o f semantic notions i n relation to the verb, w i t h aspect closest to the inherent meaning o f the verb, tense and person more distant, and negation and modalities such as the conditional the most distant. Particular languages present pieces o f this pattern; comparison across languages reveals the more general pattern.

4. M o r p h o l o g i c a l S y s t e m s a r e C o n s t r u c t e d A c c o r d i n g to Formal (Non-semantic) Criteria Examples such as those summarized earlier show that children are engaged i n mapping particular types o f semantic categories onto lexical items and gram­ matical forms i n specific and limited ways. Such examples show subtle interac­ tions o f semantic and non-semantic criteria i n the construction o f grammar. I t is also evident, however, that many systematic aspects o f early child grammar have no semantic bases at a l l . Children are equipped w i t h capacities to construct morphological paradigms and syntactic rules. T o use Annette Karmiloff-Smith's (1979) felicitous phrase, they approach language as a formal "problem-space" in its o w n right, i n addition to the acquisition o f a tool for communication and thought. Our crosslinguistic data show many attempts to w o r k on the structures o f language per se. For example, it is evident that children w o r k at constructing morphological paradigms—that is, principles for the systematic alteration o f forms o f particular morphemes. M u c h o f the classical child language literature deals w i t h children's attempts to regularize inflectional systems, reflecting universal patterns o f sys¬ tematicity, such as the familiar overregularization o f the English past tense. Such examples show a preference for uniform marking o f a grammatical category i n all instances. There are also preferences to build P A R A D I G M S on the basis o f systematic distinctions. For example, Spanish-speaking children (Clark, 1985) have little trouble i n realizing that nouns are divided into t w o classes on the basis of their endings, although these endings generally have no semantic basis. They make use o f this division i n matching the forms o f associated articles and nouns. I n Spanish, nouns that take the indefinite article un and the definite article el generally end i n -o, and nouns that take the articles una and la generally end i n -a. Children adjust inconsistencies i n the input language, saying things like *un papelo instead o f un papel and * la flora instead o f laflor. I n this way they show their grasp o f a purely formal characteristic o f their language—namely, a binary division o f nouns for grammatical purposes. Crosslinguistic comparison suggests

14

Slobin

that binary divisions may be high on a built-in hierarchy o f formal criteria for choice o f grammatical morphemes. For example, nouns i n Hebrew (Berman, 1985) are pluralized by the addition o f -im to masculine nouns and -ot to femi­ nine nouns. A t first, children use the single form -im to pluralize all nouns, indicating that the semantic notion o f ' p l u r a l ' is basic, taking precedence over formal, nonsemantic variation. However, at the next stage o f development, Israeli children divide their nouns into t w o classes on an idiosyncratic basis. I n the adult language, words that end i n stressed -a and unstressed -et and -at generally take the feminine plural. Little children, however, arrive at a simpler binary split, adding the feminine -ot to all nouns that end i n -a, whether or not the final v o w e l is stressed, retaining the masculine -im for all other nouns. This results i n many errors i n regard to the adult system, w h i c h has more complex means o f distinguishing masculine and feminine nouns. But the children's solu­ tion shows a basic ability to deal w i t h a formal basis for suffixing i f it is based on a highly salient and binary division o f words on a single criterion, such as the final sound o f the w o r d . Sometimes children's regularizations even reflect patterns that are not model­ ed i n the input language, suggesting quite general preferences. For example, Bowerman (1974) and others have noted tendencies to use the sentence frame alone to define the valence o f the verb, without special lexical or inflectional marking o f transitivity or causativity, as i n familiar English examples such as " K e n d a l l fall that t o y " ( = ' d r o p ' ) and " I come it closer" ( = 'bring/make come'). There are examples i n the child language literature o f such constructions in English, French, Portuguese, Polish, Hebrew, Hungarian, and T u r k i s h . What is striking is the fact that children attempt such forms even i n those languages i n which the input does not model this possibility. As another example, consider children's preferences for analytic over synthetic expressions. Where the input language provides both options, the analytic forms are used early on, such as the French possessive de moi i n place o f the synthetic monlmalmes. Note that the child's option allows for separate expression o f possession, person, and number, reflecting a general tendency towards one-to-one mapping. I n instances where the language provides synthetic forms, children often invent their o w n analytic equivalents, such as the early use i n Hebrew o f a separate P R E P O S I T I O N + P R O N O U N construction i n place o f adult fused forms (e.g. al + ani ' o n + V instead o f alay 'on:me'). In sum, the acquisition literature is full o f examples o f children's attempts to simplify or restructure the input language along universal lines o f c h i l d grammar formation, supporting the null hypothesis that language acquisition is every­ where the same, regardless o f particularities o f individual languages. However, language acquisition takes place i n a web o f universals and particulars, and much recent crosslinguistic w o r k seeks to evaluate hypotheses o f the effects o f features of particular languages on the course o f acquisition. W h i l e the longterm goal remains one o f discovering general processes, the course o f language acquisition is not everywhere the same i n its details.

Introduction

15

H Y P O T H E S E S O F SPECIFIC L A N G U A G E EFFECTS 1. F o r m - f u n c t i o n I n t e r a c t i o n I n f l u e n c e s R a t e o r Sequence of Development We have already noted that conceptual development plays a key role i n providing starting points and developmental sequences o f the meanings o f grammatical forms. However, factors o f accessibility o f linguistic forms to the child also play a role i n individual cases. Locative Development. Although the overall sequence o f development o f locative adpositions follows the scheme presented by Johnston and Slobin (1979) and others, relative linguistic difficulty can cause minor changes i n developmen­ tal sequence. Consider Table 1.1 once again. Earlier, we noted that the middle section o f the table reflected crosslinguistic differences i n the order o f develop­ ment o f pre- or postpositions expressing the notions 'between' and 'front/back' w i t h featured objects. The expected order o f development on grounds o f concep­ tual complexity is 'back' < 'front' < 'between' (Johnston & Slobin, 1979). This order matches the sequence o f linguistic development i n Turkish, where the terms do not differ i n linguistic complexity: each locative expression is a single semantically transparent postposition. The English and Italian terms for 'back' and 'front' are sufficiently complex linguistically to retard the emergence o f these terms i n relation to terms for 'between', even though 'between' is a conceptually more advanced notion. I n English we have t w o terms for 'back' rather than one, w i t h the further complexity that one o f the options is mor­ phologically complex: behind and in back of. The Italian expression, dietro a, is also morphologically complex, and does not correspond to the w o r d for the body part 'back'. Apparently formal factors such as these can exert a degree o f influence on aspects o f the emergence o f linguistic forms. Johnston and Slobin conclude (1979, p . 541): "Wherever conceptual complexity fails to predict actual order o f acquisition, we find some pocket o f relative L I N G U I S T I C d i f f i c u l t y . " Crosslinguistic comparisons o f this sort help to define the sorts o f linguistic difficulty that can interact w i t h conceptual factors i n pacing acquisition. Tense/Aspect. A l t h o u g h past-tense marking o f non-punctual events is gen­ erally later than marking o f punctual events, both types o f past tense seem to be early i n Slavic languages. For example, i n Polish, where there is a clear mor­ phological distinction between perfective and imperfective verb forms, there is evidence that very young children use past imperfectives to refer to anterior nonpunctual events, while using past perfectives to refer to completed punctual events (Smoczynska, 1985; Weist et a l . , 1985). Here the presence o f a clear linguistic distinction may accelerate children's ability to mark a tense/ aspect contrast.

16

Slobin

Pragmatic Functions. I n languages w h i c h provide clear grammatical mark­ ing o f pragmatic functions it is possible to find evidence for children's early attention to functions that may not be clearly discernible i n the development o f other languages; and i t is even possible that the existence o f such marking draws children's attention to the relevant functions. For example, Clancy (1985) reports very early aquisition o f perceptually salient, sentence-final particles i n Jap­ anese—appropriately used to express pragmatic functions not as clearly evident in other sorts o f languages: Yo is used when the child is encountering resistance or lack of mutuality, and feels he can impose his information or will on the addressee; ne is used when the child is in rapport with the addressee, agreeing with him or expecting his confirmation or approval. The emotional content of no is less fixed than yo and ne; typically no seems to be fairly neutral in affect, occurring in the ordinary give and take of information shared in the speech context.

2. P a r t i c u l a r L i n g u i s t i c F o r m s a r e R e l a t i v e l y M o r e Accessible, Holding Content Constant About a decade ago (Slobin, 1973) I noted that postpositions and suffixes tend to be acquired earlier than prepositions and prefixes for the expression o f particular locative notions and grammatical cases, suggesting that children pay special attention to the ends o f words. Such comparisons can only be carried out crosslinguistically, since one must try to hold meaning and frequency o f use constant. Thus, for example, the emergence o f the first morphological marking of simple locatives like ' i n ' and ' o n ' is earlier i n postpositional and inflectional languages like Hungarian and Turkish than i n prepositional languages like En­ glish and Serbo-Croatian. Peters (1985) reports additional data on the salience o f postposed over preposed grammatical markers. T o cite another type o f example, it appears that case inflections are acquired earlier than word-order regularities for the expression o f comparable semantic relations such as agent-patient ( A m ­ nion & Slobin, 1979; Slobin, 1973, 1982; Slobin & Bever, 1982). A m m o n and Slobin suggest that it is easier for children to attend to " l o c a l cues" on individual words than to process and store patterned configurations o f words i n clauses. Crosslinguistic comparison thus reveals general language acquisition strategies which have different effects on the course o f acquisition o f particular languages.

3. S e p a r a t e M a r k i n g o f N o t i o n s i n P a r t i c u l a r Languages Reveals a Conceptual S u b s t r a t u m The early acquisition o f pragmatic particles i n Japanese reveals children's ability to attend to the relevant underlying distinctions. Similar evidence is provided wherever a particular language divides up a conceptual area i n more detail than

Introduction

17

others. For example, Japanese has several distinct negative markers, and c h i l ­ dren show only certain types o f confusions i n acquiring these markers (Clancy, 1985). The directions o f confusion reveal an underlying conceptual substratum which could not readily be seen i n a language w i t h less elaborate negative marking. The arrows i n the f o l l o w i n g chart represent the directions o f children's overextensions o f the meaning o f one form to include that o f another: jfiya ('rejection')

\

nai ( ' non-existence' )

The early confusion o f iya ('rejection') and dame ( ' p r o h i b i t i o n ' ) suggest " a n undifferentiated semantic complex [of] rejecting, demanding, commanding, pro­ hibiting, and i n s i s t i n g " (Clancy, 1985), along w i t h independent status for chigau ('denial') and nai ('non-existence'). Note also that 'non-existence' and 'rejec­ t i o n ' can be overextended to expressions o f 'denial', while 'prohibition' is not extended to this function. What emerges is a more subtle map o f a terrain that can be compared w i t h languages that make fewer, more, or different distinctions.

4. L a n g u a g e - s p e c i f i c C o - o c c u r r e n c e s R e v e a l P a t t e r n s of Conceptual Relevance Earlier we noted ways i n w h i c h combinability and placement o f grammatical morphemes reveal general notions o f semantic affinity and relevance. Co-occur­ rence patterns i n particular languages can facilitate or impede acquisition. For example, noun suffixes expressing case notions are easily acquired (e.g., ergative, accusative, dative, and other case notions discussed above). But chil­ dren have varying degrees o f difficulty i n acquiring conflations o f casemarking w i t h other semantic notions. Children easily learn to use different direct-object inflections for whole and partial objects, as i n Russian (Gvozdev, 1949) and Finnish (Toivainen, 1980), but they have difficulty w i t h distinctions o f the accusative on the basis o f animacy or natural gender, as i n Russian and Polish (Smoczyiiska, 1985). Children readily make the verb agree w i t h the object o f resultant activity, even i f not allowed by the language, as i n Italian (Antinucci & M i l l e r , 1976); but agreement o f the verb w i t h the definiteness o f the object is difficult, as i n Hungarian ( M a c W h i n n e y , 1985). I t is easier for children to learn to change the form o f the verb stem for aspect than for tense (Polish), and easier for tense than for person (Spanish, Portuguese). Detailed crosslinguistic com­ parison o f this sort reveals universal patterns o f conceptual relevance i n semantic space, allowing for language-specific predictions o f ease or difficulty o f acquisi­ tion o f particular conflations.

18

Slobin

5. C r o s s l i n g u i s t i c D i f f e r e n c e s i n D e g r e e o f C o d i n g a n A r e a Reveal Linguistic Capacities Finally, the crosslinguistic method is useful i n instances where languages differ in the degree o f elaboration o f paradigms. For example, languages can be con­ trasted i n terms o f the dimensions that play a role i n their verb paradigms. I n our sample we have a range—from English, w i t h m i n i m a l marking o f personnumber, through Portuguese, w i t h extensive marking o f person-number, to Hebrew, w i t h extensive marking o f person-number-gender. W e need this full range i n order to better understand the means available to children for construct­ ing paradigms. Using data from a language like Portuguese, Bybee was able to discover generalization o f person-marking across tenses and the use o f 3SG as the base for generalizations w i t h i n tenses (Bybee Hooper, 1979; Simoes & StoelGammon, 1979), w h i l e such patterns could simply not exist i n a language like English. Hebrew developmental data were needed to show precocious acqusition of sex-marking o f verbs i n first and third persons, w i t h difficulty i n second person (Berman, 1985; L e v y , 1983), since gender is not marked on the verb i n most other languages i n our sample. Similar comparisons could be made w i t h respect to crosslinguistic differences i n the degree o f elaboration o f other sys­ tems, such as casemarking, temporal and locative expressions, and gender and noun classifiers.

THE P L A N OF THE B O O K B y the late seventies, linguists and psycholinguists i n many countries were convinced o f the need for gathering data on the acquisition o f different types o f languages and using the crosslinguistic method for purposes such as those sum­ marized above. I t was clearly time to draw together available data as a resource for constructing more detailed and universally valid theoretical accounts o f child language development. I n November 1979 the authors o f this volume were invited to prepare interpretive summaries o f the course o f acquisition o f various languages. W e were limited i n our choice o f languages on the basis o f the richness o f available data—but l u c k i l y a fair range o f diversity is represented i n our sample. The authors were asked to contribute to " a book i n w h i c h findings on the acquisition o f different types o f languages w i l l be examined i n terms o f their contribution to language acquisition theory generally." Thus the chapters are intended to be selective reviews, rather than exhaustive summaries o f the acquisition o f each language. Each author was asked to approach his or her partucular language "as a case study i n a potential crosslinguistic typology o f acquisitional p r o b l e m s , " considering those data w h i c h "contribute to an issue o f general theoretical concern i n developmental psycholinguistics." W i t h these ends i n m i n d , a common framework was suggested, as reflected i n the f o l l o w i n g

Introduction

19

outline and guiding questions (which we hope w i l l be used for future acquisition summaries as w e l l ) : Introductory

Materials

1. Brief grammatical sketch of the language or language group, presenting those linguistic facts which are relevant to the following developmental analysis. 2. Summary of basic sources of evidence, characterizing methods of gathering data, and listing key references. (Note that the sources of evidence vary widely, ranging from ethnographic and longitudinal study of several children in the early years, as in Kaluli, to longitudinal and experimental studies of many children across a large age range, as in English.) 3. Brief summary of the overall course of linguistic development in the lan­ guage or language group. (The reader interested in obtaining a rapid crosslinguistic survey of the materials covered in the book is advised to read sections 1 and 3 of each chapter.) Language

Acquisition

Data

Sections 4-6 present detailed summaries of the acquisition of linguistic forms, orienting to issues of ease and difficulty of acquisition of various systems of the language. 4. Typical errors: What sorts of errors typically occur? How can they be ac­ counted for? What are the most general problems posed to the child in acquiring a language of this type? 5. Error-free acquisition: What systems are learned relatively free of error? How can this be accounted for? In what ways is a language of this sort well-suited to the child's language-acquisition capacities? 6. Timing of acquisition: What systems are acquired strikingly early or strik­ ingly late in comparison with general crosslinguistic or English-based expecta­ tions? What constitutes linguistic complexity for the learner? Data on the Setting

of Language

Acquisition

Sections 7-10 situate language acquisition in regard to issues of cognitive and social development, exploring interactions between linguistic form and conceptual representation, interpersonal communication, and individual patterns of acquisition. 7. Cognitive pacesetting of language development: What evidence is there for the influence of cognitive development on problems of acquiring linguistic forms? (This issue is the special topic of Johnston's chapter in Vol. 2.) 8. Linguistic pacesetting of cognitive development: What evidence is there for the role of language development in guiding or shaping the course of cognitive or conceptual development? 9. Input and adult-child interaction: In what ways do patterns of communica­ tion and social interaction influence the course of linguistic development?

20

Slobin

10. Individual differences: What evidence is there for distinct patterns of ac­ quisition within a language, based on individual characteristics of the type of learner? Are there data on individual differences that cast light on developmental processes? Theoretical

Conclusions

11. Reorganizations in development: At what points in development is there evidence for significant linguistic reorganization on the part of the child? What do reorganizations suggest about underlying processes of acquisition? 12. Theoretical implications: What general principles of language development are suggested by study of a language of this type? (The chapters, in Vol. 2, by Bowerman, MacWhinney, Peters, and Slobin seek general theoretical implications across the array of languages surveyed here.) 13. Suggestions for further study: What issues could be illuminated by further study of languages of this type, or in explicit compairson with other types of languages?

Preliminary versions o f chapters on individual languages were studied i n a seminar on crosslinguistic approaches to language acquisition at the 1980 Sum­ mer Linguistic Institute at the University o f N e w M e x i c o , followed by a con­ ference at Berkeley i n November 1980, supported by the Sloan Foundation Program i n Cognitive Science. T a l m y G i v o n and Melissa Bowerman were invit­ ed to be discussants at that conference, examining the data from the perspectives of linguistics and developmental psycholinguistics. Judith Johnston explored the issue o f cognitive prerequisites for language acquisition; and workshops were held on individual differences (Elena Lieven) and cultural determinants (Susan Ervin-Tripp). Some o f the participants took part i n an additional workshop i n the summer o f 1981 at the Max-Planck-Institut fiir Psycholinguistik i n Nijmegen, The Netherlands ( A k s u - K o g , Berman, Bowerman, Clark, E r v i n - T r i p p , L i e v e n , MacWhinney, Slobin, Smoczynska). The National Science Foundation sup­ ported Slobin i n systematic summary and interpretation o f the body o f crosslinguistic data ( 1 9 8 0 - 8 3 , Grant B N S 80-09340). Chapters were revised and re-written i n 1 9 8 2 - 8 3 , w i t h a fair degree o f communication between authors. Although some writers have chosen alternative means o f organization, all o f the chapters address the basic issues i n the above outline (to the extent that relevant data are available for the given language). What emerged - and what we offer to our readers - is a compendium o f detailed surveys o f the acquisition o f a number o f different types o f languages, drawn from a fair sampling o f language groups: (1) I N D O - E U R O P E A N : (a) G E R ­ M A N I C : English, German; (b) R O M A N C E : French, Italian, Portuguese, Ruma­ nian, Spanish; (c) S L A V I C : Polish; (2) S E M I T I C : Hebrew; (3) F I N N O - U G R I C : H u n ­ garian; (4) U R A L - A L T A I C : Turkish; (5) J A P A N E S E - R Y U K Y U A N : Japanese; (6)

Introduction T R A N S - N E W G U I N E A NON-AUSTRONESIAN: Kaluli; (8) S I G N L A N G U A G E :

ASL.

21

(7) P O L Y N E S I A N : Samoan;

5

Given the states o f the art i n both developmental psycholinguistics and l i n ­ guistics, our data are limited i n particular ways. I n almost all instances, we have data on production rather than comprehension—simply because i t is easier to taperecord speech than to construct and evaluate measures o f comprehension. Furthermore, much o f the speech has been recorded i n semi-controlled situa­ tions, or only i n a small sample o f situations i n which speech naturally occurs. Ethnographic data, such as those o f Schieffelin i n New Guinea and Ochs i n Samoa, show how much more can be learned from a detailed study o f languagein-use; and experimental data, such as those reported by de Villiers and de Villers for English, how what can be learned from detailed and systematic probes of children's competence. The analyses are heavily weighted towards issues o f the acquisition o f morphology and word-order patterns—issues that can be dealt w i t h fairly easily i n terms o f traditional grammar. Studies o f English acquisition show what can be achieved by investigators who are equipped w i t h more subtle and complex theoretical models o f the language i n question, but most o f the languages o f our collection have not yet been submitted to the sort o f detailed syntactic study w h i c h w o u l d allow for acquisition data that w o u l d be relevant to issues o f universal grammar. W e are only beginning to glimpse the outlines o f general syntactic theories that w o u l d lead to interesting crosslinguistic develop­ mental questions. We are also forced to rely heavily on error data, though all investigators raise caveats about the status o f such data. What is o f interest, o f course, is not that the child was " w r o n g , " but that the way i n w h i c h he or she was w r o n g may illuminate some underlying attempt to structure language i n a particular direc­ tion. Thus we are interested i n the ways i n w h i c h child speech regularly, fre­ quently, and systematically deviates from the speech which the child hears. Error data thus must be evaluated i n the light o f the input, and often we lack sufficient input data to fully evaluate the source o f children's " e r r o r s . " The authors also pay close attention to the T I M I N G o f development: early errors mean something different than late errors; errors occurring together at a point i n development mean something different than isolated errors; errors that reorganize previous systems are different from errors at the onset o f acquisition o f a system; and so forth. Wherever possible, we also try to find elusive but highly significant " n o n occurring errors"—that is, places i n the grammar where our theories lead us to expect children to restructure the system, but where they quickly master adult

The individual chapters, all appearing in this series, were written by the following authors: E N G L I S H : Jill and Peter de Villiers, G E R M A N : Anne E . Mills, R O M A N C E L A N G U A G E S : Eve V . Clark, POLISH: Magdalena Smoczyriska, with assistance of Richard M. Weist, H E B R E W : Ruth A. Berman, HUNGARIAN: Brian MacWhinney, T U R K I S H : Ayhan A. Aksu-Koc, and Dan I . Slobin, JAPANESE: Patricia M. Clancy, K A L U L I : Bambi B. Schieffelin, SAMOAN: Elinor Ochs, ASL: Elissa L . Newport and Richard P. Meier. 5

22

Slobin

forms without error. Error-free acquisition, and developmental sequences i n acquisition, thus supplement inferences drawn from errors. This range o f data provides quite a rich base for crosslinguistic theorizing. I n the background, however, we must always be aware o f the fact that we have not only studied a limited sample o f languages, but also a very small number of children, and that considerable individual differences exist between children acquiring a given language. The authors are w e l l aware o f the many limitations o f their material, yet they are able to lay out broad patterns o f acquisition o f their particular languages, suggesting general developmental principles i n relation to other languages. I t is our hope that the ways i n w h i c h we have presented our materials and our ideas w i l l stimulate further crosslinguistic study along similar lines, contributing to a psychological and developmental explanation for the phenomenon already noted by Roger Bacon i n the thirteenth century: " G r a m m a r is substantially the same i n all languages, even though i t may vary accidentally" (Lyons, 1968, p. 15).

ACKNOWLEDGMENTS It is a pleasant task to acknowledge the many friends and colleagues who contributed to the project published here. The first—and continuing—encouragement came from my friend and publisher, Larry Erlbaum, who stood by patiently and supportingly as so many writers tried to do so much, taking more more time than we had hoped. All of the authors were of immeasurable help to me and to each other, in meeting together, corresponding, evaluating and re-evaluating their own work and the work of the others. This was truly a collegial endeavor. Laurie Wagner kept the correspondence and the production and cir­ culation of manuscripts alive; while working space, computer facilities, and financial aid were provided by the Institute of Human Learning and Department of Psychology of the University of California at Berkeley, the Max-Planck-Institut fur Psycholinguistik in Nijmegen, and the Summer Linguistic Institute at the Department of Linguistics of the University of New Mexico in Albuquerque, with grant support from the National Science Foundation (Grant BNS 80-09340), the Sloan Foundation Program in Cognitive Science, and the Committee on Research of the University of California at Berkeley. My children, Heida and Shem, and my wife, Ay§egiil, graciously accepted the many long hours and days when I was lost in my study and seemingly fused with my terminal. They will no doubt be pleased when "the book" comes to mean an object on a shelf once again. —Dan Isaac Slobin Berkeley REFERENCES Aksu, A. A. Aspect and modality in the child's acquisition of the Turkish past tense. Unpublished doctoral dissertation, University of California, Berkeley. 1978. Aksu-Kog, A. A . , & Slobin, D. I. The acquisition of Turkish. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Ammon, M. S., & Slobin, D. I. A cross-linguistic study of the processing of causative sentences. Cognition, 1979, 7, 3-17.

Introduction

23

Antinucci, F . , & Miller, R. How children talk about what happened. Journal of Child Language, 1976, 3, 167-189. Argoff, H. D. The acquisition of Finnish inflectional morphology. Unpublished doctoral dissertation, University of California, Berkeley, 1976. Badry, F . Acquisition of lexical derivation rules in Moroccan Arabic: Implications for the development of Standard Arabic as a second language through literacy. Unpublished doctoral dissertation, University of California, Berkeley, 1983. Berman, R. The acquisition of Hebrew. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Bloom, L . , Lahey, M . , Hood, L . , Lifter, K . , & Fiess, K . Complex sentences: Acquisition of syntactic connections and the semantic relations they encode. Journal of Child Language, 1980, 7, 235-261. Blount, B . G . Acquisition of language by Luo children. Unpublished doctoral dissertation, University of California, Berkeley, 1969. Bowerman, M. F . Early syntactic development: A cross-linguistic study with special reference to Finnish. Cambridge: Cambridge University Press, 1973. Bowerman, M. Learning the structure of causative verbs: A study in the relationship of cognitive, semantic and syntactic development. Papers and Reports on Child Language Development (Department of Linguistics, Stanford University) 1974, 8, 142-178. Bowerman, M. Beyond communicative adequacy: From piecemeal knowledge to an integrated system in the child's acquisition of language. Papers and Reports on Child Language Development (Department of Linguistics, Stanford University), 1981, 20, 1-24. Bybee, J. L . Morphology: A study of the relation between meaning and form. Amsterdam: John Benjamins, 1985. Clancy, P. M. The acquisition of narrative discourse: A study in Japanese. Unpublished doctoral dissertation, University of California, Berkeley, 1980. Clancy, P. The acquisition of Japanese. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 1), Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Clancy, P., Jacobsen, T . , & Silva, M. The acquisition of conjunction: A cross-linguistic study. Papers and Reports on Child Language Development (Department of Linguistics, Stanford University), 1976, 12, 71-80. Clark, E . The acquisition of Romance, with special reference to French. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Eisenberg, A. R. Language acquisition in cultural perspective: Talk in three Mexicano homes. Unpublished doctoral dissertation, University of California, Berkeley, 1982. Erbaugh, M. S. Coming to order: Natural selection and the origin of syntax in the Mandarin speaking child. Unpublished doctoral dissertation, University of California, Berkeley, 1982. Gheorgov, I. A. Ein Beitrag zur grammatischen Entwicklung der Kindersprache. Leipzig: En¬ gelmann, 1908. [Also in Archiv fiir die gesamte Psychologie, 1908, 11, 242-432.] Gvozdev, A. N. Znaèenie izucenija detskogo jazyka dlja jazykovedenija. Rodnoj jazyk i literatura v trudovoj skole, 1928, No. 3, 4, 5. [Reprinted in A. N. Gvozdev, Voprosy izucenija detskoj reci; Moscow: Izd-vo Akademii Pedagogiceskix Nauk RSFSR, 1961.] Gvozdev, A. N. Formirovanie u rebenka grammaticeskogo stroja russkogo jazyka. Moscow: Izd-vo Akademii Pedagogiceskix Nauk RSFSR, 1949. Hopper, P. J . , & Thompson, S. Transitivity. Language, 1980, 56, 251-299. Johnston, J. R., & Slobin, D. I. The development of locative expressions in English, Italian, SerboCroatian and Turkish. Journal of Child Language, 1979, 6, 529-545. Karmiloff-Smith, A. A functional approach to child language. Cambridge: Cambridge University Press, 1979. Kernan, K . T. The acquisition of language by Samoan children. Unpublished doctoral dissertation, University of California, Berkeley, 1969. Levy, Y . It's frogs all the way down. Cognition, 1983, 75, 75-93.

24

Slobin

Lyons, J . Introduction to theoretical linguistics. Cambridge: Cambridge University Press, 1968. MacWhinney, B. How Hungarian children learn to speak. Unpublished doctoral dissertation, Uni­ versity of California, Berkeley, 1973. McCarthy, D. Language development. In W. S. Monroe (Ed.), Encyclopedia of Educational Re­ search. American Educational Research Association, 1950. [Reprinted in A. Bar-Adon & W. F . Leopold (Eds.), Child language: A book of readings. Englewood Cliffs, NJ: Prentice-Hall, 1971.] MacWhinney, B. Hungarian language acquisition as an exemplification of a general model of grammatical development. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisi­ tion (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Mills, A. E . The acquisition of German. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Mitchell-Kernan, C . Language behavior in a black urban community. Unpublished doctoral disser­ tation, University of California, Berkeley, 1969. Parisi, D . , & Antinucci, F . Lexical competence. In G. B. Flores d'Arcais & W. J . M. Levelt (Eds.), Advances in psycholinguistics. Amsterdam: North-Holland, 1970. Peters, A. M. Language segmentation: Operating principles for the perception and analysis of language. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Pye, C . L . The acquisition of grammatical morphemes in Quiche Mayan. Unpublished doctoral dissertation, University of Pittsburgh, 1979. Radulovic, L . Acquisition of language: Studies of Dubrovnik children. Unpublished doctoral disser­ tation, University of California, Berkeley, 1975. Savasir, I . , & Gee, J. The functional equivalents of the middle voice in child language. Proceedings of the Berkeley Linguistic Society, 1982. Schieffelin, B. B. The acquisition of Kaluli. In D. I . Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Simöes, M. C . P., & Stoel-Gammon, C . The acquisition of inflections in Portuguese: A study of the development of person markers on verbs. Journal of Child Language, 1979, 6, 53-67. Slobin, D. I. (Ed.) A field manual for cross-cultural study of the acquisition of communicative competence. Berkeley, CA: A S U C Bookstore, 1967. Slobin, D. I. Cognitive prerequisites for the development of grammar. In C . A. Ferguson & D. I. Slobin (Eds.), Studies of child language development. New York: Holt, Rinehart & Winston, 1973. Slobin, D. I. The origins of grammatical encoding of events. In W. Deutsch (Ed.), The child's construction of grammar. London: Academic Press, 1981. Slobin, D . I . Universal and particular in the acquisition of language. In E . Wanner & L . R. Gleitman (Eds.), Language acquisition: The state of the art. Cambridge: Cambridge University Press, 1982. Slobin, D . I . Crosslinguistic evidence for the language-making capacity. In D. I . Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum Associ­ ates, 1985. Slobin, D. I . , & Bever, T. G . Children use canonical sentence Schemas: A crosslinguistic study of word order and inflections. Cognition, 1982, 12, 229-265. Smoczynska, M. The acquisition of Polish. In D. I . Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Stern, C , & Stern, W. Die Kindersprache: Eine psychologische und sprachtheoretische Unter­ suchung. Leipzig: Barth, 1907. Stross, B. Language acquisition by Tenejapa Tzeltal children. Unpublished doctoral dissertation, University of California, Berkeley, 1969. Toivainen, J . Inflectional affixes used by Finnish-speaking children aged 1-3 years. Helsinki: Suomalaisen Kirjallisuuden Seura, 1980. Weist, R. M . , Wysocka, H . , Witkowska-Stadnik, K . , Buczowska, E . , & Konieczna, E . The defec­ tive tense hypothesis: On the emergence of tense and aspect in child Polish. Journal of Child Language, 1984, 11, 347-374.

SPOKEN LANGUAGES

The Acquisition of English

Jill G. de Villiers Peter A. de Villiers Smith

College

Contents Introduction 27 Descriptive Sketch of English 27 Sources of Evidence 36 Overall Course of Development 37 Data 38 Word Order and Basic Relations in Stage 1 Grammatical Morphemes 66 Modalities of the Simple Sentence 77 Dative and Passive 92 Coordination 103 Relative Clauses 111 Verb Complements 118 Conclusion 124

38

INTRODUCTION 1. D e s c r i p t i v e S k e t c h o f E n g l i s h The English language has attracted more attention than virtually any other, and has been described i n extensive detail. The wealth o f linguistic analysis presents the writer o f a thumbnail sketch w i t h something o f a problem. I t is difficult to present facts about English from an atheoretical perspective, especially some o f its more complex aspects. W e have selected to describe the major forms o f sentence construction from the barest simple sentence to complementation, focusing upon the processes that have been most heavily researched i n child language. I n this selection process we owe a substantial debt to B r o w n ' s (1973) readable survey o f English i n the introduction to his book A First Language. The seven sections on English also constitute the major subheadings o f the material 27

De Villiers a n d De Villiers

28

on child language to f o l l o w . W e have chosen to focus not at all on phonology, derivational morphology, or w o r d meanings. Each area is rich and fascinating i n its o w n right, but syntax is o f central importance i n linguistics and as a defining quality o f human language. 1.1. Word

Order

The basic w o r d order i n English is S V O , and i n contrast to some languages, word order is affected little by pragmatic factors. Stress is a preferred device for signaling changes i n focus w h i l e preserving the S V O structure. However there are structural devices such as the passive w h i c h allow the logical object to be the focus, ^e.g. John was chased by a wasp. and constructions such as the cleft and pseudocleft allow various constituents to be highlighted, e.g. What he wanted was a drink, or: It was Sally that John liked. These constructions are rare even i n adult speech, however. Left or right dislocation o f an N P is another device used for pragmatic pur­ poses, but i t is also relatively uncommon except i n certain dialects, e.g. That man I would like to meet, or: She loves oranges, my mother. Imperatives can also follow the verb w i t h a subject specification after a slight pause, and hence violate the S V O order, e.g. Move over, son. or: Help me, Bill. Articles invariably precede the noun, and most often other determiners and adjectives are i n prenominal position. However they can be placed i n predicate position w i t h the copula be, e.g. The book is red. The house is ours.

1.

The A c q u i s i t i o n o f English

29

when they constitute the comment rather than a specification. The relation o f possession can be marked b y an inflection on the possessor, i n which case the order is possessor-'s + possessed, or by the preposition of, i n which case the order is possessed-o/-possessor. Similarly there is some freedom o f order w i t h locative expressions, one can say both: There is the book. The book is there. w i t h little change o f meaning. As far as language development goes, one could therefore expect the input to be heavily biased towards canonical S V O structures, though from the child's perspective there might appear to be variation i n adjective and locative order. 1.2. Noun

and Verb

Inflections

Compared to other languages, English has a relatively impoverished mor­ phology. I n terms o f noun markings, there exist inflections for plurality (one versus many) and possession. These are phonologically governed but have no semantic variations. Articles are usually required before common nouns. They mark the dimension definite/indefinite {the versus a/an) but are not affected by gender or other properties o f the object, w i t h the exception o f the mass/count and proper/common noun distinctions. The rules governing these variations have not been completely worked out, sometimes to the exasperation o f foreigners. For instance, mass nouns cannot take a, but can take the plural form some, even though they are not marked by a plural inflection. So one has: I would like some soup, or: I would like soup. However i f i t is a definite referent, the is used: I like the soup. The mass /count distinction is not easily identified w i t h a semantic distinction between substance and thing, but is largely defined by its distributional priv­ ileges w i t h articles and quantifiers. The proper/common noun distinction is semantically defined, but some proper nouns must take articles c.f. the Eiffel Tower; the King Dome. Finally, even the distinction definite/indefinite referent is enormously complex and dependent upon the listener's knowledge as w e l l as the speaker's knowledge (Maratsos, 1975). Verbs i n the present tense require only one inflection,-s, for the third person singular. There are only a limited number o f exceptions, e.g. has and does. The

De Villiers and De Villiers

30

present progressive tense is used for ongoing actions, and has t w o required markings; the inflection -ing on the verb and the auxiliary be before i t , e.g. He is going They are talking. The form o f be varies w i t h person, number, and tense but not gender. The plural forms are not distinguished by person. The past progressive is the same form w i t h the past tenses o f be. There is no irregularity i n the progressive f o r m . The past tense is the third major verb inflection, and exceptions are the rule for common verbs i n English. The regular past tense -ed is phonologically conditioned by the final consonant o f the root w i t h the allomorphs ft/, id/, and / i d / but it is insensitive to number or person. However, irregular past tenses abound and take a variety o f forms e.g.

but

fall-fell bring-brought sing-sang and string-strung, tell-told, stand-stood, eat-ate

Virtually all o f these must be learned by rote. The perfect form o f the verb, used to mark current relevance, consists o f the modal have plus the past participle inflection on the verb, w h i c h can be either -ed, -en, or some irregular f o r m , e.g. I have eaten lunch. I have spilled it. I have sung in a choir. Have has one variant, the third person singular has. The past tense form had creates the past perfect form. There are virtually no verbs (but see prove) w h i c h are perfectly regular i n English i n taking present -s, past -ed, and perfect -en together. Hence for an English language learner, person and number distinctions are rarely marked on verbs, but are needed nonetheless for the common verb be. Gender is irrelevant, and verbs do not incorporate other constituents such as pronominal objects, indirect objects, directions, or locations w i t h i n their bound­ aries. However, several linguists have proposed that such elements as C A U S E are incorporated i n some verbs without overt marking, c.f. The door opened. John opened (cause to open) the door. (e.g. Jackendoff, 1972). A n account o f the morphological prefixes such as re-, mis-, dis- and so on w o u l d take us beyond the bounds o f our present task.

1.

The A c q u i s i t i o n o f English

31

Finally, verbs can be modified by an extensive modal auxiliary system i n English, w h i c h uses preposed modals to indicate possibility/ability, e.g. I can swim, hypothetical, e.g. I may go. obligatory, e.g. I must leave, future tense, e.g. I will leave, conditional tense, e.g. I would go if you came too. The modal auxiliaries can combine w i t h be and have to make more complex expressions, e.g. I might have been eating supper when you called. The phrase structure rule for expansion o f the auxiliary component is complex, and can be written as: Aux-H»tense (modal) (have) (be) Some linguistic treatments control the affixes required f o l l o w i n g each compo­ nent by coding them w i t h that f o r m and allowing a transformation called affixhopping to move them to the appropriate verbal element, e.g. Aux—>tense (modal) (have + en) (be 4- ing) (Akmajian & H e n r y , 1975). As one final complexity: i n cases where there is no auxiliary i n the present form, a

4 4

d u m m y " f o r m do appears i n the corresponding emphatic, negative, or

question f o r m , e.g. He likes me. He does like me!

De Villiers and De Villiers

32

He doesn't like me. Does he like me? Obviously, the auxiliary component i n English is rich and complex. 1.3.

Modalities

of the Simple

Sentence

The declarative sentence form described above can be mapped onto other sentence modalities: yes-no questions, constituent (wh) questions, imperatives, negatives, and tag questions. The auxiliary component plays an essential role i n these modalities. In yes-no question forms, the distinctive feature is an inversion o f the subject and the first auxiliary component f o l l o w i n g tense, e.g. You can come. Can you come? She must have been doodling. Must she have been doodling? When tense is the only component, do is introduced, e.g. He ate it. Did he eat it? This aux-inversion occurs also i n w h - or constituent questions, such as When can you come? What did he eat? However it does not occur i n embedded wh-forms, e.g. I asked you when you can come. I told him where I must go. but these do not have interrogative force. Wh-questions have the questioned constituent preposed to the front, though there is a less common variant i n w h i c h it is not preposed, e.g. You want to go where? For negatives, generally the first auxiliary component after tense is marked w i t h n't, or not is inserted after i t , e.g. You can't come. She mustn't have been doodling.

1.

The A c q u i s i t i o n o f English

33

A g a i n , i f tense is the only component, do is inserted: He didn't eat it. However unlike questions, there is some optionality about where i n the auxiliary the negative element can go. I t is permissible to say: She must have not been doodling. or: He might have been not listening on purpose. but the placement makes subtle differences i n meaning that reflect the different scope o f the negative element. The imperative form i n English has no subject or auxiliary on the surface, though reflexive forms such as Wash yourself! and tag questions, e.g. Help me, won't you! have suggested an " u n d e r l y i n g " structure contains those elements (e.g. Jacobs & Rosenbaum, 1968). Tag questions are complex i n English and require a series o f rules that are governed by the preceding declarative: pronominalization o f the subject, trunca­ tion o f the predicate to the first auxiliary, aux-subject inversion as i n questions, and negation/affirmation, e.g. John might be coming, mightn't he? versus: John shouldn't have gone, should he? Simple variants such as " r i g h t ? " " O . K ? " or " H e y ? " also exist i n more infor­ mal speech. 1.4.

Dative

and

Passive

Before turning attention to complex sentences, consider for a moment t w o other sentence forms i n English that involve some variations i n the canonical SVO order. The passive form has been widely discussed i n linguistics and has served as a cornerstone o f many transformational accounts (e.g. Chomsky, 1957). I n the passive, the usual order o f logical subject and object is reversed, and the logical subject (or agent) is marked w i t h by. I n addition, the f o r m o f the

De Villiers and De Villiers

34

verb changes: the auxiliary be or get is required, and the past participle o f the verb, e.g. John got beaten by Bill in the race. Controversy exists over whether the passive should derive by transformation from the active form or whether passive verbs should have their o w n listing i n the lexicon w i t h a redundancy rule mapping them to the active verbs (Bresnan, 1978), and the agent by-phrase generated i n the base as a prepositional phrase. Regardless o f the linguistic treatment, passive sentences stand as a major excep­ tion to the canonical sentence order. Datives also involve a prepositional phrase, i n this case marking the indirect object as distinct from the direct object: Harry sent the book to his teacher. The optional form o f the dative omits the marking, and re-orders the direct and indirect object: Harry sent his teacher the book. A g a i n , whether the t w o forms are transformationally related or should be treated as distinct lexical rules (Roeper et a l . , 1981) is a matter o f dispute. The first form is normally considered basic, and the second version " t r a n s f o r m e d . " The trans­ formed version not only constitutes a violation o f S V O order, but may be block­ ed for some verbs: * Harry transmitted his teacher the message. and w i t h pronominal objects (in most English dialects): * Harry sent his teacher it.

1.5.

Coordination

Joining t w o propositions together w i t h and is perhaps the simplest way to generate a complex sentence i n English. Redundant elements o f the t w o proposi­ tions need not be specified, so some linguistic treatments consider the phrasal coordination to be derived from the full sentential f o r m by a deletion schema (Harries, 1973). The rules include deletion i n a forward direction (null site i n the right conjunct), e.g. Birds like strawberries and ( • ) hate cucumbers.

1.

T h e A c q u i s i t i o n of English

35

and i n a backward direction (null site i n the left conjunct), e.g. Birds ( • ) and children like strawberries. The direction o f deletion depends on the constituents that are redundant. Further details o f English coordination are provided later. 1.6.

Relative

Clauses

Relative clauses are one o f the major forms o f sentence embedding i n English, and may modify noun phrases playing any role i n the basic sentence. The sentence can be embedded i n a variety o f places and the head noun may also play a wide variety o f roles w i t h i n the embedded sentence (the focus). Keenan and Comrie (1977) surveyed over 40 different languages and concluded that there exists a universal hierarchy o f " a c c e s s i b i l i t y " o f noun phrases. For example, some languages allow the head noun to be the subject or the object o f the embedded clause, but not other types o f focus. The proposed hierarchy is: subject < direct object < indirect object < object o f a preposition < possessive noun phrase < object o f a comparative particle. English is very tolerant relative to other languages, even permitting the head noun to be the object o f a com­ parative particle, e.g. I met the kid that Paul is smarter than. The complementizer w i t h relative clauses may be a w h - w o r d , or that, but i t is commonly deleted: I saw the man you met. I know the reason you didn't go. The head noun is not redundantly marked, though i t was i n seventeenth century English, e.g. I saw the man which man you met. The absence o f clear marking o f the embedded S might be expected to cause children some problems i n acquisition, and it does cause adults processing d i f f i ­ culties i n experimental tasks (Fodor, Bever, & Garrett, 1974). The configural properties o f relative clauses i n English w i l l be described i n more detail on page 112. 1.7.

Complements

Embedded sentences o f a variety o f types occur as elaborations o f the V P , and the system o f sentential complements i n English is extremely rich. For example, one can have to infinitives:

36

De Villiers and De Villiers

John likes to swim in rivers. Jane promised Sally to come home. or " b a r e " verbs, e.g. Andrew saw Bill drop his coffee, gerundive complements: Carol likes Sam watching T . V . and tensed clauses: Mary knew Sally picked the rose. The complementizer may be absent as i n the above, or take the form for before a to-infinitive, e.g. Harry likes for Sally to rub his back. that before a tensed clause, or possessive 's marking the subject o f a gerundive complement: Alice enjoyed Bill's playing the national anthem. To add to the complexity, different matrix verbs allow different forms o f comple­ mentation, so that Bresnan (1978) has proposed that the possibilities must be lexically encoded. For instance, only perception verbs take the bare form o f complement verb. One cannot have: Andrew liked Bill drop his coffee. The potential for error and overgeneralization w o u l d be everywhere i n the En­ glish complement system i f the rules were not coded w i t h the particular verbs.

2. S o u r c e s o f E v i d e n c e Children learning English as a first language have been extensively studied i n natural environments. Several studies have observed small groups o f children longitudinally for a number o f years (e.g. B l o o m , 1970; Bowerman, 1974; Braine, 1963; B r o w n , 1973; B r o w n & B e l l u g i , 1964; Leonard, 1976; L i m b e r , 1973; M i l l e r & E r v i n , 1964; Wells, 1979). The data come chiefly from A m e r i -

1.

The A c q u i s i t i o n of English

37

can English, w i t h some material from British and Australian English. The focus of most o f the studies was on early stages o f grammar. M o r e complex grammar has been studied primarily using experimental tasks to test groups o f preschool or gradeschool children. The data base on the acquisition o f English is thus rich and varied, and is growing at a rapid rate.

3. O v e r a l l C o u r s e o f D e v e l o p m e n t In the material that follows, the acquisition o f English syntax is explored i n the order that it was discussed for adult English, since that roughly parallels the course o f development. W e begin w i t h the child at Stage I , when his utterances are still less than 2.0 morphemes long ( B r o w n , 1973). The focus o f interest there is the rules for ordering basic relations i n a sentence, and the research on that question has been extensive. The next accomplishment for the child is to begin adding the grammatical elements—articles, tenses, plurals, prepositions—that modulate the meaning o f the content words. This process begins at the t w o w o r d stage but is not completed for several years. There is considerable orderliness i n the way these morphemes are added to the child's speech, and an account o f it occupies the next section. A d d i t i o n o f auxiliary verbs also begins at the t w o or three word stage, and the child begins to use auxiliary verbs to signal questions and negatives, using a limited set at first. Interest here lies i n how the child's rules are being formulated, and what kinds o f errors occur. There is little or no evidence o f transformed datives and passives i n the speech of 3- and 4-year-olds, yet they are beginning to understand these deviations from the canonical w o r d order i n English, especially when semantic and contextual clues are available. The fourth section considers the literature on these forms i n acquisition. The acquisition o f the rules for j o i n i n g sentences together is the next focus o f attention, since coordination o f sentences seems to occur before sentence embeddings, although the full syntax and semantics o f conjunction is not complete at age 4 years. Relative clauses are discussed i n the section following coordination. Though the rudiments o f embedding may be found i n the speech o f 2-year-olds, most researchers agree that the child's knowledge o f relative clauses is not complete until much later. Finally, verb complements are discussed as another example o f embedding. A g a i n , the earliest complements are found i n 2-year-olds, but it is many years before children understand the full grammar o f the English complement system in all its complexity. For each area o f acquisition, we have attempted to answer a similar set o f questions. These are:

38

De Villiers and De Villiers

1. What is the course o f acquisition o f the form? Is there a reliable order o f acquisition o f the construction? 2. H o w does the order o f acquisition relate to the cognitive complexity o f the form? A r e there related cognitive achievements? 3. What are the typical errors made i n acquisition o f the construction? 4. A r e there individual differences i n order o f acquisition, or i n the kinds o f errors made? 5. What is k n o w n about the information i n the input concerning that form? Do input variations affect acquisition? 6. H o w does the semantics o f the construction interact w i t h the syntax i n acquisition? 7. What is k n o w n about the pragmatics, or function, o f the construction? A r e pragmatic factors significant i n its acquisition? There were several instances i n w h i c h we could find no data bearing on one or more o f these questions, hence the sub-headings under each area vary to some extent. A l s o , the order o f consideration o f the questions is constrained by either the particular history o f investigation o f each f o r m , or the way the story best unfolds. Nevertheless, we hope sufficient uniformity exists to allow the reader to pursue particular interests throughout.

THE

DATA

4. W o r d O r d e r a n d B a s i c R e l a t i o n s i n S t a g e I Although some investigators have claimed to find the roots o f grammatical development i n children's single-word utterances (e.g. Greenfield & Smith, 1976; Ingram, 1971), the use o f clear syntactic devices i n English begins w i t h the earliest m u l t i - w o r d utterances. This first stage, defined by B r o w n as the period i n which mean length o f utterance ( M L U ) i n morphemes grows from 1.0 to 2.0 ( B r o w n , 1973; B r o w n & B e l l u g i , 1964), has drawn a great deal o f attention. I n part this stems from a natural interest i n the beginning stages o f the mastery o f English grammar; i n part i t arises from the possibility o f exploring the links between the emergence o f grammar and cognitive development at the end o f the sensorimotor period. Finally, children typically spend a few weeks i n the stage o f one- and t w o - w o r d utterances, w i t h longer utterances rare, so that researchers can collect fairly stable and sizeable samples o f speech that promise a good chance o f capturing crucial developments during this stage. Historically there have been several shifts i n emphasis i n attempts to charac­ terize language development during Stage I .

1.

4 . 1 . Syntactic

The A c q u i s i t i o n of English

39

Approaches

Accounts o f Stage I speech during the early 1960s tried to give purely formal, context-free descriptions o f the child's utterances, f o l l o w i n g the prevailing model i n linguistics o f w r i t i n g grammars for the language that were independent of semantic or pragmatic considerations. Some descriptions were fairly general, such as B r o w n and B e l l u g i ' s (1964) characterization o f Stage I speech as tele­ graphic—made up predominantly o f content words (nouns, verbs, and adjec­ tives), and omitting function words like prepositions, auxiliaries, and copulas (see also B r o w n & Fraser, 1963). Others described the phrase and clause struc­ ture o f the child's early sentences i n terms o f traditional grammatical categories like nouns, verbs, determiners, and adjectives (or modifiers i n general), and examined the elaboration and expansion o f noun and verb phrases as the child developed ( B r o w n & B e l l u g i , 1964; M e n y u k , 1969; M i l l e r & E r v i n , 1964). Most o f these approaches imposed syntactic categories and phrase structure rules from the adult language onto the child's utterances, although M i l l e r and Ervin (1964) distinguished between three classes o f operator words (words like this I that, althe) that frequently combined w i t h nouns or verbs. These were categorized separately on the basis o f a distributional analysis o f the sentence positions i n w h i c h they occurred and the lexical items w i t h which they com­ bined. A t approximately the same period o f time Braine (1963) performed dis­ tributional analyses o f Stage I speech i n an attempt to determine the child's o w n rules governing his initial w o r d combinations. Noting that for several children early t w o - w o r d utterances consisted o f a few high frequency words that appeared in fixed positions combined w i t h a much larger category o f less frequent words that appeared i n varying positions, Braine specified what he called a pivot grammar o f early child language. The high frequency words were fixed pivots around which simple sentences were organized: some ( P I ) always appeared i n initial position; others (P2) always followed the words w i t h w h i c h they were combined. The less frequent open class words (O) could combine w i t h either type o f pivot w o r d or w i t h each other; pivot words could not occur alone or combined w i t h each other. It was soon apparent that there were problems w i t h a pivot grammar as a complete account o f the rules underlying Stage I speech. The grammar could not appeal to semantic considerations to restrict the productivity o f the system, so i t predicted combinations that d i d not occur and made little or no sense (e.g. nightnight see, get say, and more away). I n the early 1970s research began to appear (e.g., Bowerman, 1973) demonstrating that the t w o - w o r d utterances o f some children i n early Stage I d i d not conform to the pivot and open pattern—words defined as pivots on the basis o f frequency and rigidity o f sentence position occurred alone and i n combination w i t h each other. But the strongest argument against pivot grammar was also an argument against any purely formal, context-

40

De Villiers and De Villiers

free account o f Stage I speech, namely that such analyses underrepresented the child's linguistic knowledge. B l o o m (1970) presented several examples o f struc­ tural h o m o n y m s — t w o - w o r d utterances that w o u l d be given the same formal description i n their surface forms, but which had clearly different meanings when considered i n context. The best k n o w n o f these is A l l i s o n ' s use o f Mommy sock, said once while A l l i s o n ' s mother was putting A l l i s o n ' s sock on, and once when A l l i s o n found her mother's sock. I f this difference i n underlying meaning was to be reflected i n the child's grammar, a different approach from a distributional analysis at the level o f surface structure was needed. B l o o m argued for an interpretive analysis (what B r o w n (1973) called a " r i c h " interpretation) o f chil­ dren's utterances based on what they said, the context i n w h i c h they said i t , and such other clues as the responses o f the adult listeners. Nevertheless, B l o o m d i d not abandon g i v i n g a formal syntactic description o f the early sentences. She suggested that the meanings o f the homonymous sentences be represented as distinct deep structures i n a transformational grammar, much as surface ambigu­ ous sentences i n the adult grammar were disambiguated i n deep structure (Chomsky, 1965). The same surface forms arose because elements from the deep structure were automatically deleted under cognitive processing limitations (by a deletion transformation) when the child produced the utterances. B l o o m there­ fore continued to describe t w o - and three-word utterances i n terms o f abstract linguistic categories like subject, verb, and object, plus transformational rules that derived surface forms from underlying deep structures. Interpretive analysis was used to decide what the deep structure should be. During the 1970s, although they were no longer the dominant approach to Stage I speech, formal syntactic analyses continued to be performed. B r o w n (1973) refined his earlier description o f telegraphic speech by specifying that not all the morphemes classified linguistically as functors (Gleason, 1961; Hockett, 1958) were reliably missing i n Stage I . Several phonological, syntactic, and semantic features o f those morphemes seemed to determine the likelihood that the child w o u l d use them. Fairly frequent and perceptually salient (syllabic and occasionally stressed) functors that express basic semantic or grammatical rela­ tionships i n sentences (e.g. pronouns or demonstratives) tend to be used produc­ tively quite early i n Stage I , especially i f they are not phonologically or gram­ matically conditioned by the linguistic contexts i n w h i c h they appear. I n contrast, inflections that merely modulate basic meanings and are usually un­ stressed, appear rarely, and i f at a l l , i n prefabricated routines that are not productive. Several studies i n the late 1960s had reported that predicate constructions seemed to emerge first i n Stage I , w i t h V + N sequences being the earliest and most frequent patterns ( M c N e i l l , 1970; M e n y u k , 1969). M e n y u k noted that the child herself was usually the presupposed subject o f such sentences and proposed that the child could describe her o w n actions before she could comment on the actions o f others. Sinclair (1976) similarly suggested that the distinction between

1.

The A c q u i s i t i o n o f English

41

child-as-agent and other-as-agent might only be acquired after the onset o f t w o word utterances. I n more detailed longitudinal studies by B l o o m (1970) and Ramer (1976), V E R B + O B J E C T sentences emerged before S U B J E C T + V E R B and

SUBJECT + OBJECT for some o f their children, but for other children all three forms seemed to emerge at the same time. Garman (1979) (following Crystal, Fletcher, & Garman, 1976) described the clausal and phrasal structure o f early m u l t i w o r d utterances i n terms o f noun and verb sequences plus other structural elements such as pronouns, adjectives, and prepositions, and more specific lexical items like this I that, here I there, and num­ bers. H i s analysis o f the data from Braine's (1976) study yielded a syntactic profile for each child at different stages and revealed variations i n clause and phrase structures across the children. For D a v i d , V + N predominated over N + V combinations i n both samples 1 and 2; for Kendall, N + V combinations predominated i n both samples; while for Jonathon, the t w o patterns were o f roughly equal frequency. Earlier emergence and greater frequency o f V + N sequences therefore does not appear to be a consistent pattern i n Stage I speech. On the other hand, toward the end o f Stage I when clause and phrase structures begin to " b l e n d " together into longer sentences (Crystal et a l . , 1976), the object noun phrase seems to be elaborated before the subject N P (Garman, 1979; Limber, 1973, 1976). This could reflect the predominance o f personal pronoun or proper name subjects i n Stage I , w h i c h do not allow for expansion or elabora­ tion (Limber, 1976). Considerable disagreement remains over the extent to which the Stage I child should be credited w i t h knowledge o f syntactic categories that are formally defined. For example, Bowerman (1973,1974) and Braine (1976) questioned the appropriateness o f analyzing Stage I sentences i n terms o f such abstract linguistic categories as subject and predicate, although they thought these may be appropri­ ate i n later stages. I n opposition, B l o o m (1970) and B l o o m , L i g h t b o w n and Hood (1975) argued that such categories are needed to capture the structural properties o f early sentences. Bowerman (1973) pointed out that the structural information that defines these abstract categories is absent from Stage I , and the characteristics o f early t w o - and three-word utterances are more compatible w i t h a description that is semantically based. Some researchers have even questioned the attribution o f such categories as noun, verb, and adjective to the c h i l d i n Stage I . Maratsos (1979) proposes that the child begins by learning specific information about the sentence position and other syntactic properties o f specific lexical items. Only after a considerable amount o f information has been collected i n this way does the child form more general categories o n the basis o f the distributional patterns shared by groups o f lexical items. Macnamara (1977) and Bates and MacWhinney (1982) argue that initially the child defines nouns, verbs, and adjectives semantically, as things, actions, and properties. I n sharp contrast to this, B l o o m (1970) and Garman (1979) assume the appropriateness o f these syntactic categories—if a lexical

42

De Villiers and De Villiers

item used by the child w o u l d be a noun i f so used by an adult, then it is classified as a noun. Valian (1984) tested for the presence o f the categories determiner, adjective, noun, and noun phrase i n the spontaneous speech o f six children. The tests consisted o f examining whether lexical items that prima facie fall into those categories were i n fact used i n appropriate distributional patterns that specify category membership for the linguist. Valian concluded that all o f the children had syntactic knowledge o f the four categories ( w i t h the possible exception o f one child who showed borderline performance on adjectives), and that k n o w l ­ edge o f the ways languages divide up such categories must be innate, so that not much syntactic evidence is needed for their formation. Unfortunately, the least developed o f Valian's subjects had an M L U o f nearly 3.0, so the appropriateness of using these categories to describe the syntactic knowledge o f the child i n Stage I remains i n doubt. Indeed, i t is possible that much o f the distributional evidence necessary for testing for the presence o f these categories is missing i n Stage I . Valian points out that very few syntactic tests allow reliable attribution o f the category V P , for example, and those that do are missing from early child speech. Different approaches to the child's acquisition o f linguistic categories like subject and object, or noun, verb, and adjective w i l l be discussed i n more detail later i n the chapter. 4 . 2 . Semantic

Approaches

I n her 1970 book, B l o o m had added to her formal syntactic description o f Stage I speech an analysis o f the semantic relationships holding between ele­ ments o f the sentence. She pointed out that syntactically similar negative sen­ tences can convey very different meanings when seen i n context. For example, one o f the children i n her study produced both no dirty soap and no pocket; the first while pushing away a bar o f soap, the second after searching her mother's skirt i n vain for a pocket. F r o m the context, B l o o m interpreted the first as a rejection o f the object and the second as an expression o f the nonexistence o f the object. She also reported that the different meanings o f the negative emerged at different points i n the course o f syntactic development—nonexistence first, then rejection, and finally denial o f the truth o f a preceding proposition. The child's acquisition o f underlying semantic notions could therefore interact w i t h syntactic learning i n determining the course o f language development. Schlesinger (1971) went still further i n proposing that all o f the child's t w o and three-word utterances derived directly from underlying semantic rela­ tionships (encoded i n what he called intention-markers or I-markers). He be­ lieved that the child perceived the w o r l d i n terms o f agents, actions, objects, and locations and learned to map these categories directly onto the syntactic forms available i n his language. Realization rules ( i n Stage I English, primarily w o r d order rules) derived the surface forms o f utterances from the underlying semantic I-marker. Thus while B l o o m resolved the ambiguity o f A l l i s o n ' s Mommy sock by

1.

T h e A c q u i s i t i o n of English

43

postulating t w o different formal deep structures, Schlesinger derived these utter­ ances from t w o different I - m a r k e r s — A G E N T + A C T I O N and M O D I F I E R + X (more specifically POSSESSOR + POSSESSED).

Applications o f semantic-based rule systems to Stage I speech dominated research on Stage I during the 1970s. Some researchers took linguistic systems like those o f Fillmore's (1968,1971) case grammar, Chafe's (1970) analysis o f verb relations, or generative semantics (Lakoff, 1972; M c C a w l e y , 1968), and applied them to c h i l d language. Others attempted to derive the child's o w n semantic categories from an analysis o f her speech i n context together w i t h assumptions about the cognitive and/or semantic distinctions available to her at this stage. Table 1.1 shows some o f the more frequent semantic notions or relations applied to Stage I speech by different researchers. There are apparent differences between the respective classifications, usually as one researcher divides into subcategories a semantic relation applied i n more general form by another. Schlesinger (1971) has a single category o f M O D I F I E R + X ; B l o o m et al. (1975) and Leonard (1976) subdivide that into R E C U R R E N C E , A T T R I B U T I O N , and POSSESSION. Schlesinger and Leonard use a single locative category; B r o w n has t w o , one i n v o l v i n g action and the other not; and B l o o m et al. suggest three locative categories. I n general, however, these researchers converged on roughly t h é same set o f semantic notions that described the com­ municated meaning o f almost all o f the early m u l t i w o r d sentences. B r o w n (1973), for example, reported that his limited set o f eight predominant types accounted for some 70% o f his subjects' t w o - w o r d utterances. Furthermore, the same semantic notions captured the meanings o f Stage I utterances across a wide range o f languages (see B r o w n , 1973; Bowerman, 1973), leading Slobin (1970) to suggest that " t h e rate and order o f development o f semantic notions are fairly constant across languages, regardless o f the formal means o f expression em­ p l o y e d . " Another positive aspect o f this approach was the continuity i t stressed between language acquisition and cognitive development i n general. I f the child in Stage I has knowledge o f abstract linguistic categories like subject and object that are defined by deep structure configurations, any theory o f learning must incorporate a substantial language-specific innate mechanism (Fodor, Bever, & Garrett, 1974; M c N e i l l , 1970). I n contrast, the semantic notions so prevalent i n Stage I speech reflected just those cognitive distinctions among agents, their actions, and the things acted upon, that developmentalists regarded to be the culmination o f the sensorimotor period—the period that ends at about the same age as syntax begins (Ingram, 1975a; Piaget, 1963; Sinclair-de Z w a r t , 1971). (The issue o f the relationship between cognitive development and the acquisition of semantic relations w i l l be discussed i n more detail i n section 4.4.) On the other hand, i t was apparent to several researchers that application o f semantic relations to children's speech was not without its problems. N o t i n g that some theorists subdivided semantic categories that were generic to others, B r o w n (1973) wrote that:

44

De VillJers a n d De Villiers TABLE 1.1 The M o r e Frequent S e m a n t i c N o t i o n s o r Relations A p p l i e d t o Stage I Speech b y Different Researchers

Schlesinger (1971)

Brown (1973) a

Bloom et al (1975) Existence

Introducer + X (Ostensive sentence)

Demonstrative + Entity

Negation + X Modifier + Head

Agent + Object Agent + Action Action + Object X + Locative

Notice^ Negation Recurrence Attribution Possession c

Entity + Attribute** Possessor + Possessed Agent + Object Agent + Action Action + Object Action + Locative Entity + Locative

X + Dative

Action Locative action Place of action Locative State State Intention Instrument Dative

Leonard (1976)

Example

Designated/object Nominative

This is a Teddy

Notice Negation Recurrence Attribution Possessor Agent Action Place

Experience Experiencer Instrument

Hi, Teddy No pocket More juice Red car Daddy pipe Adam ball Doggie bite Push car Walk store Swim pool Cup table Like candy I need it Wanna swim key open Give Mommy

Brown also notes uses of instrumental (Sweep broom), benefactive (For Daddy), dative (Give me book), experiencer (Adam see), comitative (Go Mommy), conjunction (in the sense of sequential naming of two present objects) (Kimmy Phil), and classificatory (Mommy lady) semantic notions, but they were infrequent and only appeared in a few of the children in Stage I. Vocatives (Notice), questions and negatives were placed in Brown's category of "other" constructions, the first because he regarded them to be of little grammatical or semantic interest, the latter two modalities because most of their syntactic development took place in Stage III and they seem to be a rather different class of semantic relation than the others. As major modalities they can be superimposed on almost any of the other types of semantic relation. ^The category of notice for Leonard differed markedly from that of Bloom et al. For Leonard any vocative counted as an expression of notice, but for Bloom et al. a verb of attention (e.g. see, hear, or look at) had to appear in the utterance. Bloom (1970) subdivided negation into expressions of nonexistence, rejection and denial. In their 1978 book Bloom and Lahey distinguish between five categories of negation—nonexistence, disappearance, nonoccurrence, rejection, and denial. Brown does not use a separate category of recurrence. Utterances involving more + TV were counted as a form of attribution. a

c

d

This . . . makes clear that the relations or roles are abstract taxonomies applied to child utterances: That it is not known how finely the abstractions should be sliced and that no proof exists that the semantic levels hit on by any theorist, whether Bloom, Schlesinger, Fillmore or whomever, are psychologically functional. Nor is this a nonsense question. It is an empirical question awaiting a technique of investi­ gation, (p. 177)

1.

The A c q u i s i t i o n of English

45

Bowerman (1976) outlines the myriad ways i n w h i c h action verbs i n a typical 2year-old's vocabulary can be subdivided on semantic grounds—for example, into those that refer to actions that are momentary (throw, hit) versus those that continue for a period o f time (carry, fly); those that result i n a change o f location (put, throw) versus those that result i n a change i n state (break, open); those that create something (build, draw) versus those that act on existing objects, and so on. A n extreme case o f fine distinctions made i n semantic notions is found i n Wells (1974), where attributive is subdivided into physical, quantity, class, evaluative, substance, and dispositional; experiences are further specified as physical, affective, cognitive (perceptual), cognitive (mental), and wanting; and locative becomes movement, directional, directional movement, target, and d i ­ rectional target. As B r o w n (1973) points out: Meanings of sentences may be described so generically as to make no distinctions (a sentence expresses a complete thought) or so specifically as to make as many distinctions as there are sentences or with any number of distinctions in between these extremes. The question is, which of the possible distinctions is, in fact, functional or "psychologically real." (p. 146) Howe (1976) critically examined the approaches o f Schlesinger (1971), B l o o m (1973), and B r o w n (1973), and concluded that the evidence from Stage I supports the existence o f only three broad semantic categories: action o f concrete object, state o f concrete object, and name o f concrete object. She notes that not only do the above researchers differ i n how they divide up some categories, but also i n how they categorize specific utterances even when they agree on catego­ ries. This arises from t w o fundamental problems: there are few i f any surface syntactic features that distinguish the specific semantic relations from each other in adult English; and when a particular syntactic rule like w o r d order is used to interpret an utterance, the nonlinguistic context may not specify a single i n ­ terpretation. Daddy hat, produced i n a context i n w h i c h the child's father is wearing a hat, could be intended to communicate either a possessive or a locative or a simple association between the t w o objects. Finally, H o w e (1976; see also 1981) questions whether there is sufficient developmental evidence to attribute to the child an understanding o f semantic roles like agenthood, possession, or attribute and all that they entail. W h i l e there is evidence from Piagetian tasks that the child at the onset o f syntax may be able to distinguish actors from their actions and objects from things done to them, that is not sufficient to ascribe to the child knowledge o f a general semantic role like agent or patient (see section 4.4 for further discussion o f this issue). These problems w i t h a rich interpretation o f children's utterances based p r i ­ marily on the nonlinguistic context are not as worrisome i f the researcher is trying to describe the emergence o f the communication o f different meanings by early single- or m u l t i w o r d utterances. Parents continually interpret the meaning of what their toddlers say (Cross, 1978; Nelson, 1973), and communication

46

De Villiers a n d De Villiers

between them is usually quite effective. Indeed, as we shall see when we discuss the issue o f the order o f emergence o f semantic notions, the kinds o f things little children talk about and are able to communicate may change systematically w i t h development. B u t i f the researcher is trying to determine the child's linguistic knowledge i n the form o f what categories enter into his earliest syntactic rules, then the criteria for attribution o f a category need to be stronger. Schlesinger ( 1 9 7 4 ) proposed t w o criteria for deciding the appropriate level at which to describe a child's utterances. Utterances that ( 1 ) enter the child's productive speech at about the same time, and ( 2 ) that share the same syntactic device, e.g. a particular w o r d order, are reasonably classified together as having the same underlying semantic relation and entering into the same rule. Braine ( 1 9 7 6 ) followed these criteria i n a study o f the Stage I language o f eleven children learning either English, Finnish, Samoan, Hebrew, or Swedish, and concluded that, 'the first productive structures are formulae o f limited scope for realizing specific kinds o f meaning. They define how a meaning is to be ex­ pressed by specifying where i n the utterance the words expressing the meaning should be p l a c e d . " According to Braine the categories upon w h i c h the ordering rules operate are semantic rather than syntactic, but they are narrower than categories like agent or action or possessor. A n illustration o f Braine's analysis is provided by Jonathan's use o f t w o - w o r d utterances using big and little. I n the first sample o f his speech Jonathan used big + X and little + X w i t h the w o r d for size consistently i n initial position. This suggested to Braine that a w o r d order rule like SIZE + X was operating. However, since utterances w i t h other adjec­ tives like hot + X and old + X appeared around the same time, Braine enter­ tained the possibility that the rule operated on a more general category like property or attribute. I n this case Braine argued that the more specific category was appropriate for the w o r d order rule because Johathan also produced utter­ ances w i t h wet i n a variable pattern wet + X or X + wet, as i f the ordering rule did not yet apply to them. Braine found some ordering formulae that were fairly common across children, e.g. patterns that draw attention to things (see + X, here I there + X), or location (X + PREPOSITION + here/there), and patterns indicating possession (my/PROPER N A M E 4- X ) . Others were less common and appeared i n few children. V E R B + OBJECT sentences were infrequent i n Braine's subjects and were usually o f variable w o r d order or reflected ordering rules applying to either individual lexical items (e.g. see + X or want + X) or small sets o f verbs w i t h a common semantic context like eat I bite I drink ( " o r a l con­ s u m p t i o n " ) . Surprisingly, Braine found that a broad A G E N T 4- A C T I O N pattern seemed to be productive for many children. 4

Braine's study was thus considerably more cautious i n attributing knowledge of semantic categories to children than were many earlier studies. A particular word order had to be used consistently above chance level, even i f the adult language could potentially use both orderings i n expressing a semantic relation e.g. The dog is here or Here is the dog to express the locative. The variety o f

1.

The A c q u i s i t i o n o f English

47

lexical items used i n a pattern was also important i n determining the semantic categories attributed to the c h i l d . Bowerman's (1976) study o f her o w n children, Christy and Eva, contained data that were even more valuable i n establishing the appropriate categories entering into early rules. I n addition to the earliest t w o word utterances, Bowerman had a record o f the one-word utterances preceding the onset o f syntax as w e l l as overlapping w i t h i t . Taped samples were recorded weekly and there were extensive daily entries, so the emergence o f syntax could be traced i n great detail. Bowerman could thus establish not only what t w o - w o r d combinations occurred but also w h i c h d i d not but might have done because the lexical items appeared i n single-word utterances at the same time. I n this way she showed that Eva's initial combination rules seemed to be lexically-based, pro­ ducing patterns like want + X, more + X, no + X, and here + X (while handing over an object). The pattern more + X d i d not spread rapidly to semantically similar words like again, no more, or allgone; i n fact, these terms only started to combine w i t h other words about a month after the appearance o f more + X and then d i d so one at a time over a period o f weeks. Rather mysteriously, Eva went rapidly from this system o f lexical-item-based combination rules to a much more mature system that seemed to Bowerman to be best described by context-free syntactic categories like subject and object. Bowerman could find no evidence for an intermediate stage o f rules based on broad semantic categories like agent or action. Christy, on the other hand, seemed to begin w i t h more broadly based seman­ tic categories and proceeded w i t h less discontinuity i n development. For exam­ ple, although she used many adjectives as one-word utterances and knew the names for the objects she seemed to be describing, Christy d i d not use M O D I F I E R + M O D I F I E D sequences until roughly t w o months after she produced her first two-word combination. W i t h i n a few days she then expressed many sentences like That wet, Daddy hot, and Bottle allgone, combining that or the name o f an object w i t h hot, wet, allgone, and alldone, w i t h the modifier always i n second position. A t about the same time she began to produce predicate nominatives like that airplane, forms that share w i t h predicate adjectives the copula construction in adult English, and may also be described as attributing something (a name) to an object (Bowerman, 1976). W e discuss individual differences i n the acquisi­ tion o f English syntax i n Stage I i n more detail i n section 4.7, but these examples from Braine and Bowerman suggest that many children begin w i t h syntactic rules of more limited semantic scope than those rules have i n the adult system. They also illustrate the k i n d o f caution necessary i n attempts to determine the child's syntactic knowledge i n Stage I . A different approach to children's knowledge o f w o r d order rules and seman­ tic relations has examined their comprehension o f sentences exemplifying a particular rule or relationship. Thus, de Villiers and de Villiers (1973a) investi­ gated Stage I children's ability to act out reversible active sentences like The boy pushed the girl i n w h i c h the only clue to the semantic roles o f agent and object is

48

De Villiers and De Villiers

the w o r d order. Children w i t h an M L U < 1.5 were unable to act out the sentences above chance level. Both Bowerman (1973) and de Villiers and de Villiers (1974) noted that children's agents i n early agent-object and agent-action sentences were overwhelmingly animate nouns and pronouns, so their ordering rules at this stage could apply to a more restricted semantic category than agent. They w o u l d then understandably fail to comprehend sentences i n w h i c h agenthood could not be identified w i t h animacy. Only i n late Stage I or Stage I I did children succeed at this task. Several other studies established that children rely on event probabilities i n interpreting agent-action-object sentences before they use w o r d order information alone (Chapman & M i l l e r , 1975; Strohner & Nelson, 1974). G o l i n k o f f and Markessini (1980) studied comprehension o f three types o f possessive relationships—alienable (the mommy's ball), intrinsic (the mommy's face), and reciprocal/reversible (the mommy's baby). For each o f 24 possessive sentences that their mother produced, the child had to point out the possessed object i n one o f t w o pictures. Both sensible and anomalous (e.g. the face's mommy) possessives o f each type were tested. A l l o f the late Stage I ( M L U > 1.5) children and three out o f six o f the early Stage I children were at or above 75% correct on picking out sensible intrinsic and alienable possessives when semantic probability constrained the interpretation. Reversible possessives, i n which w o r d order was the only cue, d i d not reach that criterion o f accuracy until M L U was over 4.0 (Stage V ) . These results suggest that Stage I children have a basic understanding o f the notion o f what objects are likely to be possessions and possessors, but cannot use w o r d order alone to comprehend a possessive rela­ tionship. H o w e (1981) criticized G o l i n k o f f and Markessini for assuming the chance probability o f pointing to the correct referent was 25% when the pictures contained t w o possessors and t w o possessions. She points out that the crucial possessor—mommy i n the mommy's shoe—only appeared i n one picture even though the possessed object occurred i n both. The child could thus reduce the chances o f being w r o n g by simply p i c k i n g the picture i n w h i c h the possessor referent was shown. Nevertheless, even i f chance is thereby reduced to 5 0 % , consistent performance above 75% by the late Stage I children suggests under­ standing o f the constrained possessives at this point. I n her o w n study i n 1981, H o w e investigated comprehension o f the semantic roles o f beneficiary and locative i n a similar picture-cued comprehension pro­ cedure. The target referent appeared i n both pictures but i n different roles. For example, when asked, " W h e r e is the hat on the b o y ? " the child had to choose between a picture o f a girl putting a hat on a boy and a picture o f a boy putting a hat on a g i r l . For " W h e r e is the tea for the l a d y ? " the choice pictures depicted a man giving tea to a woman versus a woman g i v i n g tea to a man. Stage I children could not pick out the appropriate picture representing a beneficiary or locative, even though control conditions revealed that they knew the lexical items and succeeded i n the task when the object (hat or tea i n the examples above) was

1.

T h e A c q u i s i t i o n o f English

49

varied rather than the beneficiary or locative. Late Stage I I and Stage I I I children did show an understanding o f these semantic roles. In conclusion, recent studies o f both production and comprehension question early Stage I children's use o f broad semantic roles like agent and beneficiary and the syntactic means o f expressing them i n English. Early rules o f production and strategies o f comprehension are frequently based on more specific semantic categories or particular relationships between referents. W e address the issue o f the child's conceptual understanding o f agent, action, location and so on, later i n this section. 4.3.

Order

of

Acquisition

In their longitudinal studies o f the emergence o f semantic notions B l o o m et al. (1975) and Leonard (1976) both conclude that the notions are acquired i n a consistent order i n Stage I and Stage I I . I n sharp contrast, Braine (1976) ob­ served no consistency across children i n the order i n which syntactic patterns to express semantic notions became productive. I n fact, for t w o o f his Englishspeaking subjects, A n d r e w and Kendall, there was no overlap i n productive patterns i n Stage I . Several factors seem to determine whether or not a common order o f acquistion is observed w i t h i n and across studies. Aside from sample size and frequency of sampling, a major determinant is whether evidence for a consistent and pro­ ductive syntactic rule is required before a semantic notion is attributed to the child. Braine required that consistent w o r d order be used to mark a semantic notion, so a fairly substantial number o f utterances o f a given type were needed to establish productive control o f the rule. B l o o m et al. and Leonard relied more heavily o n a rich interpretation o f the utterances i n context and allowed w o r d order to vary. I n B l o o m ' s study the child was credited w i t h productive mastery o f a semantic notion when five utterances expressing that notion had been observed; for Leonard the first nonstereotypical m u l t i w o r d expression o f a notion sufficed. These methodological differences clearly affect the reported orders o f acquisi­ tion. For example, B l o o m (1970), Ramer (1976), and Leonard (1976) all report either earlier expression o f A C T I O N + O B J E C T than A G E N T + A C T I O N , or s i m u l ­

taneous emergence o f these relations; Braine (1976) reports that A G E N T + A C T I O N constructions were consistently productive before constructions express­ ing A C T I O N + OBJECT. For Braine that means that consistent positional formulae for agent + action preceded the formulae for A G E N T 4- OBJECT. Leonard (1976) reanalyzed the data from the subjects on w h i c h Braine had longitudinal data, using his o w n criterion for attribution o f a semantic notion to the child. The emergent order o f acquisition o f agent, action, and object categories as w e l l as other semantic notions was then much more consistent w i t h the other studies. Leonard therefore distinguishes between the child's ability to communicate a semantic notion i n a m u l t i w o r d utterance and his mastery o f a syntactic device to encode that notion.

50

De Villiers a n d De Villiers

Another factor influencing convergence on a common order o f acquisition is the level o f specificity o f the proposed semantic notions. Braine's set o f mean­ ings expressed by early syntactic formulae are quite narrowly defined, and vary from child to child. B l o o m et al. and Leonard applied a simple system o f relativity broad semantic notions to all o f their subjects, although there remain some differences i n how these notions are defined i n the t w o studies. Table 1.2 shows the overall order o f acquisition for each o f these t w o studies produced by averaging the ranked acquisition order o f the notions across the children. There was substantial similarity i n the ordering o f notions among the children i n each study. For B l o o m et al.'s four subjects the concordance was greater than .8; for Leonard's eight subjects i t was over .5 (p < .001 i n each case). Grouping the notions together into broader categories produces an order o f acquisition that is very similar across these and other studies (e.g. Wells, 1974). Operations o f reference that name an object or note its presence, absence, disap-

TABLE 1.2 Order of A c q u i s i t i o n of M a j o r S e m a n t i c Relations in Early M u l t i w o r d Utterances Bloom et al. (1975) Existence (Nomination) Recurrence ! Negation Action Locative Action Attribution Possession State Notice Locative State Intention Instrument Dative

Leonard (1976) > a

b

Notice Nomination Action Place Agent Negation Attribution Possessor Recurrence Experience/Experienced Instrument c

"This ordering leaves out Leonard's category of designated/ object, which is so defined that it must appear when either notice or nomination occur. Any such referring expression must involve designation of an entity. Experience and experiencer are also conflated. Leonard scored them separately, but in fact any utterance containing reference to an experience also spec­ ified the experiencer e.g. / need letter please. I wanna see that. \n cases in which a semantic notion did not occur in any sample from a child (they were only recorded up to an M L U of 2.3) it was ranked as the last to be acquired. Specification of an instrument only appeared in the sample of speech from two of b

1. pearance, or recurrence

T h e A c q u i s i t i o n o f English

51

are usually the first to appear, and they are closely

followed by expressions encoding actions. Specification o f the properties or static relationships o f objects (attribution, possession, location) emerge next, then reference to experiences or internal states o f the subject. F i n a l l y , construc­ tions i n v o l v i n g instruments or dative recipients o f actions become productive very late i f at all i n Stage I . As the categories are more n a r r o w l y defined, however, the agreement be­ tween studies begins to disappear. The major discrepancies between the orderings o f the semantic notions i n Table 1.2 involve the categories o f notice, negation, and recurrence. The later acquisition o f notice i n B l o o m et al.'s c h i l ­ dren is readily explained by differences i n the definition o f the category. Leonard accepted any vocative (e.g. Hi, camel) plus utterances l i k e Here box and Hey shoe as an expression o f notice; B l o o m et a l . required a verb o f attention such as TABLE 1.2 (continued) the eight children in Leonard's study; expressions of recurrence were missing for two of the children. Leonard's place category subsumes Bloom et al.'s notions of locative action, locative state, and place of action. ^Leonard's experience and experiencer categories incor­ porate several of Bloom et al.'s categories—state utterances (those with verbs like need and like), notice utterances (those with verbs of attention like see, hear, and look at), and inten­ tion utterances (e.g. / want go park). Actually there are some ambiguities about Leonard's classification of experience/experiencer utterances. Consider the following three utterances: c

See that (Lynn turns the cash register over, bends and looks.) I see hat (Alec sees a toy barrel on the floor.) I see in here (Morton hits the cash register drawer, which opens. Morton looks inside the cash register.) Leonard classifies the first as notice, the second as AGENT + ACTION,

and the third as E X P E R I E N C E R + E X P E R I E N C E . Pre­

sumably the grounds for the difference in categorization have to do with the point in development at which these utterances were observed, but when one is attempting to determine the order of acquisition of the notions, that is a dangerous degree of flexibil­ ity to have. There may be grounds for classifying initial sen­ tences with want and see separately, since several researchers (e.g. Wells, 1974; Braine, 1976) have found utterances with these verbs early in Stage I even though reference to other feelings and perceptions is late. But evidence of the limited semantic nature of such combinations and the availability or nonavailability of other words for experiences should be gathered to justify such a decision. Many of the initial instances of EXPERIENCER + E X P E R I E N C E in Leonard's data involved use of / want X.

De Villiers and De Villiers

52

see or look at. The relatively late acquisition o f recurrence and negation i n Leonard's subjects is not as easily accounted for, though it may be a sampling artifact. These notions were infrequent i n the speech samples from Leonard's subjects (for t w o subjects expressions o f recurrence were not observed at all i n Stage I ) , and their rank orderings across the eight subjects were particularly variable. Frequent constructions w o u l d be more likely to appear i n a sample and so be counted as acquired (see Bowerman, 1975). I n Leonard's study the rank order correlation between the order o f acquisition o f the semantic notions and their frequency i n the children's speech was greater than .90. The relative sali­ ence o f the notions may also depend on the situation or may vary across children (see the discussion i n section 4.8 o f pragmatic factors that may affect the ex­ pression o f semantic notions). Examining more narrowly defined semantic relations i n other studies reveals disagreement about order o f acquisition. B l o o m (1970) wrote that m u l t i w o r d negative sentences emerged i n the order nonexistence, then rejection, and then denial; M c N e i l l (1968) found denial before rejection i n the child he studied; and Ramer (1976) observed great variability across her seven subjects i n the order i n which the negations were produced. B l o o m et al. (1975) and Wells (1974) reported locative actions ( A moves to B ) before locative states ( A is at B ) , but Braine (1976) found productive expressions o f locative states before locative actions i n some children and simultaneous emergence o f these t w o notions i n others. Finally, Wells (1974) reported that utterances concerned w i t h the func­ tions o f people (eat, play, sing, kiss, etc.) are produced before utterances noting changes o f state (break, open, cut, etc.). Braine (1976) found just the opposite. A l l o f this suggests that only when meanings are defined i n general terms is there orderliness i n the child's changing ability to communicate meanings that may be tied to his cognitive development. There is less orderliness i n the mastery of specific syntactic rules to express those meanings. W e w i l l now turn to the issue o f cognitive constraints on the expression o f semantic relations i n Stage I , and w i l l return to consideration o f individual differences i n the acquisition o f syntactic devices at a later point. 4.4.

Cognitive

Development

and Semantic

Relations

Three different issues have dominated discussions o f the relationship between cognition and semantic or syntactic development i n Stage I : 1. What are the prerequisite cognitive skills for the emergence o f m u l t i w o r d utterances? 2. What is the relationship between the order o f acquisition o f semantic notions and the cognitive distinctions that may underlie them? 3. D o semantic categories like agent, action, and object that enter into early syntactic rules simply map onto existing cognitive categories o f the same scope?

1.

The A c q u i s i t i o n of English

53

Several theorists have argued that the internalization o f action schemas into the mental representations that are the culminating achievements o f Piaget's sixth stage o f sensorimotor development (at approximately 18-months-of-age) repre­ sent the necessary prerequisites for w o r d combinations (see for example, Ingram, 1974,1978; M c N e i l l , 1974; M o e r k , 1975; Morehead & Morehead, 1974; Sin¬ clair-de Z w a r t , 1971). The cognitive achievements postulated to occur at this stage include the concept o f permanent objects that continue to exist independent of actions upon them, together w i t h an understanding that whole categories o f objects can be used for the same action and a variety o f actions can be performed w i t h one object. I n addition the child comes to understand the causal relationship between objects that can initiate actions and changes i n the objects that are acted upon. The argument that these achievements must underlie any ability to produce operations o f reference, or utterances expressing semantic relations among agents, actions, or objects, is plausible on the basis o f logic alone, but it has proven difficult to establish a clear synchrony between the cognitive and l i n ­ guistic skills. M o s t o f the earlier accounts o f cognitive prerequisites for the acquisition o f syntax supported the logical argument by appeals to correlations between the age norms for the mastery o f the different Piagetian tasks and the emergence o f t w o - w o r d utterances. M o r e recent empirical studies that tested the relationship between performance on Piagetian tasks and various measures o f language acquisition i n the same children have produced only equivocal support for such a strong correspondence between skills i n the t w o domains. Some children appear to pass through Stage 6 o f sensorimotor operations as much as 6 months before the onset o f syntax; others produce t w o - w o r d utterances before they succeed on Stage 6 cognitive tasks. Sixty-five percent o f the children i n a study by Bates, Benigni, Bretherton, Camaioni, and Volterra (1977) passed the Stage 6 object permanence task around 13 months o f age and many o f them showed other Stage 6 behaviors like symbolic play, deferred imitation, or memory for absent objects, long before m u l t i w o r d utterances emerged i n their speech. Bates (1976) therefore suggested that it is not the onset o f mental representation, but increases i n the span o f the child's representational capacity, allowing h i m to represent and operate on groups o f objects or events, w h i c h are crucial for the onset o f syntax. This is similar to B l o o m ' s (1973) notion that the relations among objects and events come to be represented i n the late sensorimotor period. However, outside o f informal behavioral observations like those o f Piaget, it is not clear how to assess the achievement o f this cognitive ability. Ingram (1975a) and Corrigan (1976) both reported that most o f the children i n their longitudinal studies began to produce their first m u l t i w o r d utterances just after showing Stage 6 performance on object permanence tasks. But one child i n each study produced m u l t i w o r d utterances before reaching Stage 6. Ingram noted that the maverick child i n his study was the most imitative o f the children and suggests that her early sentences may have been imitated routines and not syntac­ tically productive.

54

De Villiers and De Villiers

Corrigan (1979) points out that much o f the confusion i n this area stems from different operational definitions o f representation and object permanence. A l l measures o f mental representation may not be equivalent, and there is no sub­ stantial synchrony i n children's mastery o f the tasks Piaget used to define stages of sensorimotor development—in fact, decalage is the rule rather than the excep­ tion (Fischer, 1980; Uzgiris, 1976). A close correspondence between cognitive and linguistic skills is therefore only likely to be found when the same skills are required and the task demands are similar i n the t w o domains. It does not follow from the fact that understanding o f semantic notions may depend on cognitive abilities that they must emerge i n child language i n the same order as they are acquired nonlinguistically (Fodor, Bever, & Garrett, 1974). Nevertheless, a plausible account o f the order o f acquisition o f general semantic categories can be given i n terms o f their cognitive difficulty, defined by the order of attainment i n cognitive development (Leonard, 1976). Thus operations o f reference like notice, nomination, and recurrence seem to require only the level of understanding o f objects that is achieved i n the third and fourth stages o f the sensorimotor period ( B r o w n , 1973), and they are the earliest semantic notions to appear. Specification o f agents and objects, however, requires encoding and relations between objects and events, plus an understanding o f causality i n the sense that the child appreciates that other people as w e l l as herself are potential sources o f actions as w e l l as recipients o f actions. Similarly, semantic notions like attribution and possession seem to reflect the child's ability to represent the properties and relationships between objects independent o f relevant action. These cognitive achievements emerge i n Stages 5 and 6 o f sensorimotor intel­ ligence (Piaget, 1954). Finally, notions o f experience and experiencer do not deal directly w i t h overt activity but rather encode internal psychological state. Since the primary source o f information i n the sensorimotor period is overt action, these more abstract notions w o u l d be expected to emerge relatively late i n Stage I (Leonard, 1976). The same assessment o f cognitive difficulty based on nonlinguistic development cannot be used to account for the time lag between reference to agents and reference to instruments, however. Piaget (1954) sug­ gests that the understanding o f causality that underlies both o f these notions emerges i n Stage 5, when the child is observed to use objects as tools and also seems to appreciate the agency o f other people. Some writers have suggested that linguistic semantic categories map directly onto cognitive categories o f the same scope formed i n prelinguistic cognitive development (Sinclair-de Z w a r t , 1971). Early meanings are seen as representing a set o f internalized sensorimotor schemata rather than as a set o f linguistic deep structures (Bates, 1976). Others have distinguished more sharply between con­ ceptual and linguistic knowledge. Thus, although B l o o m (1973) stated that "children learn a language as a linguistic coding o f developmentally prior con­ ceptual representations o f experience" (p. 16), she does not believe that cog­ nitive categories develop " i n a one-to-one correspondence w i t h eventual l i n ­ guistic categories" (p. 121). According to B l o o m et al. (1975) the cognitive

1.

The A c q u i s i t i o n of English

55

representations o f the relationship among objects and events are more global, not distinguishing between the respective semantic roles o f the elements, but rather representing the entire relationship. " S u c h differential semantic categories as agent, place, affected object, etc. are linguistic inductions that the child has made on the basis o f his linguistic experience" ( B l o o m et a l . , 1975, p . 30). A similar point is made by Bowerman (1975) when she argues a child's general cognitive understanding o f object permanence, causality, or actions as distinct from the objects acted upon, does not tell the child how finely or broadly to slice the semantic categories that rules o f w o r d combination might act upon. Schlesinger (1974,1982) proposes that exposure to the way the language encodes various events and relationships interacts w i t h the child's nonlinguistic experience i n forming semantic categories. For example, " b y hearing sentences in w h i c h all agents are treated the same w a y , the child acquires the agent concept w i t h rules for realizing it i n his speech" (1974, p . 45). There is little empirical evidence to support either position, but recent devel­ opmental research has set out to show that infants possess general concepts o f agent and recipient (Golinkoff, 1981). G o l i n k o f f and co-workers investigated the habituation and recovery o f visual fixation to films o f agents engaging i n actions on objects. For example, G o l i n k o f f and Kerr (1978) showed films o f a man ( A ) pushing another man ( B ) to 15- to 18-month-old infants who were not yet producing t w o - w o r d utterances. The positions o f the actors and the direction o f the movement were continually changing, but A always pushed B . After several presentations visual fixation time to the f i l m declined, then the agent and recip­ ient roles were reversed—now B pushed A — a n d the recovery o f fixation time was measured. Significant recovery was observed, and G o l i n k o f f and Kerr i n ­ terpret this as showing that the infants had perceived the change o f actor and recipient. However, Schlesinger (1982) has rightly argued that this experiment only shows that the infant could discriminate between " A as pusher" and " A as pushed," not between categories as broad as agent and recipient. A subsequent experiment by G o l i n k o f f (1981) is more promising because i t varied the actions that A performed on B i n the habituation series. I n the dishabituation test, either A performed a new action on B (action change), or B performed the same new action on A (action change plus role reversal). Unfortunately, only the younger girls (16 months) and older boys (24 months) showed more recovery from habituation for the role reversal w i t h an action change than to the action change alone. The other subjects dishabituated to both test films to the same degree. Differential recovery is needed to argue that the child can detect the additional role change. Clearly more research is warranted to investigate the k i n d o f agent and action events that presyntactic infants can discriminate. 4.5. Relation

to

Input

There is considerable disagreement about the ways i n which modifications i n parental speech to children might facilitate early language acquisition. Studies that have correlated differences i n mothers' speech to their children i n Stage I

56

De Villiers and De Villiers

w i t h the children's rate o f language development have come out w i t h d i ­ ametrically opposed results and conclusions. For example, Newport, Gleitman, and Gleitman (1977) studied the language o f 15 mother-daughter dyads at t w o points 6 months apart. A t the first sample the children fell into three age groups—12-15 months, 1 8 - 2 1 months, and 2 4 - 2 7 months. Partialling out age and the initial language competence o f the c h i l d , Newport et al. correlated the types o f constructions (e.g. declarative, imperative, yes/no questions, and w h questions) i n the parental speech at the time o f the initial sample w i t h measures of the child's language development over the 6 month period. They distinguished between "language general" (i.e. universal) aspects o f the children's speech that included major structural variables like the number o f verb or noun phrases per utterance, and "language specific" aspects such as use o f auxiliaries or noun inflections. The obtained correlations suggested that variations i n the input had only a limited effect on language specific aspects o f the child's development— for example, frequent use o f initial auxiliaries i n yes/no questions was positively related to the later use o f auxiliaries i n the children (see p . 80 for a fuller discussion o f this finding). There were no significant effects on more general structural aspects o f the child's language acquisition. Newport et al. concluded that "whether mothers speak i n long sentences or short ones, restricted or wideranging sentence types, complex sentences or simple ones—none o f these plau­ sible candidates for a teaching style have a discernible effect on the child's language growth during the six month interval we investigated" (p. 136). I n sharp contrast to this, F u r r o w , Nelson, and Benedict (1979) found that several aspects o f mothers' speech reflecting the use o f a simpler communicative style were positively correlated w i t h productive language growth i n their c h i l ­ dren. Seven first-born children (four boys and three girls) were recorded i n verbal interaction w i t h their mothers at 18 months and again at 27 months. A t 18 months six o f the children had no word-combinations i n their speech ( M L U = 1.0) and the seventh had very few ( M L U < 1.4), so both the age and level o f language production at the first sample o f mothers' speech was controlled for. Several factors that added semantic or syntactic complexity to the mothers' speech when the children were 18 months o l d were negatively correlated w i t h basic measures o f the children's language production at 27 months. For example, a mother's use o f more verbs, more pronouns relative to nouns, and more copulas and contracted forms correlated negatively w i t h her child's M L U , verbs per utterance, and noun phrases per utterance nine months later. The major difference between these t w o studies lies i n how they controlled for differences in the children's language abilities when the initial sample o f parental speech was taken. Newport et al. partialled it out statistically; Furrow et al. recorded the sample when the children were at the same age and linguistic competence. Furrow et al. argue that the weakness o f the statistical procedure is that i t assumes that variations i n the input are going to have similar effects regardless o f the age and levels o f language development over w h i c h the 6 months o f growth

1.

The A c q u i s i t i o n of English

57

are measured. Furthermore, Nelson (1982) points out that statistical adjustments for overall language competence i n terms o f utterance length may w o r k un­ equally for different aspects o f syntax. For example, i n his subjects, axiliary use correlated only weakly w i t h utterance length; noun phrase complexity correlated more strongly. Hence, adjustments for initial M L U may not control for level o f syntactic ability i n specific domains. Finally, H a r d y - B r o w n (1983) points out that the likelihood o f replicating a significant correlation is very l o w w i t h small sample sizes like those i n the above studies. O f 24 mother-child correlations i n common between Newport et al. and Furrow et al. only one was significant i n both studies—half o f the others were i n the same direction, but the remainder were i n opposite directions. W e must conclude that evidence for a strong relation between the parental input i n Stage I and rate o f syntactic development i n the child's language production remains equivocal. These studies primarily explored syntactic aspects o f parent and child lan­ guage, but others have looked more closely at semantic aspects. Wells (1974) suggested that children learn linguistic structure from hearing adults describe situations and events that they clearly understand i n cognitive terms. For h i m an ideal situation for the child acquiring the linguistic encoding o f semantic rela­ tions w o u l d be ' a shared activity w i t h an adult i n which the adult gave linguistic expression to just those meanings i n the situation which the child was capable o f intending and to w h i c h he was at that particular moment paying attention" (p. 267). Several studies o f mothers' linguistic interaction w i t h their young children have commented on the extent to w h i c h mothers restrict themselves to talking about the here-and-now for the c h i l d (see Snow, 1977, 1979 for reviews). 4

I n her study o f 16 children aged 19 to 32 months, Cross (1977) noted that 7 2 % of the maternal utterances encoded the child's or mother's ongoing activities or referred to present persons or objects. Furthermore, there was a close referential and semantic contingency between the child's utterances and the mother's utter­ ances. Some 55% o f the mothers' utterances referred back to topics i n the children's speech, or incorporated them exactly. Overall, the less mature chil­ dren i n comprehension and production received a greater proportion o f these semantically related utterances, suggesting that the parental speech was tailored or " f i n e - t u n e d " to the linguistic competence o f the child. I n a different analysis of the same data, Cross (1978) discovered that children receiving a greater number o f semantically related utterances (expansions and semantic extensions of their o w n utterances) were likely to be linguistically accelerated i n M L U relative to children o f the same age who received fewer o f them. In a study o f nine Dutch mothers, Snow (1977) found that 65 to 87% o f their utterances to their toddlers could be described by B r o w n ' s (1973) eight prevalent semantic relations from Stage I speech. Van der Geest (1977) reported a still more dramatic case o f fine-tuning o f parental input to children's level o f develop­ ment. V a n der Geest and co-workers scored the speech o f eight Dutch motherchild pairs for reference to a range o f underlying " s e m a n t i c " notions (including

58

De Villiers and De Villiers

tense, modals, locative, possession, and definite and indefinite reference) based on a rich interpretation o f the utterances i n context. I n general, the frequency o f the semantic notions peaked i n the children's speech before they peaked i n the parental input, but frequency o f the syntactic realization o f those notions peaked first i n the mothers' speech. V a n der Geest concludes that " i n mother-child interaction the child somehow determines how complex the daily conversation w i t h h i m may be i n semantic cognitive terms, and the mother takes the oppor­ tunity to provide the child w i t h the correct realization rules to cover the semantics of the conversation" (1977, p . 91). Retherford, Schwartz and Chapman (1981) set out to test this claim o f finetuning by the mother by examining the relative frequency o f 15 semantic roles and five broader syntactic categories i n the speech o f six English speaking mothers and their toddlers at t w o points i n development. The M L U ' s o f the children varied from 1.0 to 1.95 (mean = 1.32) at the first sample, and from 1.38 to 2.92 (mean = 1.91) when the second sample was taken 3 to 6 months later. Mother and child utterances from a half-hour play session were coded for the expression o f the f o l l o w i n g semantic roles, using a rich interpretation from context: action, agent, object, locative, demonstrative, recurrence, possessor, quantifier, experiencer, recipient, beneficiary, comitative, created object, instru­ ment, and state. Also coded were five grammatically defined categories: oneterm expressions o f entity, multiterm expressions o f entity, negation, attribute, and adverbial. Note that both one-word and m u l t i w o r d utterances were included in the scoring, and several o f the "syntactic categories" were used as semantic notions by B l o o m et al. (1975) and Leonard (1976). Cross-lagged panel correla­ tions on the relative frequency o f use o f the semantic roles and syntactic catego­ ries were used to determine the probable direction o f effect—did the mother tune in to the child or vice versa? The pattern o f correlations suggested that the mothers were ahead o f the children. The children's use o f the categories from sample one to sample t w o changed to become more like the mothers' both i n which semantic roles were present and i n how frequently they were used. I f a mother used a semantic role w i t h relatively high frequency i n sample one, the child was more likely to use it frequently i n sample t w o : the reverse was not true. Overall, the mothers used a fairly wide range o f semantic categories i n both samples, and differences i n frequency were quite stable. B r o w n ' s (1973) preva­ lent semantic relations accounted for around 70% o f the multiterm utterances that were categorized. Retherford et al. argue against the fine-tuning hypothesis and suggest that the relative frequency o f the different semantic roles i n the mothers' speech, and presumably the children's speech as they became more similar to their mothers, is determined by pragmatic factors to do w i t h the here-and-now topics o f conversation and the redundancy o f certain roles i n dyadic interaction. The effects o f pragmatic factors on the frequency o f semantic relations is consid­ ered i n section 4.8. There are many differences between the studies o f V a n der Geest et al. and Retherford et al. that could account for the discrepancy i n their results; not the

1.

The A c q u i s i t i o n of English

59

least being the difference i n languages studied, differences i n the semantic no­ tions scored and i n the way the data were analyzed. Thus it is fairly w e l l established that mothers' speech to young children is to a large degree limited to the objects and events o f the ongoing interaction, tends to maintain the topic o f conversation (especially to linguistically less developed children), and i n that sense provides an input that could enhance learning o f the way the language encodes different semantic relations. What remains unresolved is the extent to which particular simplifications i n the input do or do not facilitate acquisition o f those semantic and syntactic forms. Finally, the way i n which such pragmatic factors as the topic o f conversation and the desire to communicate effectively control the relative frequency o f different semantic and syntactic features i n mother-child conversations also needs further specification. 4.6.

Typical

Errors

The most obvious error i n Stage I speech is the omission o f elements that are obligatory i n adult language, either entire constituents o f the sentence or gram­ matical morphemes that modulate its meaning. The gradual acquisition o f gram­ matical morphemes like articles, inflections, and prepositions is the topic o f the next major section o f the chapter. Here it w i l l suffice to note B r o w n ' s (1973) characterization o f the '/telegraphic" look o f early utterances. I n English, as well as i n German, Finnish, and Hebrew ( B r o w n , 1973), the morphemes left out in Stage I are primarily those that have little phonetic substance (i.e. are nonsyllabic and/or unstressed) and do not encode primary semantic relations. Slobin's (1973,1985) "operating p r i n c i p l e s " supplement B r o w n ' s characteriza­ tion by specifying processing variables that contribute to the early or late emergence o f grammatical morphemes i n many different languages. Several explanations have been proposed for the omission o f major constitu­ ents like agent, action, or object. The least interesting case is when the child does not k n o w the lexical item needed to express one o f the semantic roles; however, many times the child possesses the requisite vocabulary but still deletes constitu­ ents. The question that arises i n these cases is to what extent the missing element should be represented i n the proposition underlying the child's utterance (whether that is conceived o f as a semantic I-marker or an abstract syntactic deep structure). The question is posed most forcefully by replacement s e q u e n c e s successive t w o - and three-word utterances i n w h i c h all the elements o f an agentaction-object sentence are expressed, but never all i n a single utterance. For example, B l o o m (1970) notes the sequence: Lois read. Read book. Lois book. This leads her to suggest that all the obligatory constituents are present i n the deep structure o f the sentence, but because o f cognitive processing limitations

60

De Villiers and De Villiers

the child often cannot express them all i n one utterance. One or more constituents is automatically deleted i n production. Several factors that plausibly increase information processing load seem to increase the likelihood that an element is deleted. The addition o f negation is one such factor, and B l o o m reports the following sequence from Kathryn: Me like coffee. Daddy like coffee. Lois no coffee. Other factors are the use o f two-part verbs like turn on and take off, the addition of modifiers to the object noun phrase (e.g. Drive blue car), or the use o f words that the child has just acquired, especially new verbs or pronouns. Use o f articles or verb inflections d i d not seem to make constituent omissions more l i k e l y . Other theorists (e.g. Antinucci & Parisi, 1973; Greenfield & Smith, 1976; and Parisi, 1974) do not accept B l o o m ' s notion o f underlying syntactic deep struc­ tures and deletion transformations, but they too argue that the child intends to communicate an entire proposition. Thus, an elaborate semantic structure under­ lies both one- and t w o - w o r d utterances, but it is not fully expressed because o f cognitive limitations. Those elements that are semantically most informative are the ones produced, while elements that are more redundant i n context are omit­ ted. For example, when it is obvious that the child herself is the agent o f an action on an object, the agent w i l l frequently not be specified. B r o w n (1973) offers a complementary explanation. He suggests that the Stage I child may consider all sentence constituents to be optional, not i n the strictly syntactic sense that prepositional phrases or adverbial modifiers are optional elements i n predicate noun phrases, but i n the sense that elliptical utterances are allowed i n discourse. He points out that parental speech models many nonsentence fragments that are usually well-formed constituents o f sentences. For example, rules o f ellipsis allow answers to Wh-questions to specify just the constituent being queried, What did the boy do?-. Who hit the dog?

Hit the dog. The boy did.

Parents also often repeat a final constituent or some other element o f a sentence to get the child to respond or to emphasize what is being communicated, e.g. Give me the red truck. The red truck. Give it to me. That truck. In addition, imperatives i n which the subject is understood (e.g. Go to your room) and truncated yes-no question forms like Want your juice? are quite

1.

The A c q u i s i t i o n of English

61

frequent i n the input to the child. Finally, Maratos (1984) notes that i n de­ clarative sentences i n English it is acceptable to delete the object for some verbs but not others. Thus, He ate something.

He ate.

but not, He made something.

He made.

So the child must k n o w w h i c h verbs must take objects and w h i c h need not. B r o w n argues that the child may develop the impression that ellipsis is more freely allowed than it is, that a full sentence or any part o f it can be produced i n discourse. The child has to learn when more o f the sentence must be expressed i n order to communicate effectively, as w e l l as when i t is not optional to delete elements even though communication w o u l d not be impaired. However, Maratsos points out that B r o w n ' s theory about the optionality o f constituent elements must be supplemented by some specification o f w h i c h ele­ ments are pragmatically more important to the child i f it is to account for the pattern o f deletions i n Stage I speech. A l l constitutents are not deleted equally often; i n particular, deletion o f the verb is relatively rare. O f the children i n Brown's (1973) sample, only Eve showed an equal frequency o f agent-action, action-object, and agent-object utterances; agent-object utterances d i d not appear in either A d a m or Sarah i n Stage I . B l o o m et al. (1975) also recorded agentobject sequences i n only t w o o f their four subjects. For some children agentaction predominate (e.g. Kendall (Bowerman, 1973) for others action-object strings are most frequent (e.g. A d a m ( B r o w n , 1973)). (Individual differences and pragmatic factors are discussed i n more detail i n later sections.) Another error i n Stage I speech that has been much discussed is the use o f incorrect w o r d order, primarily because o f its apparent rarity. As we outlined on p. 28, English has a normal S V O (agent-action-object) order, and B r o w n (1973) remarks that " i n all the samples o f Stage I English listed i n Table 9 (some 17 samples from 12 children) the violations o f normal order order are triflingly f e w " (p. 189, emphasis added). Many researchers have therefore used word order as a cue to the intended semantic relations i n a rich interpretation o f early m u l t i w o r d utterances. Stage I children do seem to adopt fairly consistent w o r d order to express most semantic relations, even when English does allow some variation i n the ordering o f semantic roles, for example i n the case o f possessor and pos­ sessed, or attribute and object (Braine, 1976). But it is questionable whether many o f the variations i n w o r d order that do appear i n early utterances should be considered as " e r r o r s " since English does allow alternative orders. I t is rare that an appropriate English gloss cannot be derived for a child's t w o - w o r d utterance, regardless o f the ordering o f the words. Nevertheless, some variation i n the w o r d

62

De Villiers a n d De Villiers

order o f early utterances has been reported, especially i n the early stages o f the emergence o f a semantic relation. Braine (1976) referred to " g r o p i n g patterns" of variable w o r d order before the child fixed on a consistent ordering to express a semantic notion. Thus Gregory produced Gregory fix it and Fix it Gregory; Fall down rabbit and Rabbit fall down. A n d for many weeks Bowerman's daughter Eva observed consistent subject-verb and possessor-possessed ordering only when / was the subject and my was the possessor. Other possessives and agentobject relations occurred i n variable order. English-speaking children may differ from each other i n the prevalence o f variable w o r d order i n their speech. Ramer (1976) found that the three children i n her study who acquired syntax more slowly made almost no errors i n w o r d order, but the four faster developing children used odd w o r d order i n 3 to 4 % o f their m u l t i w o r d utterances. She suggested that the faster developing children were more likely to experiment and take risks before they were sure o f a rule (see p . 64 for further discussion o f Ramer's study and her hypothesis about styles o f syntax acquisition). A third k i n d o f error that shows up i n early syntactic development involves inappropriate segmentation o f the input so that functors are incorporated into the word that precedes them. C o m m o n examples are the presence o f -a or -s i n words like this-a, have-a, get-a, that's or it's. Sometimes the chunks are larger, adding a neutral pronoun like it to a verb; as i n get-it or have-it. They reveal themselves as "prefabricated routines" ( B r o w n , 1973) rather than productive uses when utterances like This-a Bonnie pants, Have-a pants, Mommy get-it ladder, Have-it juice, and It's went appear. R . Clark (1974) reports quite substantial segments o f previous parental utterances being produced as unanalyzed wholes b y a child. Many examples o f rote-learned routines are discussed b y M a c W h i n n e y (1982,1985). 4 . 7 . Individual

Differences

M a n y researchers have commented o n the presence o f semantically limited pivot patterns i n the early m u l t i w o r d utterances o f some children, but other children do not seem to produce as many o f them. I n accounting for this discrep­ ancy, B l o o m (1973) contrasts t w o approaches to the acquisition o f syntax—a " p i v o t a l " approach and a " c a t e g o r i c a l " approach. M o s t o f the early t w o - w o r d utterances o f children who adopt the pivotal strategy use a small set o f words that express a constant functional relationship to the words they combine w i t h . Exam­ ples o f these include more + X to express recurrence, my + X to encode possession, and this 4- X to draw attention to an object. Children who adopt a categorical strategy, o n the other hand, produce sentences i n w h i c h the gram­ matical or semantic relations between the words do not depend o n the lexical items used. Thus a noun like Mommy may not only enter into an A G E N T + A C T I O N relation but also be used as a possessor i n a POSSESSOR + POSSESSED

pattern. B l o o m et al. (1975) expanded on B l o o m ' s original observation by noting

1.

The A c q u i s i t i o n of English

63

that Peter and Eric, children w h o initially followed the pivotal strategy, also used pronouns and other proforms i n place o f nouns to express semantic roles. A l ­ though they knew many o f the names o f the objects or people referred to, and used those words i n single w o r d utterances, i n w o r d combinations they tended to us it, this one or that for patients, my for possession, and here or there for location. They concluded that the grammar the t w o boys were learning "consist­ ed o f relations between different verb forms and a number o f constant functional f o r m s " (p. 19). W h e n later semantic relations entered into their speech, they began w i t h the same approach as they had for earlier relations, using proforms to combine w i t h content words. I n contrast, Gia and Kathryn followed the cate­ gorical strategy, and their early sentences contained far fewer pronouns. Seman­ tic roles i n their speech were filled by different nouns. Over a period o f some months the t w o groups o f children began to look more alike, until at an M L U o f about 2.5 morphemes their expression o f the major semantic relations was very similar. B l o o m et al. proposed that " c h i l d r e n can break into the adult linguistic code i n one o f (at least) t w o ways: w i t h a system o f formal markers, or w i t h a system o f rules for deriving grammatical categories" (p. 35). Nelson (1975) also found that some children used few pronouns i n their early constructions, while others used more o f a balance o f nouns and pronouns. She traced this to differences i n their approach to the language i n the one-word stage. Children w h o m she classified as "referential"—because names for objects pre­ dominated i n their early vocabulary—used a much higher proportion o f nouns than pronouns i n early m u l t i w o r d utterances. " E x p r e s s i v e " children, so termed because their early vocabulary was made up primarily o f verbs and socialexpressive words like hi, more, allgone, used a more evenly balanced proportion of nouns and pronouns i n w o r d combinations. U n l i k e B l o o m et a l . , Nelson d i d not observe children whose early constructions were marked by almost exclusive use o f proforms i n major semantic roles, possibly because she averaged pronoun usage across her groups o f children and possibly because some o f her speech samples were close to the M L U o f 2.5 by w h i c h B l o o m et al. reported the disappearance o f differences i n pronoun use (Bowerman, 1978). Both o f these studies tried to classify children into one o f t w o major types o f developmental pattern. Nelson clearly regarded them as points along a con­ tinuum, but B l o o m et al. went so far as to suggest that the children " w e r e learning t w o different systems o f semantic-syntactic structure that were virtually mutually exclusive i n the b e g i n n i n g " (p. 20). I n a review o f B l o o m et al. (1975), Maratsos (1975) argues that by mid-Stage I there is evidence for no more than a preference for pronoun use rather than the exclusive use o f one system over another. A t M L U 1.69 Eric used pronouns for patients 32 times (e.g. Get it) and nouns 33 times (Get ball). B y comparison Kathryn at M L U 1.89 used 81 nouns and 41 pronouns for patients. Maratsos concluded that when pronoun use was really dominant i n Eric's and Peter's speech the relevant semantic relations were marginally productive. Bowerman's (1978) study o f her o w n daughter Eva em-

64

De Vllliers and De Villiers

phasizes the difficulty i n characterizing different developmental strategies. Eva was classified as a " r e f e r e n t i a l " speaker on the basis o f her first 50 vocabulary items. However, like the "expressive" speakers i n Nelson's study she used a pivotal strategy i n her early t w o - w o r d utterances, w i t h many semantically limited formulae like want + X, more + X, and here + X, but she d i d not later depend on pronominal forms i n fixed syntactic frames. A t 18 months almost all V E R B + OBJECT constructions used nouns as objects, w i t h it only beginning to substitute for nouns around 20 months; but SUBJECT + V E R B and possessive constructions used both nouns and pronouns as soon as they appeared at around 19 months. The exclusivity o f B l o o m et al. 's strategies is therefore questionable, though they may capture a dimension along w h i c h children differ. Another possible dimension underlying individual differences was described by Ramer (1976). She followed seven children from when they began to combine words until 20% o f their m u l t i w o r d utterances were subject-verb-complement constructions. The children were divided into slower learners (the three boys) and faster learners (the four girls) on the basis o f how many months i t took from the onset o f w o r d combinations until the 20% subject-verb-complement criterion was reached. The slower learners took from 6V2 to 9 months, the faster learners took Wi to AVi months, so none o f the children were particularly slow learners. The syntactic development o f the t w o groups o f children differed i n several notable respects. Slower learners used many more d u m m y forms like a + N, reduplications o f the same w o r d , or combinations o f meaningful words w i t h empty forms for w h i c h no lexical meaning could be established. They also acquired V E R B + C O M P L E M E N T constructions w e l l before SUBJECT + V E R B and

SUBJECT + C O M P L E M E N T forms. For the faster learners all three o f these gram­ matical relations seemed to be productive from the beginning. Finally, the faster learners were freer w i t h their use o f w o r d order, w i t h three to four percent o f their utterances deviating from the expected English order. Deviations from expected w o r d order almost never occurred i n the slower learners. Ramer sug­ gests that the variability i n w o r d order represents a greater predilection on the part o f the faster learners to take risks w i t h syntactic rules they had not yet completely mastered. Though speculative, such a difference i n cognitive (or personality?) style is an interesting notion. What is needed is more detailed (and longer) study o f i n d i ­ vidual children i n order to establish i f such a tendency is found i n a particular child's acquisition o f several syntactic rules, or i n other domains besides lan­ guage acquisition. F r o m a grand sample size o f one we can report that our o w n son showed very systematic errors i n several aspects o f language acquisition— from overextensions i n early w o r d use, the creation o f expected but unacceptable verb forms (ham from hammer and hoove from hoover, after the model o f mow from mower and run from runner) and inappropriate causatives (He fell me down), to very persistent use o f overregularized -ed past tenses.

1.

The A c q u i s i t i o n of English

65

Thus while individual children may differ i n their reliance on pivot patterns for their initial t w o - w o r d utterances, i n the proportion o f nouns and pronouns that they use, and i n the variability o f their w o r d order, a great deal more w o r k is needed to establish the existence or absence o f consistent or coherent styles o f language acquisition. 4.8.

Pragmatics

Most o f the research on the relationship between pragmatic factors and early language acquisition has concentrated on prelinguistic interaction and the emergence o f words (e.g. Bruner, 1975,1977; Bates, 1976; Bates et a l . , 1977,1979). But some writers have noted effects o f pragmatic considerations on the prevalence o f different semantic notions i n t w o - w o r d utterances and on the ordering or differential stress o f elements i n those utterances. For example, the context i n w h i c h the conversation takes place is going to influence the frequency of particular relations. The mother o f B r o w n ' s subject Sarah spent a great deal o f time naming or eliciting names for objects from her child. Sarah therefore shows relatively frequent use o f forms expressing nomination ( B r o w n , 1973). Lieven (1978) reports that one o f her children seemed preoccupied w i t h attracting her mother's attention, so notice, nomination, and recurrence were frequent seman­ tic notions; the other c h i l d was much more concerned w i t h exploring the environ­ ment and produced far more utterances encoding locative action or attribution. Children i n constant interaction w i t h siblings o f a similar age are likely to d w e l l on possession (see Schaerlaekens (1973) for a study o f Dutch triplets). Finally, even very rare semantic notions can be increased dramatically i n frequency i n an appropriate pragmatic context—for example, increases i n reference to created objects where children are playing w i t h play-dough (Retherford et a l . , 1981). However, while these observations suggest that the relative frequency o f seman­ tic notions i n conversation varies w i t h pragmatic constraints, they do not show that pragmatic factors affect the order o f emergence o f those notions i n children's speech. Bates and M a c W h i n n e y (1982) provide a detailed analysis o f the relationship between the topic-comment function i n language and grammatical devices like surface subject positioning, pronominalization, ellipsis, and prosodic stress. Se­ lection o f topics and comments depends on such factors as the givenness or newness o f information (topics tend to be given information and comments provide new information, though the correlation is not exact), and the salience o f different elements o f the proposition. I n spontaneous speech and elicited produc­ tion (picture description) studies i n w h i c h given and new information was manip­ ulated, M a c W h i n n e y and Bates (1978) investigated the use o f w o r d order and other syntactic devices to mark topic and comment i n Italian, Hungarian, and English-speaking children. Though Italian children began by placing comments ( " n e w " information) i n initial position, most English children at the correspond-

De Villiers and De Villiers

66

ing stage o f development seemed to place agents or actors ( B r o w n , 1973) at the beginning o f agent-action strings regardless o f pragmatic role. The English c h i l ­ dren initially used contrastive stress to mark new information (Wieman, 1976), and ellipsis and pronominalization were used to indicate given information (Mac­ Whinney & Bates, 1978) Nelson (1978) has suggested that individual differences i n children's initial orientation to the communicative functions o f language may also have implica­ tions for the pragmatic uses o f early t w o - and three-word utterances. She ana­ lyzed the first ten multiterm utterances recorded from t w o children, one o f w h o m had been categorized as extremely referential (object oriented) and the other o f w h o m was categorized as extremely expressive (a social, interpersonal orienta­ tion) on the basis o f early vocabulary. I n terms o f Halliday's (1975) functional categories, the early constructions o f the referential child were primarily avail­ able for mathetic (or ideational) functions o f speech, those related to learning, k n o w i n g and informing. The expressive child's utterances were primarily prag­ matic and interpersonal i n function, regulating social interaction between the interlocutors. The expressive child also seemed to have a wider variety o f seman­ tic relations available for expression. Hence the communicative context i n w h i c h the child's early language is acquired and possible cognitive or personality char­ acteristics o f the child and parents may also influence the pattern o f acquisition.

5. G r a m m a t i c a l M o r p h e m e s 5.1.

Brown's

Fourteen

Morphemes

In the period when M L U is between 2.0 and 2.5 ( B r o w n ' s Stage I I ) another major syntactic development begins. Function words and inflections that do not seem to have independent meaning on their o w n but i n B r o w n ' s felicitous phrase " m o d u l a t e " the meaning o f the utterance now begin to appear. These include a few prepositions, articles, plural and possessive inflections on nouns, inflections marking progressive, past, and third person present on the verb, and the occa­ sional copula form o f be. I n B r o w n ' s (1973) words, " A l l these, like an intricate sort o f i v y , begin to grow up between and upon the major construction blocks, the nouns and verbs, to w h i c h Stage I is largely l i m i t e d " (p. 289). B r o w n points out that they only " s p r o u t " i n Stage I I , however, and the acquisition o f many o f them is not complete by Stage V , when M L U exceeds 4.0 morphemes. B r o w n chose to focus on 14 grammatical morphemes for w h i c h obligatory contexts could be identified, that is, contexts i n w h i c h the morpheme w o u l d be obligatory for an adult speaking English. These were the present progressive inflection -ing; the third person singular, regular -s and irregular (e.g. has); past tense, both regular -ed and irregular (e.g. went or fell); the copula and auxiliary be; t w o noun inflections, the regular plural and possessive -s; and t w o preposi­ tions, in and on. M o s t o f these are quite frequent i n adult English and appear i n a

1.

The A c q u i s i t i o n of English

67

variety o f lexical or syntactic settings (the exceptions being the irregular forms o f the past and third person present that must be learned by rote). The obligatory contexts for each morpheme were identified on the basis o f the linguistic context, both the child's utterance itself and the preceding or f o l l o w i n g adult utterances, or the non-linguistic context. Thus a linguistic obligatory con­ text for the plural could be a sentence i n which a plural quantifier was used, two " or "some " ; or one i n which the number o f the subject con­ strained the number o f the predicate, "Those are . " A non-linguistic context 44

that required a plural noun might consist o f the child requesting items that normally occur i n the plural, e.g. shoes, or referring to a set o f identical objects, e.g. blocks. Occasionally it is impossible to tell which morpheme is missing, especially when the child uses unmarked verbs, but these cases were infrequent in B r o w n ' s longitudinal sample (1973) and de Villiers and de Villiers crosssectional sample (1973b), and very few data had to be discarded. The use o f obligatory contexts has obvious advantages over simply scoring the frequency or point o f first (or fifth, or whatever number) use o f a morpheme. Since the morpheme is required by the grammar, its presence or absence indicates what the child is able to say rather than what he chooses to say. Topic o f conversation affects the frequency o f contexts, but not the accuracy o f supplying the mor­ phemes. B r o w n took a£ his criterion for acquisition the presence o f a morpheme in 90% o f its obligatory contexts i n three successive 2-hour-long speech samples from a child, the samples being collected approximately bi-weekly. T w o major findings emerged from B r o w n ' s longitudinal study o f A d a m , Eve, and Sarah. First, the morphemes were gradually acquired. Even i n the same linguistic context they d i d not rapidly go from never being supplied to always being supplied. For some morphemes, the time from first appearance to 90% accuracy was as long as a year. Second, although there was some individual variation i n the chronological rate o f acquisition o f particular morphemes, the overall order o f acquisition o f the morphemes was remarkably similar across the three children. 5.2.

Order

of

Acquisition

B r o w n ranked the morphemes i n the order i n which they reached the 90% criterion (the first o f the three successive samples) for each child. Morphemes that had not reached 90% accuracy by Stage V for a particular child were ranked in terms o f the accuracy w i t h w h i c h they were supplied at that point. The rank order correlation between orders o f acquisition for A d a m and Sarah was + .88, between A d a m and Eve + . 8 6 , and between Eve and Sarah + . 8 7 . The average rank-ordering is given i n Table 1.3. A very similar order o f acquisition was reported by de Villiers and de Villiers i n their cross-sectional study o f the same morphemes. Whether the morphemes were ranked according to the lowest M L U sample at w h i c h each o f them first appeared i n 90% or more obligatory contexts

68

De Villiers and De Villiers TABLE 1.3 Rank Order Correlation Between Orders of Acquisition

The 14 Grammatical Morphemes Present Progressive on in Plural Past Irregular Possessive Uncontractible Copula Articles Past Regular Third Person Regular Third Person Irregular Uncontractible Auxiliary Contractible Copula Contractible Auxiliary

Average Order of Acquisition from Brown (1973)

Order of Acquisition from de Villiers and de Villiers (1973b) (Method I)

1 2.5 2.5 4 5 6 7 8 9 10 11 12 13 14

2 2 4 2 5 7 12 6 10.5 10.5 8.5 14 8.5 13

(Method I ) , or i n terms o f the average accuracy o f each morpheme across all o f the children (Method I I ) , the resulting ordering was highly correlated w i t h Brown's mean ranking (Rho = + .84 w i t h Method I and + .78 w i t h M e t h o d I I ) . B r o w n explored several possible determinants o f the order o f acquisition. First, differences among the allomorphs o f some inflections are found i n elicited production studies w i t h older children (e.g. Berko, 1958; Bryant & A n i s f e l d , 1969). One class o f allomorph is governed by a simple phonological rule o f voicing assimilation: the voiceless allomorph is attached to a stem w i t h a voice­ less terminal consonant (e.g. bik-biks) and the voiced allomorph is attached to a voiced terminal consonant (e.g. wug-wugs). Allomorphs f o l l o w i n g this rule are significantly easier for children on elicited production tasks than other al­ lomorphs. A second class o f stems ending i n III, Iml, Inl, / r / , or any v o w e l , requires the voiced allomorph but B r o w n argues that the rule is not a pho­ nological necessity since there are English words w i t h voiceless terminal conso­ nants f o l l o w i n g the same stems (e.g. her-hers, but see hearse). These al­ lomorphs are i n turn supplied more readily by children than the third class o f stems, those ending i n the same consonant as the inflection, or a closely similar one. This type o f stem requires a v o w e l , lil, before the voiced allomorph (e.g. gutch-gutches). B r o w n did not score allomorphs o f morphemes separately be­ cause o f insufficient frequency i n some cases and uncertain transcription i n others. But phonological factors cannot be a major determinant o f the ordering o f the 14 morphemes as a group. The plural, possessive, third person present

1.

The A c q u i s i t i o n of English

69

singular, and some forms o f the contractible copula and auxiliary be share the same phonological f o r m , but they are mastered at very different times. T w o different factors were highly correlated w i t h the order o f acquisition. B r o w n calculated the syntactic complexity o f the morphemes ( w i t h the exception of the possessive) i n terms o f the number o f transformations involved i n their derivation i n a representative transformational grammar o f English (Jacobs & Rosenbaum, 1968). He also considered several authors' characterizations o f the meanings o f the morphemes and determined the number o f unitary meanings involved i n their correct use by the children. Rank order correlations between syntactic and semantic complexity and the orders o f acquisition determined by B r o w n and de Villiers and de Villiers were over + . 8 . However, assessing complexity i n terms o f the sheer number o f transformations or dimensions o f meaning involved i n a morpheme's use is a dubious procedure since i t presumes that each transformation or meaning adds a constant amount o f complexity. B r o w n therefore suggested that cumulative complexity be used as an index o f grammatical or semantic difficulty. I n this metric a construction i n v o l v i n g rule X would be easier than one i n v o l v i n g X + Y , which w o u l d i n turn be simpler than one i n v o l v i n g X + Y + Z . The predictions made by partially ordering the morphemes on the basis o f cumulative syntactic and semantic complexity were confirmed i n both studies. The major discrepancy between the t w o studies was i n the order o f acquisition of the contractible and uncontractible forms o f the copula and auxiliary. B r o w n found the uncontractible forms were better supplied than their contractible equiv­ alents by all o f these children. I n our study we generally found the opposite, although only the copula forms could be compared w i t h i n children since none o f the subjects reached a sampling criterion o f five or more contexts for both the contractible and uncontractible auxiliary. Kuczaj (1979) suggested that the dif­ ference could result from the relative frequency o f yes/no questions versus w h questions and declaratives i n the samples from the t w o studies. I n his o w n study Kuczaj found no reliable difference i n order o f acquisition o f the t w o forms, but uncontractible forms o f be were more consistently supplied i n some linguistic contexts than i n others. 5.3.

Cognitive

Complexity

B r o w n (1973) and de Villiers and de Villiers (1973) concluded that the rela­ tive contributions o f syntactic and semantic complexity could not be teased apart, since the t w o variables made almost identical predictions. Using a more sophisitcated multiple regression procedure, B l o c k and Kessell (1980) discovered that each o f the variables accounted for a significant percentage o f the variance i n the order o f acquisition o f the morphemes when it was entered into the regression analysis first. However, adding the other variable d i d not add significantly to the variance accounted for. A commonality analysis revealed that most o f the vari­ ance was common to both syntax and semantics, and each alone accounted for

70

De Villiers and De Villiers

very little extra variance. They suggest that this is because a single factor under­ lies both syntactic and semantic complexity, namely cognitive complexity. As Pinker (1981) has pointed out, however, the postulation o f such a third underly­ ing factor, while not forbidden by their analysis, is by no means implied by i t either, especially since they give no good ancillary arguments as to how such a factor should be characterized. Indeed, they conclude that there are " a number of ways to define the underlying construct implied by this reanalysis. I n terms o f Piagetian sensorimotor intelligence as developed by Bates (1976); i n terms o f communicative functions and speech acts as developed by Bruner (1975) and Dore (1975); i n terms o f cognitive and perceptual strategies as developed by Clark (1977); and so o n " (p. 188). But B l o c k and Kessell give no indication o f how the different morphemes could be characterized according to any o f these approaches. Finally, we might add that an account i n terms o f cognitive com­ plexity, as distinct from the semantic complexity described by B r o w n , might be expected to predict the order o f emergence o f obligatory contexts, i.e. reference to the notions underlying the morpheme, rather than the syntactic realization o f these notions i n English. Crosslinguistic studies o f the acquisition o f inflections and other grammatical morphemes are more l i k e l y to reveal the factors that determine an order o f acquisition (see Johnston & Slobin (1979) for an analysis of the linguistic and conceptual factors that contribute to the order o f acquisition of locative expressions i n English, Italian, Serbo-Croatian and Turkish). 5.4. Relation

to

Input

B r o w n calculated the relative frequency o f the 14 morphemes i n the speech o f the parents to their children immediately preceding Stage I I I . The rank orders o f the frequencies were very similar across the three parental pairs, suggesting that this frequency profile was quite stable. However, when the average order o f acquisition was correlated w i t h their average order o f frequency i n the parental input, rho was calculated as + . 2 6 , a nonsignificant value. F r o m this general analysis and more specific analyses o f prepositional phrases and allomorphs o f the contractible copula, B r o w n concludes that the frequency o f the morphemes i n the input to the child is not a significant determinant o f the order o f acquisition. More recently, however, a couple o f studies have claimed to find effects o f input frequency on use o f some o f the morphemes. Newport et al. (1975) found that the children o f parents who used frequent deictic expressions like " T h a t ' s a d o g " tended to produce more plural inflections per noun phrase. But they scored changes i n the frequency o f the inflections, not their presence and absence i n obligatory contexts. M o e r k (1980) reanalyzed some o f B r o w n ' s data. He criticized B r o w n ' s use o f a single parental frequency profile taken some months or years before the c h i l ­ dren mastered many o f the morphemes. He argued that the input frequencies for particular morphemes may change even w i t h i n a single recording session and are bound to change over a period o f months as the child matures. Furthermore, the

1.

The A c q u i s i t i o n of English

71

effect o f the input may depend on the child's cognitive and linguistic level. Therefore, M o e r k selected out just those morphemes acquired during Stage I I for A d a m and Sarah and Stages I I and I I I for Eve (because she went through the stages so q u i c k l y ) . These morphemes were acquired relatively soon after the parental sample was taken. He was left w i t h the present progressive, in, on, and the plural for A d a m ; w i t h the same four plus the possessive and past irregular for Sarah; and w i t h those four plus the possessive and past regular for Eve. For each child he correlated the approximate age ( i n months) at w h i c h the morphemes reached the 90% criterion w i t h the absolute frequency o f those morphemes i n their respective parents' speech. For all three children M o e r k reported sizable negative correlations (Pearson's r) between input frequency and the age at w h i c h a morpheme was mastered. They varied from - .56 for A d a m to - .76 for Eve. Although none o f these correlations were significant w i t h an N as small as four or six, M o e r k was impressed by the size o f the effects and the agreement across children. Sarah was slower than the other t w o children i n acquiring the mor­ phemes, and her parents also used them relatively less frequently i n speech to her. M o e r k found a substantial correlation between how impoverished Sarah's input was w i t h respect to particular morphemes and how far she was behind the other t w o children ( i n months) i n acquiring them. Finally, M o e r k noted that the frequency w i t h w h i c h Eve used particular prepositional phrases w i t h in was related to their frequency i n her parents' speech. He concluded that the rela­ tionship between input frequency and the acquisition o f the grammatical mor­ phemes is worth another look, and that B r o w n was wrong when he dismissed i t as an insignificant variable. M o e r k ' s reanalysis o f B r o w n ' s data has been strongly criticized by Pinker (1981). I n our view Pinker's most effective arguments concern the selection o f the morphemes that were correlated w i t h parental frequency. I n particular, Moerk leaves out the irregular past tense that was acquired by Sarah i n Stage I I "Since it appears i n many and widely differing f o r m s " (p. 109) and presumably each form has to be learned by rote. But i f i t is included, the correlation between frequency and Sarah's order o f acquisition drops from - .66 to - . 4 8 , and the percentage o f variance accounted for decreases from 4 3 % to 23%. Similarly, Moerk included t w o constructions acquired i n Stage I I I by Eve on the grounds that she passed through the stages faster than the other children, but Pinker notes that i f the correlation used only morphemes acquired i n Stage I I , the percentage of the variance accounted for drops from 57% to 9%. W e think M o e r k ' s reasons for including and excluding morphemes are plausible, but the extent to w h i c h the correlations change is alarming. M o e r k also ends up w i t h a rather strange subset of morphemes, t w o prepositions that are much more like lexical items, plus one or t w o noun and verb inflections. The primary reason w h y the overall correlation performed by B r o w n was nonsignificant was because the t w o most frequent morphemes, the articles and the contractible copula, were somewhat late ac­ quired, the articles i n the middle o f the order and the copula towards the end.

72

De Villiers a n d De Villiers

These morphemes are always the most frequent i n the parental input and do not change substantially i n frequency between Stage I I and Stage V , yet they d i d not enter into M o e r k ' s correlations. There are also problems w i t h M o e r k ' s analysis of Sarah's comparative linguistic deprivation and her relative delay i n acquiring those morphemes. A l t h o u g h A d a m and Eve's inputs are the most similar i n frequency, w i t h several morphemes actually being more frequent i n the input to A d a m , A d a m was much more like Sarah i n his rate o f acquisition o f the mor­ phemes than he was like Eve. He varied from 9 to 15 months behind Eve i n mastering the morphemes, but only 4 to 6 months ahead o f Sarah. I f Sarah's slower development is tied to the input frequency, what do we make o f A d a m ' s delay relative to Eve? Finally, M o e r k and Pinker differ on the significance o f the data on the acquisition o f in (see also M o e r k ' s [1981] reply to Pinker). M o e r k made much o f the difference i n frequency o f the phrases in a minute and in a while i n Eve's speech. Despite their semantic similarity, in a minute appeared frequently i n both the input and Eve's speech; in a while was modeled by the parents only once i n the early samples and on the few occasions on w h i c h Eve produced i t , the in was missing. Pinker noted that frequent in phrases like in a minute may have been learned as idiomatic routines. Furthermore, 19 phrases w i t h in were used correctly by Eve even though they never appeared i n any sample o f the input and were presumably rare. However, there are other suggestive data on input frequency and the order o f acquisition o f a linguistic subset o f B r o w n ' s morphemes. M o e r k cites a study by Forner (1977) that found extremely high correlations between the order o f ac­ quisition o f bound morphemes and their frequency i n parental speech. Table 1.4 shows this pattern for the five bound noun and verb inflections. I t also gives the rank order correlations relating each child's order o f mastery o f these morphemes w i t h the ranked input frequency. The concordance (Kendall's W ) between the children on order o f acquisition is .89 (p < .01); between the parents on ranked frequency o f use i t is .82 (p < .025). Finally, the average ranked frequency o f the morphemes i n the input correlates almost perfectly w i t h the average order o f acquisition across the three children i n B r o w n ' s study (Rho = + . 9 0 , p < .05), as w e l l as w i t h the acquisition order o f these morphemes i n de Villiers and de V i l l i e r s ' (1973) cross-sectional study (Rho = + . 9 7 5 , p < .05). It is interesting to note that B r o w n ' s measure o f cumulative syntactic com­ plexity makes no predictions about the order o f acquisition o f these forms (only four o f them are represented i n the Jacobs and Rosenbaum grammar). Cumulative semantic complexity only orders the third person regular, predicting that it should follow the plural and the past regular—the first o f these predictions is supported, the second is not for A d a m and Sarah. M o e r k suggests that the bound morphemes add semantic information, but the auxiliary and copula forms o f he, and the articles, for w h i c h input frequency seems to have no effect, add only m i n i m a l information. B u t this is questionable, especially for the articles. A far more likely explanation o f the later acquisition o f articles concerns their

1.

T h e A c q u i s i t i o n of English

73

TABLE 1.4 Pattern f o r t h e Five B o u n d N o u n a n d Verb Inflections

Morphemes in Order of Acquisition Present Progressive Plural Possessive 3rd person regular Past regular

Eve

Sarah

Adam

Parental Frequency 65 57 25 25 28

(1) (2) (4.5) (4.5) (3)

Morphemes in Order of Acquisition

Parental Frequency

Plural Present Progressive Possessive 3rd person regular Past regular Rho = +.90 (p < .05)

Rho = +.675 (n.s.)

57 28 16 19 9

(1) (2) (4) (3) (5)

Morphemes in Order of Acquisition

Parental Frequency

Present Progressive I Plural ( Possessive Past regular 3rd person regular

67 33 30 7 7

(1) (2) (3) (4.5) (4.5)

Rho = +.95 (p < .05)

cognitive complexity since they require some appreciation o f the knowledge o f the listener (Maratsos, 1976). I n short, B r o w n dismissed parental input as a major variable determining the acquisition o f the grammatical morphemes, but for at least some classes o f morphemes that seems to have been premature. (See also p . 74 for an inverse relationship between the input frequency o f different irregular verbs and the likelihood that they w i l l be regularized by preschoolers.) 5.5. Typical

Errors

The prototypical error i n the acquisition o f English inflections is the overregularization o f plurals and past tenses. I n each case, when the regular inflection begins to be mastered it is overgeneralized to irregular forms, resulting i n errors like foots, sheeps, goed, and eated ( B r o w n , 1973; Cazden, 1968; E r v i n , 1964; Kuczaj, 1977; Slobin, 1971). I n the case o f the past tense, children usually begin by correctly using a few irregular forms like fell and broke, perhaps because these forms are frequent i n the input and the child learns them by rote. A t first they may not be fully analyzed i n the sense o f being syntactically related to their corresponding generic forms (Kuczaj, 1977). But as soon as the child begins to produce regular past tense endings, they apply that form more widely than they should, and irregular forms that were previously correctly produced may now be regularized (e.g. failed and breaked). E r v i n (1964) and Slobin (1971) even report cases i n w h i c h forms like corned, buyed, and doed appeared before any correct regular past tenses were produced. T w o kinds o f past-tense overgeneralization errors are observed: one i n w h i c h the -ed ending is attached to the generic form o f the irregular verb (e.g. eat—eated), and another i n which it is added to the irregular past form o f the verb (e.g. went—wented). Kuczaj (1977, 1978a) showed that there are age differences i n the relative frequency o f the t w o types o f errors, w i t h the double-marked past generally being more common and judged acceptable i n older children (around 5 years o l d ) . He suggests that the difference

74

De Villiers and De Villiers

reflects an increasing emphasis on syntactic regularity as the child gets older, so that the child w i l l add the regular past marker even to verbs he has analyzed as expressing earlierness. Slobin (1971,1973) made the strong prediction that overgeneralized past tenses for a time completely replace irregular pasts. This was derived from an operating principle stating that there is a universal preference not to mark a semantic category by a zero morpheme. However, Kuczaj (1977) reports that some children produce forms such as goed, went, and wented at the same point i n development. A n d some children do not overgeneralize the -ed ending to all irregular verbs. O f the 14 children i n Kuczaj's cross-sectional sample, eight never applied -ed to irregular verbs like hit and cut that have the same present and past form, even though they had many opportunities to do so. Thus overgeneralization does not seem to be an all-or-none process at any point i n development. Frequency o f usage i n the input to the child appears to be an important determinant o f w h i c h irregular verbs w i l l be overregularized. I n an analysis o f past-tense overregularization errors i n several sets o f spontaneous speech sam­ ples from preschool children, Bybee and Slobin (1982) found a significant nega­ tive rank order correlation (—.67) between the number o f times the adult care­ givers used irregular verb forms and the number o f times those forms were overregularized by the children. This finding is i n keeping w i t h the notion that the irregular forms must i n large part be learned by rote from the input. Finally, Slobin (1971) suggested that irregular verbs i n w h i c h there is a systematic phonological change relating the generic to the past tense f o r m (e.g. lose—lost; make-made) w i l l be more resistant to overgeneralization since they are partially regular. Other irregular verbs that either make a dramatic change (e.g. go—went), change only the internal vowel (e.g. bite—bit; fall—fell), or make no change at all (e.g. hurt—hurt; hit—hit) w i l l be much more susceptible to the addition o f -ed. Kuczaj d i d find that overgeneralizations were most likely to occur w i t h verbs making only a v o w e l change, but for many children verbs that do not change at all were least likely to be overgeneralized. Recently, Bybee and Slobin (1982) have suggested that i n addition to the rotelearning o f irregular past-tense forms, young children form phonological gener­ alizations or schemas about them. These are not strictly phonological rules that relate base to derived forms, but are somewhat looser statement about similarities between the phonological properties o f past tenses that help i n organizing and accessing the lexicon. A n example o f such a schema w o u l d be the statement that past-tense verbs typically end i n [t] or [ d ] . I n a sentence-completion task that elicited past-tense forms, Bybee and Slobin observed that preschoolers systemat­ ically failed to add the regular -ed suffix to verbs already ending i n [t] or [ d ] , even i f they were regular verbs like melt or pat. The application o f such a schema about past tense ending i n [t] or [d] accounts for Kuczaj's (1977, 1978a) finding that verbs that do not change at all i n the past tense are most resistant to overregularization, since those verbs end i n [t] or [ d ] . Verbs making only a

1.

T h e A c q u i s i t i o n of English

75

vowel change typically do not end i n [t] or [ d ] , so are much more likely to be overregularized. The preschoolers thus seem to be w o r k i n g w i t h both a regular suffixation rule to form the past tense and a schema for the phonological proper­ ties o f past forms. Similar observations were made by Berko (1958) for both plural and past-tense formations i n her elicited production tasks. Preschoolers and first-grade children rarely added the plural suffix to nonsense syllable names ending i n silibants, perhaps because these forms f i t a schema about the pho­ nological properties o f plurals. The child's problem i n mastering the past tense seems to be i n relating the present and past forms o f an irregular verb to each other, and i n learning that the irregular past is the only past tense form for that verb (Kuczaj, 1977,1981). Even when the child seems to understand that a verb like went is used only i n the past, he may also produce goed as the past for go, and even wented as an overmarked past. The learning process is complicated by the fact that parents rarely correct ungrammatical past tenses so the child has to learn from the mismatch between his o w n production and the speech o f the adults he hears, a still more difficult task as his peers are also overgeneralizing the regular past tense. Furthermore, the child has to learn the relation between irregular pasts and their generic forms in a rather piecemeal fashion, since there are few regularities, so overgeneralization errors w i t h -ed only, disappear over a period o f years. 5.6.

Relation

to

Semantics

T w o interesting interactions between semantics and the acquisition o f mor­ phemes have recently been proposed. Leonard (1976) scored the first emergence of the articles, copula and present progressive i n utterances expressing different underlying semantic notions. The articles, w h i c h can appear i n the widest range of semantic relations, were first provided i n utterances reflecting Leonard's earlier emerging group o f semantic notions (cf. Table 1.2). Similar findings (though w i t h a much smaller number o f observations) were noted for the copula and progressive. The copula tended to appear first i n utterances expressing nomination, negation and recurrence, rather than i n utterances expressing lateremerging semantic notions like attribution, location, or possessor. The present progressive emerged i n utterances expressing agents and actions before it was supplied i n EXPERIENCER + E X P E R I E N C E or instrumental semantic relations. These results are i n keeping w i t h the notion that there is less processing-load i n earlier-acquired semantic relations, allowing the child to apply a new mor­ phological rule. Unfortunately, Leonard recorded first emergence o f the mor­ phemes rather than percent supplied i n obligatory contexts, so his findings are confounded by the much greater frequency o f the earlier emerging semantic notions. I t w o u l d be interesting to examine the types o f semantic relations i n which the morphemes were first reliably supplied. B l o o m , Lifter and Hafitz (1980) studied four verb inflections—present pro­ gressive, third person singular, and regular and irregular past tense—as they emerged i n obligatory contexts i n the speech o f four children observed longitudi-

76

De Villiers and De Villiers

nally. They found that the early usage o f the inflections was dependent on the semantics o f the verbs to w h i c h they were attached. The progressive -ing was correctly supplied most often w i t h action verbs that have continuous duration and no clear result, e.g. singing. The past tense was used initially to inflect actions that are punctual and have a clear end result, e.g. closed. A n d the third person occurred most reliably w i t h verbs indicating the placement o f an object, e.g.//to, sits or goes. Since B l o o m et al. scored accuracy i n obligatory contexts, this result was independent o f the sheer frequency o f the morphemes or contexts—at the same time as a past tense was supplied for a verb like eat or close, i t w o u l d be omitted i n an obligatory context w i t h a verb like sing or walk. B l o o m et al. argue from this finding that the inflections mark aspect rather than tense when they first emerge. Progressive -ing marks imperfective action or duration o f an activity; the past tense marks perfective or punctate action. Bronckart and Sinclair (1973) made a similar, but even stronger claim about the present and past tense i n French-speaking children. They had their subjects describe different events enacted i n front o f them. L o n g duration events w i t h no clear conclusion tended to be described i n the present. Shorter, perfective actions were described i n the past tense. Bronckart and Sinclair conclude that the past encodes aspect rather than temporal earlierness until the child is 6 years o l d . However, Smith (1980) analyzed samples o f spontaneous speech (the children from B r o w n , 1973 and Kuczaj, 1978b), and repeated Bronckart and Sinclair's experiment w i t h English-speaking children. She found that even the youngest children ( M L U 2.94, age 29 months) used the past tense for verbs w i t h both perfective and imperfective aspect. The children i n the study o f B l o o m et al. (1980) were still far short o f B r o w n ' s criterion for mastery o f the morphemes. Even i n the most advanced stage o f development that was examined ( M L U 2 . 5 - 3 . 0 ) , the morphemes as a set were only supplied i n an average o f 54% o f the obligatory contexts, and three o f the four children had very little use o f the regular past. Thus, the use o f the past tense to mark perfective aspect may be a short-lived phenomenon. For example, at the time Kuczaj's son A b e was 31 months o l d ( M L U 3.5), Smith (1980) found that between 25 and 4 0 % o f his use o f the regular past was on imperfective verbs, even though he was not yet supplying the -ed inflection i n 90% o f obligatory contexts. Further research needs to explore how the inflections spread across verb types as the child approaches consistent use o f morphemes. A t a later point i n the acquisition o f the regular past, when the child begins to overgeneralize the use o f -ed, i t seems to be applied across all verb types, and the child ignores some semantic distinctions that he clearly makes w i t h respect to other inflections. Children acquiring American English do not generalize the use o f the progressive -ing inflection from action or process verbs like hit or break to state verbs like want, think, and know ( B r o w n , 1973; Kuczaj, 1981). I n the input they receive -ing only occurs on action and process verbs. I n their overgeneralizations o f -ed to irregular verbs, however, errors like knowed, thinked,

1.

The A c q u i s i t i o n of English

77

and seed are as common as forms like runned or breaked. N o r is it the case that overgeneralizations o f -ed begin w i t h action verbs and then spread to state verbs (Maratsos, Kuczaj, Fox & Chalkley, 1979). Thus, children observe semantic restrictions on inflection use at certain peri­ ods o f time and for particular inflections, but at other times there appear to be no semantic limitations. The manner i n w h i c h those semantic restrictions are loosened must still be specified. 6. M o d a l i t i e s o f t h e S i m p l e S e n t e n c e 6.1.

Auxiliaries

Fletcher (1979) provides a useful discussion o f the knowledge an Englishspeaking child must acquire o f the auxiliary system. There are many forms o f the auxiliary, they must appear i n certain positions relative to one another, and there are restrictions on co-occurrence. Chomsky's phrase structure rule (1957) was: Aux-»Tense (modal) (have+en) (be+ing). The three areas o f meaning expressed i n the verb phrase are time, aspect, and mood, though many o f the forms are plurifunctional, e.g. past-tense forms can refer to unreal events, as i n the conditional. Even for declaratives, then, the auxiliary system presents some complexity to the language learner, and data concerning that development w i l l be reviewed before turning to the modalities o f the simple sentence. 6 . 1 . 1 . Course of acquisition. Fletcher (1979) uses Leopold's diary data from Hildegard to describe the acquisition o f the verb phrase. Her first auxiliaries were won't and can't, at the age o f 2;0-2;2 years. She d i d not have the positive auxiliaries will or can i n her speech at that time. About a month later, she acquired will i n yes-no questions and responses to them, but never used the contracted form 7/. K l i m a and Bellugi (1966) report a similar phenomenon: although the mothers o f their three subjects used the contracted form '11 almost always, the children used will (see also Kuczaj & Maratsos (1975)). Hildegard used will i n sentences not dependent on a question w i t h i n about one more month, at 2;4. She also acquired / may at that time. B y 2;5, she was using the pro­ gressive form be+ing but i t was "variable i n use," and the details o f that acquisition have already been discussed ( B r o w n , 1973; see page 68). A t 2;6, Hildegard used will to refer to future, along w i t h the expression going to. Late acquisitions were the modals should, would, and could, and the perfect forms w i t h have (see also Cromer, 1968). Hildegard seems to have been a rapid learner, but the overall course o f her development concurs w i t h other accounts. Kuczaj and Maratsos (1975) used elicited imitation to explore one child's knowledge o f the auxiliary system before he used it productively i n spontaneous

78

De Villiers and De Villiers

speech. A m o n g other sentences, A b e was given declaratives to imitate that were either correct, or ungrammatical i n one o f t w o ways: the aux was misplaced, e.g. The slow turtle wash did his dirty hands. or a tensed verb followed the aux, e.g. The monkey's mommy will yelled at the lion. Abe's imitations showed a sensitivity to the rules o f aux placement although his spontaneous speech showed no auxiliary use. For instance, he imitated 38/48 o f grammatical declaratives correctly, failing primarily on sentences w i t h un­ stressed positive do. W i t h misplaced auxiliaries, he either deleted the aux or put it i n the correct sentence position (but never w i t h do). W i t h tensed verbs, he either de-tensed the verb or deleted the aux, both o f w h i c h w o u l d result i n a grammatical sentence. Before producing auxiliaries i n spontaneous speech, A b e seemed to k n o w that modal auxiliaries (except unstressed do) could occur before the main verb i n declaratives, and that the main verb must not be tensed. A t this point, Abe d i d not k n o w how to imitate yes/no questions correctly. I t is interesting that t w o months later, when he d i d show some ability to imitate aux i n yes/no questions, he also began using auxiliaries i n his spontaneous declaratives. Elicited imitation may taneous speech, but i t is children ( B l o o m , 1974). promise, however, so the

thus reveal knowledge that is not expressed i n spon­ not a technique that is equally informative w i t h all Comprehension studies o f auxiliary use hold little information base is not likely to become richer.

6.1.2. Typical Errors. Errors i n auxiliary use seem surprisingly rare. One would expect children not to k n o w co-occurrence restrictions, and so produce: He must will go. Or, the affixes might not be properly attached to the preceding aux, so one should hear: He have going. or He will eaten it. Tense could be attached to the main verb rather than the first aux, so the child should produce: He does ate it.

1.

The A c q u i s i t i o n of English

79

When double or triple auxs are produced, they could be scrambled: He have been might going there. These errors are not c o m m o n l y observed, however. One that has been reported is overmarking, e.g. I did broke it. I did rode my bike. Jenni did left with Daddy. (all from Mayer, Erreich, & Valian, 1978). The researchers argue that the error results when the child copies the tense marker to its correct position f o l l o w i n g the main verb, but fails to delete i t from its original position i n the aux. A stranded tense marker normally triggers ^ - i n s e r t i o n , and so the child ends up w i t h t w o tensed morphemes. The account rests on the assumption that the transformation of tense-hopping is composed o f t w o "basic operations"—copying and dele­ tion. Mayer et al. argue that the child misformulated the rule as consisting only of tense-copying, hence generating the errors. They make a similar claim about errors found i n aux-inversions, to be discussed later under "questions." Maratsos and Kuczaj (1978) argue against the transformational account, pointing out that many other errors o f this type should occur but do not. The overmarking error occurred i n their subjects primarily w i t h do, and only i n cases w i t h a main verb. One w o u l d expect to find mistakes such as: He did could have it. but no-one has reported them. Maratsos and Kuczaj contend that the error occurs because children have not w o r k e d out when do-support is necessary. They also report the overmarking i n negative utterances such as: It didn't broke, and It didn't disappeared. yet these are not readily explained by the basic operations hypothesis. A further possibility that they raise is that the error is more likely w i t h irregular past tenses, which may more closely resemble unmarked verbs or be misanalyzed as such by the child. Further data on overmarking errors are needed to decide among the alternative accounts, but the error does not seem to be a typical one, or to be as prevalent w i t h i n the grammar o f a child w h o does produce these errors, as one w o u l d expect from the basic operations hypothesis.

De Villiers and De Villiers

80

6.1.3. Relation to Input. The child's acquisition o f the auxiliary system seems to be one development i n which input variation makes a difference. Three different studies have found an impressive correlation between the mother's use of auxiliaries and the child's rate o f development i n auxiliaries per verb phrase. Newport et al. (1977) found that the number o f yes/no questions asked o f 15 1and 2-year-olds by their mothers predicted the children's change i n auxiliary use 6 months later. They argue that the salient initial position o f the auxiliary draws the child's attention to the form. I n the only convergence between the t w o studies, Furrow et al. (1979) also found a correlation o f similar magnitude ( + . 8 5 ) between mother's overall use o f yes/no questions to seven children o f M L U 1.0, and the children's o w n auxiliary use at 2;3 years. Unfortunately the convergence is less evident upon closer inspection. I n Furrow et al.'s study, the mother's use o f initial auxiliaries i n yes/no questions d i d not correlate signifi­ cantly w i t h aux development i n the child. Use o f " o t h e r " yes/no questions was the variable resulting i n the high correlation, and the majority o f these contained no auxiliary! Furrow et al. argue that a mother who uses both regular yes/no questions and these other forms that contain no auxiliary but have the same intonation, e.g. You want to go home? causes the child to correctly analyze the auxiliary when i t is present. I n both studies, however, a high use o f imperatives was negatively correlated w i t h auxiliary growth. Possibly, then, i t is the optionality o f the initial auxiliary, not its overall frequency, that enhances the child's understanding o f the forms. Nelson (1982) also found mothers w h o used verbs rich i n auxiliaries to their 22-month-olds had children who advanced rapidly i n auxiliary use from 22 to 27 months (r = + . 5 2 ) . Even more impressive was that the mother's initial use was not at all tied to the child's use at 22 months (r = — .03). Newport et al. (1977) however, found no significant relation between overall use o f auxiliaries by the mother and children's rate o f auxiliary development. Hence i n finer detail, the studies are not i n close agreement.

6.2.

Negatives

6 . 2 . 1 . Course of Acquisition. F o l l o w i n g detailed analysis o f the negative sentences produced by A d a m , Eve and Sarah, Bellugi (1967) described the stages o f development o f negative syntax. These stages, she argued, followed the stages i n the linguistic derivation o f a negative sentence ( K l i m a , 1964). The grammar postulated an initial stage when the deep structure o f a negative sen­ tence consisted o f a negative marker external to the sentence, and the earliest m u l t i w o r d utterances o f children seemed to be simple propositions preceded or followed by no or not.

1.

The A c q u i s i t i o n of English

81

In the second derivational step, the negative marker was placed adjacent to the verb stem, and i n the second stage o f negative development ( M L U 2.8 to 3.0) the three children used four negative elements internal to the sentence: no, not, can't, and don't. Since the positive auxiliaries can and do d i d not appear at this point, Bellugi argued that can't and don't were unanalyzed wholes. A f f i x i n g the negative element to the appropriate auxiliary verb w o u l d be the third derivational stage, and the children next acquired a range o f auxiliaries and their negative counterparts, between M L U 3.4 and 3.9. M o s t o f the elementary syntax was then mastered, though i t took many more months or even years to master the correct use o f double negatives, and indefinites, such as not-any or no-one. Maratsos and Kuczaj (1976) looked at one further aspect o f negative auxili­ aries: do young children perceive the contracted n't form as equivalent to not? They used an imitation task to elicit production o f double and triple auxiliary sentences w i t h various placements o f not and n't in the models, e.g. The dog would have not been barking. Little consistency was observed across children, but individual children were quite consistent. For instance, t w o children consistently placed not before the verb and after all auxiliaries, yet the same children correctly imitated n't on the first auxiliary. Such observations suggest that children may believe the t w o forms have different privileges o f occurrence, w h i c h is true i n adult speech. Data from Kuczaj's o w n son A b e , similarly suggest that children may not relate A U X + not to A U X + n't. Abe followed his parents' practice i n showing an over­ whelming preference for don'tower do not, and can't over cannot. However w i t h the auxiliary be, though the parents again preferred the contracted form isn't, Abe at an early stage preferred the full form such as is not. Maratsos and Kuczaj propose that not is learned as an early negative for various constituents: adjec­ tives, NPs, and progressive VPs, w h i c h all can take be before them. W h e n the child learns to produce be i n such sentences, the full negative marker not may be retained. 6.2.2. Typical Errors. M u c h subsequent w o r k on the development o f nega­ tion has focused on B e l l u g i ' s first stage, i n which the child produces negatives external to the sentence. M c N e i l l (1970) surveyed data from a number o f lan­ guages, and concluded that this was a first step not just i n English, but univer­ sally. Other investigators have searched for an alternative account, or questioned the existence o f such a stage. B l o o m (1970) pointed out that the critical data came from sentences i n w h i c h the subject is expressed, e.g. No the sun shining. Not Fraser read it.

(Adam)(='The sun's not shining.') (Eve)(= "Fraser mustn't read it.")

82

De Vilh'ers a n d De Villiers

I f the subject is not expressed, then it is not clear that the subject has been " d e l e t e d " from the left or the right o f the negative morpheme, e.g. ? not ? hold it. I n her o w n longitudinal study, B l o o m (1970) identified a small number o f sen­ tences w i t h subject expressed, but i n every case the negative morpheme was anaphoric, i.e. referring back to an earlier utterance rather than negating the proposition to w h i c h i t was attached, e.g. No mommy do it. (meaning: 'No, let Mommy do i t ' . ) Wode (1977) proposed that anaphoric negatives o f that sort might provide the false analogy for the child to produce nonanaphoric negatives external to the sentence. However his data from English are anecdotal, and the stages he de­ scribes for German are refuted by Park (1979). We reanalyzed the early negatives o f A d a m , Eve, and Sarah, and agreed that very few critical examples existed i n w h i c h the negative element was nonanaphoric, and the subject was expressed (de Villiers & de V i l l i e r s , 1979). Our motivation for the reanalysis derived from observation o f our son, Nicholas, w h o produced plenty o f examples o f the critical sentences between 23 and 29 months. Clearly then, the initial-no negatives do not seem to constitute a universal first step, but individual children may adopt such a strategy. Before turning to an account o f these individual differences, i t may be useful to consider the range o f meanings that are expressed v i a negative sentences. 6.2.3. Relation to Semantics. A c t u a l l y , i n the case o f negation i t seems especially difficult to determine what falls i n the domain o f "semantics" and what constitutes " p r a g m a t i c s . " The negative meanings that have been described by M c N e i l l (1968) and B l o o m (1970), may also be classified as illocutionary forces, that is, the functions that the utterances serve. B l o o m ' s major categories are: 1. nonexistence: where the child expects something but does not see i t . For instance, a child might look at a place setting and say " n o f o r k . " 2. rejection: the child opposes some action, event or object that is either present, proposed or imminent. For instance, the child says " n o b r u s h ! " to a mother approaching w i t h a hairbrush. 3. denial: the child negates the truth o f a proposition uttered by someone, e.g. " I am not a b a b y " i n response to " Y o u ' r e a b a b y . " Pea (1979) adds others such as " s e l f - p r o h i b i t i o n " and " u n f u l f i l l e d expecta­ t i o n , " as do B l o o m and Lahey (1978).

1.

The A c q u i s i t i o n of English

83

In terms o f the interaction between syntactic development and the semantics of negation, B l o o m (1970) found that sentences indicating nonexistence ap­ peared before those expressing rejection, w i t h denial last to appear i n sentences. Furthermore, when auxiliaries began to appear, variation i n auxiliary forms appeared i n sentences expressing nonexistence before it occurred i n sentences expressing rejection ( B l o o m & Lahey, 1978).

6.2.4. Individual Differences. W e have reported individual differences concerning the frequency w i t h w h i c h various negative functions were expressed in sentences. For instance, Nicholas had a clear preference for the expression o f rejection, by a stereotyped and primitive f o r m , namely no + S. However, he produced rarer but well-formed expressions o f denial during the same time period, w i t h correctly placed negatives. Hence his initial-no sentences were reserved for a limited function. Eve showed a similar pattern: her initial-no sentences were primarily for rejection. I n contrast A d a m used don't for rejection: a form not used by Eve or Nicholas during the early period.

6.2.5. Relation to Input. The source o f individual differences i n negative use can be traced to the input the children received. For example, A d a m ' s mother used don't i n her rejection negatives, and sentences w i t h an initial no were predominantly used for denial. A d a m ' s speech showed the same pattern. Eve's mother and Nicholas' parents used initial-no sentences for rejection, usually an initial emphatic no, followed by a polite negative, such as: No, I don't think you should do that. Very rarely did they use don't imperatives for rejection, paralleling the children's use. Eve, and especially Nicholas, seem to have learned the initial no as a convenient way to express rejection, but not denial. Interestingly, the patterns o f anaphoric negation revealed a similar parallelism: A d a m ' s mother used ana­ phoric negatives to deny an earlier statement, as d i d A d a m ; Eve's mother used anaphoric negatives to reject actions, as d i d Eve (de Villiers & de V i l l i e r s , 1979). Maratsos and Kuczaj (1976) also reported close matching to the input use o f not and n't by A b e , except i n the case o f the verb be. Finally, Pea (1979) traced idiosyncratic forms o f his subjects to particular parental forms, e.g. mustn't bite used by one child i n self-prohibition. Maratsos and Kuczaj found double auxiliaries very rare i n the input to A b e , and always w i t h the contracted negative. A s usual, i t is difficult to imagine how children master the more complex aspects o f syntax given the rarity o f the crucial examples.

84

De Villiers a n d De Villiers

6.2.6. Relation to Cognition. Pea (1979) analyzed the three major func­ tions o f negation—rejection, nonexistence and truth-functional (denial) nega­ t i o n — i n terms o f their cognitive complexity. According to Pea, rejection should be the simplest, since i t expresses the child's emotional attitudes towards some­ thing present i n the context, hence no internal representation is required. Nonex­ istence does require internal representation since the object is not present to the child's sense experience but expected given the context. Truth-functional nega­ tion needs internal representation o f a proposition, hence i t is even more cognitively complex. B l o o m (1970) and Pea (1979) found single-word expressions o f rejections to come i n first, then nonexistence, w i t h denial expressed last. However the evidence that this order is recapitulated as the first sentences emerge, is equivocal (see p . 83). 6.2.7. Pragmatics. As mentioned earlier, the distinction between seman­ tics and pragmatics is difficult to make for negation, but there are some phe­ nomena that appear to fall clearly i n the traditional concern o f pragmatics. For instance, take the effect o f contextual factors on denial. Adults find it relatively easy to process the affirmatives concerning an object, e.g. " T h i s is an a p p l e . " They are slower, but still accurate, w i t h a false affirmative, e.g. " T h i s is a pear." The false negative " T h i s is not an apple" takes a little longer, but the true negative " T h i s is not a pear" is the most difficult o f all. Hence there is a reliable interaction between truth/falsity and affirmation/negation, that has been found also i n children (e.g. K i m , 1980). The phenomenon has been accounted for i n a variety o f ways, including a logical processing model by Clark and Chase (1972). Nevertheless, Wason (1972) has a convincing argument that the use o f a negative requires a plausible context: there has to be some reason for someone to believe the opposite, or the felicity conditions are not met. I n the above situation, the listener is shocked by the true negative, as were the preschool children i n K i m ' s study: i t might be true, but one w o u l d n ' t say i t . Naturalistic speech is a poor domain for investigating children's knowledge o f felicity conditions for negation, but Antinucci and Volterra (1973) argue that early denials are i n keeping w i t h the constraint on plausibility, de Villiers and Tager-Flusberg (1975) adapted Wason's (1965) experimental task for use w i t h young children, and provided evidence o f enhanced production o f true denials under conditions o f plausibility, albeit i n a very limited domain. Their subjects viewed arrays o f seven objects, six o f w h i c h were alike and the seventh was different. The children could much more easily complete the statement " T h i s is not . . . " about the odd item, than about one o f the identical items. Hence even for 2-year-olds, contextual factors influence the difficulty o f negative sentences. 6.3.

Questions

6 . 3 . 1 . Course of Acquisition. Questions i n English involve three linguistic devices: rising intonation, inversion o f the usual order o f subject and auxiliary, and initial placement o f the w h - w o r d . Rising intonation seems to be the earliest

1.

The A c q u i s i t i o n of English

85

device used by children, but there may be individual differences i n this regard (for review, see de Villiers & de V i l l i e r s , 1978; Crystal, 1978). K l i m a and Bellugi (1966) undertook the first systematic study o f interrogative syntax i n the speech o f A d a m , Eve and Sarah. A t an average M L U o f 1.75, the three children marked yes/no questions w i t h rising intonation, but had no auxil­ iary verbs at that time. Wh-questions i n this period consisted o f routines such as: What's that? Where (NP) go? and What (NP) doing? More complex questions were not comprehended. I n the second period, when the children averaged 2.75 i n M L U , they understood more wh-questions and more variants appeared i n their o w n speech, suggesting that they understood that the wh-word stood i n place o f a specific sentence constituent. B y the third period ( M L U 3.5), auxiliary verbs were produced, and correctly placed i n initial posi­ tion i n yes/no questions. However wh-questions retained the subject-aux order o f declaratives. Hence the children produced correct forms such as: Does the kitty stand up? at the same time as errors on wh-questions, e.g. Why kitty can't stand up? The discrepancy was accounted for by a theory o f transformational complexity: wh-questions involved a second transformation o f wh-fronting, and it was the­ orized that the children had some l i m i t on the number o f transformations they could use i n one sentence. Several subsequent studies have questioned the gener­ ality o f this finding, and the debate w i l l be discussed under " t y p i c a l errors" below. A second theoretical debate concerns the nature o f the child's formulation o f the category " a u x i l i a r y . " That is, do children regard all auxiliaries as one general class, w i t h equivalent privileges o f occurrence? I f so, one might expect the initial auxiliaries o f yes/no questions to be as varied as the auxiliaries i n declaratives, once the child has discovered how to form yes/no questions. I f , however, the child has not discovered the general category, auxiliary placement in yes/no questions might be acquired i n a piecemeal fashion. Unfortunately the data are not convincingly i n favor o f either alternative. Kuczaj and Maratsos (1983) collected longitudinal data on t w o children: A b e and his younger brother Ben, as w e l l as cross-sectional samples from 14 other children. A variety o f

86

De Villiers and De Villiers

initial auxiliaries emerged simultaneously rather than piecemeal, but t w o auxili­ aries—haven't and couldn't—were used several months i n declaratives before appearing i n yes/no questions. Furthermore, i f children are using the declarative contexts o f auxiliaries to predict w h i c h elements can occur i n initial position i n yes/no questions, Kuczaj and Maratsos argue that they should make certain errors, e.g. produce Better you go? i f they say, You better go. No such overgeneralizations were observed. Children appear to have a represen­ tation o f auxiliaries as a class, but it is one that fails to include certain members (e.g. haven't, couldn't) at the same time as it correctly excludes other words (e.g. better). However to be certain, i t w o u l d be necessary to have evidence that the excluded forms d i d appear i n questions but not i n the correct place. Having established that the opportunity occurred, arguments about incomplete scope o f the auxiliary rule w o u l d be on firmer ground. A clearer example o f piecemeal acquisition o f a rule concerning auxiliaries was reported by Kuczaj and Brannick (1979). Using imitation and judgment tasks, they assessed young children's knowledge o f the rule o f auxiliary place­ ment i n different types o f wh-questions. They discovered that the auxiliary placement rule was acquired at different points for different wh-questions, rather than being an across-the-board rule. Labov and Labov (1978) reported the same phenomenon: wh-words for their subject influenced the correct placement o f the auxiliary i n the order: how, which, who, where, what, when, and why. Unfortunately there is little consistency across studies i n this ordering (Erreich, 1980). Erreich's subjects also had different rates o f correct auxiliary placement, but i n her study the order o f difficulty was: most often correct w i t h which(one), who, what, and where; less often correct w i t h how, why, and when. Erreich also argues against Kuczaj and Brannick's view that the rule is a piecemeal one, despite these equivalent data. Instead, she believes the rule has been formulated in abstract terms, e.g. referring to the category o f wh-words and auxiliaries, but the child has not yet learned the membership o f the category. O n either account, the child has some reason for caution, since the phrase how come is an exception to the general rule i n that the auxiliary remains i n declarative position: How come you didn't go? Kuczaj and Brannick nevertheless found some children i n their tasks over* generalized the inverted auxiliary to how come questions, but the prevalence o f

1.

The A c q u i s i t i o n of English

87

that error i n spontaneous speech has not been established. I f children do overgeneralize to the exceptional case, the notion that they are piecing the rule (or category membership) together from positive evidence is not w e l l supported. The final issue about question acquisition that has attracted the attention o f researchers concerns the development o f the wh-forms themselves. M a n y studies have reported consistency i n the order o f wh-question emergence, among the first being Ervin-Tripp (1970) using a comprehension study. Wooten, M e r k i n , Hood, and B l o o m (1979) described the sequence o f spontaneous speech acquisi­ tion i n seven subjects studied longitudinally. The earlier appearing forms were what, where, and who, but why, how and when were later to emerge. Wooten et al. point to a possible linguistic difference between the early and late forms. The early forms are wh-pronominals, that request information about major sentence constituents: noun phrases and prepositional phrases. I n contrast, they argue, why, how, and when question the semantic relations i n the sentence as a whole, they are wh-sententials that do not simply stand for a missing constituent. For this reason, the c h i l d might not be able to produce them until later. Wooten et al. found that when the wh-sententials d i d emerge, they appeared i n more complex constructions than what, where, and who. The latter tended to occur w i t h verb proforms o f very general character, such as do or go or be. Although the distinc­ tion between wh-pronominals and wh-sententials has intuitive appeal, i t runs counter to traditional accounts (e.g. B r o w n , 1973) that describe wh-questions all as replacing sentence constituents, and i n particular the distinction between where and when is difficult to recognize since both could replace prepositional phrases: Where did you go? When did you leave?

To the shop. After dinner.

However i t is clear that there is some interaction between the wh-words and sentence complexity that deserves further exploration. Other researchers have analyzed the order i n which wh-questions are under­ stood. The structure o f the questions and the transitivity o f the verb are both influences on the comprehension o f wh-questions (Cairns & Hsu, 1978; Tyack & Ingram, 1977), as w e l l as the semantics o f the verb (Winzemer, 1981). Hence the order o f difficulty o f wh-questions varies across studies o f comprehension, but is still i n keeping w i t h the order o f emergence i n spontaneous speech. The relative contributions o f cognitive complexity, structural variables and semantic factors i n the verb, w i l l be discussed further i n the material to f o l l o w . 6.3.2. Typical Errors. K l i m a and Bellugi (1966) reported the existence o f a stage during w h i c h children inverted auxiliaries i n yes/no questions, but failed to invert i n wh-questions. Three other studies have failed to identify such a stage. Hecht and Morse (1974) looked at inversion rates for the t w o types o f questions

88

De Villiers and De Villiers

in spontaneous speech samples from 12 children at the same age (30 months) but varying widely i n linguistic development. N o child had a higher rate o f inversion for yes/no questions than for wh-questions. Ingram and Tyack (1979) analyzed samples o f questions recorded by parents from 21 children aged between 2 and 4 years. A g a i n , rates o f inversion were equivalent at all stages o f development. The most complete study o f this phenomenon was performed by Erreich (1980). Eighteen subjects ranging i n M L U from 2.66 to 4.26 provided spontaneous questions and questions elicited by the experimenter over a series o f sessions. For ten o f the children, the rate o f inversion was equal for w h - and yes/no questions. Five other children presented an unexpected pattern: a higher rate o f inversion i n wh-questions than i n yes/no questions! Therefore, the hypothesis that children are more likely to invert the auxiliary in a yes /no question than i n a wh-question, does not seem well-supported. Only one study, using elicited imitation, confirmed the original finding (Kuczaj & Maratsos, 1975) but only w i t h a single child. Furthermore, the experimental task may have biased the child: half o f the models for wh-questions contained un¬ iverted auxiliaries, but none o f the models for yes/no questions contained uninverted auxiliaries (see de V i l l i e r s , 1984). Overmarking errors have also been reported i n yes/no questions by Hurford (1975). His subject produced sentences such as: Did you came home? What's that is? One explanation offered is that the first aux element has been copied to the initial position, but not deleted from its original declarative position. Once again, such errors have been held as evidence for the basic operations hypothesis (Valian et al., 1981). However Maratsos and Kuczaj (1978) d i d not find them to be fre­ quent i n their data, and favor a processing account o f the error. 6.3.3. Individual Differences. Evidently, individual variation is the rule i n the development o f interrogative syntax. The rate o f inversion i n yes/no versus wh-questions is different across individuals (Erreich, 1980; Tyack & Ingram, 1977). I n addition the wh-words that have the most correct auxiliary placement vary across subjects (Labov & Labov, 1978; Erreich, 1980; Kuczaj & Brannick, 1979). Unfortunately no research has explored the reasons for these individual differences either i n the input the children receive, or by exploring the functions of the forms, their relative frequency o f use by individuals, and such factors as the productivity o f the different questions w i t h different verbs or subjects. A l l o f these factors might be influences on the course o f individual acquisition patterns in interrogative syntax, just as they are i n other areas o f grammar (e.g. negation, see p . 83).

1.

The A c q u i s i t i o n of English

89

6.3.4. Relation to Cognition. The reason most often given for the order o f emergence o f the wh-forms is that they vary i n cognitive complexity, since why and how questions encode more abstract ideas than questions such as what or who (Ervin-Tripp, 1970). Wooten et al. (1979) argue that cognitive complexity is but one variable i n the order o f emergence, and perhaps not so important as linguistic factors. For example, L i g h t b o w n (1978) demonstrated that secondlanguage learners o f English acquire wh-questions i n the same order as first language learners, even though they are cognitively more advanced and have sufficient conceptual ability to ask the questions i n their first language. Sec­ ondly, Wooten et al. point out that the conceptual notions held to cause d i f f i ­ culty, such as time and causality, were encoded i n other linguistic devices such as connectives ( B l o o m et a l . , 1980) at an earlier point. I t is therefore the l i n ­ guistic complexity o f the question forms, not just the concepts they encode, that delays their appearance i n spontaneous speech. Yet children may use questions spontaneously before they understand their meaning. Blank (1975) discusses the special nature o f why questions, w h i c h encode a complex o f abstract notions o f natural laws, human motivation and logical reasoning that a young child could not possibly grasp. She provides anecdotes from the spontaneous speech o f a young c h i l d , Dusty, who follows the common practice o f asking a great many why questions, many o f which have no sensible interpretation, e.g. Adult Dusty

"That's the garage door." "Why the garage door?"

Blank argues that i t is only by engaging i n a great deal o f discourse i n v o l v i n g why questions and their answers, that the child can unravel the meanings that why encodes and begin to determine the circumstances under which the question becomes appropriate. Given all o f these convincing linguistic and cognitive complexity arguments for the piecemeal emergence o f wh-questions, one result i n the literature be­ comes especially anomalous. Genie, the child w h o received little linguistic or cognitive stimulation before her rescue at 13 years 7 months (Curtiss, 1977) appeared able to answer all wh-questions at the same point i n time. However the fact that she never produced a syntactically marked question, and the discrepancy between her responses to " r e a l - l i f e " versus test questions, suggest that she may have been dependent upon contextual clues rather than a true understanding o f the forms. 6.3.5. Relation to Semantics. W i t h respect to a particular scene, some questions seem more plausible than others. For instance, i f a picture contains a girl eating an orange, the question

De Villiers and De Villiers

90

What is the girl eating? sounds more reasonable than: Where is the girl eating? Winzemer (1981) suggests i t is i n the semantics o f the verb that this plausibility resides. Her model proposes that verbs have certain implied constituents as components o f their meaning, e.g. eat has an object as an expected constituent, but not a location. Drive, however, has a location as an expected constituent. I f children k n o w these properties o f verbs, then questions that ask for expected constituents w o u l d be easier than questions that ask for an unexpected constitu­ ent. Her prediction was confirmed w i t h preschool children, who often made the error o f answering an unexpected question w i t h the expected component. I n a picture-cued comprehension test, verb choice could have a significant impact on question difficulty. I n real life discourse however, there may be cues other than the verb semantics to the meaning o f a question. 6.3.6. Interaction with Pragmatics. The functions served by interrogative forms i n English are numerous, and several writers have developed classification schemes for them. Holzman (1972) analyzed the functions o f questions i n the speech o f A d a m , Eve, and Sarah and their mothers, and found five broad categories: 1. 2. 3. 4.

requests for information requests for behavior " t e s t " questions for w h i c h the speaker already knows the answer interrogatives such as what to indicate a lack o f understanding

5. other purposes, such as threats or suggestions, e.g. " W h a t do you keep asking me f o r ? " Shatz (1979) identified 11 categories o f interrogative function i n parental speech, including challenge, calling attention to something, and a variety o f conversa­ tional fillers. Despite the fine categorization, she found coder agreement to be around 80%. I n terms o f children's use, Griffiths (1979) argues that the early-appearing functions are requests for actions and object, and calling attention. True requests for information may come i n relatively late i n the one- or t w o - w o r d period. Holzman's results indicate that children w i t h an M L U o f 3.0 use many but not all of the major functions o f the interrogative, including questions serving as re­ quests for information, indicating a lack o f understanding, and suggestions for behavior. T w o o f the children also used test questions, though Sarah d i d not. A t this stage they d i d not use interrogative forms as requests for behavior or as

1.

The A c q u i s i t i o n of English

91

negative evaluations o f another person's behavior. Garvey (1975) found requests expressed indirectly as interrogatives i n dyads o f 4- and 5-year-old children. Furthermore, the forms reflected the variety found i n adult discourse: they questioned ability: Can you hold this? desires: Do you want to catch me? future actions: Will you get me one? reasons: Why don't you hold it? just as described for adult polite requests (e.g. Searle, 1975). The mapping between f o r m and function i n questions is therefore particularly rich, and yet disappointingly little data exist on how individual children enter the system. The research to date has focused almost exclusively on form (e.g. K l i m a & B e l l u g i , 1966; Kuczaj & Maratsos, 1983) or on function (Holzman, 1972) i n an individual's development. D o wh-question forms become immediately avail­ able for all functions, or are they tied to particular functions at the start? The limited data that have been reported suggest that at least the wh-words them­ selves originate i n restricted interaction routines (Johnson, 1980) but little is known beyond the earlies stages. 6.3.7. Relation to Input. Researchers have focused on parental speech to children to determine whether or not the input is especially revealing about the structure o f questions, and arguments can be found for both positions. For instance, parents frequently omit the initial auxiliary o f a yes/no question: You want juice? Newport et al. (1975) used this as an illustration o f the problems that "motherese" presents to the language learner. A s discussed earlier, Furrow et al. (1979) find a correlation between parental omission o f initial auxiliary, and children's rate o f auxiliary development, so they argue that the optionality draws attention to the auxiliary! Snow (1972) found mothers made use o f occasional questions—wh-forms i n which the w h - w o r d is not preposed but occupies the place o f the missing constit­ uent, e.g. You're trying to find what? These structures might reveal the role o f the w h - w o r d more clearly than the preposed question. Interestingly, children themselves do not seem to use occa­ sional questions. Shatz (1979) studied the questions that 17 mothers used w i t h their young children during play sessions. Some o f the children were linguistically advanced

92

De Villiers and De Villiers

(three or four words per sentence), and others were less advanced (less than t w o words per sentence). Shatz's interest was i n the stereotypicality o f the questions w i t h regard to the functions they served, to explore whether mothers used limited form-function pairings that might enhance the child's decoding o f language. However the stereotypicality o f the questions was defined on a group rather than an individual level: her objective was to discover form-function pairings that would facilitate understanding o f language i n general, not o f a particular moth­ er's speech. Shatz d i d find more stereotypical questions used by mothers o f less linguistically advanced children, but even their speech revealed substantial varia­ tion i n the forms for particular functions. Furthermore, there was only weak evidence that the children responded more successfully to characteristic pairings of form and function than to noncharacteristic pairings. The only statistically significant result was that o f the mothers o f less advanced children, those who tended to produce stereotyped " t e s t " questions, had children who understood them readily. Presumably these questions developed as routines between particu­ lar mother and child pairs (see Snow, 1977). Data from Holzman (1972) confirm Shatz's claim that questions even to young children serve many distinct func­ tions. Bellinger (1979), however, describes parents' increasing use o f inter­ rogative forms rather than imperative forms serving as directives as children increase i n age, suggesting that parents may diversify the use o f interrogative forms as their children mature. Nevertheless, Shatz (1978b) has demonstrated that even 2-year-olds w i l l re­ spond appropriately to requests veiled as questions, such as " C a n y o u shut the d o o r ? " Children do not seem to process the form as a question, then reason that it must have an indirect meaning (e.g. Clark & L u c y , 1975). Rather, Shatz proposed that they respond w i t h action whenever they can, and often overgeneralize this strategy i n responding w i t h action to real questions such as: Do you brush your teeth?

7. D a t i v e a n d P a s s i v e 7. /. Parallel

Concerns

There are parallel concerns for research on the dative and passive construc­ tions i n child speech, and before considering each type it might be helpful to point to those issues: 1. English-speaking children develop a heuristic o f sentence interpretation based on the canonical sentence order o f S - V - 0 found i n English declaratives. I n the dative and passive constructions, such a strategy could lead the child astray when there are no semantic clues to interpretation, e.g. The boy gave the dog the duck. The boy was kissed by the girl.

1.

The A c q u i s i t i o n of English

93

When do these strategies come into play? 2. Both the passive and the dative constructions have been argued to be lexical rules rather than syntactic rules, w i t h the possibilities o f transformations or alternative phrase structures coded directly w i t h the verbs i n the lexicon (Bresnan, 1978). There are restrictions for both constructions on the verbs that can occur i n them, for instance, there is no passive equivalent o f The bottle contained shampoo, cf * Shampoo was contained by the bottle. and there is no transformed dative o f The man announced the winner to the crowd, cf * The man announced the crowd the winner. Hence, as Baker (1979) has argued, perhaps the child should be cautious about generalizing either the passive or the dative to verbs without positive evidence from the input that they can appear i n those forms. H o w specific or general is the child's formulation o f the dative and passive rules? 3. For both construction types, there are biases i n the frequency o f use i n the language, possibly reflecting the a priori likelihood o f agents being animate, and of recipients being animate. Hence the models that the child receives are not semantically neutral, but could be given an alternative, semantic description. These biased sentences can be readily interpreted by the child because they do not depend upon his k n o w i n g the syntax, only the likelihood o f certain events i n the w o r l d . So for instance, John gave Mary the book. has only one plausible reading, as does The car was driven by the girl. Hence children's early understanding o f these constructions may not be syntac­ tically based. H o w prevalent and how lasting are these semantic biases? 7.2.

Datives

1.2.1. Order of Acquisition. Roeper, Lapointe, B i n g and Tavakolian (1981) and Osgood and Zehler (1981) confirmed earlier reports (e.g. Fraser, Bellugi & B r o w n , 1963) that the transformed dative, w h i c h violates the can­ onical S V O order, is more difficult for children to understand than the simple dative when no semantic supports are provided. Both studies found a prevalent

94

De Villiers a n d De VilNers

error o f acting out the S - V - I O - D O sentence as i f i t had a missing to, interpreting:

e.g.

The dog gave the cat the puppy. as: The dog gave the cat to the puppy. Osgood and Zehler argue that the canonical order is S - V - D O - I O i n adult processing (Waryas & Stremel, 1974) and furthermore, I O - m a r k i n g is more common than D O - m a r k i n g i n the w o r l d ' s languages. Verb incorporation o f pronominal direct objects is also a common occurrence crosslinguistically, while virtually nonexistent for pronominal indirect objects (Sedlack, 1975). I n production, Osgood and Zehler found the basic form to be heavily pre­ ferred i n their task, but the transformed version was beginning to be used by the 4- and 5-year-olds for describing selected event types. There are plenty o f exam­ ples i n the spontaneous speech o f much younger children o f transformed datives, e.g. Give me that. Show me the book. but they may be frozen forms w i t h particular pronominal indirect objects. The full analysis o f the transformed dative may take considerably longer. I t may also be important to look at a wider variety o f verbs i n experimental studies: both Roeper et al. and Osgood and Zehler used only the verb give. 7.2.2. Relation to Input. N o studies exist on the natural input children receive on the dative, but one study used a training technique to explicitly test the hypothesis that children should be cautious about generalizing the dative trans­ formation i n the absence o f positive evidence. W i l s o n , Pinker, Zaenen and Lebeaux (1981) taught 4- and 5-year-olds new words for novel bitransitive actions (e.g. passing an object through a tube to a recipient). Subjects heard the word either i n the basic or transformed dative construction, and they were subse­ quently tested on their ability to comprehend and produce both forms w i t h the novel verb. Despite being taught the verb i n the transformed version, these subjects showed virtually no production or understanding o f that f o r m . Instead, they heavily preferred the basic dative even for verbs not heard i n that construc­ tion. The subjects i n this age group may have had too little knowledge o f the transformed dative to provide an adequate test population. Perhaps older subjects who are i n control o f both variants o f the dative w o u l d provide a better test o f the hypothesis o f lexical specificity.

1.

The A c q u i s i t i o n of English

95

7.2.3. Typical Errors. I n comprehension tests, the typical error made by young children is to rely on the canonical S - V - D O order, i n the absence o f semantic constraints, and hence to misinterpret the transformed dative as de­ scribed earlier. For sentences w i t h semantic constraints, children tend to rely on animacy as a clue to identify the recipient (see below under "semantics"). I n production, Bowerman (1982) has reported errors made by her daughters that suggest overgeneralizations can occur i n the dative, the type o f error that Baker (1979) proposed w o u l d not be seen. For instance, she reports such utter­ ances as: She said me the answer. occurring late i n the preschool years. Errors o f this sort violate the predictions made from a lexicalist position (Roeper et a l . , 1981) but we do not k n o w how prevalent or long-lasting the overgeneralizations might be. 7.2.4. Relation to Semantics. The most thorough study to date o f 3-5-yearold children's comprehension and production o f both forms o f the dative was conducted by Osgood and Zehler (1981). Previous research (e.g. Cromer, 1975; Fraser, B e l l u g i , & B r o w n , 1963) had suggested that the transformed dative was difficult and acquired late by children, but these investigators had carefully controlled the sentences to remove semantic clues. Osgood and Zehler employed sentences w h i c h varied semantically, including nonhuman agents/recipients and both inanimate and animate objects: e.g. 1. 2. 3. 4.

Brother gives block to sister. Mother gives puppy to father. Dog gives ball to cat. Tiger gives puppy to cat.

In addition, they had four levels o f sentence complexity; e.g. A. B. C. D.

Mother gives the hat to the girl. Father gives the hat and the book to the girl. Mother gives the hat to the girl and mother gives the book to the boy. Father gives the bird to the boy and then the boy gives the bird to the girl.

Osgood and Zehler argued that the semantics o f the situation i n (1) are pro­ totypical transfers, and their subjects performed best i n producing and com­ prehending datives describing such situations. A d d i t i o n a l l y , the children were best able to use the transformed dative for prototypical events. They reverted to the basic to-form when the semantics became less typical. Comprehension was predictably impaired for the transformed sentences w i t h animate transfer objects, e.g.

De Villiers and De Villiers

96

The tiger gives the cat the puppy. Osgood and Zehler (1981) argue that the prototypicality o f the transfer events provided a support for the children's initial use o f syntactic forms. " O n l y later does a syntactic ability that is independent o f contextual support d e v e l o p " (p. 381). Roeper et al. (1981) studied older children (kindergarten, second and fourth grade) and found a similar dependence upon semantic cues i n comprehension. They constructed questionnaires that counterbalanced animacy so that i t could not provide a systematic cue to interpretation, but their second graders i n particu­ lar relied on animacy as an indicator o f the indirect objects. The kindergartners also relied on animacy i n cases where the prototypical semantics was reversed, e.g. The dog gave the spoon the cow. hence misinterpreting those sentences more than the other varieties. It is one thing to argue that children's comprehension o f dative sentences is dependent upon the probabilities o f events i n the real w o r l d . Such dependence reflects a lack o f understanding o f the syntax, particularly o f the " t r a n s f o r m e d " version. However Osgood and Zehler wish to go one step further, and to argue that children's syntactic analysis o f the transformed dative is complete first for the typical events. Then, the children are able to produce transformed datives as long as the event they are describing is a typical transfer. There are t w o possible alternatives: one, that children's knowledge o f the syntactic form is complete for different semantic domains at different times; t w o , that children's ability to produce/comprehend the syntactic form is subject to processing constraints, such that when semantic supports are lacking, their performance declines. These alternatives have not been w e l l distinguished i n the literature to date.

7.3.

Passives

7 . 3 . 1 . Order of Acquisition. The course o f acquisition o f the passive con­ struction is not entirely worked out, primarily because i t is a rare construction i n spontaneous speech even i n adulthood. Truncated passives, w i t h the agent un­ specified, are much more frequent than full passives. Watt (1970) made the proposal that despite the linguistic treatment o f truncated passives as deriving from full passives, i n acquisition truncated passives should appear first. He argued that psychologically, truncated passives might be analogous to predicate adjectives, e.g. He was kicked. may be perceived as the same structure as:

1.

The A c q u i s i t i o n of English

97

He was sore. Horgan (1978) presented evidence on the acquisition o f the passive i n the speech of 2- to 4-year-olds describing pictures, and argued that the truncated passive had a distinct course o f acquisition from the full passive. Her evidence was that the truncated passives almost all had inanimate surface subjects, e.g. The lamp was broken. whereas the majority o f the full passives had animate surface subjects, e.g. The girl is chased by the boy. Furthermore, the truncated passives had verbs such as break that were also used as statives; the full passives contained a wide variety o f action verbs. So topics and verbs were distinct for truncated versus full passives. The conclusion that truncated precede full passives i n development is disputed by Maratsos and Abramovitch (1975) who studied comprehension o f the t w o forms. T o match the sentences on semantics, they tested comprehension o f full passives w i t h an unspecified agent, e.g. The cow was pushed by someone. Their preschool subjects performed equally w e l l on truncated and full passives. I t seems l i k e l y , however, that w i t h a specified agent, children might show more errors because o f the processing load. W h e n the semantics are controlled, trun­ cated and full passives seem to be perceived as equivalent structures. Maratsos and Abramovitch argue that truncated passives may appear earlier i n samples o f spontaneous speech because o f their greater frequency than full passives i n adult speech. They point out that for a rare construction like the full passive, the time of appearance i n spontaneous samples should not be taken as the time o f acquisi­ tion. Wells (1979) makes a similar point; i n 18,000 utterances taken from 60 children aged 3 6 - 4 2 months, only 19 passive verbs were recorded. However, i n elicited production also, truncated passives are more often produced than full passives (Baldie, 1976). Knowledge o f the form o f the agentive Z?y-phrase was also investigated by both Horgan (1978), and Maratsos and Abramovitch (1975). I n Horgan's study o f spontaneous speech, 75% o f logical subjects were inanimate NPs for the 2- to 4-year-olds, but older children (5 to 13 years) used inanimate logical subjects less often ( 3 9 . 5 % ) . I t should be noted, however, that the populations were describing different picture sets. Furthermore, individual differences were found among the 2- to 4-year-olds in the types o f passives used. A b o u t half o f the children used only reversible passives, such as

98

De Villiers and De Villiers

The cat was chased by the girl. However most o f these had the w o r d order reversed, i.e. the picture was o f the cat chasing the g i r l . The other children used only nonreversible passives, and then only w i t h an instrument as the logical subject, e.g. The lamp was broken by the ball. No nonreversible passives w i t h agents as logical subjects, e.g. The lamp was broken by the girl. appeared before 9 years. The possibility is raised by Horgan that individual children might develop the passive forms along different lines. I t was only at age 11 that Horgan found the same children producing both reversible and nonrever­ sible passives. However i t is difficult to believe that this result reflects k n o w l ­ edge rather than encoding preferences, given the data on comprehension from preschoolers. Nevertheless the bias against having an inanimate N P chosen over an animate N P as the sentence topic seems to be a pervasive one (see also Baldie, 1976; de V i l l i e r s , 1980). Horgan argued that the nonreversible passives might be produced by analogy to other constructions also found i n the speech samples, i n particular reflexives and sentences w i t h from. Both constructions were used i n a restricted way similar to the nonreversible passives e.g. the reflexives had inanimate nouns f o l l o w i n g by, as i n The ball bounced by itself. The earliest uses o f from were locative, or encoded non-agentive causation e.g. Snowman is melting from the sun. Hence for all these constructions only nonagentive causation was signaled, and the noun phrase f o l l o w i n g the preposition was inanimate. Horgan speculated that the child assumes that by signals nonagentive causation on analogy w i t h the other constructions, and adds the Z?y-phrase or/rom-phrase onto the truncated structure that she already uses to comment on the state o f things. Maratsos and Abramovitch (1975) add to the picture o f children's syntactic knowledge o f the full passive. They gave children sentences to act out that were anomalous i n several ways, e.g. no be form: The cat licked by the dog. or no by preposition:

1.

The A c q u i s i t i o n of English

99

The cat is licked the dog. and demonstrated that young children require both features to be present i n order for the sentence to be treated as a passive. The failure when the preposition was absent also serves as a demonstration that the children were not processing the full passives simply by processing the truncated passive i n the first portion. Evidently they k n o w that a prepositional phrase should be part o f the structure. Substitution o f the preposition of"for by resulted i n the sentence being treated as a passive also, though children tended to imitate it as from, i n keeping w i t h Horgan's observation. However, the children d i d not regard a sentence w i t h a fake preposition as a passive, e.g. The girl was chased po the squirrel. The subjects thus had a fairly well-defined knowledge o f the permissible struc­ ture o f the prepositional phrase found i n the passive, and the closed class items that could appear therein. I n comprehension studies, questions about the productivity o f the activepassive relation have only recently been asked. I n attempts to control the seman­ tic cues to comprehension, investigators have used reversible passive sentences, generally w i t h t w o animate N P ' s . For both act-out and picture-cued comprehen­ sion, verbs that referred to a clear action were selected for the sentences. A s a result, our knowledge o f the generality o f the active-passive relation was impoverished. Maratsos et al. (1979) tested children's understanding o f passive sentences w i t h either action verbs or nonaction verbs. The latter included: remember, forget, know, like, miss, see, hear, and watch. T o examine children's com­ prehension, they employed t w o different procedures w i t h the same results. I n one procedure, they told the child a sentence, e.g. Donald was liked by Goofy. and then asked " W h o d i d i t ? " W h i l e recognizing the peculiarity o f such a question about a nonaction, the authors argued that the question should have been equally peculiar for the equivalent active sentence, but their subjects gener­ ally answered i t correctly i n that case. The results revealed a clear interaction between sentence type (active versus passive) and verb type, w i t h passive sen­ tences containing action verbs being w e l l understood by the 4- and 5-year-old subjects, but passives containing nonaction verbs not being understood. I n a second task, the passive sentences were embedded i n a story context i n w h i c h answering a question e.g. " W h o really d i d n ' t like the other one?" depended upon understanding a crucial sentence, e.g. The cat was hated by the raccoon.

100

De Villiers and De Villiers

A g a i n , the children understood the active sentences w i t h nonaction verbs but not the corresponding passives. Maratsos et al. argue that children o f 4 and 5 years do not therefore have a full knowledge o f the passive-active relation. They contend that this relationship provides one o f the major reasons for positing the abstract concepts o f logical subject and logical object i n English, and children are not able to formulate the relation i n such general terms. Rather, the learning o f the passive proceeds piecemeal. Using picture-cued comprehension, de V i l l i e r s , Phinney and A v e r y (1982) have confirmed that children fail to understand nonaction verb passives while they do understand both action verb passives and nonaction verb actives. H o w e v ­ er they also find that at an earlier developmental stage, the nonaction verbs are poorly understood i n the active sentences. Maratsos et al. d i d not find this result, though they used older subjects. The significance o f this finding is that the presence o f nonaction verbs could increase processing load, w h i c h may not matter i f the sentence is active, but might be sufficient to interfere w i t h process­ ing o f the passive sentences by a child w i t h only a fragile hold on the syntax. Hence i t is not clear whether the results should be interpreted as piecemeal acquisition o f knowledge, or as complete knowledge but a processing limitation. These alternatives are particularly difficult to distinguish i n preschool children. 7.3.2. Typical Errors. Bever (1970) studied the comprehension o f semantically reversible and irreversible passive sentences by children between 2 and 5years-of-age. Between V/i and 4, w i t h girls i n advance o f boys, the subjects performed at better than chance level w i t h semantically reversible sentences. However the group o f children who were slightly older showed a worse perfor­ mance, systematically reversing the passive sentences w i t h no semantic con­ straints, i.e. interpreting The boy was kissed by the girl. as: The boy kissed the girl. Bever suggested that these children were using a perceptual strategy by w h i c h any N - V - N sequence was treated as the active order agent-action-object. Marat­ sos (1975) replicated Bever's finding o f a regression i n performance at around 4 years, when subjects tend to reverse passives. Strohner and Nelson (1974) failed to find evidence o f an earlier stage when children perform above chance level on reversible passives. Their 2-year-olds performed at random on such sentences, their 3-year-olds tended to reverse them, using the noun-verb-noun strategy.

1.

The A c q u i s i t i o n of English

101

de Villiers and de V i l l i e r s (1973) attempted to relate the comprehension o f reversible active and passive sentences to M L U rather than age. Their sample was younger, w i t h a mean o f only 26.8 months. W i t h reversible passive sen­ tences, the group o f children i n early Stage I V ( M L U 3 . 0 - 3 . 5 0 ) showed the greatest tendency to choose the first noun as agent. A g a i n , there was no evidence that the less advanced children were acting out passives correctly. 7.3.3. Individual Differences. Bridges (1980) claimed that the grouping o f children i n previous studies was concealing individual response patterns to pas­ sive sentences. Her large-scale study was concerned w i t h identifying individual response patterns to active and passive sentences i n four variants o f the com­ prehension task. I n all the tasks, response patterns based on w o r d order were infrequent (10% overall) compared to "situationally d e r i v e d " strategies, for instance, where the child used the toy closest to his dominant hand to perform the action. Although Bridges' point is an important one, that "strategies" should be considered at an individual rather than at a group level, her study resulted i n a rather l o w estimate o f the prevalence o f word-order strategies. Unfortunately she does not present data separated by sentence type, so active and passive sentences are considered together i n the tables. The l o w incidence might be due to the more stringent criterion that one must apply i n judging the existence o f a strategy for an individual c h i l d . O n the other hand, her tasks included sentences o f a more homogeneous variety than other studies: the only verb studied was push, and all the nouns referred to toy vehicles. Possibly i n this particular area, young children are responsive to more sources o f contextual bias than i n the usual testing situation where subjects may have little to rely on other than w o r d order i n the instructions. A t present then, the prevalence o f the word-order strategy is not clear. 7.3.4. Relation to Input. O n all estimates, the number o f passives that children hear i n the input is small. B r o w n (1973) found no full passives i n the speech samples o f 713 utterances from the parents o f A d a m , Eve and Sarah, but the children were i n Stage I . Maratsos (1984) claims that passives containing action verbs are more frequent i n parental speech than passives containing nonac­ tion verbs. However, no direct study o f passives i n the natural input to children has been done. Several training studies, however, have manipulated the conditions o f model­ ing passive sentences and observed the effects on children's use o f the f o r m . Strohner and Nelson (1974) conducted an interesting training study w i t h the passive using 100 4-year-old subjects divided into five treatment groups. Pre­ vious research (e.g. Turner & Rommetveit, 1967) had shown that i f attention were drawn to the patient i n an event, the possibility that a subject w o u l d employ the passive voice was increased. The function o f the English passive is to high­ light the patient by placing i t i n the salient sentence position. Strohner and

102

De Villiers and De Villiers

Nelson's subjects were pretested on passive comprehension, then four groups received training that consisted o f presenting either a passive or an active sen­ tence along w i t h a picture. One passive and one active group had their attention drawn to the patient, e.g. " L o o k at this c a t . " For the other groups the actor was highlighted. The fifth group received no training. The hypothesis was that active sentences were a better match to the situation drawing attention to the actor; passive sentences were a better match when the patient was the focus o f interest. After modeling 12 sentences w i t h appropriate pictures, a comprehension posttest w i t h novel sentences was given. The group that received passive sentences matched to an event w i t h the patient highlighted, showed the greatest (and significant) improvement i n comprehension o f passives from pretest to posttest. No corresponding improvement occurred for the matched active group. Apparently even this brief (12 sentence) exposure to passive sentences i n appropriate felicity conditions was enough to decrease the 4-year-olds' depen­ dence on an actor-action-patient strategy. B r o w n (1976) found a similar i m ­ provement after modeling passive sentences to children i n appropriately m o t i ­ vated contexts i n a story. Providing a suitable context for use o f the passive voice by calling attention to the patient i n a sentence can thus enhance children's decoding o f the roles o f the noun phrases i n passive sentences. The standard comprehension test, i n w h i c h passives are presented i n inappropriate felicity conditions, may i n fact underestimate what children k n o w , and increase the likelihood o f use o f a processing strategy. T w o further training studies i n v o l v i n g the passive have been concerned w i t h the productivity o f the passive rule. I n opposition to the claims o f Maratsos et al. (1979) , Lebeaux and Pinker (1981) argued that the child's passive rule works as w e l l for sensory predicates like see as i t does for action predicates like push. Their study involved teaching a new verb to 4-year-old children, one that re­ ferred to a sensory predicate e.g. looking through a tube. They found their subjects able to use the new verb i n passive sentences w i t h no difficulty, even though they had only heard i t i n the active voice. Hence there was no restriction of the passive voice to action verbs, and also the subjects produced passives containing the new verb i n the absence o f positive evidence that it could occur i n that construction.' O f course, their subjects may have considered the new activity to be an action. I n the second study that examined productivity o f the passive rule, de Villiers (1980) attempted to teach the passive to 3-year-old children who failed on a pretest o f passive comprehension. The children heard modeled either " p r o ­ t o t y p i c a l " passives i n v o l v i n g an action verb and an animate patient, e.g. The dog is being licked by the cat. or " a t y p i c a l " passives containing a nonaction verb and an inanimate patient, e.g.

1.

The A c q u i s i t i o n of English

103

The book is being read by the lion. Passive sentences were modeled w i t h pictures, and the child had to produce sentences about other interspersed pictures that included actions and nonactions, animate, and inanimate patients. The subjects produced few passives, but those that they d i d produce were " p r o t o t y p i c a l " regardless o f the modeling they received, de Villiers argued that those events constituted the most appropriate conditions for use o f the passive i n a picture-cued task, and that the subjects came into the experiment w i t h that bias from previous experience. However, a follow-up study (de V i l l i e r s , 1984) which elicited many more full passives failed to find an effect o f differential training on the range o f passive types produced by 3-and 4-year-old children. 7.3.5. Relation to Semantics. Quite apart from the issue o f semantic limita­ tions on the passive rule discussed above, there is a question about the extent o f children's reliance on semantic cues to comprehension. Bever (1970) reported that 2-year-olds performed at random on passive sentences even when semantic constraints were available to help them, e.g. The dog was patted by the boy. Bever argued that this contradicted the claim that young children rely on nonlinguistic, contextual information i n their decoding o f language (e.g. M a c namara, 1972). I n contrast, the 3-year-old subjects placed heavy reliance on the probability o f the events i n decoding passive sentences. Contradicting this report, Strohner and Nelson (1974) systematically varied event probability i n their stimulus sentences presented to 2- to 5-year-old c h i l ­ dren, and found even 2-year-olds relied on event probability. However Chapman (1978) argued that 2-year-olds may not know the differential likelihood o f certain events, such as dogs chasing rats rather than vice-versa. Hence there may be an improvement i n using event probabilities over the preschool years. B y 6 years, however, children can generally use the syntactic form o f the passive even when the event probability is very l o w , (Bever, 1970) as i n The policeman was eaten by the candy.

8. C o o r d i n a t i o n 8.1. Issues

in

Acquisition

O f all varieties o f combining sentences together, coordination is the simplest. Yet there are rules governing coordination w i t h and that involve a complex interplay o f syntax, semantics and pragmatics. According to the standard trans­ formational treatment (e.g. Harries, 1973), phrasal coordinations such as:

104

De Villiers and De Villiers

Bill and John loved apples. are derived from the full sentential coordination i n deep structure: Bill loved apples and John loved apples. I f the redundant elements are deleted from the left conjunct, i t is referred to as backward deletion; forward deletion occurs i f the redundant elements are deleted from the right conjunct. I n English, subject and verb redundancies are deleted i n a forward direction, verb phrase and object redundancies are deleted i n a back­ ward direction, e.g. forward deletion subject redundancy: John loved apples and hated lemons, verb redundancy: John loved apples and Bill lemons, subject and verb redundancy: John loved apples and lemons. backward deletion verb phrase redundancy: object redundancy:

John and Bill loved apples, John loved and Bill hated apples.

Backward deletion is less common i n the w o r l d ' s languages, and is considered to be a more controversial process i n English than is forward deletion (Harries, 1973). M o r e recent grammatical treatments (e.g. Lasnik, 1976) do not transforma­ tionally derive phrasal coordination from sentential forms, but talk o f " g a p s " i n the surface structure o f phrasal coordinations which the semantic interpretive rules f i l l i n by mapping to the full f o r m . Hence the relationship to the structural form is still considered essential to the phrasal form's interpretation. Other accounts (Dougherty, 1967), less favored i n general, generate phrasal conjunc­ tions as base phrase structures on the grounds that this is necessary for certain phrasal conjunctions, e.g. John and Sally are a married couple. for which there is no sentential counterpart. I t is easier to write phrase structure rules to generate conjunctions o f like constituents, however, than i t is to include conjunction o f nonconstituents e.g. ( S V + S V ) 0 or S V O + S O . I n contrast, the deletion schema can deal w i t h all structural varieties o f phrasals. It should be noted that i n order for the sentential form to be reduced, the redundant noun phrases ought to be coreferential, so for example, John patted his dog and Bill stroked his dog. could not reduce to:

1.

The A c q u i s i t i o n of English

105

John patted and Bill stroked his dog. Whether or not one should consider this a syntactic constraint or a pragmatic one, the child clearly must learn the conditions under which phrasal and sentential forms mean the same thing, and when the alternative forms are each appropriate. Finally, the semantic relationship between the t w o propositions i n a conjunction can be very varied, even w i t h the conjunction and. For English language acquisition, the issues for coordination are: 1. What is the order o f acquisition o f the sentential versus phrasal forms, and of backward versus forward deleted forms? 2. D o children respect the conditions o f use o f phrasal versus sentential forms? 3. H o w does the semantics o f the conjunction affect acquisition?

8.2. Order

of

Acquisition

Lust (1977; Lust & M e r v i s , 1980) has made a strong argument that i n acquisi­ tion, sentential forms have primacy over phrasal forms. I n elicited imitation, phrasal forms were more difficult for children than sentential forms matched i n syllable length. Furthermore, children demonstrated that they understood the mapping by elaborating phrasals to sententials, and reducing sententials to phrasals when imitating. Lust argued i n addition for the primacy o f deletion i n a forward direction i n child speech, since forward phrasals were imitated more readily, and forward sententials more often reduced, than their backward coun­ terparts. Lust and M e r v i s (1980) make a similar case based on spontaneous speech samples from a cross-sectional sample o f young children. Other researchers (Tager-Flusberg, de Villiers & Hakuta (1982); B l o o m , Lahey, Lifter & Feiss (1980); and Ardery (1980) have disputed the fundamental finding o f primacy o f the sentential form. I n data on comprehension, elicited imitation and elicited production tasks w i t h preschool children, and i n spon­ taneous speech analysis from the three subjects B r o w n (1973) studied longitudi­ nally, Tager-Flusberg et al. found no clear evidence that phrasals on the whole are more difficult than sententials. T w o particular phrasal forms, ( S V + S V ) 0 and S V O + S O , gave children inordinate trouble i n the experimental tasks, and those are precisely the forms that could not easily be generated by phrase struc­ ture rules i n the base. Hence Tager-Flusberg et al. argue that children's phrasal conjunctions are generated directly by phrase structure rules, but they leave open the possibility that later i n childhood, children might learn a mapping rule be­ tween the sentential and phrasal forms. A l l the researchers concur, however, i n finding children's earliest phrasal coordinations to be forward deletions rather than backward deletions. Primarily this is because subject conjunctions are rare i n child speech. W h e n appropriate

106

De Villiers and De Villiers

conditions were provided for subject conjunction i n the elicited production task used by Tager-Flusberg et a l . , it was used 75% o f the time compared to appropri­ ate object conjunction which was used 95% o f the time. The other opportunity for backward deletion involves the conjunction o f the nonconstituents SV and SV, and such constructions are vanishingly rare even when opportunities are provided. 8.3. Relation

to

Cognition

Greenfield and Dent (1982) explored the influence o f various aspects o f the nonlinguistic context on children's use o f conjunction reduction, using an older population o f subjects (6 years and 10 years). The subjects had to describe an action sequence o f putting differently colored beads i n a cup for the benefit o f a listener behind a screen who had to repeat the action. I n one condition, the subjects described the action as it was occurring ( " s i m u l t a n e o u s " ) , i n the second ( " p o s t " ) the subjects communicated after the action was completed. The utter­ ances were scored as forward or backward deletions and for the constituent that was conjoined. Their hypothesis was that certain features o f the communication situation have salience and motivate the use o f different syntactic structures. For instance, in the simultaneous condition, the subject might focus on the redundancy o f the repeated action, and hence omit redundant mentions o f the verb: She put the red in the yellow and the blue in the green. As a result, forward deletions should predominate, and they d i d . I n fact, some­ times this resulted i n nonstandard forms, such as: She putting a red bead in the yellow and a green one a blue one. where the locative phrase in the yellow

has been forward deleted by mistake.

I n contrast, when children were describing a completed action, the salience o f the perceptual grouping supposedly came into play, resulting i n more object conjunction and backward deletion o f the locative phrase, e.g. She put the blue red and green into the yellow one. Interestingly, forward deletions still predominated. Greenfield and Dent argue that the production process for phrasal coordina­ tions could not plausibly derive them from the corresponding sentential forms. Instead, the referential situation is directly encoded into the surface structure o f the sentence as i t is produced. The salience o f perceptual groupings, and the referential redundancy are responsible for the choice o f linguistic expression.

1.

The A c q u i s i t i o n of English

107

The question o f a production model for coordination is an interesting one, but the subjects i n Greenfield and Dent's study were already competent i n the requi­ site linguistic structures. Hakuta, de V i l l i e r s , and Tager-Flusberg (1982) propose a tentative model for younger children's production o f coordinate structures, from the assumption that the process underlying phrasal coordination consists o f conjunction o f constituents and proceeds i n a left-to-right fashion. A second area where cognitive achievements are likely to be significant lies i n the meaning relationships that hold between the connected propositions i n a conjunction. W o r k on acquisition o f these meanings by B l o o m et al. (1980) is discussed later under Interaction w i t h Semantics. 8.4.

Relation

to

Input

Data on the types o f coordination found i n parental speech to children are scarce. Only one study examined the forms i n the detail necessary, and then only for one mother-child pair (de V i l l i e r s , Tager-Flusberg & Hakuta, 1977). Nev­ ertheless, the result was intriguing. I n the same way as the researchers cate­ gorized Eve's coordinations for percentages o f sentential versus phrasal, forward versus backward deletions, they also categorized her mother's coordinations. I n plotting changes over time i n the proportions o f the different types, there was a striking resemblance between Eve and her mother. Forward phrasals heavily predominated, but the proportion dropped w i t h time, and forward sententials increased. Backward phrasals similarly outnumbered the corresponding senten­ tial forms. Finally, the number o f coordinations i n Eve's samples suddenly increased, and the mother's use o f coordination increased i n tandem. A trivial explanation w o u l d either be that Eve m i m i c k e d her mother's sentences, or her mother glossed Eve's, resulting i n a parallel change. However the contents o f their coordinated sentences were not closely matched: the parallelism existed at a more global level than individual sentences. U n t i l further data o f the same sort are analyzed, it is not clear whether Eve's mother was responding to changes i n Eve, or leading the way. Whichever is true, these preliminary data suggest that the biases i n children's spontaneous speech are closely mirrored i n the input they receive. A report from B l o o m et al. (1980) provided some evidence on children's ability to connect their sentences to preceding input sentences from the adult i n discourse. Their focus was on the meaning relationship encoded by the syntactic forms (see below) and their subjects proved more able to express complex meanings entirely i n their o w n speech, rather than i n conversational interchange w i t h an adult. I n contrast to the view that discourse provides the learning oppor­ tunities for linguistic expressions, B l o o m et al. claim that the child learns the linguistic forms o f the connective first i n his o w n speech, and can only then extend them to encode meanings expressed i n discourse between child and adult. However, an examination o f how the adult connected her response to the child's

108

De Villiers and De Villiers

statement, might display an increasing complexity as the child matured. These data are not presently available. 8.5.

Typical

Errors

and Individual

Differences

I n imitation tasks, the typical " e r r o r s " that children make i n coordinated sentences are elaborations o f phrasal forms into sententials, and reductions o f sentential forms into phrasals. However, these errors are not evenly distributed across sentence types: both Lust (1977) and Tager-Flusberg et al. (1982) reported an effect o f direction o f deletion, such that children do not reduce backward sententials, but more often reduce i n a forward direction. Even the latter may not be too frequent: 8% overall for Lust's subjects, 4 % for Tager-Flusberg et al.'s subjects. I n the latter study, the majority o f changes were made for sentences that had one redundant element deleted, but the potential for another deletion, e.g. Susy bought a necklace and bought a bracelet, or: Stephen chased the balloon and hit the balloon. Such linguistic structures sound particularly poorly motivated out o f context, as in an imitation task. I n a comprehension task, Tager-Flusberg et al. found t w o particular structures attracted many errors, the S V O + S O form, and the S V + S V O sentence. For example, i n response to: The zebra pushed and the zebra licked the alligator. 7 1 % o f the errors involved selecting another, unmentioned animal as the object o f push. I n response to: The sheep patted the kangaroo and the pig the giraffe. children acted out the first N V N sequence correctly, then were at a loss to k n o w what to do w i t h the remaining creatures! Errors i n spontaneous speech have not been reported, except by Greenfield and Dent from older subjects. I f children are w o r k i n g w i t h a deletion schema, it is surprising that there are not more errors o f deletion i n the w r o n g direction, such as: Mary went home and John, or: Likes apples and John lemons. The paucity o f errors favors the view that children are not producing phrasal coordinations by deleting elements from the sentential forms. Individual differences are a factor i n the spontaneous speech patterns reported here. For instance, i n Tager-Flusberg et al.'s study o f B r o w n ' s three subjects,

1.

The A c q u i s i t i o n of English

109

Eve had many more backward phrasals than A d a m or Sarah. B l o o m et al. (1980) found that for three o f their subjects, sentential and phrasal forms emerged simultaneously, for the fourth, phrasals emerged earlier. Explanations for these variations w i l l only emerge after more detailed study o f the input and contexts o f use to which individual children are exposed. 8.6.

Relation

to

Semantics

B l o o m et al. (1980) described the acquisition o f connective forms, and the meaning relations between connected clauses. The data came from the spon­ taneous speech o f four children studied longitudinally between 2 and 3 years. B y using contextual information, the researchers noted the meanings encoded by the conjoined sentences, identifying them as one o f eight varieties: 1. additive, e.g. Maybe you can carry that and I can carry this. 2. temporal, e.g. Jocelyn's going home and take her sweater off. 3. causal, e.g. She put a bandaid on her shoe and it maked her feel better. 4. adversative, e.g. 'Cause I was tired but now I'm not tired. 5. object specification, e.g. It looks like a fishing thing and you fish with it. 6. epistemic, e.g. I think that that's where the baby will go. 7. notice, e.g. Watch what I'm doing. 8. other, e.g. Tell Iris that I wet my bed.

110

De Villiers and De Villiers

The latter three types were mostly expressed as complements, to be discussed later. The mean rank order o f development was as above for the four children, w i t h some variation i n the order o f epistemic relations. The order o f emergence o f the connectives: and, and-then, when, because, what, so, then, if, but and that, was more variable across children. And, the first connective and the most frequent, was used w i t h the most different meaning relations. I t was used to encode conjunction w i t h the different meaning relations in the order: additive, temporal, causal, and then adversative. The other connec­ tives were less productive and tended to encode particular meaning relations. Finally, the syntactic structure used to encode these connected meanings was used first for coordination, then for relativization, and then for complementization. Clearly then, the meaning relation between the connected propositions is a significant variable i n the acquisition o f conjunction i n English, and the rela­ tionship o f the orders described here to cognitive development is a rich topic for crosslinguistic investigation. 8.7.

Pragmatics

As mentioned i n the beginning o f the section on coordination, certain condi­ tions must obtain before a phrasal form can be used rather than sentential f o r m . In particular, the redundant noun phrases must be identical i n reference i n order for one o f them to be omitted. Tager-Flusberg et al. note that one difference between Lust and Mervis (1980) and their o w n study o f spontaneous speech lies in the different way that sentential forms were treated. They argued that a true test o f the primacy o f sentential forms over phrasal forms requires counting only sentential forms w i t h a potential for deletion that is not realized. Lust and Mervis argue that any linguistic redundancy w i l l suffice; Tager-Flusberg et al. contend that the condition o f semantic redundancy must also h o l d . As one example o f dispute, take the sentence: This is Daddy's and this is Mommy's. where the referents o f this are distinct. Tager-Flusberg et al. contend that there is no possibility for phrasal coordination i n such a statement. I n an elicited produc­ tion task, they controlled the relevant referent variables that influence the like­ lihood o f sentential versus phrasal forms, to see whether young children were sensitive to this dimension. For instance, t w o contrasting pictures contained: A frog watching television and a turtle watching another television. and: A frog and a turtle watching the same television.

1.

The A c q u i s i t i o n of English

111

For all age groups from 3 to 6 years, the percentage o f phrasal coordination was much higher for the second type o f picture (73%) than for the first (13%). Furthermore, the children rarely made errors that involved deleting a l i n ­ guistically, but not referentially, redundant item, but instead produced sentences such as: A frog and a turtle watching television sets. Referential context thus had an important effect i n determining whether a sentential or phrasal f o r m was produced, w i t h sentential forms being used to mark the fact that linguistically identical NPs have different referents. This functional distinction between phrasal and sentential forms seems to operate early i n child language.

9. R e l a t i v e C l a u s e s Research on children's knowledge o f sentence embeddings has been motivated by a number o f concerns, not all o f them from linguistic theory. For example, some have tried to determine when the first attempts at sentence embedding occur. Other researchers have tried to describe general processing strategies that preschool children might use w i t h complex sentences. There has been no system­ atic work on the input children receive concerning relative clause constructions. Neither has there been much w o r k on prerequisite cognitive achievements, ex­ cept Greenfield et al. (1972) w h o argued for formal parallels i n the nonlinguistic domain o f embedding containers. Since this achievement occurs years before, it is not clear that i t is related to the corresponding linguistic achievement. The following issues for English revolve primarily around acquisition stages and typical errors. 9 . 1 . Order

of

Acquisition

H o w do relative clauses come into child speech? Hamburger and Crain (1982) argue that they appear i n the early speech o f 2 year olds, albeit i n structurally impoverished f o r m . For instance, they report sentences such as This is my did it. Look-a my made. as representing the first sentence embeddings. Initially the child embeds a V P i n a N P to restrict the referent set, then later learns to embed an S, still later includes the head noun, and eventually adds the complementizer. I n keeping w i t h these claims, F l y n n and Lust (1980) found headless relatives, e.g. Cookie monster hits what pushes Big Bird.

112

De Villiers and De Villiers

to be more successfully imitated than headed relatives, e.g. Big Bird pushes the balloon which bumps Ernie. in an elicited imitation task. They argue that the effect was not due to the semantic indeterminacy o f the headless relative, since indeterminate heads such as the thing which Mary's got were difficult for their subjects to imitate. The children seemed better able to nominalize a sentence when no separate N P head was involved, and could embed that nominalization as a noun phrase. Only by 6V2 years could children differenti­ ate the head and complementizer i n relative clauses, so F l y n n and Lust claim that full structural knowledge o f relative clause constructions does not come i n until late i n the preschool years. These results are i n keeping w i t h other data demonstrating the paucity o f relative clauses i n children's free speech before the age o f 6 years. I n fact Ingram (1971) proposed that the first spontaneous relative clauses may be stereotyped forms such as " a girl called C i n d e r e l l a , " but Hamburger and Crain's data stand in contradiction to that claim. Others (Limber, 1973, M e n y u k , 1971) have re­ ported that the first relative clauses are elaborations o f the object noun phrase rather than the subject noun phrase. However, that asymmetry could result from the fact that children's sentence subjects are generally pronouns (Limber, 1973) which do not require further specification. Tager-Flusberg (1982) used an elic­ ited production task w i t h preschool children, and demonstrated that when oppor­ tunities for object and subject relativization are equalized, children show no differential difficulty i n producing subject relatives. I n dative sentences, howev­ er, her subjects could only produce relatives on the last N P , the indirect object. The data on comprehension o f relative clauses partially support the argument that preschool children k n o w little o f the adult structure. However there are clearly different variables entering into these tasks, such that it is difficult to generalize results across performances. A s one example, i n F l y n n and Lust's parallel study o f the comprehension o f free versus headed relatives, the children found determinate, headed relatives easier than headless or nondeterminate head­ ed relatives. Hence the structure was not as important as the semantics for the comprehension task, the reverse o f the imitation task. Questions about the sensitivity o f the comprehension measure to structural variables pervade research on relative clause constructions, leading one re­ searcher (Maratsos, 1984) to discount that performance as providing useful data about syntactic knowledge. Maratsos argues that the strategies children adopt i n experimental tasks mask their true knowledge o f the construction under investi­ gation. I t is true that the experimental sentences are decidedly unnatural, since

1.

The A c q u i s i t i o n o f English

113

semantic clues are removed to study syntactic variables. Most investigators would agree w i t h Goodluck and Tavakolian (1982), who take the position that children's competence can be obscured by performance variables. I t is worth­ while to consider the rather long list o f task variables together w i t h the structural variables that influence comprehension. The t w o major structural variables are embeddedness, namely the role o f the complex N P i n the sentence, and focus, or the role o f the head noun w i t h i n the complex N P . Table 1.5 indicates the possible combinations that can occur w i t h only t w o roles: subject and object. The conventional labels for these types are the abbreviations listed i n the table. M a n y early investigators investigated only one structural variable, e.g. em­ beddedness, without controlling the second (for review, see de V i l l i e r s , TagerFlusberg, Hakuta & Cohen, 1979). However even among those researchers who carefully controlled embeddedness and focus, there is disagreement about the relative difficulty o f the four sentence types i n Table 1.5, as w e l l as the typical errors that are found. Sheldon (1974) theorized that sentence types SS and 0 0 w o u l d be more readily understood than SO and OS, since i n the former the head noun has a parallel function i n the complex N P and i n the matrix sentence. I n an act-out procedure w i t h toy animals, preschool children found SS and 0 0 easier than SO and OS. Tavakolian (1981) investigated comprehension o f the same four sentence types and argued against the parallel function claim. Instead, she argued that her subjects imposed a conjoined-clause analysis on the complex sentences, treating them as coordinates rather than embeddings. I n particular, SS sentences were well understood, since a conjoined-clause analysis corresponds to the correct interpretation i n this task. OS sentences were poorly understood, and i n fact were responded to as i f they were SS sentences. 0 0 and SO sentences were intermedi­ ate i n difficulty. Hence the order was: SS, O O , SO, and OS, though the latter three types were generally poorly understood.

TABLE 1.5 Types of Relative Clause

Role of Complex NP in Matrix Sentence (Embeddedness) subject subject object object

Role of Head Noun in Relative Clause (Focus) subject object subject object

Example

Abbreviation SS SO OS OO

The The The The

cat cat cat cat

that bit the dog that the dog bit bit the dog that bit the dog that

chased chased chased the cat

the rat. the rat. the rat. chased.

114

De Villiers a n d De Villiers

de Villiers et al. (1979) conducted a study rather similar to Sheldon's and Tavakolian's, using the same sentence types and act-out procedure, w i t h a larger sample o f subjects. Their data present a different order o f difficulty, primarily i n that OS was no more difficult than SS. The order for their subjects was: O S = S S , O O , SO. The error data supported neither parallel function nor a conjoinedclause analysis, but was best accomodated by a set o f processing heuristics, such as the N - V - N strategy, that children might adopt i n such a task. A g a i n , the percentages o f correct responses were all quite l o w . 9.2. Typical

Errors

As just described, there are several proposals concerning the typical errors that children make i n understanding sentences w i t h relative clauses. Sheldon (1974) contended that children were prone to maintain the same role o f a noun phrase i n both the relative clause and the matrix sentence. L e g u m (1975) sug­ gested this may not have a structural basis, but may instead be an artifact o f the act-out procedure, i n which the child uses a " b i r d - i n - t h e - h a n d " strategy to perform t w o actions w i t h the same toy. However, such errors were relatively rare in the studies by Tavakolian (1981) and de Villiers et al. (1979). Errors o f interpreting the relative clause as a conjoined clause are frequent i n Tavakolian's study, resulting i n very poor performance on OS sentences, de Villiers et al. found that to be the most prevalent error on OS sentences, but correct responses outnumbered those errors. I n general, they found children were more likely to ignore the syntactic marker that than to interpret it as and. The discrepancies among the studies might stem i n part from the averaging procedures: age is probably not the most revealing basis for grouping the sub­ jects. None o f the studies found significant interactions between age and perfor­ mance on the various sentence types. Perhaps correlating the child's performance to some other linguistic measure w o u l d reveal whether particular strategies or hypotheses are more prevalent at different stages o f development. T w o further studies have explored the variables that influence the conjoinedclause interpretation o f OS sentences. Goodluck and Tavakolian (1982) investi­ gated the influence o f the animacy o f the object, and the transitivity o f the verb i n the subordinate clause, i n an act-out procedure. Goodluck (1978) had earlier observed that OS relatives w i t h no animate object i n the subordinate clause were well comprehended, e.g. The cow kisses the horse that jumps over the fence. For the subjects tested by Goodluck and Tavakolian, sentences w i t h inanimate objects proved easier than sentences w i t h animate objects, and intransitives were easier than transitives. The most common error was the subject coreference error predicted by the conjoined clause analysis. Thus they speculated that a process­ ing load i n the standard comprehension test could mask children's competence i n

1.

The A c q u i s i t i o n of English

115

relative clause comprehension. I t is fairly obvious that restricting possible i n ­ terpretations o f N P roles w o u l d result i n improved performance. Their second experiment was designed to show the inadequacy o f a processing strategies account o f OS relatives and V P complements e.g. The dog tells the horse to knock over the table. and argued that their evidence demonstrates that even 4-year-olds analyze rela­ tive clauses as constituents o f the N P node. I n particular the authors argue that children's interpretations are governed by the principle o f c-command, a struc­ tural condition. The node A on a phrase structure tree is said to " c - c o m m a n d " a second node B , i f the first branching node that dominates A also dominates B , and neither A nor B dominates the other. The c-command condition, held to be universal, states that a lexical N P must c-command an empty node w i t h w h i c h it is co-indexed. U p o n hearing either a relative clause sentence or a complement sentence, the child must find a coreferent N P for the empty N P i n the subordinate clause. Goodluck and Tavakolian argued that i f one assumes that their subjects' choice o f coreferent N P was governed by the principle o f c-command, then their responses on the tasks reveal adult-like structures for the relative clause construc­ tion. As w i l l be discussed later under complements, the assumption o f the ccommand condition also constrains the authors to the position that children's temporal complements are mis-analyzed as V P constituents rather than being attached to the S-node as i n adult grammar. I t is not clear from the present evidence that the assumption is warranted. Solan and Roeper (1978) also argue strongly that preschool children's com­ prehension o f relative clauses is constrained by structural principles, not just by processing strategies. A g a i n exploring OS sentences only, they compared the likelihood o f a subject coreference error i n the t w o sentence types: The cat put the cow that kicked the dog in the barn. versus The cat pushed the cow that kicked the dog in the barn. Notice that the first sentence containing the verb put, requires specification o f three N P roles: the subject, object, and object o f the preposition. Hence i n that sentence, the PP is clearly attached to the verb phrase node. Their subjects were pretested on sentences w i t h put, and preselected to k n o w that subcategorization requirement. However, then the relative clause that kicked the dog could not be attached to the subject N P without creating crossed branches i n the phrase struc­ ture tree (see F i g . 1.1). I f children were sensitive to this no-tangle constraint on phrase structure rules, they should not make a subject coreference error for the put sentence. I n fact, the children made no such errors, but failed to act out the relative clause at a l l . Conjoined clause responses were frequent for the sentence

116

De Villiers and De Villiers

s

the cat

put

the cow

that

A

kicked

the dog

that

A

kicked the dog

in

the barn

S

the cat pushed the cow

in

the barn

FIG. 1.1.

containing pushed, i n w h i c h the PP can attach to the S-node too (see F i g . 1.1). Solan and Roeper claim that the experiment demonstrates that the typical error depends on structural considerations, rather than being a processing strategy o f no structural import. I t could be argued, however, that their subjects simply ignored the relative clause while searching for the final N P argument for put, that is, albeit ad hoc, a processing account o f the phenomenon. 9.3.

Pragmatics

The above experiments all suffer from one major weakness, according to Hamburger and Crain (1982). I n the toy manipulation paradigm i t is customary to supply only one instance o f each animal mentioned i n the sentence, and to ignore the order i n w h i c h the child acts out the clauses. The sentences then

1.

The A c q u i s i t i o n of English

117

become nonrestrictive relative clauses, akin to simple conjunction. The real function o f the relative clauses, i n contrast, is to disambiguate the N P referent by giving it further specification. Hamburger and Crain argue that violation o f these felicity conditions for relative clause use might be an additional factor i n pre­ school children's failure to understand them. These authors contend also that the use o f the present tense is unnatural also, so they use the past tense, like Solan and Roeper (1978), Sheldon (1974), and de Villiers et al. (1979). Unfortunately the only sentences considered by Hamburger and Crain were OS relatives, on the rationale that this type "presented the most difficulty for children i n all the previous studies" (p. 264). I n addition, the order o f mention i n an OS sentence is opposite to the conceptual order, and they wanted to see whether under improved felicity conditions, children w o u l d act out the subordi­ nate clause first. I n fact 3-year-olds i n their study used an order-of-mention strategy, but their 5-year-olds acted out the relative clause first. Subject coreference errors were not frequent. Although a major conclusion o f the study is that the appropriate felicity conditions resulted i n improved performance by their 5-year-olds, their data reveal that 55% o f the " c o r r e c t " responses made by 5year-olds involved acting out only the main clause o f the OS relative. Hamburger and Crain counted this as correct, on the argument that the presupposition en­ coded by the relative clause could be assumed to have occurred already, so only the assertion needed to be acted out. I f other researchers had counted matrix clause responses as correct, their subjects w o u l d look equally proficient. There is thus no guarantee that improved felicity conditions are as beneficial as one might suppose. 9.4.

Summary

In sum, a large number o f structural and task variables interact i n determining children's performance on relative clause constructions. Some children make subject coreference errors as i f they treated the relative clause structure as a flat, coordinate structure rather than an embedding. However this type o f response is not universal (de Villiers et a l . , 1979, Hamburger & Crain, 1982) and appears more prevalent under a variety o f processing overloads or pragmatic infelicities. Researchers are i n substantial disagreement about the extent o f children's struc­ tural knowledge i n the preschool years, w i t h Goodluck and Tavakolian, Solan and Roeper taking the position that children's knowledge o f relatives is w e l l developed, obedient to various structural constraints (c-command, no-tangle constraint) but masked b y performance factors. Others (Flynn & Lust, 1980; de Villiers et a l . , 1979) stress the incompleteness o f children's structural knowledge before age 6 or so, and Hamburger and Crain emphasize the piecemeal, step­ wise character o f relative acquisition i n early childhood. The time is ripe for a detailed longitudinal study o f a number o f children, tested on a range o f tasks at frequent intervals, to elucidate the paths children can take to an adult knowledge of this form o f sentence embedding.

118

De Villiers a n d De Villiers

10. V e r b C o m p l e m e n t s Recently, increased interest has developed i n how children deal w i t h complex sentences. M u c h o f the w o r k i n this area is motivated by considerations o f learnability w i t h i n current linguistic theory, for example, deriving tests o f c h i l ­ dren's knowledge o f putative universal constraints (e.g. Tavakolian, 1981; Phinney, 1981; Otsu, 1981). A number o f investigators have used various kinds o f complements as an avenue o f exploration, and some o f those studies are re­ viewed here. Primarily because o f the motivation for the research, but also because o f its recency, there has been virtually no w o r k on the nature o f the input, on related cognitive achievements, on individual differences, on semantic constraints or on the pragmatics o f the forms. Attention has instead been directed towards the course o f acquisition, questions o f reorganization i n the grammar, and typical errors made at different ages. 1 0 . 1 . Course

of

Acquisition

Apart from the w o r k on relative clauses, research has concentrated on the development o f infinitival complements, w i t h only a little on gerundive comple­ ments and embedded tensed clauses. B l o o m , Tackeff and Lahey (1982) studied the development o f to i n predicate complement constructions i n four children studied longitudinally, whose spon­ taneous speech comprised the data base. A l l o f the utterances containing se­ quences o f V - t o - V were considered, e.g. I'm going to see Mommy. The purpose o f the study was to examine the distributional regularities o f to: i f it were being learned as a marker o f the infinitive, then the regularities i n distribu­ tion should occur w i t h the complement verbs. However, to instead occurred regularly w i t h particular matrix verbs, and i n no pattern w i t h the complement verbs. Initially, the antecedent verbs were a small set that appeared to function as modal verbs, e.g. I want open it. I gonna get it. The to was either absent, or marked w i t h a schwa. Later, nonmodal matrix verbs e.g. try, like, entered the children's speech, and these usually had to after them. Only then d i d the modal verbs begin to be used w i t h to rather than schwa. I n contrast, the complement verbs d i d not seem to be conditioning use o f to at a l l . That is, the presence o f to was not affected by the f o l l o w i n g verb but by the verb preceding i t . For instance, like to might be established as a frame, but to eat

1.

The A c q u i s i t i o n of English

119

w o u l d not be. Furthermore, the sequence V - r o - V was only very rarely inter­ rupted by an intervening N P (only 4 % o f contexts) and even then the forms were highly constrained, e.g. Want you to V. B l o o m et al. point out that this course o f acquisition calls into question the traditional account o f infinitives i n w h i c h to is the marker o f that f o r m . Instead, it is i n keeping w i t h Bresnan's (1978) analysis i n w h i c h verbs are subcategorized to take an infinitival complement, since the children seemed to learn the basic structure V E R B + to. Hafitz, Gartner and B l o o m (1980) report similar findings on other forms o f verb complements f o l l o w i n g the verbs see, look, know, think, say and tell, i n w h i c h the complementizers (e.g. wh) were acquired specific to individual verbs and only later generalized. T w o other studies suggest that children's knowledge o f verb subcategorization for complements is not complete until much later. Phinney (1981) studied the emergence o f the distinction between verbs such as wish, which block subject extraction from the infinitival complement, and verbs such as want w h i c h allow it. Compare the possible subjects o f kick i n : Who did the bear wish to kick? versus: Who did the bear want to kick? I n the second sentence there is the possibility that the bear wants someone else to do the k i c k i n g . Phinney found that 6-year-olds had not yet made a distinction between the verb types though older children were beginning to differentiate them. Goodluck and Roeper (1978) tested children's knowledge o f a distinction between perception and nonperception verbs, namely that i n the sentence John saw Bill sitting on the beach. the subject o f sitting

is ambiguous, but i n

John hit Bill sitting on the beach. there is no ambiquity. Five-year-olds had not acquired this distinction, even though it is one based on semantics. I t should be pointed out that some adult native speakers have doubts about the existence o f the distinction. Admittedly these distinctions among verb types are more subtle than the information about w h i c h verb takes which form o f complement. Children may encode the latter i n the lexicon, but they clearly make generalizations across matrix verbs i n their rules for interpreting complement subjects. I t is to an account o f those generalizations that we now turn.

120

10.2.

De Villiers and De Villiers

Typical

Errors

Most o f the other research on verb complements has addressed the issue o f interpreting the " n u l l " subject o f the complement, beginning w i t h the experi­ ment by Chomsky (1969). She asked children questions i n v o l v i n g the matrix verbs ask, tell, and promise. A common tendency among children between 5and 9-years-of-age, was to interpret the subject o f the complement verb as coreferential w i t h the object o f the matrix verb, hence making an error on the exceptional verb promise. Compare: Bill told Bert to mow the lawn. Bill promised Bert to mow the lawn. The children w o u l d believe that Bert m o w e d the lawn i n both sentences. Further­ more, they had most difficulty w i t h the verb ask, w h i c h is inconsistent i n assign­ ment o f reference: compare e.g. Bill asked to go. Bill asked Bert to go. Bill asked Bert where to throw the ball. Chomsky interpreted her findings i n terms o f a m i n i m a l distance principle, or M D P (Rosenbaum, 1967) such that children take the nearest noun to be the subject o f the complement verb. In later w o r k Maratsos (1974) challenged Chomsky's account by using pas­ sive versions o f the same sentences i n which the M D P could not account for the results. Even i n sentences such as: Bill was promised by Bert to mow the lawn. his subjects believed Bill, subject o f mow.

the logical object but not the nearest N P , to be the

Goodluck (1981) argued that the available data on children's interpretation o f complement sentences can be understood i f i t is assumed that children under­ stand the constraint k n o w n as c-command. The c-command constraint described earlier specifies how NPs are to be co-indexed i n a phrase structure, hence it is a constraint on the interpretation o f anaphora (Solan, 1981) and null subjects o f complements and relative clauses (Goodluck & Tavakolian, 1982). I n particular it specifies that the subject o f the complement can only be coreferential w i t h the object o f the matrix sentences i f the complement is attached to the V P , not the Snode. So, for instance, B i l l can be c o - i n d e x e d w i t h the empty subject node o f mow i n : John told Bill to mow the lawn.

1.

The A c q u i s i t i o n of English

121

since the infinitival complement is part o f the V P (see F i g . 1.2). But i n the sentence: John hit Bill after mowing the lawn. the participial complement is attached to the S-node i n adult grammar, so John must be its subject, not Bill. Verbs like promise constitute an exceptional case. The second ramification o f the c-command condition is that the object o f a prepositional phrase cannot be coreferential w i t h the subject o f the complement, e.g. i n John was asked before Bill to mow the lawn. Bill could not be the subject o f mow. I f passive by-phrases are base-generated prepositional phrases (Bresnan, 1978), then interpretation o f Bill as mower w o u l d also be blocked i n :

122

De Villiers and De Villiers

John was asked by Bill to mow the lawn. hence accounting for Marastos' results. Goodluck explored 4-, 5- and 6-year-old children's interpretation o f par­ ticipial complements such as: The boy hits the girl after jumping over the fence. as well as sentences w i t h prepositional phrases, and passive matrix sentences. Unfortunately, i n apparent violation o f c-command, children tended to interpret the girl as the subject o f jumping i n the above sentence. Goodluck believes the independent evidence for c-command conditions i n children's grammar (Solan, 1978 on anaphora) is so compelling that she prefers to argue that children's phrase structure for that sentence is not the same as adults'. That is, rather than contend that children are violating c-command, she claims that they have misanalysed that participial complement as a constituent o f the V P rather than the S node. O f course then, their interpretation does not violate c-command. However given the weight o f evidence i n favor o f children's phrase structures being flatter than adults', w i t h a strong tendency to attach constituents to the S-node (Tavakolian, 1981) it is difficult to accept this argument. When the complements were preposed, e.g. After jumping the fence, the boy hits the girl. children persisted i n interpreting the girl as the subject o f jump, again demon­ strating the inadequacy o f the M D P account. The second consequence o f ccommand, that the object o f a prepositional phrase should not be the subject o f the complement, was not entirely borne out either. Some o f the younger subjects allowed that interpretation, though such responses declined w i t h age. Goodluck's position is that the appropriate phrase structures for participial comple­ ments are only beginning to emerge at age six, but that position is a consequence of assuming that children are constrained by c-command conditions. Evidence independent o f that assumption w o u l d be more convincing. It does appear that young children prefer to interpret the object o f the matrix sentence as the subject o f the complement verb. Their individual spontaneous complements are all subject-controlled, but have no surface object ( B l o o m et a l . , 1982) e.g. I want to go home. I need to have that. Perhaps children restrict subject control to sentences w i t h no direct object on the surface. I f a direct object is present, even i f it is not adjacent to the complement

1.

The A c q u i s i t i o n of English

123

verb, it tends to be interpreted as the complement's subject. Even i n ' i n order t o ' sentences such as: Daisy hits Pluto to put on the watch. Goodluck found a sizeable proportion o f object control interpretations. What is missing at present is a careful analysis o f individual, rather than group, response patterns, to check on the consistency o f individuals i n their interpretations across different sentences. A t present then, children's errors i n interpreting complement subjects have several competing explanations, but Goodluck's account i n terms o f c-command is potentially more integrative than the M D P or Maratsos' (1974) account i n terms o f semantic relations. 10.3. Nonoccurring

Error

I n one last study using complement constructions as a vehicle for exploring children's knowledge o f constraints, Phinney (1981) looked at embedded tensed clauses, such as the sentence: The boy believed the girl swam in the pond. A constraint operates against extracting subjects from tensed complements when a complementizer e.g. that is present but extracting objects is not blocked (Bresnan, 1972). The test compared children's responses to sentences such as: la. Who did the lion know swam in the pond? lb. Who did the horse know that ate the hay? which have a gap i n the subject position o f the embedded clause. I n l a , however, the wh-pronoun refers to the subject o f the embedded verb. I n l b , the w h prounoun must refer to the object o f the matrix verb know, and the clause at the end is an extraposed relative. There is no difference, though, between the corre­ sponding questions concerning the object, regardless o f the presence o f that. 2a. Who did the bear see the cow kicked? 2b. Who did the dog notice that the rooster kissed? The sentences i n the test were all w e l l motivated by short cartoon stories w h i c h offered several alternatives for the referent o f who, and allowed the researcher to determine how the subjects were interpreting the sentence. For example, they might process only the simple sentence Who did the bear see?

124

De Villiers and De Villiers

and give one answer. Or, they could process it as a sentential complement, and give a different answer. A third alternative indicated that they had processed it as an extraposed relative clause. The hypothesis was that l b could not be given a sentential complement reading as this w o u l d violate the constraint on extraction. Indeed, the proportion o f sentential readings was very l o w compared to the other three sentences. Y o u n g children tended to process it as a simple sentence; only by grade t w o d i d a significant number (60%) o f subjects respond to it as a relative clause, the adult reading. These results suggest that children do know the specific constraint blocking the sentential reading o f the critical sentnece. Phinney argues that this is not explicable i n terms o f a more general constraint against extraction, since l a , 2a, and 2b were w e l l understood. The claim that children are i n possession o f the appropriate constraint (not easily derivable from input evidence) is an impressive one that deserves further exploration. For present purposes, it is clear that the full grammar o f complementation including extraposed relatives, is not completed until at least grade t w o .

CONCLUSION I f we liken the research on the acquisition o f English to a gigantic jigsaw puzzle, how near is it to completion? Armies o f graduate students are j o i n i n g pieces together i n the vast middle terrain, where no clear benchmarks exist. Their mentors, meanwhile, have joined the easy edge pieces, but have come up w i t h a bewildering array o f competing frames. Particular areas o f the puzzle, like Stage I speech, promise to hold the key and have thus attracted an undue amount o f attention. Unfortunately the workers do not seem agreed upon w h i c h side o f the pieces represent the puzzle. Some prefer to w o r k w i t h the cardboard side where false clues are minimized; others use the picture side and cut the pieces to suit their o w n needs. I n other areas (later grammar?) the pieces have not all been found, but the researchers have painted the surface beneath the puzzle to create the illusion o f completion. In a task as large and multifaceted, each researcher must find a set o f guiding principles, or theoretical questions, to make sense o f the whole enterprise and to which new findings can be assimilated. W e present here our o w n personal view of the field, not as a theoretical stance but as a framework o f questions to w h i c h we continually seek answers. Coloring the entire enterprise is the issue o f nativism and empiricism. Chomsky is responsible for the modern revival o f this ancient debate, and his views find representation i n all the areas we have discussed. For example, i f children have an innate propensity to select structure dependent rules, they might be prepared from the start to encounter abstract linguistic categories such as subject or N P , and to ignore the meaning o f the terms i n search o f syntactic

1.

The A c q u i s i t i o n o f English

125

regularities. The child should entertain a limited set o f hypotheses about gram­ mars, constrained by innate knowledge o f the forms that rules can take, and the input should provide no especial assistance i n this regard over and above " e x ­ posure to the language." Certain errors predictable on any theory o f "gener­ alization" or " a n a l o g y " should never appear, and for all significant purposes, children could be regarded as linguistic replicas o f one another. On the other hand, i f the child is equipped w i t h very little i n the way o f linguistic hypotheses, induction o f the rules from the input language must receive support from nonlinguistic achievements. Since a prelinguistic child has some skills i n interpreting the w o r l d and engaging i n social interaction, perhaps he can use this knowledge to help crack the linguistic code. Such a learner w o u l d be searching the input for convenient ways to express the meanings he already knows, and an account o f his acquisitions w i l l be derived from a description o f the input and his nonlinguistic concepts. I n this search the child may go through several early rule systems, revising them as the data base, the child's cognitive ability, and communicative needs grow. O f course, individual children w i l l be exposed to different inputs, use language for different purposes, and hence variation w i l l be the n o r m . This is the way the boundaries have been drawn, but not all researchers are restricted by them. One doesn't have to be a nativist to be intrigued by nonoccurring errors, or be a learning theorist to be interested i n the input. Nonetheless, the theoretical positions have influenced the way research has proceeded over the last ten or fifteen years, and provide the background to our discussion o f the remaining questions. The first very general question concerns the independence o f semantics and syntax i n an account o f language acquisition. D o syntactic categories have a semantic basis i n early childhood? Does the child exploit meaning i n his search for early rules? M a n y different theorists have been attracted to the notion that children beginning to learn the language w i l l formulate rules on the basis o f the semantics o f the terms involved, rather than abstract categories. I n Stage I , the special attraction is that such rules w o u l d capitalize on the cognitive achieve­ ments o f the sensorimotor period, the termination o f which coincides w i t h the onset o f language. However problems remain w i t h semantic descriptions o f early speech. One is empirical: some children do not seem to go through a stage that can be described i n general semantic terms; like Bowerman's Eva, w h o went from word-specific formulae to apparently adult-like, nonsemantic rules. The second is methodological: any particular utterance can be given an infinite number o f semantic descriptions: what are the constraints i n w r i t i n g semantically based grammars? The third is theoretical: i f we agree that children begin w i t h semantically based grammars, do we agree that they have to give them up at a later point? (see Gleitman & Wanner, 1982). I n later stages o f development, the data suggesting children make use o f semantically based rules are not terribly strong, and can be given alternative

126

De Villiers and De Villiers

accounts i n terms o f processing or vocabulary limitations. For instance, the data on restrictions on the passive rule are equivocal at present (Maratsos et a l . , 1979; Lebeaux & Pinker, 1981). I f children's formulation o f the category " v e r b " for the purpose o f marking w i t h the past tense is initially semantic, it seems to be rapidly revised (Maratsos & Chalkley, 1980). Finally, there are even instances where the child could make use o f a broad semantic distinction that does corre­ late w i t h adult grammatical distinctions, but fails to notice it (Goodluck & Roeper, 1978). Theoretical considerations have led other researchers to consider the possibility not that the child formulates rules i n semantic terms, but that she uses semantics as a " b o o t s t r a p " (Pinker, 1982) or " t r i g g e r " (Roeper, 1982) to the grammatical categories themselves. A second issue relates to the first i n its concern w i t h general rule formula­ tions. D o children seek very general rules, or do they acquire the rule i n a piecemeal fashion, perhaps verb-by-verb? (Kuczaj, 1982). Maratsos (1984) ar­ gues that there is plenty o f evidence that adult grammars require a great deal o f term-specific information, so both mechanisms must be i n operation throughout life. Roeper et al. (1981) argue that children ought to be cautious i n extending rules i n the absence o f positive evidence (see also Baker, 1979). Evidence from early Stage I supports the view that many children are learning specific formulae for making sentences containing particular terms (Braine, 1976). I n inflectional morphology, the evidence is less clear cut: generalization across terms is prevalent, but the positive evidence is probably present i n the input too. Learning o f auxiliary placement i n wh-questions seems to be a piece­ meal acquisition (Kuczaj & Brannick, 1979). Children's formulation o f the dative and passive rules may not await positive evidence on particular verbs (Lebeaux & Pinker, 1981; W i l s o n et a l . , 1981), though learning w h i c h comple­ mentizers can follow particular verbs may be verb-specific (Hafitz et a l . , 1980). To address the question properly, we need to accumulate data on when overgeneralization occurs, and when undergeneralization o f a rule is the typical pattern. D o different constructions or rules show different tendencies i n this regard, or are methodological differences i n the way they are studied responsible for the different results? A final important consideration is the possibility that individual children might be bold or cautious i n rule generalization, and this style might be found throughout the language system. Researchers need to be sen­ sitized to the possibility o f undergeneralization i n rule learning just as they were to the phenomenon o f underextension i n vocabulary acquisition ( A n g l i n , 1977). Both domains have benefitted from the introduction o f systematic tests o f gener­ alization to test the limits o f the child's knowledge. I n the area o f complex syntax, clever tests o f generalizations that should be limited by universal con­ straints are just beginning to be devised (Otsu, 1981; Phinney, 1981). W h e n more o f these data are available, it w i l l be possible to evaluate the various theories that predict generalization o f different sorts.

1.

The A c q u i s i t i o n of English

127

On some accounts, the child might have established abstract categories and thus i f a syntactic construction is exemplified w i t h a k n o w n member o f the category, the rule should be formulated i n general terms and apply to all mem­ bers. A lexicalist perspective w o u l d predict more caution: just because one verb, for instance, permits that construction does not guarantee that other verbs w i l l . Hence a child should be alert for lexical information or subcategorizations for the rule. A third possibility, hypothetical at present, is that syntactic categories contain degrees o f membership based not on semantics, but on distributional properties, for example as proposed by Ross (1972) (see Table 1.6). The rows refer to verbs, ordered by the number o f different rules (the columns) that they enter. The rules are ordered i n terms o f choosiness, that is, the variety o f verbs that they permit. Note that once such a matrix o f possibilities is learned, the verbs become differentially informative about new rule generalizations. I f a new rule is learned w i t h respect to V I , the child does not k n o w whether it can be generalized like R l , or is choosy, like R4. However i f a new rule is learned w i t h respect to V 4 , the child can be certain that V I , V 2 , and V 3 are also allowed. Thus asymmetries o f generalization could be predicted depending upon the " t y p ­ i c a l i t y " o f the input examples to the category (see de V i l l i e r s , 1979, and 1980). A l l o f these theoretical possibilities await new data. The equivalence o f children's and adults' grammars is also a central question for language acquisition research. Gleitman and Wanner (1982) point out that i f we accept that early grammars are semantically based rather than syntactically based, we then require a discontinuous theory o f learning. They take i t as given that adult grammar is not semantically based, though there is dispute about that too (e.g. Bates & M a c W h i n n e y , 1982; Schlesinger, 1982). As pointed out ear­ lier, even i f children's early grammars are semantically grounded, at least by age four or so their language presents evidence o f abstract categories. However the acquisition o f complex syntactic constructions continues w e l l beyond the pre-

TABLE 1.6 Choosiness Rl

R2

R3

R4

R5

VI

+

+

+

+

?

V2

+

+

+

V3

+

+

V4

+

fypica

t

(After Ross, 1972.)

? 7

?

128

De Villiers a n d De Villiers

school years, and children's grammars are still not equivalent to adults' for embedded structures at age six or seven (see pp. 1 1 4 - 1 2 4 ) . Nonetheless, claims have been made that although the phrase structure may not be correct, children's rule formulations are obedient to universal constraints (c-command, no-tangle constraint, A - o v e r - A principle) and hence not significantly deviant from adult grammars. However since many o f these discoveries are quite new, and heavily dependent upon the methodology o f comprehension tasks (see Maratsos, 1984), it w i l l be some time before they are accepted as established facts about language development. The phenomena i n this area w o u l d seem a particularly appropriate area for crosslinguistic investigation. Finally, we must return to the problems o f competence versus performance. I n language acquisition, the competence/performance distinction is used by re­ searchers o f all persuasions to try and define what constitutes data and what constitutes noise. T o take a couple o f examples: the error He did broke it is regarded by one group o f researchers (Mayer et a l . , 1978) as a significant piece of evidence that transformations are acquired i n units o f basic operations. T o others, it is a mere slip, a random error o f the speech mechanism. When the first subjects that children use i n sentences turn out to refer primarily to animate agents, it is seen as evidence that their rules are based on such categories as agent and action rather than subject and verb (e.g. Bowerman, 1973). However, it could be argued that the restriction is a performance limitation: very young children don't make reference to more abstract subjects or have the vocabulary to do so, but their rule formation may not be semantically tied (Valian, 1984). Obviously, theoretical convictions determine when and how the "performance" criticism is raised, but the possibility o f that criticism should keep researchers methodologically sophisticated and alert to alternative accounts o f their favorite phenomena.

REFERENCES Akmajian, A . , & Henry, F . Introduction to the principles of transformational syntax. Cambridge, MA: MIT Press, 1975. Anglin, J . M. Word, object and conceptual development. New York: Norton, 1977. Antinucci, F . , & Parisi, D. Early language acquisition: A model and some data. In C . A. Ferguson & D. I. Slobin (Eds.), Studies of child language development. New York: Holt, Rinehart & Winston, 1973. Antinucci, F . , & Volterra, V . Lo svilippo della negazione nel linguaggio infantile: Un studio pragmatico. In Studi per un modello del linquaggio: Quaderni della Ricerca Scientifica, Rome, 1973. Ardery, G . On coordination in child language. Journal of Child Language, 1980, 7, 305-320. Baker, C . L . Syntactic theory and the projection problem. Linguistic Inquiry, 1979, 10, 533-583. Baldie, B. J . The acquisition of the passive voice. Journal of Child Language, 1976, 3, 331-348. Bates, E . Language and context. New York: Academic Press, 1976.

1.

T h e A c q u i s i t i o n of English

129

Bates, E . , Benigni, L . , Bretherton, I . , Camaioni, L . , & Volterra, V. From gesture to first word: On cognitive and social prerequisites. In M. Lewis & L . Rosenblum (Eds.), Interaction, conversa­ tion and the development of language. New York: Wiley, 1977. Bates, E . , Benigni, L . , Bretherton, I . , Camaioni, L . , & Volterra, V . The emergence of symbols: Communication and cognition in infancy. New York: Academic Press, 1979. Bates, E . , & MacWhinney, B. Functionalist approaches to grammar. In E . Wanner & L . R. Gleitman (Eds.), Language acquisition: The state of the art. New York: Cambridge University Press, 1982. Bellinger, D. Changes in the explicitness of mother's directives as children age. Journal of Child Language, 1979, 6, 443-458. Bellugi, U . The acquisition of negation. Unpublished doctoral dissertation, Harvard University, 1967. Berko, J . The child's learning of English morphology. Word, 1958, 14, 150-177. Bever, T. G . The cognitive basis for linguistic structure. In J . R. Hayes (Ed.), Cognition and the development of language. New York: Wiley, 1970. Blank, M. Mastering the intangible through language. In D. Aaronson & R. W. Rieber (Eds.). Developmental psycholinguistics and communication disorders. Annals of the New York Acade­ my of Sciences, 1975, 263, 44-58. Block, E . M . , & Kessell, F . S. Determinants of the acquisition order of grammatical morphomes: A re-analysis and re-interpretation. Journal of Child Language, 1980, 7, 181-188. Bloom, L . M. Language development: Form and function in emerging grammars. Cambridge, MA: MIT Press, 1970. Bloom, L . M. One word at a time: The use of single word utterances before syntax. The Hague: Mouton, 1973. Bloom, L . M. Talking, understanding and thinking. In R. L . Schiefelbusch & L . L . Lloyd (Eds.), Language perspectives: Acquisition, retardation and intervention. Baltimore: University Park Press, 1974. Bloom, L . M . , & Lahey, M. Language development and language disorders. New York: Wiley, 1978. Bloom, L . M . , Lahey, M . , Hood, L . , Lifter, K . , & Fiess, K . Complex sentences: Acquisition of syntactic connectives and the semantic relations they encode. Journal of Child Language, 1980, 7, 235-261. Bloom, L . M . , Lifter, K . , & Hafitz, J. Semantics of verbs and the development of verb inflections in child language. Language, 1980, 56, 386-412. Bloom, L . M . , Lightbown, P., & Hood, L . Structure and variation in child language. Monographs of the Society for Research in Child Development, 1975, 40 (2), Serial No. 160. Bloom, L . M . , Takeff, J . , & Lahey, M. Learning to in complement constructions. Unpublished manuscript, 1982. Bowerman, M. F . Structural relationships in children's utterances: Syntactic or semantic? In T. E . Moore (Ed.), Cognitive development and the acquisition of language. New York: Academic Press, 1973. Bowerman, M. F . Discussion summary—development of concepts underlying language. In R. L . Schiefelbusch & L . L . Lloyd (Eds.), Language perspectives: Acquisition, retardation and inter­ vention. Baltimore, MD: University Park Press, 1974. Bowerman, M. F . Commentary on "Structure and variation in child language" by L . Bloom, P. Lightbown, & L . Hood. Monographs of the Society for Research in Child Development, 1975,40 (2), Serial No. 160. Bowerman, M. F . Semantic factors in the acquisition of rules for word use and sentence construc­ tion. In D. Morehead & A. Morehead (Eds.), Directions in normal and deficient child language. Baltimore, MD: University Park Press, 1976.

130

De Villiers a n d De Villiers

Bowerman, M. F . Words and sentences: Uniformity, individual variation, and shifts over time in patterns of acquisition. In F . D. Minifie & L . L . Lloyd (Eds.), Communicative and cognitive abilities: Early behavioral assessment. Baltimore: University Park Press, 1978. Bowerman, M. F . Reorganizational processes in lexical and syntactic development. In E . Wanner & L . R. Gleitman (Eds.), Language acquisition: The state of the art. New York: Cambridge University Press, 1982. Braine, M. D. S. The ontogeny of English phrase structure: The first phase. Language, 1963, 39, 1¬ 13. . Braine, M. D. S. Children's first word combinations. Monographs of the Society for Research in Child Development, 1976, 41(\): Serial No. 164. Bresnan, J . Theory of complementation in English syntax. Unpublished doctoral dissertation, MIT, 1972.. Bresnan, J . A realistic transformational grammar. In M. Halle, J . Bresnan, & G . A. Miller (Eds.), Linguistic theory and psychological reality. Cambridge, MA: MIT Press, 1978. Bridges, A. SVO comprehension strategies reconsidered: The evidence of individual patterns of response. Journal of Child Language, 1980, 7, 89-104. Bronckart, J . , & Sinclair, H . Time, tense and aspect. Cognition, 1973, 2, 107-130. Brown, I. Role of referent concreteness in the acquisition of passive sentence comprehension through abstract modeling. Journal of Experimental Child Psychology, 1976, 22, 185-199. Brown, R. A first language: The early stages. Cambridge, MA: Harvard University Press, 1973. Brown, R . , & Bellugi, U . Three processes in the child's acquisition of syntax. Harvard Educational Review, 1964, 34, 133-151. Brown, R . , & Fraser, C . The acquisition of syntax. In C . N. Cofer & B . Musgrave (Eds.), Verbal behavior and learning: Problems and processes. New York: McGraw-Hill, 1963. Bruner, J . S. The ontogenesis of speech acts. Journal of Child Language, 1975, 2, 1-19. Bruner, J . S. Early social interaction and language acquisition. In H. R. Schaffer (Ed.), Studies in mother-infant interaction. London: Academic Press, 1977. Bryant, B . , & Anisfeld, M. Feedback versus no-feedback in testing children's knowledge of English pluralization rules. Journal of Experimental Child Psychology, 1969, 8, 250-255. Bybee, J . L . , & Slobin, D. I. Rules and schemas in the development and use of the English past tense. Language, 1982, 58, 265-289. Cairns, H. S., & Hsu, J . R. Who, why, when and how: A developmental study. Journal of Child Language, 1978, 5, 477-488. Cazden, C . The acquisition of noun and verb inflections. Child Development, 1968, 39, 433-448. Chafe, W. L . Meaning and the structure of language. I L : University of Chicago Press, 1970. Chapman, R. S. Comprehension strategies in children. In J. Kavanagh & W. Strange (Eds.), Speech and language in the laboratory, school and clinic. Cambridge, MA: MIT Press, 1978. Chapman, R. S., & Miller, J . F . Word order in early two and three word utterances: Does produc­ tion precede comprehension? Journal of Speech and Hearing Research, 1975, 18, 355-371. Chomsky, C . The acquisition of syntax in children from 5 to 10. Cambridge, MA: MIT Press, 1969. Chomsky, N. Syntactic structures. The Hague: Mouton, 1957. Chomsky, N. Aspects of the theory of syntax. Cambridge, MA: MIT Press, 1965. Clark, E . Strategies and the mapping problem in first language acquisition. In J. Macnamara (Ed.), Language learning and thought. New York: Academic Press, 1977. Clark, H . H . , & Chase, W. G . On the process of comparing sentences against pictures. Cognitive Psychology, 1972, 3, 472-517. Clark, H . , & Lucy, P. Understanding what is meant from what is said: A study in conversationally conveyed requests. Journal of Verbal Learning and Verbal Behavior, 1975, 14, 56-72. Clark, R. Performing without competence. Journal of Child Language, 1974, 1, 1-10. Corrigan, R. Relationship between object permanence and language development: How much and

1.

The A c q u i s i t i o n o f English

131

how strong? Paper presented at the Eighth Annual Stanford Child Language Research Forum, Stanford, 1976. Corrigan, R. Cognitive correlates of language: Differential criteria yield differential results. Child Development, 1979, 50, 617-631. Cromer, R. F . The development of temporal reference during the acquisition of language. Un­ published doctoral dissertation, Harvard University, 1968. Cromer, R. F . An experimental investigation of a putative linguistic universal: Marking and the indirect object. Journal of Experimental Child Psychology 1975, 20, 3-80. Cross, T. G . Mothers' speech adjustments: The contribution of selected child listener variables. In C . E . Snow & C . A. Ferguson (Eds.), Talking to children: Language input and acquisition. New York: Cambridge University Press, 1977. Cross, T. G. Motherese: Its association with rate of syntactic acquisition in young children. In N. Waterson & C . E . Snow (Eds.), The development of communication. New York: Wiley, 1978. Crystal, D. The analysis of intonation in young children. In D. Minifie & L . L . Lloyd (Eds.), Communicative and cognitive abilities: Early behavioral assessment. Baltimore, MD: University Park Press, 1978. Crystal, D . , Fletcher, P., & Garman, M. The grammatical analysis of language disability. New York: Elsevier, 1976. Curtiss, S. Genie: A psycholinguistic study of a modern-day wild child. New York: Academic Press, 1977. de Villiers, J . G . Prototypes in grammatical rule learning. Paper presented at the Biennial meeting of the Society for Research in Child Development. San Francisco, March, 1979. de Villiers, J . G . The process of rule learning in child speech: A new look. In K . E . Nelson (Ed.), Children's Language: Volume 2. New York: Gardner Press, 1980. de Villiers, J. G . Form and force interactions: The development of negatives and questions. In R. L . Schiefelbusch & J . Pickar (Eds.), Communicative competence: Acquisition and intervention. Baltimore, MD: University Park Press, 1984. de Villiers, J. G . Learning the passive from models: Some contradictory data. Paper presented at the 9th Annual Boston University Conference. de Villiers, J . G . , & de Villiers, P. A. Development of the use of word order in comprehension. Journal of Psycholinquistic Research, 1973, 2, 331-341. (a) de Villiers, J . G . , & de Villiers, P. A. A cross-sectional study of the acquisition of grammatical morphemes in child speech. Journal of Psycholinguistic Research, 1973, 2, 267-278. (b) de Villiers, J. G . , & de Villiers, P. A. Competence and performance in child language: Are children really competent to judge? Journal of Child Language, 1974, 1, 11-22. de Villiers, J . G . , & de Villiers, P. A. Semantics and syntax in the first two years: The output of form and function and the form and function of the input. In F . Minifie & L . L . Lloyd (Eds.), Communicative and cognitive abilities: Early behavioral assessment. Baltimore, MD: University Park Press, 1978. de Villiers, J . G . , Phinney, M . , & Avery, A. Understanding passives with non-action verbs. Paper presented at Eighth Annual Boston University Conference on Language Acquisition, October, 1982. de Villiers, J . G . , & Tager-Flusberg, H . Some facts one simply cannot deny. Journal of Child Language, 1975, 2, 279-286. de Villiers, J . G . , Tager-Flusberg, H . B . , & Hakuta, K . Deciding among theories of the develop­ ment of coordination in child speech. Papers and Reports on Child Language Development, (Department of Linguistics, Stanford University), 1977, 13, 118-125. de Villiers, J . G . , Tager-Flusberg, H . B . , Hakuta, K . , & Cohen, M. Children's comprehension of relative clauses. Journal of Psycholinguistic Research, 1979, 8, 499-518. de Villiers, P. A . , & de Villiers, J . G . Form and function in the development of sentence negation.

132

De Villiers a n d De Villiers

Papers and Reports on Child Language Development (Department of Linguistics, Stanford University), 1979, 17, 56-64. Dore, J. Holophrases, speech acts and language universals. Journal of Child Language, 1975, 2, 21-40. Dougherty, R. A. Coordination conjunction. Unpublished paper, MIT, 1967. Erreich, A. The acquisition of inversion in wh-questions: What evidence the child usesl Unpublished doctoral dissertation, City University of New York, 1980. Ervin, S. Imitation and structural change in children's language. In E . H . Lenneberg (Ed.), New directions in the study of language. Cambridge, MA: MIT Press, 1964. Ervin-Tripp, S. Discourse agreement: How children answer questions. In J. R. Hayes (Ed.), Cognition and the development of language, New York: Wiley, 1970. Fillmore, C . J . The case for case. In E . Bach & R. T. Harms (Eds.), Universals in linguistic theory. New York: Holt, Rinehart & Winston, 1968. Fillmore, C . J. Some problems for case grammar. In R. J . O'Brien (Ed.), Georgetown University Round Table on Language and Linguistics. Washington, D . C . : Georgetown University Press, 1971. Fischer, K . W. A theory of cognitive development: The control and construction of hierarchies of skills. Psychological Review, 1980, 87, 477-531. Fletcher, P. The development of the verb phrase. In P. Fletcher & M. Garman (Eds.), Language Acquisition. New York: Cambridge University Press, 1979. Flynn, S., & Lust, B . Acquisition of relative clauses: Developmental changes in their heads. In W. Harbert & J . Hershenson (Eds.), Cornell University Working Papers in Linguistics, 1980, 7. Fodor, J . A . , Bever, T . , & Garrett, M. The psychology of language. New York: McGraw Hill, 1974. Forner, M. The mother as L A D : Interaction between order and frequency of parental input and child production. Paper presented at the Sixth Annual University of Minnesota Linguistics Symposium, Minneapolis, 1977. Fraser, C , Bellugi, U . , & Brown, R. W. Control of grammar in imitation, comprehension and production. Journal of Verbal Learning and Verbal Behavior, 1963, 2, 121-135. Furrow, D . , Nelson, K . , & Benedict, H. Mothers' speech to children and syntactic development: Some simple relationships. Journal of Child Language, 1979, 6, 423-442. Garman, M. Early grammatical development. In P. Fletcher & M. Garman (Eds.), Language acquisition. New York: Cambridge University Press, 1979. Garvey, C . Requests and responses in children's speech. Journal of Child Language, 1975, 2, 41¬ 63. Gleason, H. A . , Jr. An introduction to descriptive linguistics. Rev. ed. New York: Holt, Rinehart & Winston, 1961. Gleitman, L . R . , & Wanner, E . Language acquistion: The state of the state of the art. In E . Wanner & L . R. Gleitman, (Eds.), Language acquisition: The state of the art. New York: Cambridge University Press, 1982. Golinkoff, R. M. The case for semantic relations: Evidence from the verbal and nonverbal domains. Journal of Child Language, 1981, 8, 413-437. Golinkoff, R. M . , & Kerr, L . J . Infants' perceptions of semantically defined action role changes in filmed events. Merrill Palmer Quarterly, 1978, 24, 53-61. Golinkoff, R. M . , & Markessini, J . "Mommy sock": The child's understanding of possession as expressed in two-noun phrases. Journal of Child Language, 1980, 7, 119-136. Goodluck, H . Linguistic principles in children's grammar of complement interpretation. Un­ published doctoral dissertation, University of Massachusetts, Amherst, 1978. Goodluck, H . Children's grammar of complement-subject interpretation. In S. L . Tavakolian (Ed.), Language acquisition and linguistic theory. Cambridge, MA: MIT Press, 1981. Goodluck, H . , & Roeper, T. The acquisition of perception verb complements. In H. Goodluck & L .

1.

T h e A c q u i s i t i o n of English

133

Solan, (Eds.), Papers in the structure and development of child language. University of Mas­ sachusetts, Occasional Papers in Linguistics, vol. 4, 1978. Goodluck, H . , & Tavakolian, S. Competence and processing in children's grammar of relative clauses. Cognition, 1982, 11, 1-27. Greenfield, P. M . , & Dent, C . H. Pragmatic factors in children's phrasal coordination. Journal of Child Language, 1982, 9, 425-444. Greenfield, P. M . , Nelson, K . , & Saltzman, E . The development of rulebound strategies for manip­ ulating seriated cups: A parallel between action and grammar. Cognitive Psychology, 1972, 3, 291-310. Greenfield, P. M . , & Smith, J . H . The structure of communication in early language development. New York: Academic Press, 1976. Griffiths, P. Speech acts and early sentences. In P. Fletcher & M. Garman (Eds.), Language acquisition. New York: Cambridge University Press, 1979. Hafitz, J . , Gartner, B . , & Bloom, L . Giving complements when you're two: The acquisition of complement structures in child language. Paper presented at the Fifth Annual Boston University Conference on Language Development, October, 1980. Hakuta, K . , de Villiers, J . G . , & Tager-Flusberg, H. Sentence coordination in Japanese and En­ glish. Journal of Child Language, 1982, 9, 193-207. Halliday, M. A. K . Learning how to mean. In E . H. Lenneberg & E . Lenneberg (Eds.), Founda­ tions of language development: A multidisciplinary approach (vol. 1). New York: Academic Press, 1975. Hamburger, H . , & Crain, S. Relative acquisition. In S. A. Kuczaj (Ed.), Language development: Syntax and semantics. New York: Lawrence Erlbaum Associates, 1982. Hardy-Brown, K . Universals and individual differences: Disentangling two approaches to the study of language acquisition. Developmental Psychology, 1983, 19, 610-624. Harries, H. Coordination reduction. Stanford University Working Papers on Language Universals, 1973, 11, 139-209. Hecht, B. F . , & Morse, R. What the hell are dese? Unpublished paper, Harvard University, 1974. Hockett, C . F . A course in modern linguistics. London: Macmillan, 1958. Holtman, M. The use of interrogative forms in the interaction of three mothers and their children. Journal of Psycholinguistic Research, 1972, 1, 311-336. Horgan, D. The development of the full passive. Journal of Child Language, 1978, 5, 65-80. Howe, C . J . The meanings of two-word utterances in the speech of young children. Journal of Child Language, 1976, 3, 29-48. Howe, C . J . Interpretive analysis and role semantics: A ten-year mesalliance? Journal of Child Language, 1981, 8, 439-456. Hurford, J. A child and the English question-formation rule. Journal of Child Language, 1975, 2, 299-301. Ingram, D. Transivity in child language. Language, 1971, 47, 888-910. Ingram, D. Stages in the development of one-word utterances. Paper presented at the Sixth Annual Child Language Research Forum, Stanford, 1974. Ingram, D. Language development during the sensori-motor period. Paper presented at the Third International Child Language Symposium, London, 1975. (a) Ingram, D. If and when transformations are acquired by children. In D. P. Dato (Ed.), Developmen­ tal psycholinguistics: Theory and application. Twenty-sixth Annual Georgetown University Roundtable. Washington, D . C . : Georgetown University Press, 1975. (b) Ingram, D. Sensorimotor intelligence and language development. In A. Lock (Ed.), Action, gesture and symbol: The emergence of language. New York: Academic Press, 1978. Ingram, D . , & Tyack, D. Inversion of subject NP and auxiliary in children's questions. Journal of Psycholinguistic Research, 1979, 8, 333-341.

134

De Villiers a n d De Villiers

Jackendoff, R. S. Semantic interpretation in generative grammar. Cambridge, MA: MIT Press, 1972. Jacobs, R. A . , & Rosenbaum, P. S. English transformational grammar. Waltham, MA: Blaisdell, 1968. Johnson, C . The ontogenesis of question words in children's language. Paper presented at the Fifth Annual Boston University Conference on Language Acquisition, October, 1980. Johnston, J. R . , & Slobin, D. I. The development of locative expressions in English, Italian, SerboCroatian and Turkish. Journal of Child Language, 1979, 6, 529-545. Keenan, E . L . , & Comrie, B . Noun phrase accessibility and universal grammar. Linguistic Inquiry, 1977, 8, 63-99. Kim, K . Development of the concept of truth-functional negation. Unpublished doctoral disserta­ tion, Harvard University, 1980. Klima, E . S. Negation in English. In J . Fodor & J . J . Katz (Eds.), The structure of language: Readings in the philosophy of language. Englewood Cliffs, NJ: Prentice-Hall, 1964. Klima, E . S., & Bellugi, U . Syntactic regularities in the speech of children. In J . Lyons & R. J . Wales (Eds.), Psycholinguistic papers. Edinburgh: Edinburgh University Press, 1966. Kuczaj, S. A . , II The acquisition of regular and irregular past tense forms. Journal of Verbal Learning and Verbal Behavior, 1911, 16, 589-600. Kuczaj, S. A . , II Children's judgments of grammatical and ungrammatical irregular past-tense verbs. Child Development, 1978, 49, 319-326. (a) Kuczaj, S. A . , II Why do children fail to overgeneralize the progressive inflection? Journal of Child Language, 1978, 5, 167-171. (b) Kuczaj, S. A. II Influence of contractibility on the acquisition of be: Substantial, meager, or unknown? Journal of Psycholinguistic Research, 1979, 8, 1-11. Kuczaj, S. A . , II More on children's initial failure to relate specific acquisitions. Journal of Child Language, 1981, 8, 485-487. Kuczaj, S. A . , II On the nature of syntactic development. In S. A. Kuczaj II (Ed.), Language development (Vol. 1). Syntax and semantics. Hillsdale, NJ: Lawrence Erlbaum Associates, 1982. Kuczaj, S. A . , I I , & Brannick, N. Children's use of the Wh question modal auxiliary placement rule. Journal of Experimental Child Psychology, 1979, 28, 43-67. Kuczaj, S. A . , I I , & Maratsos, M. P. What children can say before they will. Merrill-Palmer Quarterly, 1975, 27, 89-111. Kuczaj, S. A . , I I , & Maratos, M. P. The initial verbs of yes-no questions: A different kind of general grammatical category. Developmental Psychology, 1983, 19, 440-444. Labov, W., & Labov, T. Learning the syntax of questions. In R. N. Campbell & P. T. Smith (Eds.), Recent advances in the psychology of language: Language development and mother-child in­ teraction. London: Plenum Press, 1978. Lakoff, G . Generative semantics. New York: Holt, Rinehart & Winston, 1972. Lasnik, H . Remarks on coreference. Linguistic Analysis, 1976, 2, 1-22. Lebeaux, D . , & Pinker, S. The acquisition of the passive. Paper presented at the Sixth Annual Boston University Conference on Language Development, October, 1981. Legum, S. Strategies in the acquisition of relative clauses. Southwest Regional Laboratory Tech­ nical Note, No. T N 2-75-10, 1975. Leonard, L . B. Meaning in child language. New York: Grune & Stratton, 1976. Leopold, W. F . Speech development of a bilingual child: A linquist's record (vol. 3). Grammar and general problems in the first two years. Evanston, I L : Northwestern University Press, 1949. Lieven, E . Conversations between mother and young children: Individual differences and their possible implications for the study of language learning. In N. Waterson & C . E . Snow (Eds.), The development of communication. New York: Wiley, 1978. Lightbown, P. Question form and question function in the speech of young French L 2 learners. In M. Paradis (Ed.), Aspects of bilingualism. Columbia, SC: Hornbeam Press, 1978.

1.

T h e A c q u i s i t i o n of English

135

Limber, J . The genesis of complex sentences. In T. E . Moore (Ed.), Cognitive development and the acquisition of language. New York: Academic Press, 1973. Limber, J . Unravelling competence, performance and pragmatics in the speech of young children. Journal of Child Language, 1976, 3, 309-318. Lust, B. Conjunction reduction in child language. Journal of Child Language, 1977, 4, 257-288. Lust, B . , & Mervis, C . Development of coordination in the natural speech of young children. Journal of Child Language, 1980, 7, 279-304. Macnamara, J . Cognitive basis of language learning in infants. Psychological Review, 1972, 79, 1¬ 14. Macnamara, J. From sign to language. In J . Macnamara (Ed.), Language learning and thought. New York: Academic Press, 1977. MacWhinney, B . Levels of syntactic acquisition. In S. A. Kuczaj II (Ed.), Language development: Syntax and semantics. Hillsdale, NJ: Lawrence Erlbaum Associates, 1982. MacWhinney, B . Hungarian language acquisition as an exemplification of a general model of grammatical development. In D. I. Slobin (Ed.). The crosslinguistic study of language acquisi­ tion (Vol. II). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. MacWhinney, B . , & Bates, E . Sentential devices for conveying giveness and newness: A crosscultural developmental study. Journal of Verbal Learning and Verbal Behavior, 1978, 17, 539¬ 558. Maratsos, M. P. How preschool children understand missing complement subjects. Child Develop­ ment, 1974, 45, 700-706. Maratsos, M. P. Commentary on "Structure and variation in child language" by L . Bloom, P. Lightbown, and L . Hood. Monographs of the Society for Research in Child Development, 1975, 40(2), No. 160. Maratsos, M. P. The use of definite and indefinite reference in young children. New York: Cambridge University Press, 1976. Maratsos, M. P. How to get from words to sentences. In D. Aaronson & R. Rieber (Eds.), Perspectives in psycholinguistics. Hillsdale, NJ: Lawrence Erlbaum Associates, 1979. Maratsos, M. P. The acquisition of syntax. In J . Flavell & E . Markman, (Eds.), CarmichaeVs handbook of child psychology, 1984. Maratsos, M. P., & Abramovitch, R. How childen understand full, truncated and anomalous pas­ sives. Journal of Verbal Learning and Verbal Behavior, 1975, 14, 145-157. Maratsos, M. P., & Chalkley, M. A. The internal language of children's syntax: The ontogenesis and representation of syntactic categories. In K . E . Nelson (Ed.), Children's language (Vol. 2). New York: Gardner Press, 1980. Maratsos, M. P., & Kuczaj, S. A . , II Preschool children's use of not and n't: Is not (isn't) n't. In Papers and Reports on Child Language Development, (Department of Linquistics, Stanford University), 1976. Maratsos, M . , & Kuczaj, S. II Against a transformationalist account: A simpler analysis of auxiliary overmarkings. Journal of Child Language, 1978, 5, 337-346. Maratsos, M. P., Kuczaj, S. A . , II, Fox, D. M . , & Chalkley, M. A. Some empirical studies in the acquisition of transformational relations. In W. Collins (Ed.), Children's language and commu­ nication. The Minnesota Symposia on Child Psychology, Vol. 12. Hillsdale, NJ: Lawrence Erlbaum Associates, 1979. Mayer, J . , Erreich, A . , & Valian, V . Transformations, basic operations and language acquisition. Cognition, 1978, 6, 1-13. McCawley, J . The role of semantics in grammar. In E . Bach & R. Harms (Eds.), Universals in linguistic theory. New York: Holt, Rinehart & Winston, 1968. McNeill, D. A question in semantic development: What does a child mean when he says "no"?. In E . Zale (Ed.), Conference on language and language behavior. New York: Appleton-CenturyCrofts, 1968.

136

De Villiers a n d De Villiers

McNeill, D. The acquisition of language. New York: Harper and Row, 1970. McNeill, D. Semiotic extension. Paper presented at the Loyola Symposium on Cognition. Chicago, 1974. Menyuk, P. Sentences children use. Cambridge, MA: MIT Press, 1969. Menyuk, P. The acquisition and development of language. Englewood Cliffs, NJ: Prentice-Hall, 1971. Miller, W . , & Ervin, S. The development of grammar in child language. In U . Bellugi, & R. Brown, (Eds.), The acquisition of language. Monographs of the Society for Research in Child Development, 1964, 29, 9-34. Moerk, E . L . Piaget's research as applied to the explanation of language development. MerrillPalmer Quarterly, 1975,27, 151-170. Moerk, E . L . Relationships between parental input frequencies and children's language acquisition: A reanalysis of Brown's data. Journal of Child Language, 1980, 7, 105-118. Moerk, E . L . To attend or not to attend to unwelcome reanalyses? A reply to Pinker. Journal of Child Language, 1981, 8, 627-631. Morehead, D. M . , & Morehead, A. From signal to sign: A Piagetian view of thought and language during the first two years. In R. L . Schiefelbusch & L . L . Lloyd (Eds.), Language perspectives: Acquisition, retardation, and intervention. Baltimore, MD: University Park Press, 1974. Nelson, K . Structure and strategy in learning to talk. Monographs of the Society for Research in Child Development, 1913,38. Nelson, K . Individual differences in early semantic and syntax development. In D. Aaronson & R. W. Rieber (Eds.), Developmental psycholinguistics and communication disorders. Annals of the New York Academy of Sciences, 1915,263, 132-139. Nelson, K . Early speech in its communicative context. In F. Minifie & L . L . Lloyd (Eds.), Commu­ nicative and cognitive abilities: Early behavioral assessment. Baltimore, MD: University Park Press, 1978. Nelson, K . E . Experimental gambits in the service of language acquisition theory: From the Fiffin project to operation input swap. In S. A. Kuczaj, II (Ed.), Language development (Vol. 1). Syntax and semantics. Hillsdale, NJ: Lawrence Erlbaum Associates, 1982. Newport, E . Motherese: The speech of mothers to young children. In N. Castellan, D. Pisoni, & G . Potts (Eds.), Cognitive Theory, (Vol. II). Hillsdale, NJ: Lawrence Erlbaum Associates, 1976. Newport, E . , Gleitman, L . , & Gleitman, H. A study of mother's speech and child language acquisi­ tion. Papers and Reports on Child Language Development, (Department of Linguistics, Stanford University), 1975, 10, 111-116. Newport, E . , Gleitman, L . , & Gleitman, H. Mother, I'd rather do it myself: Some effects and noneffects of maternal speech style. In C . E . Snow & C . A. Ferguson (Eds.), Talking to children. Cambridge: Cambridge University Press, 1977. Osgood, C . E . , & Zehler, A. M. Acquisition of bitransitive sentences: Pre-linguistic determinants of language acquisition. Journal of Child Language, 1981, 8, 367-383. Otsu, Y . Grammatical constraints and syntactic development in children: Toward a theory of syntactic development in children. Unpublished doctoral dissertation, MIT, 1981. Parisi, D. What is behind child utterances? Journal of Child Language, 1974, 1, 97-105. Park, T. Z. Some facts on negation: Wode's four-stage developmental theory of negation revisited. Journal of Child Language, 1919,6, 147-151. Pea, R. D. The development of negation in early child language. In D. R. Olson (Ed.), The social foundations of language and thought: Essays in honor of Jerome S. Bruner. New York: W. W. Norton, 1979. Phinney, M. Syntactic constraints and the acquisition of embedded sentential complements. Un­ published doctoral dissertation, University of Massachusetts, Amherst, 1981. Piaget, J . The construction of reality in the child. New York: Basic Books, 1954.

1.

T h e A c q u i s i t i o n of English

137

Piaget, J . The origins of intelligence in children. New York: W. W. Norton, 1963. Pinker, S. On the acquisition of grammatical morphemes. Journal of Child Language, 1981, 8, 477-484. Pinker, S. A theory of the acquisition of lexical-interpretive grammars. In J. Bresnan (Ed.), The mental representation of grammatical relations. Cambridge, MA: MIT Press, 1982. Ramer, A. L . H. Syntactic styles in emerging language. Journal of Child Language, 1976, 3, 49¬ 62. Retherford, K . S., Schwartz, B. C , & Chapman, R. S. Semantic roles and residual grammatical categories in mother and child speech: Who tunes in to whom? Journal of Child Language, 1981, 8, 583-608. Roeper, T. On the importance of syntax and the logical use of evidence in language acquisition. In S. A. Kuczaj, II (Ed.), Language development (vol. 1). Syntax and semantics. Hillsdale, NJ: Lawrence Erlbaum Associates, 1982. Roeper, T . , Lapointe, S., Bing, J . , & Tavakolian, S. L . A lexical approach to language acquisition. In S. L . Tavakolian (Ed.), Language acquisition and linguistic theory. Cambridge, MA: MIT Press, 1981. Rosenbaum, P. The grammar of English predicate complement constructions. Cambridge, MA: MIT Press, 1967. Ross, J . R. Act. In D. Davidson & G . Harman (Eds.), Semantics of natural languages. Dordrecht, Holland: Reidel, 1972. Schaerlaekens, A. M. The two-word stage in child language development: A study based on evi­ dence provided by Dutch-speaking triplets. The Hague: Mouton, 1973. Schlesinger, I. M. Production of utterances and language acquisition. In D. I. Slobin (Ed.), The ontogenesis of grammar. New York: Academic Press, 1971. Schlesinger, I. M. Relational concepts underlying language. In R. L . Schiefelbusch & L . L . Lloyd (Eds.), Language perspectives: Acquisition, retardation and intervention. Baltimore, MD: Uni­ versity Park Press, 1974. Schlesinger, I. M. Steps to language: Toward a theory of native language acquistion. Hillsdale, NJ: Lawrence Erlbaum Associates, 1982. Searle, J. Indirect speech acts. In P. Cole & J . L . Morgan (Eds.), Syntax and semantics (Vol. 3). New York: Academic Press, 1975. Sedlak, P. A. S. Direct/indirect object word order: A cross-linguistic analysis. Working papers on Language Universals, 1975, 18. Language Universals Project, Stanford University. Shatz, M. Children's comprehension of their mother's question-directives. Journal of Child Lan­ guage, 1978, 5, 39-46. (a) Shatz, M. The development of communicative understanding: An early strategy for interpreting and responding to messages. Cognitive Psychology, 1978, 10, 271-301. (b) Shatz, M. How to do things by asking: Form-function pairings in mother's questions and their relation to children's responses. Child Development, 1979, 50, 1093-1099. Sheldon, A. The role of parallel function in the acquisition of relative clauses in English. Journal of Verbal Learning and Verbal Behavior, 1914, 13, 272-281. Sinclair-de Zwart, H . Sensorimotor action patterns as a condition for the acquisition of syntax. In R. Huxley & E . Ingram (Eds.), Language acquisition: Models and methods. New York: Academic Press, 1971. Sinclair, H . The role of cognitive structures in language acquisition. In E . H. Lenneberg & E . Lenneberg (Eds.), Foundations of language development. New York: Academic Press, 1976. Slobin, D. I. Universals of grammatical development in children. In G . B. Flores d'Arcais & W. J . M. Levelt (Eds.), Advances in psycholinguistics. Amsterdam: North-Holland, 1970. Slobin, D. I. On the learning of morphological rules. In D. I. Slobin (Ed.), The ontogenesis of grammar. New York: Academic Press, 1971.

138

De Villiers a n d De Villiers

Slobin, D . I . Cognitive prerequisites for the development of grammar. In C . A. Ferguson & D. I. Slobin (Eds.), Studies of child language development. New York: Holt, Rinehart & Winston, 1973. Slobin, D. I. Crosslinguistic evidence for the Language-Making Capacity. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. II). Hillsdale, NJ: Lawrence Erlbaum Associ­ ates, 1985. Smith, C . S. The acquisition of time talk: Relations between child and adult grammars. Journal of Child Language, 1980, 7, 263-278. Snow, C . E . Mother's speech to children learning language. Child Development, 1972, 43, 549¬ 565. Snow, C . E . Mother's speech research: From input to interaction. In C. E . Snow & C . A. Ferguson (Eds.), Talking to children. Cambridge: Cambridge University Press, 1977. Snow, C . E . Conversations with children. In P. Fletcher & M. Garman (Eds.), Language acquisi­ tion. New York: Cambridge University Press, 1979. Solan, L . Anaphora in child language. Unpublished doctoral dissertation, University of Mas­ sachusetts, Amherst, 1978. Solan, L . The acquisition of structural restrictions on anaphora. In S. L . Tavakolian (Ed.), Lan­ guage acquisition and linguistic theory. Cambridge, MA: MIT Press, 1981. Solan, L . & Roeper, T. Children's use of syntactic structure in interpreting relative clauses. In H . Goodluck & L . Solan (Eds.), Papers in the structure and development of child language. University of Massachusetts Occasional Papers in Linguistics (Vol. 4), 1978. Strohner, H . , & Nelson, K . E . The young child's development of sentence comprehension: The influence of event probability, non-verbal context, syntactic form and strategies. Child Develop­ ment, 1974, 45, 567-576. Tager-Flusberg, H. The development of relative clauses in child speech. Papers and Reports on Child Language Development (Department of Linguistics, Stanford University), 1982, 21, 104¬ 111. Tager-Flusberg, H . , de Villiers, J . S., & Hakuta, K . The development of sentence coordination. In S. A. Kuczaj, II (Ed.), Language development (Vol. 1). Syntax and semantics. Hillsdale, NJ: Lawrence Erlbaum Associates, 1982. Tavakolian, S. L . The conjoined-clause analysis of relative clauses. In S. L . Tavakolian (Ed.), Language acquisition and linguistic theory. Cambridge, MA: MIT Press, 1981. Turner, E . A . , & Rommetveit, R. Experimental manipulation of the production of active and passive voice in children. Language and Speech, 1967', 10, 169-180. Tyack, D . , & Ingram, D. Children's production and comprehension of questions. Journal of Child Language, 1977, 4, 211-224. Uzgiris, I. Organization of sensorimotor intelligence. In M. Lewis (Ed.), Origins of intelligence. New York: Plenum, 1976. Valian, V . Noun phrases in young children's speech. Paper presented at the Ninth Annual Boston University Conference on Language Development, October, 1984. Valian, V . , Winzemer, J . , & Erreich, A. A "Little Linguist" model of syntax learning. In S. L . Tavakolian (Ed.), Language acquisition and linguistic theory. Cambridge, MA: MIT Press, 1981. Van der Geest, T. Some interactional aspects of language acquisition. In C . E . Snow & C . A. Ferguson (Eds.), Talking to children: Language input and acquisition. New York: Cambridge University Press, 1977. Wary as, C . L . , & Stremel, K . On the preferred form of the double object construction. Journal of Psycholinguistic Research, 1974, 3, 271-280. Wason, P. C . The contexts of plausible denial. Journal of Verbal Learning and Verbal Behavior, 1964, 4, 7-11. Wason, P. C . In real life negatives are false. Logique et Analyse, 1972, 57-58, 17-38.

1.

T h e A c q u i s i t i o n o f English

139

Watt, W. C . On two hypotheses concerning psycholinguistics. In J. R. Hayes (Ed.), Cognition and the development of language. New York: Wiley, 1970. Wells, G . Learning to code experience through language. Journal of Child Language, 1974, 7, 243¬ 269. Wells, G . Variation in child language. In P. Fletcher & M. Garman (Eds.), Language acquisition. New York: Cambridge University Press, 1979. Wieman, L . A. Stress patterns of early child language. Journal of Child Language, 1976, 3, 283¬ 286. Wilson, R . , Pinker, S., Zaenen, A . , & Lebeaux, D. Productivity and the dative alternation. Paper presented at the Sixth Annual Boston University Conference on Language Development, October, 1981. Winzemer, J. A. A lexical-expectation model for children's comprehension of wh-questions. Un­ published doctoral dissertation, City University of New York, 1981. Wode, H. Four early stages in the development of L I negation. Journal of Child Language, 1977,4, 87-102. Wooten, J . , Merkin, S., Hood, L . , & Bloom, L . Wh-questions: Linguistic evidence to explain the sequence of acquisition. Paper presented to Biennial meeting of the Society for Research in Child Development. San Francisco, March, 1979.

2 Anne

E.

Universität

The Acquisition of German

Mills Tübingen

Contents Introduction 141 Descriptive Sketch of German 141 Sources of Data 151 Overall Course of Development 153 The Data 158 Typical Errors 158 Error-Free Acquisition 214 Timing of Acquisition 227 Reorganization in Development 226 Lexical Innovations 227 Cognitive Pacesetting of Language Development Linguistic Pacesetting of Cognitive Development Input and Adult-Child Interaction 234 Individual Differences 237 Theoretical Implications 239 Suggestions for Further Study 246

229 233

INTRODUCTION 1. D e s c r i p t i v e S k e t c h o f G e r m a n 1 . 1 . Regional

and Social

1

Variation

German is an Indo-European language and w i t h i n that f a m i l y belongs to the Germanic group. I t is spoken over a large area and has many different dialects. The languages w h i c h are spoken i n some areas, for example Austria, German Switzerland, and Italian T i r o l , are variants o f German w h i c h are, however, often *I am indebted to David A. Reibel, Rudolf Thiem, Christine Meinecke, and Fritz Hermanns for their helpful suggestions in the writing of this paper, and especially to Gisa Neumann for her constructive typing of it. 141

142

Mills

considered as being quite separate from German. The dialects w i t h i n Germany also show considerable variation i n phonology, lexical items, etc., so that d i f f i ­ culties o f mutual understanding can arise. Standard H i g h German, w h i c h is consistent w i t h the written form o f the language, and described i n compendious grammars such as Helbig and Buscha (1972) and Curme (1904), is i n fact an abstraction. The German spoken i n the Hannover area is usually considered to be closest to this standard form. The dialect a child is exposed to i n his surroundings is extremely important when considering acquisition data, since the dialect form may be at variance w i t h the standard f o r m . Where it is appropriate, this has been discussed i n the evalua­ tion o f the data presented i n this chapter; unfortunately, though, i n most cases too little is k n o w n about the linguistic environment o f the children investigated and the linguistic features o f the local dialect. This is an important point to be carefully considered i n the planning o f all future observational and experimental research. 1.2.

Structural

Features

A brief summary o f the main structural features o f Standard H i g h German w i l l be presented. A necessary selection o f the structures to be described has been made w i t h respect to the acquisition data to be presented below. 1.2.1. Word Order. W o r d order is variable but rule governed, especially w i t h respect to the position o f the verb (Bierwisch, 1973). Grammar books refer to a canonical w o r d order o f subject-verb-object from w h i c h they derive other word orders. Finite verb i n second position is an obligatory rule i n declarative sentences. The canonical status o f this w o r d order w i t h regard to the position o f the other elements, however, is not clearly reflected i n an overwhelming fre­ quency o f this w o r d order i n written or spoken language. I n main clauses, elements other than subject may occur i n initial position associated w i t h topicalization and given-new distinction but the verb is then obligatorily i n second position followed by the subject. Sentences 1-4 are examples o f w o r d order variants o f the same sentence: 'the man gave the woman a bunch o f flowers yesterday'. (1) der DEF.ART: MASC:SG:NOM

Mann

gab

der

man

gave

DEF.ART: FEM:SG:DAT

Frau

gestern

ein

-en

woman

yesterday

INDEF.ART

MASC:SG:ACC

Blumenstrauß* bouquet

* [The German letter fi represents historical sz, pronounced in the same way as ss (and sometimes written as ss or sz). German nouns are capitalized as in standard orthography. Where child speech is quoted, the orthographic conventions of the individual authors have been adopted. Ed.]

2.

The A c q u i s i t i o n o f G e r m a n

143

(2) einen Blumenstrauß gab der Mann gestern der Frau (3) der Frau gab der Mann gestern einen Blumenstrauß (4) gestern gab der Mann der Frau einen Blumenstrauß I n subordinate clauses the finite verb is generally i n clause final position. For example: (5) das DEF.ART: NEUT:SG:NOM

Kind,

das

im

child

REL.PRO: NEUT:SG:NOM

in+DEF.ART: MASC:SG:DAT

Garten

spiel

-t

garden

play

3SG:PRES

'the child who is playing in the garden' After an auxiliary or modal verb the main verb is also i n final position i n main sentences. For example: (6) das DEF.ART: NEUT:SG:NOM

Mädchen

muß

20

Stunden

arbeit

-en

girl

must: MOD

20

hours

work

INF

'the girl must work 20 hours' (7) das DEF.ART: NEUT:SG:NOM

Mädchen

wird

20

Stunden

arbeit

-en

girl

FUT.AUX

20

hours

work

INF

'the girl will work 20 hours' (8) das DEF.ART: NEUT:SG:NOM

Mädchen

hat

20

Stunden

girl

PAST.AUX

20

hours

ge

-arbeit

PAST.PART.PREFIX

work

-et PAST. PART. SUFFIX

'the girl has worked 20 hours'

1.2.2. Noun Phrase

Morphology.

A l l nouns have a syntactic gender w h i c h

determines the set o f inflections used to mark case relations. There are three genders i n the singular: masculine, feminine, and neuter. The plural paradigm is common to a l l genders. The case inflections are present i n the definite and indefinite articles, demonstrative adjectives, and attributive adjectives. Tables 2.1 and 2.2 illustrate these paradigms. The noun itself is marked for case only i n

Mills

144

the genitive case o f masculine and neuter singular nouns and i n the dative plural. In some set phrases the dative is marked on masculine and neuter singular nouns. Otherwise the noun occurs i n citation form (MacWhinney, 1985). Examples 9-11 illustrate this for the nouns die Hunde 'the dogs' and das Haus 'the house'. (9) Peter Peter

gab

es

den

Hund

-e

-n

gave

if.ACC

DEF.ART: DAT: PL

dog

PL

DAT: PL

'Peter gave it to the dogs' (10) die DEF.ART: FEM:SG:NOM

Tür

des

Haus

-es

door

DEF.ART: NEUT:SG:GEN

house G E N

'the door of the house' (11) zu at

Haus

-e

house D A T

'at home' Adjectives i n attributive function have inflectional endings, as Table 2.2 indicates. These endings vary, however, according to the presence or absence o f an article and according to the type o f article. I f the adjective is preceded by a definite article or demonstrative adjective i n the noun phrase, the adjective takes what are called the " w e a k " endings. I f the adjective has no preceding article, the endings are " s t r o n g . " I f the adjective is preceded by an indefinite article, the endings are " m i x e d . " For example: (12) der DEF.ART: MASC:SG:NOM

gut

-e

Käse

good

MASC: SG:NOM

cheese

"weak"

'the good cheese'

TABLE 2.1 Definite and Indefinite Article Declension Neuter

Masculine

Feminine

Plural

Def.

Indef.

Def.

Indef

Def.

Indef.

Def.

Nominative

der

ein

das

ein

die

eine

die

Accusative

den

einen

das

ein

die

eine

die

Genitive

des

eines

des

eines

der

einer

der

Dative

dem

einem

dem

einem

der

einer

den

TABLE 2.2 Declension o f A d j e c t i v e Gross ' B i g ' : " S t r o n g " W h e n N o t Preceded b y Article, " M i x e d " W h e n Preceded b y Indefinite Article, or Similar F o r m , " W e a k " W h e n Preceded b y Definite Article Singular

Plural

FEM

MASC

NEUT

Case

strong

mixed

weak

strong

mixed

weak

strong

mixed

weak

strong

mixed

weak

NOM ACC GEN DAT

grosser grossen grosses grossem

grosser grosse« grosse« grosse«

grosse grosse« grosse« grosse«

grosse grosse grosser grosser

grosse grosse grosse« grosse«

grosse grosse grosse« grosse«

grosses grosses grosses grossem

grosses grosses grosse« grosse«

grosse grosse grosse« grosse«

grosse grosse grosser grosse«

grosse« grosse« grosse« grosse«

grosse« grosse« grosse« grosse«

146

Mills

(13) ein INDEF.ART: MASC:SG:NOM

gut

-er

Kase

good

MASC: SG:NOM

cheese

"mixed"

'a good cheese' (14) gut good

-er

Käse

MASC:SG:NOM

cheese

"strong"

'good cheese' As can be seen from Tables 2.1 and 2.2, there is a considerable multiplicity o f form, even w i t h i n one gender paradigm. This presents a considerable problem for the child. He must rely on w o r d order rules and semantic and pragmatic information to interpret strings before he has mastered this system. This is true to some extent even when the system has been learned, since formal ambiguity exists i n some cases. This system o f inflections and w o r d order rules is a complex area for language acquisition i n German and is o f major interest. 1.2.3. Prepositions. Prepositions require the use o f particular cases, pre­ dominantly accusative and dative. Some prepositions can amalgamate w i t h a following masculine or neuter definite article i n the dative case and a f o l l o w i n g neuter definite article i n accusative case, for example, von dem can become vom from the', in + das can become ins 'into the'. The contracted form is obligatory in some set phrases. This is an additional complexity for the child i n the casemarking system. 4

Some prepositions can be used w i t h either dative or accusative case to convey a stative or directional meaning. (15) ich I

gehe

in

das

go

in: DIRECT

D E F . ART: NEUT: S G: A C C

Haus house

'I go into the house' (16) ich I

arbeite

in

dem

Haus

work

in: STAT

D E F . ART: NEUT: S G: DAT

house

'I work in the house' (17) die D E F . ART: FEM:SG:NOM

Ameise

kriecht

auf

den

Tisch

ant

crawls

on: DIRECT

D E F . ART: MASC:SG:ACC

table

'the ant crawls onto the table' (18) die D E F . ART: FEM:SG:NOM

Ameise

kriecht

auf

dem

Tisch

ant

crawls

on: STAT

D E F . ART: MASC:SG:DAT

table

'the ant is crawling on the table'

2.

The Acquisition of German

147

I n most contexts the verb determines whether the preposition is being used i n the stative or directional sense, as examples 15 and 16 illustrate. I n sentences where the verb can allow both senses, however, the case information becomes crucial. The child's comprehension o f such sentences and production o f all sentences w i t h prepositions relies on his control o f the inflectional system. 1.2.4. Verb Morphology. Verb morphology is not particularly complex. Verb tenses are formed by on the whole postfixed inflections on the verb root, or by using auxiliaries, as examples 1 9 - 2 2 illustrate w i t h the verb zeigen 'to show'. (19) ich I

zeig

-e

show

1SG

T show' (20) ich I

zeig

-t

-e

show

PAST

1SG

'I showed' (21) ich I

hab

-e

ge-

zeig

-t

AUX

1SG

PAST

show

PAST

'I have shown' (22) ich I

werd

-e

zeig

-en

AUX

1SG

show

INF

'I will show' Some verbs behave irregularly i n the formation o f tenses i n v o l v i n g changes i n the root forms as w e l l as the inflections. One group, the " s t r o n g " verbs, change

TABLE 2.3 Verb Inflections in the Present Tense of Sin gen 'to Sing' 1SG 2SG 3SG

ich sing- e du sing- st e

r

}

sie > sing- t es ) 1PL 2PL 3PL

wir sing- en ihr sing- t sie sing- en

1

The third person plural form is identical in the pronoun and verb form with the second person polite form singular and plural. !

148

Mills

the root form i n some tenses and have different inflections; another smaller " m i x e d " group change the root form but have the regular inflections. For example, the simple past tense form o f ich bring-e T b r i n g ' is ich brach-t-e but the past tense o f ich zieh- e T p u l l ' is ich zog. These exceptions present the main source o f difficulty. Person is also marked on the verb by postfixed inflections, as Table 2.3 indicates. There is some mutiplicity o f form i n the inflections associated w i t h different persons, first and third person plural are identical for example. I t is obligatory to use a personal pronoun i f no full noun is present. Some verbs are made up o f a stem and a verbal particle. Some o f these verbal particles are like prepositions and separate from the verb stem i n some construc­ tions. I n a declarative sentence, for example, the particle i n such verbs w i l l be i n clause final position as sentence 23 illustrates w i t h the verb an-ziehen 'to put o n ' . (23) er he

zog

seine

Jacke

an

pulled

his

jacket

on

'he put his jacket on' In the past participle the verbal particle is placed i n front o f the past prefix ge- so that the past participle o f an-ziehen is an-ge-zogen. M a n y verb stems f o r m a number o f different verbs i n combination w i t h different verbal particles, so the particle w o u l d seem to be contributing a major part to the meaning o f the verb. There are 60 different verbs (Mater, 1967) w i t h such particles i n combination w i t h the stem ziehen 'to p u l l ' . For example: abeinüberweg-

ziehen ziehen ziehen ziehen

'pull down' 'draw in' 'cover' 'pull off

1.2.5. Questions. Questions are o f t w o types: yes-no questions and w h questions. Yes-no questions can be formed by a rising intonation only or by subject-verb inversion. (24) Peter Peter

kauft buys

Brötchen? rolls

'is Peter buying rolls?' (25) kauft Peter buys

Peter

Brötchen? rolls

'is Peter buying rolls?'

2.

The Acquisition of German

149

Tag questions can be formed on declarative sentences by adding a w o r d or phrase, as sentence 26 illustrates. The exact form varies according to dialect. There can be a variation i n intonation to indicate different functions o f the tag as in English (see M i l l s [1981] for fuller discussion). (26) Peter Peter

kauft Brötchen,

oder?

buys

or

rolls

'Peter is buying rolls, isn't he?' Wh-questions involve a fronting o f the interrogative pronoun and subject-verb inversion. (27) wo

kauft Peter

LOC.INT

buys

Peter

Brötchen? rolls

'where is Peter buying rolls?' 1.2.6. Negation. Negation on the verb requires the placement o f nicht 'not' after the finite verb i n most instances. Where subject-verb inversion has oc­ curred, nicht is placed after the subject. I n subordinate clauses, where the finite verb is i n clause-final position, nicht is placed before the verb. There is however considerable variation i n these rules. Negation on the noun requires the use o f kein ' n o ' as an article. (28) Johann

geht

nicht

ins

Kino

Johann

goes

not

into+the

cinema

'Johann isn't going to the cinema' (29) wir we

haben keine

Milch

have

milk

no

'we haven't any milk' 1.2.7. Passive. The passive is formed much as i n English, that is by using the auxiliary werden, the preposition von ' b y ' before the agent ( w i t h dative case), and the past participle o f the verb. This last element is however positioned at the end o f the sentence, f o l l o w i n g the general rule i n all complex verb forms. Sentences 30 and 31 illustrate the active and passive forms. (30) der DEF.ART: MASC:SG:NOM

Mann

öffne

-t

die

Tür

man

open

3SG:PRES

DEF.ART: FEM:SG:ACC

door

'the man opens the door'

150

Mills

(31) die DEF.ART: FEM:SG:NOM

Tür

wird

von

dem

door

PRES.AUX

by [PREP]

DEF.ART: MASC:SG:DAT

Mann

ge-

öffne

-t

man

PAST

open

PAST

'the door is opened by the man' 1.2.8. Subordinate Clauses. Subordinate clauses also have the verb i n clause final position, as mentioned above. Relative clauses branch principally to the right o f the head noun. The relative pronoun, w h i c h is i n most cases identical to the definite article, is placed i n clause initial position next to the head noun. German also allows a left branching relative construction, w h i c h requires no relative pronoun. The verb has the form o f an adjectival present or past partici­ ple. Sentences 32 and 33 illustrate these t w o types o f relative clause. The latter type, however, is mainly restricted to formal written language. (32) der DEF.ART: MASC:SG:NOM

Brief,

den

der

Mann

letter

REL.PRO: MASC:SG:ACC

DEF.ART: MASC:SG:NOM

man

ge-

schrieben

hat,

ist

unklar

PAST

write:PAST

AUX

is

unclear

'the letter which the man wrote is unclear' (33) der DEF.ART: MASC:SG:NOM

von

dem

Mann

by [PREP]

DEF.ART: MASC:SG:DAT

man

ge-

schrieben

-e

Brief

ist

unklar

PAST

write:PAST MASC:SG:NOM [ADJ.SUFF]

letter

is

unclear

'the letter written by the man is unclear' Many verbs allow verbal complements w h i c h can be formed either w i t h daß 'that' and a finite verb i n clause final position (34) or w i t h an infinitive construc­ tion (35). A s i n English, the subject o f the infinitival complement is usually the object o f the main clause except w i t h a few verbs such as versprechen 'to promise' etc. (34) der DEF.ART: MASC:SG:NOM

Beamte

empfiehlt,

daß

die

official

recommends

that

DEF.ART: FEM:SG:NOM

2.

The Acquisition of German

Universität

den

Antrag

university

DEF.ART: MASC:SG:ACC

application

rechtzeitig

stell

-t

in:good:time

submit

3SG:PRES

151

'the official recommends that the university submit the application in good time' (35) der DEF.ART: MASC:SG:NOM

Beamte

empfiehlt

der

official

recommends

DEF.ART: FEM:SG:DAT

Universität, university

den

Antrag

rechtzeitig

zu

stell

-en

DEF.ART:

application

in: good: time

to

submit

INF

MASC:SG:ACC 'the official recommends the university to submit the application in good time' 2. S o u r c e s o f D a t a

2

The richest sources o f data are still the diary studies made at the end o f the last and at the beginning o f this century (Preyer, 1882; Lindner, 1898; Schadel, 1905; Scupin & Scupin, 1907, 1910; Stern & Stern, 1928; Leopold, 1949). These have all the limitations o f diary studies i n that only selected material is recorded, for example, and i n that full contextual information is usually missing. Studies such as Preyer and Scupin and Scupin d i d not concentrate on language but recorded utterances where they reflected another developmental aspect o f the child. A l t h o u g h this is a major disadvantage i n any analysis o f the material, Scupin and Scupin contains such a wealth o f examples (from one child up to the age o f 6 years) that many insights can be gained from the study. Stern and Stern recorded their t w o children up to the ages o f 5 and 6 but i n their summarizing and analysis o f their material they concentrated on vocabulary and morphology and neglected syntax. None o f the studies describe the dialect the child or children were exposed to, so that it is impossible to detect the influence this might have had on some forms. Despite these limitations, the data from the diary studies can usefully be pooled when no other more detailed w o r k is available. Later w o r k has often i n fact shown that the observations have more general validity. It is not the purpose of this chapter to review all the work that has ever been done on child language in German. The aim is rather to discuss the data which are of interest when considering which general principles may be involved in language acquisition and which problems languagespecific features pose in the acquisition process. Within that framework I hope that I have not made too many omissions. A. E . M. November, 1981. 2

152

Mills

Apart from these studies, little w o r k has been done on spontaneous produc­ tions, especially from the point o f view o f syntax and morphology. M a n y o f the studies (e.g. K o n i g , 1972; Rickheit, 1975) have been o f a statistical nature comparing the quantitative use o f specific constructions i n the written and spoken language o f older children. Such studies are not very useful for present purposes since they trace changes i n use only i n quantitative terms and contain practically no information about errors or conditions o f use o f a construction. G r i m m (1973) gives more detail o f use but the data are still predominantly numerical w i t h little analysis o f structure. Park (1970, 1971a, 1971b, 1974) collected data from three Swiss-German children aged 2 - 3 years and analyzed various aspects o f acquisi­ tion, for example verb morphology, use o f the a u x i l i a r y . The data presented are often few, however, and the proposed analysis is not always clearly supported. M i l l e r (1976) collected a considerable amount o f data from three children aged 1-2 years and analyzed the types o f semantic constructions occurring i n the, predominantly t w o - w o r d , utterances. His a i m was to show the importance o f semantic and pragmatic information i n providing a description o f such utter­ ances. He does not therefore enter into a highly detailed discussion o f his results in terms o f the principles o f acquisition but his data are a useful source o f information. Few researchers, Stephany (1976), Wode (1971, 1977), Clancy et al. (1976), use data from spontaneous utterances o f German children to make explicit comparison w i t h data from other languages. This has been done i n the areas o f w o r d order, negation and question formation, and complex sentences. 3

Experimental studies on German child language are however far more com­ mon than studies o f spontaneous productions and are too numerous to mention here individually. These studies test imitation, comprehension, and production of many different structures i n varying situations. The children tested are prin­ cipally older than 3-years-of-age. Since there are few data from spontaneous utterances from children o f this age, however, i t is frequently difficult to con­ struct a balanced picture o f acquisition. Very little information is available on the structural aspects o f caretakers' speech to their children. The w o r k that has been done by M i l l s ( i n prep.) and Oksaar (1977) is based on a small sample and needs to be verified by data from a larger sample. Although there has been a steadily growing interest i n child language over the last 10 years i n Germany, there is still relatively little research done i n this area, especially i n the way o f longitudinal studies. The growing interest i n c h i l d language has also coincided w i t h a main focus i n German linguistics on pragmat­ ics and semantics, w i t h the result that more w o r k has been done on these aspects of child language than on the acquisition o f structures. V e r y little is k n o w n about

Park (1981) has analyzed the data from these papers, especially from one child, more exten­ sively in a recent publication. Unfortunately this was received too late to be discussed here in detail. 3

2.

The Acquisition of German

153

the acquisition o f many constructions i n German, so that crosslinguistic com­ parison can only be made i n a few areas.

3. O v e r a l l C o u r s e o f D e v e l o p m e n t 3. /. One-word

Stage

The child's first utterances typically are produced around 1;0. The type o f semantic functions i n the child's first productions follow the pattern found i n English; that is, they are predominantly constative and volitional, involving one entity rather than t w o . They most commonly refer to an object, action, ap­ pearance/existence/demonstrative, negation, and position. Utterances referring to agent, benefactor, possession, and disappearance are less frequently encountered. Nouns predominate as a grammatical class, being produced i n the unmarked singular except for those nouns w h i c h are most commonly used in the plural, such as Schuh-e 'shoes'. The child's productions usually contain many onomatopoeic forms and nursery words, w h i c h is probably a function o f the amount o f baby talk i n the input speech, for example adda 'go for a w a l k ' , wau­ wau ' d o g ' . These forms and the child's production o f standard forms are com­ monly reduplicative, for example mi-mi for the name Melanie. Consonant cluster reduction is also typical, for example *lafen for schlafen 'to sleep'; this con­ tinues until roughly age 3;0. Verbal particles are c o m m o n , especially w i t h the semantic function o f loca­ tion, for example auf ' u p ' , runter ' d o w n ' . They also frequently stand for a full verb, for example an from an-ziehen 'to put on ( c l o t h i n g ) ' . These occur clausefinally i n adult speech and are stressed, w h i c h probably explains their early acquisition. W h e n the child produces full verbs, they usually have the -en ending like the infinitive. This is probably attributable to the frequent use i n adult speech of modals and auxiliaries w h i c h send the main verb i n infinitive form to the end of the clause, infinitival imperatives, and syntactic baby talk i n which no modal or auxiliary is used but the main verb is final i n infinitive form. The verbs i n the child's speech which do not have this -en ending are usually part o f a set formula such as schmeck-1 from es schmeckt ' i t tastes g o o d ' . This is the answer to the frequent question from the adult,when any food is given to the c h i l d , schmeckt's ? 'does i t taste good?'. Modification is variable i n the amount it occurs as a function. W h e n adjec­ tives are used, they are used i n predicative f o r m , that is without the inflectional endings w h i c h are necessary i n attributive position, for example heiß ' h o t ' . Other frequent forms i n this function are the adverbs auch ' t o o ' , nochmal 'again', mehr ' m o r e ' . The negative is expressed by the sentence external negator nein ' n o ' . Demonstrative function is usually expressed through the deictic terms da 'there' and hier 'here'.

154

Mills

Some t w o w o r d sequences are used at this stage but these appear to be unanalyzed forms which are frequent i n the input speech, for example guck mal 'just l o o k ! ' . 3.2. Two-word

Stage

The first functional relations expressed at this stage are demon­ strative/appearance/existence, i n agreement w i t h the findings for English. A c ­ tions are more commonly referred to than states. Nouns are still predominantly used i n the unmarked singular but some plurals are used—usually correctly, for example Bucher 'books'. These appear to be examples o f rote learning, howev­ er, rather than a production by rule. Articles are rare at this stage and, when they do occur, they are often reduced i n form: de for the definite article and n for the indefinite article. These reduced forms cannot mark case or gender, w h i c h is necessary i n the full forms o f the article. The use o f this common form may be due to the child's lack o f control o f the inflectional system but these reduced forms also occur i n adult fast speech. Adjectives are used i n attributive position, that is before the noun, and are usually inflected to agree w i t h the noun i n case and gender. Since no article precedes the adjective at this stage, strong endings are appropriate and there are recorded examples o f this correct use i n children's production. A common error is to overgeneralize the ending -e, however, w h i c h is the most frequent across nominative and accusative case i n the strong and weak paradigms. W h e n the strong endings are used, they are usually correct for gender, for example grofi-es Loch ' b i g hole' from a child aged 1;10. W o r d order at this stage is most frequently verb-final, when a verb is ex­ pressed. This is probably attributable to the frequency o f the main verb i n final position i n the input speech, as discussed above. The verb is most frequently i n the infinitive f o r m , that is w i t h the ending -en. When the agent is expressed before the verb, i t becomes more common for the verb to have the third person singular ending -t. This indicates the development o f the l i n k between subject and verb. Past participles are produced at this stage, although i t is frequently not clear that a reference is being made to past time. W h e n these are produced, the past affixes, especially the prefix ge-, are frequently missing, for example *nommen instead o f ge-nommen 'taken'. The past participles are also i n verb-final position as is correct i n adult speech. Auxiliaries and modal verbs are very rare at this stage. Verb particles are common, as i n the one-word stage, but although these frequently have the same forms as prepositions, prepositions themselves are rare. I n negation a development takes place during this stage. The first construc­ tions have nein ' n o ' i n initial position, for example nein trinken ' ( I ' m ) not d r i n k i n g ' . Later the negator changes to nicht ' n o t ' , which is then correctly placed after the verb i n most cases. Intonation questions are produced at this stage, for

2. e x a m p l e M u t t i komment

155

'(is) M u m m y coming?'. Locative questions also start to

appear, for example wo Teddy! 3.3. Three

The Acquisition of German

and More

3 . 3 . 1 . Noun Phrase.

'where (is) teddy?'

Words

(to Age

4;0)

Articles start to be used more regularly, although they

are still omitted on occasion. W h e n used, they are usually correct for gender which is possibly due to sensitivity to morphophonological regularities, though this is not clear from the data. I f an error is made, it is usually to overgeneralize the feminine definite article die, probably since i t is the most frequent f o r m i n nominative and accusative case across the singular and plural paradigms. W i t h the indefinite article, the feminine form eine is overgeneralized, probably be­ cause it is salient through being polysyllabic. The marking o f case i n the nominative and accusative is only apparent i n the masculine gender paradigm. The distinctive marking o f nominative and ac­ cusative is sporadic before age 3;0; otherwise the nominative case f o r m is used. This can probably be attributed to an attempt to regularize the paradigm since i n the feminine, neuter, and plural paradigms there is no distinction. Dative case appears around age 3;0 and is usually marked correctly except after prepositions. Genitive case does not appear marked on the article i n any o f the data reported. The possessive relation is indicated by the suffix -s added to the noun referring to the possessor, w h i c h is ordered before the thing possessed. This is correct i n adult speech w i t h proper names but children overgeneralize this to all nouns, for example *Eisenbahn-s

Wohnung

'railway's house'. Possession

is also expressed by placing the possessor before the thing possessed and using a possessive pronoun, w h i c h occurs i n some dialect forms. This double marking makes the possessive relationship very clear. A third way to mark possession is by a preposition; von ' o f is used but also the incorrect zu ' t o ' . The pronouns ich T

and du ' y o u ' appear around 2;6 and occasionally are

used i n case-marked forms. These uses appear to be formulaic rather than i n d i ­ cating a control o f the inflected forms, for example fiir

dich ( A C C ) 'for y o u '

when g i v i n g someone something, gib mir ( D A T ) 'give m e ' when requesting something. Adjective endings are fairly w e l l established i n the strong and weak para­ digms and case endings by the age o f 4;0. Prior to that, there is some evidence that children attempt to mark the same endings on all forms preceding the nouns, for example mein-*er

gut-er

Papa

' m y good D a d d y ' . The comparative is also

produced w i t h overgeneralization o f the root i n irregular forms, for example guter from gut ' g o o d ' instead o f besser. Prepositions start to appear regularly, predominantly i n locative use, around age 3;0. Accusative case is frequently overgeneralized after prepositions. This is probably due to the easy confusion o f n (marking accusative) and m (marking dative) i n the masculine gender paradigm. F r o m experimental evidence

the

156

Mills

stative meaning appears to be learned before directional meaning w i t h those prepositions which can have both meanings. 3.3.2. Verb. A r o u n d age 3;0 considerable developments take place i n the verb. The verb-final rule is not as prevalent as i n the t w o - w o r d stage, although some children use it even w i t h longer sentences for a while. W i t h the placing o f the verb i n second position after the first element i n the sentence, the marking o f person becomes common and is usually correct. Auxiliaries and modals begin to appear and usually are placed correctly i n second position, w i t h the main verb then i n final position. Few errors occur w i t h this w o r d order. The prefix ge- on the past participle starts to be used but it is still frequently omitted. Future tense starts to be marked w i t h the auxiliary werden. Overgeneralization i n parts o f the verb which are irregular becomes c o m m o n . I n the past participle the regular " w e a k " form is overgeneralized to the irregular " s t r o n g " verbs to produce errors such as *ge-geh-t instead o f ge-gang-en from geh-en 'to g o ' . The irregular forms i n the present tense are also regularized, for example er *lauf-t from lauf-en 'to r u n ' instead o f er lauf-t. The simple past tense starts to be used and here too the regular form is overgeneralized to the irregular forms, for example sie *stehl-t-en from stehl-en 'to steal' instead o f sie stahl-en. I n all these regularizations the child is using the present tense as the basis o f the f o r m , w h i c h indicates that he is able to establish what the present tense is i n each case. 3.3.3. Complex Structures. I n questions, the subject-verb inversion neces­ sary after a question w o r d is acquired. M o r e question words are used, first was ' w h a t ' , later wer ' w h o ' , then wie ' h o w ' and warum ' w h y ' . Tag questions are learned around age 3;0, apparently earlier than i n English. This is possibly due to the simpler structure o f the German form. The passive is sporadic i n spontaneous use until age 4;0 and is only ever agentless. I n comprehension, children o f this age rely heavily on a w o r d order strategy, that is that the noun i n first position is interpreted as being agent. Relative clauses are rare. W h e n they are produced, the verb is correctly i n final position but the relative pronoun is frequently omitted or a form is used which is unmarked for case and gender, for example das Madchen, wo in die Schule geht 'the g i r l that goes to school'. This form wo 'where' occurs as the relative pronoun i n some dialects, but the influence o f dialect on the children's use is unclear. A r o u n d age 4;0 the children w h o have used the wo pronoun, begin using a double form w i t h the standard relative pronoun and wo, for example ein Heinzelmann, der *wo so machen kann 'a Heinzelmann that can go like t h i s ' . Children apparently have difficulty w i t h the relative pronoun since i t is similar to the definite article and must be marked for gender and case. U n t i l they have mastered these features, they omit the pronoun, select a simple form or use the simple together w i t h the more complex f o r m . I n comprehension at age 4;0

2.

The Acquisition of German

157

children are still having difficulty perceiving the structure o f a relative clause and grouping the elements w h i c h belong to i t . They w i l l frequently assume the first noun o f the sentence to be agent i n both the main and relative clause, even when the relative clause occurs at the end o f the sentence. 3.3.4. Analytic Ability. F r o m the questions children ask about language and the lexical innovations they produce, they start to w o r k hard at analyzing struc­ tures between the ages o f 3;0 and 4;0. I t becomes important that parts o f a w o r d contribute clearly to the semantic content o f the whole. False analyses are also made. This fits together w i t h the general tendency to overregularize the most transparent rules. 3.4. Later

Development

(Age 4;0

Onwards)

Constructions which were sporadic previously appear w i t h greater frequency, for example the future tense, relative clauses, and the passive, although this is still predominantly agentless. I n the comprehension o f these last t w o structures under experimental conditions, semantic information is relied on much more for interpretation. O n l y after the age o f 7;0 can structure alone be used to establish agent and object relations independently o f semantic information. Children now seem to be able to group together the elements o f the relative clause but have problems establishing case relations w i t h i n the clause. Where there is no cue from semantic information, they rely on a w o r d order strategy associating subject or agent w i t h first position i n the clause, as do adults i n the interpretation o f ambiguous relative clauses. T i m e prepositions become more common as w e l l as the temporal interrogative pronoun wann ' w h e n ' . Children are still having difficulties w i t h all those parts o f the language w h i c h are highly irregular, celebrated by M a r k T w a i n i n " T h e A w f u l German Lan­ guage." Errors still occur i n the use o f nominative case for accusative case. Plural mistakes are common w i t h overgeneralization o f the most regular form -en. Nouns w h i c h take the zero plural allomorph are often marked for plural, indicating a tendency to mark all forms clearly. The comparative adjective is also commonly marked twice, for example grdfi-er-*er, possibly because the ending is identical to a case/gender inflection i n the masculine paradigm. The possessive pronoun sein (for 3rd person masculine and neuter gender) is overgeneralized to all singular 3rd person singular nouns, whereas ihr (feminine and plural) is used for plural only. A g a i n one form is preferred for one clear function. Although the morphological rules o f noun compounding are complex and have many excep­ tions, children make few errors i n this area. Apparently they learn through the high frequency o f compounds, w h i c h forms are possible for each noun when it occurs as determinans. Errors are still frequent i n the past tense o f irregular verbs. Children tend to overgeneralize the use o f the auxiliary haben to those verbs which should have sein, since this use cannot be predicted by any clear rule. The mistakes w i t h the

158

Mills

simple past and past participle o f irregular verbs seem now to be interrelated. The errors reported tend not to be overgeneralizations o f the regular form but rather an error i n selection o f the change i n the root, for example the past participle ge-ging-t takes the simple past form ging from the verb geh-en 'to go' as the changed root, instead o f the correct ge-gang-en. Although w o r d order errors are rare, errors do occur i n the ordering o f verbal elements i n a subordinate clause where there is more than one verbal element. The rules here are complex and are apparently learned later than age 6;0.

THE

DATA

4. T y p i c a l E r r o r s 4.1.

Word

Order

4 . 1 . 1 . Word Order in the Two-word Phase. W o r d order is an area i n w h i c h there is a considerable amount o f error. A t the stage when t w o - w o r d utterances predominate, many researchers have observed that the dominant w o r d order involves the placement o f the verb i n final position (Preyer, 1882; Stern & Stern, 1928; Park, 1970; G r i m m , 1973; M i l l e r , 1976; Stephany, 1976). I n the t w o children aged 1-2 years, recorded by M i l l e r (1976), this verb-final rule ac­ counted for 70% o f the utterances i n v o l v i n g a verb. The f o l l o w i n g examples are taken from M i l l e r ' s data and collected from one child, M e i k e . They reveal that nouns i n many different functions occur i n first position before a verb. (36) Miller (1;7)

mddi

lafen

girl:AGENT

sleep

'the girl is sleeping' (37) Miller (1;8)

teddy

holen

teddy:OBJ

fetch

'fetch teddy' (38) Miller (1; 10)

hause

gehen

home:LOC

go

'go home' (39) Miller (1;10)

meike

ab-

machen

Meike:BENEF ACTOR

off

take

'take (it) off Meike'

2.

The Acquisition of German

159

I n t w o - w o r d utterances it is not clear that the verb-final position is an error, since it is a rule o f adult German that the verb must occur as second element i n main clauses, whatever the first element may be. Since the verb also occurs i n final position i n three-word utterances at this stage, as the f o l l o w i n g examples show, i t is clear that the child is using a verb-final rule. I t is therefore correctly classified as an error. (40) Miller (1;8)

teddy

sofa

fahren

teddy:AGENT

moped:OBJ drive

'teddy drives the moped' (41) Miller (1; 10)

meike

fenster

gucken

Meike: AGENT

window:LOC

look

'Meike is looking out of the window' W o r d order is clearly not being used as a basic means o f indicating semantic functions at this stage, as Stephany (1976) observed. I t is not, for example, the predominance o f agent function i n the preceding noun which accounts for the data. Roeper (1973) provides an explanation for this rule linked to linguistic theory. He argues that German is an S O V language i n its deep structure and that clauses w i t h the elements i n this order are easier to process. Firstly, however, i n terms o f this linguistic description, i f accepted, it is not at all clear that German is an S O V language or that the question is even relevant (Bartsch & Vennemann, 1972). Secondly, i t is not clear that this underlying w o r d order should necessarily have any relevance for language acquisition. M o r e important for language ac­ quisition seems to be the w o r d order w h i c h is perceived as basic to the language and from w h i c h other w o r d orders seem to be derived by rule. For German, this canonical w o r d order w o u l d be S V O . Roeper's explanation cannot be dismissed, however, since i t does provide an account o f the data, but it can only be accepted i f many other theoretical assumptions are met. It has also been suggested that the verb-final rule may be related to the language input (Roeper, 1972; Stephany, 1976; M i l l e r , 1976). N o full survey o f caretakers' speech to children has yet been carried out for German but from m y o w n data this suggestion seems to have some validity. I n adult German, certain constructions involve positioning the main verb at the end o f the clause. I n subordinate clauses the verb occurs finally but such clauses are less rather than more frequent i n adults' speech to children. W i t h a modal or future auxiliary verb, the main verb is i n infinitive form at the end o f the clause, as sentences 6 and 7 above illustrate. The use o f modals and auxiliaries is common i n care­ takers' speech related to the structure o f the discourse. Characteristic o f adultchild discourse is the large number o f indirect commands, suggestions, etc.

160

Mills

which involve the use o f modals and auxiliaries. Sentences 4 2 - 4 4 taken from m y data illustrate such use. (42) möchtest

du

want [MOD] you

ein

Haus

bau

INDEF.ART: NEUT:SG:ACC

house

build INF

-en?

'would you like to build a house?' (43) willst

du

want [MOD] you

pipi

mach

-en?

wee-wee

do

INF

'do you want to do a wee-wee?' (44) du you

sollst

nicht

should [MOD]

N E G cry

wein

-en INF

'you shouldn't cry' A n alternative imperative f o r m also involves positioning the verb at the end o f the clause. Sentences 45.a and 46 illustrate this i n comparison w i t h the standard 2nd person singular imperative form i n 45.b and 47. This infinitive-like form is particularly common w i t h negative commands. (45. a) nicht NEG

beiss

-en!

bite

INF/IMP

'don't bite!' (45.b)beiss bite

-e

nicht

2SG:IMP N E G

'don't bite!' (46) jetzt now

auf-

stehen!

up

stand

'now stand up!' (47) steh stand

-e

jetzt

auf!

2SG:IMP

now

up

'stand up now!' A construction w h i c h occurs i n regional dialects but w h i c h is not an accepted standard German f o r m , involves the use o f the verb tun ' d o ' as an auxiliary i n a declarative sentence. Its use seems to involve no change i n meaning, however, so that sentence 48 is to be considered as identical i n meaning to sentence 49.

2. (48) wir we

The Acquisition of German

tu

-n

hier

Bilder

mal

AUX

1PL:PRES

here

pictures

paint INF

161

-en

'we're painting pictures here' (49) wir we

mal

-en

hier

Bilder

paint

1PL:PRES

here

pictures

'we're painting pictures here' The use o f the auxiliary tun again sends the main verb to the end o f the clause. In m y o w n research, samples o f speech from three mothers have been ana­ lyzed. I t was found that the number o f constructions i n v o l v i n g the main verb i n final position was extremely high. A l l three mothers used the dialectal tun, although one mother was not a speaker o f a dialect w h i c h allows this construc­ tion. Her use o f this construction was limited to speech addressed to her child. I t is therefore possible that this construction is a baby-talk feature, but this needs to be investigated further. A l l three mothers d i d produce syntactic baby-talk, how­ ever, i n that they produced utterances w i t h a verb final rule and w i t h the verb i n infinitive form, although no modal or auxiliary was used. Sentence 50 taken from m y data illustrates this, i n contrast to the standard form o f 5 1 . (50) Mama mummy

Bonbon

ess

-en?

sweet

eat

INF

'mummy eat the sweet?' (51) soil

ich

should [MOD] I

das

Bonbon

ess

-en?

DEF.ART: NEUT:SG:ACC

sweet

eat

INF

'should I eat the sweet?' The mothers varied i n their use o f such baby-talk. I n one mother the number o f such utterances was extremely high. This verb-final rule i n the children is captured therefore by the universal suggested by Slobin (1973, p . 197) " w o r d order i n child speech reflects the w o r d order o f the input language." Particular emphasis is necessary i n the case o f German on I N P U T L A N G U A G E . The Operating Principle related to this universal: "pay attention to the order o f the words and morphemes" w o u l d also be sup­ ported i n the case o f German by the generalized operating principle: " p a y attention to the ends o f words and sentences.'' I t w o u l d appear that the role o f the latter principle must be contributory rather than providing a total explanation, since i t w o u l d not explain, for example, English data. It has been argued that the structure o f the caretakers' discourse is an explana­ tion, at least i n part, for the high frequency o f main verb i n final position. This

162

Mills

was related to the adults' use o f command, suggestions, etc. I t w o u l d be interest­ ing to know whether the children also are producing a large number o f com­ mands, questions, etc. I f this were the case, it is possible that the verbHfinal rule is linked to the absence o f the use o f auxiliaries and modals i n the child's o w n speech. 4.1.2. Word Order in Longer Sentences. German has complex word-order rules, as has already been discussed. Unfortunately, apart from the data present­ ed above, little information is available about word-order rules at later stages and in other constructions. The verb-final rule was still being applied w i t h three-word utterances i n M i l l e r ' s data, w h i c h is an error i n terms o f adult word-order rules. The rule then appears to be quickly learned that the verb is i n second position whatever element is placed first, although one child has been recorded i n m y o w n data using the verb-final rule i n four- and five-word utterances. There is no clear evidence that the canonical S V O order predominates i n simple sentences. A n investigation o f the relationship o f input to the child's utterances needs to be carried out i n this area on the lines o f Bowerman (1973). Stern and Stern (1928) discuss w o r d order in longer utterances to some extent. They attribute deviant w o r d order on the one hand to " i m i t a t i o n " and on the other to "spontaneity." I n the case o f imitation, they claim, the child's order has been affected by the immediate input speech. (52) Stern (ca. 2;6) ADULT:

wo-

mit

ha

-st

Du

ge-

pfiffen?

INT.PRO

with

AUX

2SG

you

PAST

whistle:PAST

'what did you whistle with?' CHILD:

*

4

Mund

mit

mouth

with

'with (my) mouth' The correct answer should be mit dem Mund. The principle o f spontaneity is related to topicalization and o l d and new information, but without a greater amount o f systematic data and a detailed linguistic account o f adult speakers' usage i n this respect, this principle cannot usefully be discussed further here. 4.1.3. Word Order in Questions. Wode (1971) describes one systematic error i n the acquisition o f w o r d order i n questions. The use o f a question w o r d requires the inversion o f subject and verb, as described earlier. One o f the first The children's utterances quoted here and throughout the chapter are often deviant in more than the particular part marked with an asterisk. For the purposes of simplicity, the asterisk is used to mark only the deviant usage which is immediately under discussion. 4

2.

The Acquisition of German

163

question words is the interrogative pronoun wo 'where', as i n English. First o f all Wode's t w o children used wo i n initial position but w i t h no verb. Later they produced well-formed questions using wo, that is w i t h subject-verb inversion, but only w i t h forms o f the verb sein 'to be' as i n sentence 53. (53) Mama, mummy

wo

bist

du?

INT.PRO:LOC

be:2SG

you

'mummy, where are you?' No other verbs were recorded. Before producing well-formed questions w i t h all verbs, one child had a long intermediate stage using the order: wo SUBJECT V E R B . The other child had only a few utterances i n this order. Unfortunately all the utterances quoted contain only three words so that it is impossible to decide whether the child is using a word-order rule appropriate to a main clause but w i t h the question w o r d i n initial position or a verb-final rule. The addition o f a fourth linguistic unit X w o u l d indicate w h i c h interpretation was correct. The first rule would produce: woSVX and the second rule: woSXV Wode suggests i t is the latter, on the basis that this one child also had a lengthy stage o f verb-final rule i n noninterrogative utterances. W h e n he comes to sum­ marize the general features o f development, he describes word-order rules i n questions by saying: " M u l t i w o r d I(ntonation) Q(uestions) at first do not reflect adult interrogative w o r d order i f i t differs from the non-interrogative

one"

(Wode, 1971, p . 308). He does not however specify w h i c h w o r d order is then preferred. I n interrogative pronoun questions, the first rule could be captured by Slobin's (1973) Universal D l : "Structures requiring permutation o f elements w i l l appear first i n nonpermuted f o r m . " I f the child were using the second rule mentioned above, however, this w o u l d appear to be an overgeneralization o f the child's o w n noninterrogative order, that is w i t h a verb-final rule. Unfortunately Wode does not present an analysis o f the acquisition o f yes-no questions i n German w h i c h involve inversion o f subject and verb. Data on this area might suggest w h i c h explanation is correct. These rules all f i t together w i t h the acquisi­ tion o f w o r d order-rules i n noninterrogative sentences, where any element X at the beginning o f a sentence w h i c h is not subject causes an inversion o f verb and subject. Unfortunately data are not available on this area as a whole. Wode does note, however, that his impression is that inversion first occurs w i t h yes-no

164

Mills

questions and that once the other interrogative pronouns appear, that is those acquired after wo, the inversion rule appears to have been fully learned. 4 . 1 . 4 . Word

Order

in Subordinate

Clauses.

There are few data on the

acquisition o f the verb-final rule i n subordinate clauses. Stern and Stern (1928), for example, do not report any w o r d order errors i n subordinate clauses. I n other diary studies the first utterances containing a subordinate clause have the verb already i n clause-final position. (54) Scupin (2;2)

mal sehen,

daß fort

ist

just

that

is

see:INF

away

'let's see that it has gone away' (55) Preyer (2;8)

weiß

nicht,

wo

es ist

know

N E G INT.PRO:LOC

it

is

'(I) don't know where it is' A w e l l - k n o w n children's rhyme Backe, backe Kuchen places an infinitive after a modal i n non-clause-final position i n order to obtain a rhyme: der muß habenIsieben Sachen. A t age 2; 10 Scupins' child corrects this line to produce the grammatical utterance: (56) Scupin (2; 10)

der

muß

he must [EMPH] [MOD]

sieben

Sachen

hob

-en

seven

things

have INF

'he must have seven things' The verb-final rule i n subordinate clauses w o u l d seem to be w e l l established at an early age and could be seen as a continuation o f a general verb-final order dominant at the t w o - w o r d stage. It w o u l d be interesting to k n o w h o w the acquisition o f the verb-final rule i n subordinate clauses relates to the acquisition o f subject-verb inversion i n de­ clarative main clauses w h i c h do not begin w i t h the subject. One could be seen as a greater permutation than the other b y t w o lines o f argumentation. A n y rule which removes subject from initial position is more complex and acquired later than a rule w h i c h preserves i t i n initial position, that is subject/verb inversion is acquired later than verb-final rule. Or the reverse, any rule w h i c h moves an element a greater distance from its original position is more complex and ac­ quired later than one w h i c h moves an element a shorter distance, that is the verbfinal rule is acquired later than subject-verb inversion. Park (1976), i n an imitation experiment, tested the ability o f 3-, 4- and 5-yearolds to imitate, amongst other things, subordinate clauses w i t h verb-final rule

2.

The Acquisition of German

165

which are grammatical strings (sentence 57). He also tested imitation o f ungrammatical versions, that is w i t h S V O order (sentence 58) and w i t h subject-verb inversion order (sentence 59). (57) die

Kinder

sind froh,

weil

der

Vater

are

because

the

father

the

children

ein

Auto

kauf

-t

happy

a

car

buy

3SG

'the children are happy, because their father is buying a car' (58) *Die Kinder sind froh, weil der Vater kauft ein Auto. (59) *Die Kinder sind froh, weil kauft der Vater ein Auto. He used the subordinating conjunctions weil 'because' and warum

'why'.

Imitation o f ungrammatical utterances is a problematic test procedure but, nevertheless, i t is significant that the majority o f children i n this age range correctly imitated the grammatical form w i t h verb-final rule. The behavior w i t h the ungrammatical clauses was different according to w h i c h subordinating con­ junction was used. M o r e children changed the ungrammatical clauses w i t h weil than they d i d w i t h warum, transforming them into the grammatical form using the verb-final rule. The verb-final rule thus seemed far better established w i t h weil than w i t h warum. Park suggests, and I think this is clearly correct, that the difference is attributable to the different status o f the t w o conjunctions. Weil can only be used w i t h verb-final rule, whereas warum can also be used as an inter­ rogative pronoun where subject-verb inversion rules apply. T w o children con­ verted the subject-verb inversion sentences into the S V O f o r m . T w o children, however, used the subject-verb inversion rule i n their imitations o f the gram­ matical and the SVO-rule sentences; one child used this rule i n his imitation o f all forms. These results, although they only test t w o conjunctions w i t h a small sample o f children ( n = 7 ) , suggest that, although the hypothetical principles discussed ear­ lier may be significant, an important factor is the status o f the conjunction i n the child's grammar. The t w o functions o f the same w o r d warum have to be dis­ tinguished. Park (1971b) has one example, from a spontaneous utterance, o f a word-order error. (60) weil because

ha

-st

du

das

ge-

sag

-t

AUX

2SG

you

DEM.PRO:ACC

PAST

say

PAST

'because you said that' The child is treating weil as i f i t were an X element occurring at the beginning o f the sentence rather than as a subordinating conjunction. This one error might

166

Mills

suggest that subject-verb inversion has been i n this case learned before the verbfinal rule, or that the subject-verb inversion rule may dominate at one stage when the verb-final rule is perceived as inappropriate i n declarative main clauses. Far more data are needed from many children before any f i r m conclusions can be drawn. Although there are generally few word-order errors reported i n subordinate clauses, older children still make errors i n subordinate clauses where t w o or three verbal elements have to be ordered. This is to be expected since, i n this case, the rules o f ordering i n adult German are more complex than the simple verb-final rule. I f a subordinate clause contains t w o verbal elements, for example a modal and infinitive or an auxiliary and participle, the finite verb is ordered after the other verbal element. I f the clause contains three verbal elements, the finite verb is ordered before the other t w o . (61) weil because

du

gestern

komm -en

you

yesterday

come

konn

INF can

-te

-st

PAST

2SG

'because you could come yesterday' (62) weil because

du

gestern

hätte

-st

komm -en

you

yesterday

have:PAST

2SG come

können

INF camPAST

'because you could have come yesterday' The material o f both Scupin and Scupin and Stern and Stern contains examples o f mistakes i n such clauses as late as four and five years. I n subordinate clauses containing t w o verbal elements, the finite verb is sometimes placed before the other verbal element instead o f after i t . (63) Scupin (2;5) *weil

ich

dir

den

Apfel

because

I

you: DAT

theACC

apple

hab

(ge)geben

AUX

give:PAST

'because I have given you the apple' (64) Scupin (4;6) *daß

man

nicht

that [CONJ]

one

N E G down

runter

darf

fall

may

fall INF

'that one may not fall down' (65) Stern (5;1) *wenn

du

if

you:NOM

uns it:ACC

nicht

us:DAT N E G

-en

2. wiirde

-st

would

2SG give

geb

The Acquisition of German

167

-en INF

'if you wouldn't give it to us' In other clauses the finite verb is ordered soon after the subject. (66) Scupin (4;7) *wenn

du

mich

hätte

if

you:NOM

me:ACC

have:COND 2SG

einen

Pfennig

in den

INDEF. ART: A C C

penny

in lassen

Automaten

steck

-en

slot:machine

put

INF

-st

DEF.ART:ACC

allow:PAST

'if you had allowed me to put the penny in the slot machine' (67) Stern (5;5) *wenn

ihr

würd

-et

if

you:PL:NOM

would

2PL always

in

Berlin

ge-

blieben

in

Berlin

PAST

stay:PAST

immerfort

'if you were always to stay in Berlin' These errors occur mainly i n conditional clauses where it is most common to have three verbal elements and therefore the more complex rule i n adult speech. It w o u l d seem as though the child has associated a different ordering w i t h this type o f clause or possibly w i t h all clauses containing more than one verbal element but has not yet learned all the details o f the rule w i t h regard to the restrictions on its application and the positioning o f the finite verb. F r o m the data, the children involved show awareness o f the fact that the rule involves multiplicity o f verbal elements i n the context o f subordinate clause and the moving o f the finite element out o f final position. This indicates a quite high level o f analysis i n the child. 4.2.

Verb

Morphology

4 . 2 . 1 . Compound Past Tense. The verb morphology o f the past tense is a fairly complex area and is the source o f considerable error. M a n y errors occur i n the formation o f the compound past tense which requires the use o f an auxiliary and the past participle w h i c h is derived from the verb root. T o this root is added the prefix ge- ( w i t h a few exceptions) and the suffix -t. Irregular or " s t r o n g "

168

Mills

verbs generally have the suffix -en added to the root w h i c h frequently has a different form, for example: regular—kauf -en irregular—geh -en lauf -en sing -en

'to 'to 'to 'to

buy' go' run' sing'

gegegege-

kauf gang lauf sung

-t -en -en -en

The past participle is placed at the end o f the clause. There is some evidence that the past participle appears first, without the auxiliary and prefix ge-, at a relatively early age. (68) Scupin (1;11) *Decke

putt

mach

-t

ceiling

broken

make

PAST

'the ceiling has been broken' (?) The context is often unclear, however, so that it cannot be established w i t h certainty that a reference is being made to past time. The auxiliary does not appear until around age 3;0 but then i t is i n frequent use. The prefix ge- on the past participle also appears around this time but seems to present some difficulty in that its acquisition lasts a good deal longer (Preyer, 1882; Park, 1971b). MacWhinney claims that this is attributable to the fact that the auxiliary and ge­ nre parts o f a discontinuous morpheme and the auxiliary is the part w h i c h is "subject to the most easily controlled v a r i a t i o n " ( M a c W h i n n e y , 1978, p . 53). This property he claims to be an important factor i n the acquisition o f mor­ phology. This may w e l l be a contributory factor i n the earlier acquisition o f the auxiliary but the fact that the prefix ge- is an unstressed syllable also seems to be important. I n other words beginning w i t h this syllable, but not the past mor­ pheme, pronuciation is reported where this syllable is omitted. (69) Preyer (2;3) *fährlich instead of gefährlich 'dangerous' (70) Grimm (1973) *rage instead of Garage 'garage' Omission o f unstressed syllables is reported i n the acquisition o f many languages and w o u l d seem to be the most likely explanation here. The early forms o f the past participle, w h i c h the child produces, i f they can be clearly classified as such, are often correct i n terms o f any v o w e l changes i n the strong verbs and the selection o f the -t or -en suffix. A r o u n d age 3;0 however there is a frequent overgeneralization o f the weak form o f the past participle. A l l

2.

The Acquisition of German

169

the diary studies report such overgeneralization. The root used most often is the root which appears i n the infinitive. This root M a c W h i n n e y calls the " c i t a t i o n a l l o m o r p h . " For example: (71) Scupin (2;8)

*ge- geh -t from geh-en 'to go' instead of ge- gang -en *ge- denk -t from denk-en 'to think' instead of ge- dach -t

MacWhinney claims that it is a general principle i n the acquisition o f mor­ phology that the citation allomorph be taken as a base for inflections. There are some examples however o f the simple past form being taken as the root for the past participle: (72) Scupin (5;0)

*ge- ging -t from (ich) ging T went' instead of ge- gang -en

Since these examples are mostly later, this suggests that the child has established that a different root is used i n the past o f the strong verbs but has not sorted out which root form belongs to w h i c h past tense f o r m . Some o f the strong verbs have the same root i n the simple past as i n the past participle, so that the child may be overgeneralizing this class. Further evidence o f the child attempting to regularize the past forms comes from Neugebauer's (1915) child who at one stage formed irregular past partici­ ples (without ge-) o f verbs w h i c h are regular using the vowel lol or / o / i n every case. For example: (73) Neugebauer (2;2) *um-kopp-en

from um-kipp-en 'to tip over'

*klob-en

from kleb-en 'to stick'

instead of um-ge-kipp-t instead of ge-kleb-t A l l the examples Neugebauer recorded were from verbs which contain the vowel /e/, / I / , or Izl i n the infinitive. A possible hypothesis is that the child was attempting to regularize the past participle forms on the basis o f the vowel i n the stem, like the correct examples: ge-nomm-en from nehm-en 'to take', geschwomm-en from schwimm-en 'to s w i m ' . 4.2.2. Auxiliary in the Past Tense. I n the compound past tense either the auxiliary hab-en 'to have' or sein 'to be' is used. The selection is determined by rule (Helbig & Buscha, 1972, p . 119). The verbs w h i c h take the auxiliary sein i n

170

Mills

the past tense are relatively few and most commonly are perfective or a dynamic verb which involves a change o f location. For example: (74) der Patient ist gestorben

the patient has died'

4

(75) der Sportler ist ge-laufen

'the sportsman has run'

A few verbs can use both auxiliaries and the difference usually reflects the transitive versus the intransitive use or the durative versus the perfective aspect. For example: (77) ich habe den Bus ge-fahren

'I drove the bus'

(78) ich bin zu schnell ge-fahren

'I drove too quickly'

(79) sie hat jede Woche ge-tanzt

'she went dancing every week'

(80) sie ist durch den Saal ge-tanzt

'she danced across the room'

Stern and Stern observed that Hildegard always used the correct auxiliary at age 3;2. Park (1971a,b) observed no errors i n the choice o f the correct auxiliary i n his data either and suggests the child is using the rule that haben is linked w i t h transitive verbs and sein w i t h intransitive verbs. A strict application o f this rule would produce many errors but Park's data contain no such errors. Scupin and Scupin's observations, however, contain examples o f auxiliary errors from age 2; 10 until age 5;0. I n every case the child selected haben rather than sein as would be expected on the basis o f frequency o f use. This is possibly an area where individual differences i n the tendency to overgeneralize as opposed to learning by rote could explain the variation. On the other hand, Ramge (1973) reports frequent self-corrections i n the auxiliary i n 6-year-olds, so that i t w o u l d seem rather to be an area o f general uncertainty. 4.2.3. Simple Past Tense. The formation o f the simple past tense is an area which produces many errors, since the group o f strong verbs take different endings from the weak verbs and often have a different root. For example: regular—kauf-en irregular—geh-en lauf-en sing-en

ich ich ich ich

kauf- te 'I bought' ging 'I went' lief 'I ran' sang T sang'

The simple past tense appears later than the compound past tense and so the reported errors occur correspondingly later. Only Scupin and Scupin report er­ rors i n this tense; other studies do not comment on its acquisition. Scupins' child

2.

The Acquisition of German

171

overgeneralized the weak ending -te w i t h strong verbs, sometimes using the root from the infinitive and sometimes making a change. For example: (81) Scupin (4;4)

er *kam-te

from komm-en 'to come' instead of er kam

(82) Scupin (5;0)

er *gang-te

from geh-en 'to go' instead of er ging

(83) Scupin (5;5)

sie *stehl-ten from stehl-en 'to steal' instead of sie stahl-en

The same child produced also the f o r m *ging-te (3;7) and *gang (4; 10) for the simple past o f gehen. There are t w o separate problems here for the child: w h i c h verbs take the -te endings and w h i c h verbs have a changed root and what is the form o f that root. The overgeneralization o f -te can be explained by its being the most frequent form but also by its being a clear marker o f simple past. There is one example o f the same v o w e l being overgeneralized i n the simple past. Scupins' child i n a highly excited state (5;0) produced all the simple past forms o f weak and strong verbs i n a narrative so that they contained the v o w e l /a/. I t is not clear how this can be explained, but the child is obviously w o r k i n g on the basis that the v o w e l is changed i n the simple past but has not restricted the application o f this rule reliably. The strong verbs are often those most frequently used. The notion o f frequency must include here the number o f different verbs and the amount o f use o f the verbs. M o r e data are needed to determine whether vowel change is frequently used as a basis for overgeneralization. 4.2.4. Present Tense. N o t many errors are reported i n the present tense. The few that are reported are i n connection w i t h the irregular forms, as might be expected. The irregular forms occur only i n the singular o f some " s t r o n g " and " m i x e d " verbs, and involve a change o f stem and, i n modal verbs, the omission of the inflections marking first and third person (compare Table 2.3). For example: from lauf-en 'to run'

from woll-en 'to want'

ich du es/sie/er wir ihr sie

lauf-e läufst läuf-t lauf-en lauf-t lauf-en

ich will du willst eslsieler will

T run' 'you run' 'it/she/he runs' 'we run' 'you:PL run' 'they run' ' I want' 'you want' 'it/she/he wants'

172

Mills

wir woil-en ihr woll-t sie woll-en

'we want' 'you:PL want' 'they want'

The verb sein 'to be' is irregular i n every person. Scupin and Scupin's child correctly produced the 3rd person plural o f the verb sein at age 2;0 and i n fact corrected himself when he erroneously produced the singular. Also at age 2;0 he regularized the 3rd person singular forms o f irregular verbs. (84) Scupin (2;0)

*hab-t from hab-en 'to have' instead of ha-t

(85) Scupin (2;0)

*lauf-t from lauf-en 'to run' instead of lauf-t

(86) Scupin (2;0)

*hal-t

from halt-en 'to hold' instead of hdl-t

Around age 3;0, however, he overgeneralized the irregular singular root to the 3rd person plural and added -n. For example: (87) Scupin (2; 11)

*kann'n

from konn-en 'to be able' instead of konn-en

(88) Scupin (3;0)

*is-n

from sein 'to be' instead of sind

(89) Scupin (3;2)

*will'n

from woll-en 'to want' instead of woll-en

These errors are only w i t h modal verbs or sein 'to be'. I t is significant that this overgeneralization occurs rather than the overgeneralization o f the regular form. It is likely that frequency is part o f the explanation o f these errors, that is the form most frequent w i t h i n the paradigm is taken as a basis for generalization. W i t h the modals, however, frequency o f use is probably significant i n that the child uses singular more often than plural. 4.3. Noun

Phrase

Morphology

4 . 3 . 1 . Gender. M o r p h o l o g y i n the noun phrase is a highly complex area and is one o f the most frequent sources o f error. Both case and gender must be marked, as has been discussed earlier, but they are marked by one form o n l y , so that i t is difficult to establish whether an error is due to the selection o f w r o n g gender or w r o n g case or both. The system is also potentially very confusing (see

2.

The A c q u i s i t i o n o f G e r m a n

173

Table 2 . 1 ) , since w i t h i n a gender paradigm the same form can mark different cases, for example, for feminine and neuter nouns the nominative and accusative cases i n articles are marked i n the same way. The same forms occur across gender paradigms marking the same cases, for example the dative case i n mas­ culine and neuter nouns. Finally, identical forms occur marking different cases i n different gender paradigms, for example the nominative o f the definite article w i t h masculine nouns der is identical to the dative o f the definite article w i t h feminine nouns. Adjectives are used i n attributive position before articles are i n frequent use. When the adjective is used without a preceding article, the strong endings are appropriate (see Table 2 . 2 ) , w h i c h distinctively mark gender—at least i n n o m i ­ native and accusative cases. These cases, as associated w i t h subject and object function, predominate i n early usage and the gender marking is frequently correct: (90) Preyer (2;7)

ganz

-es

whole

Battalion battalion:NEUT

'whole battalion' (91) Scupin (2;2)

groß

-er

big

Ball ball:MASC

'big ball' (92) Scupin (2;2)

so

klein

such

little

-es Pappa daddy

-le DIM:NEUT

'such a little daddy' (93) Scupin (2;2)

ganz

klein

very

little

-e

Großmama

grandma: F E M

'very little grandma' The ending -e is also overgeneralized but this need not necessarily be a gender mistake. This ending is amongst the most frequent w i t h i n the paradigms and is probably overgeneralized for that reason (see later discussion). I t is rare that the -er and -es endings are used w i t h nouns o f a different gender. Scupins have only one such example: (94) Scupin (2;0)

klein

* - e s Durst

little

thirst:MASC

'little thirst' Gender w o u l d appear to be established therefore at an early stage.

174

Mills

The first occurrences o f the definite and indefinite articles appear to be as part of an amalgam w i t h the nouns. These first forms are also frequently reduced. I n the definite articles, the f o r m de is reported by Preyer at age 2;6 (Preyer, 1882, p . 330) and by Stern and Stern i n both their daughter and son at age 1;10 and 1;2 respectively (Stern & Stern, 1928, pp. 4 4 , 86). This form can be seen as a reduction o f the nominative forms der ( M A S C ) und die ( F E M ) but is quite distinct from the neuter f o r m das. I n fact, no errors are reported o f this form being produced w i t h neuter forms. The indefinite article is commonly reduced to n or e w h i c h could apply to all genders. Stern and Stern report the insertion o f this form on a large scale. Hilde (1;10) and Gunther (2,4) also inserted i t i n inappropriate contexts, for example before adverbs, so that its status is unclear. The correct choice o f gender only becomes clear when the full form o f the articles is produced. The three gender paradigms are distinct i n the nominative case o f the definite article, so that the use o f the definite article provides the clearest evidence as to the correct selection o f gender. The definite articles appear gradually from around age 2;0 onwards but i t is common for the article to be missing i n an utterance until age 2;6. W h e n the definite article is produced i n its full f o r m , i t is most frequently correct i n gender. A g a i n , at this age, n o m i ­ native and accusative cases i n association w i t h subject and object function pre­ dominate, where there is a clear distinction between all these gender paradigms. Even where nominative case may be used instead o f accusative case i n the masculine paradigm (see 4 . 3 . 2 ) , the form is marked for gender distinctively from the others, so that one cannot talk o f a gender error. In Preyer's data, no gender mistakes occur but it is not a point on w h i c h he comments. Examples o f the three genders used correctly appear by age 2;9. Stern and Stern claim the correct distinction o f the masculine and feminine forms by age 2;6 i n H i l d e , but they neglect to report whether she used the w r o n g article w i t h neuter nouns or whether the article was omitted. Scupin and Scupin, howev­ er, claim that when the definite article first was i n frequent use i n their son at age 2;3, he made many errors, using all three forms quite unsystematically. They do not give examples o f such errors however and from the utterances reported over the following period i n the diary, a different picture emerges. I t must not be forgotten that Scupins' diary consists o f almost daily entries w h i c h are not summarized or collated at any point, so that i t is not clear over what period o f time the observations apply. I n the majority o f utterances recorded, Scupins' son does not make mistakes w i t h the gender o f the definite article. The errors w h i c h do occur w i t h the definite article suggest an overgeneraliza­ tion o f the form die. Scupins' data include three such errors over the period 2;3 to 2;8: *die

Hochstein

( M A S C ) 'the s u m m i t ' , *die

Mann

( M A S C ) 'the m a n ' ,

*die

Truthahn ( M A S C ) 'the turkey'. Only one other mistake is reported and, i n this instance, the neuter is used *das Kopf ( M A S C ) 'the head'. Stern and Stern's data contain five gender errors (up to age 2;8), all o f w h i c h indicate overgeneraliza­ tion o f die. Gunther's reported utterances contain t w o errors: *die bebau (= Wau-

2.

The A c q u i s i t i o n of German

175

Wau, Hund ( M A S C ) ) 'the d o g ' at age 1;2 and *die ding ( N E U T ) 'the thing' at 2;7. Three errors made by H i l d e were when using the definite article die as a demon­ strative pronoun but quite clearly i n the context o f referring to a non-feminine noun Keks ( M A S C ) 'cake', Papier ( N E U T ) 'paper', Papa ( M A S C ) ' D a d d y ' . Data from spontaneous utterances o f D o w n ' s Syndrome children support this finding that die is overgeneralized (Schaner-Walles, personal communication). In an experimental study ( M i l l s , 1978), children aged 5 to 10 were required to assign the definite article to ten familiar nouns (all names o f toys). This test was carried out i n the context o f a study examining the relationship between natural and syntactic gender. They were presented w i t h the three gender possibilities for example *der Buch, das Buch, *die Buch 'the book' and asked to choose the correct f o r m . Even under these highly artificial conditions, the children made very few mistakes. Mistakes were only made by the 5- and 6-year-olds. The feminine gender nouns had the highest percentage o f correct responses, as can be seen from Table 2.4. A n overgeneralization o f die was also relatively c o m m o n , and more frequent than the overgeneralization o f der and das. The selection o f the correct gender seemed also to be dependent on the noun. The greatest number o f mistakes were made on the masculine and neuter animate nouns, as opposed to the masculine and neuter inanimate nouns and feminine animate and inanimate nouns. This difference is possibly due to the referents being animate but not clearly o f one sex or the other, therefore the semantic possibility o f female sex is increased. This point w i l l be taken up again later i n a discussion o f natural and syntactic gender. Inspection o f the scores o f individual children also reflected regularities on this task. T w o children (boys) substituted der for all their errors, 3 children (1 boy, 2 girls) substituted das and 5 children (3 girls, 2 boys) substituted die. Only

TABLE 2.4 Percentage of Correct Responses and Substitutions in Choice of Grammatical Gender by German Children A g e d 5 - 6 Years (from Mills, 1978) Correct Grammatical Gender der n = 144

die n = 144

das n = 192

der

82

3

10

die

10

94

8

das

8

3

82

Response

176

Mills

one child made a large number o f completely mixed substitutions. Three factors seem to influence the results i n this task: a tendency to overgeneralize die, an influence o f the animacy o f the noun, and an individual overgeneralization strategy. There is not enough evidence from the diary studies to suggest the role the latter t w o factors might play i n spontaneous production. They may possibly be artifacts o f the task. MacWhinney (1978) tested gender assignment to real and nonce words i n children aged 3 years to 12 years (see Table 2.5). The words were selected i n order to examine the effects o f inherent semantic gender and phonological end­ ings, the real and nonce words being given the same structure. They were presented i n three conditions: w i t h no overt cuing, and w i t h overt cuing through previous mention o f either the indefinite article i n accusative case or the pronoun in accusative case. M a c W h i n n e y found that age improved performance on the task. The older children were more able to make use o f overt cuing and pho­ nological information, especially i n the nonce words. The older children used das more frequently, especially w i t h the nonce words. Phonological information was most readily used i n the feminine nonce words. MacWhinney concludes that children i n German make little use o f inherent semantic gender to determine syntactic gender, since this is o f limited ap­ plicability. He also compares German children's use o f analogy w i t h that o f Hungarian children and draws the conclusion that German children use analogy less, since the regularities are fewer. He does not offer any explanation for the TABLE 2.5 Real a n d Nonce W o r d s Used b y M a c W h i n n e y (1978), Endings U n d e r l i n e d Real Noun

Meaning

Nonce

Meaning

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

man soldier hammer rug basket tree woman cow pipe mess hand clock child horse monster chair bed glass

der Gann der Molat der Fammer der Leppich der Norb der Faum die Lau die Puh die Neife die Teinerei die Gand die Muhr das Pind das Nerd das Heusal das Nülchen das Rett das Sennas

samurai archer barbapapa doll turtle green man orange figure dinosaur fairy barbapapa doll june bug noise maker ichthysaurus tyranosaurus rex octopus stick figure fire engine robot kiwi mouse

der Mann der Soldat der Hammer der Teppich der Korb der Baum die Frau die Kuh die Pfeife die Schweinerei die Hand die Uhr das Kind das Pferd das Scheusal das Stühlchen das Bett das Glas

2.

The A c q u i s i t i o n of G e r m a n

177

overgeneralization o f das by older children, w h i c h does not fit w i t h the results reported earlier. M a c W h i n n e y ' s material is however problematic w i t h respect to his classifica­ tion o f the test words. Some words w h i c h he claims to be linked w i t h a particular gender because o f their phonological or morphological structure cannot be so clearly classified. I n the group o f words w h i c h should have arbitrary gender assignment, some regularities were overlooked. The explanation o f the selection by the older children o f das, as opposed to die as might be expected from previous results, can be explained by the monosyllabicity o f the majority o f items. Monosyllabicity is strongly associated w i t h masculine and neuter genders, not w i t h feminine genders. Masculine and neuter gender also seem closer to each other and quite distinct from feminine gender. Some loan words are used w i t h different genders, but disagreement occurs main­ ly as to whether nouns should be masculine or neuter, not as to whether they should be masculine or feminine, neuter or feminine (Carstensen, 1980). M u c h more investigation needs to be made o f the phonological regularities i n German and their accessibility to children on the lines o f the w o r k i n French by Tucker et al. (1977) and by Karmiloff-Smith (1979). W o r k is being currently conducted by Z u b i n and Kopcke on the existence o f regularities i n gender assign­ ment ( Z u b i n & Kopcke, 1981; Kopcke, 1982; Kopcke & Z u b i n , 1982). They have already established the existence o f rules based on semantic, mor­ phological, and phonetic properties. The results o f their w o r k w i l l make i t easier to look for the use o f regularities i n children ( M i l l s , 1984). The early overgeneralization o f die noted i n the spontaneous production data can be accounted for by the frequency o f the form die. F r o m Table 2 . 1 , it can be seen that this form accounts for 50% o f the nominative and accusative case forms. Children w o u l d also seem to be aware o f the influence o f syllabicity i n gender assignment. M a c W h i n n e y ' s results indicated no use o f semantic cues by children, but until this is correctly controlled for i n variation w i t h other pho­ nological and morphological regularities, it cannot be eliminated. The evidence for use o f phonological cues by children i n assigning gender is as yet unclear. Since the regularities as yet described (e.g. Z u b i n & Kopcke, 1981) relate to relatively small categories o f nouns, i t could w e l l be expected that the use o f phonological cues w i l l develop slowly i n accordance w i t h M a c W h i n n e y ' s princi­ ple o f production applicability. The size o f the child's lexicon and the status o f particular items i n the lexicon w e l l may influence the regularities to be first used. Parallel to the overgeneralization o f the feminine definite article, the feminine form o f the indefinite article eine also appears to be overgeneralized i n the early stages. Park (1976) reports this and there is supporting evidence from the diary studies. Scupin and Scupin note *eine strump ( M A S C ) 'a sock' at age 2;0. The possessive adjectives are declined i n the same way as the indefinite article, and here too the feminine form is overgeneralized. Stern and Stern record *meine appele

(NEUT)

4

m y little apple' at age

1;10

and meine

referring to

Mantelchen

( N E U T ) 'little coat' at 2;6. The dominance o f this f o r m , however, cannot be

178

Mills

explained by the frequency principle. Eine and related forms may be more salient than ein because they are polysyllabic, for example. Quite obviously, Greenberg's (1966) proposed universal that masculine forms are unmarked and the associated acquisition principle that unmarked forms are acquired before marked forms is inappropriate here. This principle may be rele­ vant when syntactic gender is related to natural gender (see

Deutsch

&

Pechmann, 1978, and subsequent discussion). Although it seems more than a coincidence that the feminine gender forms i n three forms should be overgeneralized, namely the adjective ending -e, articles die and eine, I do not think the explanation has any basis i n the form being that o f feminine gender. Frequency and saliency w o u l d seem to provide a sounder explanation (see M i l l s , 1984 for a detailed discussion o f gender i n German). 4.3.2. Nominative

and Accusative

Case.

Case is marked on articles, adjec­

tives, and pronouns as an integral part o f the f o r m . The form o f case marking is dependent on the gender o f the noun i n the noun phrase (see Tables 2.1 and 2.2). The marking o f nominative, accusative, dative, and genitive w i l l be discussed i n the subsequent sections, followed by case marking after prepositions since this is a more complex issue. The case f o r m o f the article w h i c h first appears is the nominative case f o r m . Nominative case marking is overgeneralized to accusative case (this is only observable i n masculine gender nouns) w i t h considerable frequency. There is evidence, however, that the t w o cases are sometimes distinguished before arti­ cles are regularly used, through the correct use o f adjective endings. Scupin and Scupin report, at age 2;2, the t w o sentences: (95) Scupin (2;2)

groß

-en

Ball

big

MASQACC

ball:MASC

sehen

see

'(I want) to see the big ball' (96) Scupin (2;2)

groß

-er

Ball

big

MASCNOM

ball:MASC

kommt

comes

'the big ball is coming' Stern and Stern also report early use o f accusative case. The frequent error o f substituting nominative for accusative case persists, however, until quite late. Sterns' data still contain such errors at age 3;2. (97) Stern (3;2) hab

*der

Stuhl

(g)e-

hau

-t

PAST.AUX

DEF.ART: MASC:SG:NOM

chainMASC

PAST

hit

PAST

'I have hit the chair'

2.

The Acquisition of German

179

Scupins' data contain such errors as late as 4;7. Even w i t h i n the same sentence the child can produce the correct and incorrect f o r m . (98) Scupin (4;4) da

muß

ich

andermal

there

must

I

anothentime

*der

DEF.ART:MASC:SG:NOM

Operngucker

mit-

opera:glasses:MASC:SG

with [PARTICLE]

nehmen

take

T must take the opera glasses with me another time' (99) Scupin (4;7) jetzt

hab

ich

*der

auch

tot

now

PAST.AUX

I

DEF.ART: MASC:SG:NOM

also

dead

ge-

schössen,

weil

er

PAST

shot

because

he

den

ander-

DEF.ART:

other

MASC:SG:ACC Soldaten

tot

ge-

schössen

hat

soldienMASC

dead

PAST

shot

PAST.AUX

en

MASCSG: ACC

'I shot that one dead now, because he shot the other soldier dead' As the above examples show, the overgeneralization o f nominative case is not limited to nouns i n initial position, so that w o r d order does not appear to be the sole factor. I n the last example, the fact that the article der is used as a demon­ strative pronoun and i n a position o f emphasis may w e l l be factors contributing to the error. There is only one report o f accusative articles being overgeneralized to nominative and that is i n Hildegard Leopold around 5;0. The fact that only articles w i t h masculine gender nouns are marked distinctively for nominative and accusative case is probably the main factor i n this error. The child may w e l l take longer to perceive the rules governing use o f accusative case because o f its less frequent m a r k i n g , and secondly he may w e l l regularize the system so that it is symmetrical across all three gender paradigms. I n the indefinite article (and the negative kein and possessive adjectives mein, dein, etc., which follow an identical declension pattern) the same overgeneralization o f the nominative form occurs. G r i m m (1973, p . 99) also notes this. I t is reported as late as 5;6 by Scupins and 5;1 by Sterns. Where the article appears w i t h an adjective f o l l o w i n g , however, w h i c h is comparatively rare i n the data, the adjective is marked for accusative case. I t is rare that such a marking occurs i n nominative case forms. This suggests that this error is not due to a lack of distinction between nominative and accusative cases but that a regularization of the article system produces the error. The first example o f the correct ac­ cusative form o f the indefinite article noted by Scupin and Scupin is as late as 4;2.

180 (100)

Mills Scupin (2;4)

Bubi

hat

*kein

Bubi

has

NEG.ART: MASC:SG:NOM

Hunger

hungenMASC

'Bubi isn't hungry' (101) Scupin (2;6) hab'

ich

so

*ein

groß

-en

have

I

such

INDEF.ART: MASC:SG:NOM

big

MASC:SG:ACC

bös

-en

Finger

bad

MASC:SGACC

fingenMASC

'I have such a big bad finger' (102) Scupin (4;2) ich

hab'

einen

Spazierstock

und

du

I

have

INDEF.ART: MASC:SG:ACC

walking: stick: MASC

and

you

hast

have

keinen

NEG.ART:MASC:SG:ACC

T have a walking stick and you have none' (103) Scupin (5;3) hat

die

*ein

so

voll

-en

has

she

INDEF.ART: MASC:SG:NOM

such

full

MASC: SG:ACC

Bauch

stomach:MASC

'she has such a full stomach' (104) Scupin (5;1) für

den

Vater

ein

for

DEF.ART: MASC:SG:ACC

father

INDEF.ART: MASC:SG:NOM

*schönen

-en

beautiful

MASC:SG:ACC

Gruß

greeting:MASC

'a big hello to daddy' MacWhinney's claim that it is the citation form w h i c h is overgeneralized might apply i n the instances o f articles occurring alone w i t h nouns but i n the instances where the adjective is correctly declined, this explanation cannot apply. Unfortu­ nately there are no recorded instances o f the definite article, inappropriately i n nominative case, w i t h a f o l l o w i n g adjective. Scupins' data also contain a few examples o f the adjective i n nominative case after the indefinite article having an

2.

The Acquisition of German

181

incorrect -n ending. The child might w e l l be regularizing the masculine gender paradigm at this stage to ein A D J + -n for both nominative and accusative case. I n adult fast speech the accusative -en ending on the indefinite article is often contracted so that it is very difficult to perceive the ending except as a lengthened consonant n. This w o u l d support the child's regularization o f the paradigm. A n early distinction is made between nominative and accusative forms o f the first pronouns to appear: ich T and du ' y o u ' . Case forms are marked quite distinctly: ich

(NOM) mich

(ACC)

mir

(DAT)

du

(NOM) dich

( A C C ) dir

(DAT)

The first uses o f the accusative are very limited i n context, that is, they are used only w i t h certain verbs, frequently as a reflexive. These w o u l d appear to have been learned by rote. (105)

Scupin (2;1)

schäm

mich

nicht

embarrass

me: A C C N E G

'I am not embarrassed' (106)

Scupin (2;2)

stoß

dich

bang:IMP

you:ACC N E G

nicht

'don't bang yourself This is also true o f the first instances o f the dative form mir. Most authors report the early use o f gib mir 'give to m e ' . The nominative forms, especially ich, are not so limited to the context o f particular verbs, and the first uses are correct for case. (107)

Lindner (2;1)

ich

fort-jagen

I:NOM

away-chase

Fliege

fly

T will chase the fly away' (108)

Scupin (2;3)

ich

gehe

in

die

Schule

I:NOM

go

in

the

school

'I go to school' The child is producing utterances w i t h those forms at the same time as utterances using the proper names and third person for himself and addressee; Scupin's child, for example, was still regularly using his o w n name (Bubi) to refer to himself at the age o f 2;8. Since the pronoun forms are not i n wide use or productive, i t is difficult to k n o w how much analysis to attribute to the child. The

182

Mills

case functions are, however, to some extent differentiated, and this might indi­ cate that the notion o f case is available early on. Since the pronouns are mono­ syllabic and carry more stress, it may be easier to learn the forms appropriate for use i n the appropriate function. As mentioned above, the articles are often contracted, especially i n fast speech, so that the information distinguishing the case-marked forms is lost. The paradigms for pronouns do not overlap i n forms marking case, so that it is easier to associate form w i t h function. The articles are particularly problematic i n this respect. The importance o f nominative/accusative case marking on the article, as opposed to w o r d order, for determining agent and object i n active declarative sentences was tested i n a comprehension experiment w i t h children aged 6 to 9 years ( M i l l s , 1977a). A c t i v e sentences were presented i n a picture selection task. These sentences had a high degree o f reversibility (Slobin, 1966). The sentences varied according to the relative order o f the agent and object and according to the position o f the article (masculine gender only) w h i c h was marked for case. The other article was unmarked (feminine or neuter): (109)

der

Hund

sieht

die

DEF.ART: MASC:SG:NOM

dog

sees

DEF.ART: FEM:SG:NOM/ACC

Katze

cat

'the dog sees the cat' (110)

das

Mädchen

schlägt

den

DEF.ART:

girl

hits

DEF.ART:

NEUT:SG:NOM/ACC

Mann

man

MASC:SG:ACC

'the girl hits the man' (111)

den

Jungen

ruft

das

Mädchen

DEF.ART: MASC:SG:ACC

boy

calls

DEF.ART: NEUT:SG:NOM/ACC

girl

'the girl calls the boy' (112)

das

Mädchen

ruft

der

DEF.ART: NEUT:SG:NOM/ACC

girl

calls

DEF.ART: MASC:SG:NOM

Mann

man

'the man calls the girl' A l l the children found the sentence w i t h agent i n first position, that is w i t h S V O order, easiest to interpret correctly. I n the sentences w i t h O V S order, those sentences w i t h the marked article i n initial position (like sentence 111) were easier to interpret than those sentences w i t h the marked article i n final position. Performance on the O V S sentences improved w i t h age, so that it can be seen that a reliance on a word-order strategy decreases as the child is able to use case

2.

The Acquisition of German

183

information to determine relations. The fact that the marked article i n initial position facilitates comprehension can be explained by the saliency o f initial position. When the unmarked form is i n initial position, the word-order strategy creates the expectation o f a f o l l o w i n g accusative, so that the case information is not so readily perceived. These results strongly suggest the importance o f the canonical w o r d order ( S V O ) i n comprehension (see Slobin & Bever, 1982), even i f it is not clear that it is more frequent i n spontaneous productions (see earlier discussion). F r o m the few examples Ramge observed i n school beginners, self corrections i n case reveal an association o f initial position w i t h nominative case, that is the child w i l l begin w i t h a noun i n nominative case but then change the case as he realizes it is inappropriate for the construction. For example, when referring to der Blumentopf 'the flower pot': (113)

(ca. 6;0)

und

*der

and

D E M . PRO: MASC:SG:NOM

de

gar

nicht

EMPH

N E G pour

den

kann

DEM. PRO: MASC:SG:ACC

can

gießen

'and that one you can't water at all' 4.3.3. Dative Case. The dative case is commonly used to mark the bene­ factor. This usage is established relatively early. F r o m one example o f an adjec­ tive being used i n the dative case, without a preceding article, it w o u l d appear as though the form chosen is the one appropriate when an article is present. The weak -n form is used. Scupin (2;4) Papa

gibt

nich

Daddy

gives

N E G poor

arm-

*en

Mama

Mummy

Kussel

kiss + DIM

'Daddy won't give poor Mummy a kiss' The first examples o f dative case i n articles suggest that the accusative case is overgeneralized. Caution i n interpretation is advisable here, however, since i n several dialects, for example that o f B e r l i n , dative and accusative case are frequently interchangeable, (see K u r t Schwitters (1965) poem " A n A n n a B l u m e " : O du, Geliebte

meiner

siebenundzwanzig

Sinne,

ich liebe

(114) Scupin (2;6) ich

will

Briefe

ins

Haus

I

want

letters

into:the

house

*die

DEF.ART: FEM:SG:NOM/ACC

dir).

184

Mills Mama

schicken

Mummy

send

'I want to send letters to mummy in the house' (115) Scupin (2;9) mach

*den

Mann

make:IMP

DEF.ART: MASC:SG:ACC

man

Beine

legs

'make legs for the man' The correct dative form o f the definite article appears quite early. Scupin and Scupin record many utterances around 2;9 and Stern and Stern at 3;2. Scupin and Scupin, interestingly enough, appear not to have noted these utterances as con­ taining the dative, because they first comment on the use o f the dative at age 3;5. A t this stage, their son overgeneralized the masculine/neuter form to the femi­ nine which he had previously correctly distinguished. F r o m this one recorded utterance, it is not clear that this was a common occurrence: (116) Scupin (3;5) ich

bin

*'m

Mama

sehr

gut

I

am

DEF.ART: MAS C/NEUT: S G: D AT

Mummy :FEM

very

good

und

'm

Papa

ganz

gut

and

DEF.ART: MASC:SG:DAT

Daddy:MASC

quite

good

'I am very good to (for) Mummy and quite good to (for) Daddy' Plural nouns i n the dative require a suffixed -n on the noun itself. Utterances containing such dative plurals are rare i n the diary material. W h e n Scupin and Scupin first record such a sentence, at age 4 ; 1 1 , the usage is correct. This is an area which often produces error i n adults, however, not only i n spoken language. The phrase Schokolade mit *Nusse 'chocolate w i t h nuts' has been seen i n print. Errors could be expected to be frequent i n child speech for some time therefore. It is rare for the dative to be used instead o f the accusative. Stern and Stern report one use at age 3;2. (117)

Stern (3;2)

haste

*mir

ge-

kratz

-t

you:have

me: DAT

PAST

scratch

PAST

'you have scratched me' Verbs w h i c h require dative case i n the object, such as helfen first used w i t h accusative case.

' h e l p ' , appear to be

2. (118)

The Acquisition of German

185

Scupin (2;8) ich

werde

*dich

helfen

auf-

zu-

I

FUT.AUX

you:ACC

help

up

to

stehen

stand

'I will help you stand up' The dative case, when generally learned, appears here too, however. (119)

Scupin (2; 11)

he If

mal

'm

Bubi

help:IMP

just

DEF.ART: MASC:SG:DAT

Bubi:MASC

'help Bubi' There are no recorded instances o f nominative case being used instead o f dative case, i n so far as that is determinable from the data. This might be attributable to the fact that dative case is distinctively marked from nominative and accusative in all gender paradigms. The context for this form may be more easily learned. Certainly the dative form is acquired only a few months after the use o f articles becomes more frequent. Unfortunately there is no body o f data on the group o f verbs which have the direct object i n the dative case. 4.3.4. Genitive Case. Genitive case is very rare amongst the diary data and is limited to nouns preceding a noun phrase and is then marked by a suffixed -s. M a r k i n g o f genitive i n articles is not recorded at a l l , which implies for Scupins' child that this emerged after age 6;0, when the diary finishes. The first use o f genitive case involves proper names suffixed by -s, which corresponds to adult usage. This use is acquired quite early, around age 2;6. (120)

Scupin (2;8)

in Papa-s

Kragen

(121)

Scupin (3;0)

in Mama-s

Kopf

(122) Stern (2;4)

Hildes

Suppe

'in daddy's collar' 'in mummy's head' 'Hilde's soup'

Scupins' son later extends this form erroneously to nouns which are not proper names. (123)

Scupin (3;1)

das

ist

Männer

*-s

that

is

men

G E N car

Wagen

'that is the men's car' (124) Scupin (4;3)

hier here

ist is

Männer railway

*-s

Wohnung

G E N house

'this is the railway's house'

186 (125)

Mills Scupin (4;4)

da

tut

Männern

*-$

there

make

men

G E N stomach

Bauch

weh

hurt

'the men's stomachs hurt' The correct form here is to mark the case on an article and to position the noun phrase i n genitive case after the first noun phrase. For example: (126)

der

Wagen

der

DEF.ART: MASC:SG:NOM

car:MASC

DEF.ART: PL: GEN

Männer

men:PL

Some noun compounds are also formed by adding an -s to the first noun, for example: Räumung -s -verkauf 'clearance sale' (see later discussion 4 . 3 . 6 ) . This would be another formal pattern, N O U N + s + N O U N , w h i c h could lead to the child's overgeneralization, although there is no possessive relationship involved. Over the same period o f time, Scupins' son also uses the prepositions von 'from, o f , w h i c h is correct, and zu ' t o ' , to express the possessive relationship. (127)

Scupin (3;3)

wo

sind

von

den

aliens

where

are

of

the

all

die

Mamales

the

Mummies

Kindeln

children

'where are all the children's Mummies?' (128)

Scupin (4;7)

das

ist

der

Schornstein

von

that

is

the

chimney

of

brauner

Baukasten

brown

building:box

mein

my

'that's the chimney from my brown building box' (129)

Scupin (3;7)

wo

ist

denn

aber

die

where

is

then

but

the

*zw

den

to

the

Großmama

grandma

Affe?

monkey

'but where's the monkey's grandma?' The first system is clearly an overgeneralization o f a standard pattern. The second system o f marking, through use o f a preposition, makes the possessive relationship explicit, w h i c h w o u l d support the principle that semantic rela­ tionships should be clearly and explicitly marked.

2.

The Acquisition of German

187

It is permissible i n standard German to use the dative case to express the possessive form, especially where the relationship is close to the benefactive. The noun phrase i n the dative case is usually ordered before the noun phrase i n accusative case. (130)

ich

putzte

dem

I

cleaned

DEF.ART: MASC:SG:DAT

den

Mann

mamMASC

Tisch

DEF.ART: MASC:SG:ACC

table:MASC

'I cleaned the man's table'/T cleaned the table for the man' As i n the above example, this use o f dative case can express either the possessive or the benef active w i t h no possessive relation. I n some dialects, this form is very common, especially w i t h a proper name followed by a redundant marking o f the possessive pronoun. I f no article precedes the proper name, there is no marking of dative case; i t is only indicated by position. (131)

das

ist

Heinz

seine

that

is

Heinz: DAT

his

Tasse

cup

'that is Heinz' cup' Sterns' data contain t w o examples o f this form. The semantic relationship is clearly indicated by the ordering o f the nouns and the possessive pronoun. (132) Stern (2;6) is'n

de

Hilde

is

DEF.ART

ihr

Hilde

Stiihlchen

her

chainDIM

'it's Hilde's chair' (133) Stern (3;6)

mit

with

klein

little

Hilde

Hilde

seine

his

Schere

scissors

'with little Hilde's scissors' Again the redundancy o f the marking, through the noun i n dative case and the possessive pronoun, means that the semantic relationship is explicitly and clearly marked. I n all these examples from children the possessor is ordered before the thing possessed, unless the relationship is made explicit through the use o f a preposi­ tion. W o r d order w o u l d seem to be important here for marking this relationship.

188

Mills

In a repetition experiment, w i t h children aged 2 ; 6 - 6 , G r i m m (1973, p . 182) found that this w o r d order was always preserved, although up to 4 years o l d the morphology was often greatly changed. 4.3.5. Cases After Prepositions. This is a complex area since the child must know which case is appropriate after w h i c h preposition A N D must be able to mark the case i n the correct way. Firstly, it is common for articles to be omitted after prepositions, even at an age when they are i n frequent use. This observation was confirmed i n m y o w n data and i n the data o f Stern and Stern (1928), Scupin and Scupin (1910), and G r i m m (1973). This was mainly observed w i t h the prepositions an and in, but since these are commonly used i n the early stages, errors o f omitting the article were frequent. (134)

Scupin (3;1)

da regnet

's an

*Schlafzimmerfenster

'there it's raining at (the) bedroom window' I think that this is not simply due to an uncertainty o f case or simplification o f a complex construction. A n additional factor is the phonetic similarity o f the prepositions an, in, and the amalgam, PREP + A R T : D A T ( M A S C / N E U T ) , name­ ly am and im. There are t w o arguments i n support o f this explanation. Firstly, the omission o f the article occurs w i t h the prepositions an and in only when followed by a masculine or neuter noun. Secondly, i n one utterance from Scupin's son, the dative was marked on the noun by suffixing -e w h i c h is an optional f o r m . (135)

Scupin (3;6)

sind die Lockerle

in Halse

drin?

'are the curls inside the neck?' The marking o f the dative on the noun w o u l d seem to indicate that the child was not omitting case marking i n this context. Paprotte (1977) also suggests that the phonetic similarity o f n and m as endings played a role i n the comprehension o f case endings, as w i l l be discussed below. W i t h prepositions which require accusative case only, there appear to be few errors, although i t must be remembered that there are few prepositions i n this category. Preyer (1882) reports one example o f dative after filr, as do Stern and Stern: (136) Preyer (2;4) fur

*'m

Axel

fur

*dir

'for you'

(137) Stern (3;2)

'for Axel'

It cannot be said that this is a general trend, however, from the few examples available. W i t h prepositions w h i c h require dative case only, many errors occur, until

2.

The Acquisition of German

189

age 5 at least. Scupins' data provide the fullest picture o f development and the following analysis is based on their son's acquisition pattern. Caution i n i n ­ terpretation is advised again here, however, because o f the possible influence o f dialect, which i n certain circumstances may allow a substitution o f accusative for dative case and vice versa. Errors occur w i t h Scupins' child i n all gender para­ digms at first until age 2; 11, w h i c h is the time the dative case as benefactive becomes established. Correct forms are used about 50% o f the time. It might be expected that the nominative case form as the " c i t a t i o n all o m o r p h " w o u l d first be used after prepositions. Scupins' data contain only one such example. This occurs at age 4;7 (see sentence 128 above), however, w h i c h makes it seem likely that the error had occurred earlier but had not occurred i n the utterances recorded. G r i m m ' s data (1973, p . 100) contain one such example, which she cites as a typical error o f lack o f case agreement, but it is not clear how often this occurred w i t h masculine nouns. (138) Grimm (ca. 3;0) der

geht

auf

*der

he

goes

on

DEF.ART: MASC:SG:NOM

Stuhl

chair:MASC

'he climbs onto the chair' More data are needed before i t can be claimed that the citation form is used i n this context. For the purposes o f simplifying explanation, it w o u l d be nice i f i t was. The errors indicate an overgeneralization o f accusative case, since w i t h mas­ culine nouns the distinct accusative form is used. (139) Scupin (2;5) will

bei

*die

Lottel

want

at

DEF.ART: FEM:SG:NOM/ACC

Lottel

spielen

play

'(I) want to play at Lottel's house' (140) Scupin (2;8) die

Milchtasse

von

the

milkcup

of

Piepvögele

DEF.ART: NEUT:SG:NOM/ACC

'the bird's milkcup' (141) Scupin (2;8) ich

mach

neben

*'n

Kopf

I

make

near

DEF.ART: MASCSG.ACC

head

bird: DIM :NEUT

190

Mills eine

feine

Frisur

a

fine

hairstyle

'I am making a fine hairstyle on the head'

After age 2 ; 1 1 , only errors w i t h masculine and neuter nouns occur. This suggests that the phonetic similarity o f accusative n and dative m i n the masculine para­ digm may cause some confusion. The neuter nouns should provide evidence o f whether the continuing errors are due to the overgeneralization o f accusative or due to nlm confusion. I n fact, o f the five errors i n v o l v i n g neuter nouns i n Scupins' data, both explanations are supported. T w o errors involved overgeneralization o f the accusative, as sentences 140 and 142 indicate. Interestingly enough, this second sentence reflects the uncertainty about the correct f o r m , since the correct and the incorrect forms are side by side i n parallel constructions.

(142) Scupin (4;11) da

denke

ich

mir

immer

ein

viel

there

think

I

me: DAT

always

a

much

schöneres

nicer

Bild

mit

einem

gut

-en

picture

with

INDEF.ART: NEUT:SG:DAT

good

NEUT:SG:DAT

Schwein

aus,

bloß

wenn

du

mir

so

ein

out

just

when

you

me:DAT

such

a

pig:NEUT

häßliches

ugly

Bild

mit

*ein

bös

-es

picture

with

INDEF.ART: NEUT:SG:NOM/ACC

bad

NEUT:NOM/ACC

Schwein,

da

kann

ich

das

nicht

pig:NEUT

then

can

I

that N E G

'I always invent a much nicer picture with a good pig, just when you show me such an ugly picture with a bad pig, then I can't do it!'

This example, together w i t h the f o l l o w i n g example from m y o w n data, suggest that the prepositional phrase may be learned as a unit as a k i n d o f formulaic speech. The errors occur when a new noun has to be used i n the context.

(143)

MOTHER:

mit wem hast du gespielt

heute?

'who did you play with today?' CHILD (3;7)

mit

dem

Rainer,

mit

with

DEF.ART: MASC:SG:DAT

RainenMASC

with

2.

The Acquisition of German

dem

Frank

and

mit

DEF.ART: MASC:SG:DAT

FrankiMASC

and

with

der

191

Julia

DEF.ART: FEM:SG:DAT

Julia:FEM

'with Rainer, with Frank, and with Julia' MOTHER:

und

Barbara?

'and Barbara?' CHILD:

nee,

mit

*die

no

with

DEF.ART: FEM:SG:NOM/ACC

Barbara

hab

ich

nicht

have

I

NEG

Barbara:FEM

gespielt

played

'no, I didn't play with Barbara' The child has the correct case i n the article used w i t h names o f children he regularly plays w i t h . He has probably frequently reported activities w i t h these friends. W i t h the name o f a different child he produces the article i n citation form, possibly because he has never used the name w i t h the preposition before. This is a very tentative explanation o f such errors, however, and far more data are needed. The remaining three errors w i t h neuter nouns, o f the five errors referred to earlier, suggest that the nlm confusion is also important. (144) Scupin (2;8)

mit with

*den

Schießgewehr

DEF.ART: MASC:SG:ACC

gun:NEUT

'with the gun' (145) Scupin (4;9)

von from

*den

Häusel

DEF.ART: MASC:SG:ACC

house:DIM:NEUT

'from the house' (146)

Scupin (5;1)

war

das

aber

nicht

was

that

but

N E G funny

*den

DEF.ART: MASC:SG:ACC

ulkig

von

of

Vögerle

bird:DIM:NEUT

'wasn't that funny of the bird?'

192

Mills

A l l nouns were being clearly used i n the singular. A gender mistake is possible but, as we discussed above, gender mistakes are rare and i n sentence 146, the child correctly produced das Vögerle i n the adjoining linguistic context. I n support o f the nlm confusion is the fact that zum (zu + dem) is also frequently produced as *zun (zu den), w h i c h is an impossible f o r m . A n example o f the indefinite article i n dative case supports this nlm confusion, since the form einen was used w i t h a neuter noun. (147) Scupin (3;4) du

hast

sie

*einen

klein

-en

you

have

them

INDEF.ART: MASC:SG:ACC

little

M ASC: ACC/NEUT: D AT

Kindel

gegeben *

child:DIM:NEUT

gave

'you gave them to a little child' Neugebauer's child revealed his awareness o f the possible confusion o f m and n at age 3;0: (148) Neugebauer (3;0) unten sind ganz kleine

Schäumerle­

ich meine keine Scheune,

ich meine

Schaum

'down there are really little bubbles I don't mean "Scheute" ('barns'), I mean "Schaum" ('foam')' T w o factors seem to be involved, therefore: the overgeneralization o f the ac­ cusative case and the confusion between n and m forms. The case-marking f o l l o w i n g prepositions which can take accusative or dative case according to directional or stative meaning follows the same pattern as described for accusative and dative only prepositions. That is, the directional use (accusative) was always correct except where the article was omitted. Preyer reports his son describing the movements o f a f l y all w i t h the correct accusative form at age 2;9: (149) Preyer (2;9) in die Zeitung (150) Preyer (2;9) in die Milch

'into the paper' 'into the milk'

(151) Preyer (2;9) unter den Kaffee

'under the coffee'

Far more errors occur when dative case must be used w i t h stative meaning. A s described above, the early use seems to indicate overgeneralization o f accusative and therefore no indication o f marking a distinction i n the t w o meanings. The later errors, however, can all be attributed to the m/n confusion. I t is not clear

2.

T h e A c q u i s i t i o n of G e r m a n

193

that the distinction i n meaning is mastered, although Scupins' son produced one contrast w i t h the same noun i n the same context.

(152) Scupin (3;2) wenn

wir

auf

'n

Molkefelsen

gehen

when

we

up:DIRECT

DEF.ART: MASC:SG:ACC

Molkefelsen

go

. . .

'when we go up the Molkefelsen . . .' (153) Scupin (3;2) auf

'm

Molkefelsen

gibt

's

on: STAT

DEF.ART: MASC:SG:DAT

Molkefelsen

gives

it

keine

no

Semmel

roll(s)

'there are no rolls on the Molkefelsen'

Paprotte (1977) experimentally tested the comprehension o f this difference i n meaning. He tested the comprehension o f prepositions in ' i n ' , auf ' o n ' , über 'over', and unter 'under', w h i c h can all involve both directional and stative meaning. He used verbs w h i c h were either stative or dynamic. W i t h the dynamic verbs, he used the prepositions followed by both cases to give a stative and directional reading. Nouns were used i n the singular and plural, but only mas­ culine nouns w i t h the zero plural morpheme were tested so that ambiguous sentences resulted. Sentences 154-159 give examples o f the type o f sentences used.

(154) der Clown sitzt in dem Kuchen

(STAT. V E R B + DAT.SG)

'the clown is sitting in the cake' (155)

der Clown

(156)

der Clown

hüpft in den Kuchen

( D Y N . V E R B + ACC:SG)

'the clown hops into the cake' hüpft in dem Kuchen

( D Y N . V E R B + DAT:SG)

'the clown is hopping in(side) the cake' (157) der Clown sitzt in den Kuchen

(STAT. V E R B + DAT:PL)

'the clown is sitting in the cakes' (158)

der Clown

(159)

der Clown

hüpft in die Kuchen

( D Y N . V E R B + ACC:PL)

'the clown hops into the cakes' hüpft in den Kuchen

( D Y N . V E R B + DAT:PL)

'the clown is hopping in(side) the cakes'

194

Mills

Sentences 155 and 159 are ambiguous. The subjects aged 6 to 12 years had to match the spoken sentences to picture material. A hierarchy o f complexity emerged, the sentence type i n first position being easiest: 1. stative verbs + D A T 2. dynamic verbs + ACC:PL 3. dynamic verbs + DAT:SG On the ambiguous sentences, the preferred interpretation was that o f dynamic verb + accusative singular. I t is concluded that the stative verbs + dative are easiest because the information from the verb makes the information from case marking redundant. This implies that children are already using the category " s t a t i v e . " The marking on the article was important i n these cases for dis­ tinguishing number, and the majority o f errors were to interpret the plural noun as singular. This w o u l d f i t i n w i t h Operating Principle E (Slobin, 1973) accord­ ing to w h i c h underlying semantic relations should be marked overtly and clearly. Since the realization o f the plural i n this case is the zero morpheme, the child easily assumes that this is the singular f o r m . The forms o f the definite article are also phonologically similar dem and den, as discussed above, so that confusion may be easier here. W i t h the dynamic verb the correct comprehension o f A C C : P L was easiest. The form o f the article die is distinctly marked as plural, i f the gender o f the noun is k n o w n to be masculine or neuter. I t is interesting that the largest number o f errors w i t h this sentence type was to interpret i t as A C C : S G . Since die can be accusative for feminine gender singular as w e l l as for plural, i t w o u l d appear that the gender is not so w e l l established. The lack o f clear plural marking (see above) may make this confusion easier. The dative singular and accusative singular and dative plural have similar forms o f the article, so that phonological confusion can explain the frequency o f mistakes. The direction o f interpretation was significant, however, i n that the largest number o f errors w i t h the dative singular and the most frequent perception of the ambiguous sentences was as accusative singular. A g a i n this may be explained by Principle E mentioned above. A further experiment using nouns o f different genders and nouns w i t h marked plural forms w o u l d possibly disen­ tangle the strategies involved here. I t is not possible to conclude that the direc­ tional interpretation is the first learned and overgeneralized w i t h dynamic verbs until the other factors have been accounted for. N o data are available from natural dialogue to investigate whether this complexity leads to misunderstand­ ing or whether pragmatic and semantic information are regularly used to deter­ mine meaning i n these instances. I w o u l d guess that the latter is the case. 4.4. Adjective

Morphology

4 . 4 . 1 . Case and Gender. Adjective declension involves even more com­ plex rules than articles. The endings are determined by what precedes the adjec-

2.

The Acquisition of German

195

tive w i t h i n the noun phrase. I f the adjective is not preceded by any article or demonstrative adjective, the endings are " s t r o n g . " I f the adjective is preceded by a definite article, demonstrative adjective, or alle ' a l l ' , they have " w e a k " endings. When preceded by the indefinite article, the negative article kein and the possessive adjective mein, etc., the adjective has a mixture o f weak and strong endings (see Table 2.2). I n predicate position, endings are rarely used, which supports the first root allomorph principle o f M a c W h i n n e y . I n attributive position, the errors reported are quite varied. I t is rare that an ending is omitted altogether but a few such errors occur i n the early stages. I n M i l l e r ' s (1976) data the adjective ending is commonly left o f f ander 'other', possibly because the root ends i n -er and appears to have an ending. W i t h other adjectives there are only t w o examples o f an uninflected adjective. Firstly, as described earlier, adjectives are used i n attributive position before articles appear i n the noun phrase. I n these circumstances, that is, preceded by no article, the strong endings are appropriate. Preyer has examples o f correct strong endings i n nominative case at age 2;7 i n his son, before articles appeared i n frequent use at 2;9. (160)

Preyer (2;7)

dumm

-es

Ding

stupid

NEUT:SG:NOM/ACC

thing:NEUT

'stupid thing' (161) Preyer (2;9)

neu

-er

new

MASC:SG:NOM daddy:MASC

Papa

'new daddy' (162)

Preyer (2;9)

du

lieb

-e

you

dear

FEM:SG:NOM/ACC

Zeit

time:FEM

'goodness!' M i l l e r has such examples at age 1; 10, Scupins at age 2;2, Sterns at age 2;6 for Hilde and 2;4 for Giinther. A common error recorded, however, is that the adjective ending -e is overgeneralized i n pre-nominal position at this stage, when articles are missing ( M i l l e r , 1976; Park, 1974; Scupin & Scupin, 1910; Stern & Stern, 1928). M a c W h i n n e y seeks to explain this overgeneralization by the fre­ quency principle, since the -e ending accounts for 5/8 o f the endings i n the weak paradigm and 2/8 i n the strong paradigm. A n alternative explanation, w h i c h would need careful investigation to substantiate, w o u l d be that the ending is selected according to a presupposed definite or indefinite article but that the article is omitted. T o explore this, detailed information w o u l d be needed about context to attempt to ascertain w h i c h article w o u l d be appropriate. I t w o u l d explain the presence o f strong endings marking gender and case correctly on the one hand, since those are appropriate after indefinite articles, and, on the other

196

Mills

hand, the overgeneralization o f -e neglecting case and gender w h i c h is appropri­ ate after the definite article. The more likely explanation w o u l d , however, seem to be that o f the frequency principle. When articles are used, mistakes i n adjective morphology seem to be fairly rare. Stern and Stern claim the weak and strong forms were distinguished i n Hilde by age 3;2. I n Scupins' son, the t w o paradigms are used and mainly correctly i n accusative and nominative case, although t w o errors occur w i t h the ending -n after the nominative indefinite article: (163) Scupin (2; 10)

ein

groß

*-en

INDEF.ART: MASC:SG:NOM

big

MASC:SG:ACC

Stein

stone:MASC 'a big stone' (164) Scupin (3;9)

das

ist

kein

that

is

NEG.ART:MASC:SG:NOM

richtig

*-en

correct

MASC:SG:ACC

Löffel

spoon:MASC

'that is not a proper spoon' This may be part o f the attempt to regularize the system making nominative and accusative i n the masculine paradigm identical. S c h ä d e l (1905) and G r i m m (1973) report overgeneralization o f the weak form after indefinite articles. Lindner (1898) on the other hand reports the overgeneralized use o f strong endings i n early usage: (165) Lindner (2;2) das

groß

-*es

Glas

DEF.ART: NEUT:SG:NOM/ACC

big

NEUT:SG:NOM/ACC [ 'strong']

glass:NEUT

'the big glass' Lindner's son also added strong endings to the indefinite article itself and the similar forms (kein, mein, etc.). I have similar examples i n m y data. (166) Lindner (2;3) mein

-*er

gut

-er

my [POSS.ADJ]

MASC:SG:NOM

good

MASC:SG:NOM

'my good Daddy'

Papa

Daddy:MASC

2.

The Acquisition of German

197

(167) Lindner (2;3) das

ist

ein

-*er

that

is

INDEF.ART: MASC:SG:NOM

MASC:SG:NOM

Brief

lettenMASC

'that is a letter' (168) Mills (5;0) eur

-*es

yourrPL [POSS.ADJ]

NEUT:SG:NOM

Baby

baby:NEUT

'your baby' These forms actually occur i n pronominal use i n adult speech, i.e. when not followed by a noun. The next sentence illustrates this.

(169)

Gib

mir

den

give:IMP

me: DAT

DEF.ART: MASC:SG:ACC

Stift.

pencil.MASC

Er

ist

dock

mein

-er.

PRO:MASC: SG:NOM

is

EMPH

mine [POSS.PRO]

MASC:SG:NOM

'Give me the pencil. It is mine.' This appears to be an alternative attempt to regularize the system, that is that all elements preceding the noun should be marked for gender and case i n the same way. The dative case form after indefinite and definite articles is i n every paradigm (masculine, feminine, neuter, and plural) -n. This appears to be quickly learned. I could find only one example o f an adjective i n dative case not preceded by an article where strong endings w o u l d be appropriate (see p . 183) from Scupins' child at age 2;4. This is an early use but I w o u l d guess that -n w o u l d commonly be generalized i n this position. Adjective endings are overgeneralized to other forms occurring i n pre-nomi¬ nal position, for example numbers and the quantifier all w h i c h should have no endings.

(170) Scupin (4;6) die

all

DEF.ART:PL

all

-*en

instead of all die Männer

Männer

men 'all the men'

198

Mills

(171) Lindner (2;10) gibt

mir

die

zwei

give:IMP

me: DAT

DEF.ART:PL

two

-*en

Äpfel

apples

'give me the two apples' Both numbers and all appear i n certain contexts to behave like adjectives, so that this regularization is quite likely as w e l l as indicating the strong tendency to establish patterns i n the system. On the whole the marking o f the adjective as related to case and gender is fairly w e l l established by age 3;0 to 3;6. 4.4.2. Comparative. A different morphological problem related to adjec­ tives (and adverbs) is the formation o f the comparative. Comparative adjectives are constructed, as i n English, by taking the root and adding -er, often w i t h an umlaut form o f the root. There are exceptions however, as i n English. Mehr 'more' is the first comparative reported (Preyer, 1882; Lindner, 1898; Stern & Stern 1928), but i t is not entirely clear that it is used i n the cases where an -er ending on the adjective w o u l d be appropriate. Other forms used to intensify the adjective are noch and viel. S c h ä d e l also reports the reduplication o f the adjective as a method o f intensifying. (172) Schädel (2;6) eine

klein-e

INDEF.ART: FEM:SG

small

klein-e

small

Puppe

doll

'a very small doll' It is not clear, however, that the concept o f strict comparison is involved i n the use o f any o f these forms. Since these forms appear first, however, this supports Wode's claim (1978) that free forms are acquired before bound forms. W h e n the morphological form o f the comparative is used, i t is also often commented by the observer that no clear comparison is evident but rather a general form o f inten­ sification. Pregel (1970) however found that 50% o f the comparatives used by children aged 6 - 9 years were formed w i t h mehr w h i c h suggests that this form is preferred since it is semantically clear. Overgeneralization o f the regular form o f the comparative occurs frequently, for example *gut-er *hoch-er

from gut 'good' from hoch 'high'

instead of bess-er 'better' instead of höh-er 'higher'

I n the last case there is a difference between the predicative form hoch and the attributive form hoh-. I t is significant that it is the predicative form o f the adjective w h i c h is taken as the root, not the attributive f o r m . This supports

2.

The Acquisition of German

199

MacWhinney's claim that citation forms are taken as first root allomorphs. Frequency o f use also seems to play a part here though. According to Pregel's (1970, p. 137) analysis o f 6 - 9 year olds' language, adjectives can be divided into those used primarily i n attributive position and those used i n predicative position, although the predominant use overall is attributive. Hoch was used primarily i n predicative position, hence possibly the error i n forming the comparative. The comparative ending -er is also extended to adverbs or verb particles which are not derived from adjectives and therefore cannot have this ending. Commonly the intensifying adverb noch ' s t i l l ' precedes such use and a com­ parative sense is clear. (173)

Stern (3;2)

noch

*rein

-er

still

into [VERB.PARTICLE]

COMPAR

'still further in' (174)

Stern (3;3)

noch

*zusammen

-er

still

together [ADV]

COMPAR

'still more together' (175)

Scupin (4;6)

noch

still

*rauf

up [VERB. PARTICLE]

-er

COMPAR

'still more up' This usage, w h i c h is quite clear semantically, indicates that the child has not separated out adverbs into the t w o classes necessary for correct use o f the suffixed f o r m . Scupins' son also reduplicated the suffix -er at around age 4;6 e.g. grofi-er-er. It is not clear under what circumstances this occurred, whether for example additional intensification was intended. The problem may have been one o f determining the adjective root; -er clearly is identified as an intensifier over this period but its relationship to the concept o f comparison is unclear where the comparison remains unexpressed. 4.5.

Prepositions

Children under 3;0 frequently omit prepositions completely (Stern & Stern, 1928; G r i m m , 1973). W h e n prepositions are used, certain ones predominate, leading to incorrect use i n locative and temporal expressions. According to the data reported by G r i m m (1975), the prepositions most frequently used i n locative expressions by children between 2;6 and 6;0 match those which are most fre­ quent i n adult speech (Meier, 1964). Table 2.6 shows the frequency o f the children's use based on 1102 occurrences. Frequency o f usage may be a factor i n

200

Mills TABLE 2.6 Frequency o f Occurrence of Spatial Prepositions in Children A g e d 2 ; 6 - 6 ; 0 (Adapted f r o m G r i m m , 1975, p. 101) Preposition in auf zu bei an nach von aus durch neben vor hinter über unter bis um gegen 1

Gloss

Frequency

'in' 'on' 'to' 'at' 'at' 'to' 'from' 'out of 'through' 'beside' 'in front of 'behind' 'above' 'under' 'to' 'around' 'against'

AAA 19.6 9.4 8.9 4.3 2.9 2.8 1.8 1.1 0.8 0.8 0.7 0.7 0.7 0.4 0.4 0.1

1 have put together the figures for the preposi­ tions in and im which Grimm listed separately, im is not a distinct preposition but is an amalgam of in and the definite article dem which can only be used with stative meaning in association with mas­ culine and neuter singular noun, in can be used in both a directional and stative sense. 1

explaining which prepositions are first acquired. Cognitive simplicity o f the relations to be expressed also seems relevant, as w i l l be discussed i n more detail below (see 8.6). The prepositions most frequently used were overgeneralized to produce errors in usage. According to G r i m m ' s classification o f the prepositions using semantic features, the children generally use a simpler preposition than a more complex one, for example rubergehen *in Strafie 'go over the street' intead o f uber die Strafie gehen. This overgeneralization shows that the restrictions on the use o f the simpler prepositions have not been acquired, even though the spatial relations they encode may be the simpler to learn. Temporal prepositions were much rarer i n G r i m m ' s data than locative prepo­ sitions, and appeared later. The errors that were made consisted mainly o f using the most frequent locative prepositions, for example, gehen *in Samstag instead

2.

T h e A c q u i s i t i o n of G e r m a n

201

of am Samstag gehen 'to go on Saturday'. G r i m m attributes these errors to the spatial/temporal undifferentiation i n the preoperational child as defined by Piaget. The child is attempting to impose the categories used to distinguish locative prepositions on to the temporal dimension. This can also be explained by the fact that the temporal prepositions also are used as locative prepositions; the homonymity makes it more difficult to learn the specifically temporal features involved. G r i m m ' s data also include uses o f verbal particles which have the same form as prepositions, for example Angst haben vor 'to be afraid o f . The use o f these verbal particles frequently has little relationship to the meaning o f the preposition in its locative or temporal usage. The most usual errors w h i c h the children make are to substitute a particle w h i c h fits the semantics o f the verb, for example Kopf zerbrechen

*gegen

literally

' h i t head against' instead o f Kopf

zerbrechen

iiber

'rack one's brains over'. This suggests that children want semantic relationships to be marked clearly, here also concretely, and for each form to have a clear semantic function. 4.6.

Passive

The passive construction is used quite rarely by children. I n German, the passive is distinct from an adjectival construction w i t h the past participle through the use o f a different auxiliary, werden as opposed to sein. For example: (176)

die

Tur

wird

ge-

the

door

AUX

PAST

stricken

paint:PAST

'the door is being painted' (177)

die

Tiir

ist

ge-

the

door

AUX

PAST

stricken

paint:PAST

'the door is/has been painted' In the case o f an agentless passive, the distinction between these t w o construc­ tions depends entirely on the use o f the correct auxiliary. U n t i l the child starts using auxiliaries, therefore, i t is difficult to be certain o f the use o f the passive i n spontaneous productions. G r i m m (1973) reports only one passive i n the group up to 3;0. The passive was used much more frequently by the group aged 4;7 to 5;0, but i n all the groups recorded up to age 6;0 only agentless passives were reported. I n the repetition experiment G r i m m conducted w i t h these children, a frequent error i n repeating an agentless passive was to replace the auxiliary werden w i t h a form o f sein, w h i c h w o u l d suggest that the passive and the adjectival construction are perceived as similar, i f not identical. I n the task o f comprehension i t has been shown that German children behave very similarly to English children i n their understanding o f passive sentences

202

Mills

w i t h agents. That is, firstly their comprehension is affected by the semantic and pragmatic content o f the sentences i n their choice o f w h i c h noun should be agent regardless o f the formal structure. Secondly, a word-order strategy based on the order o f active sentences is evident i n younger children. G r i m m ( G r i m m et a l . , 1975) tested children between the ages o f 3;0 and 7; 12 in a manipulation task. The test material consisted o f reversible and irreversible passive sentences (Slobin, 1966) w i t h a variation i n the degree o f reversibility. For example: (178) reversible: Hans

wird

von

Ursula

ge-

Hans

AUX

by

Ursula

kissed

küß-

t

'Hans is kissed by Ursula' (179) less reversible: das

Baby

wird

von

der

Mutter

the

baby

AUX

by

the

mother

ge-

pfleg-

t

cared for

'the baby is cared for by the mother' (180) irreversible: der

Boden

wird

von

Hans

the floor A U X by Hans 'the floor is kissed by Hans'

ge-

küß-

t

kissed

(181) semantically anomalous: *Hans

Hans

wird

von

dem

Boden

ge-

AUX

by

the

floor

kissed

küß-

t

*'Hans is kissed by the floor' It was found that the youngest group o f children (mean age 3;6) relied very heavily on a word-order strategy w h i c h assigns the functions actor, action, object to an N V N sequence (Bever, 1970). I n a passive sentence, this strategy leads to an interpretation which reverses the roles o f agent and object. This w o r d order strategy was so dominant i n the youngest group o f children that they made the actor-action-object interpretation even i n irreversible sentences. The reliance on this strategy declined i n the older children w h o started to use semantic informa­ tion to determine interpretation. That is, irreversible passive sentences became easier to understand, since the semantic information supported the passive i n ­ terpretation. The completely reversible passive sentences were the most difficult, since the semantic information offered no cue to interpretation. Only the oldest age group (mean age 7;2) was able to interpret the semantically anomalous passive sentences correctly, that is they were using only the formal structure to determine their interpretation. M i l l s (1977a) also obtained similar results i n a

2.

The Acquisition of German

203

comprehension task using only reversible passive sentences. Children up to the age o f 7;0 were still using the actor-action-object interpretation i n some o f these sentences. These results are very similar to those obtained i n English (Bever, 1970; Maratsos, 1974; Turner & Rommetveit, 1967a,b), where it was found that c h i l ­ dren around age 4 relied more heavily than younger children on the actor-actionobject strategy w h i c h meant that their performance on passive sentences was worse than that o f the younger children. Since the youngest German children tested were around age 3;6, i t is impossible to determine whether such a dip i n performance also occurs. G r i m m also tested imitation o f passive sentences containing agents w i t h the same group o f children as i n the comprehension test. She reports that the youn­ gest children (up to age 4;5) made a large number o f random mistakes. Older children made more systematic errors i n their imitations. They frequently changed the positions o f agent and object i n the passive sentences, even though they reproduced the auxiliary, preposition, and past participle. This was done even w i t h irreversible passive sentences. For example: (182)

*der

the

Vater

wird

vom

Teppich

father

AUX

by: the

carpet

aus-ge-klopf-t

beaten

*'the father was beaten by the carpet' was the i m i t a t i o n o f the sentence der Teppich

wird

vom Vater

ausgeklopft

'the

carpet was beaten by the father'. G r i m m explains this through a tendency to place the agent o f the sentence i n first position. The children aged 4;6 to 5;0 performed worse on the imitation test than younger children, but i t is not clear from the results what the explanation for this might be. The same group o f children were asked to transform active sentences into passive sentences and vice-versa. The younger children failed to understand the task. The children older than 5;6, when they understood the task, primarily changed the positions o f agent and object preserving the original active or pas­ sive structure. This w o u l d suggest that the w o r d order difference is the first aspect o f the active-passive relationship to be learned. G r i m m fails to point out, however, that active sentences can have an order other than S V O . The children's rule o f changing the positions o f subject and object does not necessarily lead to a difference i n meaning, especially i f case marking is ambiguous. The use o f this strategy i n this test i n German might be increased by the fact that the semantic content is then preserved. This could not be the case i n English, for example, and so it might be the case that English children w o u l d not use such a strategy. 4.7. Relative

Clauses

The first relative clauses reported i n the diary studies are from Sterns' children at the age o f 2;6. These relative clauses do not contain a relative pronoun, so that their classification as relative clauses could be i n doubt. The verb is i n final

204

Mills

position, however, as is correct for a subordinate clause and semantically the interpretation as a relative clause is acceptable. Stern and Stern also claim that the intonation was appropriate for a relative clause. For example: (183)

Stern (2;6)

Papa

sieh

mal

Daddy

look

just

*

Hilde

mach-t

Hilde

do:PAST A U X

hat

'Daddy, look (what) Hilde has done' (184)

Stern (2;6)

siehe

*

see

auf-

(g)e-

hängen

hat

up

PAST

hang:PAST A U X

'look, (what) (I) have hung up' Their son G ü n t h e r went on to insert a meaningless syllable instead o f the relative pronoun until around age 3 years. For example: (185)

Stern (2; 10)

meine

Blume

eigentlich

ist,

my

flower

really

is

Hans

(g)e-

gib- t

Hans

PAST

give:PAST A U X

*e

hat

'that's really my flower that Hans gave' (186)

Stern (3;2)

das

ist

ein

Pilz

that

is

a

mushroom

*mm

in

Walde

ist

in

wood

is

'that is a mushroom which is in the wood'

Obviously no information is available about cases from these meaningless forms or where the relative pronoun is omitted. I t w o u l d appear as though the semantic and pragmatic information i n the clause is sufficient to make the meaning clear. The child does not seem to be hindered i n any way i n his production o f relativ clauses by not being able to produce the correct case inflected forms o f the relative pronoun. This suggestion is supported by other observations from G r i m m (1973). She reports that the younger children i n her sample ( 3 - 4 years) predominantly used wo as the relative pronoun i n the clauses they produced. I n Standard German, wo is the form o f the locative interrogative pronoun 'where' and is also a possible form o f the relative pronoun used w i t h a preposition. For example:

(187)

das

Thema,

wo

the

topic

REL.PRO

-ruber

about [PREP]

'the topic he is talking about'

er

spricht

he

talk:PRES

2.

The A c q u i s i t i o n of G e r m a n

205

In some dialects, however, wo is also used as the relative pronoun i n subject and object function. I t is this use w h i c h predominates i n the children's relative clauses. For example: (188) Grimm

das

ist

ein

Mädchen,

wo

in

die

that

is

a

girl

REL.PRO

in

the

Schule

geht

school

goes

'that is a girl who goes to school' (189)

Grimm

aber

die

Puppa,

wo

die

Uschi

hat

but

the

doll

REL.PRO

the

Uschi

has

'but the doll which Uschi has' G r i m m took her sample o f children from the Heidelberg area w h i c h uses wo i n this way i n the local dialect. The same usage is reported i n children from G ö t t i n g e n , but I have not been able to establish whether the local dialect has this form. I t is obviously relevant to establish what the influence o f the local dialect is. I t could be the case that all children use this f o r m , independent o f dialect usage, since i t is uninfluenced by case inflections. As i n the early examples from Stern and Stern, no case information is available from this relative pronoun form. Around age 4, G r i m m reports, the children produce an intermediate form combining wo and the standard relative pronoun der, die, das, etc. For example: (190) Grimm ich

hab

ein

Heinzelmann,

der

*wo

I

have

a

Heinzelmann: MAS C

REL.PRO:

REL.PRO

MASC:SG:NOM so

mach

-e(n)

thus

do

INF

kann

can [MOD]

'I have a Heinzelmann who can go like this' A possible explanation for this double use is that the children are unsure o f the formal correctness o f the der, etc., forms, since their case inflections make these more complex. B o t h are used until the child feels confident o f the case inflected forms according to the principle that semantic relations should be overtly and clearly marked. One error w i t h the relative pronoun, observed i n an older child (5 years) , suggests that the child has more problems when the relative pronoun should have a case form different from a preceding definite article. I n the follow5

5

I am grateful to Werner Deutsch for this example.

206

Mills

ing example, the child has repeated the f o r m o f the preceding definite article as the relative pronoun. (191)

(ca. 5;0)

ich

habe

das

mit

der

I

have

that

with

DEF.ART: FEM:SG:DAT

Schere,

*der

ganz

scissors:FEM:SG

REL.PRO FEM:SG:DAT/ MASC:SG:NOM

very

scharf

schneidet,

sharp

cuts

gemacht

made

T have made that with the scissors that cut really sharp' This could be interpreted as a gender mistake, since der is a nominative f o r m i n the masculine paradigm. Possibly the preceding article and the existence o f this form i n the nominative both contribute to produce this error. The child may adopt a strategy o f simply copying a preceding definite article as the relative pronoun, as i n fact was the case i n O l d English, for example. M o r e data are necessary on error i n this area. Experimental w o r k w i t h relative clauses has produced evidence that the posi­ tion o f the relative clause, the function o f the relative pronoun w i t h i n the clause, and the position o f case-marking inflections are all relevant i n the stage o f acquisition. Park (1976) conducted an imitation experiment i n w h i c h the children were tested w i t h three types o f relative clause: final right-branching, initial leftbranching, and initial right-branching. The following sentences illustrate these structures: (192) final right branching— ich

brauche

eine

Sekretärin,

die

gut

I

need

a

secretary :FEM

REL.PRO: FEM:SG:NOM

well

tippt

types

T need a secretary who can type well' (193) initial left branching— wer

das

schreibt,

muß

vollkommen

besoffen

REL.PRO

that

writes

must

completely

drunk

sein

be

'whoever writes such things must be completely drunk' (194) initial right branching— der

Mann

,der

Würste

verkauft,

ist

the

man:MASC

REL.PRO: MASC:SG:NOM

sausages

sells

is

mein

my

2.

The Acquisition of German

207

Vater

father 'the man who is selling sausages is my father' Park only considers the factors o f position o f the relative clause i n respect o f the main clause and the direction o f the branching. His results indicated that children found it easier to imitate relative clauses i n final position rather than i n initial position. I n the initial clauses, left-branching clauses were easier than rightbranching. These results can be attributed to the principle that an interruption i n the elements o f the same clause w i l l cause that structure to be more difficult to interpret. G r i m m and Wintermantel ( G r i m m et a l . , 1975) tested comprehension o f relative clauses i n a toy-moving task w i t h younger children ( 4 ; 0 - 5 ; 4 ) and i n an interview procedure w i t h older children ( 6 ; 0 - 7 ; 4 ) . The clauses tested were all right-branching. The conditions which they varied i n the material were position of the relative clause and the relative pronoun being i n subject or object function. The f o l l o w i n g examples are taken from the test material:

(195) final,REL.PRO.subject— der

Bär

beißt

den

Igel

the

bear

bites

the

hedgehog

,der

den

REL.PRO:

the

Vogel

bird:ACC

MASC:SG:NOM streift

touches 'the bear bites the hedgehog that touches the bird' (196) final,REL.PRO.object die

Kuh

haut

die

Katze

,die

der

the

cow

hits

the

cat: F E M

REL.PRO:

the

Vogel

bird:NOM

FEM:SG:ACC anstößt

bumps 'the cow hits the cat that the bird bumps into' (197) initial,REL.PRO.subject— der

Elefant

the

elephant:MASC REL.PRO: MASC:SG:NOM

tritt

den

kicks

the

,der

den

Hasen

the

rabbit:ACC

Esel

donkey

'the elephant that strokes the rabbit kicks the donkey'

streichelt,

strokes

208

Mills

(198) initial,REL.PRO.object die

Giraffe

,die

der

Hase

packt,

beißt

the

giraffe:FEM

REL.PRO: FEM:SG:ACC

the

rabbit:NOM

grabs

bites

die

Ente

the

duck

'the giraffe that the rabbit grabs bites the duck' To summarize their results: the responses to the interview were difficult to evaluate but i n general agreed w i t h the results o f the manipulation test. Initial subject clauses were easier than final subject clauses, final object clauses were easier than initial object clauses. Sentences where the relative pronoun was subject o f the clause were easier than where i t was object o f the clause. I n these results the position o f embedding, initial or final, was not shown to influence interpretation independently. There are several possible explanations o f these results. The ordering o f initial subject before final subject and final object before initial object is explained by G r i m m and Wintermantel by the extent to w h i c h the sentences deviate from what they regard as the standard order, namely actoraction-object. I t is not clear from the argument presented how this deviation is measured i n order to predict this sequence o f difficulty. These results also match the results obtained by Sheldon (1974) w o r k i n g w i t h English. Her proposed explanation is that o f a "parallel f u n c t i o n " hypothesis. According to this hypothesis, it is easier to interpret sentences i n w h i c h the relativized noun phrase has the same function i n the main clause as it does i n the relative clause, hence the order obtained above. Results w h i c h do not support this hypothesis, however, have been obtained for English (de Villiers et a l . , 1979) and for German ( M i l l s , 1977a,b) (see later discussion). The result that relative clauses w i t h the relative pronoun as subject o f the clause are easier than where i t is object o f the clause can also be explained i n several ways. The explanations are also possibly related to one another. Keenan and Comrie (1977) proposed an Agreement Hierarchy according to w h i c h the possibility o f relativization from a particular function is ranked i n relation to other functions. Subject function is i n first position on the hierarchy, followed by object, indirect object, object o f preposition, possessive, and object o f com­ parison. I n German, relativization is only allowed from the first five positions on the hierarchy. I t is claimed that relativization becomes more difficult the further down the hierarchy you go. I f the Agreement Hierarchy has psychological real­ ity, i t w o u l d predict that relative clauses w i t h the relative pronoun i n subject function should be easier to process. Since nominative case is closely associated w i t h subject function, i t w o u l d predict i n this study that relative pronouns i n nominative case w o u l d be easier than those i n accusative case. Nominative case also seems to be associated w i t h first position i n the clause, for w h i c h some evidence has been presented earlier. Since the relative pronoun is

2.

The A c q u i s i t i o n of G e r m a n

209

the first element i n the clause, this association w o u l d explain the correct i n ­ terpretation o f clauses w i t h the relative pronoun as subject. I f relative pronouns occur w i t h high frequency i n nominative case, this fre­ quency may facilitate interpretation. The use o f the most frequent form may be quite independent o f the factors w h i c h cause that form to be the most frequent. I n this case, the complexity predictions o f the Agreement Hierarchy or the associa­ tion o f nominative case w i t h first position may account for a high frequency o f nominative case i n relative pronouns, but it may be only the frequent occurrence of nominative case i n the relative pronoun w h i c h facilitates this interpretation. I n order to examine the explanation more closely, it had to be ascertained whether nominative case d i d occur frequently i n relative pronouns. For this purpose three different types o f written language from adults (philos­ ophy text, newspaper, and popular novel) were analyzed and compared w i t h one another for frequency o f function, etc. The results o f this analysis are reported i n M i l l s (1981a). T o summarize: Relative pronouns occur far more frequently i n nominative case than i n any other case. A l t h o u g h the Agreement Hierarchy's prediction is met for subject function (associated w i t h nominative case), the other cases are not i n the order predicted, so that the Hierarchy as an explanation o f the order o f complexity i n interpretation is put into doubt. The frequency o f relative pronouns i n nominative case and the general association o f subject w i t h first position are clearly related and both probably contribute towards making the interpretation o f subject relative clauses easier. The most common error i n G r i m m and Wintermantel's results was to take the first noun i n the sentence to be the actor i n both the main and relative clause. This strategy leads to a correct interpretation i n initial subject clauses (see sentence 197) w h i c h might w e l l explain w h y this type o f sentence was the easiest to interpret. This interpretation strategy leads to error i n the other three sentences types and this error was made i n all three types. I n sentence 195, for example, the bear was made to bite the hedgehog and touch the b i r d . This strategy suggests that the children have difficulty recognizing the relative clause structure, possi­ bly because o f the similarity between the relative pronoun and the definite article. The interpretation by the children that the first noun is actor o f both clauses supports the suggestion made earlier that first position is strongly associ­ ated w i t h actor/subject. G r i m m and Wintermantel also tested the children's comprehension o f the same four types o f sentence but w i t h wo inserted after the relative pronoun. This compound form is the deviant form w h i c h the children i n the Heidelberg area themselves produce i n spontaneous utterances (see earlier discussion). This com­ pound f o r m , i n general, improved comprehension. The insertion o f wo clearly marks the previous form as a relative pronoun and eliminates the possibility o f i t being a definite article, w h i c h should make the subsequent processing o f the clause easier. The comprehension o f the same four types o f sentence was tested i n a picture selection task w i t h older children (5; 11—8; 11) ( M i l l s , 1977a). As w e l l as testing

210

Mills

for the influence o f the position o f the relative clause and for the influence o f the case o f the relative pronoun, the test material was constructed so that the form clearly marking nominative and accusative case (masculine gender) was used EITHER as the relative pronoun O R as the definite article i n the second noun phrase i n the relative clause. For example: (199) initial, REL.PRO.SUBJ, R E L . P R O marked, D E F . A R T unmarked der

Mann,

der

the

man:MASC REL.PRO: MASC:SG:NOM

ruft,

folgt

dem

calls

follows

the

das

Mädchen

DEF.ART: NEUT:SG:NOM/ACC

girhNEUT

Jungen

boy

'the man who calls the girl follows the boy' (200) initial, REL.PRO:SUBJ, R E L . P R O unmarked, D E F . A R T markea die

Frau,

the

woman:FEM REL.PRO: FEM:SG:NOM/ACC

die

erschießt,

schlägt

den

shoots

hits

DEF.ART: MASC:SG:ACC

den

Mann

DEF.ART: man:MASC MASC:SG:ACC Polizisten

policeman

'the woman who shoots the man hits the policeman' The results indicated that all three factors were relevant i n the interpretation o f the clauses. Finally-embedded clauses were easier to interpret than initiallyembedded clauses. This is the same result as Park found i n his imitation test and can be explained by the non-interruption principle. Clauses where the relative pronoun was subject o f the clause were easier than where i t was object, as G r i m m and Wintermantel also found. The exact order o f complexity o f the four types was: final subject < initial subject < final object = initial object. These results do not fit the prediction o f Sheldon's parallel function hypothesis. The difference between G r i m m and Wintermantel's and these results can be ex­ plained i f i t is the case that the older children i n this study were not relying on the interpretation strategy o f taking the first noun as actor for both clauses. The older children seem to have fewer problems i n separating the relative clause from the main clause. The older children still have problems using the case information w i t h i n the relative clause. For all sentence types, the marked article rather than the marked relative pronoun facilitated interpretation o f the clause. A possible explanation o f this is based on a surface structure interpretation model w h i c h is discussed i n detail i n M i l l s (1977a, p. 145). To summarize: The relative pronoun can only be identified as such when it is obvious that it is not a definite article, that is relatively late i n the clause. The definite article i n these clauses, however, can be

2.

The Acquisition of German

211

easily identified as such because o f the noun f o l l o w i n g immediately after i t , and can therefore be analyzed more easily for case information than the relative pronoun. The importance o f the position o f the relative clause and the tendency to associate subject w i t h first position i n a clause were examined again using ambiguous relative clauses ( M i l l s , 1977b). The clauses were constructed so that the relative pronoun and the definite article w i t h the second noun were unmarked for case. Adults and children aged 6 to 13 years were tested on their interpretation o f these relative clauses. Semantic and pragmatic information w h i c h might disam­ biguate the clause was reduced to a m i n i m u m . The f o l l o w i n g sentences are taken from the test material and illustrate the high degree o f reversibility (Slobin, 1966): (201) initial, R E L . P R O and D E F . A R T ambiguous die

Katze,

die

DEF.ART: FEM:SG:NOM

cat: F E M

REL.PRO: FEM:SG:NOM/ACC

das

Mädchen

DEF.ART: NEUT:SG:NOM/ACC

girl:NEUT

beißt

den

bites

DEF.ART: MAS:SG:ACC

sieht,

sees

Hund

dog:MASC

(202) final, R E L . P R O and D E F . A R T ambiguous der

Hund

DEF.ART:

dog:MASC sees

sieht

das

Mädchen,

das

DEF.ART: NEUT:SG:NOM/ACC

girl:NEUT

REL.PRO: NEUT:SG:NOM/ACC

die

Krankenschwester

ruft

DEF.ART: FEM:SG:NOM/ACC

nurse: F E M

calls

I f semantic and pragmatic information cannot guide interpretation, it w o u l d be expected that 50% o f the subjects w o u l d choose nominative case and 50% accusative case.

212

Mills

I n every age group, the relative pronoun was interpreted as being i n n o m i ­ native case significantly more often than i n accusative case. The frequency o f this preferred interpretation was not affected by the position o f embedding o f the relative clause. These results again suggest that subject is strongly related to first position i n the clause, possibly i n association w i t h the canonical w o r d order S V O o f main clauses, as discussed earlier. This is also the case i n French (see Clark, 1985). Frequency o f usage must also be taken into account, possibly as an independent factor, or possibly related to the principle above. 4.8. Infinitival

Complements

The only data on acquisition o f these structures come from comprehension experiments conducted by G r i m m and S c h ö l e r ( G r i m m et a l . , 1975). Children were tested on their comprehension o f sentences containing the causative verbs lassen 'to cause', erlauben 'to a l l o w ' , and befehlen 'to order'. These three verbs can have an infinitival complement i n w h i c h the verb can be intransitive or transitive w i t h an object o f the main verb, i n dative or accusative case according to the verbs used. The f o l l o w i n g sentences illustrate these possi­ ble sentence types w i t h the verb erlauben. (203) Maria

Maria

erlaubt

Hans

zu

arbeit

-en

allows

Hans

to

work

INF

'Maria allows Hans to work' (204) Maria

Maria

erlaubt

Hans

Uschi

zu

hol

-en

allows

Hans

Uschi

to

fetch INF

'Maria allows Hans to fetch Uschi' The verb lassen does not have the marker zu before the infinitive. The verbs erlauben and befehlen also allow a complement construction w i t h daß 'that', so that sentences 203 and 204 can also have the f o l l o w i n g form w i t h very similar semantic content: (205) Maria

Maria

erlaubt,

daß

Hans

allows

that

Hans

arbeitet

works

'Maria allows Hans to work' (206) Maria

Maria

erlaubt,

daß

Hans

Uschi

holt

allows

that

Hans

Uschi

fetches

'Maria allows Hans to fetch Uschi' The verb lassen can also be used w i t h an active infinitive but w i t h the agent and object reversed to produce a passive meaning. For example:

2. (207)

The A c q u i s i t i o n of G e r m a n

Maria

läßt

Uschi

von

Hans

hol

Maria

makes

Uschi

by

Hans

fetch INF

213

-en

'Maria causes Uschi to be fetched by Hans' G r i m m and S c h ö l e r tested comprehension o f such structures i n a manipulation task w i t h children aged 5;0 to 6 ; 4 . Reversibility was controlled for across all construction types by including a noun i n the complement clause which could not semantically function as agent. The verbs erlauben and lassen were tested w i t h the infinitive complement and the verb erlauben w i t h the daß complement. The verb versprechen 'to promise' was also included w i t h an infinitival complement. As i n English, this verb requires the selection o f the agent o f the main clause as agent i n the infinitival complement. 6

Reversibility was shown to be an important factor i n correct interpretation o f all construction types. The non-reversible clauses were easier i n every case. The complements w i t h daß were easier than the infinitival complements. The sug­ gested explanation is that daß clearly marks the beginning o f a subordinate clause, so that the elements o f that clause can be more easily identified as belonging to a unit for the purposes o f analysis. The infinitival complement has no such marking at the beginning, although a pause between the main clause and complement clause can mark the subordination. This w o u l d fit w i t h the proposed explanation for difficulties i n processing relative clauses (4.7) which are not clearly marked at the beginning because o f the similarity between the relative pronoun and definite article. G r i m m and S c h ö l e r state that the performance w i t h lassen i n non-reversible infinitival complements was worse than w i t h erlauben. They attribute this to the more general semantic content o f lassen since it can have the meaning o f ' a l l o w ' and 'cause'. However, they overlooked a formal ambiguity i n their test sentences which may have produced the different result. It is also possible that the particle zu before the infinitive helps to mark the clause as a subordinate clause. Lassen has no zu before the infinitive, and i n this respect is like modal verbs, where the agent o f the modal is always the agent o f the infinitive. As suggested above, the main problem i n comprehension may be the grouping o f elements i n infinitival clauses into a unit. This lack o f marking w i t h lassen could lead to errors i n interpretation. W i t h o u t detailed results these possibilities cannot be unravelled. The sentences w i t h versprechen were the most difficult. G r i m m and S c h ö l e r attribute this difficulty, i n part, to the fact that the construction w i t h this verb cannot be interpreted according to the M i n i m a l Distance Principle (Rosenbaum, 1965; C. Chomsky, 1969) w h i c h , they say, can be used to interpret infinitival complements i n German. According to this principle, the noun immediately Grimm and Scholer conducted a second experiment with these verbs using an interview tech­ nique. The results are however unclear because of difficulties in experimental design and are there­ fore not discussed here. 6

214

Mills

preceding the verb i n an infinitival complement is taken to be the agent o f that verb. This principle produces a correct interpretation for English and for German sentences o f type 203 w h i c h have no other nouns i n the clause. I t does not w o r k for sentences such as 204 however, where the principle w o u l d make Uschi the subject o f the verb holen. I t cannot be stated therefore that versprechen is an exception to this principle, since the principle only works for a limited structure in German. Since no information is given about the type o f errors made i n this and other verbs, it is difficult to discuss possible explanations except hypothetically. I f the M i n i m a l Distance Principle is involved, it w o u l d be expected that the object o f the complement clause w o u l d be selected as agent, since it is next to the verb. This d i d not happen w i t h relative clauses, however (see 4 . 7 ) , w h i c h have the same ordering o f object before the verb. I f the errors follow those o f the relative clauses, it w o u l d be expected that younger children take the first noun o f the sentence, that is the m a i n clause, to be agent o f every verb according to the principle o f associating subject w i t h first position, since they have difficulties indentifying a subordinate clause w i t h i n the main sentence, as discussed earlier. This w o u l d i n fact lead to a correct interpretation o f the versprechen sentences. When they can identify the elements o f the subordinate clause, it w o u l d be expected that they take the noun nearest to the subordinate clause to be subject. This leads to a correct interpretation for the majority o f verbs; versprechen is then an exception. A detailed study o f these constructions i n spontaneous speech and through careful experimentation w o u l d throw some light on the dependence o f interpreta­ tion strategies on the w o r d order possibilities allowed and most frequently used w i t h i n a language. 5. E r r o r - F r e e A c q u i s i t i o n 5.1.

Verb Morphology:

Person

The acquisition o f the inflections marking person on the verb, appears to involve relatively little error. This may seem a surprising claim when the data are first examined. Stern and Stern (1928) refer to the infinitive stage o f the verb i n the t w o and three w o r d utterances, since the verb frequently has the -en ending characteristic o f the infinitive. I n the discussion above on w o r d order, i t was mentioned that when the child first produces t w o - and three-word utterances, he commonly uses a verb-final rule. The suggested explanation for this was the frequency o f final infinitive forms, that is -en forms, i n caretakers' speech to children. The prevalence o f this ending i n children therefore suggests that i t is a directly imitated form w h i c h is unanalyzed. W h e n the verb occurs i n non-final position, the verb is finite and must therefore have the ending determined by the person and number o f the subject. A n analysis o f M i l l e r ' s data ( 1 9 7 6 ) collected 7

7

I am indebted to Ingrid Münnig for her help with this analysis.

2.

The Acquisition of German

215

TABLE 2.7 Distribution of Verb Endings in T w o - and Three-Word Utterances f r o m Miller's Data (Miller, 1976, p. 171-196) Order of Elements

Endings of Verb -INF ni naru and V-(r)areru; the f o l l o w i n g examples illustrate their usage: gakusei student

ga SUBJ

Hon book

o yom - u. DO read NONPAST

(Plain)

'The student will read the book'. sensei teacher

ga

Hon 0 0 - yomi HON read

ni to

nar - u. become NONPAST

(Honorific)

'The teacher will read the book'. sensei

ga

Hon o

yom - are - ru. read HON NONPAST

(Honorific)

'The teacher will read the book'. The -(r)are suffix, w h i c h is somewhat less formal, is homophonous w i t h the passive, and for vowel-final stem verbs, also w i t h the potential. Sonkeigo lexical substitutions affect some o f the most common verbs, such as iu 'say', w h i c h is replaced by ossharu to show deference to the subject referent. Substitutions are also made for other parts o f speech, for example, kata for hito 'person' and donata for dare ' w h o ' . Kenjoogo 'humble words' include similar linguistic devices, such as lexical substitutions and the verbal inflection o - I N F sum, w h i c h uses the honorific prefix plus infinitive w i t h sum ' d o ' . The f o l l o w i n g example illustrates the for­ mation o f the 'humble' verbal inflection, w h i c h is used when the subject referent is o f lower status than the non-subject referent:

4. gakusei student

ga SUBJ

sensei teacher

ni 10

Hon book

o DO

The A c q u i s i t i o n of Japanese o HON

kaeshi return

449

su - ru. do NONPAST

'The student will return the book to the teacher'. Table 4.11 lists some o f the most common lexical substitutions used i n honor­ ific language. A s this table shows, there are many irregularities i n the honorific forms. As S. I . Harada (1976) points out, either an entire predicate, or just the infinitive, may have a lexical replacement (e.g. go-ran instead o f o-mi from miru 'see'). Honorific forms are also sometimes irregular semantically, neutralizing semantic distinctions marked by plain verbs. Thus nomu ' d r i n k ' and taberu 'eat' both become meshi-agaru, and iru 'exist', kuru ' c o m e ' , and iku ' g o ' may all become irassharu. Matters are further complicated by the existence o f polite and hyper-polite speech styles for showing deference to an honored addressee; i n these styles the polite -mas inflections are added to honorific verb stems. Thus there are many structural and semantic features w i t h i n the system o f Japanese honorifics w h i c h probably contribute to the late acquisition o f these forms. As this very incomplete discussion o f honorifics indicates, the system i n ­ volves many morphological, semantic, and social complexities which could delay acquisition. Irregularities and exceptions i n the use o f the honorific prefix and lexical substitutions require that many honorific forms must simply be mem­ orized. Moreover, the morphology o f honorific inflections runs counter to certain basic operating principles w h i c h have been proposed. As K . Harada (1980a) has pointed out, the subject honorific pattern 6>-INF ni nam involves discontinuous morphemes; Slobin (1973) has proposed that children tend to avoid such inter­ ruption o f linguistic units. I n the case o f subject honorifics, t w o different inflecTABLE 4.11 S o m e C o m m o n Honorific Lexical Substitutions (Adapted f r o m S. I. Harada, 1976 and M a r t i n , 1975, pp. 3 4 8 - 5 1 )

da iru iru kuru iku shiru miru sum nomu taberu iu

'be' (COP) 'be' (AUX) 'exist' 'come' 'go' 'know' 'see/look' 'do' 'drink' 'eat' 'say'

SONKEIGO

KENJOOGO

de irassharu irassharu irassharu 1 o-ide ni nam irassharu 1 o-koshi ni nam irassharu 1 o-ide ni nam go-zonji da go-ran ni nam nasaru

de gozaru oru oru maim maim zonjiru haiken sum itasu

meshi-agaru

itadaku

ossharu

moosu

450

Clancy

tions are available to f i l l a single function, whereas children tend to prefer a oneto-one correspondence between form and function (Slobin, 1985). Honorific forms also violate the principle o f one-to-one mapping i n another way: the construction N o u n / A d j . + ni nam and the inflection -(r)are have both already been acquired at an early age i n different meanings, w h i c h probably makes their acquisition as honorifics more difficult. The pattern w i t h ni nam, w h i c h is similar to the honorific pattern, is acquired by 2-year-olds as an inchoative construction; -{r)are is mastered by 2- and 3-year-olds as the passive, and for certain verbs, also as the potential. Furthermore, as a passive inflection, -ir)are has actually been acquired w i t h diametrically opposed semantic functions. I n non-honorific speech, -(r)are is one o f the cues o f a passive sentence, i n w h i c h the subject should be interpreted as the patient rather than the agent. As an honorific, -{r)are is i n the active voice, and therefore the subject w i l l often be an agent. Not only is the system o f honorifics grammatically difficult, i t also presents the child w i t h a new concept, namely, that linguistic expression o f deference is required upon the mere mention o f a socially superior person and not just when addressing that person face-to-face. The special feeling o f speaking w i t h a supe­ rior must be extended to contexts where that person is not even present, i n w h i c h the social constraint which this presence might create may not be felt very deeply, i f at a l l . Thus the pragmatic and affective basis for the acquisition o f polite speech is lacking i n the case o f grammatically based honorifics, and this may be an additional barrier to their acquisition. Furthermore, it is probably difficult to separate the notions o f politeness toward an addressee and respect toward a socially superior referent, since i n many situations the honored person being referred to i n a sentence is also the addressee. This overlap may contribute to the trend noted by Harada (1980a) for the younger generation to use the polite -mas inflection instead o f honorifics, saying, for example, Sensei ga kaki-mashi-ta instead of Sensei ga o-kaki ni nat¬ ta 'The teacher w r o t e ' . As Harada points out, compared w i t h the o - I N F ni nam pattern, the -mas inflection is more regular and fully productive, and so these young people are substituting regular for irregular patterns. This change is also an instance o f Slobin's (1973) principle that " n e w functions are first expressed by old f o r m s , " since the -mas inflection is acquired much earlier than honorific inflections. I n accounting for the late emergence o f honorifics, i t is also important to consider the nature o f the input. I n m y data, there appeared to be ample oppor­ tunity for children to acquire the polite -mas level o f speech i n direct interaction, but I d i d not find mothers i n m y sample using honorifics to their children, even i n role-playing. This is i n accordance w i t h Okuda's (1979) findings on maternal expectations w i t h respect to the time at w h i c h honorifics should be acquired. I n Okuda's survey o f mothers, only 1 0 . 1 % thought that honorifics should be used before 6 years-of-age; 7 5 . 1 % felt that honorifics should start being used between

4.

The A c q u i s i t i o n of Japanese

451

6- and 10-years-of-age. I t may w e l l be that by the time a mother might feel her child is linguistically and socially advanced enough to deal w i t h honorifics, the kind o f role-playing w h i c h fosters acquisition o f -mas forms no longer is part o f mother-child interactions. Perhaps the main exposure to honorifics which young children receive is passive, that is, observing interactions among adults w h o are using keigo. I t is not clear how frequent such input is or how much direct instruction children receive from their parents on the usage o f honorifics. I t is easy to imagine that there w o u l d be extensive individual differences i n exposure to honorifics, depending upon the social class and kinds o f interactions typically engaged i n by a child's family members. I t is likely that out-of-context instruc­ tion and second-hand input are less effective sources o f input than the face-toface exchanges typical o f the child's early experience w i t h -mas. Thus the nature of children's exposure to honorifics, as w e l l as the linguistic difficulties inherent in the system itself, contributes to their late acquisition. 6.2.2. Pronouns. Compared w i t h the Indo-European languages, the devel­ opment o f pronouns is very late i n Japanese. B r o w n (1973, p . 210) reports that i n Stage I , at l ; 6 - 2 ; 3 years-of-age, the three American children i n his sample were using the personal pronouns / , you, it, and my. I n m y data, i n contrast, none o f the three children aged 1 ;11—2;4 were using second person pronouns. The t w o little boys occasionally referred to themselves as boku T , but at 2;2 years-old, the girl still had not spontaneously referred to herself w i t h a pronoun, although her linguistic development was certainly no less advanced than that o f the boys. These children usually referred to themselves by name, as Japanese children continue to do, at least occasionally, until the age o f 6 years (Ide, 1977). This is a very general finding i n the data on Japanese language acquisition. For example, Horiguchi (1979b, 1981a), based on longitudinal records o f three children, re­ ports that the earliest and by far most common form o f self-reference was nickname + -chan, the diminutive suffix. This form o f self-reference accounted for 99% o f all sampled self-references for one boy during his third year, and 94.2% for one g i r l . M o s t little boys, however, use boku T for self-reference (Ide, 1978), as the t w o boys i n m y sample were beginning to do. Fischer (1970) regards this usage o f names for self-reference as one o f several instances i n w h i c h Japanese families differ from American families i n permitting developmentally earlier ways o f speaking to be continued for a longer time. T o understand this belated acquisition o f pronouns, it is important to recognize that the system o f Japanese personal reference is very different from that o f IndoEuropean languages. For example, w i t h i n the family, parents and older siblings are addressed w i t h kinship terms, such as otoosan 'father', okaasan 'mother' (papa and mama i n the earliest stages o f development), oniisan 'older brother' and oneesan 'older sister'. These family members also often refer to themselves by these kinship terms where appropriate w i t h i n the family. One's children and younger siblings are usually addressed and referred to by their first name or

452

Clancy

nickname plus the diminutive suffix -chart. Thus w i t h i n the Japanese f a m i l y , names and kinship terms often substitute for personal pronouns. (See Fischer [19641 for a description o f personal reference w i t h i n Japanese families.) Not only do nouns frequently substitute for pronouns i n Japanese, but often any overt form o f personal reference w i l l be omitted. I n Japanese, person is not marked i n the morphology o f verbs, and pronouns are entirely optional. Ide (1977, 1978, 1978-9) has analyzed the forms used for first and second person reference i n the speech o f 18 children from Wi to 6 years-of-age. After these children's speech was observed at a nursery school, detailed questionnaires were sent out to 150 mothers o f children from 1 to 6 years o l d , and filled out during one week o f observing their children's use o f first and second person references in a variety o f contexts. Ide found that the most frequent form o f reference to first, second, and third persons was ellipsis; person deixis is thus usually under­ stood from context. This creates many difficulties for speakers o f Indo-European languages (confusions between first and second person can be particularly disas­ trous!). However, Ide (1977) reports that a dozen Japanese children can play together for half an hour at a time without using a single first or second person reference. Similarly, Horiguchi (1979b) reports that overt first person references occurred w i t h increasing frequency i n the speech o f her t w o children during their third year, but the percentage was only 6 . 5 - 1 3 . 2 % i n the boy and 1 3 . 5 - 2 4 . 9 % i n the g i r l . (Second children i n Japanese families apparently feel more need to use explicit self-references; see also Okubo, 1981a.) Since titles and names, which Westerners typically regard as third person references, generally substitute for first and second person pronouns i n Japanese, the frequency o f personal pronouns i n the input to Japanese children is extremely l o w compared w i t h IndoEuropean languages. Moreover, i n Japanese there are several different options available for first and second person pronouns, and these are conditioned, not grammatically, as i n English, but socially, by such factors as the age, sex, and relative social status o f speaker and hearer, and the nature o f the conversational context. Children hear different pronouns from different people, and hear the same people referring to themselves w i t h different pronouns i n different contexts. Thus Japanese pro­ nouns constitute a good example o f many forms being used for a single function; their late acquisition is consistent w i t h Slobin's (1985) proposal that children prefer a one-to-one mapping o f f o r m and function. Table 4.12 lists some o f the most common Japanese pronouns, f o l l o w i n g Uyeno ( 1 9 7 1 , p . 16). I t is also possible to add plural markers, such as -tachi, to several o f these pronouns. The third person pronouns seem similar to English ones i n marking number and gender. However, they are used very infrequently, and have somewhat different functions. For example, kare 'he' and kanojo 'she' may mean girlfriend and boyfriend, although this usage is less common nowa­ days (Okubo, personal communication). A s more neutral referential forms, Jap­ anese third person pronouns tend to be used i n quasi-academic or formal style

4.

The A c q u i s i t i o n of Japanese

453

TABLE 4.12 Some Common Japanese P r o n o u n s (From U y e n o , 1971) FIRST

SECOND

THIRD

PERSON

PERSON

PERSON

watakushi watashi atashi boku ore

anata anta kimi omae

kare 'he' kanojo 'she' karera 'they' ( M ) kanojotachilkanojora 'they' ( F )

( A o k i , personal communication). None o f the children in m y sample ever used third person pronouns, and Okubo's daughter at the age o f 6 years had still not acquired them (Okubo, 1967, p . 72). As Uyeno ( 1 9 7 1 , Chap. 2.2.2) discusses, the first and second person pro­ nouns listed above are restricted i n usage by the sex o f the speaker and some­ times also o f the hearer, as w e l l as by their social relationship, relative status, degree o f intimacy, and the speech context. F o l l o w i n g Uyeno, the pronouns are listed i n order o f decreasing formality. Certain pronouns are paired at the same stylistic levels; thus ore T and omae ' y o u ' are at the same, extremely casual level o f speech. The more formal pronouns, wata(ku)shi T and anata ' y o u ' can be used by either male or female speakers to addressees o f either sex, although actually these pronouns are used much more frequently by women. Atashi T is used only by women to addressees o f either sex; boku, kimi, ore, and omae are used only by men. W o m e n use an(a)ta to their husbands, boyfriends, to their female friends, and, i n certain contexts, to their children. Omae ' y o u ' is how a man addresses his wife or girlfriend, his buddies, and, sometimes, his children. Although wata(ku)shi T is formal enough to use i n conversation w i t h a superior, anata ' y o u ' is not; superiors are always referred to by title or last name plus title. Boku T may be used by boys even i n formal situations, although they learn to use wata(ku)shi by the time they enter j u n i o r high school (Ide, 1977). Kimi ' y o u ' is appropriate for a man to use only when addressing social equals or inferiors w h o m he knows fairly w e l l . Both ore T and omae ' y o u ' are limited to particular contexts w i t h peers or inferiors. There is thus an interesting discrepancy between the sexes i n their use o f pronouns. W o m e n are limited to the more formal options, and i f they address their husband and children w i t h pronouns, they use the more formal anata. I n contrast, men's options cover the full range o f for­ mality/informality, and the pronoun they use to their wives, girlfriends, or children is the least formal one available. I n m y data, the earliest and by far the most common form o f reference used when speaking to a child was his or her name, usually nickname, plus the

454

Clancy

diminutive suffix -chan. For example, the mother o f a boy o f l ; l l - 2 ; 0 years always addressed h i m using this form as a second person reference; at 2;1 yearsof-age, she began calling h i m boku T , as i n the f o l l o w i n g example: (while reading her child a Winnie the Pooh story) boku,

Puu

I

Pooh

- san

HON

mi

look

- te.

IMP

Lit., ' I , look at Pooh'. As Ide (1977) and Fischer (1970) have pointed out, boku or boku-chan tend to function as alternate proper names for little boys. Fischer (1964) proposes that this usage originated as a way o f teaching little boys to call themselves boku. I n m y sample, about 2 months after the mother i n the above example began to use boku as a second person reference, her son also began to refer to himself as boku T . This boy's mother also w o u l d model the use o f boku as a first person pronoun for h i m . For example, one time when the child (at 2;2) had been awaiting m y arrival, his mother said to h i m , BOKU matte ta n da yo tte iwanakya 'Say, " I was w a i t i n g " ' . The mothers o f 2-year-old boys i n m y sample used boku frequently as a vocative and second person reference, and both boys were beginning to use boku for self-reference. I n contrast, the mother o f the 2-year-old g i r l i n m y sample rarely referred to her as watashi

T,

and this child at 2;3 years had not

spontaneously used watashi to refer to herself. I n her survey o f Japanese moth­ ers, Ide ( 1 9 7 8 - 7 9 ) found that girls generally started using watashi later than boys used boku. The use o f watashi as an alternative name seems to be more l i m i t e d than boku; for example, the suffix -chan w h i c h appears w i t h names cannot be used w i t h watashi, although i t occurs w i t h boku. Iwabuchi and Muraishi (1968) report acquisition o f watashi as late as 2;8 years i n one g i r l . However, there are apparently great individual differences i n this domain. Okubo (1967, p . 71) reports that her daughter was using atashi T for self-reference at 1;10 years o l d , and Fujiwara (1977, p . 130) cites this form i n the speech o f a g i r l o f the same age. O n the other hand, Horiguchi's son was not yet using boku for self-reference by 2; 11 years, although most boys acquire this f o r m much earlier. The way pronouns are used i n the input to different children probably plays an important role i n the t i m i n g o f their acquisition. Except for the type o f modeling described above for boku, parents' speech does not seem to be as important for the acquisition o f pronouns as the speech o f other children. This is because, as Horiguchi notes, parents do not usually refer to themselves w i t h pronouns i n front o f their children. I f a child has older siblings, their usage w i l l lead to earlier acquisition o f pronouns i n the case o f same-sex siblings, and to imitation o f inappropriate pronouns when the sibling is o f the opposite sex. For children w h o do not have older siblings o f the same sex, entering nursery school or kinder­ garten and coming into contact w i t h children outside the family circle leads to

4.

The A c q u i s i t i o n of Japanese

455

acquisition o f the appropriate pronouns. Horiguchi reports that her son began to use ore at 2; 10 years, when a boy w i t h an older brother j o i n e d his class; her daughter began using watashi at 2;2 years, when she entered nursery school and heard this form i n the speech o f a classmate w i t h an older sister. Horiguchi proposes that consistently correct use o f Japanese pronouns for self-reference is not established until about 3 years, when the child's consciousness o f sex dif­ ferences is also f i r m l y established. As different forms o f self-reference are acquired, children also begin to use them selectively i n different contexts. For example, one boy o f 2;4 years i n m y sample, who had an older sister, called himself oniichan 'older brother' when playing w i t h his toy Pooh bear. Okubo (1967, p. 71) reports that her daughter called herself by name when speaking to her parents, but used atashi T when speaking to her friends. Examples cited by Fujiwara (1977', p. 246) from a girl o f 3;0 years o l d indicate that she called herself by name i n sentences in which she was using plain verb forms, but used watashi T i n sentences w i t h polite inflec­ tions. Similarly, Horiguchi's 2-year-old daughter frequently used watashi i n sentences ending w i t h the feminine sentence-final particle wa; this child also used certain non-verbal components o f female style i n these cases. A l t h o u g h this child did not always combine elements o f female style consistently (watashi also occurred w i t h the more masculine sentence-final n da yo), early examples o f watashi reflect some preliminary awareness o f the different features appropriate to a particular speech style. This k i n d o f code-switching has been documented i n detail by Ide (1977, 1978). Ide found that the use o f different forms o f first and second person reference was governed by both inter-personal factors such as relative age, sex, social status, and personal relationship between speaker and addressee, and by "intra-personal" factors, i.e. psychological or behavioral factors such as selfassertion or dependency. Boys and girls used their name plus -chan for selfreference (boys also used boku-chan) w i t h i n their families i n a mood o f amae or dependency. (See D o i , 1973, for a discussion o f this fundamental Japanese concept, which refers to a wide class o f behaviors and feelings which presume upon the indulgence o f another.) Girls i n Ide's study switched to (w)atashi T or oneesan 'older sister' to express an attitude o f kidori 'affected maturity' es­ pecially w i t h younger friends or children. Boys switched to ore T to show solidarity w i t h a peer group. Ide states that " i n a group situation, the w o r d ore becomes a k i n d o f 'password', a verbal reflection o f an important psychological attitude." Younger boys use ore to gain acceptance into a play group o f older boys, who use this form i n active play, when showing off, bossing their peers around, or seeking the admiration o f younger children. Horiguchi's son also used ore when showing o f f or calling attention to his strength, when he was still only 2-years-old and occasionally using the female pronoun. Second person pronouns seem to be acquired even later than first-person pronouns, probably at about the time children enter kindergarten or begin spend­ ing time i n peer groups. For example, when a boy is referring to himself as ore,

456

Clancy

he w i l l typically be addressing his friends as omae. Errors i n the use o f second person pronouns help provide insight into the factors underlying their acquisi­ tion. Omae, for example, is sometimes used inappropriately by kindergarten boys. The 4-year-old son o f a friend o f mine, w h o usually referred to me as Patorishiya-san w o u l d switch to omae ' y o u ' when involved i n active play and rough-housing. His mother said that he d i d the same w i t h her, and she invariably corrected h i m , asking Omae tte dare? ' W h o is " o m a e " ? ' . I n these cases, the child was apparently allowing the affect associated w i t h the behavioral context to take precedence over inter-personal social factors, such as the age and status o f the addressee, i n selecting a second person reference. As w i t h the use o f polite -mas inflections,it seems that young children may at first rely heavily upon their o w n affect i n a particular context i n deciding what linguistic forms to use. The second person pronoun an(a)ta also emerges late. Okubo (1967, p . 71) reports that her daughter began using anta ' y o u ' to her friends regularly at 4 years-of-age. I n m y data, mothers o f 2-year-olds used an(a)ta to their children very rarely, typically when instructing, criticizing, or otherwise distancing them­ selves somewhat, as i n the f o l l o w i n g cases: (telling her boy of 1; 11 years to give my assistant an apple) doozo

tte.

anata

sanzan

saw at - ta

no

please

QUOT

you

completely

touch

one

ja

PAST

- naku - te.

COP+TOP

not

CONT

'(Say), "Help yourself". Not one that you completely touched'. (to her daughter of 2;1 years who calls all colors pink) anta

pinku

shika

wakara

you

pink

except

know

- na

- i

N E G NONPAST

no.

EP

'You only know pink'. Horiguchi (1979b, 1981a) reports only a few instances o f an(a)ta i n the speech o f her son and daughter during their third year, to one another and, inappropriately, to their mother. Ide's data indicate that boys learn to identify anata as a woman's w o r d , perhaps because they are never addressed this way by their fathers. Ide points out that anata ' y o u ' , the second person pronoun w h i c h girls acquire, reflects a formal attitude, whereas omae, the second person pronoun w h i c h boys first acquire, reflects a coarse, extremely casual attitude. Kimi, the more formal second person pronoun for boys, is not acquired until about 5 - 6 years-of-age, later than girls acquire anata. Apparently, as Ide suggests, boys first learn to relate to their peers i n an attitude o f rough solidarity, the ore 'YI omae ' y o u ' mode, and girls i n an attitude o f affected maturity and formality, the atashi T V

4.

The A c q u i s i t i o n of Japanese

457

an{a)ta ' y o u ' style. Boys must proceed to acquire another, more formal level, whereas there are no less formal options for girls to master. The late acquisition o f pronouns i n Japanese can be attributed to their l o w frequency o f occurrence, the large number o f forms, and most importantly, to the social and psychological complexity o f appropriate usage. Pronouns form part o f the sociolinguistic structure o f Japanese; their use requires code-switching, just like selection o f appropriate verbal inflections. The type o f input required may preclude the acquisition o f pronouns outside o f certain contexts, such as peer group, same sex situations, w h i c h may not be available to some children until kindergarten. Japanese children must learn to be sensitive to the relevant at­ tributes o f the addressee vis-a-vis themselves, and must also learn to use different forms o f self-reference i n different moods and contexts. Japanese pronouns do present some problems for the language learner, but i n contrast w i t h the acquisi­ tion o f honorifics, there do not seem to be any cases o f children w h o fail to acquire pronouns. This is partly because pronouns occur more frequently than honorifics and i n a wider variety o f contexts, but also because pronouns i n Japanese are strongly linked to the child's developing sense o f identity, both personal and social.

7. R e o r g a n i z a t i o n s i n D e v e l o p m e n t 7. /. Prenominal

Modifiers

The development o f prenominal modifiers i n many Japanese children undergoes an interesting reorganization i n the course o f development. I n Japanese, many types o f relations between nouns, such as possessive, appositional, and some­ times locative, are expressed by the genitive particle no i n constructions o f the form N no N , i n w h i c h the modifier plus no precede the head noun. I n contrast, adjectives precede the noun they modify w i t h no special grammatical marker; the same is true o f relative clauses, since there are no relative pronouns i n Japanese. The following examples illustrate these three types o f prenominal modifiers: Noun

otoosan father

no kuruma G E N car

'father's car' Adjective

ookii big

kuruma car

'a big car' Relative clause

otoosan father

ga SUBJ

kat - ta buy PAST

'the car that father bought'

kuruma car

Clancy

458

One o f the earliest case particles w h i c h children acquire is the genitive no, which they first use to indicate possession i n sentences o f the form N no ' I t ' s N ' s ' . Okubo (1967, p . 89) and Miyahara (1974) both found that no emerged at 1;8 years-of-age, as i n the f o l l o w i n g example from Miyahara: Noriko-chan

no.

GEN 'It's Noriko's (= mine)'. I n the next stage o f development, many children begin to produce both the modifier and the head noun i n t w o - w o r d constructions. I n m y sample, one child of 1; 11 years was at this stage. This child used t w o types o f nominal modifiers: nouns and adjectives. Both were formed simply by juxtaposing the related words in the order: MODIFIER - HEAD. Thus this child correctly produced adjectival modifiers, such as, akai

red

buubuu

car

'a red car' but omitted the obligatory genitive marker i n constructions w i t h nouns, such as: *neechan

buubuu

older sister car 'older sister's car' This child even omitted no i n repeating N noN constructions from the i m m e d i ­ ately preceding adult utterance at this stage. A t 2;2-2;4 years-of-age, this child began to use no productively i n N no N constructions indicating possession. No was never omitted where appropriate, and began to occur i n constructions w i t h body parts, locations, and other types o f relations between nouns, such as: Yotchan

no

Yotchan

G E N penis

o-chinchin

'Yotchan's penis' Oosaka

no

Osaka

G E N grandpa

ojiichan

'grandpa who lives in Osaka' During the same period, this child also produced the f o l l o w i n g constructions w i t h adjectival modifiers:

4. *aoi

blue

no

T h e A c q u i s i t i o n of Japanese

459

buubuu

G E N car

'a blue car' *chitchai

tiny

no

buubuu

G E N car

'a tiny car' It seems to be very common for Japanese children to go through a stage o f overgeneralizing no w i t h prenominal adjectives. Fujiwara (1977, pp. 1 2 4 - 5 , 157, 218) notes this error i n the speech o f three different children, aged l ; 1 0 - 2 ; 4 years, and Iwabuchi and Muraishi (1968) cite similar examples from a child o f 1;8 years o l d . Okubo's daughter, however, is an exception to this trend ((Okubo, 1967, p. 108). 27

Harada (1980b) found the same course o f reorganization w h i c h appeared i n my data. Her daughter first used the N no N pattern without no; at a later stage, when she was using N no N correctly, she produced A D J no N constructions as w e l l . Harada's daughter continued this mistake even after acquiring relative clauses, and added no between relative clauses and head nouns, as i n : *kaijuu

monster

ni

nat

to

become

- ta

PAST

no

onna no ko

G E N girl

'the girl who became a monster' *usachan

rabbit

ga

tabe - ta

no

SUBJ

eat

G E N carrot

PAST

ninzin

'the carrot that the rabbit ate' Thus many Japanese children initially treat nominal and adjectival modifiers in the same way, simply placing them before the head n o u n . W h e n N no N constructions have been acquired, children often apparently notice the discrep­ ancy between the formation o f modifying constructions w i t h nouns and adjec2 8

Nominal adjectives such as kiree 'pretty/clean' take the na form of the copula rather than no as prenominal modifiers, e.g. kiree na hana 'pretty flowers'. At first, like no, na may be omitted. Okubo (1976, p. 216) cites * kiree hana at 1;8 years in her daughter's speech; na was used correctly by 2;2 years. Iwabuchi and Muraishi (1968, p. 157) found omission of na, and also overgeneralization of no, as in *kiree na no densha 'clean/pretty streetcar' at over 2Vi years old. Apparently, as with true adjectives, an early pattern of simple juxtaposition, Modifier + Noun, is used, followed by overgeneralization of no in some children. Not all children go through a stage of producing *NN genitive constructions, however. Komura (1981) describes the development of one child who developed N n o N constructions directly from N no constructions, without an intermediate stage of *NN. (See section 10.2 for a fuller discussion of Komura's findings.) 27

28

460

Clancy

tives, and overgeneralize no so that all prenominal modifiers are again formed i n the same way. A t this point, i t seems that these children have recognized the single grammatical category o f nominal modifiers, and see the functional sim­ ilarity between the t w o patterns o f noun modification which have been develop­ ing, one w i t h nouns and one w i t h adjectives. The reorganization o f adjectival modifiers w i t h no provides strong support for Slobin's operating principle (1985) that children prefer to use a single form for a single function. A t this stage o f development, it seems that the N no N pattern is too f i r m l y established i n familiar constructions, such as the possessive, for the child to resolve the discrepancy i n form between nominal and adjectival modifiers by returning to juxtaposition for nouns. I n m y data, N no N constructions are more frequent than prenominal adjectives, and this difference i n frequency may also favor the use o f no w i t h adjectives. The overgeneralization o f no is also consistent w i t h Slobin's operat­ ing principle that " u n d e r l y i n g semantic relations should be marked overtly and c l e a r l y , " and provides strong support for his prediction that " i f a category is sometimes marked by 0 and sometimes by some overt phonological form, the latter w i l l , at some stage, also replace the 0" (Slobin, 1973). Obviously, i n achieving adult competence, Japanese children who overgeneralize no have to go through another phase o f reorganization, i n which they distinguish nominal from adjectival and sentential modifiers. 7.2. Word

Order

and Case

Particles

I n English, the subject and object o f sentences are indicated by w o r d order; i n many languages w i t h freer w o r d order, such as Turkish, inflections are used to indicate grammatical relations. Japanese is interesting i n that i t represents a k i n d of middle ground, using both options, but neither one entirely consistently. The canonical w o r d order is S O V , but O S V is also p o s s i b l e , and either subject or object may optionally be placed after the verb. Grammatical relations are marked by postpositions such as ga (subject) and o (direct object). However, certain postpositions, including ga and o, may be omitted i n casual conversation, and this ellipsis occurs even more frequently i n the speech addressed to very young children (see section 4 . 2 ) . The strategies w h i c h Japanese children develop for sentence production and comprehension reflect the importance o f both w o r d order and case particles i n the adult language. As children figure out the rela­ tionship between these devices, various reorganizations take place. These have been documented primarily i n experimental studies. 29

7 . 2 . 1 . Simple Active Sentences. Since case particles typically do not emerge until after m u l t i - w o r d utterances are being produced, and then appear only gradually, the developmentally earlier strategy for marking grammatical Kuno (1978) cites a large-scale study of modern Japanese sentence patterns in journalistic writings which found that the ratio of SOV to OSV word order was 17:1. 29

4.

The A c q u i s i t i o n of Japanese

461

relations is w o r d order. Miyahara (1973) documents an interesting reorganization which occurred i n the speech o f her daughter. A t 1; 11 years, this child used object-verb utterances i n fixed order, although i t is possible to postpose objects after the verb i n colloquial speech. R i g i d O V order was maintained until the child acquired the direct object marker o; only then d i d postposed objects begin to appear, as i n : ire

put in

- ta

PAST

yo

aiai

o.

EMPH

monkey

DO

'I put it in, the monkey'. It w o u l d be grammatical to omit the object marker i n postposed position, as i n pre verbal position. This child was apparently at first using w o r d order to indicate the object-verb relation; when the object marker was acquired, the word-order strategy could be relaxed. Miyahara's finding is very similar to Slobin's observa­ tion that one Russian child preserved a more r i g i d subject-object w o r d order than the adult language until he acquired the accusative inflection, at which point he began producing both subject-object and object-subject sentences (Slobin, 1966), and similar to findings o f Radulovic (1975) i n Serbo-Croatian. Experiments on the comprehension o f simple active sentences document re­ organizations i n the roles o f w o r d order and case particles i n children's strategies for processing sentences. Hayashibe (1975) and Hakuta (1982) found that at approximately 4 years-of-age, Japanese children w i l l impose a consistent i n ­ terpretation upon strings o f t w o nouns and a verb, taking the first noun to be the agent. I n each o f these studies, children were presented w i t h semantically revers­ ible and irreversible sequences o f the form N N V , N V N , and V N N w i t h no case particles, and were asked to act them out. Testing 60 children between 3;0 and 5;11 years-of-age, Hayashibe found that an agent-patient strategy for interpreting the nouns appeared at about 4 years-of-age, regardless o f the position o f the verb. Hakuta, testing 30 children between 3;3 and 6;2 years-of-age, reports a statis­ tically significant increase i n preference for interpreting the first noun as agent at 4;2-5;2-years-old. I n both studies, even the 3-year-old children had no difficulty acting out semantically irreversible sequences; thus the earliest strategy appears to be semantic/pragmatic, based upon the likelihood o f a certain referent being the agent o f a particular action. The strategy o f taking the first noun as agent, which emerges at about 4 - 5 years-of-age, appears to be derived from the d o m i ­ nant adult S O V w o r d order o f Japanese. However, this strategy is not sensitive to the position o f the verb; Hakuta found that N N V sequences were no more likely to receive an agent-patient interpretation than N V N or V N N strings. Apparently, the focus o f the syntactic strategy is the first noun itself, rather than its position relative to the other noun or verb. Thus by about 4 years-of-age, Japanese children have acquired a sensitivity to w o r d order i n the absence o f semantic or morphological cues.

462

Clancy

Experiments w i t h simple active sentences incorporating case particles reveal the early preferences o f Japanese children for the standard adult S O V w o r d order. Various studies have assessed the performance o f children i n acting out semantically reversible and irreversible simple active sentences i n S O V and O S V order, such as Hayashibe (1975), testing 30 children from 3;0 to 5;11 years o l d ; Hakuta (1982), w h o tested 48 children between 2;3 and 6;2-years-old; and Sano (1977), who studied 80 children between 3;3 and 6;8 years. A g a i n , semantically irreversible actives were generally w e l l understood by even the younger children, but developmental trends were apparent i n the interpretation o f reversible sen­ tences. Each o f these studies found that i n the dominant S O V order, sentences were acted out correctly more frequently. I n each study, there were a few children who consistently interpreted all O S V sentences as i f they were S O V , ignoring the case particles completely. Analyzing errors, Hayashibe found that only the older children i n his sample (mean age 5;0 years) could interpret both SOV and O S V reversible actives correctly, showing that they were able to rely upon case particles even when these d i d not occur i n the expected w o r d order. Hayashibe concludes that w o r d order is a developmentally earlier strategy than case particles; the former seems to develop at about 4 years-of-age, the latter at about 5. Similarly, Hakuta (1977) found that O S V sentences were acted out correctly by about 5;4 years-of-age. Thus, one reorganization w h i c h must take place i n the development o f processing strategies is a shift i n emphasis away from word order toward case particles. It is interesting to compare these findings w i t h crosslinguistic data on sentence comprehension. Hakuta (1982) contrasts his results w i t h those o f Slobin (1982) on Turkish. Unlike Japanese children, from the age o f 2 years, Turkish children are able to comprehend sentences w i t h all six possible orders o f subject, object, and verb. I n Turkish, w o r d order is quite free and inflections are obligatory. Thus inflections provide a more consistent cue than i n Japanese, where the direct object marker o is frequently omitted, and the subject marker ga may also occasionally be omitted. Moreover, since Turkish w o r d order is so variable, children are forced to rely primarily upon inflections to interpret sentences, whereas i n Japanese, the dominant S O V w o r d order leads children to form a fallible agent-patient-action strategy. Hakuta concludes that i n languages w h i c h utilize both w o r d order and inflections to mark grammatical relations, children formulate processing strategies w h i c h incorporate the t w o variables, and find it difficult to comprehend sentences w h i c h violate the expected correlation between them. The findings on Japanese children's strategies for processing sentences are borne out by studies o f imitation, reported by Hakuta (1982) and Sano (1977). Hakuta had 14 children aged 3;8-6;8 imitate simple active sentences w i t h case particles, i n S O V and O S V orders. There were no errors w i t h S O V sentences, whereas only 6 3 . 4 % o f the O S V sentences were correctly imitated. O f the errors, 85.4% retained the order o f nouns, but switched the particles to create canonical

4.

The A c q u i s i t i o n of Japanese

463

SOV sentences, or changed the first particle to the subject marker ga, so that both nouns were marked by ga. Clearly, Japanese children expect the first noun of a sentence to be a ga-marked subject. The children i n Sano's study also imitated S O V actives correctly, but changed the particles i n O S V sentences to create canonical S O V sentences. These findings are supported by the results o f a production task conducted by Hakuta (1982), i n w h i c h 48 children between 3;3 and 6;2 produced only 2 . 2 % O S V actives. M o s t sentences were canonical S O V actives, w i t h ga marking the subject 90% o f the time by the age o f 3;9 years. Thus imitation data provide further evidence for Hakuta's proposal that Japanese children expect agreement between case particles and their location w i t h i n a sentence. 7.2.2. Passive Sentences. A s i n English, Japanese passive sentences re­ verse the semantic roles o f sentence subject and object, and provide special casemarking for the agent, and a special inflection on the verb. The f o l l o w i n g examples illustrate the contrast between active and passive sentences i n Japanese: Active:

Taroo

ga SUBJ

Hanako

o DO

but - ta. hit PAST

'Taroo hit Hanako'. Passive:

Taroo

ga SUBJ

Hanako

ni but - are - ta. A G T hit PASS PAST

'Taroo was hit by Hanako'. In the active sentence, ga marks the subject, which is also the agent, and o marks the direct object or patient. I n the passive sentence, the patient is now the subject, and the agent is marked w i t h ni, which also marks indirect objects, locatives, certain sources, as w e l l as the agents o f causative sentences. The verb takes the passive suffix -(r)are before tense marking. L i k e active sentences, passives may appear i n both S O V and O S V order; for example, the above passive could also be phrased Hanako ni Taroo ga butareta ' B y Hanako, Taroo was h i t ' . As expected, Japanese children acquire active sentences, at least i n the stan­ dard S O V order, before passive ones. I n the earliest stage documented by experi­ mental research, Japanese children are able to discriminate passive sentences from active ones, but do not k n o w how to interpret the passive sentences cor­ rectly. Harada (1977) reports that at 3; 11 years-of-age, her daughter's com­ prehension o f passive sentences was still random. Both Sano (1977) and Hakuta (1982) found that children o f about 3Vi years o l d performed randomly when acting out passive sentences, but acted out S O V actives correctly about 70% o f the time. The fact that very young children do discriminate between actives and

464

Clancy

passives shows some early sensitivity to the special morphology o f passive sentences: the case particle ni and the passive inflection on the verb. I n the next stage o f development, many Japanese children begin to interpret passive sentences consistently, but incorrectly, as i f they were actives. I t is interesting that this development takes place at about the same age at w h i c h children w i l l impose an S O V interpretation on strings o f t w o nouns and a verb, approximately 4 years o l d . Hakuta (1982) found a U-shaped curve for the ac­ quisition o f passive sentences by Japanese children. Grouping his 48 subjects by their mean length o f morphemes i n a production task (a measure w h i c h correlated highly w i t h age), Hakuta found that the group w i t h the lowest production scores interpreted passives randomly, but the next group, w h i c h had a mean age o f 3;9 years, consistently misinterpreted S O V passives as S O V actives. I n the third group, interpretations again became random; the fourth group, w i t h a mean age of 5;4 years, interpreted S O V passives correctly over 70% o f the time. Thus, as in the development o f passive sentences i n English (see Bever, 1970), there is a stage i n w h i c h the interpretation o f S O V passives i n Japanese is consistently reversed. However, as Hakuta's analysis reveals, this misinterpretation o f Japanese passives is not merely the result o f applying an SOV w o r d order strategy to all passive sentences. As Hakuta points out, i f Japanese children were relying solely upon a w o r d order strategy, i n w h i c h N N V sequences were interpreted as agentpatient-action regardless o f case-marking, then we w o u l d expect them to perform w e l l on O S V passives, since an agent-patient-action interpretation w o u l d be correct. Instead, Hakuta found that performance on O S V passives was signifi­ cantly worse than on S O V actives. Furthermore, the children d i d not perform any better on O S V passives than on O S V actives and S O V passives, w h i c h have the reverse semantic sequence, patient-agent-action. Sano (1977) made similar find­ ings i n a study o f 80 children aged 3;3-6;8; their ability to act out S O V actives correctly was much superior to their performance on O S V actives, but there was no such difference between S O V and O S V passives. Hakuta concludes that i f the subject marker ga is not on the first noun o f a sentence, children cannot take advantage o f the semantic information i n the particle i n interpreting O S V actives. I f the semantic sequence is agent-patient-action, but the agent is not marked by ga, as i n O S V passives, children are unable to benefit from the correspondence of the sentence to the canonical semantic sequence. Apparently, the presence o f the agent marker ni after the first nominal i n O S V passives blocks the agentpatient-action interpretation assigned to S O V actives beginning w i t h a gamarked nominal. Thus Japanese children's reversal o f passive sentences is not based upon a fixed w o r d order strategy, as appears to be the case i n English, but rather, Hakuta concludes, upon the correlation between w o r d order and case particles i n expected positions. I t is interesting to note that Slobin and Bever (1982) have made similar findings on the processing o f sentences by SerboCroatian children. These children performed better on O V S sentences w i t h ac-

4.

The A c q u i s i t i o n of Japanese

465

cusative marking on the first noun than on O V S sentences w i t h an unmarked first noun (neuter zero) and a marked subject (feminine nominative) on the second noun. These authors also conclude that a noun w i t h nonstandard case-marking i n initial position blocks the canonical sentence-processing strategy. To acquire passive sentences, Japanese children must learn to re-interpret ga as marking the patient rather than the agent when passive morphology occurs later i n the sentence. A critical factor i n this reorganization may be focusing upon the ^/-marked nominal, w h i c h indicates the agent i n passive sentences. Sano (1977) found that when imitating sentences w i t h passive verbs, children often supplied the particle ni i f it was omitted from the model sentence, or substituted ni for some other particle i n the sentence. As noted above, both Hakuta and Sano failed to find the large difference between S O V and O S V order w i t h passives that is found w i t h actives. I t is interesting that a number o f children acquire O S V passives before S O V passives, although this is extremely rare w i t h actives. For example, i n Hakuta's data, a higher percentage o f children i n group three, w i t h an average age o f 5 ; 1 , controlled O S V than S O V passives. Perhaps this is because O S V order i n passives places the ra-marked nominal i n sentence-initial position, where it can serve as an early trigger for applying a different processing strategy. The role o f specific morphological cues i n developing processing strategies can be seen by comparing Japanese children's comprehension o f passive sen­ tences w i t h their performance on O S V actives i n w h i c h the initial object nominal has been topicalized. A s noted earlier, i n passive sentences, children must learn to take the ga-marked nominal to be the patient; this reinterpretation takes place in the context o f the passive inflection on the verb, and, frequently, a ^/-marked nominal somewhere i n the sentence. Thus, there are morphological cues to indicate that ga w i l l not have its typical interpretation as the agent marker i n a transitive sentence. Experimental data from Sano's (1977) study suggest that it may be more difficult to re-interpret a familiar case particle when there are no such overt morphological cues to rely upon. I n Japanese, any argument o f a predicate may be topicalized w i t h the particle wa. W h e n subject or direct object is topicalized, the case particles ga and o are omitted and only wa appears on the surface. ( I n other cases, the original particle is retained and followed by wa.) Sano reports that the children i n her sample tended to treat wa as an alternate subject marker; they understood S O V actives i n w h i c h the subject was marked by wa just as w e l l as sentences i n w h i c h the subject was marked by ga. I f instead an initial object nominal were topicalized, children had tremendous difficulty i n ­ terpreting the sentence. The number o f correct responses on O-waSY actives was lower than other actives across the entire 3;3-6;8 year age range i n Sano's study. Moreover, the responses on these sentences were even lower than on any type o f passive sentence. M o s t probably, this is because N - w a N V sequences, w h i c h are common enough, typically do not have a patient-agent-action interpretation. For example, i n what may be the most common sentence type w i t h this configura-

Clancy

466

tion, wa marks the subject and ga marks what is sometimes analyzed as the underlying object o f a transitive stative predicate, as i n the following: mama

wa

mama

TOP shrimp

ebi

ga

SUBJ

suki.

like

'Mama likes shrimp'. (Lit., 'As for mama, shrimp are likeable'.) In these sentences, wa marks the experiencer, which is semantically somewhat similar to an agent, and ga marks the object o f experience. Thus O S V actives having the semantic sequence patient-agent-action w i t h wa marking the patient force the child to re-interpret, i n fact almost reverse, the meaning o f this familiar configuration o f case particles, although there are no overt morphological cues to trigger a new strategy. The extreme difficulty o f these sentences as compared with passives suggests that it is easier to reorganize processing strategies when there are new local morphological cues which can be taken as signals to switch to the new strategy, as there are i n passive sentences. 7.2.3. Relative Clauses. Experiments on the comprehension o f sentences with relative clauses also indicate that children must undergo reorganizations i n their strategies for sentence comprehension. Several studies i n w h i c h children acted out and/or imitated sentences w i t h relative clauses give insight into the strategies which children follow during this process o f reorganization: Harada et al. (1976), K . I . Harada (1976), Hakuta (1981), Asano (1979). I n Japanese, relative clauses precede the head noun w i t h no relative pronoun. There are only m i n i m a l cues as to the syntactic structure o f sentences w i t h TABLE 4.13 S a m p l e SOV Sentences w i t h Relative Clauses (From Harada et al., 1976) SS:

(kirin giraffe

o DO

taoshita) knocked down

zoo elephant

ga SUBJ

shika deer

o DO

nadeta. patted

'The elephant that knocked down the giraffe patted the deer.' SO:

(zoo elephant

ga SUBJ

taoshita) knocked down

kirin giraffe

ga SUBJ

shika deer

o DO

nadeta. patted

'The giraffe that the elephant knocked down patted the deer.' OS:

zoo elephant

ga SUBJ

(kirin giraffe

o DO

taoshita) knocked down

shika deer

o DO

nadeta. patted

'The elephant patted the deer that knocked down the giraffe.' 00:

zoo elephant

ga SUBJ

(kirin giraffe

ga SUBJ

taoshita) knocked down

shika deer

'The elephant patted the deer that the giraffe knocked down.'

o DO

nadeta. patted

4.

The A c q u i s i t i o n of Japanese

467

TABLE 4.14 Surface Configuration of Case Markers in Complex Sentences w i t h SOV and OSV W o r d Order in the Matrix Clause (From Hakuta, 1981) MATRIX S O V ORDER SS: SO: OS:

00

V ) N-ga (N-ga V ) N-ga N-ga (N-o V ) N-ga (N-ga V )

(N-o

MATRIX O S V ORDER

N-o V

SS:

N-o

(N-o

N-o V

SO:

N-o

(N-ga V )

N-o V

OS:

(N-o

V)

N-o

N-o V

OO:

(N-ga V )

N-o

V)

N-ga N-ga N-ga N-ga

V V V V

relative clauses: the embedded verb can only appear i n plain, not polite, inflec­ tions, the topic marker wa is avoided, and the subject marker ga may be replaced w i t h the genitive particle no. The examples on Table 4.13, taken from Harada et al. (1976), illustrate the formation o f sentences w i t h relative clauses i n Japanese. There are four types o f sentences, those w i t h embeddings on the matrix subject, which also serves as the subject o f the relative clause (SS), those w i t h embed­ dings on the matrix subject, w h i c h also serves as the object o f the relative clause (SO), those w i t h embeddings on the matrix object, w h i c h also serves as the subject o f the relative clause (OS), and those w i t h embeddings on the matrix object, which also serves as the object o f the relative clause ( 0 0 ) . The surface structure o f these sentences, w h i c h have S O V w o r d order i n the matrix clause, is presented on the left side o f Table 4.14. Experimental studies conducted i n English have tested children's ability to process sentences o f these four types, but conflicting results have been obtained in different experiments, such as B r o w n (1971), deVilliers et al. (1979), Sheldon (1974), Smith (1974), and Tavakolian (1981). See Hakuta (1981) for a summary contrasting the results o f these studies. The theories which have been proposed to account for these findings have been based on grammatical and/or perceptual factors, and make different predictions about comprehension. For example, Sheldon (1974) has set forth the "parallel function hypothesis," according to which sentences w i l l be easier to understand i f the relativized noun serves the same grammatical function i n both the matrix and the embedded clauses. A c ­ cording to this hypothesis, SS and O O sentences should be easier to comprehend than OS and SO. A n account i n terms o f perceptual constraints has been pro­ posed by Slobin (1973), i n the f o r m o f the operating principle " a v o i d interrup­ tion or rearrangement o f linguistic u n i t s . " This " a n t i - i n t e r r u p t i o n " principle predicts that sentences i n w h i c h the relative clause is center-embedded, interrupt­ ing the matrix clause, w i l l be more difficult to process. Thus i n English, SS and SO sentences, w h i c h have center-embedding, should be harder to comprehend than the right-branching OS and 0 0 sentences. Since consistent results have not been obtained i n English, Japanese data provide an interesting test o f these hypotheses.

468

Clancy

Results o f experiments i n Japanese indicate that the first types o f embeddings which children can process are left-branching subject embeddings. Using semantically irreversible sentences, K . I . Harada (1976) found that her daughter was able to comprehend and imitate SS sentences correctly at 2;9-2;10 years-of-age. Using semantically reversible sentences, Harada et al. (1976) tested 98 children between 3;6 and 10; 11 years-of-age, and found that children over 5 years-old performed correctly on sentences w i t h subject embeddings, that is, SS and SO sentences, much more frequently than on OS and O O sentences. I n the great majority o f cases, children's errors were based upon failure to recognize a clause boundary between the first and second noun phrase i n the center-embedded OS and O O sentences. Instead, they interpreted these as i f they were conjoined sentences, for example, interpreting the OS sentence given above as i f i t were: T h e elephant knocked d o w n the giraffe and patted the deer'. A s the table o f Japanese surface structures shows, the initial segment o f OS and O O sentences follows the w o r d order o f simple N N V sentences; the children processed the initial N N V sequence as i f i t were an independent clause. This was more frequent in the case o f OS than O O sentences, since OS sentences m i m i c not only the N N V w o r d order o f a simple sentence, but also the N-ga N-o V sequence o f particles i n a standard S O V active sentence. Harada et al. conclude that these findings are consistent w i t h Slobin's anti-interruption principle, and also w i t h the "juxtaposition hypothesis," namely, that complex sentences are at first pro­ cessed as i f they were juxtaposed clauses. (See Tavakolian's "conjoined-clause analysis," 1981.) Apparently, one way o f avoiding interruption is to apply a conjoined-clause i n t e r p r e t a t i o n . 30

T w o further studies have confirmed these findings w i t h statistical analyses. Asano (1979) tested 45 Japanese children aged 2;11 to 6;7 years, and found that children over 4;6 years o l d correctly acted out sentences w i t h left-branching relative clauses, SS and SO, significantly more often than center-embedded OS

I n Japanese, the occurrence of ellipsis, especially of subject ellipsis, means that even complex sentences which do not begin with an NNV sequence may be interpreted as conjoined. Thus sen­ tences with the configuration NVNNV may be analyzed as conjoined clauses of the form NV-NNV, with ellipsis in the first rather than the second conjunct. Taking into consideration the naturalness of this kind of ellipsis and of the resulting configuration of case particles, SS/SOV sentences seem the most likely candidates for this analysis. Thus the SS/SOV sentence from Table 4.13, Kirin o taoshita zoo ga shika o nadeta 'The elephant that knocked down the giraffe patted the deer', may receive a conjoined-clause interpretation, meaning approximately 'Having knocked down the giraffe, the elephant patted the deer'. This interpretation will yield the correct assignment of nominal arguments to predicates, and is probably responsible for the superior performance on SS/SOV sentences which has been noted in the Japanese studies of relative clause processing. It would be difficult to formulate predictions based on the application of a conjoined NV-NNV analysis to other sentence types, however, since certain combinations of first conjunct ellipsis and case particles are quite unnatural, and probably do not occur frequently enough in the input to children to serve as the basis for development of a conjoined-clause processing strategy which could be applied to sentences with relative clauses. 3 0

4.

The A c q u i s i t i o n of Japanese

469

and O O sentences. Hakuta (1981) tested 12 children between 5;3 and 6;2 years old and also found significantly better performance on left-branching than centerembedded sentences. A g a i n , there is support for the anti-interruption and j u x ­ taposition hypotheses, whereas the superior performance on SO than O O sen­ tences runs counter to Sheldon's parallel function hypothesis. These findings leave open the possibility that grammatical role is the crucial factor i n comprehension, since subject relatives i n Japanese are consistently easier to process than object relatives. Since both S O V and O S V w o r d orders are permitted i n Japanese, it is possible to analyze the effects o f grammatical role and center-embedding separately. Table 4.14 above, right side, summarizes the surface structure configuration o f sentences w i t h relative clauses which have OSV w o r d order i n the matrix clause. A s the table shows, i n these sentences, it is the subject relatives w h i c h are center-embedded and the object relatives w h i c h are left-branching. Harada et al. (1976), Asano (1979), and Hakuta (1981) have all investigated the role o f grammatical vs. perceptual factors i n experiments using sentences w i t h O S V matrix order. A g a i n , the results o f these studies show that i t is the sentences w i t h center-embedding w h i c h are the most difficult; i n each experiment, OS and O O sentences were correctly interpreted significantly more often than SS and SO sentences. A g a i n , Slobin's perceptual anti-interrup­ tion principle, rather than an explanation based on grammatical role, is sup­ ported. The authors o f these three investigations conclude that Slobin's proposal is the only one to date which can claim crosslinguistic validity. The processing difficulty w h i c h sentences w i t h relative clauses pose for the Japanese child can be understood by examining the sequences o f case particles which occur i n these sentences. Hakuta has proposed that i n interpreting sen­ tences w i t h relative clauses, i t is the " s t a c k i n g " o f case particles which children find difficult, rather than center-embeddedness per se. I n SOV sentences, O O embeddings create a sequence o f t w o ga-marked nouns; i n O S V sentences, SS embeddings create a sequence o f t w o o marked nouns. Asano reports that some children over A h years-of-age i n her study, when trying to act out O O sentences w i t h matrix S O V order, w o u l d point out that they could not figure out which toy should perform the action because there were t w o ga's i n a r o w . The lack o f particle stacking i n SS/SOV sentences probably played a part i n K . I . Harada's (1976) finding that her daughter acquired SS sentences before SO sentences. I n SOV matrix order, only SS sentences allow the child to process constituents as they are heard, without being confused by repeated particles, or sequences o f particles w h i c h m i m i c simple S O V sentences. x

Hakuta (1981) investigated the effects o f stacking i n an experiment i n w h i c h the second matrix noun o f each stimulus sentence was postposed. This un-stacks the particles i n complex sentences so that no t w o nominals w i t h the same casemarking occur i n immediate succession. Table 4.15, w h i c h presents configura­ tions o f surface structures from Hakuta's experiment, illustrates how postposing un-stacks particles; i n these cases, the second matrix noun w i t h its embedded

470

Clancy TABLE 4.15 Surface Configuration of Case Markers in Complex Sentences W i t h and W i t h o u t Postposing (From Hakuta, 1981) STANDARD

ss/osv oo/sov

(N-o N-ga (N-ga

N-o

V)

N-ga

V ) N-o

POSTPOSED V ) N-ga (N-ga V ) N - o

V

N-o

V , (N-o

V

N-ga

V,

clause, is postposed. Hakuta analyzed the performance o f 5;4-6;3-year-old children on SS/SOV, S S / O S V , O S / S O V , and O S / O S V sentences i n both stan­ dard and postposed conditions. Comprehension o f the center-embedded struc­ tures O S / S O V and SS/OSV was significantly better i n the postposed condition, in which the particles were un-stacked. Hakuta concludes that i t is the stacking o f case particles which makes sentences w i t h center embeddings difficult for Jap­ anese children to process, leading them to make erroneous assignments o f nouns to verbs i n sentences o f the configuration N ( N V ) N V . Thus between the ages o f 5 and 10 years, Japanese children reorganize their strategies for processing complex sentences, abandoning early attempts to pro­ cess them as simple, conjoined sentences. This reorganization probably involves learning to recognize that successive noun phrases marked by the same case particle may be a local cue for embedding, and that finite verbs inflected for tense which do not occur at a clause-final intonation contour are always a sign o f embedding. This reorganization must await development o f short-term memory to the point where the assignment o f nouns to verbs can be suspended until t w o clauses have been stored.

THE SETTING OF L A N G U A G E ACQUISITION 8. C o g n i t i v e P a c e s e t t i n g o f L i n g u i s t i c D e v e l o p m e n t 8.1.

Locatives

One type o f evidence for the cognitive pacesetting o f linguistic development can be found by comparing the order i n w h i c h similar concepts emerge across different languages. Johnston and Slobin (1979) have analyzed the acquisition o f locative expressions by children acquiring English, Italian, Serbo-Croatian, and Turkish, and found the f o l l o w i n g order o f acquisition i n the four languages: I N , ON, UNDER, BESIDE < B A C K , FRONT , BETWEEN < B A C K , FRONT. (Objects w i t h inherent fronts and backs, such as trucks and houses, were called "featured objects.") This order o f acquisition, the authors propose, reflects both cognitive complexity and language-specific linguistic complexities. f

f

Although Japanese data on this topic are rather sparse and unsystematic, there is some evidence that Japanese children acquire locatives i n a stable sequence.

4.

The Acquisition of Japanese

471

M y o w n data on one child between l ; l l - 2 ; 4 years-of-age, as w e l l as Miyahara (1974) and Okubo (1963, 1967, Chap. 2) provide some information on the order of emergence o f case particles, including locatives. Usually the first locative to be acquired is ni, w h i c h covers the semantic range o f English in and on, as w e l l as at and to. Shortly afterwards i n m y data, but at the same time i n some children (Okubo, 1967), the directional e ' t o / t o w a r d ' begins to be used. I n my data, the next locative to emerge was de, which marks the place at which an action is being performed. The next t w o locatives which appear are made ' u n t i l / u p t o ' and kara ' f r o m ' . Kara emerged first i n m y data, Miyahara's c h i l d , and one child cited by Okubo; these t w o locatives were acquired i n reverse order by Okubo's daughter. I n m y data, after several simple postpositions had emerged, locatives w h i c h are linguistically and cognitively more complex began to appear, such as the Japanese equivalents o f 'beside', 'inside', etc. These expressions are formed by using the genitive marker no plus a series o f nouns referring to specific locations, such as ue ' t o p ' , naka ' m i d d l e , inside', shita ' b o t t o m ' , soba 'side', and tokoro 'place', followed by the locative ni, as i n : N

no

N

G E N bottom L O C

shita

ni

'under N \ Since these forms are constructed i n a manner quite similar to the corresponding Turkish locatives, Japanese w o u l d provide an interesting test case for the course of acquisition proposed by Johnston and Slobin (1979). Unfortunately, systemat­ ic data on the acquisition o f these expressions by Japanese children are not yet available. 8.2.

Conjunctions

On the basis o f a variety o f longitudinal and cross-sectional data from En­ glish, German, Italian, and T u r k i s h , Clancy et al. (1976) hypothesized that there is a consistent, cognitively based order o f emergence o f the notions underlying sentence conjunction across languages. The authors found that the earliest rela­ tions between sentences included notions o f symmetric coordination, antith­ e s i s , sequence, and causality. N e x t , conditional notions emerged, followed by 31

C l a n c y et al. (1976) categorize as antitheses cases in which the child rejects one proposition and asserts an alternative, as in No mangiae gioco io 'no eat I play', from an Italian child of 1;6 years. Since the authors were concerned with the meaning relations between clauses rather than specific conjunctions, they did not analyze the emergence of other types of adversative relations, which appear later. In contrast, Bloom et al. (1980) did not analyze such antitheses, which they classified as "anaphoric negation." In their English data, adversative relations were late in emerging, and the connective but did not appear until 2;8-3;2 years-of-age. Okubo (1967) does not cite examples of antithesis; her data on the emergence of explicit adversative connectives in Japanese are consistent with the findings of Bloom et al. 3

472

Clancy

conditional and temporal uses o f conjunctions corresponding to English when. The last development was the acquisition o f before and after. Using detailed longitudinal data from four English-speaking children, B l o o m et al. (1980) also found a consistent order o f emergence o f semantic relations i n conjoined sen­ tences (considering only those w i t h explicit connectives): additive < temporal < causal < adversative. The first conjunction acquired by all the children was and; the other connectives, however, emerged i n different orders i n the different children. A t present, there is comparatively little longitudinal information available on the order o f emergence o f conjunctions i n Japanese. However, the longitudinal data i n Okubo (1967, Chap. 2) and Fujiwara (1977), as w e l l as m y o w n sample, are generally consistent w i t h the crosslinguistic findings. The first conjunction acquired by the Japanese 2-year-olds i n m y sample, and by Okubo's daughter, was -te, w h i c h , like English and, covers the broadest semantic range o f any connective. I n adult speech, -te 'and/then/so' includes the meanings o f coordi­ nation, temporal sequence, and causality, as w e l l as instrumental and manner relations between clauses. I n m y data, the earliest examples o f -te occurred i n sentences conjoining a main action w i t h a clause expressing the manner or means i n w h i c h the action was performed. For example, describing how to place one piece o f a game on top of another, a child o f 2; 1 years said: koo

shi - te

nose

like this

do

put on

and

- ru

NONPAST

no.

EP

'You put it on like this'. These sentences seem semantically similar to certain English sentences w i t h object complements cited by L i m b e r (1973), such as, / show you how to do it, which he found at about the same time as the earliest conjoined sentences. Other examples o f the earliest uses o f -te, such as the f o l l o w i n g sentence from Fujiwara (1977, p. 169), are instrumental: basu

n ( = ni)

bus

on

not - te

ride

and

kaer

return (home)

- u

NONPAST

no.

2;0 yrs

EP

F

'I'll go home by bus'. Apparently, certain o f the earliest functions o f conjoined sentences w i t h -te i n Japanese w o u l d not be expressed by conjoined sentences i n languages such as English, making it difficult to compare the t i m i n g o f their acquisition. Other early instances o f -te express the kinds o f temporal and causal rela­ tionships which can be found among the early uses o f English and, and the first conjunctions i n other languages. For example, Fujiwara (p. 121) cites the f o l l o w ­ ing case, which is strongly causal/temporal as w e l l as instrumental:

4. ocha

non

tea

drink

- de

and

naor

T h e A c q u i s i t i o n of Japanese

473

- u.

recover

NONPAST

'I'll drink tea and recover'. This sentence was produced b y a child o f 1;10, who had eaten something sour. Such use o f -te to express sequence and causality is consistent w i t h the findings of B l o o m et al. (1980), w h o report that and was used for temporal and causal meanings before any other conjunctions were acquired. Shortly after the emergence o f -te, kara 'because' also begins to appear i n conjoined sentences. A s discussed earlier i n section 6.1.5, both -te and -temo 'even though' may occur at this stage to express permission and -tara ' w h e n / i f to express prohibition. Ignoring these early uses, the next conjunction to emerge in Okubo's data and mine was -tara ' w h e n / i f expressing temporal sequence and conditionality, as i n the f o l l o w i n g example from Okubo at 2;1 years: 32

kaze

hii

cold

catch

- tara

if

komar

- u

trouble

deshoo.

NONPAST

COP:PRESUM

'It will be too bad if I catch a cold'. This initial phase o f development, w i t h -te, kara, and -tara, typically occurs during the first months o f the child's third year, or somewhat earlier. The meanings expressed—manner, sequence, and causality—are similar to those found i n other languages at this stage. The lack o f coordination, or " a d d i t i v e " meanings w i t h -te, w h i c h does serve this function i n adult speech, is unexpected. Perhaps symmetric coordination w i t h -te is not very frequent i n Japanese com­ pared w i t h other languages, or perhaps this negative finding simply reflects the paucity o f data on Japanese connectives at this point. The use o f -tara i n condi­ tionals is precocious compared w i t h English and other Indo-European languages, but such examples do occur very early i n the more typologically similar Turkish (see Clancy et a l . , 1976). The lack o f adversative connectives (other than -temo 'even though' i n permission sentences) is also consistent w i t h the findings o f B l o o m et al. The next development w h i c h Okubo reports is the emergence o f -tari, the alternative or representative inflection (see Jorden, 1963, Part 2, 27.1) used i n presenting a series o f similar actions or states. I t is like additive and i n English, except that it strongly implies 'and so o n ' . The earliest examples cited by Okubo use the same verb, as i n : omame

ka

bean

buy

- i

ni

INF to

it

go

- tari,

nanka

R E P something

kat - tari,

buy R E P

0 n e child in my sample was using kara 'because' at the end of single clauses even before the first spontaneous examples of S-te S 'S and S \ As discussed in section 12.1.4, kara typically appears on single clauses before being acquired as a conjunction between two clauses. 32

474

Clancy

oume

kat

plum

buy

- tari,

jidoosha

kat

R E P car

- tari

buy

su - ru

REP do

NONPAST

no

yo.

EP

EMPH

T will go to buy beans, and buy something, and buy plums, and buy a car'. Thus this child's first use o f -tari is quite similar to English and as a simple coordinator o f interchangeable actions. The late emergence o f -tari reflects its infrequency and greater semantic complexity compared w i t h and. -Tari is also more difficult grammatically; it is typically used i n a series o f clauses, each marked by -tari, w i t h sum ' d o ' as the main, tensed verb at the end o f the list, as in the above example. I n the next stage o f development, more sophisticated temporal connectives are acquired, including to w h e n / w h e n e v e r / i f , which expresses both sequence and conditionality, and toki, a noun meaning ' t i m e ' w h i c h functions as a conjunction similar to English when. The earliest use o f toki which Okubo (p. 228) cites at 2;3 years-of-age expressed both temporal sequence and conditionality, as do the early uses o f when i n English and other Indo-European languages (see Clancy et a l . , 1976): 4

mata

kureyon

na

again

crayon

not exist

- ku

nat

- ta

become

PAST

toki

kat

time

buy

- te

ne.

IMP AGR

'When the crayon is used up, buy (one) again, alright'. The earliest use o f to at 2;5 years was similar: Yachiyo

( = Yachiko)

Yachiko naka - na

cry

mama

soba

ne

mama

side

sleep

- i

- teru

to

PRES. PROG

when

no.

N E G NONPAST

EP

'When mama sleeps beside me, I don't cry'. A very late acquisition was -nagara ' w h i l e ' , at 3;8 years. Thus, the sequence o f development for temporal connectives is generally consistent w i t h the crosslinguistic finding that ' w h e n ' appears first i n temporal sequences and condi­ tionals, w i t h later acquisition o f conjunctions expressing simultaneity, such as 'while'. The adversative connectives noni 'although' and kedo 'but' emerged at 2;4 and 2;5, respectively, Okubo reports. The examples cited indicate that both were used to express semantic opposition and/or denial o f expectation, as i n : Yachiyo

Yachiko

( = Yachiko)

takusan

tabe - ru

noni

a lot

eat

although

NONPAST

akachan

baby

4. tabe - na

eat

- i

N E G NONPAST

The A c q u i s i t i o n of Japanese

475

no.

EP

'Although Yachiko eats a lot, the baby doesn't eat'. The emergence o f adversative connectives after coordination, sequence, and causality is consistent w i t h the findings o f B l o o m et al. (1980) on English. Although the data are sparse and somewhat inconsistent (Okubo cites various differences i n the order o f emergence o f conjunctions i n another child), the development o f these clause-final connectives i n Japanese seems similar enough to findings from other languages to warrant further investigation from the stand­ point o f cognitive universals. (See, also, Okubo, 1983, for further information on the acquisition o f conjunctions.) It is interesting that the order o f emergence o f the optional, clause-initial connectives i n Okubo's daughter followed the same semantic sequence, at a somewhat later age. The first set o f clause-initial conjunctions to be acquired were those expressing temporal sequence, soshite (2;2 years) and soshitara (2;3) 'and then'. Next came sorede (2;4), w h i c h implies causal and temporal se­ quence. A t 2;7 years, dakara 'and so/therefore' appeared. The conditional sorenara ' i f so' emerged at 3;2, and at 3;5 the temporal/conditional suruto. The last group o f clause-initial connectives was the adversatives: soredemo 'even so' at 3;5 and dakedo and demo ' b u t ' at 4;0 (Okubo, 1967, Chap. 1.4.4). A g a i n , this parallels the sequence found by B l o o m et a l . , w i t h temporal, causal, and adver­ sative conjunctions emerging i n that order. Thus the development o f clauseinitial connectives i n Japanese provides further evidence for cognitive pacesetting o f the acquisition o f conjunctions. 8.3.

Questions

For many years, crosslinguistic evidence has been accumulating for a cognitively based order o f emergence o f question types. Studies o f English-speaking children, such as Ervin-Tripp (1970) and Tyack and Ingram (1977) have sug­ gested the f o l l o w i n g sequence o f development: what < where < why < how < when. Ervin-Tripp proposes the sequence what do < whose < who between where and why; Tyack and Ingram suggest a later acquisition for who. Using detailed longitudinal data from Serbo-Croatian, Savic (1975) divided her t w i n s ' acquisition o f question words into four stages. Each child followed the general sequence: I what < I I where < I I I who < how < I V why < which < when, w i t h what kind and whose appearing i n Stage I I I i n one child and stage I V i n the other. One child acquired how much i n stage I V . Fortunately, there are detailed longitudinal data available from Japanese for comparison. Okubo (1965, 1967) provides a very complete description o f her daughter's acquisition o f different types o f questions over the course o f more than 3 years. I n acquiring different types o f questions, i t is clear that this child

476

Clancy TABLE 4.16 Order o f Emergence of Question Forms in One Child's Speech (Based on Okubo, 1967, p. 167) nani doko nani shiteru dare dorè doo donna dooshite dolchi {dare no naze ikura dono itsu

i;8

1;11 2;1 2;3 2;5 2;6

a

3;0 4;0 4;10

'what' 'where' 'what doing' 'who' 'which' (of more than two) 'how' 'what kind of 'why' (can also mean 'how') 'which' (of two) 'whose') 'why' (more formal, less frequent than dooshite in speech to young children) 'how much' 'which' (ADJ) 'when'

This form is not from Okubo's daughter, but occurred in other data (Okubo, personal communication). a

followed a developmental sequence similar to that w h i c h has been found i n other languages. Table 4.16 gives the age and order o f emergence o f these forms i n Okubo's daughter. O n the basis o f data from several children, Murata (1968, p . 145) reports that the first question asked is nani ' w h a t ' , w h i c h typically appears around 1 ;9— 1; 11 years-of-age. Iwabuchi and Muraishi (1968), examining the emergence o f a more limited set o f questions i n the speech o f one c h i l d , found a somewhat different sequence o f development from Okubo. The child i n their study used nani ' w h a t ' , doko 'where', and dare ' w h o ' at 1;8 years-of-age, and used doo ' h o w ' as early as 1;7 years. However, doo ' h o w ' was apparently not used again after its first occurrence until the child was 2;0-years-old. Dore ' w h i c h ' emerged at 1;9, dotchi ' w h i c h ' at 1;11, and donna 'what k i n d ' at 2;8 years. Obviously, further research w i l l be necessary to determine the extent o f individual differences i n the acquisition o f different question types i n Japanese. However, i n general, the Japanese data do f i t the crosslinguistic order o f emergence quite w e l l . The data also suggest that there may be language-specific, as w e l l as individual, differences i n the emergence o f question types. Thus i n English why seems to precede how, whereas i n Japanese (as i n Serbo-Croatian), the reverse order holds. A l s o , although Ervin-Tripp proposes the sequence whose < who for English, the Japanese data show a clear lag between who and whose. As Okubo has pointed out (personal communication), i n Japanese dare no 'whose' is morphologically complex, combining the question form dare and the genitive case particle no. Therefore, it seems unlikely that a Japanese child would acquire dare no 'whose' before the simple dare ' w h o ' . A l t h o u g h data from a larger number o f Japanese children w o u l d be necessary before any defini-

4.

The A c q u i s i t i o n of Japanese

481

rather violently against the social behavior characteristic o f atashi, preferring to identify herself as boku, and allowed to engage i n the loud, active behavior which kindergarten boys enjoy. Despite discussions and arguments such as the above, which certainly brought the issue to her conscious attention, this g i r l used boku very frequently and naturally i n her conversation, just as the other child used atashi. Clearly, she was not just using boku to get a reaction from adults, but had acquired the form instead o f atashi as her ordinary means o f selfreference. Use o f boku by females sometimes continues w e l l past the age o f primary language acquisition. I have observed girls i n the third grade refer to themselves as boku when playing active games w i t h the boys. There is much anecdotal evidence that girls o f even high school or college age refer to themselves as boku, usually among friends. Very rarely, a woman regularly refers to herself as boku; this is interpreted as a personal mannerism and/or an indication o f pro-liberation sentiments. I t w i l l be interesting to see whether the pronoun boku gradually yields to this pressure from discontent females, who are reluctant to be limited to atashi. The current situation may reflect an early stage o f linguistic change, which may lead to the addition o f another first person option for Japanese women, who now have a more limited range o f accepted linguistic (and social) choices than men.

10. I n d i v i d u a l D i f f e r e n c e s Although individual differences have not been a major focus o f Japanese re­ search i n language acquisition, there are many examples o f such differences i n the literature. Unfortunately, it is usually very difficult to determine whether these are due to input differences, subtle linguistic differences between the examples being cited, as i n the area o f case particles, or differences i n linguistic level across children o f similar age. However, there are a few studies document­ ing different acquisition strategies used by individual children; recent papers, such as Okubo (1980, 1981a) and K o m u r a (1981) may reflect an increasing interest i n this topic among Japanese scholars. 1 0 . 1 . Noun-Dominant

vs. Verb-Dominant

Children

Okubo (1981a) proposes that there are t w o main strategies w h i c h children follow i n the transition process from the one-word to the t w o - w o r d stage. This fundamental difference can be seen i n the relative proportion o f nouns to verbs i n the speech o f children during this period. The data for Okubo's study consist o f approximately 14 hours o f speech recorded during the course o f a typical day from t w o siblings when they reached 2 years-of-age. The motivation for the comparative analysis came from Okubo's observation that the language develop­ ment o f a boy she had been studying was quite different from that o f his younger sister.

482

Clancy

The words o f each child were divided into 15 parts o f speech; nouns were the most frequent type for both children. However, the boy was clearly " n o u n dominant" compared to his sister. I n a type/token analysis, 72% o f all his vocabulary items were nouns, and 4 5 . 5 % o f the total number o f words he used i n the sample were nouns. I n contrast, only 58.0% o f his younger sister's vocabu­ lary consisted o f nouns, and only 38.3% o f the total number o f words i n her sample were nouns. The comparative frequencies o f verbs were also strikingly different; the " v e r b - d o m i n a n t " girl used a total o f 23.9% verbs, whereas the boy used only 5.2%. Thus when the g i r l wanted to be read a storybook, she w o u l d say Yonde 'Read', using the imperative inflection; her older brother at that stage made the same request w i t h the babytalk nominal Jiji 'Characters/writing'. Thus the same situation can be expressed using either a noun or a verb. I t is interesting to note that Tanouye (1979) has found Japanese children to be verb-dominant compared w i t h American c h i l d r e n . Language-specific patterns o f encoding situations and structural factors such as w o r d order may cause languages to differ along the dimension o f noun vs. verb dominance, w i t h individual differences existing w i t h i n the range for any particular language. 33

These different strategies had implications for the acquisition o f various other linguistic features. The g i r l , for example, had a greater variety o f verbal inflec­ tions, which she used more frequently; she was also more advanced w i t h respect to sentence-final particles, which typically follow the verb i n adult Japanese. The boy, although more limited i n the predicate position o f his sentences, had more complex nominal arguments, including modifiers such as color terms. These differences i n noun vs. verb dominance also had interesting effects upon the nature o f early t w o - w o r d utterances i n the children's speech. Okubo (1980) reports that the verb-dominant girl tended to use kore 'this' and people's names as fixed " p i v o t s " i n her t w o - w o r d utterances, w i t h a wide range o f predicates and sentence-final particles. I n contrast, the noun-dominant boy used a few simple predicates, such as nai 'does not exist' and akai 'red' as " p i v o t s , " and predicated these o f a wide range o f different nouns. Thus the patterns established during the one-word period continued into the next stage, and pro­ vided each child w i t h a different means o f breaking into the creation o f produc­ tive t w o - w o r d combinations. These findings are quite similar to those made by

Tanouye (1979) analyzes the utterances of two children acquiring Japanese, and gives the percentage of "function," "action," and "substantive" forms. Although various studies of En­ glish-speaking children have found that verbs are acquired later than nouns, Tanouye found that at the one-word stage (MLU 1.1 for one child, 1.08 for the other), only 16% and 22%, respectively, of all utterances were substantives. Furthermore, in Japanese, the "function" of existence is expressed by the verbs irularu 'exist', and non-existence by the adjective nai. Also, in contrast with English, the great majority (94%) of all "action" forms in Japanese were verbs. Thus there is a very high frequency of verbs in early Japanese language development. (At a later stage, the frequency of substantives in the children in Tanouye's sample increased to a level comparable with the children of 2;0 in Okubo's study.) 33

4.

The A c q u i s i t i o n of Japanese

479

but also i n how people refer to themselves. Therefore, acquisition o f this system w i l l probably affect the child's developing sense o f identity, especially i n the area o f social and sexual roles. Ide concurs w i t h the view that "personal identity is more consistent i n the case o f speakers o f Indo-European languages, because T is T wherever one is and whomever one is talking to, whereas i n Japanese personal identity is flexible and dependent on the position o f speaker i n relation to hearer w i t h i n a given s i t u a t i o n . " The Japanese system o f self-reference, by allowing different options, lets a person function i n several different roles, and the existence o f a special linguistic form for a particular range o f social and emotional situations probably delineates each persona for the speaker. Since the form o f self-reference chosen depends i n part upon the addressee, there is i n Japanese a very literal identification o f oneself i n terms o f others. Based on her findings on the forms o f self-reference used by children from 1 Vi to 6 years-of-age, Ide (1977, 1978) analyzes how the Japanese system o f pro­ nouns shapes the child's developing sense o f identity. Children's earliest sense o f self, Ide assumes, is expressed i n the forms which they first hear and use i n referring to themselves: NAME + -chart for boys and girls, and also boku-chan T for boys. (Recall that Japanese adults use boku T as a form o f second person reference i n speaking to little boys, saying, for example, 'Does boku want to play?'). Ide considers the first developments away from these basic forms o f selfreference to be indications o f how the child's sense o f identity develops, shaped by the available linguistic options. I n her data, for little boys the next form o f self-reference to develop was ore, which is based upon awareness o f both speak­ er and hearer as masculine, emphasizes solidarity w i t h a same-sex peer group, and occurs frequently i n contexts where the child is swaggering and showing off, or involved i n active play. For little girls the next form o f self-reference to be acquired was (w)atashi, w h i c h reflects a formal attitude, at least compared w i t h ore, and/or an affectation o f maturity, as w e l l as an awareness o f themselves as female. Girls referred to themselves as oneechan 'older sister' outside the fami­ l y , whereas boys d i d not extend oniichan 'older brother' i n this way i n Ide's data. Thus the familial, nurturing role was emphasized i n the system o f selfreference for girls. The affect and behavior associated w i t h the pronouns girls and boys acquire is quite different: atashi is mature and well-behaved, ore is rough and rather w i l d . Ide points out that these developments away from the child's initial forms o f self-reference lengthen the psychological distance be­ tween speaker and hearer for girls and shorten i t for boys, since the girls' option, (w)atashi, is more formal than ore, w h i c h emphasizes a boy's solidarity w i t h his peers. A t first, both boys and girls may occasionally use inappropriate first person pronouns w h i c h they hear frequently i n their surroundings. For example, Horiguchi (1979a) reports that when they were 2-years-old, her son and daughter each sometimes used the first person pronoun appropriate for the opposite sex; thus, hearing her older brother use ore, the daughter imitated his usage, although only very rarely. Beyond such early imitations, especially after becoming aware

480

Clancy

of the sex differences governing correct usage at about 3 years-of-age, children usually stick to the first person pronoun appropriate for their sex. Ide (1977) has suggested that the Japanese system o f self-reference plays an important role i n supporting and reinforcing stereotyped notions o f male and female i n Japanese society, since acquisition o f this system fosters a definition o f oneself i n terms o f these notions. What happens to the child w h o finds, i n acquiring this system, that the linguistic options available do not coincide w i t h his or her favorite persona? There is evidence that children are aware o f the implications o f the available forms o f self-reference for the shaping o f one's identity and behavior, namely, the existence o f children w h o resist acquiring sexappropriate forms o f self-reference or who deliberately choose to use the first person pronoun o f the opposite sex, at least on occasion. I t is interesting that this seems to happen mostly w i t h girls, who sometimes prefer to refer to themselves as boku, the most common first person pronoun for boys, rather than using the feminine (w)atashi. Ide (1977) found that girls o f 5 - 6 years-of-age used boku to draw attention to themselves, especially when talking to their mothers. I n m y data, there was a much more extreme case, a little girl o f 3V2 years w h o steadfastly insisted upon calling herself boku or boku-chan all the time, and not just for special effect. This g i r l had no brothers, just a very ladylike older sister; she herself was a tomboy w i t h many friends among the little boys at kinder­ garten. The f o l l o w i n g dialogue was part o f a continuing struggle i n w h i c h her mother tried to get her to refer to herself as (w)atashi (atashi is a less formal variant, used only by women). Mother:

w atashi. boku.

Child: Mother:

watashi

I

deshoo.

COP:PRESUM

'It should be Child:

boku

watashV.

deshoo.

'It should be

boku\

As this example suggests, the use o f appropriate male and female pronouns for self-reference is one o f the few areas o f grammar i n w h i c h Japanese parents correct their children's errors. Fujiwara (1977, p . 151) also cites a case i n w h i c h a girl o f 1; 11 years called herself boku and was immediately corrected by her father. The other 3-year-old girl i n m y sample used atashi very frequently, i n fact, somewhat more often than an adult probably w o u l d . This child apparently enjoyed being atashi and liked to carry a toy pocketbook, imitate her mother's speech, etc. I n contrast, the girl o f the above example seemed to be rebelling

4.

The A c q u i s i t i o n of Japanese

481

rather violently against the social behavior characteristic o f atashi, preferring to identify herself as boku, and allowed to engage i n the loud, active behavior which kindergarten boys enjoy. Despite discussions and arguments such as the above, w h i c h certainly brought the issue to her conscious attention, this g i r l used boku very frequently and naturally i n her conversation, just as the other child used atashi. Clearly, she was not just using boku to get a reaction from adults, but had acquired the form instead o f atashi as her ordinary means o f selfreference. Use o f boku by females sometimes continues w e l l past the age o f primary language acquisition. I have observed girls i n the third grade refer to themselves as boku when playing active games w i t h the boys. There is much anecdotal evidence that girls o f even high school or college age refer to themselves as boku, usually among friends. Very rarely, a woman regularly refers to herself as boku; this is interpreted as a personal mannerism and/or an indication o f pro-liberation sentiments. I t w i l l be interesting to see whether the pronoun boku gradually yields to this pressure from discontent females, w h o are reluctant to be limited to atashi. The current situation may reflect an early stage o f linguistic change, which may lead to the addition o f another first person option for Japanese women, who now have a more limited range o f accepted linguistic (and social) choices than men.

10. I n d i v i d u a l D i f f e r e n c e s Although individual differences have not been a major focus o f Japanese re­ search i n language acquisition, there are many examples o f such differences i n the literature. Unfortunately, it is usually very difficult to determine whether these are due to input differences, subtle linguistic differences between the examples being cited, as i n the area o f case particles, or differences i n linguistic level across children o f similar age. However, there are a few studies document­ ing different acquisition strategies used by individual children; recent papers, such as Okubo (1980, 1981a) and K o m u r a (1981) may reflect an increasing interest i n this topic among Japanese scholars. 1 0 . 1 . Noun-Dominant

vs. Verb-Dominant

Children

Okubo (1981a) proposes that there are t w o main strategies w h i c h children follow i n the transition process from the one-word to the t w o - w o r d stage. This fundamental difference can be seen i n the relative proportion o f nouns to verbs i n the speech o f children during this period. The data for Okubo's study consist o f approximately 14 hours o f speech recorded during the course o f a typical day from t w o siblings when they reached 2 years-of-age. The motivation for the comparative analysis came from Okubo's observation that the language develop­ ment o f a boy she had been studying was quite different from that o f his younger sister.

482

Clancy

The words o f each child were divided into 15 parts o f speech; nouns were the most frequent type for both children. However, the boy was clearly " n o u n dominant" compared to his sister. I n a type/token analysis, 72% o f all his vocabulary items were nouns, and 4 5 . 5 % o f the total number o f words he used i n the sample were nouns. I n contrast, only 58.0% o f his younger sister's vocabu­ lary consisted o f nouns, and only 38.3% o f the total number o f words i n her sample were nouns. The comparative frequencies o f verbs were also strikingly different; the " v e r b - d o m i n a n t " girl used a total o f 23.9% verbs, whereas the boy used only 5.2%. Thus when the g i r l wanted to be read a storybook, she w o u l d say Yonde 'Read', using the imperative inflection; her older brother at that stage made the same request w i t h the babytalk nominal Jiji 'Characters/writing'. Thus the same situation can be expressed using either a noun or a verb. I t is interesting to note that Tanouye (1979) has found Japanese children to be verb-dominant compared w i t h American c h i l d r e n . Language-specific patterns o f encoding situations and structural factors such as w o r d order may cause languages to differ along the dimension o f noun vs. verb dominance, w i t h individual differences existing w i t h i n the range for any particular language. 33

These different strategies had implications for the acquisition o f various other linguistic features. The g i r l , for example, had a greater variety o f verbal inflec­ tions, which she used more frequently; she was also more advanced w i t h respect to sentence-final particles, which typically follow the verb i n adult Japanese. The boy, although more limited i n the predicate position o f his sentences, had more complex nominal arguments, including modifiers such as color terms. These differences i n noun vs. verb dominance also had interesting effects upon the nature o f early t w o - w o r d utterances i n the children's speech. Okubo (1980) reports that the verb-dominant girl tended to use kore 'this' and people's names as fixed " p i v o t s " i n her t w o - w o r d utterances, w i t h a wide range o f predicates and sentence-final particles. I n contrast, the noun-dominant boy used a few simple predicates, such as nai 'does not exist' and akai 'red' as " p i v o t s , " and predicated these o f a wide range o f different nouns. Thus the patterns established during the one-word period continued into the next stage, and pro­ vided each child w i t h a different means o f breaking into the creation o f produc­ tive t w o - w o r d combinations. These findings are quite similar to those made by

Tanouye (1979) analyzes the utterances of two children acquiring Japanese, and gives the percentage of "function," "action," and "substantive" forms. Although various studies of En­ glish-speaking children have found that verbs are acquired later than nouns, Tanouye found that at the one-word stage (MLU 1.1 for one child, 1.08 for the other), only 16% and 22%, respectively, of all utterances were substantives. Furthermore, in Japanese, the "function" of existence is expressed by the verbs irularu 'exist', and non-existence by the adjective nai. Also, in contrast with English, the great majority (94%) of all "action" forms in Japanese were verbs. Thus there is a very high frequency of verbs in early Japanese language development. (At a later stage, the frequency of substantives in the children in Tanouye's sample increased to a level comparable with the children of 2;0 in Okubo's study.) 33

4.

The A c q u i s i t i o n of Japanese

483

Lieven (1980), w h o has also investigated individual differences and the con­ tinuity between one- and t w o - w o r d stages, using data from three children acquir­ ing English. Certain differences between the t w o children i n Okubo's study may have been the result o f birth order. For example, w i t h i n the category o f nouns, the boy more frequently labeled objects and their colors; i n contrast, the girl had a much higher frequency o f references to people. This suggests an early pattern o f mother-child interaction focusing on objects w i t h the first c h i l d , but more social contacts w i t h the second; such a difference could obviously result from the presence o f another child and his friends. Similarly, the girl frequently used the quotative particle tte as a sentence-final form; at 2 years, her brother d i d not use this particle even once. As Okubo points out, the older brother probably d i d not have as much opportunity to report the speech o f others to a third party as his sister d i d , who would frequently be i n a context including her brother and another person. The girl referred to herself by name 145 times as compared w i t h only 25 examples i n her brother's data; she also had a much higher frequency o f request forms i n -te. Both differences Okubo attributes to birth order. A s the second c h i l d , the girl probably had a greater need to distinguish herself from her brother and empha­ size her o w n wishes. I n fact, this motivation for acquiring the -te inflection, which marks requests, may have contributed to the girl's early focus on verbs. Thus early environmental differences may foster the development o f different approaches to language acquisition. Once the child has focused upon a particular type o f language use, such as labeling objects, reporting speech, or requesting actions, this may lead to the development o f certain grammatical areas over others. Thus a child w h o reports speech w i t h a sentence-final particle may become more interested i n other sentence-final particles; a child w h o is not interested i n objects may not be motivated to acquire grammatical means o f expressing object attributes w i t h modifiers. Clearly, the social context o f lan­ guage acquisition, as w e l l as personality and cognitive differences, must be taken into account i n understanding individual differences i n language development. 10.2.

Genitive

Constructions

Komura (1981) analyzes the differences between t w o children i n their acquisi­ tion o f genitive N noN constructions, focusing primarily upon possessives. Each child was recorded at 1-2 week intervals from 1 year-of-age to somewhat over 2 years. T w o different courses o f development were observed. One child first expressed possession by single nouns referring to possessors while pointing at the object possessed. N e x t , the genitive particle no emerged and was used after the possessor, as i n Toochan no ' D a d d y ' s ' . Then constructions w i t h t w o nouns emerged, first simply juxtaposed, as i n *neechan tokei 'older sister's w a t c h ' , or w i t h a vague sound between the nouns. Later, the full adult form neechan no tokei 'older sister's watch' emerged. Thus i n this c h i l d , the acquisition o f geni­ tive constructions w i t h t w o nouns d i d not seem to be derived directly from the

484

Clancy

temporally prior stage o f N no. I n this c h i l d , the sequence o f development was: N < N no < N N < N no N . I n contrast, the other child started using both single possessor nouns and N no at the same time. Next, full N no N forms emerged without the intermediate step o f N N . Juxtaposed nouns d i d appear, but only after N no N had been acquired; they were very rare and dropped out quickly. I n each child, other types o f genitive constructions, such as those expressing apposition, time, place, quantity, etc., emerged later than possessives, but followed the same sequence o f development. Thus for the first c h i l d , the acquisition o f genitive constructions apparently involved an extra step, i n w h i c h adult N noN was first simplified by omission o f the l i n k i n g no before the full construction emerged. A s noted i n section 7 . 1 , this is apparently a very common course o f development i n Japanese children. I t is interesting to note that the second c h i l d , who skipped this stage, showed an extremely rapid pace o f acquisition. The first child used single nouns at 1;5 years, N no at 1;6, N N at 1;7 and N no N at 1;8, whereas the second child used both N and N no at 1;4, and was using N no N later i n the same month. Moreover, the frequency o f the construction i n each child's speech was quite different. The first child d i d not have five instances o f N no until 1;8, or o f N no N until 1; 11, whereas the second child used at least five instances o f both N no and N no N w i t h i n the first sample i n w h i c h each construction appeared. Thus i n the slower, and perhaps more typical, course o f development, child­ ren gradually learn to add explicit surface markers o f semantic intention to unmarked nominals; i n these children, the semantic intention clearly precedes the ability to express that intention grammatically (see Slobin, 1973). The omission of no i n two-noun constructions after i t has been acquired w i t h single nouns suggests the continued pressure o f a slowly lifting cognitive ceiling on the linguistic means o f expressing semantic intentions i n production. I n children w h o are developing more slowly, the limits o f this ceiling apparently apply first to the production o f an unmarked semantic relation, then to the production o f the grammatical marker, creating four stages i n the acquisition o f genitive construc­ tions: N < N no < N N < N no N . W h e n the acquisition process is greatly accelerated, the constraint on production seems to apply i n only t w o stages, to single nouns and then to two-noun constructions. I n this case, overt surface marking o f the intended semantic relation does not seem to appreciably increase the cognitive demands, and the grammatical marker appears as soon as the "content w o r d s . " This makes sense, since grammatical morphemes often, as i n Japanese genitive constructions, merely mark redundantly what is already being 3

4

A s expected, this production constraint seems to apply only at the early stages of development. Komura reports that both children acquired N no before N no N in sentences which contained only a single constituent, such as Watashi no '(It's) mine' or Watashi no hon '(It's) my book'. In each child's speech, N no also emerged before N no N in two-constituent sentences, e.g. ones including both subject and predicate, but N n o N became frequent earlier in these sentences. By the time threeconstituent sentences were being used, N no N appeared before N no. 3 4

4.

The Acquisition of Japanese

485

adequately conveyed i n context by more telegraphic means (see B r o w n , 1973). As the individual differences i n K o m u r a ' s study suggests, the addition o f a content term i n a semantic relation presents children w i t h a more difficult cog­ nitive task than the addition o f redundant morphological markers. 10.3. Numeral

Classifiers

As noted earlier i n section 4.7, one o f the problem areas for Japanese children is the acquisition o f numeral classifiers. Matsumoto (1984c) reports on i n d i ­ vidual differences w h i c h he found i n children's use o f numeral expressions i n various experimental tasks. I n these counting tasks, children generally tended to prefer Sino-Japanese numerals and classifiers to the native Japanese forms. H o w ­ ever, individual children differed i n their degree o f preference for the SinoJapanese forms. I n one counting task, children were asked to count cups and erasers, which can take either Japanese or Sino-Japanese forms, and mountains and clouds, w h i c h take only Japanese numerals w i t h the unmarked -tsu classifier. On the basis o f their responses, Matsumoto divided the children into t w o groups: a 'Sino-Japanese g r o u p , " who used Sino-Japanese numerals, alone or w i t h the classifier -ko, w h i c h Matsumoto analyzes as the unmarked Sino-Japanese classi­ fier for objects, on 6 or more o f the 8 items, and a "Japanese g r o u p , " who used Japanese numerals w i t h the unmarked Japanese classifier -tsu on 6 or more o f the 8 items. O f the 25 T o k y o first-graders i n one experiment, 11 preferred Japanese forms, and 16 Sino-Japanese; i n an experiment on 31 K y o t o children aged 5 ; 6 6;5 years, 6 preferred Japanese forms and 12 Sino-Japanese. The children's preferences were manifest i n certain o f the other counting tasks as w e l l . The "Japanese g r o u p " correctly attached the Japanese classifiers -ka for days and -tsubu for small, grainlike objects to Japanese numerals, whereas the " S i n o Japanese g r o u p " was more likely to make errors overgeneralizing Sino-Japanese forms, saying, for example, san-nichi instead o f mik-ka 'the third day ( o f the m o n t h ) ' , or attaching a Japanese classifier like -tsubu to a Sino-Japanese numeral. 4

Counting i n Japanese provides an interesting case i n which more than one form is available for the same function. Matsumoto's findings indicate that, faced w i t h t w o different series o f numeral expressions for counting, children tend to stick to either the Japanese series or the Sino-Japanese one. Thus there is support for Slobin's claim (1985) that children prefer a one-to-one correspon­ dence between form and function. Sanches (1977) also reports that some o f the young children i n her study o f numeral classifiers stuck to the Japanese forms w i t h -tsu; others, however, began using Sino-Japanese numerals and classifiers for lower numbers while relying on the Japanese forms w i t h -tsu for numbers above t w o or three. This finding, as w e l l as the differences i n degree o f prefer­ ence among Matsumoto's subjects, suggest that some children may adhere more strongly than others to the one form-one function principle. I t w o u l d be interest-

486

Clancy

ing to investigate whether different children's degree o f adherence to the princi­ ple o f one-to-one mapping results i n consistent profiles o f individual differences in the acquisition o f various linguistic subsystems. 10.4. Processing

Strategies

Both Hakuta (1982) and Sano (1977) report individual differences i n the sequence i n which different types o f sentences are acquired. Both authors found that most children had acquired S O V active sentences by about VA years-of-age. However, o f children w h o controlled t w o types o f sentences, some had mastered OSV passives, but not O S V actives, whereas other children had mastered O S V actives, but not passives. These individual differences may reflect cognitive differences shaping the kinds o f processing strategies which children formulate. Some children may be more sensitive to specific local morphological markers, whereas others rely more heavily upon w o r d order i n processing sentences. Children w h o acquire passive sentences before O S V actives apparently find it easier to formulate an entirely new processing strategy, using different mor­ phological cues, such as the agentive particle ni and the passive inflection, than to change the w o r d order o f their existing agent-patient-action strategy. Children who acquire O S V actives before passives apparently experience greater difficulty in dealing w i t h a new configuration o f morphological markers and/or greater flexibility i n changing the w o r d order o f an existing processing strategy. There is some striking evidence o f differing degrees o f dependency upon word order from the experiments on sentence processing. Hayashibe (1975) reports that three children i n his study o f active sentences consistently interpreted all O S V sentences as i f they were S O V , ignoring case particles completely. I n their studies o f active and passive sentences, Sano (1977) found t w o children, and Hakuta (1982) found three, who correctly interpreted all S O V actives and all OSV passives, and consistently reversed the interpretation o f all O S V actives and SOV passives. Clearly, these children had formulated an extremely r i g i d agentpatient-action strategy for processing sentences and were oblivious to mor­ phological cues such as the case markers on the nouns and the inflections on the verbs. The other children i n Hakuta's study were usually sensitive to at least the subject marker ga when i t appeared on the first noun i n a sentence. O f course, longitudinal data w o u l d be necessary to verify that these findings represent real individual differences rather than different stages i n the same course o f development or simply experimental " n o i s e . " However, at this point it seems likely that some children focus upon more global, gestalt aspects o f linguistic input, such as relative w o r d order, whereas others rely more upon local cues. Such individual differences w o u l d probably have definable influences throughout the course o f acquisition. Ultimately * they might prove to be depen­ dent upon very general aspects o f cognitive style, such as field dependency, which could be measured independently and correlated w i t h linguistic findings.

4.

The A c q u i s i t i o n of Japanese

487

1 1 . Input and Adult-Child Interactions A t this stage, there is various research available on the nature o f adult input to Japanese children, but little on the effects o f typical patterns o f adult-child interaction upon language development. I t is often pointed out that Japan, like the United States, is a "child-centered" society (see Fischer, 1970), and certain aspects o f the linguistic input to very young children, such as frequent use o f directives, questions, and expansions i n both countries, reflect this orientation. As i n America, i t seems that the primary source o f verbal input to Japanese children is usually their mother; Okayama (1979) reports that 83.9% o f the mothers o f 31 2-year-olds i n her study said that the person who talked to the child most was the mother. In keeping w i t h the greater emphasis i n Japan upon differentiating social status linguistically, the babytalk register is much more developed than i n Amer­ ica, and is used by mothers to a later age (Fischer, 1970). Fischer suggests that this is indicative o f the Japanese indulgence o f dependency, as contrasted w i t h the American push for precocity, w h i c h reflect the very different attitudes toward children i n each country. A s Fischer proposes, extensive use o f the babytalk register i n Japan emphasizes the status o f the child as such, w h i c h is consistent w i t h the more hierarchical nature o f family and social structure i n Japan. Chew (1969) analyzes the structure o f the babytalk lexicon i n Japanese, and Murata (1960) examines the use o f babytalk words i n a large sample o f 1-year-olds. Sanches (1968, Chap. 3) discusses the morphology and acquisition o f babytalk forms by three 2-year-olds. In this section, I w i l l consider those features o f mothers' speech to Japanese 2year-olds w h i c h may have an influence upon their acquisition o f grammar, and their development o f the characteristically Japanese indirect communicative style. 1 1 . 1 . Expansions

and

Paraphrases

T w o features o f English-speaking mothers' speech w h i c h are frequently as­ sumed to facilitate children's grammatical development are paraphrases and expansions. As B r o w n (1973, p . 106) points out, expansions provide children w i t h a complete, grammatical version o f their semantic intention at the moment when they are concentrating on expressing that intention w i t h their current, limited grammatical means. Similarly, Cross (1977) suggests that mothers' selfrepeating paraphrases, w h i c h frequently juxtapose both full and abbreviated versions o f the same message, provide the child additional opportunity to process the original utterance, and also illustrate ways o f encoding the same intention at different degrees o f complexity. Hearing these different options w h i l e trying to understand a particular message may be important i n revealing to the child the functions o f certain forms, such as verbal inflections.

488

Clancy

Research on mother-child interaction i n Japan indicates that both expansions and paraphrases occur i n the input to 1-year-old children. For example, Murata and Ohara (1966) recorded 41 mothers interacting w i t h their 1-year-olds, and found that they repeated their children's speech very frequently. Almost all o f these were "corrective i m i t a t i o n s " or expansions. Mothers o f children aged 1;31;5 expanded 14.5% o f their children's utterances, and mothers o f children aged 1;6-1;8 expanded 25.4% o f their utterances. Thus as children reached the stage where grammatical development usually begins, at about V/z years-of-age, the frequency o f maternal expansions increased. Studies o f English-speaking moth­ ers have found different frequencies o f expansion: Cross (1977) reports that expansions constituted 20.5% o f maternal input to the l ; 7 - 2 ; 8 - y e a r - o l d children in her sample, whereas Newport et al. (1977) found only 6% expansions i n the speech addressed to children o f l ; 1 0 - 2 ; 6 years. Although i t is not possible to compare frequencies w i t h the data available, i t is clear that expansions are an important feature o f mothers' speech to Japanese, as to English-speaking, children. Data on English-speaking mothers also indicate a high frequency o f selfrepetitions or paraphrases. Newport et al. (1977) report that 23% o f the mothers' utterances i n their sample were exact or partial renditions o f the same content; Cross (1977) found a rate o f 2 8 . 2 % . I n m y sample, many such sequences occurred i n the mothers' speech to 2-year-olds. A s Newport et al. and Cross also found, paraphrases were especially common i n sequences o f imperatives, where they served to ensure comprehension, or, as Gleason (1977) has discussed, to guide and direct the child's behavior. Thus sequences such as the following were typical: (to her daughter of 2;1 years old) ja, well

sono that

teeburu table

katazuke - te clean CONT

soko ja - nai, hako there COP+TOP N E G box

choodai. please

no naka G E N inside

ni in

ire - te put CONT

ki - te. come IMP hako box

no naka G E N inside

ire - te put CONT mada, still

ni in

mada mada mada, still still still

hako box

no naka G E N inside

ko - nakya. come O B L I G

motto, more

asoko there

no ue G E N top

katazuke - te clean CONT

choodai. please

ni in

4.

ja, well

The A c q u i s i t i o n of Japanese

489

o-katazuke shi - te. cleaning do IMP

'Alright, please clean up that table. Not there, put them inside the box. Inside the box, still still still, you must put them inside the box. Still, more, please clean up on top of that. Alright, clean up'. I n this sequence, the imperative i n -te is modeled repeatedly, w i t h and without the polite choodai 'please'. Both the ordinary imperative katazukete 'clean up' and the special babytalk form o-katazuke shite 'do cleaning' w i t h the o- nominal prefix, verb stem, and sum ' d o ' , appear. The concatenated VERB 4- kuru 'come' construction appears twice, both i n the affirmative irete kite 'put i n and come' and the negative o f obligation irete konakya 'must put i n and come'. (The form ko-nakya is short for ko-nakya ike-nail nara-nai I dame 'not coming w o n ' t d o ' , which roughly corresponds to English must.) As this example suggests, the verbal inflections appearing i n sequences o f paraphrases are extremely varied. I n two hours o f tape, I found the f o l l o w i n g verb forms exemplified i n sequences o f paraphrases i n the speech o f the mother o f the above example to her 2; 1-year-old daughter: the imperatives -te and -nasai, optionally w i t h yo (emphatic) or choodai (casual) and kudasai (formal) 'please'; the non-past V-rw ' y o u w i l l V ; benefactives w i t h agerulkureru ' g i v e ' ; conditionals w i t h -tara and -(r)eba ' i f ; aspectual concatenated verbs, such as Y-te miru ' V and see' or its babytalk equivalent V-te goran; conjoined sentences w i t h V-cha dame ' V - i n g is no good' and W-nakute mo ii 'not V - i n g is alright'; questions o f the type V n ja nai ' i s n ' t it that one V ' s ' and V deshoo, the presumptive form o f the copula meaning approx­ imately 'shouldn't one'; the cohortative -oo ' l e t ' s ' . Thus sequences o f para­ phrased directives served as a k i n d o f substitution d r i l l for verbal inflections, w i t h several major forms contrasted i n sequence for the child. Since these are precisely the inflections w h i c h are being mastered by young 2-year-olds, the tendency to paraphrase directives w i t h these forms may w e l l facilitate their acquisition. Paraphrases may also play a role i n the acquisition o f different speech styles. I n m y data, paraphrases addressed to young 2-year-olds w h o were not yet using polite inflections productively sometimes juxtaposed both plain and polite verb endings. For example, as the above mother was playing "hostess" w i t h her daughter using polite inflections, she said: tsukat - te use CONT

ii good

desu COP:POL

ka? Q

tsukat - te use CONT

'Is it alright to use it? Is it alright to use it?'

ii? good

490

Clancy

In the first question, the polite form o f the copula, desu, is used after the adjective // ' g o o d ' , which is how adjectives are inflected for politeness. The second question simply uses the plain style ii ' g o o d ' . The question marker ka appears i n the first sentence. As noted i n section 5.1., is used by women when they are using polite inflections; i n plain style, i t is common i n men's speech. Thus this sequence also exemplifies the use o f ka i n polite style and its omission in plain style i n women's speech. 11.2.

Directives

A striking feature o f the speech o f Japanese mothers to their 2-year-olds i n m y sample was the high frequency o f explicit directives, as illustrated above w i t h respect to paraphrases. Murata and Ohara (1966) report that an average o f 12.7% of mothers' utterances to children o f 1;3—1;5 years o l d and 17% o f utterances to children o f 1;6-1;8 were yookyuu 'requests', a category which included ques­ tions as w e l l as imperatives. These frequencies seem roughly comparable to the 18% frequency o f directives reported by Newport et al. (1977) and the 7.4% found by Cross (1977). Okayama (1979), however, found a higher frequency o f directives, 22.5%, i n the speech o f mothers to 31 2-year-olds who were taped during the course o f an entire day's activities. Apparently Japanese mothers are, i f anything, even more directive than American mothers. The high frequency o f directives i n Japanese mothers' speech, along w i t h their high percentage o f questions (see section 11.4 b e l o w ) , reflects a style w h i c h seeks to keep children involved i n interaction and to control their behavior. Azuma et al. (1979) have analyzed the relationship between a highly directive maternal communicative style and the performance o f Japanese children on a number o f cognitive measures. I n an analysis o f " c o n t r o l strategies," 58 Jap­ anese mothers were given a set o f hypothetical situations calling for adult inter­ vention, and were asked to respond w i t h exactly what they w o u l d say i f their child were present. Direct physical intervention and demands for compliance made w i t h no rule or reason were categorized as "appeals to a u t h o r i t y . " Use o f this authoritative strategy correlated positively w i t h the children's performance on seven different cognitive tasks, such as number conservation, literacy, and spatial ability. I n contrast, i n a sample o f 67 American mother-child pairs from the same study, reliance upon "appeals to a u t h o r i t y " correlated negatively w i t h children's performance on six out o f the seven tasks. F r o m a social point o f v i e w , one might speculate that i n a society which emphasizes conformity and hier­ archical relationships, maternal input w h i c h is highly directive is a good prepara­ tion for successful functioning, whereas i n a society w h i c h emphasizes indi­ vidualism and independence, excessive emphasis upon obedience to authority figures puts children at a disadvantage. On the personal level, the meaning o f an authoritarian style may be quite different i n each culture, reflecting different attitudes toward the c h i l d , w h i c h i n turn have different effects upon cognitive development.

4.

11.3. Verbal

The A c q u i s i t i o n of Japanese

491

Instructions

I n m y sample, many o f the directives w h i c h Japanese mothers addressed to their 2-year-olds consisted o f explicit instructions i n what the children should say in various situations. These instructions occurred either i n the course o f ordinary interactions or w i t h i n role-playing routines. M u c h o f this instruction was aimed at teaching children the many polite verbal formulas i n Japanese, w h i c h are much more numerous than i n English, and cover a much broader range o f situations. The verbal formulas taught to 2-year-olds include greetings such as Ohayoo (gozaimasu) ' G o o d m o r n i n g ' , Konnichi wa 'Good day', Oyasuminasai ' G o o d night', and the polite expressions choodai or kudasai 'please' for requests. Japanese children also learn many formulas w i t h no fixed English equivalents, such as Itadakimasu ( L i t . ) T receive i t ' , said before eating, Gochisoosama deshita ' I t was a fine m e a l ' , said after eating, Doozo 'Please/help yourself/go ahead', Tadaima ' I ' m h o m e ' , and Itte kimasui'mairimasu T go and w i l l come back', said upon leaving the house. Mothers i n m y sample said these formulas for their children i n appropriate situations, attributed them to characters i n picturebooks, and called their children's attention to the forms when they were used by other people. The mothers frequently prompted their children to use polite formulas when the occasion arose. I n the f o l l o w i n g case, when m y assistant had picked up something for a child o f 1; 11 years, his mother said, Horn, ochichatta yo. Arigatoo wa? Oneechan ni arigatoo tteyuu deshoo ' L o o k , it fell. What about thank you? Y o u should say "thank y o u " to older sister'. Mothers seemed to take every opportunity to comment on and reinforce the use o f polite formulas, as i n the following case, when the same child at 1; 11 was talking about his father's departure that morning: Child:

Bai-bai tte itta no. 'He said "bye-bye" '.

Mother:

Itta no ne. Papa nante itta? Itte mairimasu tte itta deshoo. Itte mairimasu. 'He said it, didn't he. What did papa say? He said, " I go and will come back," didn't he. " I go and will come back" '.

The finding that mothers o f 2-year-olds provide so much instruction i n the use o f these verbal formulas is consistent w i t h the research reported i n Hess et al. (1980). I n this study, Japanese mothers expected earlier mastery o f social cour­ tesy, such as greetings, than American mothers. Japanese mothers also provide children w i t h a considerable amount o f instruc­ tion i n appropriate verbal behavior which is somewhat less stereotyped than polite formulas. I n role-playing routines i n m y data, mothers created certain types o f social interactions, and then demonstrated to their children, w i t h explicit instructions, how to behave and what to say. For example, after starting to eat, it is polite to comment that the food is delicious. I f a child does not say Oishii ' I t ' s

492

Clancy

delicious' spontaneously, the mother may ask whether the food is delicious. I n the following exchange, the mother demands that her child o f 2;1 years make this comment, when she had been eating without saying anything: Mother:

Oishiku nai no? Tabenai de ii yo. Moo gochisoosama shite choodai. Sore moo ii kara. 'It's not delicious? You don't have to eat it. Please finish eating. You've had enough of that'.

Child:

Iya! 'No!'

Mother:

Soshitara oishii tte iwanakya. 'Then you must say, "It's delicious" '.

Child:

Oishii. 'It's delicious.'

Mother:

Oishii nee. 'It's delicious, isn't it?'

Children are also instructed i n the proper behavior for a host/ess serving guests. I n m y data, 2-year-olds were taught to ask their guests what they wanted to eat, to say Doozo ' H e l p y o u r s e l f when offering food, to ask Oishii? 'Is i t delicious?' after the guests began to eat, to ask i f the food was too hot or cold, and whether the guests wanted more to eat. I n role-playing routines, children were also taught how to speak on the telephone. For example, one mother w o u l d demonstrate to her child what to say to his grandmother on the phone, teaching h i m to inquire about her health and to invite her to visit. Such play included instruction i n the use o f polite style through modeling, as i n the f o l l o w i n g example to a child o f 2;2 years: Mother:

o - genki HON healthy

desu COP:POL

ka Q

tte QUOT

chanto properly

iwa - nakya, say OBLIG

hora. listen 'Listen, you must say, "How are you?" '. Here the mother uses the honorific prefix o- on the w o r d genki 'healthy', and the polite form o f the copula, desu, w i t h the question particle ka, as is typical o f questions i n polite speech. Thus instructions i n what to say i n particular situa­ tions incorporate information about which speech register should be used as w e l l .

4.

11.4.

Question-Answer

The A c q u i s i t i o n of Japanese

493

Sequences

Another major feature o f the child-centered verbal input to English-speaking children is a very high frequency o f questions. Newport et al. (1977) cite a frequency o f 4 4 % , and Cross (1977) gives 33.4%. Figures for Japanese indicate a somewhat lower frequency o f questions: Okayama (1979) reports 18.7% i n her sample o f mothers' speech to 2-year-olds. Since the studies on English found lower frequencies o f imperatives than Okayama, it may be that English-speaking mothers phrase directives i n question form more frequently than Japanese mothers. Aksu-Kog & Slobin (1985) have pointed out the role o f question-answer sequences i n exemplifying grammatical morphology to Turkish children. Noting that inflectional suffixes remain the same across both question and answer, they propose that this k i n d o f interaction may direct the child's attention to the morphology. I n Japanese, the importance o f question-answer sequences i n ex­ emplifying verbal inflections may even be greater, since questions often consist of a single predicate, w i t h the nominal arguments ellipted. Moreover, an ex­ tremely common response to a yes/no question i n Japanese is a full predicate, rather than a single-word 'yes' or ' n o ' answer, as is common i n English. Jap­ anese children, for example, i f asked Taberu? " W i l l you eat?', very often reply Taberu T w i l l eat' and not just Un ' Y e s ' . Children apparently learn to answer questions this way very early. Tanouye (personal communication) has found that mothers o f Japanese children just over 2 years o l d often answer their o w n questions w i t h full predicates, w h i c h may be how children acquire this pattern o f response. Having learned to answer questions i n this w a y , Japanese children incidentally receive a tremendous amount o f practice i n a variety o f verbal inflections simply by answering their mothers' questions. I n m y transcripts o f 2year-olds, most imitative utterances were answers to yes /no questions. Further­ more, these answers seemed to be grammatically progressive i n the youngest child i n m y sample, a boy o f 1; 11 years. N e w inflections tended to appear first and to be most common i n answers to questions, w i t h spontaneous usage appear­ ing a few weeks later. The Japanese pattern o f asking questions about alternatives may also contrib­ ute to the acquisition o f verbal inflections. I n Japanese, the typical way o f asking questions about t w o contrasting possibilities does not involve verbal ellipsis as i n English sentences, such as Will you eat an apple or a tangerine?. Rather, t w o distinct sentences repeating the predicate are used, as i n the following: ringo apple

tabe - ru? eat NONPAST

mikan tangerine

tabe - ru? eat NONPAST

'Will you eat an apple? Will you eat a tangerine?' This pattern serves to increase the frequency o f verbs and verbal inflections i n the input to Japanese children. I n cases o f affirmative and negative alternatives, the

494

Clancy

Japanese child receives explicit modeling o f the contrast between the difficult negative inflections and the more common affirmative forms, as i n the f o l l o w i n g example from m y data: oishi - i? delicious NONPAST

oishi - ku delicious

na - i N E G NONPAST

no? EP

Ts it delicious? Is it not delicious?' Clearly, such questions could be valuable to the child who is having difficulty acquiring the negative inflection o f adjectives w i t h -ku. Question-answer exchanges may also play a part i n the acquisition o f case particles i n Japanese. For example, i n m y data, it was clear that these sequences were fostering acquistion o f the topic marker wa i n one girl o f 2;1 years. This child at first used wa only i n one-word questions, such as Papa wa? ' W h a t about/where is papa?'. I n declarative sentences, this child used wa only i n the answers to questions w i t h wa, such as the f o l l o w i n g : (Child is giving out blocks) Mother:

mama wa? mama TOP 'What about mama?'

Child:

mama wa kore. mama TOP this 'Mama gets this one'. Lit., 'As for mama, this one'.

(pointing to characters in Heidi storybook) Adult:

kotchi this

wa dare? TOP who

'Who is this?' Child:

kotchi this

wa neesan. TOP older sister

'This is older sister'. Thus i n her replies to questions w i t h wa, this child builds her sentence upon the preceding utterance. This strategy produces grammatical results, since the topic of a sentence i n Japanese, unlike the subject, is not restricted to a limited set o f semantic relations w i t h the predicate. The child probably developed this strategy by observing the responses to her o w n N wa? ' W h a t about N ? ' questions.

4.

T h e A c q u i s i t i o n of Japanese

495

Question-answer sequences o f this type exemplify the grammatical structure o f sentences w i t h topics, and probably facilitate their acquisition. 11.5.

Communicative

Style

Compared w i t h Western languages, the typical style o f communication i n Japanese is intuitive and indirect. The basis o f this style is a set o f cultural values which emphasize omoiyari 'empathy', and are so w i d e l y shared that overt verbal communication often is not required. Thus i n Japan, the ideal interaction is not one i n w h i c h speakers express their wishes or needs adequately and addressees understand and c o m p l y , but rather one i n w h i c h each party understands and anticipates the needs o f the other and fills them before any verbal communication becomes necessary. Silence is more highly valued i n Japan than i n the West; i f all is going w e l l , there should be no need for speech. Ito (1980) points out that excessive verbalism has traditionally been looked d o w n upon i n Japan, es­ pecially for men. This attitude, he notes, is revealed i n traditional sayings, such as Iwanu ga hana 'Silence is better than speech'. I f verbal communication enters i n , it w i l l not be explicit; rather the speaker w i l l rely upon the hearer's ability to realize what s/he means, often i n spite o f what is actually said. This style o f communication can cause tremendous problems for Americans, who discover to their frustration that 'yes' often means ' n o ' , but cannot figure out when. (See Loveday, 1982, and Ueda, 1974, w h o gives "Sixteen ways to avoid saying ' n o ' in Japan.") I n Japan, people are often consciously aware o f these differences i n commu­ nicative style, and most Japanese w h o have had contact w i t h Americans can articulate these values explicitly, on the basis o f many misunderstandings. For example, Japanese visitors to this country often point out that i n Japan, an offer of food should ideally be refused three times before accepting. O f course, the host must realize that the guest is actually hungry and merely exhibiting appro­ priate enryo 'reserve'. M o s t Japanese w h o come to America can expect to suffer a period o f hunger before learning that offers o f food w i l l be made only once, and w i l l probably raise a few eyebrows by their unseemly boldness i n accepting offers quickly upon returning to Japan. Clearly, the Japanese system o f commu­ nication can w o r k only i n a homogeneous society, i n w h i c h people actually can anticipate each other's needs and wants. Universal expectation o f empathy fos­ ters amae, dependency or relying upon the indulgence and patronage o f others. D o i (1973) has analyzed how this concept pervades both individual and social psychology i n Japan. How does a Japanese child learn this intuitive, indirect style o f communica­ tion? Studies o f early mother-child interaction have revealed patterns emphasiz­ ing non-verbal communication at an extremely early stage. For example, i n a study o f 30 Japanese and 30 American infants o f 3 - 4 months o l d , Caudill and

496

Clancy

Weinstein (1974) have found that Japanese mothers talk to their children signifi­ cantly less often than American mothers, and Japanese children had significantly lower rates o f "positive v o c a l i z a t i o n " than the American children. On the other hand, the Japanese mothers were together i n the same r o o m w i t h their children, even while they were sleeping, significantly more often than American mothers, and responded quickly to soothe and care for their child's needs upon any negative vocalization. Caudill and Weinstein conclude that " i t is as i f the major­ ity o f the American pairs had reached an 'agreement' to be talkative, while the majority o f the Japanese pairs had reached an 'agreement' to be s i l e n t . " Thus, as early as 4 months-of-age, Japanese infants have developed a pattern o f silent togetherness w i t h their mothers, whereas American infants have learned to i n ­ teract w i t h their mothers vocally. M y data on a much later stage o f development reveal patterns o f verbal interaction w i t h 2-year-old children w h i c h could foster acquisition o f an intu­ itive, " m i n d - r e a d i n g " style o f communication. Japanese mothers teach their children to pay attention to the speech o f others, to intuit and empathize w i t h their feelings, to anticipate their needs, and to understand and comply w i t h their requests, even when these are made indirectly. For example, i n m y data, the 2-year-olds w o u l d sometimes become engrossed in their o w n actions and fail to notice the attempts o f others to engage them i n conversation. The mothers d i d not allow this to continue; when their children failed to respond to someone, they w o u l d explicitly focus the children's attention upon the person who was trying to interact. Mothers often repeated other peo­ ples' utterances, indicating to their children that they must pay attention and respond to the speech w h i c h is addressed to them. I t was common for mothers to repeat questions which their child had failed to answer, as i n the f o l l o w i n g case: 35

(Child is serving imaginary food on toy dishes) Adult:

Mahochan nanika tabeten no? Koko nani ga haitten no? 'Are you eating something? Is there something in here?'

Child:

(doesn't respond)

Mother:

Nani ga haitten no kana. Oneechan nani ga haitten no tte kiiteru yo. T wonder what's in it. She's asking, "What's in it?" '

I f someone made a request, mothers frequently repeated i t , indicating that the child must comply immediately.

Pye (1982) reports that Mayan mothers also often repeat the speech addressed to their children. The Mayan mothers also frequently spoke for their children, and this occurred in my data as well, although not as frequently as attributing speech to others. 35

The A c q u i s i t i o n of Japanese

4.

497

Indirect speech is especially characteristic o f certain types o f polite interac­ tions, such as between host/ess and gue?t. I n m y data, 2-year-olds received practice and instruction i n the use and interpretation o f indirect speech i n roleplaying host/ess-guest routines w i t h their mothers. I n these routines, a mother w o u l d often explicitly tell her child what to say to the guests or to prepare and offer a certain type o f imaginary food. I n expressing requests for themselves, however, mothers frequently simply stated a w i s h , such as, M a m a suupu mo hoshii naa 'Gee, I ' d like soup t o o ' . The sentence-final particle naa often follows expressions o f inner feelings and wishes, and conveys a sense o f monologue. Sing-song intonation w i t h a very long vowel also occurs, adding to the impres­ sion o f talking to oneself. Mothers sometimes even affected indifference to the presence o f the c h i l d , even gazing away. However, the 2-year-olds seemed quite aware that these wishes were actually to be treated as imperatives, perhaps because they were sometimes accompanied by an explicit directive, as i n the following case: mama supagechi mama spaghetti

ga SUBJ

ii good

naa. EXCLAM

tsukut - te make CONT

kudasai. please

'Gee, I'd like spaghetti. Please make some'. I f a child failed to understand an indirect request, the mother might "trans­ l a t e " i t into a more explicit expression. For example, on one occasion m y assistant wished to end a routine i n w h i c h a child o f 2;1 years kept serving her food, and the f o l l o w i n g interaction took place: Adult:

moo already

ii good

desu. COP:POL

(Lit., 'It's already good'.) 'I've had enough'. (Child continues serving imaginary food) Mother:

moo already

ii good

tte QUOT

oneesan. older sister

'She said, "I've had enough" '. (Child continues serving) Adult:

hai. moo onaka yes already stomach 'Thank you. I'm full'.

Child:

(serving)

ippai full

mii. milk '(Here's some) milk'.

desu. COP:POL

498

Clancy

Adult:

doomo very wa TOP

gochisoosama fine meal mii milk

deshi - ta. COP:POL PAST

desu COP:POL

a, oh

kondo this time

kore this

ka? Q

Thank you for the fine meal. Oh, now is this milk?' Child:

suupu soup

na no. COP EP

suupu. soup

suupu. soup

'It's soup. Soup. Soup'. Adult:

hai, yes

hai, yes

hai. yes

'Thank you. Thank you. Thank you'. Child:

jaa, alright

jaa, alright

jaa. alright

'Here. Here. Here'. Mother:

moo already

oneesan older sister

iya no

tte, QUOT

moo already

ii good

tte QUOT

mii wa. milk TOP 'She said, " I don't want anymore, I've had enough milk" '. A t this point, the child discontinued the game. I n this sequence, the mother begins by repeating the phrase moo ii without the polite desu for the child. W h e n the child ignored this, and continued serving despite the more explicit Moo onaka ippai desu ' M y stomach is f u l l ' , and the polite Gochisoosama deshita ' I t was a fine m e a l ' , w h i c h is said at the end o f a meal, the mother intervenes again. I n her last utterance, she quotes m y assistant as having said iya, w h i c h is a very strong refusal meaning approximately T don't want i t ' . She juxtaposes this w i t h the less direct expression moo ii 'already g o o d ' , m y assistant's first utterance, clarifying for the child that this should have been interpreted as iya T don't want i t ' . Thus Japanese mothers teach their children to read behind the polite state­ ments o f others, understanding them as expressions o f strong feelings and wishes w i t h which the child must comply. Not only do children learn to comply w i t h explicit and i m p l i c i t requests, they are also taught to anticipate the needs o f others, before anything is actually said. One common behavior for all three mothers o f 2-year-olds i n m y sample was to attribute speech to people w h o had not actually spoken, i n order to indicate to the child what they might be thinking and feeling. For example, when children played host/ess, their mothers w o u l d attribute requests for food to other people.

4.

The A c q u i s i t i o n of Japanese

499

They also often d i d this when the child was eating something alone. For exam­ ple, when one child o f 1; 11 years was eating a tangerine, his mother suddenly said, Oneesantachi mo tabetai tte T h e girls also say, " W e want to eat," although we had not said anything. Thus Japanese mothers train their children to anticipate the unspoken wishes o f others. Empathy is especially important i n order to avoid inconveniencing or annoy­ ing other people. The mothers i n m y sample were quick to point out cases i n which a child had caused someone trouble; again, this was often done by attribut­ ing thoughts or speech to someone who might appear to be silently content. When a child imposed upon someone, even i f the person w i l l i n g l y complied, the mother might indicate that such behavior is not appreciated. For example, when a child o f 1; 11 years asked m y assistant to peel a tangerine for h i m , and she was doing this very amiably, his mother said, Oneechan jibun no muite taberu tte ne 'Older sister says she'll peel and eat her o w n ' . Such examples indicate to the Japanese child that there may be a difference between what people seem to feel and what they are really thinking. Mothers frequently attributed feelings o f pain to others, especially i f the child was responsible but had failed to notice or apologize for causing the pain. For example, when the toy dishes o f a child o f 2;1 years fell on m y assistant, her mother immediately said, Neechan itai-itai tte 'Older sister says " o u c h , o u c h " ' before anything had been said. Feelings were even attributed to inanimate ob­ jects. When a child o f 1;11 years kept dropping apples on the floor, his mother said, Sonna koto sum n dattara ringo-san itai itteru wa yo ' I f you do that k i n d o f thing, M r . A p p l e says " o u c h " ' . As these examples suggest, mothers often attributed speech to others as a way o f correcting their child's behavior. For example, when a boy o f 2;4 years was getting very loud, pretending that he was firing guns, his mother attributed a request that he stop to his stuffed animal Kikochan: Kikochan bikkuri shitem yo. Kikochan yamete kudasai tte. Gomennasai tte. Kikochan ga itai tte ' K i k o c h a n is amazed. Kikochan says, "Please stop.'' Say, ' ' I ' m s o r r y . ' ' K i k o c h a n says ' ' o u c h . ' ' ' Mothers often pointed out to children when they were misbehaving that others d i d not approve o f their behav­ ior, attributing very direct statements o f disapproval or correction to people w h o had not spoken. For example, on one occasion when a child o f 2;1 years was loudly refusing to sing a song and yelling Dame! ' N o ! ' , his mother said: O, kowai, Yotchan, oneechan kowai tte. Yotchan kowai naa, Yotchan dame nante yuu kara, kowai, kowai ' " O h , I ' m afraid o f Y o t c h a n , " older sister says, " I ' m afraid. Gee, I ' m afraid o f Yotchan because he says ' n o ' . I ' m afraid, I ' m afraid." ' Consistent w i t h these findings, A z u m a et report that "appeals to f e e l i n g " function as mothers, who invoke the feelings o f others behavior. I n a study o f 58 Japanese and 67

al. (1979) and Conroy et al. (1980) a control strategy among Japanese as the rationale for a child's good American mother-child pairs, each

500

Clancy

mother was given a set o f hypothetical situations, such as being i n the super­ market w i t h a disruptive c h i l d , and was asked to respond as she w o u l d i f the child were actually present. I n analyzing the results, A z u m a et al. report, the category "appeals to feelings" had to be created because o f its frequency among the Japanese mothers, w h o used this strategy at a frequency o f 2 2 % , compared w i t h only 7% among the A m e r i c a n mothers. I n these responses, Japanese mothers said that they w o u l d ask the child to consider how they felt as the mother, or how the child w o u l d feel i f someone else d i d the same thing. As i n m y sample, this study found that Japanese mothers appealed to the feelings o f third parties and even o f inanimate objects, as i n " T h e w a l l w i l l feel sad." Although these appeals to feelings often function as control strategies, they can also be viewed as providing children w i t h explicit training i n empathy, lessons i n how to guess what others are thinking and feeling even when they have not spoken. I t is consistent w i t h the very different styles o f verbal communication i n Japan and America that American mothers thought o f using this strategy so much less frequently than Japanese mothers. Through their training i n empathy, Japanese mothers teach children not only to be sensitive to the needs and desires o f others, but also to fear their criticism and disapproval. Benedict (1946) has discussed how Japanese children are incul­ cated w i t h a fear o f ridicule and ostracism. A s Ito (1980) points out, familial approval is dependent upon more general social approval, upon children's con­ ducting themselves i n a manner w h i c h w i l l uphold their family's good name. Japanese mothers, Ito notes, often threaten a misbehaving child w i t h the ridicule of other people, saying Hito ni warawareru ' Y o u w i l l be laughed at by other people'. This strategy locates the source o f disapproval and constraint outside the mother, i n society at large. Japanese mothers also frequently convey their o w n reactions directly to their children, yet i n these cases too, the reactions tend to incorporate the opinion o f Japanese society. I n m y sample, the clearest examples o f this were mothers' use of the words okashii 'strange' and hazukashii 'shameful/embarrassing'. These reactions label for the child the behaviors w h i c h are not expected and/or socially disapproved, and indicate how to feel i n such situations. For example, when a 2year-old boy was asked to bring over some toy cars, he refused, whined, and started to cry. His mother said, okashii yo, strange EMPH

nai - tara. cry if

'It's strange, if you cry'. The w o r d hazukashii frequently means ' s h y ' , and children are often asked i f they feel hazukashii i n situations where they are the focus o f attention, as when they have been asked to perform for guests. The w o r d is also used w i t h a sense closer

4.

The A c q u i s i t i o n of Japanese

501

to 'ashamed' when a child has misbehaved i n front o f other people. I n such cases, mothers w i l l simply say hazukashii. Typically no explicit subject is used; hazukashii conveys both the mother's o w n feeling and the strong implication that the child should feel the same way. For example, when a girl o f 2;2 years i n m y sample was about to leave the house w i t h her parents and guests, she wet her pants, and then ran away from her mother as she was trying to wash her. Her mother pointed out that everyone was watching, and repeatedly said hazukashii. Kasahara (1974) suggests that Japanese child-rearing practices, which attempt to control the child's behavior by appealing to the disapproving gaze o f others, may contribute to a fear o f eye-to-eye confrontation, which is common among young adults i n Japan, but virtually u n k n o w n i n the West. I n this neurosis, people experience a phobia o f being stared at by others, and i n severe cases, fear that they cannot control their o w n eyes, and prevent their stares from inflicting undue pain upon others. The strong emphasis on Hito ga miteru 'People are watching' from early childhood, Kasahara proposes, may foster the development of personalities which incessantly watch, and dread being watched, by those outside their family circle. In this light, i t is interesting to note that very young children i n Japan may at first strongly resist internalizing their mothers' hazukashii and the disapproval o f watching eyes. For example, the 2-year-old i n m y sample who wet her pants responded to her mother's hazukashii w i t h a resounding, Hazukashii chigau! ' I ' m not ashamed!'. This is probably because Japanese adults do not actually expect very young children to measure up to the social norms o f self-restraint and discipline until the school years. As Benedict has pointed out, the "arc o f l i f e " in Japan is a U-curve w i t h the greatest freedom and indulgence enjoyed by babies and the elderly, and the l o w point o f greatest restriction during the prime o f life, especially just before marriage. F r o m an American point o f v i e w , child-rearing in Japan tends to be permissive. A n American's blood w i l l b o i l to watch the boyish temper tantrums w h i c h Benedict describes, i n which a child o f 3- or 4years-old screams and pummels his mother, while other family members stand patiently by. Yet an examination o f mothers' interactions w i t h 2-year-olds re­ veals that, i n their comments and attitudes toward their children's unrestricted or selfish behaviors, they are already sowing the seeds for the social constraints to be imposed by the watching eyes o f hito 'other people'. The permissiveness o f the early years actually increases mothers' opportunities to present such evalua­ tions over a long period o f time. W h e n the period o f early indulgence is finally over, Japanese children w i l l be quite familiar w i t h w h i c h behaviors are okashii 'strange', and w h i c h should make them feel hazukashii 'ashamed'. Early train­ ing i n empathy thus leads Japanese children to understand the feelings and expectations o f others, and also to experience the expected feelings themselves. This sets the stage for successful functioning o f the Japanese indirect, intuitive mode o f communication.

502

Clancy

CONCLUSIONS 12. T h e o r e t i c a l C o n c l u s i o n s Slobin (1973) has proposed that certain constraints on children's perception and production o f speech, and on their organization and storage o f linguistic rules, can be formulated i n terms o f universal operating principles w h i c h are brought to bear i n acquiring language. The data on Japanese acquisition provide evidence for certain basic tendencies underlying the encoding and decoding o f speech, and the organization o f form and function i n developing grammars. I n addition to cognitive pacesetting, Japanese data reveal the importance o f affect and pragmat­ ic developments i n the early stages o f language acquisition. 1 2 . 1 . Encoding

IDecoding

Principles

12.1.1. Perceptual Salience. Slobin (1973) has suggested that children have a tendency to pay attention to the ends o f units, acquiring word-final elements, such as suffixes and postpositions, earlier than word-initial elements, such as prefixes and prepositions. There is considerable evidence i n the data on Japanese acquisition for a perceptual bias favoring the ends o f units. The earliest grammatical acquisitions are those w h i c h occur not only at the ends o f words, but also at the ends o f utterances, as do sentence-final particles. Verbal inflections, which become productive late i n the one-word stage, very frequently occur at the ends o f sentences, since Japanese is usually verb-final. Certain very early case particles, such as the genitive no and the topic marker wa, also frequently appear at the ends o f sentences, as i n Papa wa? ' W h a t about/where is papa?' and Papa no ' I t ' s papa's'. A t an early stage, this perceptual bias may have discernible effects throughout a child's speech production. I n m y sample, one child o f 1; 11 years produced a rather large number o f utterances w h i c h were almost completely incomprehensi­ ble except for the last syllable, usually a verbal inflection or sentence-final particle. I n a smaller number o f utterances, only the first and last syllables were clearly articulated. The frequency o f both types o f utterances decreased over the course o f about 2 months, but i t was clear that this child's initial focus was upon producing the ends o f units correctly. Selective attention to the ends o f units, and to a lesser extent the beginnings o f units, leaves medial position as the most vulnerable. The acquisition o f mor­ phology i n Japanese provides ample evidence for children's failure to process morphemes w h i c h are " s a n d w i c h e d " between familiar elements. For example, the form ja, w h i c h is a contraction o f the copula da plus the topic marker wa, is typically omitted i n sentences o f the form N ja-nai ' I t ' s not an N ' , where it is sandwiched between a familiar noun and the negative predicate nai. Similarly, the particle to may at first be omitted from the construction N to chigau ' I t ' s not an N / i t ' s different from N ' , and the genitive no is often omitted from N no N

4.

The A c q u i s i t i o n of Japanese

503

constructions. The -ku inflection, w h i c h is suffixed to the stem o f adjectives preceding the negative inflection, seems to be omitted by all children i n the early stages o f negating adjectives. The only striking exception to this general tendency to omit morphemes i n medial position is -te, the continuative inflection. This inflection appears i n medial position i n compound tense/aspects, such as the present progressive/resultative -teru, and i n concatenated verb structures, such as benefactive and aspectual constructions o f the form V-te V . I n this case, children have already acquired the -te form o f many common verbs as a verb- and often utterance-final imperative. Therefore, the Y-te form is familiar and salient, whereas the f o l l o w i n g verb may be unfamiliar, and at first even not understood. Moreover, the form o f the verb preceding -te cannot occur alone; due to pho­ nological processes such as v o i c i n g , place o f articulation assimilation, and con­ sonant gemination, this f o r m w o u l d be unpronounceable, and has never been heard i n the input. Thus familiar morphemes which form a single phonological unit w i t h the preceding element may be exempt from the general vulnerability o f medial position. Similarly, the familiarity o f no as a possessive i n N no constructions may be responsible for the fact that some children do not omit it i n N no N constructions. Thus familiarity and/or semantic salience can over-ride the filtering process, but usually children do not attend to medial position as w e l l as initial or final position when processing input.

12.1.2. Intonation. I n addition to these biases, it appears that non-segmental, gestalt aspects o f linguistic input, such as intonation and w o r d order, have a perceptual advantage over local, segmental cues i n the acquisition o f certain areas o f Japanese grammar. For example, i n the development o f questions i n Japanese, there is some evidence that intonation may be processed earlier than morphological markers. Yoshida (1977) found that one child responded to cer­ tain types o f questions at the age o f 2;4 only i f they were uttered w i t h rising intonation. Three months later, he was able to respond to these questions even i f falling intonation was used, apparently because he had learned to recognize the sentence-final morphemes as potential question markers regardless o f intonation. Yoshida proposes that i t may be easier for young children to focus upon intona­ tion than upon morphology. Intonation is probably also an important factor i n the early acquisition o f sentence-final particles i n Japanese. 12.1.3. Word Order. The data on Japanese acquisition supports Slobin's (1973) proposal that children are sensitive to the sequential order o f input words and morphemes at a very early age. Japanese children at the t w o - w o r d stage are already producing sentences w i t h constituents postposed after the verb, as w e l l as sentences w i t h the canonical verb-final order. This finding, w h i c h is consistent w i t h the crosslinguistic data, provides further support for the view that a capacity

504

Clancy

for processing and storing sequential aspects o f linguistic input may be inherent in the child's mechanisms for perceiving and storing information. There is also some evidence i n the Japanese data for the primacy o f w o r d order over case particles i n processing sentences; this is consistent w i t h findings on German acquisition ( M i l l s , 1985). Hayashibe (1975) reports that i n his study of S O V and O S V active sentences, comprehension based on case particles emerged later than w o r d order strategies. Frequent ellipsis o f the direct object marker o, and the semantic inconsistency o f the subject marker ga, probably contribute to children's greater reliance upon w o r d order i n Japanese. I n lan­ guages such as Turkish, where inflections are regular but w o r d order is much more variable than i n Japanese, children are able to rely upon inflectional end­ ings at an extremely early age (Aksu-Kog & Slobin, 1985). I n Japanese, where word order is more consistent, it is apparently preferred over morphological strategies at an early age. 12.1.4. Contextual Support. I n language production, one factor w h i c h af­ fects children's ability to encode information is the availability o f contextual support, as Slobin (1985) proposes. I n m y data, a clear example o f this can be found i n the acquisition o f the conjunction kara 'because'. I n Japanese, kara is often used elliptically; the speaker presents a reason and leaves it to the listener to deduce the statement or request w h i c h follows from i t . For example, i f the addressee is resisting accepting a favor, it is common to say / / kara 'Because it's alright', leaving implicit 'let me do i t ' . I n m y data, the first uses o f kara were o f this type. For example, one child o f 2;1 years accidentally tore a page i n his sister's book, and i n the resulting uproar, as his mother sought to appease his sister, he said: (h)att age - ru paste give NONPAST

kara, because

mama. mama

'Because I'll paste it for her, mama'. Here the child is ellipting the conclusion that everyone should stop getting excited, and gives only the reason: that he w i l l paste the page together again. This elliptical use o f kara also emerged first i n the speech o f another 2-yearold i n m y sample. For example, talking to her father, w h o was practicing English w i t h me and not paying attention to her, this child angrily said, dame no good

kara because 36

'Because it's no good'. Since dame is a nominal adjective, it should be followed by the copula da before adding a conjunction. Therefore, the correct form is actually Dame da kara, with the copula da preceding the conjunction kara. 36

4.

T h e A c q u i s i t i o n of Japanese

505

leaving i m p l i c i t the conclusion that he should stop speaking to me. Similarly, Okubo (1967) reports that her daughter used kara at the end o f a single clause at 1;10 years o l d , 1 month before its first appearance as a conjunction between t w o clauses. Thus the earliest instances o f kara are linked to a context i n w h i c h the ellipted proposition is obvious, rather than to another explicit clause. I n m y data, one 2-year-old reached the next stage, i n w h i c h kara appeared i n single-clause responses to dooshite ' w h y ' questions. I n these cases, the child's subordinate clause is linked w i t h the proposition i n the prior utterance. A k s u (1975) has found that causal connectives i n Turkish emerged i n response to adult questions before appearing i n conjoined sentences. Thus the course o f acquisi­ tion o f kara i n Japanese suggests that it is easiest for the child to l i n k a proposi­ tion to the nonverbal context, somewhat harder to conjoin a proposition to a prior utterance, and most difficult to produce both clauses o f a conjoined sentence without contextual s u p p o r t . 37

12.2.

Form I Function

Principles

The crosslinguistic data suggest that children have certain inherent prefer­ ences for the organization o f form and content i n grammars. M a n y o f the errors which occur follow directly from the basic principles that underlying semantic notions should be overtly and clearly marked, and that the use o f grammatical markers should make semantic sense (Slobin, 1973). However, the Japanese data suggest that i n certain cases, either form or content may take precedence, and the child may temporarily ignore one or the other i n the early stages o f acquisition. 1 2 . 2 . 1 . Surface Structure Configuration. For example, i n the very early stages o f acquiring case particles, Japanese children may use newly emerging particles i n a random way. A t this stage, the children seem to k n o w that i n Japanese, a noun is typically followed by a case particle, which together form a single intonational unit. W h e n children come to recognize NOUN + PARTICLE as a typical surface structure configuration, they may add case particles after nouns even when they do not k n o w w h i c h particle is required. I t is interesting that this tendency to create a canonical surface structure seems to take precedence, at least in some children, over Slobin's principle that "the use o f grammatical markers should make semantic sense." Perhaps this provides further evidence for the relative importance o f intonation compared w i t h morphology i n the early stages of language development. I n contrast, Bloom et al. (1980) report that in three English-speaking children, the use of connectives to chain a child utterance to the nonlinguistic context developed at the same time as syntactic uses linking two clauses. A fourth child acquired contextual before syntactic uses. It should be noted, however, that contextual use of kara 'because' in Japanese is much more frequent in colloquial adult speech than contextual use of because in English. Bloom et al. also found that two of these children first used because both in answer to adult why questions, and to connect clauses within a single sentence. The other two children first used because in answers to why questions, as did the one child in my sample who advanced beyond the stage of contextual sentence-final kara. 3 7

506

Clancy

12.2.2. Semantic I Pragmatic Processing. The reverse tendency can be ob­ served i n the early stages o f acquiring sentence processing strategies. Experi­ mental data on Japanese 3-year-olds show that they tend to ignore morphological markers, such as case particles, and syntactic devices, such as w o r d order, and process input on the basis o f semantic/pragmatic factors. Hayashibe (1975), Hakuta (1982), and Sano (1977) all report that even before the age at which morphological or syntactic processing strategies emerge, children interpret sen­ tences by picking out the most likely agent o f the action from the t w o nouns given. Similarly, K . I . Harada (1976) found that at 2;8-2;10 years-of-age, her daughter was able to understand sentences w i t h relative clauses which described experiences w h i c h had occurred i n her o w n life, but misinterpreted complex sentences which were not consistent w i t h her experience. These findings, which are not, o f course, limited to Japanese, support the priority o f seman­ tic/pragmatic processing over morphological/syntactic strategies i n the t i m i n g o f language acquisition. Thus whereas form may sometimes take precedence over function i n the speech production o f very young children, i n their processing o f linguistic input, formal devices may be ignored i n the search for meaning. 12.2.3. Semantically Transparent Marking. There is considerable evidence in the Japanese data for the proposal that children prefer clear and overt marking of semantic information (Slobin, 1973). The errors which occur indicate that children acquire semantically empty or opaque forms late, prefer overt to zero morphemes, replace syncretic w i t h analytic options where possible, and have difficulty acquiring discontinuous morphemes and interrupted linguistic units. I n the early stages o f acquiring Japanese morphology, children tend to e l i m i ­ nate semantically empty or opaque forms, probably at first filtering them out i n perception as w e l l as production. For example, i n the negative -kunai inflection on adjectives, the -ku has no real semantic function, since -nai marks negation and non-past tense. The data indicate that children go through a stage o f omitting -ku and suffixing -nai directly to the non-past form o f adjectives. Similarly, allomorphs which are not recognizable to the child are typically omitted, such as na, the form o f the copula da which precedes sentence-final no, and ja, the form of the copula contracted w i t h the topicalizer w h i c h precedes the negative inflec­ tion. (As noted above, the perceptual " s a n d w i c h i n g " effect contributes to the vulnerability o f these morphemes.) Similarly, the non-past inflection o f adjec­ tives,-/, is very different from the more common verbal non-past marker, -(r)w, and children do not recognize -/ as a tense marker at first, either on adjectives, or on the negative inflection na-i. This v o w e l is not omitted, since it is taken as the base form o f adjectives, but is not analyzed as the non-past inflection for a long time. 12.2.4. Overt Marking. Slobin (1973) has proposed that zero morphemes violate the "clear and o v e r t " constraint, and that " i f a category is sometimes marked by 0 and sometimes by some overt phonological f o r m , the latter w i l l , at

4.

T h e A c q u i s i t i o n of Japanese

507

some stage, also replace the 0 . " There is a striking example from Japanese acquisition w h i c h supports this proposal: the extremely common overgeneralization o f the genitive particle no. W h e n a noun is modified by another noun, it must be preceded by the genitive particle no i n the construction N no N . I n contrast, adjectival and relative clause modifiers simply precede the head noun w i t h no overt morphological marker. Japanese children very often go through a stage o f extending no to follow adjectival as w e l l as nominal modifiers. Thus no is strongly preferred to zero marking o f prenominal modifiers. Children's preference for overt marking can also be observed i n the develop­ ment o f processing strategies for different types o f sentences. The Japanese data suggest that i t is easier for children to change an established processing strategy i f there is some overt morphological cue to signal that a change is necessary. For example, passive sentences have t w o overt morphological cues indicating that the subject should not be interpreted as the agent: the agent marker ni on a nominal other than the ga-marked subject, and the passive inflection -(r)are on the verb. I n constrast, O S V actives w i t h topicalized objects do not incorporate any new particles or inflections, but do require the child to interpret a wa-marked noun as the patient, although wa does not usually mark patients. Sano (1977) has found that passive sentences are comprehended earlier than O-waSY actives; one reason may be the overt markers available to trigger passive processing strategies. 12.2.5. Analytic vs. Syncretic Marking. Japanese data also provide evi­ dence for the view that children prefer analytic over syncretic marking. For example, i n denials o f the form N janai and N to chigau ' I t ' s not an N ' , the negatives janai and chigau follow independent forms and do not require segmen­ tation and affixing. Japanese 2-year-olds w h o have learned to use janai and chigau i n denials may overextend them, adding them on after finite verb forms instead o f suffixing a negative inflection to the verb root. Similarly, i n express­ ing the negative past, 2-year-olds w i l l sometimes simply add the negative nonpast -nai after a finite verb i n the past tense rather than affixing the conflated form -nakatta to the verb root. 12.2.6. Contiguous Marking (The ' 'Anti-Interruption'' Principle). There is support i n the Japanese data for Slobin's proposal that children tend to preserve the structure o f linguistic units, and to process one unit at a time. This principle is clearly operative i n Japanese children's acquisition o f complex sentences. Since Japanese does not have relative pronouns, the only mark o f embedding is repeti­ tion o f case particles and optional genitive marking on the subject. Several studies have shown that sentences w i t h center-embedded relative clauses are interpreted as i f they were conjoined sentences. That is, children assume that they are processing an uninterrupted main clause, and treat the initial N N V segment i n sentences o f the form N ( N V ) N V as a single unit. The anti-interrup­ tion principle, w h i c h presumably reflects very general constraints on perception

508

Clancy

and short-term memory i n young children gives an explanation o f certain find­ ings on complex sentences i n English w h i c h is also applicable to the Japanese data. 12.2.7. Formal Consistency. Perhaps the simplest and clearest type o f marking w h i c h a grammar may have is a one-to-one correlation between form and function. Slobin (1985) proposes that i f a language has more than one form for expressing a particular function, or different means o f expressing more differ­ entiated aspects o f a basic notion, then the child w i l l try to use one form for all instances. O n the other hand, i f a particular form has t w o or more related functions, the child w i l l avoid that form or attempt to restructure the system so as to distinguish its functions. I n several areas o f Japanese grammar, there are many forms to express subtle differentiations w i t h i n the domain o f a single basic concept. I n these cases, children typically overgeneralize at least one o f the available forms, and may try to use a single f o r m for all functions. For example, i n the area o f negation, distinct forms are available for the concepts o f prohibition, rejection, non-exist­ ence, and denial. The development o f negation is characterized by repeated overextensions o f the earliest and most common negative forms: nai, w h i c h expresses non-existence and serves as the non-past negative inflection, and iya, which expresses rejection and is typically the earliest negative to be acquired. I t o (1981) notes that there is a tendency for nai to be used as an " o m n i b u s " negative. I n the expression o f benefactive relations, there are t w o verbs o f g i v i n g as w e l l as a verb o f receiving which may be used. The most frequently confused forms are those w i t h the most similar function, the verbs o f g i v i n g ageru and kureru. The most common verb, ageru, was overextended i n Horiguchi's data (1979a) to a large percentage o f cases where a different verb w o u l d have been appropriate, including 4 4 . 9 % o f all contexts calling for kureru. I n the acquisition of counting, some children go through a phase o f overextending Japanese numer­ als w i t h the unmarked -tsu classifier where specific classifiers w o u l d be appropri­ ate. Thus the more semantically differentiated options are temporarily avoided. Another area o f Japanese grammar w i t h a many-to-one relation between form and function is the expression o f respect or deference. Polite speech, w h i c h is marked by -mas verbal inflections, shows respect for the addresee; honorific speech, w h i c h is marked by lexical substitutions and three special verbal inflec­ tions, shows respect even for referents other than the addressee. I n the acquisi­ tion o f respect terms, polite forms are sometimes overgeneralized as honorifics. Also, some speakers are eliminating one o f the t w o available inflections for subject honorification. Thus there is a strong tendency throughout Japanese grammar for a single form to be generalized w i t h i n a semantic domain, eliminat­ ing the subtle distinctions marked by the different forms available. In other cases, alternate forms w i t h i n a particular domain express grammatical rather than semantic distinctions. For example, i n Japanese morphology, a single

4.

The A c q u i s i t i o n of Japanese

509

semantic notion may be realized by different surface forms w i t h different parts o f speech. This is true o f negation, where nouns and nominal adjectives are negated w i t h janai, adjectives w i t h -kunai, and verbs w i t h -nai. Children generally use a single f o r m , nai, to negate these different parts o f speech i n the first stage o f acquiring negatives. Similarly, Japanese grammar distinguishes true adjectives from nominal adjectives, w h i c h are treated like nouns for marking politeness, negation, etc. Japanese children go through a phase o f treating nominal adjec­ tives like true adjectives and using the same inflectional endings for both. F o r m class distinctions are also ignored i n the acquisition o f prenominal modifiers. Japanese children often overextend the genitive particle no, which precedes head nouns modified by other nouns, and use it also w i t h adjectival modifiers, which should simply precede the noun they modify. In these cases, a single form is selected and applied to all instances o f a semantic category without regard to the distinctions between different parts o f speech made by the adult grammar. The greater generality, consistency, and persistence o f these errors suggest that the principle o f one-to-one mapping o f form and function is adhered to more strictly when different forms mark purely grammatical distinctions than when they mark subtle semantic distinctions. This is consistent w i t h Slobin's principle that "the use o f grammatical markers should make semantic sense." 12.2.8. Semantic Consistency. There is some evidence from Japanese ac­ quisition that children prefer forms to have a consistent semantic function. M a n y production errors are reported i n the literature for case particles which serve a variety o f semantic functions, such as the subject marker ga. The results o f comprehension studies indicate that children tend to apply the one form-one function principle to ga during processing. I n simple active sentences w i t h tran­ sitive verbs, ga marks the agent. I n comprehension experiments using transitive sentences, some young children assume that ga on the first nominal marks the agent, even i n S O V passive sentences, where ga marks the patient, and i n benefactive sentences w i t h morau 'receive', where ga marks the recipient. S i m i ­ larly, the topic marker wa frequently marks the subject o f a sentence, and i n transitive sentences therefore often marks the agent. Simple active O S V sen­ tences i n w h i c h wa marks a sentence-initial object are very poorly understood by Japanese children. Thus it is difficult for children to attribute semantically incon­ sistent functions, such as agent and patient, to a single morpheme. However, there are also cases i n w h i c h Japanese children do not maintain the principle o f one-to-one mapping o f form and function. These examples provide evidence for the constraints w h i c h l i m i t that general principle. One example from Japanese data is the overgeneralization o f the -te form i n the acquisition o f the verbal system. I n their attempts to produce complex adult forms incorporating the continuative -te, such as the present progressive/resultative -teru and bene­ factive and aspectual V-te V constructions, children w h o are already using -te as an imperative (correctly) may begin to overgeneralize -te to non-imperative

510

Clancy

contexts as w e l l . Thus -te ends up serving more than one function i n some children's speech. The surface homophony resulting from this overgeneralization seems unavoidable i f a child w h o cannot yet produce adult complex -te structures attempts to imitate them, or substitutes -te for them i n spontaneous production. This apparent violation o f one-to-one mapping may be tolerated partly because the children's underlying semantic intent and the adult target form are actually different. That is, since the homophony occurs only i n the child's output, it may not be subject to the same k i n d o f monitoring w h i c h leads to the avoidance o f adult homophonous forms. A more striking case o f tolerated homophony occurs w i t h no, which functions both as a sentence-final particle and a possessive marker i n the earliest stage o f Japanese grammatical development. T h i s ' f o r m is plurifunctional i n the adult input, but is acquired i n both functions and used very frequently by Japanese children even before the age o f 2 years. Here, the fact that the t w o no's belong to completely different grammatical subsystems probably accounts for children's willingness to tolerate the plurifunctionality. The t w o functions o f no are ex­ tremely different pragmatically and semantically, and also differ i n their distribu­ tional patterns. Both uses occur sentence-finally at the earliest stage, but pos­ sessive no at first follows only the names o f people, whereas sentence-particle no frequently occurs after verbal and adjectival predicates. Slobin (1985) has sug­ gested that the constraint against a single form having more than one function applies to forms having similar functions; i n this case, the t w o early functions o f no are apparently sufficiently different that the constraint does not apply. Another interesting case is the acquisition o f the subject particle ga. A s discussed i n section 4 . 2 , many Japanese children do not attempt to assign a single semantic function to the subject marker ga. This particle serves a wide range o f semantic functions i n adult speech, including agent, experiencer, recip­ ient i n benefactives w i t h morau 'receive', patient i n passive sentences, causee i n causative sentences, and what is sometimes analyzed as the object o f transitive stative predicates, such as suki ' l i k e ' . I n the early stages o f acquiring ga, some children seem to ignore its semantic functions, and simply place ga after the first noun i n their sentence. Apparently, one factor w h i c h limits the tendency toward semantic consistency is a tremendous amount o f " n o i s e " i n the input. W h e n a single semantic function cannot be found for a very frequent f o r m , children may abandon semantics and attempt to formulate a consistent syntactic (positional) hypothesis. 12.3. Other

Pacesetters

1 2 . 3 . 1 . Affect. For many years, research on language acquisition has em­ phasized the role o f cognition as a pacesetter o f linguistic development. Japanese data suggest that affect, another highly developed faculty w h i c h children bring to the task o f learning language, also plays a major role i n shaping the course o f language acquisition. M a n y o f the earliest acquisitions i n Japanese grammar

4.

The A c q u i s i t i o n of Japanese

511

have a basis i n the child's expression o f internal emotional states. I n all lan­ guages, early sensitivity to affect may contribute to the acquisition o f intonation patterns. I n Japanese, correct use o f many grammatical morphemes, such as sentence-final particles, depends upon the speaker's attitude or feelings, e.g. yo (assertive, potentially conflicting) and ne (seeking or expressing agreement). Ito (1981) has emphasized the role o f affect i n the acquisition o f negation i n Japanese. The emotional basis for use o f the different negative forms i n Jap­ anese, such as iya (rejection) and dame (prohibition), evolves during the prelinguistic period, Ito suggests, as an initially undifferentiated semantic complex of emotions involved i n insisting, demanding, and forbidding. As children ac­ quire different negative expressions, the emotional similarities among certain forms contribute to the many confusion errors w h i c h occur. Later acquisitions continue to be influenced by emotional factors. For exam­ ple, the use o f polite -mas inflections, w h i c h i n m y data first appeared i n contexts of assumed formality and role-playing, has special affective connotations w h i c h may be important i n their acquisition. A t about 3 - 4 years-of-age, when children are interacting i n social groups, sensitivity to affect is necessary for mastering the distinctions among forms o f first and second person reference, such as the difference between the ordinary masculine first person pronoun boku T and the rougher, more casual ore T . As errors w i t h the corresponding rough second person pronoun omae ' y o u ' show, affect may precede and over-ride social con­ straints i n young children's acquisition o f grammatical forms conditioned both by affect and by sociolinguistic factors. A l t h o u g h affect probably plays a major, as yet largely unexplored, role i n the acquisition o f all languages, Japanese children w i l l find more obvious and consistent grammatical correlates o f their emotions than children acquiring many other languages. 12.3.2. Presupposition. Somewhat more crosslinguistic information is available on the development o f pragmatics, and the early effects on language acquisition o f children's capacity for presupposition, another faculty w h i c h ap­ parently develops i n the pre-linguistic period. There is evidence from the data on Japanese acquisition that children are sensitive to linguistic expressions o f pre­ supposition from the earliest stage o f grammatical development. For example, presupposition underlies the use o f different w o r d orders i n Japanese, and varia­ tions i n w o r d order are used appropriately i n context from the t w o - w o r d stage. Presupposition also plays a part i n the acquisition o f sentence-final particles, such as yo, w h i c h typically presupposes that new information is being conveyed, and ne, w h i c h often presupposes shared information. The early acquisition o f presuppositional devices raises an interesting question w i t h respect to egocentrism, since they seem to require taking the addressee's point o f v i e w . Per­ haps, as Greenfield and Smith (1976, Chap. 4.3) and Bates (1976, Chap. 4) suggest, children first base apparent presuppositions about old/new information upon their o w n point o f v i e w , w h i c h i n the "here and n o w " o f early interactions

512

Clancy

often is the same as that o f the addressee. Careful investigation may show this to be true o f Japanese ellipsis and postposing rules. Affect and sensitivity to differ­ ent types o f speech acts may account for many o f the early uses o f sentence-final particles. However, this w o u l d be an interesting area for further study w i t h respect to presupposition, since appropriate use o f sentence-final particles does seem to require that the children be able to recognize, at least to some extent, the similarity or difference between their o w n state o f knowledge and/or attitude compared w i t h the addressee. Children's ability to recognize and make presup­ positions seems to be an important factor i n early language acquisition, and probably continues to influence the course o f development throughout later stages as w e l l .

13. S u g g e s t i o n s f o r F u r t h e r S t u d y 13.1.

Particles

Although there are many studies o f the acquisition o f case particles, it w o u l d be very helpful for crosslinguistic comparison i f detailed longitudinal data were analyzed i n terms o f the specific semantic functions served by each particle, since many have several distinct uses. For example, the particle ni marks stative and directional locatives, datives, agents i n passive and causative sentences, adverbials, etc. Since several, but not a l l , case particles are plurifunctional, a contrastive analysis o f their acquisition w o u l d serve as an interesting test case for Slobin's hypothesis that children prefer a one-to-one mapping o f form and mean­ ing. The particle e, for example, has the single function o f marking direction towards a location, w h i c h might lead children to prefer it to ni, w h i c h has so many other functions as w e l l . Such an analysis should provide insight into the factors w h i c h may influence the operation o f the one-to-one mapping principle, such as frequency. Since data on the acquisition o f locatives i n several languages are available, an analysis o f Japanese locative particles and constructions w o u l d be a valuable addition to the crosslinguistic data. I t w o u l d be especially useful i f the meth­ odology o f Johnston and Slobin (1979) could be followed i n a study o f the comprehension o f complex locatives, such as N no ue ni ' o n top o f N ' . Since the grammatical formation o f these locatives is very similar i n Japanese and T u r k i s h , Japanese data w o u l d provide a good test o f the cognitively based sequence o f development proposed by Johnston and Slobin. A detailed longitudinal analysis o f sentence-final particles, focusing on their functions i n different types o f speech acts, w o u l d be extremely interesting from a crosslinguistic standpoint. The discussion presented i n section 6.1.2, a com­ posite picture based upon several different children, provides only a few sug­ gestions about the strategies w h i c h individual children might follow i n acquiring sentence-final particles, such as associating a particular particle w i t h a certain

4.

The A c q u i s i t i o n of Japanese

513

type o f speech act, for example, yo w i t h contradictions. Since overt mor­ phological expression o f the pragmatic and emotional functions served by Ja­ panese sentence-final particles does not occur i n languages which have been extensively studied to date, such research should provide valuable insights into the role o f affect and presupposition i n the earliest stage o f language development. 13.2. Sociolinguistic

Features

From a crosslinguistic point o f v i e w , one o f the most interesting aspects o f Japanese is the extensive sociolinguistic conditioning o f grammatical forms. This is also an area o f Japanese acquisition w h i c h has not yet received much attention. One topic deserving further analysis is the acquisition o f polite -mas inflections, which w o u l d be interesting to study i n depth w i t h longitudinal data from a pragmatic and sociolinguistic standpoint. I n section 6.1.6 it was suggested that children may first associate -mas inflections w i t h the special affect o f certain ritualized, pseudo-formal contexts, and only later develop an understanding o f the relevant interpersonal factors, such as age and social status, w h i c h govern the selection o f -mas by adults. Future research might analyze the contextual factors conditioning the use o f polite style from the appearance o f -mas forms i n 2-yearolds to code-switching i n appropriate social contexts among older children. Although it w o u l d be valuable, o f course, to study children's use o f polite style in actual social interactions, role-playing techniques could probably elicit infor­ mation about children's knowledge o f the social factors governing polite style at a very early age. For example, the methodology o f Andersen (1978), i n w h i c h child and experimenter used puppets to act out interactions between such conver­ sational partners as doctor-patient, w o u l d probably yield extremely interesting results w i t h Japanese children. I t w o u l d then be interesting to compare findings about children's knowledge o f polite speech w i t h their o w n usage i n social contexts calling for polite inflections. As noted i n section 6 . 2 . 1 , the acquisition o f honorifics has not been investi­ gated longitudinally, probably because they are used so rarely, and are not mastered until the school years. Okuda's (1979) study suggests that role-playing might be a very effective methodology for analyzing children's knowledge o f honorifics as w e l l as polite style. The children i n Okuda's study were able to invent dialogue, incorporating honorifics i n some cases, for characters depicted in cartoons. Direct verbal role-playing, such as that used by Andersen, might tap unconscious knowledge more extensively and at an earlier age. Okuda's findings suggest that investigation o f the acquisition o f honorifics should extend i n age range at least through j u n i o r high school. Another major area for sociolinguistic investigation is the acquisition o f the differences between male and female speech i n Japanese. The studies o f the acquisition o f first and second person reference by Ide (1977, 1978, 1978-79) and Horiguchi (1979b), and H o r i ' s (1981) study o f sentence-final particles, are

514

Clancy

valuable contributions on this topic. Since Japanese is so rich i n male/female differences, detailed longitudinal research on the acquisition o f those areas o f grammar w i t h sex-based options w o u l d be an important addition to the crosslinguistic literature on the acquisition o f sex-appropriate speech. A g a i n , role-playing at different ages w o u l d probably be a valuable methodology for experimental investigations i n this area, especially for eliciting information about children's knowledge o f the speech o f the opposite sex. 13.3.

Input

Another interesting area for research is a comparison o f ordinary adult speech and the speech addressed to young Japanese children. The relevant comparison here should be the speech used among adult family members or close friends, to reduce the linguistic effects o f social distance between speaker and hearer. Miyazaki's (1979) study o f the acquisition o f wa and ga, Yamada (1980) and Cook (1985) have shown that Japanese input to young children has at least one distinctive characteristic w h i c h may have a direct bearing on language acquisi­ tion, namely, an extremely high frequency o f particle ellipsis. Other differences might include the frequency o f nominal and verbal ellipsis, o f particular verbal inflections, such as -chatta, the " v i v i d " past (see Rispoli, 1981b), and o f certain patterns o f w o r d order, such as repeated, postposed nouns. Since there are so many speech styles i n Japanese, the linguistic input to children w i l l be significantly different depending upon the composition o f their families. Given the differences between male and female speech, Japanese "motherese" may be quite distinct from father's input. I f a child lives i n an extended f a m i l y , s/he w i l l receive input from the grandparents, w h i c h w i l l be different from the speech o f younger people i n various ways, e.g. perhaps a higher frequency o f polite style i n grandmother's speech. I f children have sib­ lings o f the opposite sex, they may hear forms w h i c h they w o u l d not otherwise experience so directly, such as certain pronouns used i n the siblings' peer groups. I f there are older siblings o f the same sex, their input may encourage earlier acquisition o f sex-appropriate speech. One important question w i t h respect to input and Japanese language acquisi­ tion is Oedipal. I f the primary linguistic input to both boys and girls comes from their mothers, when and how do little boys learn to use masculine style, eliminat­ ing or reducing the frequency o f female forms heard from their mothers? Little girls do not have to undergo this shift i n identification, although they may reject at least certain parts o f the female linguistic model, as problems i n the acquisition of atashi T (female) i n certain cases indicate. Since sex-appropriate speech is so important i n Japan, mothers may adapt their speech according to the sex o f their child; it w o u l d be interesting to study what differences may exist between moth­ ers' speech to sons as compared to daughters. A thorough investigation o f the input to Japanese children should also consid­ er the role o f at least t w o other types o f input: the speech o f peers, and indirect or

4.

The A c q u i s i t i o n of Japanese

515

overheard input, not addressed to the child. Interaction w i t h peers, especially once the child has entered kindergarten and is exposed to large groups o f child­ ren, probably has significant effects upon the acquisition o f male/female speech, and register differences governed by in-group vs. out-group factors such as age. In the acquisition o f polite, and especially o f honorific speech, indirect or over­ heard input may be important. Okuda's (1979) study suggests that children have extensive knowledge about polite and honorific speech derived from their obser­ vation o f adult interactions. M y data suggest that at least some Japanese mothers engage i n direct role-playing w i t h polite speech, but children's observation o f mothers' speech to visitors and other adults may also be important. One other source o f indirect or passive input w h i c h deserves investigation is the television, which has apparently become a major source o f linguistic input for Japanese children. Okuda reports that there was a significant inverse correlation between the number o f hours per day spent watching television, and children's use o f polite speech i n various tasks; modern children's programs do not typically use the -mas style. Each o f the different sources o f input w i l l contribute to the range of registers i n children's repertoires, w h i c h w i l l be an important factor i n their ability to function successfully i n Japanese society.

LIST OF

ADJ Adjective AGR Seeking/expressing AGT Agent C O H R T Cohortative C O M P L Completed past CONC Concessive C O N T Continuative

agreement

COP Copula DIM Diminutive DO Direct object EMPH Emphatic EP Extended predicate E X C L A M Exclamatory GEN Genitive HON IMP INF

ABBREVIATIONS

Honorific Imperative Infinitive

10 Indirect object NEG Negative N O N P A S T Non-past tense O B L I G Obligation

516

Clancy

PASS Passive PAST Past tense P O L Polite PRES. P R O G Present progressive PRESUM Presumptive Q Question Q U O T Quotative REP Representative SUBJ Subject T O P Topic

ACKNOWLEDGMENTS I am grateful to the many scholars in Japan and the United States who helped me with this work, corresponding with me, sending advice, suggestions for references, and copies of their own research, especially Akiyo Asano, Kenji Hakuta, Kazuko Harada, Robert Hess, Sumiko Horiguchi, Sachiko Ide, Katsutoshi Ito, Megumi Kameyama, Barbara Lust, Kooji Murata, Ai Okubo, Akiko Okuda (Kawasaki), Matthew Rispoli, Taroo Takahashi, Ellen Tanouye, Tazuko Uyeno, Akiko Yamada, and Masako Yamamoto. I greatly appre­ ciate the time and trouble of those who commented on and corrected earlier versions of this work, including Ruth Berman, Pamela Downing, Kenji Hakuta, Morio Hamada, John Hinds, Katsutoshi Ito, Megumi Kameyama, Yoshiko Matsumoto, Ai Okubo, Dan Slobin, and Masayo Yamamoto. I especially appreciate the efforts of Katsutoshi Ito, Ai Okubo, and Masayo Yamamoto, who gave me many helpful examples and comments from their own data. I also wish to thank Haruo Aoki, Kiko Yamashita, and Mamoru Saito, who answered many questions about Japanese; Masayoshi Hirose, Yumiko Yoshimura, and Akiko Okuda (Kawaski), who helped me find Japanese references; and Yoshizo Itabashi, who helped me read many of them. I am grateful to all those who helped me with my own research on Japanese child language, especially Kazuko Harada, who found subjects for me, Akiyo Asano and Miyako Namikawa for their assistance at the recording sessions, and Janet Akaike-Toste, for her patient and painstaking transcription. Special thanks are due to Yukiko Kurihata (Nishimura) and her family, for their help and friendship during my stay in Tokyo, and to the mothers and children who participated in my research. The early stages of this work were carried out at the University of California at Berkeley; it was completed at the Center for Cognitive Science at Brown University. This research was supported in part by a Fulbright-Hays Doctoral Dissertation Re­ search Abroad Fellowship, a Social Science Research Council Postdoctoral Fellowship, a National Science Foundation National Needs Postdoctoral Fellowship, and an Alfred P. Sloan Postdoctoral Fellowship in Cognitive Science.

REFERENCES Akiyama, M. Yes-no answering systems in young children. Cognitive Psychology, 1979, 11, 485¬ 504. Aksu, A. Development of the expression of cause-effect relations in Turkish. Unpublished paper,

4.

T h e A c q u i s i t i o n of Japanese

517

Department of Psychology, University of California, Berkeley, 1975. (Abbreviated version available as Aksu, A. The acquisition of causal connectives in Turkish. Papers and Reports on Child Language Development (Department of Linguistics, Stanford University), 1978, 75, 129¬ 139.) Aksu-Kog, A. A . , & Slobin, D. I. The acquisition of Turkish. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associ­ ates, 1985. Alfonso, A . , & Niimi, K . Japanese: A basic course (5th revised ed.). Tokyo: Takayama Printing Co., 1972. Andersen, E . S. Will you don't snore please. Papers and Reports on Child Language Development (Department of Linguistics, Stanford University), 1978, 15, 140-150. Aoki, H. Lecture on the history of the Japanese language. Department of Linguistics, University of California, Berkeley, Spring quarter, 1981. Asano, A. On strategies for processing Japanese relative clauses. In Attempts in linguistics and literature: Papers in honor of Kazuko Inoue (Vol. 6). Society of Linguistics and Literature, International Christian University, 1979. Azuma, H . , Hess, R. D . , Kashigawa, K . , & Conroy, M. Maternal control strategies and the child's cognitive development: A cross-cultural paradox and its interpretation. Paper presented at the International Congress of Psychology, Leipzig, 1979. Bates, E . Language and context: The acquisition of pragmatics. New York: Academic Press, 1976. Bates, E . , & MacWhinney, B . A functionalist approach to the acquisition of grammar. In E . Ochs & B. Schieffelin (Eds.), Developmental pragmatics. New York: Academic Press, 1979. Benedict, R. The chrysanthemum and the sword. New York: Meridian, 1946 Berman, R. A. The acquisition of Hebrew. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Bever, T. G . The cognitive basis for linguistic structures. In J. R. Hayes (Ed.), Cognition and the development of language. New York: Wiley, 1970. Bloom, L . Language development: Form and function in emerging grammars. Cambridge, MA: MIT Press, 1970. Bloom, L . , Lahey, M . , Hood, L . , Lifter, K . , & Fiess, K . Complex sentences: Acquisition of syntactic connectives and semantic relations in complex sentences. Journal of Child Language, 1980, 7, 235-262. Bloom, L . , Lifter, K . , & Hafitz, J . Semantics of verbs and the development of verb inflection in child language. Language, 1980, 56, 386-412. Blount, B. G . Ethnography and caretaker-child interaction. In C. E . Snow & C . A. Ferguson (Eds.), Talking to children: Language input and acquisition. Cambridge: Cambridge University Press, 1977. Bowerman, M. Early syntactic development: A cross-linguistic study with special reference to Finnish. Cambridge: Cambridge University Press, 1973. Brown, D. H. Children's comprehension of relativized English sentences. Child Development, 1971, 42, 1923-1936. Brown, R. A first language: The early stages. Cambridge, MA: Harvard University Press, 1973. Carroll, J. B . , & Casagrande, J . B. The function of language classifications in behavior. In E . E . Macoby, T. M. Newcomb, & E . L . Hartley (Eds.), Readings in social psychology (3rd ed.). New York: Holt, Rinehart, and Winston, 1958. Caudill, W., & Weinstein, H. Maternal care and infant behavior in Japan and America. In T. S. Lebra & W. P. Lebra (Eds.), Japanese culture and behavior: Selected readings. Honolulu: University Press of Hawaii, 1974. Chafe, W. L . Givenness, contrastiveness, definiteness, subjects, topics, and point of view. In C . L i (Ed.), Subject and topic. New York: Academic Press, 1976. Chew, J . J . The structure of Japanese baby talk. The Journal of the Association of Teachers of Japanese, 1969, 6, 4-17.

518

Clancy

Chomsky, C . The acquisition of syntax in children from five to ten. Cambridge, MA: MIT Press, 1969. Clancy, P. M. Referential choice in English and Japanese narrative discourse. In W. L . Chafe (Ed.), The pear stories: Cognitive, cultural, and linguistic aspects of narrative production. Norwood, NJ: Ablex Publishing Corp., 1980. Clancy, P. M . , Jacobsen, T . , & Silva, M. The acquisition of conjunction: A cross-linguistic study. Papers and Reports on Child Language Development, (Department of Linguistics, Stanford University), 1976, 12, 71-80. Clark, E . Convention and innovation in acquiring lexicon. Papers and Reports on Child Language Development (Department of Linguistics, Stanford University), 1980, 19, 1-20. Clark, E . Lexical innovations: How children learn to create new words. In W. Deutsch (Ed.), The child's construction of language. New York: Academic Press, 1981. Clark E . , & Hecht, B. Learning to coin agent and instrument nouns. Cognition, 1982, 12, 1-24. Conroy, M . , Hess, R. D . , Azuma, H . , & Kashigawa, K . Maternal strategies for regulating child­ ren's behavior. Journal of Cross-Cultural Psychology, 1980, 11, 153-172. Cook, H. Minegishi. Frequency of nominal markers in the speech of a Japanese child and his caretakers: A case study. Descriptive and Applied Linguistics, (International Christian Univer­ sity), 1985,18, 13-24. Cross, T. C . Mothers' speech adjustments: The contribution of selected child listener variables. In C . E . Snow & C . A. Ferguson (Eds.), Talking to children: Language input and acquisition. Cambridge: Cambridge University Press, 1977. deVilliers, J. G . , Tager-Flusberg, H . , Hakuta, K . , & Cohen, M. Children's comprehension of relative clauses. Journal of Psycholinguistic Research, 1979, 8, 449-518. Doi, T. The anatomy of dependency. (Tr. by J . Bester.) New York: Harper and Row, 1973. (Originally Amae no koozoo. Tokyo: Kodansha.) Downing, P. Japanese numeral classifiers: A semantic, syntactic and functional profile. Unpublished doctoral dissertation, University of California, Berkeley, 1984. Eisenberg, A . , & Renner, T. Acquisition of complex sentences in English: Similarity and variation across children. Paper presented at the Sixth Annual Boston University Conference on Language Development, Oct., 1981. Ervin-Tripp, S. Discourse agreement: How children answer questions. In J . R. Hayes (Ed.), Cognition and the development of language. New York: Wiley, 1970. Fischer, J. L . Words for self and others in some Japanese families. American Anthropologist, 1964, 66 (pt. 2), 115-126. Fischer, J . L . Linguistic socialization: Japan and the United States. In R. Hill & R. Konig (Eds.), Families in East and West: Socialization process and kinship ties. The Hague: Mouton, 1970. Fujitomo, Y . Kodomo no kotoba no kenkyuu (I). Hokkaidoo Kyooiku Daigaku Kiyoo (Dai-ichibu), 1977, 28, 59-63. Fujiwara, Y . Yooji no Gengo Hyoogen Nooryoku no Hattatsu. Hiroshima: Bunka Hyoron Publishing Co., 1977. Givon, T. On understanding grammar. New York: Academic Press, 1979. Gleason, J . B. Talking to children; Some notes on feedback. In C . E . Snow & C . A. Ferguson (Eds.), Talking to children: Language input and acquisition. Cambridge: Cambridge University Press, 1977. Greenfield, P. M . , & Smith, J . H. The structure of communication in early language development. New York: Academic Press, 1976. Hahn, K . The development of negation in one Korean child. Unpublished doctoral dissertation, Department of Linguistics, University of Hawaii, 1981. Hakuta, K . Word order and particles in the acquisition of Japanese. Papers and Reports on Child Language Development (Department of Linguistics, Stanford University), 1977, 13, 110-117.

4.

T h e A c q u i s i t i o n of Japanese

519

Hakuta, K . Grammatical description vs. configurational arrangement in language acquisition: The case of relative clauses in Japanese. Cognition, 1981, 9, 197-236. Hakuta, K . Interaction between particles and word order in the comprehension and production of simple sentences in Japanese children. Developmental Psychology, 1982, 18, 62-75. Hakuta, K . , deVilliers, J . , & Tager-Flusberg, H . Sentence coordination in Japanese and English. Journal of Child Language, 1982, 9, 193-207. Harada, K . I. Acquisition of Japanese relative clauses: A case study on a two-year-old. Annual Reports (The Division of Languages, International Christian University, Tokyo), 1976, 1, 1-16. Harada, K . I. The acquisition of passive, causative, and te moraw constructions in Japanese: A case study on a two-year-old. Annual Reports (The Division of Languages, International Christian University, Tokyo), 1977, 2, 1-16. Harada, K . I. Notes on honorifics for seminar in crosslinguistic language acquisition, Prof. D. I. Slobin. The Summer Institute of The Linguistic Society of America, Albuquerque, New Mexico, 1980. (a) Harada, K . I. Notes on the acquisition of the genitive particle no for seminar in crosslinguistic language acquistion, Prof. D. I. Slobin. The Summer Institute of the Linguistic Society of America, Albuquerque, New Mexico, 1980. (b) Harada, K . I. The acquisition of Japanese case particles: A new look. In Y . Otsu, H . van Riemsdijk, K. Inoue, A. Kamio, & N. Kawasaki (Eds.), Studies in generative grammar and language acquisition. Tokyo: International Christian University, 1983. Harada, S. I. Honorifics. In M. Shibatani (Ed.), Syntax and semantics 5: Japanese generative grammar. New York: Academic Press, 1976. Harada, S. I . , Uyeno, T . , Hayashibe, H . , Yamada, H. On the development of perceptual strategies in children: A case study on the Japanese child's comprehension of the relative clause construc­ tion. Annual Bulletin of the Research Institute of Logopedics and Phoniatrics, (University of Tokyo) 1976, 10, 199-224. Hashimoto, S. Kokugohoo yoosetsu. Tokyo: Meiji Shoin, 1934. Hatano, E . Gengo kakutoku. In Yooji shinrigaku no shimpoo. Tokyo: Kaneko Shoboo, 1980. Hatano, K . Kotoba no hattatsu shinri. In E . Iwabuchi (Ed.), Kotoba no tanjoo: Ubugoe kara gosai made. Tokyo: Nihon Hoosoo Shuppan Kyookai, 1968. Hayashibe, H. Word order and particles: A developmental study in Japanese. Descriptive and Applied Linguistics, (International Christian University), 1975, 8, 1-18. Hess, R. D . , Kashigawa, K . , Azuma, H . , Price, G . , & Dickson, P. W. Maternal expectations for mastery of developmental tasks in Japan and the United States. International Journal of Psychol­ ogy, 1980, 75, 259-271. Hinds, J . Postposing in Japanese. Eoneo (The Journal of the Linguistic Society of Korea), 1976, 7, 113-125. Hinds, J . Ellipsis in Japanese. Papers in Japanese Linguistics, 1976-1977, 5, 63-97. Hinds, J . Ellipsis in Japanese. Edmonton, Alberta: Linguistic Research, Inc., 1982. Hirayama, M. Acquisition of Japanese particles by a child: Aged from one year and nine months to two years and two months. Unpublished bachelor's thesis, International Christian University, Tokyo, 1978. Hori, M. Shuujoshi no shuurui. In F . C . Peng (Ed.); Nihongo no danjosa: Male I female differences in Japanese. Tokyo: The East-West Sign Language Association, 1981. Horiguchi, S. Nenshooji no jukyuu hyoogen. In F . C . Peng & M. Hori (Eds.), Brain function and language acquisition: Kotoba no Hattatsu. Hiroshima: Bunka Hyoron Publishing Co., 1979. (a) Horiguchi, S. Nisaiji no hanashikotoba ni mitateru danjo no sai. Paper presented at the Sixth International Christian University Linguistics Symposium, August, 1979. (b) Horiguchi, S. Ichininshooshi, nininshooshi, koshooshi. In F . C . Peng (Ed.), Nihongo no danjosa: Male/female differences in Japanese. Tokyo: The East-West Sign Language Association, 1981. (a)

520

Clancy

Horiguchi, S. Nenshooji no asupekuto. In M. Hori & F . C . Peng (Eds.), Aspects of language acquisition: Gengo shuutoku no shosoo. Hiroshima: Bunka Hyoron Publishing Co., 1981. (b) Ide, S. Sex and personal referents of Japanese children. Proceedings of the Symposium on Japanese Sociolinguistics, Summer Institute of the Linguistic Society of America, Honolulu, University of Hawaii, July, 1977. Ide, S. Nihongo ni okeru seibetsu to ninsho daimeshi—yooji no baai. In F . C . Peng (Ed.), Development in verbal and nonverbal behavior: Hattatsu to shuutoku ni okeru gengo koodoo. Hiroshima: Bunka Hyoron Publishing Co., 1978. Ide, S. A sociolinguistic analysis of person references by Japanese and American children. Language Sciences, 1978-1979, 1, 273-293. Ide, S. Keigo. In F . C . Peng (Ed.), Nihongo no danjosa: Malelfemale differences in Japanese. Tokyo: The East-West Sign Language Association, 1981. Inoue, K . Japanese: A story of language and people. In T. Shopen (Ed.), Languages and their speakers. Cambridge, MA: Winthrop Publishers, 1979. Ito, K. Yooji ni totte 'ayamari' to wa nani ka. In F . C . Peng (Ed.), Development in verbal and nonverbal behavior: Hattatsu to shuutoku ni okeru gengo koodoo. Hiroshima: Bunka Hyoron Publishing Co., 1976. Ito, K . Towards an ethnopsychology of language: Interactional strategies of Japanese and Ameri­ cans. Bulletin of the Center for Language Studies (Kanagawa University), 1980, 3, 1-14. Ito, K . Two aspects of negation in child language. In P. S. Dale & D. Ingram (Eds.), Child language: An international perspective. Baltimore, MD: University Park Press, 1981. Iwabuchi, E . , & Muraishi, S. Kotoba no shuutoku. In E . Iwabuchi (Ed.), Kotoba no tanjoo: Ubugoe kara gosai made. Tokyo: Nihon Hoosoo Shuppan Kyookai, 1968. Johnston, J . R., & Slobin, D. I. The development of locative expressions in English, Italian, SerboCroatian, and Turkish. Journal of Child Language, 1979, 16, 531-547. Jorden, E . H . , with the assistance of H. I. Chaplin. Beginning Japanese (Parts 1-2). New Haven, CT: Yale University Press, 1962-1963. Kameyama, M. Development of verb use in Japanese: Early case frames. Unpublished paper, Department of Linguistics, Stanford University, 1982. Kamio, A . , & Harada, K . Children's processing of complex sentences in Japanese: An interim report of a repetition experiment. Kenkyuu hookoku: Nihongo no kihon koozoo ni kansuru rironteki-jisshooteki kenkyuu, International Christian University, 1979. Kamio, A . , & Harada, K . Children's processing of complex sentences in Japanese. In K . Inoue, E . Kobayashi, & R. Linde (Eds.), Issues in syntax and semantics: Festschrift for Prof Masatake Muraki. Tokyo: Sanshuusha, 1983. Kasahara, Y . Fear of eye-to-eye confrontation among neurotic patients in Japan. In T. S. Lebra & W. P. Lebra (Eds.), Japanese culture and behavior: Selected readings. Honolulu: The University Press of Hawaii, 1974. Komura, A. Shoyuu hyoogen no hattatsu. In M. Hori & F . C . Peng (Eds.), Aspects of language acquisition: Gengo shuutoku no shosoo. Hiroshima: Bunka Hyoron Publishing Co., 1981. Kuno, S. The structure of the Japanese language. Cambridge, MA: MIT Press, 1973. Kuno, S. Japanese: A characteristic OV language. In W. P. Lehmann (Ed.), Syntactic typology: Studies in the phenomenology of language. Austin: University of Texas Press, 1978. Kuno, S., & Kaburaki, E . Empathy and syntax. In S. Kuno (Ed.), Harvard studies in syntax and semantics (Vol. 1). Published by the Department of Linguistics, Harvard University, Cambridge, MA, 1975. Lieven, E . Different routes to multi-word utterances? Papers and Reports on Child Language Development (Department of Linguistics, Stanford University), 1980, 19, 37-44. Limber, J . The genesis of complex sentences. In T. Moore (Ed.), Cognitive development and the acquisition of language. New York: Academic Press, 1973.

4.

T h e A c q u i s i t i o n o f Japanese

521

Loveday, L . Communicative interference: A framework for contrastively analysing L 2 commu­ nicative competence exemplifed with the linguistic behavior of Japanese performing in English. International Review of Applied Linguistics, 1982, XXII, 1-16. Lust, B . , & Wakayama, T. K . The structure of coordination in first language acquisition of Jap­ anese. In F . Eckman & A. Hastings (Eds.), First and second language learning. Rowley, MA: Newbury House, 1979. Lust, B . , & Wakayama, T . K . Word order in first language acquisition of Japanese. In P. Dale & D. Ingram (Eds.), Child language: An international perspective. Baltimore, MD: University Park Press, 1981. MacWhinney, B. Hungarian language acquisition as an exemplification of a general model of grammatical development. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisi­ tion (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Martin, S. Lexical evidence relating Korean to Japanese. Language, 1966, 42, 185-251. Martin, S. A reference grammar of Japanese. New Haven, CT: Yale University Press, 1975. Matsumoto, Y . The child's acquisition of Japanese numeral classifiers. Paper presented at the International Congress for the Study of Child Language, Austin, July, 1984. (a) Matsumoto, Y . Hito—futa—mi vs. ichi—ni—san: Kodomo no suuchi—josuushi koozoo no keitai ni mini soosa gensoku. In F . C . Peng, K . Akiyama, & T . Kondo (Eds.), Gengo no dainamikkusu. Hiroshima: Bunka Hyoron Publishing Co., 1984. (b) Matsumoto, Y . Gengo shuutoku ni okeru wakan suushi keiretsu: Kisokusei to bunsekisei. Paper presented at the 90th Meeting of the Society for the Study of the Japanese Language, Tokyo, June, 1984. (c) McNeill, D . , & McNeill, N. B. What does a child mean when he says 'no'? In C . A. Ferguson & D. I. Slobin (Eds.), Studies of child language development. New York: Holt, Rinehart, & Winston, 1973. Miller, R. A. The Japanese language. Chicago: University of Chicago Press, 1967. (a) Miller, R. A. Old Japanese phonology and the Korean-Japanese relationship. Language, 1967', 43, 278-302. (b) Mills, A. The acquisition of German. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Miyahara, K . Language development in a young Japanese child—acquisition of particles. Unpublished paper, Kyuushu University, 1973. Miyahara, K . The acquisition of Japanese particles. Journal of Child Language, 1974, 1, 283-286. Miyazaki, M. The acquisition of the two particles wa and ga in Japanese—A comparative study of LI acquisition and L2 acquisition. Unpublished master's thesis, University of Southern Califor­ nia, 1979. Mizutani, O., & Mizutani, N. An introduction to modern Japanese. Tokyo: Japan Times, 1977. Murata, K . Ikujigo no kenkyuu: Yooji no gengo shuutoku no ichijooken to shite. Shinrigaku Ken­ kyuu, 1960, 31, 33-38. Murata, K . Gengo koodoo no hattatsu I: Issaiji-K no danwa ni okeru bun. Japanese Journal of Psychology, 1961, 32, 1-13. (a) Murata, K . Gengo koodoo no hattatsu II: Hahaoya no kansatsu kiroku ni motozuku issaiji no danwa no hattatsuteki kenkyuu. Japanese Journal of Psychology, 1961, 32, 10-23. (b) Murata, K . Gengo koodoo no hattatsu III: Yookyuu katsuwa no gengo keishiki narabi ni kinoo no shoki hattatsu katei. Japanese Journal of Educational Psychology, 1961, 9, 32-41. (c) Murata, K . Gengo koodoo no hattatsu V: Shoki rengo hatsuwa no koozoo bunseki no kokoromi. Japanese Journal of Psychology, 1965, 36, 67-75. Murata, K . Yooji no gengo hattatsu. Tokyo: Baifuukan, 1968. Murata, K . , & Ohara, T. Gengo koodoo no hattatsu VII: Issaiji to no 'taiwa' ni okeru hahaoya no yakuwari. Japanese Journal of Psychology, 1966, 37, 67-73.

522

Clancy

Newport, E . L . , Gleitman, H . , & Gleitman, L . R. Mother, I'd rather do it myself: Some effects and non-effects of maternal speech style. In C . E . Snow & C . A. Ferguson (Eds.), Talking to children: Language input and acquisition. Cambridge: Cambridge University Press, 1977. Nöda, M. An analysis of the Japanese extended predicate: A pragmatic approach to the system and pedagogical implications. Unpublished master's thesis, Cornell University, 1981. Noji, J. Yoojiki no gengo seikatsu no jittai (Vols. 1-4). Hiroshima: Bunka Hyoron Publishing Co., 1974-1977. Okayama, Y . Ko to haha no taiwa ni okeru 'kotoba' no kinooteki bunrui—2, 3, 4-saiki no hikaku. Oosaka Shiritsu Tanki Daigaku Kyookai Kenkyuu Hookokushuu, 1979, 12, 109-122. Okubo, A. Yooji no joshi no hattatsu—ichiyooji no choosa kara. Kokugogaku 55. Musashi no Shoin, 1963. Okubo, A. Yooji no gimon hyoogen no keishiki to sono hattatsu—yooji no gengo choosa 3. Kokuritsu Kokugo Kenkyuujo Ronshuu 2, Kotoba no Kenkyuu Dainishuu. Shuuee Shuppan, 1965, 255-271. Okubo, A. Yooji gengo no hattatsu. Tokyo: Tokyodoo, 1967. Okubo, A. Yooji no bunkoozoo no hattatsu: Sansai—rokusaiji no baai. Kokuritsu Kokugo Ken­ kyuujo Hookoku 50, 1973. Okubo, A. Koobun no hattatsu. In J. Orii et al. (Eds.), Kotoba no hattatsu to sono shoogai. Osaka: Osaka Kyooiku Daigaku, 1976. Okubo, A. Gengo shuutoku no kojinsa—kyoomai ni arawareta chigai. Yoonen Jidai, 1980,10, 22¬ 31. Okubo, A. Gengo shuutoku no hooryaku. In M. Hori & F . C . Peng (Eds.), Aspects of language acquisition: Gengo shuutoku no shosoo. Hiroshima: Bunka Hyoron Publishing Co., 1981. (a) Okubo, A. Kotoba no kakutoku to onmanego. In M. Hori & F. C . Peng (Eds.), Aspects of language acquisition: Gengo shuutoku no shosoo. Hiroshima: Bunka Hyoron Publishing Co., 1981. (b) Okubo, A. Dooshi to sono kasetsugo no hattatsu no jittai: Danji no nisai kara sansai zenban made. Kokuritsu Kokugo Kenkyuujo Hookoku 74: Kenkyuu Hookokushuu 4, 1983. Okuda, A. Kodomo to taiguu hyoogen. Unpublished master's thesis, Tsukuba University, 1979. Peters, A. Language segmentation: Operating principles for the perception and analysis of language. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Pye, C . Mayan motherese. Paper presented at the Seventh Annual Boston University Conference on Language Development, Oct., 1982. Radulovic, L . Acquisition of language: Studies of Dubrovnik children. Unpublished doctoral disser­ tation, University of California, Berkeley, 1975. Rispoli, M. The pragmatic and paradigmatic in language acquisition: A case study from Japanese. Unpublished paper, University of Pennsylvania, 1981. (a) Rispoli, M. The emergence of verb and adjective tense-aspect inflections in Japanese. Unpublished master's thesis, University of Pennsylvania, 1981. (b) Sanches, M. Features in the acquisition of Japanese grammar. Unpublished doctoral dissertation, Stanford University, 1968. Sanches, M. Recent publications in Japanese on language acquisition. Anthropological Linguistics, 1974, 13, 120-127. Sanches, M. Language acquisition and language change: Japanese numeral classifiers. In B. G . Blount & M. Sanches (Eds.), Sociocultural dimensions of language change. New York: Academ­ ic Press, 1977. Sano, K . An experimental study on the acquisition of Japanese simple sentences and cleft sentences. Descriptive and Applied Linguistics, (International Christian University), 1977, 10, 213¬ 233.

4.

T h e A c q u i s i t i o n of Japanese

523

Savic, S. Aspects of adult-child communication: The problem of question acquisition. Journal of Child Language, 1975, 2, 251-260. Sheldon, A. The role of parallel function in the acquisition of relative clauses in English. Journal of Verbal Learning and Verbal Behavior, 1974, 13, 272-281. Shibamoto, J . Language use and linguistic theory: Sex-related variation in Japanese syntax. Unpublished doctoral dissertation, University of California, Davis, 1979. Slobin, D. I. The acquisition of Russian as a native language. In F . Smith & G . A. Miller (Eds.), The genesis of language: A psycholinguistic approach. Cambridge, MA: MIT Press, 1966. Slobin, D. I. Cognitive prerequisites for the development of grammar. In C . A. Ferguson & D. I. Slobin (Eds.), Studies of child language development. New York: Holt, Rinehart, & Winston, 1973. Slobin, D . I . Universal and particular in the acquisition of language. In E . Wanner & L . R. Gleitman (Eds.), Language acquisition: The state of the art. Cambridge: Cambridge University Press, 1982. Slobin, D . I . Crosslinguistic evidence for the Language-Making Capacity. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum Associ­ ates, 1985. Slobin, D. I . , & Bever, T. G . Children use canonical sentence Schemas: A crosslinguistic study of word order and inflections. Cognition, 1982, 12, 229-265. Smith, M. Relative clause formation between 29-32 months: A preliminary report. Papers and Reports on Child Language Development (Department of Linguistics, Stanford University), 1974, 8, 104-110. Smith, N. The acquisition of phonology: A case study. Cambridge: Cambridge University Press, 1973. Swan, M. Yamamoto. The semantic development of negation in English and Japanese in a simul­ taneously bilingual child. Unpublished master's thesis, University of Hawaii, 1981. Takahara, P. O., & Peng, F . C . Gojun toochi. In F . C . Peng (Ed.), Nihongo no danjosa: Male/female differences in Japanese. Tokyo: The East-West Sign Language Association, 1981. Takahashi, T. Yoojigo no keitaironteki na bunseki. Kokuritsu Kokugo Kenkyuusho Hookoku 55, 1975. (a) Takahashi, T. Yoojigo no rengoronteki na bunseki—Meishi no kaku no yoojoo. Kokuritsu Kokugo Kenkyuujoo, Gengo Taikei Kenkyuubu Daiichi Kenkyuushitsu, 1975. (b) Tamori, I. NP and particle deletion in Japanese discourse. In E . O. Keenan & T. L . Bennett (Eds.), Discourse across time and space. Published by the Department of Linguistics, University of Southern California, 1977. Tanouye, E . K . The acquisition of verbs in Japanese children. Papers and Reports on Child Lan­ guage Development (Department of Linguistics, Stanford University), 1979, 17, 49-56. Tanouye, E . K . Language development in two Japanese-speaking children: A cross-linguistic study. Unpublished qualifying paper, Teacher's College, Columbia University, 1980. Tavakolian, S. The conjoined-clause analysis of relative clauses. In S. Tavakolian (Ed.), Language acquisition and linguistic theory. Cambridge, MA: MIT Press, 1981. Tyack, D . , & Ingram, D. Children's production and comprehension of questions. Journal of Child Language, 1977, 4, 211-224. Ueda, K . Sixteen ways to avoid saying 'no' in Japanese. In J. C . Condon & M. Saito (Eds.), Intercultural encounters. Tokyo: Simul Press, 1974. Uyeno, T. A study of Japanese modality: A performative analysis of sentence particles. Unpublished doctoral dissertation, University of Michigan, Ann Arbor, 1971. Uyeno, T . , & Harada, S. I. Perception of syntactic structure in Japanese. Annual Bulletin of the Research Institute of Logopedics and Phoniatrics (University of Tokyo), 1975, 9, 171-192. Uyeno, T . , Harada, S. I . , Hayashibe, H . , & Yamada, H. Comprehension of sentences with giving

524

Clancy

and receiving verbs in Japanese children. Annual Bulletin of the Research Institute of Logopedics and Phoniatrics (University of Tokyo), 1978, 12, 167-185. Yamada, A. The development of syntactic devices for communication in the early stages of the acquisition of Japanese: A case study. Unpublished master's thesis, University of Hawaii, 1980. Yamamoto, M. (See Swan 1981.) Yoshida, K . The acquisition of question and request forms in a Japanese child. Paper presented at the Second Annual Boston University Conference on Language Development, Oct., 1977.

The Acquisition of Kaluli

Bambi B. Schieffelin University

of

Pennsylvania

Contents Kaluli Language and Cultural Context Overall Sketch of Development 534 Word Order 543 Personal Pronouns 549 Nominal Casemarking 555 Verbs 566 Emphatic and Other Discourse Particles Conclusions 589

525

586

KALULI LANGUAGE A N D CULTURAL CONTEXT

1. K a l u l i L a n g u a g e D e s c r i p t i o n

K a l u l i is one o f the more than 700 languages spoken i n Papua N e w Guinea. I t is a non-Austronesian language w h i c h is part o f the Central and South N e w Guinea Stock o f the Trans-New Guinea P h y l u m (Voorhoeve, 1975). W i t h i n this Phylum are 67% o f all Papua N e w Guinea languages, and 82% o f all speakers o f Papua New Guinea languages. O n the basis o f w o r d lists and phonological inventories (Shaw, 1973), K a l u l i has been classified as one o f the five languages o f the Bosavi family. There are four dialects o f K a l u l i , but they w i l l not be discussed i n this chapter. I n several important ways K a l u l i resembles other non-Austronesian languages that have been described for Papua N e w Guinea. I t is a verb-final language marking case relations postpositional^ on nouns. Subject marking is suffixed on the verb through bound person markers and follows a nominative-accusative pattern. One bound person marker may denote t w o or several different persons. 525

526

Schieffelin

This is common, especially i n distinguishing 1st from 2 - 3 person, but there is variation i n the pattern depending on sentence type. L i and Lang (1979, p . 109) point out that "those Papuan languages w h i c h have case systems are mostly ergative." I n K a l u l i there are t w o casemarking systems for nouns, neutral and ergative/absolutive. Casemarking on nouns follows a semantically motivated split ergative system. I n terms o f marking the semantic functions o f nouns, K a l u l i has a mixed word-order and inflectional system. That is, sometimes w o r d order and other times casemarking is used to indicate semantic function. K a l u l i has t w o allowable w o r d orders for agent and object, A O V and O A V , each o f which is used for different pragmatic ends. A s w i t h other split ergative languages found i n this area, there is no passive. L i k e other languages i n this P h y l u m , there is considerable morphological complexity and irregularity i n the verb system. The complex morphological system has elaborate affixation and inflection. Transparency is often veiled through morphophonemic changes. U n l i k e some o f the other languages i n the Trans-New Guinea P h y l u m , there is no object marking i n the verb through bound markers, but the verb may carry information about the action (frequency and duration). K a l u l i nouns are not marked for gender, nor are they morphologically marked for plurality or definiteness. There are t w o sets o f independent personal pronouns known as nonfocused and focused pronouns. Nonfocused pronouns are m u l t i ­ functional and are used as subjects and experiencers o f intransitive verbs, agents of transitive verbs i n A O V utterances, and as objects. Focused pronouns are found i n ( O ) A V utterances, when the agent is i n focus. Some o f these pro­ nominal forms are marked for number (singular, dual, plural), inclusive, ex­ clusive, as w e l l as for other pragmatic and semantic properties. Subordinate clauses precede the main clause. However, unlike English, which has a variety o f lexical conjunctions such as and, before, because, when, and if to j o i n clauses, K a l u l i uses a variety o f morphosyntactic constructions, called medial verb constructions. These, i n relation to final inflected verb con­ structions, express the meanings o f purposive and temporal conjunctions. These constructions, more than any other, give Papuan languages a characteristic stamp as different from those spoken i n other parts o f the w o r l d and b u i l d the types o f sentences used i n "clause-chaining" (Haiman, 1979; Longacre, 1972; Olson, 1978). Through additional morphological markers on the verb, speakers must signal whether the f o l l o w i n g clause w i l l have the same or different subject. This system, SWITCH REFERENCE, is found i n most Papuan languages. Possession is indicated i n several ways: through casemarking suffixes on nouns and demonstratives, through possessive adjectives, and through modified forms o f the personal pronouns. The genitive construction typically has the order Genitive + Head. U n l i k e the more common pattern o f noun phrase ordering i n verb-final languages (as described by Greenberg, 1966), modifiers follow the

5.

The A c q u i s i t i o n of Kaluli

527

head noun. A l l adverbial modifiers o f the verb precede i t . W h e n an adjective follows the noun, it carries all o f the inflectional material o f the noun phrase. However, i n contrast to English, K a l u l i has relatively few adverbs and adjectives in its lexicon. Negation is signaled lexically and by both prefixing and suffixing. Nominals are negated w i t h one set o f suffixes. Verbal negation varies according to the form o f the verb, w i t h imperatives taking only suffixes, and other forms taking a variety o f prefixes and suffixes, depending on pragmatic as w e l l as structural constraints. K a l u l i employs an elaborate system o f emphatic and evidential particles. These particles are suffixed i n word-final and sentence-final position, w i t h evi­ dential particles preceding emphatic particles when they co-occur. These parti­ cles encode nonrelational meanings, but are extremely important i n conveying information about how speakers feel about what they are saying, and how they know about what they are saying. K a l u l i allows a great deal o f deletion and ellipsis i n all genres o f talk. Utter­ ances may consist o f a single verb, or a verb w i t h one or more other sentence constituents. W h e n a person opens a discourse all major NPs are usually spec­ ified, but i f one N P does not change, and there is no likelihood o f ambiguity, that NP w i l l probably not be repeated. 1.1. Linguistic

Research

on

Kaluli

L i k e most o f the other languages spoken i n Papua N e w Guinea, there is no written grammar o f K a l u l i . The first w o r k on this language was done by mission­ ary linguist M u r r a y Rule i n the mid-1960s. Rule, w o r k i n g through translators speaking related languages, spent 6 weeks i n the Bosavi area and put together a short preliminary grammatical sketch. This was designed to help Australian missionaries who w o u l d be sent into Bosavi, and was the first time the language was written. Since that time the only other linguistic w o r k i n Bosavi has been carried out by myself while doing ethnographic and linguistic fieldwork on the development of communicative competence ( 1 9 7 5 - 1 9 7 7 ) . Additional insights and assistance were provided by E . L . Schieffelin and S. Feld i n the course o f their an­ thropological fieldwork among the K a l u l i . The linguistic analyses which follow must be considered w o r k i n progress since additional fieldwork is necessary to untangle a number o f remaining linguistic puzzles. Learning a language for w h i c h there is no grammatical description o f the adult language does not present any serious problems to the child who is acquiring it as a first language. Describing the language acquisition processes o f a language for which there is no grammatical description does present a serious challenge to the researcher. One not only must w o r r y about how one is getting to one's destina­ tion, but what the destination itself looks like.

528

Schieffelin

W o r k i n g back and forth between adult and child language use i n constructing grammatical and pragmatic analyses o f K a l u l i has been an instructive task. There is much to be learned from child language about the structure and function o f linguistic systems. Through children's misanalyses and oversegmentations, un­ derlying forms are made apparent. Systematic error patterns reveal the distribu­ tion o f pragmatic forms w h i c h otherwise w o u l d be difficult to elicit. A d u l t misunderstandings, corrections, and laughter offer insights into what is and is not grammatically acceptable and pragmatically appropriate language. These are some o f the essential cues that a non-native speaker must rely on i n doing this type o f research i n a monolingual situation. 1.2. Data Base for Language

Acquisition

This study represents the first child language acquisition research carried out in Papua N e w Guinea. For the present analysis I have examined t w o types o f data: diary notes and observations made from November 1975 until A p r i l 1977 on a number o f K a l u l i children o f different ages and the longitudinal taperecorded and transcribed conversations I collected from three K a l u l i children and members o f their families from January 1976 until A p r i l 1977. A t the beginning of the study these three children, t w o boys ( A b i and Wanu) and one g i r l ( M e l i ) , were about 24 months o l d . One child (Wanu) was using mostly single words, and the other t w o were using single and m u l t i w o r d utterances. Spontaneous speech samples were collected at 4 - 6 week intervals and averaged over 3 hours per sample. Eight samples were recorded for t w o children and seven samples were recorded for the third child. I n addition to the three focus children, several samples were taken o f a fourth c h i l d , Suela, A b i ' s half sister, w h o was several months younger than A b i . The language was recorded i n the everyday familial settings i n w h i c h i t was being learned, so as to document the contexts o f language acquisition as w e l l as the acquisition process itself. The language that comprises these samples there­ fore is naturalistic and spontaneous, and captures the communication o f the child w i t h those individuals, both adult and c h i l d , w i t h w h o m he or she regularly interacted. W o r k i n g from an ethnography o f speaking framework, extensive contextual notes and ethnographic observations were recorded both during and after recordings. I n addition, in-depth interviews were carried out concerning general socialization practices and beliefs about language acquisition. This infor­ mation was used to understand and interpret the meanings o f utterances as w e l l as the meanings o f interactions. After recording each sample, the mothers o f the children and other native assistants helped me understand the language that was being spoken. This lengthy transcription process included the discussion o f linguistic as w e l l as cultural material. A total o f 83 hours o f taperecorded family interactions were transcribed and annotated. These tapes, along w i t h tape recordings o f other types

5.

The A c q u i s i t i o n of Kaluli

529

of verbal genres (e.g. traditional and personal stories) supplemented by more formal linguistic elicitation w i t h adults served as the basis for the current l i n ­ guistic analysis o f K a l u l i . (For details about method and samples see Schieffelin, 1979.) Because the framework o f this study was the spontaneous use o f language in naturalistic contexts, the findings reported here concern language production. A t this point there are no comprehension data available. 1.3. Cultural

Sketch

The K a l u l i people (population 1200) live i n a tropical rain forest just north o f the slopes o f M t . Bosavi, on the Great Papuan Plateau i n the Southern Highlands Province o f Papua N e w Guinea. They are one o f four small groups who collec­ tively refer to themselves as Bosavi kalu 'people o f Bosavi' and to their language as Bosavi to 'Bosavi language'. Except for a few o f the young men w h o have been out o f the area (Bible school or through contract labor on plantations) and have learned some T o k Pisin or other local languages, the majority o f K a l u l i speakers are monolingual. They are traditionally nonliterate, though i n the last decade, a few individuals have become literate i n K a l u l i and T o k Pisin through their association w i t h the Christian mission i n the area. The Bosavi people live i n about 20 longhouse communities, each separated by an hour or so w a l k i n g distance through primary and secondary rain forest. Villages are composed o f 6 0 - 9 0 individuals who traditionally lived i n one large longhouse w i t h no internal walls. Currently (1982), w h i l e the longhouse is maintained, many families are l i v i n g together i n smaller dwellings so that one or more extended families may live together. I t is not unusual then for at least a dozen individuals o f different ages to be l i v i n g together i n one house w h i c h consists essentially o f one semipartitioned r o o m . (See E . L . Schieffelin 1976 for detailed ethnographic account.) K a l u l i are swidden horticulturalists maintaining large gardens o f pandanus, bananas, and breadfruit. Their staple starch is sago derived from w i l d palms. They hunt and fish extensively as w e l l as keep domestic pigs. K a l u l i society is generally egalitarian, lacking the " b i g m a n " patterns o f social organization so common i n the N e w Guinea Highlands. M e n and women utilize extensive net­ works o f obligation and reciprocity i n the organization o f w o r k and sociable interaction. Everyday life is overtly focused around verbal interaction; K a l u l i think o f and use talk as a means o f control, manipulation, expression, assertion, and appeal. T a l k gets you what you want, need, and feel owed. T a l k is a primary indicator o f social competence and a primary way to be social. Learning how to talk and to become independent are major goals o f socialization. Customarily K a l u l i observe a division o f labor, and men and women's ac­ tivities are for the most part cooperative, complementary, but usually separate. Child caregiving is exclusively i n the domain o f women. Since i t is important to understand something o f the cultural basis for the ways i n w h i c h K a l u l i act and

530

Schieffelin

speak to their children, I w i l l sketch selected aspects o f the cultural dimensions of language acquisition.

1.4.

The Cultural

Context

of Language

Acquisition

K a l u l i describe their infants as helpless, 'soft' (taiyo) and 'having no under­ standing'. They take care o f them, they say, because they feel sorry for them. Mothers, who are the primary caregivers i n the first 3 years, are attentive to their infants and physically responsive to them. However, while nursing their infants they are often involved i n other activities such as cooking, making netted bags, and conversing w i t h others. Given their belief that infants 'have no understand­ i n g ' , K a l u l i do not treat them as partners (speaker/addressee) i n dyadic commu­ nicative interactions. W h i l e mothers greet their young infants by name and use expressive vocalizations, they rarely address other types o f utterances to them. Mothers encourage sibling interaction w i t h infants by holding the infant up, faced outwards w h i l e they speak for the infant i n a high-pitched nasalized voice. These triadic exchanges, w h i c h continue until an infant is between 4 - 6 months, are primarily for the benefit o f the older sibling and help create a relationship between the t w o from an early age. I t is important to point out that i n these exchanges, the mother's utterances "as the b a b y " are not based on, nor do they originate w i t h anything that the infant has initiated—either vocally or gesturally. After a l l , how could someone w i t h 'no understanding' initiate appropriate i n ­ teractional sequences? However, there is an even more important and enduring cultural construct that helps make sense out o f the mothers' behaviors i n these situations and many others as w e l l . K a l u l i say that "one cannot know what another thinks or feels." N o w , while K a l u l i obviously interpret and assess one another's available behav­ iors and internal states, these interpretations are not culturally acceptable as topics o f talk. Individuals often talk about their o w n feelings. However there is a cultural dispreference for talking about or making claims about what another might think or feel, or what another is about to do, especially i f there is no evidence. These culturally constructed behaviors, w h i c h are related to other forms o f language use, have important consequences for the ways i n w h i c h K a l u l i caregivers verbally interact w i t h their children. As infants become older ( 6 - 1 2 months) they are addressed by adults to a limited extent. They are greeted by a range o f names and receive a limited set o f both negative and positive imperatives. Rhetorical questions such as " I s i t yours?!" (meaning, i t is not yours!) are used when a child reaches for something that is not theirs to take. The language addressed to the pre verbal child consists largely o f " o n e - l i n e r s " w h i c h call for no verbal response. Either an action or termination o f an action is the appropriate reaction. Other than these utterances, very little connected discourse is directed to the young child by adult caregivers.

5.

The A c q u i s i t i o n of Kaluli

531

This pattern o f adults not treating infants as communicative partners continues even when babies begin babbling. K a l u l i recognize babbling (dabedan) but say that this vocal activity is not communicative and has no relationship to speech that w i l l eventually emerge. Adults and older children occasionally repeat vo­ calizations back to the young child (ages 1 2 - 1 6 months), reshaping them into the names o f persons i n the household or into kinterms. However, adults do not expect that children this age w i l l be able to produce closer approximations o f these names, and there is no pressure on them to do so. Thus, throughout the pre verbal period very little talk is directed to the c h i l d , except for imperatives, rhetorical questions, and greetings. A child w h o by K a l u l i terms has not yet begun to speak is not expected to respond either vocally or verbally. What this means is that i n the first 18 months or so very little sustained dyadic verbal exchange occurs between adult and infant. A l l this does not mean that K a l u l i children grow up i n an impoverished verbal environment. Quite the opposite! The verbal environment o f infants is rich and varied, and from the very beginning infants are surrounded by adults and older children who spend a great deal o f time talking to one another about what they are doing. I n addition, the infants' on-going activities such as standing and what they are eating are topics o f talk between members o f the household. A l l o f this talk about the here-and-now is available to the infant though it is not talk addressed or directed to the infant. For example, i n referring to the infant's actions, siblings and adults w i l l use the infant's name or a kinterm. They w i l l say, " L o o k at Seligiwo! He's w a l k i n g by h i m s e l f . " Thus children may learn from these contexts to attend to and learn from the verbal environment i n w h i c h they live. Every society has its o w n ideology about language, including when it begins and how children acquire i t . The K a l u l i are no exception. K a l u l i claim that language begins at the time when the child uses t w o critical words, no 'mother' and bo 'breast'. The child may be using other single words, but until these t w o words are used, the beginning o f language is not recognized. Once a child has used these t w o words, a whole constellation o f interrelated behaviors is set into motion. K a l u l i claim that once a c h i l d has begun to use language, he or she then must be 'shown how to speak' (to widari) (Schieffelin, 1979). K a l u l i show their children language by using a teaching strategy. They provide a model for what the child is to say, followed by the w o r d elema, an imperative meaning 'say like that'. Mothers use this method o f direct instruction to teach the social uses o f assertive language (teasing, shaming, requesting, challenging, reporting). H o w ­ ever, object labeling is never part o f an elema sequence, nor does the mother ever use elema to instruct the child to beg or appeal for food or object. Begging, the K a l u l i say, is natural for children. They k n o w how to do i t , and need not be taught. I n contrast, a child must be taught to be assertive through the use o f particular linguistic expressions and verbal sequences.

532

Schieffelin

A typical sequence using elema is triadic, involving the mother, child (be­ tween 2 0 - 3 6 months) and other participant(s). For example, Meli 26 months, her cousin Mama 40 months, and her mother are sitting around the house. The two girls have been talking and teasing each other about a large carton in the room. 1

Mama —> Meli:

ge you 4

Meli —» Mama:

bokisi box

-ya diefeno L O C put:l:FUT

1 will put you in the box.'

ne I/me

bokisi

-ya

diefeno

? T will put in the box' Mother —» Meli =#> Mama:

gi your

bokisi box

-yo NEUTRAL

he?! what about

elema. say: IMP 'What about your box?!' say like that. Meli —» Mama:

gi

bokisi

-yo

he?!

'What about your box!' Mama (pointing to box) —> Meli:

giyolo yours

he?! what about

'What about yours?!' Mother —> Meli =#> Mama:

niyo mine

dalab have:3:PRES

elema. say:IMP

'I have mine' say like that. Meli —> Mama:

niyo

dalab

T have mine' Mother —> Meli =#> Mama:

niyo mine

halo up there

elema say:IMP

'Mine is up there' say like that

!Ages for all children are given in months and weeks. For example, 25.3 means 25 months and 3 weeks. Transcription conventions for speaker/hearer in triadic exchanges are as follows: single arrow —> indicates addressee, double arrow =#> indicates intended addressee.

5.

Meli —» Mama: niyo mine

niyo

hado-wo raw Q

he? huh?

The A c q u i s i t i o n of Kaluli

Mama!

533

giyo hado-wo yours raw Q

'Mine, mine is raw? huh? Mama! is yours raw?' (Mama and Mother laugh at Meli's mishearing and Meli starts another teasing routine, claiming that Mama's sweet potatoes (not present) are not cooked.) I n this situation, as i n many others, the mother does not modify her language to fit the linguistic ability o f the young child. Instead her language is shaped so as to be appropriate ( i n terms o f form and content) for the child's intended addressee. Consistent w i t h the ways she interacts w i t h her infant, what a mother instructs her young child to say usually does not have its origins i n any verbal or nonverbal behavior o f the child, but i n what the mother thinks should be said. The mother pushes the child into ongoing interactions that the child may or may not be interested i n and w i l l at times spend a good deal o f energy i n trying to get the child verbally involved. This is part o f the K a l u l i pattern o f fitting (or pushing) the child into the situation, rather than changing the situation to meet the interests or abilities o f the child. Thus mothers take a directive role w i t h their young children, teaching them what to say so that they may become participants i n the social group. I n addition to instructing their children by telling them what to say i n exten­ sive interactional sequences, K a l u l i mothers pay attention to the form o f their children's utterances. They w i l l correct the phonological, morphological, or lexical form o f an utterance or its pragmatic or semantic meaning. Since the goals o f language acquisition include a child becoming competent, independent, and mature sounding i n his language, K a l u l i use no Baby Talk lexicon, for they said (when I asked about it) that to do so w o u l d result i n a child sounding babyish which was clearly undesireable and counter-productive. The entire process o f a child's development, o f w h i c h language acquisition plays a very important role, is thought o f as a hardening process and culminates i n the child's use o f " h a r d w o r d s , " speech that is well-formed and situationally appropriate (Feld & Schieffelin, 1982). The cultural dispreference for saying what another might be thinking or feel­ ing has important consequences for the organization o f dyadic exchanges be­ tween child and caregiver. For one, i t affects the ways i n w h i c h meaning is negotiated during an exchange. For the K a l u l i the responsibility for clear ex­ pression is w i t h the speaker, and c h i l d speakers are not exempt from this. Rather than offering possible interpretations or guessing what a child is saying or mean­ ing, caregivers make extensive use o f clarification requests such as " h u h ? " and " w h a t ? " i n an attempt to elicit clearer expression from the child. Since responsi­ bility for communication lies w i t h the speaker, children are also instructed w i t h elema to request clarification (using similar forms) from others when they do not understand what someone is saying to them.

534

Schieffelin

Another important consequence o f not saying what another thinks is the absence o f adult expansions o f c h i l d utterances. K a l u l i caregivers w i l l put words into their children's mouths, but these words originate w i t h the caregiver. They rarely elaborate or expand utterances initiated by the child. N o r do they j o i n t l y build propositions across utterances and speakers except i n the context o f se­ quences w i t h elema i n w h i c h they are constructing talk for the child. A l l o f these patterns o f early language use, such as the lack o f expansions or verbally attributing an internal state to an individual, are consistent w i t h cultural conventions o f adult language usage. The K a l u l i avoid gossip, and often indicate the source o f information they report. They make extensive use o f quoted speech in a language that does not allow indirect quotation. They utilize a range o f evidential markers i n their speech to indicate the source o f speakers' information: for example, whether something was said, seen, heard, or gathered from other kinds o f evidence. These patterns are also found i n early child speech and affect the organization and acquisition o f conversational exchanges i n this society. (See Ochs & Schieffelin, 1984.)

O V E R A L L SKETCH OF D E V E L O P M E N T 2. E a r l y L a n g u a g e ( 2 0 - 2 4 m o n t h s ) By 20 months K a l u l i children have acquired the names o f common foods (men 'sago/food', jun 'asparagus', galen 'crayfish', magu 'banana', hon ' w a t e r ' ) , animals (gasa ' d o g ' , kabo ' p i g ' ) , body parts (dagi 'hand', bo 'breast'), and household objects (as 'bag', helebe ' k n i f e ' , / ' w o o d / s t i c k ' ) . However, this aspect o f language, the saying o f names, is downplayed i n K a l u l i families. There are no labeling games to facilitate or encourage the learning o f object names. This is primarily due to the linguistic ideology o f the culture. I t is only i n families who are acquiring literacy that one sees any attention paid to saying the names o f objects, and this activity is initiated by the child when the mother is looking at books. W h e n extended by the child to other contexts, the mother's response is disinterest. I n contrast, because o f the cultural importance placed on learning the proper names and kinterms o f the individuals w i t h w h o m they interact, K a l u l i children are consistently encouraged to master a large number o f proper names, kinterms, and other relationship terms. They are corrected i f the name is misapplied or mispronounced and told to repeat i t correctly. W h i l e children have little problem in addressing their parents (nowo ' m y mother', dowo ' m y father'), when refer­ ring to the parents o f others they have difficulty shifting to the correct forms (gowo 'your mother', goto 'your father', ene ano 'his/her mother', ene inya 'his/her father'). Errors i n reference continue w i t h this set o f kinterms, w h i c h is applied to a large set o f individuals (not just biological parents, but classificatory parents) at 30 months. Such errors include switching terms for mother and father

5.

The A c q u i s i t i o n of Kaluli

535

(3rd person) and switching 1st and 2nd person terms i n reference, but never i n address. Children are usually corrected when this occurs. I n terms o f linguistic development it is important to emphasize that K a l u l i children learn selected aspects o f particular linguistic systems, such as a re­ stricted set o f tense markers or only the 1st person o f a given pronoun set. Therefore one must talk fairly specifically about which tenses and for w h i c h persons any particular aspect o f language is being learned. One cannot say that the young c h i l d , for example, uses the verb dimina ' g i v e ' . One must specify which forms are used at any one time. K a l u l i children use the nonfocused form o f the 1st and 2nd person subject pronouns by 22 months. Most early nouns lack any casemarking. Young lan­ guage learners also use deictics and demonstratives (we 'this/here', ko 'that/there'), though errors distinguishing proximal/distal occur w e l l beyond 24 months. Possessive relationships for 1st person are expressed w i t h possessive adjec­ tives i n combination w i t h a limited set o f deictics and nouns. For 3rd person the order o f nouns expresses possession before casemarking is used. Possessive pronouns are used i n the 1st person i n assertions o f ownership (neno ' m i n e ' ) , and w i t h an interrogative particle added to 2nd person (genowo? 'is it yours?!') used in challenges o f ownership. Locative relations are always marked by a general locative casemarking suf­ fixed to a range o f lexical items including deictics and nouns. we this/here

-na en -a L O C it L O C

'on this, on here, on it' a -ya house L O C

as -a bag L O C

'in/at/to (the) house'

'in/on (the) bag'

Function words such as 'gone' (ane for animates) and 'none' (aundomu for inanimates) are frequent, both as single words and i n t w o - w o r d utterances. However, the animate/inanimate distinction is not always applied correctly. A n additional form is now a 'another'. (Animacy is not distinguished i n K a l u l i for this form.) Early negation is expressed i n a variety o f ways: as a verb o f affect/internal state (mobeab ' u n w i l l i n g ' , rejection o f object/action/proposition); as a negative imperative (diesabo ' d o n ' t take'); and w i t h several lexical items expressing negation which can only be used as single words (a ' n o ' , em, ' n o ' , mo 'no, n o t ' ) . K a l u l i children have a range o f other affective expressions using verbs (ne tagidab ' I ' m afraid') as w e l l as both positive and negative expressive words (yagidi ' y i k e s ! ' ) .

536

Schieffelin

The first verbs include a limited number o f imperatives (hamana ' g o ' , mena 'come', dimina ' g i v e ' , boba ' l o o k ' ) and a restricted set o f verbs i n selected forms such as 'take:l:PRES', 'see:2:PRES:Q', 'eatPAST, 3:PRES, P R E S T M P ' , and 'go:PAST, 3:PRES'. One never hears an uninflected verb form, though one hears incorrectly inflected forms and misanalyzed stems. Requests are important i n early language. However many requests are prag­ matically limited and are not expandable to more complex structures. Only the word order i n these constructions provides " b u i l d i n g b l o c k s " for other types o f constructions. One frequent request consists o f the 1st person dative form nelo 'to m e ' . Since this is the only dative form at this time and is restricted to requests, i t should be considered a formulaic expression that appears i n limited syntactic constructions w i t h a restricted set o f nouns (mostly foods) and deictics. It also is correctly used w i t h a single verb, nol 'eat.T:PRES' meaning T want to eat', literally 'to me I eat'. Later it is used w i t h only a few verbs, and its use is restricted to begging requests even i n adult usage. Another early request is nowo ne 'some to m e ' . This formulaic expression does not take dative casemarking. W h i l e there is no Baby Talk lexicon, there is one imperative form mo I mo ' g i v e ' used only by children when w h i n i n g and begging for food and for the breast. Children w i l l use the fuller form dimina or mina when not w h i n i n g , and select the reduced form when whining until about 27 months, at w h i c h time only the full form is used ( i n all contexts). None o f the above request forms are instructed since they are i n a begging mode. However, K a l u l i children have ample opportunity to hear other youngsters using these begging forms and proba­ bly acquire them through that exposure. I n addition to these aspects o f language, there are a number o f discourse level features that K a l u l i children use i n their early utterances. T o many o f their early single and m u l t i w o r d utterances, such as deictics, possessive pronouns, and various verb forms, they attach w o r d final (and sentence final) emphatic parti­ cles. These particles indicate several different degrees o f emphasis. H o m o phonous w i t h several nominal casemarkers (ergative, instrumental, genitive), they are phonologically conditioned (-el-yeI-we) and are stable forms across all word classes, except past tense o f verbs, where a v o w e l change indicates tense agreement. Children under 24 months only use the nonpast form. There are other discourse level features o f language that children master early on. I n K a l u l i , as i n all other societies, individuals must acknowledge when they are called, indicate compliance w i t h or refusal o f requests, reply affirmatively or negatively to questions, and request clarification. A l l o f these speech acts must be accomplished i n culturally specific ways. I n K a l u l i t w o o f these responses o 'huh?', 'okay' have the same phonological shape, and changes i n the intonational contour convey the intended meaning. K a l u l i children learn to use these single-word utterances early. W h i l e these openings, closings and acknowledg­ ments are not usually reported i n other acquisition studies, they are crucial for

5.

The A c q u i s i t i o n of Kaluli

537

children to learn i f they are to engage i n discourse w i t h others. The fact that these lexical items are explicitly instructed shows the importance K a l u l i place on their use. Lacking i n early K a l u l i speech are adjectives, colors, adverbs, and numbers. 2.1.

The Development

of Syntax

(24-30

months)

It is difficult to distinguish a clear single w o r d from a m u l t i w o r d stage i n K a l u l i since early language includes formulaic phrases (mostly requests and rhetorical questions), a range o f m u l t i - w o r d expressions such as those w i t h agents/actors accompanying verbs inflected for person/tense, nouns and deictics w i t h emphatic and evidential particles, as w e l l as w i t h limited casemarking (though locative is always marked). I n addition, given the amount o f ellipsis i n the language, and the infrequent use o f three constituent utterances i n casual adult speech, K a l u l i child language even i n its early productive usage, appears to sound relatively mature compared w i t h the utterances o f children speaking En­ glish. Single words and t w o constituent utterances are very common i n adultadult casual speech and adult-child speech. W h i l e the children's earliest samples were dominated by single words, by 25 months, m u l t i w o r d (two-word utter­ ances) were increasing. Even i n the latest samples (32.2 months) single words are frequently and appropriately used. Single words i n K a l u l i , therefore, have a somewhat different status than they do i n languages i n w h i c h deletion and ellipsis are not major discourse factors. W i t h regard to casemarking on nouns, at about 25 months most nouns i n the role o f agent/actor, instrument, or direct and indirect object lack any casemark­ ing or have neutral casemarking w h i c h is not correct by adult standards. H o w e v ­ er, between 25 and 28 months several markers were being used, i f inconsistently. The genitive casemarker (which is homophonous w i t h agent and instrument as w e l l as w i t h certain emphatic particles) was the first to be used w i t h any reg­ ularity, though i t was not used i n all obligatory contexts until about 30 months. Following its use for genitive m a r k i n g , the same morpheme was used to mark ergatives, instrumentals (body parts), and finally other instrumentals (objects such as knives, sticks), i n that order. One error that occurred during this time ( 2 6 - 3 0 months) was double casemarking on NPs that were composed o f nouns modified by demonstratives, deictics, and directionals (especially w i t h locatives, indirect objects, and less frequently, ergatives). Only the modifier should carry the inflected material for the noun phrase, but children were casemarking both the noun and its modifier. However, children very rarely put the w r o n g casemarking on nouns, and used neutral or zero marking rather than incorrect marking. The neutral or absolutive case markers (-ol-wol-yo) were often dropped i n fast speech by children and adults or difficult to hear when they d i d occur. I t is not possible at this time to present an acquisition pattern for the Neutral/Absolutive set o f morphemes.

538

Schieffelin

B y about 25 months possessive relationships are indicated simultaneously i n several ways. For 1st and 2nd person, possessive adjectives (ni ' m y ' , gi ' y o u r ' ) are followed by proper and common nouns and deictics. For example, ni my

gasa dog

gi your

babo mother's brother

ni my

ko that

For 3rd person, possession is expressed by genitive casemarking (-e w i t h phono­ logical conditioning) on nouns such as, Abi-ye GEN

Abi-yeno GEN

'Abi's'

'It is Abi's'

and by w o r d order (Noun + - genitive casemarking + Noun—Possessor + Possessed). Children during this time period do not take the 3rd person pro­ nominal options that are available i n the language and use only the possessor's name. However, segmentation was a problem w i t h some o f these forms, and both A b i and Wanu made errors o f oversegmentation. For example, Wanu 25.1

*Babi

we GEN

-yo NEUTRAL

we Babi-yeno) this GEN

we this

'This is Babi's' Abi 25.1

(= we Babi-yeno)

*Babi-ye GEN

we this 'This is Babi's'

*/ stick

-yo NEUTRAL

we INST

we this

-yo NEUTRAL

(= iye we)

'with this stick' Abi 28.2 mother *abe who

we GEN eno-wo it Q

wen o GEN

(= goweno) 'mother's'

(= abenowo) 'whose is it?'

I n the first t w o examples we see that when there is difficulty i n forming the genitive it is due to using the deictic (we ' t h i s ' ) instead o f adding the suffix -eno which is composed o f the genitive (-e) and the w o r d eno ' i t ' . This sample pattern interfered w i t h A b i ' s early formation o f the instrumental. Oversegmentation

5.

The A c q u i s i t i o n of Kaluli

539

errors, such as found i n A b i 28.2 show the child's analysis o f the underlying morphological forms. Forms segmented i n this way are never heard i n adult speech. The general locative morpheme continues to be used w i t h a range o f nouns i n syntactic constructions w i t h the specific locative meaning derived from verbs as in adult usage. A limited number o f spatial terms which take the locative suffix (wel ' t o p ' , 'above', us 'inside', 'under', and heg 'side') are added. A n addi­ tional suffix -Iowa 'to the place o f is added to proper names and kinterms. During this time period, the child acquires a large number o f place names (names of specific streams, grounds) used when talking about the whereabouts o f individuals. In addition to locatives, K a l u l i employs an elaborate set o f directionals i n d i ­ cating culturally important ways o f conveying spatial relationships. These i n ­ clude orienting notions from the perspective o f the speaker w i t h regard to a referent such as ' u p / d o w n h i l l ' , ' u p / d o w n stream' as w e l l as 'above' and 'be­ l o w ' . Through a regular set o f infixed consonants these directional terms also mark the referent that they modify i n terms o f animate/inanimate, and i f animate, whether or not it is m o v i n g . These directionals (which may occur alone) also combine w i t h the deictic system for additional specificity. A t about 28 months a limited set o f directionals is added. However, many o f these are incorrectly used, and distinctions between animate and inanimate are not made consistently. K a l u l i do not employ a large elaborated set o f adjectives, adverbs, or quan­ tifiers. W h i l e their use is infrequent between 2 4 - 3 0 months, they do occur. Adjectives (which must occur w i t h the head noun) include lesu ' s m a l l ' , badyo 'large', sambo ' l o n g ' . Quantifiers (which can be used as single words) include helu ' l i t t l e / f e w ' and alan 'large-sized'. (The w o r d modo 'many' is not used during this time.) I n addition, these terms combine w i t h one another, and chil­ dren do use them as nominal modifiers such as helulesu 'very small' and i n existential constructions such as we alan badiyo 'This is very b i g . ' The main adverbs used are hese ' s l o w l y ' , boboge ' q u i c k l y ' , wonole 'secretly'. There were occasional reversals i n the order o f words i n noun phrases. For example, children incorrectly produced constructions w i t h modifiers preceding (rather than follow­ ing) the head nouns and deictics. The majority o f interrogative constructions involve no syntactic alteration o f the w o r d order found i n declarative utterances. There are three types o f inter­ rogative suffixes which are added to the final syllable o f the last w o r d o f the utterance. One set o f suffixes is exclusively for verbs and is used for inter¬ rogatives that are direct questions and require an answer. W h i l e verbs i n de­ clarative utterances follow a two-way split i n person markers (1st and non 1st person), verbs i n interrogative utterances have a three-way distinction (1st, 2nd, and 3rd person). Because children tend to follow the pattern for declarative sentences, they have difficulty i n using these interrogative suffixes correctly, due to the additional distinction made.

540

Schieffelin

The second type o f interrogative suffix is for lexical items that take nominal morphology. A single suffix (-o w i t h morphophonemic alternations) makes any noun or pronoun an interrogative. This suffix is acquired by 25 months. The third interrogative particle is also formed by a single suffix (-ele) and is applied across all w o r d types and i n verbs after the 1st person or 3rd person tense inflection. I t has the gloss, T wonder ( i f it is mine, where he is g o i n g ' etc.) or Ts it really . . . ' . W h e n used, a listener is not under the same obligation to answer as one is w i t h the direct interrogative sentences. This form presented difficulty for one c h i l d , A b i ( 2 5 - 2 7 months). Rather than using the verbal interrogative suffixes, he attached the -ele suffix directly to the stem material o f the verb producing incorrect utterances such as, A 25.1

ge you

oba where

*han - ele go Q

(= ge oba hana-ya) go:2:PRES:Q

'Where are you going?' There are several ways to form questions i n v o l v i n g the equivalent o f ' w h ' words. One may use constructions such as, do father

oba where

'Where's father?' we this

or

do

or

'What about father?'

he what about

oba what

'What's this?' These types o f constructions require no inflections and no verb. They are fre­ quent i n casual conversation and are acquired by 24 months without errors. Other interrogative constructions may be formed using verbs such as, do father

oba where

sab -a is:3:PRES:Q

'Where is my father?'

or

do

or

T wonder where my father is?'

oba

sab-ele Q

These presented more difficulty and A b i produced utterances such as A 25.3

*aba where

-lab *

-ele Q

(= aba sabele)

'I wonder where he/she is?' *enoit

bawhat

lab-ele * Q

'I wonder what it is?'

(= enobele)

5.

*we this

abe I l*we abewho

(= ko/we that/this

abenowo whose

ele I l*ko Q that or

The A c q u i s i t i o n of Kaluli

541

abe-ele Q

ab enow ele) who:it:Q

'Whose is it/this/that?' Through these errors we see A b i ' s misanalysis o f the formation o f certain types of interrogative structures. He has suffixed a single form -ele to interrogative pronouns and to part o f the verb-stem material. I n some cases he inserted other verb-like morphemes such as -lab w h i c h are found i n K a l u l i but not i n this type of construction. W h i l e M e l i made no errors o f this type, W a n u produced utter­ ances w i t h redundant expression such as, W 26.3

*abe who

abeno-wo whose is it

(= abenowo) 'whose is it'

' W h e n ' and ' w h y ' questions (as true information seeking questions) are al­ most never asked at all by either parents or children. ' W h y ' questions are rhetorical and not meant to be answered. Children do not produce rhetorical ' w h y ' questions until they are over 5 years. Other questions used at this time are rhetorical questions and formulaic ex­ pressions w h i c h are instructed through elema sequences and produced spon­ taneously as w e l l as imitatively. Some o f these do not involve any morphological means to indicate interrogative (ge oba?! ' W h o are you?') while others do (genowo?! 'Is it yours?!'). F i n a l l y , interrogative particles were suffixed to nouns and possessive pronouns i n clarification requests f o l l o w i n g the usage patterns o f K a l u l i adults. These interrogatives used the identical morphosyntactic forms as did the rhetorical questions, but followed a different intonational contour. The number o f different verbs increased during this time as d i d the distribu­ tion o f tense inflections. Verbs (as i n the adult language) always appeared w i t h inflected material and were used as single words and i n m u l t i w o r d constructions. A t this time imperatives were used w i t h and without subject pronouns and direct objects. Children showed a predictable acquisition o f a large number o f transitive, intransitive, and stative verbs. I n addition to simple verbs ( ' c o m e ' , 'go', ' g i v e ' ) children used a restricted set o f concatenated verbs. These verbs express sequential and purposive action, and i n a number o f cases stress disam­ biguates otherwise similar surface forms due to contraction. Initially these verbs were treated as single lexical items and there was no evidence o f analysis o f the two different underlying forms o f the suffixes (sequential action and purposive action). Purposives were the first complex verbal relationship to be expressed morphologically. (Verbs used include sieniane 'went i n order to p i c k ' , and die (ni)honol ' I ' m going i n order to take'. Consecutive past action verbs, w h i c h followed, included ome or omina 'having chewed g i v e ' , and die(ge)-mena 'having taken come' or ' b r i n g ' .

542

Schieffelin

The use o f the quotative verb 'say:3' preceded by a sentential object to report the speech o f others is frequent from the age o f 26 months. Direct quotation is frequently used to report speech, and children are encouraged to contribute a turn of talk i n this manner. The verb 'say' is used i n both past and present. Aspect is expressed i n K a l u l i through prefixes, infixes, and lexical items. Infixes are infrequent before 30 months. A r o u n d 30 months, one o f several pre verbal particles became productive. This prefix, -o ' s t i l l ' is used to indicate duration. Other expressions o f aspect by lexical and morphological means were acquired after this period. Additional means o f verbal negation were acquired between 24 and 30 months. I n addition to an increase i n different negative imperatives, children correctly prefixed the negative morpheme mo- to verbs i n the past, future, and habitual tenses i n simple verbs, and correctly placed the negative prefix on the initial verb i n complex forms taking auxiliary verbs. They continued to use less complex forms o f lexical negation i n pragmatically appropriate ways. B y 30 months they were adding the negative suffix (-bal-ma) to nominals and deictics. A t the discourse level, emphatic particles marking degree o f intensification were added across a wider range o f utterance types, especially on 1st and 2nd person possessive pronouns and imperatives i n requests, and when repeating utterances that are not understood. However, between 2 8 - 3 0 months children changed the final v o w e l o f a w o r d to the emphatic suffix, w h i c h is not correct. For example, nowo 'another' *nowe dowo 'was' *dow£ I n addition, one o f a set o f evidential particles becomes productive, -lob, which indicates visual evidence, such as welob ' i t ' s here' said when finding something that was previously out o f sight. Emphatic particles may be attached to this evidential. For example, by 25 months children produced utterances such as welobc ' i t ' s really here!' B y 28 months, children use the past form o f the emphatic for verbs and f o l l o w i n g the evidential -lob T see evidence' on deictics and existential forms, such as aundoma 'there is none' to change the tense o f that word. For example, aundoma 'there is none', aundomalob, 'there is none ( I see)' and aundomalobe 'there was none ( I saw)'. B y 28 months they use several emphatic forms i n sequence to intensify their utterances. Different sets o f nominal and verbal suffixes are used depending on whether one is close or distant from the addressee. Different intonational contours are used i f someone is saying something for the first, second, or third time. These different aspects o f language use are acquired by 28 months. However, children do tend to use different evidential and discourse suffixes as general emphatics, without the correct pragmatic force. Only those w o r d orders found i n adult speech were used by K a l u l i children. Initially utterances w i t h the f o l l o w i n g w o r d orders were produced: O V , A V , S V ,

5.

The A c q u i s i t i o n of Kaluli

543

O D A T (noun/deictic + 'to m e ' ) , D A T V , L O C V , A L O C V , S L O C V . However, for three children all w o r d orders were produced w i t h higher frequen­ cy w i t h 1st person agents than w i t h 3rd person agents. Once three constituent utterances were established, O A V utterances were produced w i t h higher frequen­ cy than A O V for all children. 2.2. Later Acquisition

(30+

months)

As noted above, K a l u l i has a rather elaborate set o f directionals which through infixing also encode animate/inanimate. A further distinction is made w i t h i n the category o f animate: + / - m o v e m e n t . These forms emerged around 30 months but were not used correctly. The animacy/inanimacy distinction was usually not acquired until later. Even at 30 months (and beyond) there was a pattern o f deletion i n ergative casemarking i n A V utterances for certain verbs, but obligatory casemarking was used i n O A V utterances. After 30 months children acquired more complex forms o f negation, includ­ ing a set o f suffixes for verbs and nominals used to counter assertions and actions. Children began to use a limited set o f future imperatives, both positive and negative, including infixes that express plurality i n both present and future i m ­ perative. Aspect was beginning to be expressed morphologically (by infixing) and lexically. Rather than just using the verb 'say' preceded by the quoted speech, around 30 months children quoted speech by bracketing the sentential object (the quoted speech) w i t h the agent (marked w i t h the ergative) and the inflected verb. There were no relative clauses used before 32 months. Children used t w o separate propositions to express the coordination o f t w o actions by t w o different agents before morphosyntactic means were used. Complex constructions are limited to same agent/actor (for both dependent and independent clauses). There was no morphologically marked switch reference before 32 months. A r o u n d 32 months a fuller range o f evidential markers are used to indicate source o f evidence: 'having heard', 'inferred from evidence' emerged and chil­ dren began to use these more specifically for their evidential features, not as generalized emphatics.

W O R D ORDER 3. T h e P r a g m a t i c s o f W o r d O r d e r K a l u l i has t w o allowable w o r d orders: O A V and A O V . Except i n situations where speakers believe that listeners have not understood what they are referring to, the verb is always final. W h e n such self-repair sequences occur and a N P

544

Schieffelin

appears after the verb, there is a marked pause followed by the clarification o f the referring expression. For example, ko

ni

mono

siabulu

-wo

that

I

eat:PAST

sweet potato

ABS

T ate that . . . the sweet potato.' W o r d order is used for pragmatic focus. ' T h e constituent w i t h the most salient or important pragmatic i n f o r m a t i o n " is i n focus (Givon 1975, p . 185). I n K a l u l i focus can be thought o f from t w o perspectives. I n utterances that have t w o NPs, the second N P is i n focus and the first N P is not. Thus the N P that is i n pre verbal position indicates what is i n focus, what is at issue, or what is being contrasted by the speaker i n the discourse. I n other words, i n utterances w i t h t w o NP constituents, the second N P has the relatively new information. I t can be seen as the informational highlight o f the sentence. Therefore i n sentences i n O A V order, the agent is i n focus; sentences i n A O V order, the agent is not i n focus. Furthermore, given the pragmatics o f these w o r d orders, each one is selected and used by speakers i n different speech genres and registers, for different speech acts. O A V , w i t h the agent i n focus, is used i n a variety o f speech acts, such as requests, teasing, tattling, and i n arguments over w h o shall perform a specific action. For example, Three people are sitting around and one spots a cucumber. Speaker 1 reaches for it. Speaker 1:

yagan

-o

ni

cucumber

ABS

I

dieno.

take:l:FUT

T i l take the cucumber.' Speaker 2:

a!

no

yagan

-D

nisa

ABS

I (not you)

diol.

take:l:PRES

'No! I take the cucumber!' I n this example Speaker 1 is a potential agent and uses O A V order i n claiming the cucumber. Both agent and object are expressed since a new topic is being introduced. Speaker 2 challenges the claim and uses another focused pronoun which is more specific ( T not y o u ' ) . Since the agent is what is the issue, or informational highlight o f the sentence, O A V order is used. For emphasis, both agent and object are expressed by Speaker 2. Utterances w i t h A O V order are used to announce or report an action, i n narratives and stories, or when the speaker is focusing on or contrasting the object. For example,

5.

The A c q u i s i t i o n o f Kaluli

545

Abi's older sister (age 9 ) sees him drinking dirty water. Calling to her mother to inform her, Abi

-yo NEUTRAL

hon water

mogago -wo bad NEUTRAL

nab eat:3:PRES

-o EMPH

'Abi is drinking bad water!' It is alright for A b i to drink, but not dirty water. The object he is acting on is problematic, i n focus and placed directly before the verb. I n these examples we can see h o w the pragmatic concerns o f the speaker are reflected i n the way i n w h i c h elements o f the utterance are arranged. Pragmatic concerns and context determine the w o r d order selected by speakers. I n turn, word-order selection communicates the concerns o f the speaker to the listener. These pragmatic functions are also conveyed i n the use o f two-constituent utterances. A V utterances are those i n w h i c h the agent is less topical, informa­ tion about the agent is relatively new, i n focus, and more specified. However, from the speaker's perspective the outcome for the agent is less certain. I n O V utterances, information about agents is referenced i n the verb through bound person suffixes and further information is presumed to be given. I n addition to the order o f N P constituents, the number o f expressed constitu­ ents i n K a l u l i follows discourse constraints. Three-constituent utterances, es­ pecially those w i t h full nouns expressing agent and object, are relatively infre­ quent i n everyday conversational discourse. One 4-hour language sample was analyzed i n order to compare the relative frequency o f three-constituent utter­ ances i n child and adult speech. The sample that was chosen was the last sample from M e l i (32.2 months) since that was the most advanced child language sample. I n addition, during this taping a number o f adults were present and their conversations w i t h her as w e l l as w i t h each other were quite extensive. I n that sample five adults produced o f 1476 utterances w h i c h included 29 utterances w i t h expressed agent and object. M e l i produced 1514 utterances o f w h i c h 49 utterances had expressed agent and object. This suggests that a higher number (and proportion) o f three-constituent utterances does not mean greater linguistic competence. When three-constituent utterances do occur their use is patterned. They are used to introduce a discourse topic when talk has not yet been established. For example, A group of people return to the house having left a food packet containing yesi (a small marsupial) warming on the coals. Finding it gone and seeing a couple of bones on the floor one exclaimed, yesi

-yo ABS

gasa dog

-ye mon E R G eat:PAST

-ga EVID

'The dogs/A dog obviously ate the yesi

9

546

Schieffelin

Three-constituent utterances are often used as initial requests, threats, and tattles—as seen i n the examples above. I n all o f these speech acts the order would be O A V . Three-constituent utterances are frequently found i n counterassertions (as i n the example above). I n addition, the use o f both agent and object adds emphasis to the speaker's utterance. Three-constituent utterances are also found i n contexts where a speaker tries to change the topic o f ongoing discourse, attempts to establish his place in the conversation, or tries to break into the turn o f another speaker. They also occur when a speaker changes addressee i n conversation without the use o f a vocative. I n addition to these uses i n conversation, three-constituent utterances are also found as expansions f o l l o w i n g clarification requests and as self-repair on utter­ ances by a speaker when he has some reason to believe that his message has not been conveyed w i t h adequate clarity. A common pattern is for a single-constitu­ ent utterance (usually the inflected verb) to be expanded to a two-constituent utterance f o l l o w i n g some need for clarification, and for a two-constituent utter­ ance to be expanded to three. Almost without exception, expansion takes place to the left o f the expressed constituents, that is i n an A V utterance, expansion would be to O A V . Speakers do not insert constituents into a two-constituent utterance. Finally, i n addition to the order and number o f constituents occurring i n the discourse, the distribution and choice o f new and o l d information conveying lexical forms (nouns, pronouns, deictics) i n the expression o f N P constituents is important. The use o f t w o full nouns is usually reserved for utterances that introduce a topic, when the addressee is out o f the immediate visual context and w o u l d not be able to adequately identify the referents through deictics or demon­ stratives, or when there is reason for emphasis. Otherwise, speakers use nonfocused forms o f the pronouns i n A O V utterances and deictics, demonstratives, and other given information markers as objects i n O A V utterances, such as the following:

Two women are reaching for the same stalk of asparagus, Speaker 1:

ko that

niba I (not you)

dagumieni peel:INTENT 1

T am about to peel that.'

3 . 1 . Acquisition

of Word

Order

Patterns

K a l u l i children, in keeping w i t h the adult model o f verb-final syntax, pro­ duced only verb-final utterances. This pattern was established at the outset and continued throughout all samples. I n terms o f the order o f elements w i t h i n the

5.

The A c q u i s i t i o n of Kaluli

547

utterance, children consistently (correctly) placed locatives and adverbs before the verb when only one o f the t w o major constituents was expressed (agent/object). There were occasional errors i n order when both major constitu­ ents were expressed. W i t h only t w o exceptions, children's use o f A O V utterances was pragmat­ ically appropriate. They were used to announce intended actions and describe ongoing activities where the agent was not i n focus. O A V utterances were also used appropriately i n teasing, tattling, shaming, and requesting. Children used word order pragmatically appropriately before they used grammatical casemarking correctly. There is strong evidence from t w o - w o r d utterances that children were encoding agent and verb when the agent was i n focus. W h e n i t was not, they encoded object and verb. The reason that pragmatic usage i n w o r d order precedes grammatical casemarking is most likely due to the fact that w o r d order is tied to pragmatic function. W o r d order is more salient than casemarking, i n that alternative w o r d orders are more obvious than the presence or absence o f a one-syllable casemarker. A n y language (such as K a l u l i ) w h i c h uses w o r d order for pragmatic purposes w i l l have an interaction between discourse and syntax. The K a l u l i data suggest that children from an early age are sensitive to that interaction and the meaning conveyed by different w o r d orders. However, to become competent speakers, children must use the morphological devices o f their language as w e l l . Comparing the frequency o f O A V and A O V 1st person declarative utterances, for all children O A V utterances were o f considerably higher frequency than A O V utterances (17 o f 25 samples, 3 samples having equal frequencies). This is not surprising, considering that O A V utterances are used i n requests and situa­ tions where the child wants something for him/herself. O A V 3rd person utter­ ances were o f higher frequency i n only 9 o f 25 samples, w i t h 9 samples equal. However, O A V (combined 1st and 3rd person) utterances were consistently more frequent than A O V utterances throughout all samples (17 o f 25 samples, 2 samples being equal). G i v e n that this O A V order is used for tattling, teasing, arguments, all parts o f family interaction that include children, i t is not surprising that children's w o r d order reflects this fact. The pattern is somewhat different for two-constituent declarative transitive utterances. The A V / O V pattern is reversed for 1st person. O V were more fre­ quent than A V utterances i n 15 o f 25 samples, w i t h 2 samples equal. But for 3rd person, A V utterances were more frequent i n 10 o f 25 samples, but 9 samples equal. Comparing the combined 1st and 3rd person two-constituent utterances, O V was higher frequency i n 13 o f 25 samples, w i t h 2 samples equal. Slobin has suggested (following Lehmann, 1973) that the verb and object constitute " a k i n d o f perceptual Gestalt w h i c h resists i n t e r r u p t i o n " (1975, p . 197). Ochs's analysis o f the acquisition o f w o r d order i n Samoan confirms Slobin's hypothesis i n that " y o u n g Samoan children prefer to keep the verb and

548

Schieffelin

patient sequentially contingent, placing the agent either before or after this u n i t " (1982, p . 663). K a l u l i children's w o r d order patterns (declaratives only) were analyzed to see i f an O V strategy was preferred. The total O V and A O V utterances for each child were combined for each sample (referred to hereafter as O V ) , as were the A V and O A V utterances ( A V ) . The most striking pattern is one o f individual varia­ tion. There is no preference i n general or developmentally for (transitive) A V or O V word-order pattern. For A b i , 5 o f 8 samples showed higher frequencies o f O V w i t h no developmental trend favoring one pattern over another. For M e l i , 7 of 8 samples showed higher frequencies for A V , w i t h the 1 sample favoring O V being at 24.3. Wanu's samples split, w i t h the first three samples ( 2 5 . 1 - 2 7 . 3 ) having higher frequency O V (though numerically the samples were close) and the last three samples ( 2 9 . - 3 2 . 1 ) having higher frequency A V . Given these findings, there seems to be no overall pattern o f w o r d order preference for the three children. These findings do not support Slobin's hypothesis nor are they consistent w i t h Ochs's Samoan data regarding the O V u n i t . Considering the nature o f the pragmatic functions o f w o r d order i n K a l u l i , the child data suggest that these three different acquisition patterns have more to do w i t h the particular types o f speech acts that children were using i n a given sample rather than indicating a cognitively based word-order preference. Since requests, challenges, threats, and tattles are expressed w i t h ( O ) A V w o r d order, i f a child was i n interactions where those speech acts were being used, ( O ) A V utterances w o u l d be produced, since children d i d tend to use w o r d order i n pragmatically appropriate ways. 2

The appropriate use o f ellipsis i n speaking is another important component o f communicative competence. The fact that three-constituent utterances were less frequent than two-constituent utterances i n child speech is not indicative o f a lack of competence. Three-constituent utterances are not more adultlike than t w o constituent utterances. W h i l e there is a general developmental trend i n the i n ­ crease o f three-constituent utterances, by 31 months there is a decrease i n the use of three-constituent elements. I n these later samples the use o f three-constituent utterances tends to be governed by discourse related, rather than language ac­ quisition related issues. W h i l e children d i d not always mark new topics w i t h three-constituent utterances, when they used them they were almost always appropriate. The children d i d follow the parallelism o f question w o r d order i n their responses, answering ' w h o ' questions w i t h A V order and 'what' questions w i t h O V order as i n the f o l l o w i n g example. Meli 32.2 is climbing a tree, out of her mother's sight.

Slobin (1977) points out that in OV languages, verbal modification follows the verb and nominal modification precedes the noun, thus keeping the object and verb together as a unit. This is not the case in Kaluli and many other verb-final Papuan languages. 2

5.

Mother:

ge you

oba what

dia do:2:PRES

The A c q u i s i t i o n of Kaluli

549

-ye Q:EMPH

'What are you doing?' Meli:

ne I

adam guava

-o NEUTRAL

suhl pick:l:PRES

-o EMPH

T'm picking guava.' Depending on the type o f talk, number o f participants, spatial arrangement o f participants, the attention paid to different speakers—the number o f three-con­ stituent utterances w i l l vary. Children learn these discourse rules, including deletion and parallelism, by 30 months and produce discourse

appropriate

speech.

PERSONAL PRONOUNS 4. P e r s o n a l P r o n o u n S e t s L i k e other related languages that have been described for the Trans-New Guinea Phylum, K a l u l i has t w o sets o f independent personal pronouns k n o w n as focused and nonfocused forms ( W u r m 1975, p . 192; Rule, unpublished). I n K a l u l i the selection o f one or another o f these pronoun sets depends primarily on how the speaker wishes to express 1st or 2nd person agents or subjects. The selection o f one or the other pronoun sets co-occurs w i t h the selection o f a specific w o r d order. I n brief, nonfocused forms are multifunctional, while focused forms serve only agent functions. Nonfocused forms have distributional parallels w i t h the neutral casemarking system found i n A O V utterances. Focused forms are found only i n ( O ) A V utterances as agents, and for the most part are distributed along ergative/absolutive

dimensions.

While

there

are

dual/plural

and

in­

clusive/exclusive forms for some members o f both sets, gender is not encoded i n any o f these pronominal forms. In order to appropriately use these forms, children must be able to m i n i m a l l y assess the situation, to understand what is being asked o f them, and what is being done around them. Children have to learn when i t is correct to assert that they w i l l do something themselves, or that they, not someone else, w i l l do the task. Children must learn w h i c h forms can be used to initiate something i n the dis­ course and w h i c h forms must be used to respond to something i n the discourse. What has gone on i n the prior discourse and nonverbal context must be taken into account. 4 . 1 . Nonfocused

Pronouns

There is only one basic nonfocused pronominal f o r m for each person i n the singular. These forms are: ne T / m e \ ge ' y o u ' , e 'she/he/her/him/it'. There are additional forms marking dual/plural, inclusive/exclusive, but since they are

550

Schieffelin

rarely used by children under 3 years they w i l l not be considered here. I n addition, i n everyday conversation the singular form is used unless duality or plurality is being emphasized. A s mentioned briefly above, nonfocused forms have several functions. They are used as the agents o f transitive verbs i n utter­ ances when the agent precedes the object ( A O V ) , as subjects o f intransitive verbs, as experiencers o f verbs o f affect and internal states ( S V ) , and as objects o f transitive verbs. I n order to form the dative, the nonfocused form takes a dative suffix -lo or -mo as i n nelo 'to m e ' . Thus, nonfocused forms are not only multifunctional,

but they do not distinguish different case relations. For

example, e he/she

ne me

sandab hit 3:PRES

(AOV)

'He/she hits me.' ne I

e him/her

sondol hit:l:PRES

(AOV)

T hit him/her.' However, it is extremely rare i n adult or child speech to hear a three-constituent utterance w i t h t w o nonfocused pronouns, especially w i t h A O V order. Because subject agreement is marked on the verb, and discourse constraints operate on the number o f constituents that are expressed, nonfocused agents ( i n three-constitu­ ent utterances) may be deleted, and often are. The preferred form is O V . I n fact, the usual pattern i n conversation is to delete most NPs w h i c h are to the left o f a focused N P or a dative. Nonfocused forms are also used i n existencial construc­ tions without verbs, such as, ne I

wego here

'I'm right here.' ge you

oba who

'Who are you?'

4 . 2 . Acquisition

of Nonfocused

Pronouns

The 1st person nonfocused f o r m , ne, is used i n early syntactic expressions first as an agent o f transitive verbs ( A O V / A V ) and as a subject (actors, experien­ cers) o f intransitive verbs ( S V ) . Soon after being used w i t h transitive and intran­ sitive verbs, i t is used to express the experiencer w i t h verbs o f affect and internal state (such as ' I ' m afraid', ' I ' m h u n g r y ' ) . The 2nd person f o r m ge is used as

5.

The A c q u i s i t i o n of Kaluli

551

agent i n a l i m i t e d number o f routines and formulaic questions as w e l l as w i t h imperatives. The third person form e, w h i c h like the other forms is unmarked for number and gender, is used infrequently i n both the child and adult samples, deletion being the usual option. The nonfocused agent forms were acquired before any o f the more semantically specific focused agent forms (discussed below) were used. That is, the most general, least semantically specified pronoun forms that could be used w i t h both transitive and intransitive verbs and w h i c h are the most frequent i n adult speech, are the first to be acquired. I n A O V utterances, only nonfocused forms were used to indicate agent. That is, children d i d not use focused forms for agents followed by objects. This suggests that children i n their early syntactic expressions distinguished between focused and nonfocused forms and their cor­ rect arrangement w i t h regard to other NPs. Furthermore, as we shall see below, children treated nouns and pronouns differently i n terms o f semantic role. That is, the semantic role o f pronouns was indicated by their position i n the sentence, sometimes supported or disambiguated by agreement i n the verb. Nouns took additional casemarking to indicate semantic role. Nonfocused forms were also used as direct objects. However, i n the early samples these children d i d not use the 1st person to express direct objects i n imperatives, preferring to use the verb alone, dimina! ' g i v e ! ' much like adults. Nonfocused forms appear later i n the samples as direct objects, initially w i t h imperative forms, such as ne galima 'carry m e ' . Use o f the 1st person non­ focused form as direct object is relatively infrequent compared to the 2nd person, which is used i n threats, such as ge semenigol ' ( I ' m ) gonna hit y o u ! ' . I n fact, use o f ne ' m e ' as direct object is infrequent even i n later samples, given that it can only occur w i t h 2nd/3rd person verb forms (when the agent is not 1st person). The use o f the 3rd person form is quite rare, both for agents and objects, as third persons are indicated by other lexical forms (proper names, kinterms, etc.). Compared w i t h 1st and 2nd person, children have a much lower frequency o f 3rd person agents. A n d i n those cases when an object is specified, it is rarely the 3rd person pronoun e, but rather a full noun, deictic, or demonstrative. Trjis re­ stricted use o f nonfocused forms o f the pronoun does not hold for 2nd person, as in certain frequent utterances like ' ( I ) hit y o u ' , ' ( I ) carry y o u ' . However, throughout the samples o f adult conversational data, the frequency o f nonfocused pronouns used as direct objects surprisingly is l o w . W h i l e they are allowed i n the language, they do not seem to be a form o f choice. The child data reflect this pattern. As an indirect object ne ' m e ' is used by 20 months. I t is used both as a single word and i n syntactic expressions such as nowo ne 'some to m e ' , and nelo 'to me' (usually begging). Both children and adults use i t w i t h nominate i n requests such as men nelo 'food to me' without a verb. I t is frequently used w i t h the verb 'give' (especially 3rd person past) i n reporting who something was given to (such as nemo miabe 'gave to me, not someone else'). I n this as w e l l as other

552

Schieffelin

syntactic contexts (which are not yet fully determined) it appears as nemo. The dative form is not used for 2nd or 3rd person 'to y o u ' , 'to him/her' until the child is at least 3 years o f age since children do not seem to talk about individuals giving things to persons other than themselves. Therefore, while the nonfocused pronouns can serve a variety o f functions, they are i n fact somewhat restricted by both developmental and pragmatic con­ straints. I n the earliest samples they are predominantly used i n the 1st/2nd person to express agent/actor, and i n the samples that follow to express 1st and 2nd person direct objects. W h e n used as indirect objects, they may be marked w i t h dative casemarking depending on the social and linguistic context o f their occur­ rence (4o or -mo). One error children made in using these forms was the occasional selection o f a dual or plural f o r m , or an inclusive or exclusive form when that d i d not seem to be referentially or pragmatically correct. However, i n every case when this occurred, children selected forms from w i t h i n the same person category that w o u l d have been correct. This maintenance o f person category was probably due to the fact that a l l 1st person forms, regardless o f semantic role are n-initial, all 2nd person forms are g-initial, and the majority o f 3rd person forms are e-initial. The similarity o f phonological form lends support for learning the broad category of person i n pronominal forms. W h i l e children had no difficulty using focused forms i n three-constituent utterances, they d i d have some difficulty i n two-constituent utterances, usually w i t h imperatives. I n scattered examples until about 30 months, children w o u l d use nonfocused pronouns when there was a clear demand for focused forms, as i n the last line o f the example that follows. M e l i should have used the focused pronoun back to her mother, giba, ' y o u not I ' . However, the category o f person was always correct. Meli 28.3 with mother while cooking bananas. Meli:

no mother

siefin tongs

ko those

-me balema INST turn:over:IMP

'Mother, turn (them) with those tongs.' Mother:

no and

niyo mine

giba balemesea you (not I) turn:be:IMP

'and mine, you keep turning' ne I

de fire

-yo NEUTRAL

oASP

'I'm still lighting the fire.' Meli:

*ge you

balemesea, no my: mother

'You keep turning, mother.'

gido light: L P R E S

-ke EMPH

5.

4 . 3 . Focused

The A c q u i s i t i o n of Kaluli

553

Pronouns

I n many important semantic, syntactic, and pragmatic aspects, focused forms differ from nonfocused forms. First o f a l l , instead o f one basic form for each person, there are five focused forms. For 1st person these forms are: ni

T

ne nono niba nineli

'I emphatic' T without assistance' T not you' ' I alone'

For 2nd person initial g- substitutes for initial n- and all contrasts are between 1 st and 2nd persons. I n terms o f N P functions, focused forms are used by adults only as agents o f transitive verbs, except for one o f these forms nineli ' I alone' w h i c h may also be used as actor w i t h intransitive verbs. None o f the focused forms may be used to express experiencer w i t h verbs o f affect or internal states. Furthermore, for all forms except for ni and nineli, there are restrictions on use w i t h past tense due to pragmatic constraints. Three o f these forms (nilnelnono) also function as geni­ tives when followed by nouns. Thus they, as w e l l as other focused forms, can only be followed by locatives, adverbs, verbs, and other lexical items that cannot be modified to avoid misinterpretation as a possessive relationship. As we have seen, while focused pronouns do share many features o f a set, there is enough variation among these forms ( i n terms o f co-occurrence restrictions w i t h tense/ negation/transitivity) to indicate that they do not form a totally homogeneous set. Unlike nonfocused pronouns, focused forms are restricted both i n the seman­ tic functions w h i c h they can serve and i n their location vis-a-vis the verb. Focused pronouns are used when the agent follows the object and precedes the verb i n transitive constructions ( O A V ) and i n A V utterances i n response to ' w h o ' questions. These forms are more semantically specific and pragmatically re­ stricted than the nonfocused pronouns. They are used when the agent is the focus o f the utterance—the new information—and often appear i n utterances w i t h deleted objects ( A V ) or as single words. 4 . 4 . Acquisition

of Focused

Pronouns

Ascertaining the developmental pattern for focused pronouns was a difficult task. First, i n transcription i t was often difficult for informants as w e l l as the researcher to determine whether a child was using the unfocused form ne T or the focused form ni T appropriately. W h e n listening to the speech o f young children, native speaker judgment favored the nonfocused f o r m , even i n contexts where the focused form was appropriate. Another complication i n determining the exact form o f this pronoun was that K a l u l i adults never corrected i t i n O A V or A V utterances. Since O A V / A V utterances were often requests or protests

554

Schieffelin

uttered i n high pitched, w h i n i n g voices w i t h rapid delivery, it was difficult to determine the exact phonological shape o f the v o w e l . However, the use o f the four other focused forms was much easier to deter­ mine because o f their phonological distinctness. Before 26 months o f age, use o f these forms was infrequent. B y the age o f 27 months, nono T without assistance' and niba T not y o u ' were established for three children, and ne ' I emphatic' was being used by t w o . The t w o earliest forms (nono and niba) were high frequency forms throughout the samples for all children, while ne and nineli T alone' were relatively infrequent. Once these forms were acquired, the frequency o f their use was determined by the pragmatic demands o f the situations i n w h i c h children found themselves. The children were using nono T by m y s e l f to emphasize that they wanted to perform actions by themselves without the assistance o f others (like eating, putting objects on themselves) and used this form when refusing or resisting help from others. This form usually occurred w i t h future or intentive tense markers. The second most frequent f o r m , niba T not y o u ' was used i n arguments over who was going to act on something. The use o f this form presupposes an assertion (or action) made by someone else; it is not used as an initial assertion. I t is often preceded by a! n o ! ' . This form cannot be used w i t h past tense, and children never made this type o f error. 4

The last t w o forms (nelnineli) were used by the children less frequently, but in all cases use was appropriate. I n order to use any o f these forms correctly, the child must attend to prior situational and discourse contexts and control the different restrictions according to tense, negation, and transitivity. I n almost all cases, children used focused forms i n two-element utterances ( A V ) before using them i n three-element utterances ( O A V ) . W h e n there was an exception, use was simultaneous. The children made none o f the f o l l o w i n g errors: 1. I n using both 1st and 2nd person focused forms, adult word-order rules were followed. There were no cases where the child used a focused f o r m f o l ­ lowed by an object ( A O V ) . 2. The restricted semantic role o f focused forms was adhered to. Focused forms were used only as agents, never as direct or indirect objects. 3. Pronouns and nouns were treated differently. Casemarking suffixes were never added to these pronouns i n spite o f their analogous syntactic position and pragmatic function to nouns that received obligatory ergative casemarking. 4. The distinction between agent o f transitive verb, and subject o f intransitive verb or experiencer was made. Children d i d not use focused forms w i t h intran­ sitive verbs or verbs o f affect and internal states, but only w i t h transitive verbs. This distinction is further supported by evidence i n the use o f casemarking on nouns. 5. Children were sensitive to the tense restrictions on some o f the focused forms. These forms were not used w i t h verbs i n the past tense.

5.

The A c q u i s i t i o n of Kaluli

555

The pattern o f acquisition o f the focused forms taken together w i t h the errors that do not occur suggest that this is an area where acquisition might be "speech act specific." I f the child has some sense o f the paradigmatic nature o f w o r d order, then focused pronouns can be " m a p p e d " onto the same pragmatic condi­ tions laid out for O A V utterances. This makes their acquisition easier than i f the child had to figure out a totally new pragmatic distinction underlying focused pronoun selection. The fact that children followed adult usage regarding tense restrictions is further evidence for the acquisition o f the focused pronouns through their association w i t h particular speech acts, since correct usage w o u l d fall out from their speech act classification. For example, children d i d not use niba T not y o u ' when reporting events i n the past, but instead used that form i n pragmatically appropriate requests and challenges. I n contrast w i t h data reported on English-speaking children ( B l o o m , L i g h t bo w n , & H o o d , 1975, among others), K a l u l i children never referred to them­ selves (as either agent, object, beneficiary, etc.) using anything other than a 1st person pronoun, regardless o f verb type or speech act. As noted earlier, mothers and other caregivers never addressed their children using that child's name (such as " I s A b i eating sugar cane?") nor d i d they ever refer to themselves using anything other than one o f the 1st person pronominal forms. I n addition, i n direct instruction sequences w i t h elema, caregivers always instructed the child to use one o f the appropriate 1st person forms when speaking about h i m or herself. Perhaps the consistency i n both speaking to the child and modeling language for the child explains w h y there were no errors occurring w i t h the acquisition o f the " s h i f t e r s " — t h e first and second person pronouns—in any o f the K a l u l i child data. American mothers, by using both the child's name and personal pronouns, put their children i n a " m a n y forms for one f u n c t i o n " situation. Children, given such a choice, may stick to their o w n name, which never changes, rather than trying to use to shifters w h i c h do not consistently refer to themselves.

NOMINAL CASEMARKING 5. C a s e m a r k i n g o n N o u n s Nouns i n K a l u l i , as i n the majority o f languages i n the Trans-New Guinea Phylum ( W u r m , Laycock, & Voorhoeve, 1975), are not morphologically marked for gender, number, definiteness, or indefiniteness. N o m i n a l mor­ phology is regular, and case relations are indicated through a series o f suffixes on proper and common nouns, deictics, and demonstratives. Proper and common nouns, demonstratives, and deictics take the f o l l o w i n g casemarking suffixes. Phonologically conditioned alternations are indicated i n parentheses. Absolutive Neutral

-o -o

(-wol-yo) (-v/o/-yo)

556

Schieffelin

Ergative Instrumental Genitive Comitative Locative Dative

-e -e -e -olie -a -lol-mo

(-weI-ye) (-we/-ye) (-we/ -ye) (-wal -yal-na)

I n addition to these casemarkers, nouns, deictics, and demonstratives take dis­ course and affective particles to indicate emphasis, evidence, and repetition among other discourse level functions. 5 . 1 . Casemarking

on Agents,

Actors,

and

Objects

As i n many other non-Austronesian ergative languages, only full nouns—that is proper nouns, common nouns, deictics, and demonstratives that refer to ani­ mate agents—receive agent casemarking. Nonfocused personal pronouns re­ ceive casemarking but only for locative and dative case. One o f the focused personal pronouns, ni T (the least semantically specified) takes neutral casemarking, but only when i t is left-dislocated. K a l u l i uses t w o casemarking systems on noun phrases, each related to w o r d order. A O V is the unmarked case and agents and objects are usually marked w i t h neutral casemarking (see Comrie, 1978, p . 331 for the use o f these terms). I n addition, actors or experiences w i t h intransitive verbs may only be marked w i t h the neutral casemarking. O A V is the "agent-focus" w o r d order and ergative marking on the agent is obligatory. Object casemarking i n O A V is the same as i n A O V . The only exception to this casemarking paradigm is A O V when both agent and object o f the sentence are either proper nouns or kinterms. I n those sen­ tences, the agent is marked w i t h the ergative casemarker. This split ergative casemarking paradigm is illustrated below. (A)OVtrans word order

(O)AVtrans word order

agent:

agent: -e (ERGATIVE) object: -o (ABSOLUTIVE) subject: -o (ABSOLUTIVE) of intransitive verbs

object:

-o -e -o

(NEUTRAL) (ERGATIVE) (NEUTRAL)

phonologically conditioned alternation:

-o -e

-yol-wo -ye I-we

The neutral casemarking is the "unmarked f o r m , " identical to (homophonous with) the absolutive and the form given as the citation form i n elicitation. I n casual or fast speech, such as commonly spoken i n family interactions, these neutral casemarkers are often omitted. This is consistent i n both adults' speech to children and adults' speech to each other.

5.

The A c q u i s i t i o n of Kaluli

557

The neutral forms o f agent and object casemarking can be used on both nouns as i n the f o l l o w i n g example. Abi

-yo NEUTRAL

siabulu sweet potato

-wo NEUTRAL

menigab. eat:3:INT2

'Abi is about to eat a sweet potato.' W h i l e A O V usually follows a neutral casemarking system, ergative casemarking can occur. W h e n both the agent and object are expressed by proper names and/or kinterms, ergative casemarking is used to mark the sentence initial agent. For example, Abi

-ye ERG

Suela

-yo sandab ABS hit:3:PRES

'Abi hits Suela.' As we shall see, the selection o f one or the other casemarking systems i n A O V depends on the animacy hierarchy o f nouns used to encode agent and object. D i x o n (1979) and Silverstein (1976) have discussed what types o f lexical items are most likely to occur as agents. A hierarchy o f agents, from most likely to least l i k e l y , shows that first and second person pronouns are highest i n this animacy hierarchy; animate common nouns follow; while inanimate common nouns are the lowest ( D i x o n , 1979, p . 85). Proper nouns fall i n the middle o f this continuum, that is, they are just as likely to be acted upon by an agent as they are to be agents. I n K a l u l i A O V utterances, when both agent and object are expressed by proper names or kinterms, this can be seen as the context o f m a x i m u m ambiguity in the animacy hierarchy. The ergative paradigm is used to indicate w h i c h equally potential agent is i n fact controlling the action. The ergative marker grammatically disambiguates the function o f each proper noun i n the utterance. Only i n A O V utterances o f this type can the ergative casemarker appear, and its use is obligatory. I n recent discussions on ergativity (Comrie, 1978; D i x o n , 1979; Silverstein, 1976) it has been pointed out that i n all split ergative languages the ergative casemarking system applies under certain conditions, and not under others, e.g. w i t h certain w o r d orders, tenses, person/number o f agents. Just as the pro­ nominal forms (focused/nonfocused) are selected according to whether or not the agent is i n focus (position vis-a-vis the object), the ways i n w h i c h nouns are marked for semantic role o f agent varies according to their position i n the sentence, that is whether or not the agent is the new information or is being emphasized or focused. I n K a l u l i ( O ) A V utterances (3rd person) which have the agent i n focus, ergative casemarking is obligatory for proper and common nouns, deictics and demonstratives.

558

Schieffelin

5.2. Acquisition

of Casemarking

in AOV

Utterances

Recall that for adult speech, A O V order uses neutral casemarking on both nouns i n most contexts; that is, both nouns are marked identically and w o r d order signals semantic role. I n casual speech, one or both o f these neutral case suffixes may be omitted. O n l y when both nouns are expressed as proper nouns or k i n ­ terms must ergative casemarking be used w i t h sentence-initial agents. T o mark agents i n any other way i n A O V utterances is incorrect. Children essentially go through three stages once they start producing A O V utterances (3rd person). First, agents are not marked at all; then the ergative marker is added to sentence-initial agents (when objects are not proper nouns or kinterms), producing incorrectly marked utterances. Finally, sentence-initial agents are marked w i t h neutral or sometimes zero casemarking. A t this time, the use o f the ergative marker is restricted to O A V order, and to those infrequent occurrences o f A O V order w i t h proper noun and k i n t e r m i n both agent and object role. The f o l l o w i n g example is from the first stage, zero casemarking. Abi 25.1 is watching me eat pandanus, a tropical vegetable. Announcing this to his mother, Babi

oga pandanus

nab eat:3:PRES

'Babi eats pandanus.' This is the correct w o r d order ( A O V ) since he is reporting an event, and a threeconstituent utterance is appropriate since he is opening a new topic o f discourse. A l l three children produced this type o f utterance without any casemarking on the agent or the object. For t w o o f the three children, w i t h i n a month o f producing correctly marked agents i n O A V utterances, there was an overgeneralization o f the ergative marker to agents i n sentence-initial position. Wanu d i d not follow this pattern exactly, but he was using the ergative marker correctly w i t h two-element ( A V ) construc­ tions, w h i c h might have influenced his incorrect extension o f the ergative marker in sentence-initial position. None o f the children produced utterances i n w h i c h both nouns were proper nouns at this time. Therefore, all sentence-initial agents were being marked incorrectly. A l l three children produced this type o f error. Wanu 29.3 is with his father, who is taking salt out of a box. To his sister, *do my father

-we solu E R G salt

diab take:3:PRES

'My father takes the salt.'

5.

The A c q u i s i t i o n of Kaluli

559

A g a i n , the w o r d order is correct since Wanu is reporting his father's action, and three-constituents are appropriate since it is a new topic o f talk. However, the ergative casemarker is on the agent i n sentence-initial position, followed by the object, which is incorrect. The use o f utterances where both agent and object were expressed by proper names and kinterms was infrequent. W i t h a total o f 41 A O V utterances produced by the three children, only t w o utterances o f this type were produced and both were correctly marked. These occurred i n samples over 30 months. B y 30 months t w o o f the three children were consistently marking A O V utterances correctly. Only M e l i made occasional errors. B y 32 months all three children were consistently controlling both w o r d order and the casemarking system. These K a l u l i children used w o r d order pragmatically appropriately before using casemarking correctly. Ergative casemarking was used independently o f w o r d order to mark transitive agents. A l l three children overextended its use i n A O V utterances, producing grammatically incorrect utterances before learning to re­ strict its use to the correct w o r d order, O A V . However, the ergative marker was never extended to the actors o f intransitive verbs, keeping separate the tran­ sitive/intransitive verb distinction important i n this split ergative language. This suggests that children may not have a general notion o f actor/agent w h i c h is applied across all verb categories. 5.3. Acquisition

of Ergative

Casemarker

in OA V

Order

There are t w o issues i n the acquisition o f the ergative casemarking i n O A V and A V utterances. One concerns the use o f ergative casemarkers i n threeconstituent utterances to clearly mark the semantic roles o f agent and object. The other concerns issues o f transitivity as seen i n A V utterances where particular verbs do not " e l i c i t " the obligatory ergative casemarking on agents i n child speech. For all children, ergative casemarking was used i n A V utterances before they were producing any O A V utterances at a l l . A l l children produced ergatively marked A V utterances before producing ergatively marked O A V utterances. I n all cases, children produced more correctly marked A V utterances per sample than O A V utterances per sample. For M e l i both correctly and incorrectly marked (meaning zero casemarking) O A V utterances were produced about a month after ergatively marked A V utterances were being used. A b i produced ergatively marked A V utterances t w o months before correctly marked O A V utterances; and Wanu produced correctly marked A V utterances four months before using O A V utterances. There was individual variation i n the acquisition o f appropriate casemarking on three-constituent utterances. Once children begin producing these utterances w i t h 3rd person nominal agents, they go through t w o stages i n acquiring the obligatory ergative casemarker. One c h i l d , A b i , produced no three-constituent

560

Schieffelin

utterances until 27.2, at w h i c h time he produced utterances w i t h the ergative marker correctly placed. W i t h only t w o exceptions, he correctly marked all remaining nouns ( i n three-constituent utterances) for the duration o f the study. Wanu initially produced three-constituent utterances without any casemarking at all until 30.3 months, at w h i c h time he produced only correct forms and con­ tinued to do so thereafter. The third c h i l d , M e l i , used ergative casemarking correctly from 26 months, but w i t h i n the same sample had utterances i n w h i c h it was not added through 32.2 months when sampling terminated. I n M e l i ' s last five samples, the number o f correct markings was always greater than the number o f neutral and zero casemarkings. Her casemarking errors tended to cluster around her use o f particular verbs (discussed below), and these caused A b i problems as w e l l . W h e n children omitted an obligatory ergative casemarker, mothers often told them to say i t again properly after providing the correct model for them to imitate. Children marked not only proper nouns (names, kinterms) w i t h the ergative marker ( i n O A V utterances), but correctly marked a variety o f animate agents, referring to them w i t h common nouns, as w e l l as deictics and demonstratives. For O A V 3rd person utterances, i n only one example i n the entire corpus (77 tokens) was an incorrect casemarker, the absolutive used on the agent. I n all other examples there was either zero casemarking, w h i c h was not correct, or correct ergative casemarking. Children used ergative casemarking and the concomitant w o r d order to cor­ rectly contrast agents w i t h i n their o w n turns o f talk, as w e l l as across speaker turns. For example, Meli 27.3 is taking one sandal and her brother S is taking the other. To her mother, nodo one side

-wo ABS

nisa I

diol take:l:PRES

-we ERG

diab take:3:PRES

'I take one side.' nodo

-wo

S

'S takes one side.' Wanu 31. and his sister are talking. She is holding a net bag. Sister:

as bag

-o ABS

we this

Daibo

-we dimiabe E R G give:RECENT. PAST

'Daibo gave this bag.' Wanu:

we this

Babi

-ye dimiabe ERG

'Babi gave this.'

5.

5.4. Inconsistency Utterances

of Ergative

The A c q u i s i t i o n of Kaluli

Casemarking

561

in (O)AV

W h i l e the pattern o f inconsistency o f ergative casemarking is pervasive only for M e l i i n three-constituent utterances, i t is the rule rather than the exception i n the ergative casemarking for two-constituent utterances for all three children. I n every sample the number o f agents without ergative casemarking is greater than the ergatively marked forms. The majority o f these agents have neutral casemarking while i n every sample there are some agents that have no casemarking at all. I n some samples, even the later ones, the number o f agents without ergative casemarking is twice the number o f marked agents. This pattern strongly con­ trasts w i t h the pattern o f marking agents w i t h the ergative casemarker i n threeconstituent utterances. For example, zero casemarking on agent: *Abi dilob take:PAST:EVID ?'Abi took'

or

?'(someone) took Abi'

neutral casemarking on agent: *Abi

-yo NEUTRAL

dilob

ergative casemarking on agent: Abi

-ye ERGATIVE

dilob

'Abi took' I n examining the data for possible explanations o f this pattern, one strong possibility is that casemarking may be interacting w i t h tense. I n many ergative languages (and i n other split ergative languages) casemarking and tense interact, in that ergative casemarking may be restricted to only certain tenses. For exam­ ple, i n Georgian as w e l l as other South Caucasian languages, i n many modern Iranian and Indo-Aryan languages (Bechert, 1979) as w e l l as i n Choi (a M a y a n language o f M e x i c o ) , the tense/aspect basis o f split ergativity i n present tense follows a nominative-accusative pattern while the past tense follows an ergativeabsolutive pattern (Comrie, 1978, p . 352). This, however, is not the case for K a l u l i . Tense does not restrict the application o f ergative casemarking i n tran­ sitive verbs. I n spite o f this, once children begin to use ergative casemarking, they mark agents more frequently when those agents accompany verbs that are i n past tense forms. That is, there is a preference to mark the agents o f past action, rather than present or future actions w i t h ergative casemarking. This is most evident i n the verb ' g i v e ' , where present and future tense agents are erratically

562

Schieffelin

marked, and past tense agents are more consistently marked. I n later samples (over 28 months) other verbs i n the past tense tend to take ergatively marked agents while their corresponding present and future tense forms have agents w i t h neutral or zero casemarking, w h i c h is incorrect. These verbs include 'take', ' h i t ' , ' b i t e ' , 'light (a f i r e ) ' , ' c u t ' , 'eat', ' c o o k ' , ' t o u c h ' — a l l relatively high i n transitivity. I n addition, while there are not enough examples for a more con­ clusive analysis, i t appears that another factor affecting inconsistent marking i n child speech is negation. Verbs that are negated also tend to have neutral or zeromarked agents. For the duration o f the study (until the children were 32 months old) all children produced A V utterances w i t h neutral or zero casemarking on agents at the same time as they were correctly marking agents w i t h the ergative casemarker. A n analysis o f these A V utterances (transitive verbs requiring ergative marking on agents directly preceding them) was carried out. W h i l e there is some individual variation, there are general developmental patterns i n acquisition. The marking or nonmarking o f the ergative casemarker on nouns is not randomly distributed throughout the verbs that co-occur w i t h those nouns. Cer­ tain verbs, i n spite o f their high frequency i n both adult and child usage, seem to present difficulties for children. T w o o f these verbs are sama 'speak/say', and engab 'does like that'. The verb 'speak/say' is used frequently i n the imperative as w e l l as the habitual and past. These tenses rarely take 3rd person agents. The most fre­ quently used form taking 3rd person agent (in the context o f children) is i n threats, where 'speak' is i n the 3rd person future and the agent is specified by name. This usually takes the f o l l o w i n g f o r m , do my father

-we sameib E R G speak:3:FUT

'My father will speak/say (something)!' This common threat never takes an object, that is, unlike reported speech, no one ever says what anyone w i l l i n fact say. W h i l e adults always use the ergative casemarker on the agent w i t h this verb, i t is not until children are over 30 months that they do the same. U n t i l that time, the agent receives zero, neutral, or ergative casemarking. Similar patterns occurred for related verbs such as 'call out'. Ergative casemarking was never used when the verb was negated. Rather than input being the issue here, the verbs o f saying may appear to be l o w i n terms o f transitivity to the child speaker since they never appear w i t h an object (except in quoted speech) nor do they carry action to an affected object, like the more typically high transitive verbs. The second verb that consistently presented difficulty, and w h i c h along w i t h sama 'speak/say' accounts for a high percentage o f neutral and zero casemarked agents, was engab 'does' and elengab 'does like that'. This verb was often used

5.

The A c q u i s i t i o n of Kaluli

563

when the child d i d not remember or k n o w the more specific verb for the action he or she was referring to. W h e n the child used these less specific verbs it was sometimes difficult to identify what aspect o f the action or situation the child was referring to. For example, upon hearing the strumming o f a guitar, one child said, Seligiwo engab, 'Seligiwo does that', but i t was unclear whether that meant he hits the strings or he makes the noise. A l l children tended to mark agents when this verb was i n the past tense, while not marking agents when the verb was i n the present and future tenses. (This verb as w e l l as speak/say' was not used w i t h 1st person, so there is no evidence from pronominal usage.) Given the consisten­ cy o f child use, there seems to be a constellation o f factors w h i c h inhibit agents accompanying this verb from taking the ergative casemarkers. Clearly t w o fac­ tors to consider are the relative transitivity o f the verb and the extent to w h i c h i t takes an object. 4

Hopper and Thompson (1980) have outlined ten parameters o f transitivity, each o f w h i c h suggests a scale according to which clauses can be ranked. A n examination o f the K a l u l i c h i l d data on the use o f ergative casemarking to indicate agency suggests that K a l u l i children are sensitive to at least four o f these parameters o f transitivity. These are (1) Aspect—children tend to mark agents o f completed rather than noncompleted actions, (2) M o d e — c h i l d r e n tend to mark agents w i t h past action, rather than action that is currently or has not yet oc­ curred, (3) A f f i r m a t i o n — c h i l d r e n tend to mark affirmative but not negated clauses and (4) Kinesis—the agent o f the verb 'say' was usually not marked. Data on the acquisition o f languages w i t h ergative features w o u l d be especially interesting to examine for these relationships. 5.5.

Expressing Agents: Comparing and Ergative Casemarking

Focused

Pronouns

Focused pronouns and ergatively marked nominals have a number o f struc­ tural parallels that make them interesting to compare i n terms o f the acquisition data. First, both focused pronouns and nouns w i t h ergative casemarkers (except when both agent and object are proper nouns i n A O V utterances) occupy the same position i n the sentence, i n that they both must follow the object. Both mark the agent as the new information or the informational highlight o f the sentence. B o t h , i n terms o f core semantic roles, can only be agents. However, one important difference between focused pronouns and ergatively marked nouns is that nouns w i t h nonergative casemarking can express a range o f roles other than agent, as determined by their different casemarking, while focused pro­ nouns can only be agents. They have no other semantic function. The question is what is the relationship between the correct use o f focused pronouns and ergatively marked nouns i n the child language data. Wanu first produced A V utterances (3) at 2 5 . 1 , and at 26.3, three different focused pronouns were used. M e l i used ergative casemarking on A V utterances (2) at 24.3 and one focused f o r m appeared at 24.1 m o . I n M e l i ' s next sample,

564

Schieffelin

two focused forms became productive. A b i used ergative casemarking i n only one A V utterance at 25.1 m o . and one focused form appeared at 25.3 m o . A t 27.3 three different focused forms were productive, and A V utterances began to be produced at 28.2. There is no clear pattern to the order o f emergence o f focused pronouns and ergative casemarking. However, focused pronouns are used correctly i n A V utterances before ergative case is consistently marked on nouns i n A V utter­ ances. I n fact, w i t h two-constituent utterances, the 90% obligatory context rule cannot be met for t w o o f the three children at 32 months because o f their inconsistent use o f the ergative casemaker! (It is met however for the threeconstituent utterances for all three children.) However, i f we compare these findings w i t h the use o f 1st and 2nd person focused pronouns that occupy the same position and semantic role as the ergatively marked forms, the picture is very different. Unlike the ergatively marked nominals, focused pronoun form usage has an acquisition pattern w i t h very few errors, especially i n three-constituent utter­ ances. This is for several reasons. Each pronominal form has an independent semantic meaning, unlike the ergative suffix w h i c h has no meaning by itself, and is homophonous w i t h the instrumental, genitive, and one set o f emphatics. A s a lexical item the pronominal form may occur alone. I t is more salient than the suffix, though both are tied to w o r d order. A s w i l l be presented below, there are errors i n ergative casemarking on nouns followed by direct objects ( A O V ) . Errors o f w o r d order are not found w i t h focused pronouns i n that their use does not violate word-order rules; that is, they are not found preceding an object. I n addition, 1st and 2nd person pronouns are used more frequently than ergatively marked nouns, and are pragmatically relevant to the child over a range o f con­ texts. Thus, when we examine the acquisition o f notions relevant to ergativity (agency, transitivity) using only nominal casemarking data, we find one set o f patterns and errors and some o f these vary w i t h the number o f constituents ( t w o or three) i n the utterance. However, when we examine the focused pronouns i n terms o f errors due to agency and transitivity, a different pattern emerges—one which is a great deal more consistent, and most importantly, does not reveal some o f the potentially interesting problems concerning the interaction o f various aspects o f language (such as transitivity, tense, and negation) that are prob­ lematic for K a l u l i children. Without errors or inconsistencies, we cannot see how the child is constructing the language at any given point i n time. However, we need to examine how the different systems (pronouns/nouns) and complemen­ tary linguistic resources available i n a language present different and sometimes inconsistent types o f information about the developing linguistic system. M o r e data for further research are needed i n this area. I t is important to emphasize that i n samples before 30 months, there is lower frequency for 3rd person i n general, unlike 1st and 2nd person forms, w h i c h are frequent and more varied.

The A c q u i s i t i o n of Kaluli

5.

5.6.

Locative

and Dative

565

Casemarking

The single locative suffix (-a ' t o ' , ' i n ' , at', and o n ' ) was used i n the earliest of samples o f all three children, around 24 months. I t was added to deictics, demonstratives, and directionals, and used w i t h a restricted set o f verbs including mesea ' s i t ' , die-foma ' p u t ' , and hamana ' g o ' . Locative expressions were used w i t h high frequency w i t h intransitive verbs and i n noun phrases. M o r e specific locative words, such as 'inside' and ' t o p ' , appeared at 2 6 - 2 7 months, and they too took the general locative casemarking. The first errors i n locative marking appeared i n M e l i ' s speech at 26 months and i n A b i ' s about a month later. Errors were o f several types, but most involved overmarking o f location and/or direction. Children's errors were o f the f o l l o w i n g types and were high frequency from 26 to 29 months: 4

4

Use of two contradictory locatives: M 26

*va here

ka mesea there sit: IMP

* 'Sit here there.' Use of a locative and a deictic/direction where only one is needed, marking both: M 26

*/ wood

wel top

-a we L O C this

-na diefoma L O C put:IMP

'Put (it) on top of the wood' A 28.2

*isasub under house

or

'Put (it) on this wood'

-a holo -na L O C down there L O C

delin be:RECENT PAST

(= isasuba delin) (= holona delin)

'It was under the house'

or

*id shit

-a dowo L O C be:PAST

a house

(= / we la diefoma) (= / wena diefoma)

-ya en L O C there

'It was down there' (= id a ena dowo)

'It was at the shit house' For these children, the casemarking was added to both the noun and the unneces­ sary deictic. Other errors included m i x i n g directionals for locatives such as wena 'on this' for bono 'that one over there (to the west or south)'. W h i l e the majority o f overmarking errors occurred w i t h locatives, similar patterns occurred w i t h the dative, genitive, and, to a lesser extent w i t h the ergative. The f o l l o w i n g example illustrates an error w i t h dative marking and the w o r d order o f the noun phrase.

566

Schieffelin

Meli 27.3 is with her mother, the researcher (BBS) and several children including the researcher's namesake, Babi. Thus there are two people named Babi present. Meli is giving her mother some sugar cane and says, *hono that one over there

Babì-mo dimina DAT give:IMP

*Babi-mo DAT

bono dimina

*Babi-mo DAT

hono-mo DAT

Mother:

Babi

hono-mo DAT

dimina eie salan likeUhat say:HAB

' "Give to that Babi over there" one says like that' I n this example, the first error made by M e l i is i n the order o f demonstrative and noun, an error that was made by all children occasionally until 28 months. The dative marker is suffixed to the noun. N o one responds to M e l i ' s utterance and i n her next turn she reverses the w o r d order, which is correct, but incorrectly attaches the dative to the noun, not the demonstrative (which like all nominal modifiers should carry the casemarking information). W i t h still no pickup from her mother, she formulates another utterance, this time attaching the dative to both noun and modifier. Her mother provides the correct model for this utterance, but M e l i does not repeat i t . W h i l e these K a l u l i children d i d not correctly attach (the homophonous) ergative, genitive, or instrumental suffixes to pronouns, they correctly added dative and locative suffixes to nonfocused forms o f the pronouns.

VERBS 6. G e n e r a l B a c k g r o u n d o n K a l u l i V e r b s I n contrast to K a l u l i nominal morphology, which is regular and relatively nonelaborated, verbal morphology is irregular and complex. K a l u l i verbs are composed o f stem material plus an inflectional suffix marking person and tense. Except for imperatives, number is not marked through bound suffixes, but is indicated through the use o f separate pronominal forms. There is no single base form o f the verb nor is there an infinitive. K a l u l i distinguishes the f o l l o w i n g major tenses: past, recent past, habitual, present, intentive 1, intentive 2, future. A l l tenses follow a regular pattern o f inflectional suffixing. Person is marked i n all tenses except recent past, nonrecent past, and habitual, w h i c h have single forms. I n declarative utterances inflections for person follow a two-way split: 1st and n o n - l s t (2nd/3rd). I n present

5.

The A c q u i s i t i o n of Kaluli

567

tense interrogatives, inflections follow a three-way split: 1st, 2nd, and 3rd per­ son. I n future interrogatives, the two-way distinction that is marked is 1st/2nd versus 3rd. U n l i k e nominal morphology which follows t w o casemarking para­ digms, verbal morphology only follows a nominative-accusative pattern. Aspect is infixed between the stem and the inflectional suffix. I t may be expressed lexically and w i t h a limited set o f prefixes as w e l l . Negation is marked by both prefixation and suffixation i n all forms except imperative, where i t is marked only through suffixation. Imperatives indicate information through suffixation about (1) number (sin­ gular/dual/plural), (2) time o f expected action (present/future), and (3) duration (+/—continuous), w h i c h is marked by infixing. I n negative imperatives, the negative suffix changes according to number/time. There are t w o different types o f verb constructions: medial and final. Final forms are found i n independent clauses, receive tense-person inflection, aspect, and emphatic and other discourse-level particles. They are sentence final. M e d i a l forms are found i n dependent clauses and do not carry information about person but do carry information about tense. I f the agent/actor i n the independent clause is different from that o f the dependent clause, that change must be signalled by a particular form o f the medial verb. This type o f reference marking is k n o w n as " s w i t c h reference" and is found i n languages that connect clauses by "clause c h a i n i n g . " The particular f o r m o f the medial verb i n dependent clauses i n con­ junction w i t h final forms i n independent clauses indicates the temporal rela­ tionship between the clauses: purposive, sequential, simultaneous. Several medi­ al forms (dependent clauses) may be chained together followed by a final form (independent clause). A number o f verbs are formed by concatenation o f a medial and final f o r m . For example 'take away' is constructed from diege + hamana 'having taken + go'; ' b r i n g ' is from diege + mena 'having taken 4- come'. I n fast speech, the final syllable o f the medial f o r m is dropped. Other complex verbs are constructed by using inflected forms plus stative verbs. W h i l e the linguist is able to determine the boundaries between what is stem material and what are inflectional endings, the child is not always able to make this distinction. I n the section that follows, these t w o components o f the verb w i l l be treated separately. 6.1.

Verb

Stems

Verb stems i n K a l u l i undergo considerable change when the various verbal suffixes are added. Predicting the form o f a given verb stem is difficult for individuals learning K a l u l i as a first or second language. According to Rule (1966), i n many o f the Central and South N e w Guinea Stock languages w h i c h are geographically adjacent to K a l u l i (such as H u l i , Foe) i t is possible to obtain from one verbal f o r m , the stem or root o f the other verb forms. I n a number o f languages i n this area o f Papua N e w Guinea, present singular imperative without

568

Schieffelin

any suffix is the base form w h i c h may be taken as the stem. F r o m this f o r m , all other grammatical forms o f the verb can be determined by adding the appropriate suffix to that stem, either as i t is, or after one or t w o slight morphophonemic changes have been made. In K a l u l i , however, there are many radical changes i n the stem between one grammatical form and the next. I n his analysis o f 100 verbs ( w i t h 10 different grammatical forms for each), Rule found that certain groups o f these gram­ matical forms were determined by adding the appropriate suffix to one particular form o f the stem. Similar stem forms: —The future imperative, negative imperative, present declarative (1st and non-lst), and recent past are all based on the same form of the verb stem. The rest of the forms can be determined if the future imperative is known. —The future declarative (1st, 2nd/3rd) and the 1st person purposive forms are all based on the present imperative form of the stem. —The past tense form is completely irregular. Thus, once the linguist knows the present and future imperative, 1st person future declarative, and past tense, other tense forms can be derived for many verbs. However, there is a sizable group o f verbs for w h i c h these rules simply do not apply, where the stem is completely changed for every f o r m . These verbs include some o f the most common verbs such as mena ' c o m e ' , hamana ' g o ' , mesea'sit\ maya 'eat', melea 'stay', and doma 'have', among others. Rule tentatively described six patterns o f stem vowel and consonant change, and the variations found w i t h i n each group. Particularly complex are the changes in stems w h i c h are multisyllabic. One major reason that they are so difficult to master is that these verb groups are not organized by any apparent semantic principles. A second reason is that there is no clear phonological pattern w h i c h determines the membership o f a particular form for a given class o f verbs. T o date, no phonological analysis is adequate to describe the changes for any single group o f verbs. The f o l l o w i n g short sample o f verb paradigms o f common verbs i n six pro­ posed classes should provide the reader w i t h a sense o f the possibilities o f stem changes for each f o r m . Imperative Present 1. blow tie up

Imperative Future

1st Person Future

Past (nonrecent)

foloma melema

falebi melebi

fomeno memeno

falo melo

hamana

henebi

heneno

ane

2.

The A c q u i s i t i o n of Kaluli

5.

come sit eat

569

Imperative Present

Imperative Future

1st Person Future

Past (nonrecent)

mena mesea maya

y ehi sebi nebi

mieno meseno meno

miyo sen mono

sama kama disema

selebi kelebi diselebi

s emeno kemeno dis emeno

siyo kolo dise

sofa sanama

sodebi sandebi

sqfeno s emeno

sofe sono

dima dimina

diyebi miyebi

dieno mieno

di dimi

diefoma

die sebi

diefeno

dife

3. speak put on (clothes) put into (bags)

4. cook hit

5. take give

6. put down (objects)

6.2. Acquisition

of Verb

Stems

Imperative forms (both positive and negative) were used w i t h high frequency by all children. Imperatives were used as single words and i n m u l t i w o r d con­ structions from 25 months. Imperatives were formed even when a child d i d not know or had forgotten the stem. Taking the " w o r d " ege which is the hesitation particle but also means 'watcha ma call i t ? ' , the child w o u l d add the suffix -ma (egema for affirmative) or -sabo (egesabo for negative). (Ege is infected for tense and receives nominal cases i n syntactic constructions by children and sometimes adults when the specific w o r d is not forthcoming.) Children were not corrected when they used ege as a noun or verb. A n analysis o f the errors i n early verb usage suggests that although K a l u l i does not have a single base form o f the verb, K a l u l i children do use one form as such. Taking the present imperative as a base form they alter the final v o w e l . For example, when a child d i d not use the correct form o f the 1st person present o f the verb, one strategy was to use the imperative as the 1st person, w i t h or without the 1st person pronoun and to change the final vowel (usually -a) to -e.

*ne I

mene come: IMP

correct form:

ne I

yol come:l:PRES

T come'

or

'I am coming'

This morpheme (-e) is not one o f the possible person/tense inflectional endings, but the highest frequency verb final emphatic marker. W a n u , whose acquisition

570

Schieffelin

of verbal morphology was somewhat slower than the others, provided the data for the majority o f these errors, especially during the time he was 2 5 - 2 8 months. However the other t w o children had examples o f this type o f error, especially i n samples before 27 months. A variation on this type o f error was to use another emphatic particle, -loga (meaning T said' but because o f contraction this meaning is not available to the child), suffixed to the present imperative or all but the final syllable o f the imperative. These errors suggest the saliency o f these final particles and their possible confusion as inflectional endings. For example, while performing the following actions M e l i said,

M 24.3

ERROR

*BASE

BASE

FORM

ne I

diefoma PRES:IMP

dietebi FUT:IMP

dietol 1:PRES

tima PRES:IMP

tiebi FUT:IMP

tiol 1:PRES

* disfoga put down

T put down' M 26

ne I

* timóloga take out of fire

'I take out of fire' The second type o f evidence to support the present imperative as the child's base form comes from errors made when attempting to form 1st person present tense. These show a close approximation o f the correct inflection, added to a stem that is modeled from the present imperative rather than the future imper­ ative, which is correct. For example, while commenting on his o w n ongoing actions Wanu said, ERROR

W 25

*dagulo peel/shell

BASE

FORM

daguma PRES:IMP

dudebi FUT:IMP

dudol 1:PRES

gadama PRES:IMP

gidebi FUT:IMP

gidol 1:PRES

'I peel' W 26.3

ne *gadolo I put together T put together'

As mentioned above, the future imperative, w h i c h is the basis for the present tense, is not used w i t h high frequency by parents, nor is i t used by children under 30 months. However, there is no clear evidence that children are learning verb paradigms or the rules o f stem formation based on particular forms as i n the linguistic description above.

5.

The A c q u i s i t i o n of Kaluli

571

Other errors appeared throughout the tenses, not just first person present tense. For example, there were errors for all children i n the use o f the verb daguma 'peel' which is used to request or talk about peeling bananas and other fruits as w e l l as shelling crayfish. I n the first example ( M 27.3) the present imperative was used for the stem.

M 27.3 'peel'

*daguabeyo R E C E N T PAST

M 30.2 'peel'

*dagumeno 1:FUT

daguma PRES:IMP

dudebi FUTTMP

dudabeyo R E C E N T PAST

daguma dubzno PRES:IMP 1:FUT

I n the second example ( M 30.2) M e l i ' s guess is a good one, 1st person future is usually based on present imperative but this verb has an additional stem change. This error, made at a later period, suggests that M e l i is using the paradigm to form the future tense, but this f o r m , like many others is irregular. Other problems w i t h stem formation are due to too much or too little stem material. For example, M 27.3 *kamama from kama 'put on clothing' There are no imperatives that have the canonical shape Consonant + amama. However, numerous imperatives follow Consonant + ama and Consonant

+

anama. I n this example M e l i has lengthened the bisyllabic stem adding -ma (part o f the imperative ending) to the imperative kama. I n this example there is not enough stem material for the formation o f the 1st person present. M 26.

ne I

*fole blow

from foloma blow:PRES:IMP

(= folol) 1:PRES T blow the fire'

When these types o f errors occurred, they were almost always corrected by the adult. W h i l e children d i d make a number o f errors i n choosing the form o f the past tense, such as M 26 *dimiya for dimilmi

' g i v e : P A S T ' , there was only one

example o f the addition o f inflectional material (recent past) to that past f o r m . The utterance, w h i c h was repeated by the child several times, was not under­ stood and dropped. The lack o f overgeneralization o f inflectional material to past tense forms o f the verbs suggests that K a l u l i children may be learning these single past forms as separate lexical items, and do not involve them i n the morphological processes o f other verb tenses.

572

Schieffelm

Stem errors were no more frequent i n certain verb classes than others, though forms that were irregular produced the highest number o f errors. There was no systematic error pattern based on the phonological shape o f the verb. One prob­ lem that faces the child concerns p i c k i n g the correct form on w h i c h to base the stem. A n important point to emphasize is that i n K a l u l i the present imperative is usually the base form for the future tenses, including intentive 1 and 2. The present declarative is based on the future imperative. This " m i s m a t c h " i n the temporal basis o f stem formation is an important factor i n errors i n stem forma­ tion for the child.

W 29

ERROR

*BASE

BASE

FORM

*hal£meni 'rub':INT:l

halebi FUT:IMP

hama PRES:IMP

hameni INT:1

T am about to rub.' M 27.3

jun asparagus

kos leaf

NEUTRAL

*gidef£nigo- bale burn:l:INT:2 EMPH

(= gilimenigobale) 'I am about to burn asparagus leaves Both children chose the future stem for intentive 1 and intentive 2, w h i c h was not correct. The stem should have been based on the present imperative. W h i l e some stem errors had to do w i t h the incorrect selection o f a consonant, other errors were due to incorrect vowels as i n *bedol for bodol T see' where 3rd person is bedab. Some errors d i d show resemblance to their imperative stems; others were simply misshapen and often misunderstood by those to w h o m they were directed. 6.3. Tense

Inflections

I f one has the basic four stems that are presented above, one can add the following tense inflections to form declarative K a l u l i sentences. Recent past: Present tense: Intentive 1: Intentive 2: Future tense: Habitual:

lst/2nd/3rd 1st 1st 1st 1st lst/2nd/3rd

-abe -ol -eni -gol -menol-wnol-eno

2nd/3rd *2nd/3rd 2nd/3rd 2nd/3rd

-ab -gab -meibl-eib

As one can see from the above inflectional paradigm, the tense endings are regular and do not change according to verb class. Tense inflections may be followed by a variety o f discourse particles marking, for example, successive degrees o f emphasis ( w i t h past tense agreement), repetition, and quotation.

5.

6.4. Acquisition

of Tense

The A c q u i s i t i o n of Kaluli

573

Inflections

For this analysis, three children's spontaneously produced verbs (from both single and m u l t i w o r d utterances) were examined for patterns i n the acquisition o f tense inflection. Both transitive and intransitive verbs were analyzed and only utterances that were situationally appropriate were included. The verbs were divided into t w o categories, 1st person and 3rd person agents/actors. The 2nd person was not considered since the majority o f 2nd person utterances were present imperatives. For all children, only inflected verbs were produced and they appeared simul­ taneously as single words and i n syntactic constructions. There was no sequential pattern. This does not seem to be an artifact o f sampling since for M e l i there were no 1st or 3rd inflected intransitive forms i n the first sample, while i n the following sample, there were both, w i t h a majority o f inflected forms appearing in syntactic constructions. Perhaps the most interesting finding from this analysis is the distribution o f tense inflections w i t h i n each sample according to 1st and 3rd person verb inflec­ tion. Five tenses were considered: past (single form for 1/2/3), present, intentive 1 (or inceptive, 'about t o ' , restricted to 1st person and used when agent/actor announces o w n intentions but has not yet initiated action), intentive 2 ('starting t o ' , action has been initiated or there is evidence that i t is about to be, i n contrast to intentive 1, this is not restricted to 1st person), and future. For all children i n all samples, there are striking differences between 1st and 3rd person usage. For 3rd person forms, usage was concentrated i n t w o tenses, past and present. Out o f 23 samples (combined samples o f three children) only t w o samples had over 7% o f their utterances i n tenses other than past or present tense forms, and these exceptions were i n later samples (over 30 months). This means that most o f the time when children are using verbs and talking about others, they are talking about completed or ongoing action. They infrequently talk about what other people are about to do (intentive 2 ) , or what others w i l l do (future). W h e n they do talk about what others w i l l do, one finds threats ("someone w i l l say some­ t h i n g " ) or responses to queries about the future actions about others. I n addition, following adult patterns o f use children never extended use o f the intentive 1 tense to the third person. Thus, when talking about 3rd persons, children talked about what was certain and observable, and most o f this talk concerned the past and ongoing actions o f others. Finally there is an additional cultural issue that may be a factor. K a l u l i adults claim that one does not k n o w what another person thinks. This is realized by a dispreference for talking about what another person might think or what they w i l l be doing. Adults do not frequently speculate or talk about these topics; to do so could be considered gossip. However, without further research i t is not possible to ascertain whether this cultural pattern affects acquisition patterns, or whether cognitive and pragmatic constraints carry the explanation.

574

Schieffelin

When children talked about themselves as agents or actors i n the earliest samples, the majority o f their utterances were i n the present tense as they de­ scribed their o w n actions. However, by about 27 months use o f the five tense inflections was distributed more evenly throughout each sample. Talk o f already initiated and future actions dominated the speech o f all three children. Between 10% and 15% o f their talk concerned past actions. Talk about what they were going to do, but had not yet begun (intentive 1) ranged between an average o f 5% for M e l i and 12% for A b i and W a n u . A g a i n , these children preferred to talk about what was observable and knowable. As agents, they knew their o w n intentions and future actions best and these were the topics o f talk. What is clear from this analysis is that K a l u l i children are not using verbs inflected for all persons and all tenses. Before the age o f 30 months, there are gaps i n the verb paradigms that a linguist w o u l d construct. Due to cognitive and pragmatic constraints, children use only certain forms o f a given verb, and not others when they begin to acquire the verbal system. This is particularly clear i n a language like K a l u l i where each verb is always inflected for person and tense. It is less clear i n a language that lacks this type o f verbal morphology. The children i n this study used, for example, dimina ' g i v e ' i n the present imperative, but not i n the 1st person present or the negative imperative. The verb dima 'take' was used i n the 1st person present, w i t h the negative imperative, and w i t h a 3rd person agent i n the past, but rarely i n the imperative or the future. Thus while inflected forms are produced from the beginning o f the child's use o f language, they are restricted to a limited set o f verbs. Furthermore, for each verb there may be a particular set o f inflectional forms w h i c h the child uses, and others that he or she does not. For example, the verb omina, 'having chewed g i v e ' , only occurred in the imperative, since children asked their parents to chew and give food, but had no other pragmatic function to express w i t h that verb. Overall, the inflectional verb endings do not pose major problems i n acquisi­ tion. T w o o f the three children d i d not have difficulty w i t h their use i n de­ clarative utterances. The reasons for this relative ease o f acquisition are as follows. K a l u l i verb inflections do not change form according to verb stem class. Inflections are both w o r d and sentence final. I n addition, the use o f emphatic particles f o l l o w i n g the inflection may highlight the perceptual saliency o f the tense inflections. This is supported by evidence from early use o f tense inflec­ tions followed by different emphatic forms as illustrated by honol ' g o : l : P R E S ' showing different emphatic endings honoke and honolo. Moreover, there is a high proportion o f single inflected verbs used as sentences i n both casual adultadult verbal interaction, and adult-child speech as w e l l . However there were a number o f recurring errors i n the data for all three children. Before 28 months t w o children produced utterances o f the f o l l o w i n g type: *ne diabo T take' instead o f ne diol when referring to their o w n actions. They used 1st person pronouns and inflected the verb for 3rd person -ab adding the final particle -o used i n calling out. The reverse error (the use o f 3rd person

5.

The A c q u i s i t i o n o f Kaluli

575

forms for incorrectly referring to the self) d i d not occur. I t was clear from the context and the corrections o f adults that children were speaking about their o w n actions. This error i n inflection occurred w i t h a number o f verbs i n several different verb classes, for example *yabo (= yol) 'come' *henabo (= honol) 'go', *bedabo (= bodol) 'see/look', *alilabo (= alilol) ' l i e d o w n ' , *adulabo (= adulol) 'step o n ' , among others. There is no clear single cause o f these errors, and I suggest several possible related causes. First, this error may be related to the fact that verbs o f affect and internal state (ne tagidab ' I ' m afraid', ne mayab ' I ' m h u n g r y ' ) w h i c h occur at this time, only take 3rd person inflection and children may be overgeneralizaing from those forms. They never inflected those forms for the 1st person. Alternatively, the 1st person ending -ol may appear phonologically similar to the nominal case endings (absolutive and neu­ tral case, both signalled by -o), whereas the 3rd person -ab is more distinctive and occurs only w i t h verbs. Therefore, the 3rd person may be seen as less ambiguous than the 1st person inflection. A third possibility is that children misanalyze the stem o f the verb as i n the case o f diab 'take' (instead o f di-) and add the emphatic marker -o as the final inflectional material. W h i l e the form diab is grammatically correct i n some contexts, it is not correct i n conjunction w i t h a 1st person pronoun nor when referring to one's o w n actions. Errors regarding correct inflection for 1st person persisted i n certain verbs for A b i until 31.2 and for Wanu until 32.1 months. Another systematic error occurs w i t h inflectional endings on present tense interrogative utterances. W h i l e declarative sentences make a two-way distinction for marking person/tense inflection (1st versus n o n - l s t person), interrogatives makes a three-way distinction, adding a different 2nd person subject ending before adding the interrogative particle. I n their use o f interrogatives, K a l u l i children do not make this three-way distinction and base all o f their inflectional patterns on the present declarative. Therefore, when they are addressing some­ one and use the 2nd person pronouns they use the 2nd/3rd person declarative inflection along w i t h the 3rd person interrogative particle. This error pattern does not occur i n 2nd person formulaic interrogatives, but only i n non-formulaic expressions until about 28 months. Another departure from the most conimon paradigm ( l s t / n o n - l s t person) is found i n future interrogatives. I n these utter­ ances, 1st/2nd person is marked identically, w i t h 3rd person marked differently, unlike the usual 1st—2nd/3rd pattern o f future declarative. W h i l e only a few future 2nd person interrogatives were produced, they were incorrectly inflected for 3rd person until 29 months. Thus children were overgeneralizing from the declarative tense inflections across sentence types, w h i c h i n most cases was not correct. Other overregularization occurred as children tried to make specially inflected verb forms conform to their developing knowledge o f inflections. This is clearly seen when children try to regularize certain formulaic expressions w h i c h do not conform to the paradigmatic use o f the inflectional system.

576

Schieffelin

There is a set o f teasing formulae w h i c h one uses when offering food or an object without the intention o f ever g i v i n g i t to the person. I t is used to tease people who beg for or want something you have when you wish to taunt them w i t h the fact that they cannot have i t . Instead o f using a form w i t h a negative, such as ' y o u don't eat/take' etc., K a l u l i has a special construction without a negative, used only i n these teasing and confrontational contexts. I t is composed of a focused form o f the 2nd person pronoun, gi ' y o u ' , w i t h the 3rd person form of the verb minus the final consonant. This results i n the f o l l o w i n g forms: REGULAR FORM PRESENT FUTURE

gi gi gi gi

TEASING FORM

'you 'you 'you 'you

nab diab meib dieib

eat' take' will eat' will take'

gi gi gi gi

na dia mei diei

'you 'you 'you 'you

don't eat' don't take' will not eat' will not take'

The f o l l o w i n g teasing exchange comes from M e l i 28.3. I n this example we see how M e l i is trying to make this special form into an inflected form during a sequence o f teasing and direct instruction using elema. Meli's Grandfather has brought a small marsupial (yesi) to the house and has given it to Meli. Meli, her mother, and mother's sister, Faili are sitting around the fire. After talk about how Grandfather had killed it and given it to Meli, Faili reaches for it. (1)

G'fa —> Meli =#> Faili:

a! no!

momieno N E G give:l:FUT

-kef EMPH

elema. say:IMP

'No! I will not give! say like that.' (2)

nineli I alone

mene eat:l:FUT

nineli

meno.

Faili,

gi you

na! (don't) eat

'Only I will eat. Only I will eat. You don't eat! say like that.' (3)

Meli —> Mother:

no! mother

Faili

we sanama! this hit: IMP

'Mother! Hit this Faili!' (4)

G'fa —>- Meli =#> Faili:

gi

na!

elema.

'You don't eat! say like that.' (5)

Meli —» Faili:

gi you

nan eat:HAB

-he! EMPH

'You usually eat!'

elema. say:IMP

5.

(6)

Mother —> Meli =#> Faili:

gi you

The A c q u i s i t i o n o f Kaluli

mei! will (not) eat.

577

elema. say:IMP

'You will not eat! say like that.' (7)

Meli —» Faili:

gi you

meib eat:3:FUT

-ke! EMPH

'You will eat!' (8)

G'fa - » Meli =¥> Faili:

gi

mei!

gi

neli! will (not) keep eating

gi

na!

'You won't eat! You will not keep eating! You don't eat!' In her first turn i n this exchange ( 3 ) , M e / / does not correctly repeat the form that her grandfather has said, gi na ' y o u don't eat', but instead she tells her mother to hit F a i l i . Her grandfather repeats his instruction to M e l i (4) telling her to say the teasing expression to F a i l i . Instead, M e l i utters the phonologically similar habit­ ual form o f the verb plus an emphatic particle (5). Her mother then tells her to say gi mei ' y o u w i l l not eat' (6) and M e l i produces the 3rd person future w i t h an emphatic particle (7). None o f M e l i ' s utterances convey the appropriate mean­ ing, as she produces utterances w i t h the regular inflectional endings rather than the phonologically similar teasing forms. Then her grandfather uses this teasing construction i n three different forms ( 8 ) , future, future + continuative aspect, and present. M e l i does not repeat. The discussion then turns to the details o f the animal, followed by further exchanges instructing M e l i i n using these formulaic teasing expressions. She not repeat them, but instead produces forms w i t h the regular inflectional endings (future and habitual) as she has done before. Finally her mother makes an important l i n k for M e l i i n a discourse sequence that has been documented elsewhere for the development o f rhetorical questions used for teasing (Schieffelin, i n press). She says, naino our

yesi

-yo NEUTRAL

e -lo 3P:DAT

mie give

-sabo! NEG:IMP

'Don't give our yesi to her!' ni my

yesi

-yo NEUTRAL

gi you

mei-ke eat EMPH

elema! say:IMP

'You (won't) eat my yesil say like that!' Here we see the meaning and consequences o f the teasing form made explicit by sequencing the negative imperative and the teasing, ' w e w o n ' t give i t , and say. . . '

578

Schieffelin

The talk then turns to all the reasons w h y Faili cannot have the yesi, followed by another set o f elema instructions to repeat these formulaic expressions, w h i c h M e l i still does not do. Instead she tells her mother to hit Faili to prevent her from taking the yesi. This is not acceptable to the adults. Later when Faili starts to beg for the yesi from M e l i , her grandfather again instructs her to say, gi rial and M e l i finally says i t w i t h the appropriate force o f delivery several times i n succession. Her grandfather then yells w i t h delight, "She says gi na7" l

I n addition to this example providing evidence about the child's attempted regularization o f irregular formulaic expressions, we see how K a l u l i adults assist their children i n mastering the use o f routine expressions and how this enables the child to participate i n culturally appropriate ways. Family members provide different cues to help the child understand the meaning o f this particular ex­ pression, which has a negative force without any surface negative marking. Without actually explaining the form to her, adults are persistent i n encouraging her to say this expression correctly through the use o f elema, direct instruction. Its use is contextualized w i t h i n a recurring framework o f teasing. The child is encouraged to think that her correct performance is critical to the desired out­ come o f the immediate situation. For the K a l u l i , the appropriate use o f these expressions, and many others like them, are important indicators not only o f linguistic competence but social competence as w e l l . W h i l e the following type o f error d i d not occur frequently, it illustrates what the child does when trying to express a more complex form o f the tense without knowing the morphological devices for doing so. I n this case the child resorts to lexical means. Meli 27.3 has found a crayfish shell on the floor. Showing it to her mother: Meli:

no, mother

galen crayfish

we this

ne I

*nol eat:l:PRES

komo finish

-lob EVID

'Mother, I eat finish this crayfish' Mother:

gilo nalogo you ate: already :2nd/3rd 'You ate already'

Meli:

ne I

*nalogo ate already:3

(= nologo)

ne

* nalogo-lob 3: EVID

-e EMPH:PAST

I n this example, M e l i is trying to express something other than the simple or recent past, since that form o f the verb was already i n her productive vocabulary and used many times i n the same sample. The verb phrase she constructs is not used i n the adult language, and has several errors i n i t . I t is a combination o f 1st person present tense and the w o r d kom meaning 'the end, finish, enough' w i t h

5.

The A c q u i s i t i o n of Kaluli

579

the evidential particle meaning T see'. The present tense w h i c h indicates on­ going action cannot be used w i t h the w o r d meaning completed. The evidential cannot be used w i t h the 1st person present tense since one cannot claim that one observes one's o w n ongoing action. However, the evidential is often used w i t h the w o r d kom. Her mother's response uses a verb w i t h an inflection that M e l i is not familiar w i t h . M e l i repeats i t w i t h the 2nd/3rd person inflection and then again w i t h the evidential particle followed w i t h an emphatic particle, creating an incorrect utterance. After M e l i ' s several incorrect attempts to use this f o r m o f the verb, her mother responds w i t h a simple question: Mother:

no it

gilo you

mono eat:PAST

-wo? Q

'Did you eat it?' Child:

Ni I

mono eat:PAST

'I ate'

They then proceeded to discuss w h o had given the crayfish to M e l i and from what river i t had come, all culturally important aspects o f the topic. M e l i d i d not attempt to use that particular past form o f the verb until 30.2. I t was still not under her control at that time. However, she was w o r k i n g on i t by trying to change the stem and inflectional material but continued w i t h the verb i n the 3rd person. nowo another

ni I

*neliyogolnowo *eat:3

ni

*nalilolo

(= nologo) eat:l

'I already ate another (one)'

6.5. Other

Problems

with

Verbs

K a l u l i has many verbs w i t h i n particular semantic domains. For example, i n English the meaning o f the verb 'put' is altered by changing the preposition, put on I off I in I over I down, etc. I n K a l u l i , each o f these distinctions is indicated by different lexical items. Some o f the differences include animate/inanimate, nature and number o f objects, duration o f action, end result o f affected object, and location. Some verbs w h i c h are semantically related have minor pho­ nological differences, such as tima ' p i c k leaves', 'take food out o f fire' and tuma 'pick (certain) fruits'. Others involve single infixes, multiple infixes, prefixes, changes i n suffixes and combinations o f the above. Other forms bear no re­ semblance at all to one another i n spite o f some o f the same close semantic relationships that one finds signalled morphologically i n other verbs. The tre­ mendous amount o f complexity and variation that one finds i n the verbal system

580

Schieffelin

gives children difficulty w e l l after 30 months. W a n u , i n particular, had extensive errors involving verb semantics. Errors i n choosing the correct verb were made by all children for several reasons: The child might not k n o w the form at a l l , producing the wrong stem and ending. This resulted i n a virtually incomprehensible utterance. Alternatively, i f the child d i d not k n o w the specific verb, ege was used and inflected accordingly. Or, a single form w o u l d be overextended to cover too much o f a given semantic field. For example, W a n u used the verb fagema meaning 'untie' to refer to all acts o f opening as w e l l as closing and t y i n g . Wanu used this single verb to refer to actions covered by four different verbs at 26.3 months, listed below. He was always corrected. fagema fagema fagema fagema

for for for for

sogoma hagoma melema kolama

'untwist' 'pick off tree bark' 'tie up' 'open up (a door)'

However, by 28 months, this type o f problem was seldom encountered. Other error patterns w i t h verbs came from several sources. A child could choose a verb w h i c h was (1) phonologically close to the correct f o r m , but semantically unrelated, (2) phonologically close and semantically related, (3) semantically related, but phonologically different, (4) w i t h i n the correct seman­ tic domain but an antonym ( T i r e dies d o w n ' 'fire flares u p ' ) and (5) totally unrelated. The first pattern was a relatively early strategy and virtually disap­ peared by 28 months. A l l children produced utterances o f the 2nd, 3rd, and 4th type through the latest samples, over 32 months. These w o u l d usually be cor­ rected and the child w o u l d be asked to repeat the correct f o r m . Only W a n u produced a fair quantity o f utterances o f the 5th category, but by 27 months they were only occasional. The tasks that the linguist faces i n understanding these verbs is very different from the child's task. First, the linguist seeks regularities w i t h i n verb paradigms, and determines the morphophonological relationships between the various parts o f the paradigms. A second task is to determine whether verbs fall into major classes, w i t h the hope o f establishing some predictability o f forms. As we have seen, i f certain forms are k n o w n , such as the imperative i n both the present and future forms, other parts o f the paradigm can be filled i n . The child faces a very different set o f tasks. He or she cannot approach the task paradigmatically, as can the linguist. As a result, due to cognitive, linguistic and pragmatic factors, children are learning the verb system piece by piece. For example, they have little basis upon w h i c h they can predict the stem o f the verb from one form to another. The " m i s m a t c h " between present and future tenses for stem formation causes the child to make errors i n trying to set up a logical system based on temporal similarity. Another "non-facilitating factor" comes from the verbal input. I n clarifica­ tion sequences, the f o r m o f the 1st person declarative has little resemblance to

5.

The A c q u i s i t i o n of Kaluli

581

the 2nd person imperative or the past tense. The child cannot be assisted by the similarity o f forms across turns that one w o u l d get i n clarification sequences i n some languages. For example, i n English i f the child says " c o m e ! " to someone who then requests clarification, that speaker w i l l usually say, " c o m e ? " . I n K a l u l i , however, except for the past and the habitual forms, i f the child says, mena! come:PRES:IMP' the clarification request will be yolo? 'come:l:PRES:Q'. Before the child has the appropriate inflected interrogative form, he or she repeats the adult's utterance w i t h some modification to the final syllable w h i c h is where inflection occurs. The most common modification is to change the final v o w e l and keep everything else intact. This o f course is incor­ rect, but it does indicate that the young child is aware o f inflections before controlling them and knows that the form has to change, especially the inflec­ tional ending. O n l y through the direct instruction sequences i n v o l v i n g elema does the child get the "adjusted" f o r m . This facilitates acquisition. 4

6.6.

Aspect

Aspect is complex i n K a l u l i , and to date there is no detailed account o f this system. Aspect is predominantly signalled through infixes between the stem and inflectional material, though i t is also indicated through prefixes. U n l i k e tense inflection, w h i c h is obligatory, aspect is optional and may be used when a speaker wishes to highlight some feature o f the situation. Children under 33 months only infrequently mark aspectual information, and much o f the time they do so incorrectly. One way i n w h i c h children mark duration is through repetition of the utterance. For example, when watching a leaf float downstream, children w o u l d say, hcnabo, henabo ' i t goes, it goes' etc., until i t was out o f sight. Repetition is used i n adult narratives as a stylistic device for expressing duration, while i n conversation, morphological means o f expressing aspect are more usual. One o f the earliest morphological forms used to express aspect is the prefix omeaning ' s t i l l , i n the process o f as when commenting on one's o w n or another's actions. A t 30.2 M e l i was prefixing many verbs this way, some appropriately and some not. The prefix was clearly being overused, w h i c h was indicated by her mother asking, " W h y are you saying o- all the t i m e ? ! " 6.7.

Complex

Verbal

Constructions

So far our discussion o f verbs has concerned simple verbs and single indepen­ dent clauses. However, this is only one part o f the verbal system. I n addition to simple verbs, K a l u l i combines verbs to f o r m serial and complex verbs. These w i l l be discussed below. Another aspect o f the verbal system that w i l l be exam­ ined is the system o f clausal relationships. I n English, t w o or more clauses are joined together by conjunctions. The verbs i n these clauses are not very different from one another. Coordination and subordination are indicated through lexical means. I n K a l u l i , like other non-Austronesian languages, the organization o f combining clauses is quite different. Longacre (1972, p . 2) describes this "clause-chaining' as follows.

582

Schieffelin

In place of sentences that are composed of several coordinated clauses joined together by verbs of equal rank in all clauses, we find sentences which contain a clause with a distinctive verb form occurring but one in the entire sentence (usually at the end) preceded by other clauses with verbs of other structure. We may say we have a structure with an engine at the end and a bunch of cars hooked on preceding it. I f the agent/actor o f the t w o verbs is the same i n t w o adjoining clauses, there is one set o f verbal suffixes. I f the agent/actors are different i n the adjoining t w o clauses, another set o f suffixes is used. The verbal suffix i n the first dependent clause not only carries the temporal information about that predicate, but i n d i ­ cates that the f o l l o w i n g clause (dependent clause) has a different agent/actor. This is called switch reference. Examples o f complex clause chaining and switch reference are not found i n the speech o f children under 3 years. They are rarely used i n direct instruction sequences w i t h elsma, though they are used i n speech to children. Instead o f using the morphosyntactic devices found i n clause chaining w i t h switch refer­ ence, children from about the age o f 30 months sequence propositions as i n the following example. Meli 32.2 is opening a piece of wire used as a gate to keep puppies in a particular area of the house. gasa dog

we this

-na sab L O C sit:3:PRES

'(The) dogs stay in here' gasa

we

-na tandomeib cross:3:FUT

'(The) dogs will cross on here' asido -kc close: 1:PRES EMPH 'I close (it)' W h i l e M e l i ' s utterances are not " w r o n g , " they are not linked using K a l u l i morphosyntactic devices. They sound, i n fact, " c h i l d i s h . " A n adult w o u l d chain these propositions i n the f o l l o w i n g w a y , gasa

wena meseiki

tok

ami

tandolowabeniki

asifeno.

'So dogs stay here, so (they) won't cross (the) door, I'll close (it).' The ability to use clause chaining (and switch reference) is one o f the achieve­ ments o f a competent speaker. I t is what makes the language sound " K a l u l i . "

5.

The A c q u i s i t i o n of Kaluli

583

Speakers use clause chaining not only i n longer stretches o f discourse, but i n fairly short expressions as i n the example above. I n K a l u l i there are complex, serial verb forms that express notions o f purpose, as w e l l as the temporal relations o f simultaneity and sequence. These notions are not expressed lexically, but morphosyntactically. Y o u n g children use these ver­ bal constructions w i t h a single agent/actor i n both clauses from about 2 7 - 2 8 months. They seem to be the building blocks o f more complex syntactic con­ structions i n K a l u l i before switch reference is used. Since data are only available for children w h o used the same agent/actor, background material for only those structures w i l l be given. Data from older children are needed to document the development o f complex clause chaining, one o f the most distinctive aspects o f language development i n K a l u l i . 6.8. Subordinate

Verbs

of

Purpose

Subordinate verbs o f purpose are used to express the reason or goal o f an individual's actions. For example, oga pandanus

meni eat:PURP

menol come:IMP

'Come (in order) to eat pandanus!' Ne I

a house

-ya sqfeni L O C cook:PURP

honol go:l:PRES

'I am going to the house to cook.' The medial verb is a constant form and is identical to the tense intentive 1 (which alone can only be used for 1st person). The final verb can be inflected for any tense. I n rapid speech, the t w o verbs are concatenated and the final syllable o f the medial form (-ni) is deleted (except when followed by a vowel-initial final verb). W h e n the subject o f the t w o verbs is not the same, a different form o f the medial verb is used. 6.9. Acquisition

of the

Purposive

The first complex verbal forms produced by children express purposive rela­ tions. These forms emerged between 2 4 - 2 6 months. They were used frequently in direct instruction w i t h elema as imperatives. The children i n this study ex­ pressed purposive relations only i n utterances where the subject was the same for both verbs. Purposive utterances w i t h different subjects for each verb are more complex linguistically as w e l l as cognitively, and were relatively late acquisi­ tions (after 32 months). When purposive verbs are first used, they occur w i t h only three medial verbs, dima 'take', mesea ' s i t ' , sanama ' h i t ' . M e d i a l verbs have a single underlying form w h i c h is identical to intentive 1. W h e n children use these medial verbs i n

584

Schieffelin

combination w i t h final verbs, they contract the medial verb, deleting the final syllable (-ni) except w i t h vowel-initial final verbs. However, because o f contrac­ tion and the infrequent occurrence o f vowel-initial final verbs, it is unclear whether i n the early stages o f acquisition children realize that the forms for the medial verb and for intendve 1 are the same. They had no difficulty using this same verb form as a single verb or a concatenated one later. This may be due to the similarity o f the meanings o f the t w o , intention and purpose. W h i l e adults and older children use a variety o f final verbs, children combine medial forms w i t h only t w o final verbs, hamana g o ' and mena ' c o m e ' . The final verbs occur predominantly as imperatives, 1st person present, and 3rd person past. For example, children produced the f o l l o w i n g utterances: 4

Meli 24.3

die take:PURP

-honol go:l:PRES

'I'm going (in order) to take' Wanu 25.1

mese sit:PURP 4

Meli 26

-menol come:IMP

Come (in order) to sit!'

jun

sieni pick:PURP

-ane -yo? go:PAST Q

'Did (she) go to pick jun?' It is important to mention that when children use a concatenated verb w i t h ' g o ' , they are referring to change o f location, and as i n the example above, are talking about their o w n or other people's movement and the reason for i t . Almost all medial verbs used were transitive verbs. A s children increased their knowledge o f inflectional endings for simple verbs, these inflections were also used on the final verbs i n concatenated forms. The children i n this study d i d not add final verbs other than 'come' and ' g o ' during the course o f taping (to 33.2 months), but did increase the number o f medial verbs. For all children, i n spite o f the limited combinations, purposive relations were the most frequently expressed complex forms i n all samples. 6.10. Past Consecutive

Action

Past consecutive action is also expressed using serial verbs. For this, the medial verb is based on the future imperative minus the imperative ending plus the suffix -esege (+/—morphophonemic changes). The final verb is inflected according to the usual procedures. For example, ke fish

nescgs eat:PAST:CONSEC

yebi come :FUT: IMP

'Having eaten fish, you will come'

5.

dies ege take:PAST:CONSEC

585

mieno come: 1 :FUT

'Having taken I will come.' diesegs take:PAST:CONSEC

The A c q u i s i t i o n of Kaluli

(= 'bring')

heneno go: 1 :FUT

'Having taken I will go.' gulssege break:PAST:CONSEC

(= 'take away')

mina give:IMP

'Having broken in half, give.' These medial forms are usually contracted i n fast speech, dropping -sege, except when the final verb is vowel-initial or when the imperative is i n the future. I n a l l o f the examples, the subject is the same for both verbs. W h e n there is a switch, both forms o f the verbs undergo morphological changes. 6 . 1 1 . Acquisition

of Past Consecutive

Action

Serial verbs expressing past consecutive action emerged about one month later than purposive verbs, around 2 5 - 2 7 months. One medial verb dominated these utterances, dima 'take', especially i n combination w i t h t w o final verbs, mena 'come' and hamana ' g o ' . A s i n the examples above, these t w o verbs were used to talk about the change o f location o f objects, their being taken away 'having taken + went' and being brought to different individuals. There were no errors i n deixis. One reason for this may be that i n K a l u l i the deictic verbs 'come' and ' g o ' are part o f the surface form o f the serial verb and the meaning is explicit. As w i t h verbs o f purpose, additional medial verbs were added. Only one additional final verb, ' g i v e ' , was used w i t h high frequency for all children. 6.12. O t h e r C o m p l e x C o n s t r u c t i o n s Additional complex constructions d i d not become productive before 30 months. A r o u n d this time children used formulaic expressions, such as ' D o y o u see h i m going there?' w h i c h were composed o f an interrogative + verb. Differ­ ent constructions used correctly w i t h auxiliary verbs ( i n 3rd person), were used in both positive and negative constructions, such as, Meli 30.2 trying to chew something tough, moNEG

omeno dowab chew:l:FUT do:3:PRES

T am unable to chew' de fire

-yo NEUTRAL

gilimeno dowab light: 1:FUT do:3:PRES

' I want to light the fire'

586

Schieffelin

However, due to the lack o f data beyond 32 months i t is not possible to discuss the development o f productive forms. Clearly i n K a l u l i , language acquisition after 32 months w o u l d show the development o f additional complexity. What the current data demonstrate are the foundations o f this development. This is only the beginning o f a complex acquisition pattern.

EMPHATIC A N D OTHER

DISCOURSE

PARTICLES

K a l u l i has an extensive range o f emphatic, affective, and discourse particles. They are suffixed to almost all parts o f speech, are utterance final, and serve a range o f functions. Emphatic and other discourse particles are among the earliest morphemes to be acquired. W h i l e emphatic particles are used correctly i n some contexts from their earliest production, there are errors due to overgeneralization. As mentioned above, one set o f emphatic particles was incorrectly used w i t h imperatives as a means to inflect the verb. Another type o f error occurs when children use too many emphatics strung together. M 27.3

*ge your

bes teeth

-o NEUTRAL

dugumenigol pull:out:l:INTENT 2

*elengo likelthis

-e EMPH

-balo EMPH

'I'm going to pull your teeth out like this!' There are several errors i n this utterance. M e l i has used a subject pronoun (ge) instead o f a possessive adjective (gi). There is an error i n the deictic term concerning p r o x i m i t y o f action (=wengo). The verb is correct, but has t w o emphatics w h i c h cannot co-occur. For all the children, but especially for M e l i , the period between 27 and 30 months was a time w i t h a great deal o f overmarking. Overmarking occurred not only w i t h emphatics, but i n the casemarking system on nominals. The evidential particle -lob T see' was acquired early and i n some contexts used correctly, but it was often used as an emphatic and not w i t h its more specific evidential meaning. The particular discourse constraints o f the use o f the eviden­ tial were not always apparent i n child language. I t was used incorrectly as an emphatic co-occuring w i t h the 1st person present tense, and i n other instances i t was used incorrectly w i t h negatives and w i t h interrogatives. Meli 28.3

ne I

*mo -gelimeno -lob N E G carry: 1:FUT E V I D

* T obviously will not carry' Abi 28.2

nowo my: mother

*ane go:PAST

-lob EVID

-de Q

* T wonder if my mother obviously went?!'

5.

The A c q u i s i t i o n of Kaluli

587

Another type o f error making use o f this evidential as a more general emphatic was its occurrence w i t h the verb 'see', w h i c h is not correct. Meli 30.2 is with her cousin Mama, age 45 months. Mama:

en there

-a bebe -ke L O C see:PAST EMPH: PAST

T saw (it) on there' Meli:

ne I

bebe-lob -ke E V I D EMPH:PAST

* T saw (it) obviously' I f M e l i wanted to escalate the degree o f emphasis, she could have added an additional emphatic particle -ye to the already present emphatic particle (as is done by more mature speakers), but i t was not correct to add the evidential. Another frequent error was i n the use o f the quotative verb i n confrontational rhetorical questions. Rhetorical questions are frequently used to challenge an­ other's rights to objects or actions. Speakers (whether i n fact they have said anything or not) use utterances such as Dima elabeyo? ' D i d I say 'take'?!' (meaning T d i d n ' t say take so you can't take!'). The errors made by children were due to using this form as an emphatic w i t h the wrong pragmatic force. For example, Meli 27.3 was looking for a lost spoon. Mother:

ko there

-na sulufotinabe L O C fall:RECENT PAST

-ke EMPH:PAST

Tt fell there!' Meli:

eh? huh?

oba where

*sulufotinebi-yo fall:IMP:FUT Q

elabeyo 'did I say'

*'Huh? Where fall, did I say?' Since this is the first clarification request i n this sequence, there is no real need for M e l i to use any emphatics at all. However, she uses the formulaic expression w i t h the quotative verb as part o f her query, which is not correct. I n addition she has used the future imperative f o r m instead o f the past tense o f the verb. These types o f errors w i t h the use o f quotatives as emphatics continued through 29 months. A t 32 months they inappropriately re-emerged i n situations where M e l i was angry and she used them to mark her anger. This use was challenged by her classificatory mother w h o asked, " W h a t ' s w i t h all this ' d i d I say'? "as every­ one laughed at M e l i . This suggests that children overgeneralize to similar affec­ tive content much like they overgeneralize i n other areas o f language acquisition.

588

Schieffelin

The quotative verbs were also used correctly during this time period i n nonchallenging contexts o f reported speech. I n addition to emphatics and e v i d e n t i a l , there are other discourse based particles i n K a l u l i . For example, i f one speaker wishes to immediately repeat the words o f another to a third person (other-repetition) there are t w o ways to do so. The speaker can use a quotative construction w i t h the verb sama ' s a y ' — w h i c h is frequently done. Use o f the verb 'say' has no time constraints and all participants need not be present. A second option is less flexible. The pragmatic constraints on this f o r m , -do, are as follows. A l l speakers must be present and the repetition must be i n the next turn. A speaker cannot use this particle for self-repetition or repetition at a later time. The repetition particle, -do, is suffixed to the utterance; sometimes there are slight morphophonemic changes before the suffix is added. For example, Speaker A to Speaker B:

sagalima!

'put on:IMP'

Speaker C to Speaker B:

sagalindo!

'put on:IMP:REPET'

The pragmatic distinctions between self- and other-repetition markers were not made by young children, w h o used this other-repetition particle as an emphatic on their o w n repeated utterances. Errors such as the f o l l o w i n g occurred, Meli 26 to her mother, magu banana

die -hamana take:PURP go:IMP

'Go get a banana' (no response from mother) no mother

*die -hama-ndo! take:PURP go:IMP:REPET

'Mother, go get a banana!'

After 27 months this particle was no longer used, either correctly or incorrectly, until 30 months, when it was correctly, though infrequently, used. W e see similar error patterns i n the acquisition o f emphatic and discourse particles as we see i n other parts o f the linguistic system. Certain salient and highly frequent forms are used early and appropriately i n many contexts. H o w e v ­ er, there are errors caused by overgeneralization. I n K a l u l i , e v i d e n t i a l and other-repetition devices are used by young children as general emphatics, w i t h ­ out the specific functions that are found i n adult language use. B y 30 months, however, the majority o f these errors disappear.

5.

The A c q u i s i t i o n of Kaluli

589

CONCLUSIONS W h i l e many different aspects o f K a l u l i children's language acquisition have been discussed i n the body o f the chapter, a few concluding thoughts remain. They are of t w o types. First, support is offered for a number o f Slobin's Operating Princi­ ples (1985) based on the developmental patterns and processes presented above. Second, several questions based on the K a l u l i acquisition data w i l l be raised w i t h the hope that answers w i l l be forthcoming at a future date. K a l u l i children's early use o f variable w o r d order i n pragmatically appropriate ways provides evidence for t w o o f Slobin's (1985) Operating Principles. They are OP ( S T O R A G E ) : U N I T S and OP ( M A P P I N G ) : V A R I A B L E W O R D O R D E R . Furthermore, given the K a l u l i data, these t w o principles appear not only to interact w i t h one another, but to strengthen one another. Y o u n g children must not only store the different w o r d orders o f the language that is both directed to them and that they overhear i n their everyday experiences, but they must also " t a g " those w o r d orders w i t h information about the relevant context i n which they occur. The relevant context may include the linguistic environment (pho­ nological, morphological, etc.) or the social environment (speaker/addressee relationship, speech event, m o o d , etc.). One type o f context which determines word-order choice i n K a l u l i is the speech act that is being expressed. For K a l u l i children, this tagging process mentioned above could match different w o r d or­ ders w i t h particular speech acts. Thus K a l u l i children may draw on those con­ sistent relationships existing between w o r d order and speech act type, facilitating the acquisition process. The early acquisition o f pragmatically appropriate w o r d order may be only one o f several areas i n w h i c h contextual support is i n the form o f pragmatic function. Another concerns the acquisition o f personal pronouns. For example, the relatively error free acquisition o f focused and nonfocused pronouns may be further evidence o f the interaction o f the t w o Operating Principles (stated above) w i t h the net effect o f mutual reinforcement. The t w o sets o f personal pronouns are also tagged by specific speech act contexts and specific w o r d orders, and their acquisition may be guided through the support o f the pragmatic context. I n particular, i f one considers the errors i n pronoun acquisition that might have been made, but d i d not occur, we see the conceptual basis for certain aspects o f morphological development and the importance o f several kinds o f contextual support, including pragmatic support. Several sets o f errors i n different areas o f the linguistic system provides evidence for overextension, or Slobin's (1985) O P : E X T E N S I O N . I n terms o f casemarking on noun phrases, the young K a l u l i child tries to mark notions consistently even w i t h i n the component parts o f noun phrases. One consistent pattern o f K a l u l i child language errors involved the addition o f nominal casemarking on both the noun and its modifier (deictic, demonstrative, directional). This occurred during a limited time period after the initial emergence o f case-

590

Schieffelin

markers, especially w i t h the locative on noun phrases i n conjunction w i t h verbs of location and change o f location, w i t h the dative, and less frequently, w i t h the ergative casemarker. Overextension occurred i n other areas o f the linguistic system. For example, children overgeneralized from the declarative tense inflections across sentence types, which i n most cases was not correct. I n addition, when children encoun­ tered unfamiliar formulaic expressions that were phonologically similar to al­ ready productive forms, they used the incorrect, but similar form before produc­ ing the correct formulaic expression. The acquisition o f correct verb stems for different forms o f the verb provides strong evidence for another one o f Slobin's (1985) Operating Principles, OP ( U N I T S ) : W O R D F O R M S . K a l u l i children tried to change w o r d forms through the suffixation o f the most salient w o r d final phonological morpheme (the em­ phatic markers), producing incorrect forms. This can be seen as an example o f an interim production strategy, where the child tries to produce an appropriate sounding inflection by " s t e a l i n g " from another part o f the grammar. There are other examples o f the use o f this same strategy, also i n verb formation, where children use other less frequent but salient verb final particles as possible inflec­ tional forms i n constructing a verb-like w o r d . I n addition, i n acquiring the extremely complex system o f verb stem changes, K a l u l i children frequently picked one form as basic (the stem o f the present imperative) and attempted to add the tense endings to that form. Recall that i n K a l u l i there is a mismatch i n the temporal basis o f stem formation i n that the present declarative is based on the future imperative. F o l l o w i n g the Operating Principle that children prefer the use o f grammatical markers that make semantic sense (Slobin, 1973), i n their selection o f one form as basic, K a l u l i children incorrectly pick the present imperative as the base for present tense forms. Later, as they acquire additional verb forms, they incorrectly base future forms on the future imperative. I n languages w i t h extensive and irregular morphology one would expect to find these strategies used w e l l beyond the age o f three years. Evidence from the distribution o f verb forms used by K a l u l i children supports the view that some aspects o f the acquisition process may rely more heavily on function and context. I n fact, the acquisition o f some forms o f the verb may be seen as context specific and pragmatically based. Given that verb forms w h i c h differ i n tense/person and sentence type are morphologically marked, K a l u l i children may not initially be learning verbs i n general or verbs i n a paradigm, but verb forms i n particular. Thus we may want to be more specific when we discuss the acquisition o f verbs. I f the child does not use all o f the forms o f a given verb, we should investigate i n more detail the extent to w h i c h particular forms i n the verbal paradigm are used. Given a language that has rich and complex mor­ phology i n the verbal system, we should examine the distribution o f verbs according to person, tense and when relevant, sentence type. I n the acquisition o f verbs, the child may be learning pieces o f the paradigm, that is, specific forms

5.

The A c q u i s i t i o n of Kaluli

591

for specific functions before he or she is able to generalize to other forms i n the paradigm. Questions generated from the K a l u l i data suggest further areas o f investiga­ tion concerning not only the acquisition o f K a l u l i but concerning child language research i n general. For example, considering the early acquisition data, K a l u l i resembles those languages for w h i c h we already have a significant amount o f information. I t is the more complex verbal constructions (clause-chaining and switch-reference) that distinguish K a l u l i from other types o f languages. Once we look at more complex syntax, the differences between K a l u l i language acquisi­ tion and acquisition i n other languages w i l l probably appear greater. One ques­ tion, then, concerns how we locate the building blocks and bridges between relatively simple syntax and more complex constructions. Data focusing on these aspects o f development w i l l help us understand how the child comes to sound like a speaker o f his or her language, not just a speaker o f " C h i l d Language." I n all languages complementary linguistic resources present different and inconsistent information about the language system to the language acquiring child. I n K a l u l i , for example, verb agreement follows a nominative-accusative pattern while nominal morphology is split between a neutral system and an ergative-absolutive system The syntactic structure o f the language is not ergative, but part o f the morphological structure is. W e need to understand how these systems interact i n the development o f the child's linguistic hypotheses and the expressions reflecting those hypotheses. Another area i n which this type o f interaction is found is i n K a l u l i personal pronoun and nominal casemarking systems. A s we have seen, the acquisition o f pronouns for expressing agents is more consistent and precocious than the acquisition o f correct casemarking o f nouns as agents. Therefore, the acquisition o f expressions o f agency varies according to w h i c h part o f the linguistic system is examined. Does this mean that the nominal and pronominal systems do not interact? The K a l u l i data suggest that we should investigate how different linguistic resources are used to express the same notions i n the system. Another issue that needs be addressed i n future child language research is the role played by ellipsis. Languages w i t h extensive ellipsis, such as K a l u l i , are more difficult to analyze i n terms o f canonical clause forms, since the full forms are less common even i n the input than reduced clauses. However, the rules o f ellipsis seem to be one aspect o f the language acquired relatively early by young speakers. Finally, as this chapter and others i n this volume (for example Ochs, Clancy) have shown, researchers investigating child language acquisition need to attend to the importance o f affective and discourse components (whether mor­ phological, lexical or syntactic) i n the acquisition process. These components are important to consider for what they tell us i n contexts o f both appropriate and inappropriate use. Affective and discourse particles are linguistic resources that enable the child to more effectively communicate and accomplish some o f the

592

Schieffelin

interpersonal and expressive goals o f language. They must be considered along w i t h the phonological, morphological, syntactic and semantic aspects o f devel­ opment i n the crosslinguistic study o f language acquisition.

ACKNOWLEDGMENT I would like to thank institutions as well as individuals for the different kinds of support necessary in order to prepare this chapter. Funding for the fieldwork among the Kaluli was from the National Science Foundation (pre-doctoral research grant # G S 43769) and the Wenner-Gren Foundation for Anthropological Research, New York (grant-in-aid # 3054). A University of Pennsylvania Faculty Summer Grant (1982) facilitated the analysis and preparation of the chapter. Many non-Kaluli speaking friends generously provided detailed comments which improved all aspects of this work. In particular I want to thank Pat Clancy and Gillian Sankoff for their critical suggestions and extensive encouragement. Elinor Ochs and Dan Slobin kept certain fires going throughout the cooking process for which I am grateful. Cherie Francis assisted in the final preparation of the chapter. Of course I wish to thank my Kaluli-speaking friends for their assistance in this project—Buck Schieffelin, Steve Feld, and the Kaluli people of Bona village. And Zachary, thanks for your patience with it all.

REFERENCES Bechert, J . Ergativity in the constitution of grammatical relations. In F . Plank (Ed.), Ergativity. London: Academic Press, 1979. Bloom, L . , Lightbown, L . , & Hood, L . Structure and variation in child language. Monographs of the Society for Research in Child Development, 1975, serial no. 160, 40(2). Comrie, B. Ergativity. In W. P. Lehmann (Ed.), Syntactic typology. Austin: University of Texas Press, 1978. Dixon, R. M. W. Ergativity. Language, 1979, 55, 59-138. Feld, S. Sound and sentiment. Philadelphia: University of Pennsylvania Press, 1982. Feld, S., & Schieffelin, B. B . Hard words: A functional basis for Kaluli discourse. In D. Tannen (Ed.), Text and talk. Washington: Georgetown University Press, 1982. Givön, T. Focus and the scope of assertion: Some Bantu evidence. Studies in African Linguistics, 1975, 6, 185-205. Greenberg, J . H . Some universals of grammar with particular reference to the order of meaningful elements. In J . H . Greenberg (Ed.), Universals of language. Cambridge, MA: M.I.T. Press, 1966. Haiman, J . Hua: A Papuan language. In T. Shopen (Ed.), Languages and their status. Cambridge, MA: Winthrop Publishers, 1979. Hopper, P. J . , & Thompson, S. A. Transitivity in grammar and discourse. Language, 1980, 56, 251-299. Lehmann, W. P. A structural principle of language and its implications. Language, 1973, 49, 47¬ 66. Longacre, R. F . Hierarchy and universality of discourse constituents in New Guinea languages: Discussion. Washington, D. C : Georgetown University Press, 1972. L i , C . N., & Lang, R. The syntactic irrelevance of an ergative case in Enga and other Papuan

5.

T h e A c q u i s i t i o n of Kaluli

593

languages. In F . Plank (Ed.), Ergativity: Towards a theory of grammatical relations. London: Academic Press, 1979. Ochs, E . Ergativity and word order in Samoan child language. Language, 1982, 58, 646-671. Ochs, E . , & Schieffelin, B. B . Language acquisition and socialization: Three developmental stories and their implications. In R. LeVine & R. Shweder (Eds.), Culture theary: Essays on mind, self, and emotion. New York: Cambridge University Press, 1984. Olson, M. L . Switch reference in Barai. Proceedings of the Fourth Annual Meeting of the Berkeley Linguistic Society, 1978,4, 140-157. Rule, M. A statement of the alphabet and grammar of the Kaluli language of Bosavi, Papua. Unevangelized Field Mission, Lake Kutubu, Papua New Guinea. 1964, Unpublished. Schieffelin, B . B. Getting it together: An ethnographic study of the development of communicative competence. In E . Ochs & B. B. Schieffelin (Eds.), Developmental pragmatics. New York: Academic Press, 1979. Schieffelin, B. B . How Kaluli children learn what to say, what to do, and how to feel: An eth­ nographic study of the development of communicative competence. Unpublished doctoral disser­ tation, Columbia University, 1979. (Cambridge University Press, in press). Schieffelin, B. B. Teasing in Kaluli children's interactions. In E . Ochs & B. B. Schieffelin (Eds.), Language socialization across cultures. New York: Cambridge University Press. In press. Schieffelin, E . L . The sorrow of the lonely and the burning of the dancers. New York: St. Martins Press, 1976. Shaw, R. D. A tentative classification of the languages of the Mt. Bosavi region. In K . Franklin (Ed.), The linguistic situation in the Gulf district and adjacent areas, Papua New Guinea. Pacific Linguistics, no 26, series C , 1973, 187-224. Silverstein, M. Hierarchy of features and ergativity. In R. M. W. Dixon (Ed.), Grammatical categories in Austrialian languages. Canberra: Australian Institute of Aboriginal Studies, 1976. Slobin, D. I. Language change in childhood and history. In J. Macnamara (Ed.), Language learning and thought. New York: Academic Press, 1977. Slobin, D. I. Cognitive prerequisites for the development of grammar. In C . A. Ferguson & D. I. Slobin (Eds.), Studies of child language development. New York: Holt, Rinehart & Winston, 1973. Slobin, D. I. Crosslinguistic evidence for the Language-Making Capacity. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum Associ­ ates, 1985. Voorhoeve, C . L . Central and western Trans-New Guinea Phylum languages. In S. A. Wurm (Ed.), Papuan languages and the New Guinea linguistic scene. Pacific Linguistics, no 38, series C , 1975, 345-460. Wurm, S. A. Personal pronouns. In S. A. Wurm (Ed.), Papuan languages and the New Guinea linguistic scene. Pacific Linguistics, no. 38, Series C , 1975, 191-218. Wurm, S. A . , Laycock, D. C , & Voorhoeve, C . L . General Papuan characteristics. In S. A. Wurm (Ed.), Papuan languages and the New Guinea linguistic scene. Pacific Linguistics, no 38, Series C , 1975, 171-190.

The Acquisition of Polish

Magdalena Smoczyriska Jagiellonian Krakow,

University Poland

Contents Introduction 595 Descriptive Sketch of Polish 595 Sources of Evidence 612 Overall Course of Linguistic Development 617 The Data 620 Typical Errors 620 U-Shaped Developments 642 Timing of Acquisition 644 Conclusions 667 Universal Operating Principles and the Acquisition of Polish 667 Language Acquisition Strategies: Universal or Language Specific 677

INTRODUCTION 1. D e s c r i p t i v e S k e t c h o f P o l i s h /. /.

General

Typological

Characteristics

Polish is an Indo-European language, belonging to the Slavic branch.

Within

the Slavic branch, Polish represents one o f the N o r t h Slavic languages, as differ­ ent from South Slavic (Serbo-Croatian, Bulgarian, etc.). The N o r t h Slavic group *The present chapter includes some sections written by Dr. Richard M. Weist, who was asked to give a first-hand account of his research conducted at Adam Mickiewicz University and SUNY College, Fredonia. These sections are indicated by [WEIST] at the beginning and end of each section. Due to the communication problems caused by the situation in Poland during the completion of work on this book, Dr. Weist undertook the job of developing and editing the first draft of my chapter as a favor to me. He has contributed enormously to the final formulation of my text, and I would like to express my deepest gratitude for his cooperation. I would also like to thank my colleague and friend, Dr. Elzbieta Tabakowska from Jagiellonian University, Krakow, who helped me with the English for­ mulation of the original draft. 595

596

Smoczynska

is further subdivided into West Slavic languages including Polish, Czech, Slo­ vak, etc., and East Slavic (Russian, Ukrainian, etc.). L i k e other Slavic languages, Polish represents an inflecting, or fusional type of language, i n which single grammatical morphemes combine several functions: case, gender, and number i n noun forms; person and number i n verb forms. However, Polish is less close to the ideal model o f an inflecting language than Czech or Slovak. L o t k o (1979) gives many examples o f replacing synthetic means o f expression by analytical ones i n contemporary Polish. I n general, the Polish grammatical system closely parallels that o f Russian. Because Polish retains a rich inflectional system, morphology is the main device for expressing syntactic distinctions. W o r d order has grammatical func­ tions only to a limited extent, and hence i t can serve other purposes, namely, i t performs pragmatic functions. Deviations from the standard (unmarked) S V O order serve the purposes o f topicalization; and combined w i t h focal stress, w o r d order expresses special emphasis on certain elements, i n a way independent from discourse structure. Furthermore, the rich inflectional system makes i t possible to apply ellipsis to a much larger extent than i n English. A selective outline o f Polish morphology and syntax is provided i n Sections 1.2 and 1.3. The interested reader w i l l profit from Schenker's (1967, 1973) comprehensive and w e l l organized description o f Polish grammar and Fisiak, Lipihska-Grzegorek, and Zabrocki's (1978) Polish-English contrastive grammar. 1.2.

Morphology

There are some relevant differences between Slavic and English morphology. I n English, the basic w o r d form is the stem. The same stem can constitute the basic form o f both noun and verb, e.g. change, drink. The form STEM + 0 usually performs numerous functions, and the use o f inflectional endings is greatly limited. The form STEM + ENDING occurs i n opposition to the basic f o r m , e.g. balls!ball, Mommy-'si Mommy, drinks/drink, want-ed/want, etc. Thus, i n English one can talk about the acquisition o f particular grammatical morphemes, as is actually done i n B r o w n (1973), de Villiers and de Villiers (1973), and others. I n Slavic languages, bare stems, i.e. forms w i t h a zero ending, are rare, and i n most cases, the w o r d stem can never occur alone. Hence, the concept o f a w o r d stem is a mere abstraction w h i c h can be formed by cutting o f f the common part that occurs i n various forms o f a w o r d from the array o f possible endings. Even those forms w h i c h are functionally unmarked, e.g. nominative singular o f nouns or infinitive o f verbs, have specific endings, and a given ending explicitly spec­ ifies the grammatical category o f the w o r d . Moreover, i n view o f the rich variety of inflectional forms o f a w o r d , a given form occurs i n opposition to the whole inflectional pattern, rather than to the basic form only. Therefore, morphological development should be conceived o f as the acquisition o f the ability to REPLACE

6.

The A c q u i s i t i o n of Polish

597

grammatical morphemes according to the rules o f language, rather than the ability to ADD them to basic forms when required. I n consequence, although Polish children can very easily split a w o r d form into a stem and an ending (as illustrated by their o w n morphological formations, differing from adult forms), the basic unit o f the text is a WORD rather than a morpheme. I n Polish this is additionally strengthened by fixed penultimate w o r d stress. I n contrast to Polish, w o r d stress is mobile i n Russian and Serbo-Croatian. Free morphemes are not numerous i n Polish, and their status is entirely different from that o f bound morphemes. Furthermore, most Polish endings are syllabic, whereas i n English many o f them are nonsyllabic. Thus their presence does not affect the prosodic shape o f the w o r d (equal to stem). The purpose o f this section on morphology is to provide the reader w i t h a sketch o f the declensional and conjugational patterns o f Polish. Hence, the description which follows has been simplified i n many ways. I n some areas, the nature o f the simplification w i l l be identified so that the reader can appreciate the complexity o f the overall system. However, i n many instances the relationship between this brief sketch and the full system w i l l not be identified. The reader seeking an understanding o f the full system should consult another source, such as Schenker (1973). 1.2.1. Nouns. Nouns are marked for gender, and they are inflected for case and number. The relationship between the cases and their most typical semantic and syntactic functions is presented i n Table 6 . 1 , and the basic regular declen­ sional patterns are outlined i n Table 6.2. Gender serves as the basis for the classification o f nouns into declensional patterns. Singular nouns have one o f three genders: feminine, masculine, or neuter. The masculine and neuter patterns are relatively similar as contrasted w i t h the feminine pattern. I n the plural, there are t w o genders: virile ( V I R ) and nonvirile ( N V I R ) , where virile includes the TABLE 6.1 The Polish Cases a n d Their M o s t Typical Functions [WEIST]

Case Nominative (NOM) Accusative (ACC) Genitive (GEN) Dative (DAT) Locative (LOC)" Instrumental (INSTR) Vocative (VOC) a

Semantic Function

Syntactic Function

Subject Agent Direct Object Patient Possessive Possessor Indirect Object Receiver Adverbial Location Adverbial Instrument/Comitative The case used when addressing someone

The locative case is also referred to as the prepositional case.

Key Wh Question Who Whom Whose To whom On what/where With what/whom

598

Smoczyñska TABLE 6.2 The Basic Declensional Patterns Singular (SG)

Case NOM ACC GEN DAT LOC INSTR

Feminine (FEM)

Masculine (MASC)

Neuter (NEUT)

-a -e -y/-i [SFC] -e/-i [SFC] = DAT - a

-0 [AN] = GEN/[IN AN] = NOM -a/(-u) -owi/(-u) -e/-u [SFC] -em

-o/-e = NOM -a -u -e/-u [SFC] -em

Plural (PL) with Virile (VIR)INonvirile (NVIR) NOM ACC GEN DAT LOC INSTR

-y/-i, (-e) [SFC], -a [NEUT], -owie [VIR] [NVIR] = NOM/[VIR] = G E N -0 [—MASC] -6w/(-y/-i) [+MASC] -om -ach -ami (-mi)

[ ] = criterion, e.g. animate [AN]. SFC = stem final consonant (hard/soft or other). ( ) = less frequent suffix.

features masculine and h u m a n . I n the plural, the dative, locative, and instrumental suffixes are identical, regardless o f gender. The nominative and accusative suffixes are conditioned by the virile versus nonvirile distinction and the genitive endings depend on the masculine versus non-masculine distinction. Throughout the declensional system specific suffixes are sometimes controlled by the morphophonological properties o f the noun stem, e.g. hard versus soft stem-final consonants. Furthermore, it is important to note that the speaker's task consists not only i n the selection o f an appropriate ending but also i n performing necessary modifications o f the phonological form o f the stem. Some o f such stem alternations result from general phonological rules. I n most instances, however, they are governed by narrow morphophonological rules only. 1

1.2.1.1. [WEIST] Cases and Functions. One o f the basic problems that the child (or any other listener) faces is to extract information about semantic functions, e.g. agent versus patient, from the surface structure. I f a simple Polish seems to be the most sex-biased of all Indo-European languages. Russian has the animate/inanimate distinction in plural for all nouns and Serbo-Croatian preserves the masculine, feminine, and neuter distinction.

6.

The A c q u i s i t i o n of Polish

599

transitive sentence contains t w o feminine singular nouns, there is no ambiguity within the basic declensional pattern as shown i n sentence l . I n sentence 1, 2

(1) Dziewczynk-a girl-FEM:NOM:SG

pcha matk-e pushes mother-FEM:ACC:SG

'(A) girl is pushing (the) mother.' (2) Dziewczynk-i girl-NVIR:NOM/ACC:PL

pchajq push

'Girls are pushing mothers' (3) Chlopiec-0 boy-MASC:NOM:SG

or goni chases

matk-i mother-NVIR:ACC/NOM:PL 'Mothers are pushing girls.' ojc-a father-MASC:ACC:SG

'(A) boy is chasing (the) father.' (4) Samochdd-0 car-MASC:NOM/ACC:SG

goni chases

'(A) car is chasing (the) bus'

autobus-0 bus-MASC:ACC/NOM:SG

or

'(A) bus is chasing (the) car.'

the w o r d dziewczynk-a ' g i r l ' has the nominative singular suffix -a which i n d i ­ cates (in this context) that the g i r l is the one who is initiating the action, and the word matk% 'mother' has the accusative singular suffix -£ which specifies that the mother is somehow affected by the action. The meaning is the same i f we change the word order: Matk% pcha dziewczynka, Pcha dziewczynka matk$, etc. A m b i ­ guity can arise i f these nouns are plural nouns since the nominative suffix is the same as the accusative for nonvirile nouns, as demonstrated by sentence 2. I f the two nouns i n a transitive sentence are masculine singular and animate, the nominative and accusative suffixes are distinct, e.g. chtopiec and ojciec i n the nominative case and chiopca and ojca i n the accusative case. However, i f the masculine nouns are NOT animate the suffixes i n the nominative and the ac­ cusative case are identical as shown i n sentence 4. (5) Siostr-a sister-FEM: NOM: SG

Michal-a Michael-MASC:GEN:SG

ciast-o cake-NEUT:ACC:SG

pod under

zjadla ate

stol-em table-MASC:INSTR:SG

w in

kuchn-i kitchen-FEM:LOC:SG 'Michael's sister ate (the) cake under (the) table in (the) kitchen' (6) Chlopc-y boy-VIR:NOM:PL 2

zjedli ate

ciast-a cake-NEUT:GEN:SG

For a guide to pronunciation, see Appendix A.

z with

600

Smoczyriska

rodzynk-ami. raisin-NVIR:INSTR:PL 'The boys ate some cake with raisins' (7) Michal-0 Michael-MASC:NOM:SG

nie not

jadl ate

ciast-a cake-NEUT:GEN:SG

palc-ami. finger-NVIR:INSTR:PL 'Michael did not eat (the) cake with (his) fingers' (8) Chlopiec-0 boy-MASC:NOM:SG

poszedl went

do to

sklep-u shop-MASC:GEN:SG

z with

siostr-q. sister-FEM:INSTR:SG '(The) boy went to the shop with (his) sister' (9) Dziewczynk-a girl-FEM:NOM:SG

czytaia read

brat-u brother-MASC:DAT:SG

o about

zwierz^t-ach. animal-NVIR:LOC:PL '(The) girl read to (her) brother about animals.' The above examples serve to demonstrate that the functions o f cases are not limited to those presented i n Table 6 . 1 . The genitive, for instance, is not only the Possessor's case (as i n sentence 5) but also expresses the Partitive (sentence 6 ) , the Direct Object o f negated sentences (sentence 7 ) , the Location i n the ex­ pression 'to the shop' (sentence 8), etc. Similarly, the instrumental is not only the expression o f Instrument (as i n sentence 7: ' w i t h fingers') but also a C o m itative: ' w i t h raisins' and ' w i t h sister' (sentences 6 and 8 ) , as w e l l as the Loca­ tion 'under the table' (sentence 5). The locative not only expresses Location (as in sentence 5) but also figures i n a prepositional phrase 'about animals' w h i c h is not a Location (sentence 9 ) . As for the locative case, w h i c h is also referred to as the prepositional case, i t is w o r t h noting that both its names are misleading: i n fact, other cases serve to express location as w e l l , and other cases are used w i t h prepositions. For instance, while na ' o n ' and w ' i n ' require the locative to express static location, when used w i t h the accusative they express dynamic notions o f ' o n t o ' , ' i n t o ' . Prepositions do ' t o ' , od ' f r o m ' require the genitive, while pod 'under', nad 'over', za 'behind' govern the instrumental case. The same prepositions, however, when used w i t h the accusative express motion: 'to go under/over/behind something'. A m o n g Polish cases, it is the genitive w h i c h has the widest range o f functions, while the use o f the dative is relatively the most l i m i t e d , i.e. Receiver, Beneficiary, Experiencer. I n addition to these points concerning the relationships between the cases and semantic and syntactic func-

6.

The A c q u i s i t i o n of Polish

601

tions, sentences 1-9 show clearly the sense i n w h i c h Polish is an inflecting language. Each suffix combines the concepts o f gender, case, and number. Furthermore, the individual suffixes can only be understood w i t h i n the context o f a declensional pattern. Hence, the suffix -a [a] has the nominative singular function i n the feminine pattern, the genitive (and/or) the accusative function i n the masculine pattern, and the accusative and nominative plural function i n the neuter pattern. [WEIST] 1.2.2. Adjectives. Adjectives agree i n case, number, and gender w i t h the noun they modify. The declension o f adjectives is relatively regular: there are five different patterns corresponding to the five genders (three i n singular, t w o i n plural). Comparative and superlative degrees are formed also rather regularly by adding special suffixes and prefixes. W i t h some classes o f adjectives only analyt­ ical forms can be used. 1.2.3. Adverbs. Adverbs are primarily derived from adjectives by means o f suffixes. They are not declined. Comparative and superlative degrees are formed according to rules similar to those for adjectives. 1.2.4. Numerals. W h i l e ordinals are declined according to the regular ad­ jective pattern, the declension o f cardinal numerals is a loose collection o f exceptions rather than a system. Every honest native speaker o f Polish w o u l d confess making occasional mistakes w i t h these forms or at least admit that the choice o f the correct f o r m requires an intensive intellectual effort on his or her part. 1.2.5. Pronouns. M o s t pronouns are declined according to the regular ad­ jectival pattern, but the case forms o f personal pronouns are highly idiosyncratic. Furthermore, oblique cases o f personal pronouns have special enclitic short forms, distributed according to special rules. Another peculiarity is the existence o f special reflexive pronouns, w h i c h replace the personal or possessive pronoun when the person to w h i c h it refers coincides w i t h the sentence subject. For instance, the English sentences / think about myself, You think about yourself, They think about themselves are trans­ lated into Polish as Myslq o sobie, Myslisz o sobie, Myslq o sobie, where a single reflexive pronoun, sobie, is used w i t h no distinction for number or gender. Similarly, the f o l l o w i n g set o f sentences reveals the contrasts for possessives: On bierze moje ksiqzki ' H e takes m y : l S G books', but Bior$ swoje ksiqzki T.take my.REFL books'; and compare Bior% jego ksiqzki 'Ltake his:3SG:MASC books' w i t h On bierze swoje ksiqzki ' H e takes his:REFL books'. The use o f the third person singular pronoun i n the last example (On bierze jego ksiqzki ' H e takes his books') w o u l d specify that the owner o f the books and the agent are t w o different persons.

602

Smoczyhska

I n general, possessive pronouns are used less frequently than i n English. For instance, i n the above examples the possessive reflexive is most frequently omitted and such terms as body parts or kinship terms are usually used without possessive pronouns, unless the "possessor" happens not to be identical w i t h the agent. Thus, a literal translation o f the English, / wash my hands, w o u l d seem queer i n Polish, unless the speaker wanted to stress that he is not washing somebody else's hands. Furtheremore, many English constructions w i t h a pos­ sessive are rendered i n Polish by means o f a construction w i t h the benefactive (in dative): / wash his hands, translates as Myjq mu rqce T:wash to:him hands'. 1.2.6. Verbs. Finite and nonfinite forms o f the verb are marked for aspect. Finite forms are inflected for tense, mood, person, number, and i n some i n ­ stances gender. 1.2.6.1. Aspect. The major aspectual distinction i n Polish is between per­ fective and imperfective, and Polish also has an iterative aspect. These distinc­ tions are clearly marked i n the morphology (see Majewicz, 1982). W h i l e the meaning o f perfectivity is still debated (e.g. Ferrell, 1951; Ridjanovie, 1976), we w i l l take the position that perfective verbs specify the notion o f a completed situation, i.e. a situation w h i c h has a beginning, a continuation, and a termina­ tion (see Comrie, 1976). The imperfective aspect is neutral and the imperfective verb form does not specify whether the action is completed or not. The selection of aspectual form depends on perspective. Given that a situation is i n fact completed, the speaker may refer to that situation w i t h a perfective or imperfec­ tive verb form. The perfective form is chosen i f the speaker intends to identify the property o f completion i n the situation. However, i f reference is made to a situation w h i c h is not completed, only an imperfective verb can be used. For most verbs, the imperfective form is morphologically simpler, and the perfective form is derived from i t . This is done by means o f adding either a prefix, e.g. pisaclna-pisac 'to w r i t e ' ( I P F V / P F V ) or a suffix, e.g. kop-a-clkopnq-c 'to k i c k ' ( I P F V / P F V ) . For some aspectual pairs, both terms do not differ i n morphological complexity. Each o f them includes a stem and a suffix, and the aspectual opposition is indicated by the k i n d o f suffix used, e.g. strzel-i-cl strzela-c 'to shoot' ( P F V / I P F V ) . For many verbs o f this group the suffix alternation is accompanied by the stem alternation as w e l l , e.g. otworz-y-clotwier-a-c 'to open' ( P F V / I P F V ) . Suffix alternation usually implies the shift o f the verb to another conjugation type. For a dozen other verbs, the basic form is perfective and the imperfective form is derived from i t , e.g. dacldawac 'to g i v e ' ( P F V / I P F V ) ; wstaclwstawac 'to get u p " ( P F V / I P F V ) . As has already been said, for most verbs perfective forms are derived from the imperfective and the number o f verbs i n w h i c h the derivation goes the other way is rather limited. There is, however, another (very frequent) process w h i c h consists o f deriving imperfectives from prefixed perfective forms. I n the example

The A c q u i s i t i o n of Polish

6.

603

cited above, pisac I na-pisac 'to w r i t e ' ( I P F V / P F V ) , the prefix na- is added to the simple imperfective f o r m to make a perfective f o r m . This particular prefix does not carry any particular meaning except for that o f perfectivity. There are, however, other prefixes w h i c h can be added to the imperfective pisac and w h i c h serve t w o functions at one time. They perfectivize the verb and carry additional meanings roughly corresponding to those expressed by verbal particles i n En­ glish, e.g., (1) pod-pisac 'to sign' (PFV) pod- 'under', (2) wy-pisac 'to write out' (PFV) wy- ' o u t ' , (3) prze-pisac 'to copy' (PFV) prze- 'through', (4) przypisac 'to ascribe' (PFV) przy- 'at', (5) w-pisac 'to write i n ' (PFV) w- ' i n ' , (6) do-pisac 'to add t o ' (PFV) do- ' t o ' , and (7) od-pisac 'to c r i b ' (PFV) od- ' f r o m ' . Since all these verbs are perfective and their meanings differ from that o f the imperfective simple verb pisac, their aspectual counterparts are derived from them by means o f suffixation, e.g. pod-pis-ywa-c, wy-pis-ywa-c, etc. Such forms are called secondary imperfectives. Finally, aspectual pairs can be sup¬ pletive, each o f the terms based on a different stem, e.g., brae (IPFW)/wziqc (PFV) 'to take,' widziec (IPFV)Izobaczyc (PFV) 'to see'. 1.2.6.2. Finite Forms. The conjugation o f the Polish verb is illustrated i n Table 6.3, where all the finite forms o f the verb pisac I napisac 'to w r i t e ' are given. There are three moods i n Polish: indicative, conditional, and imperative. The tense system is very simple and i t involves three tenses only: present, past, and future. The tense distinction appears i n the indicative mood o n l y , and its full realization concerns imperfective verbs. Perfective verbs lack the forms o f pre­ sent tense, since the notion o f perfectivity precludes the simultaneity o f action w i t h the moment o f speaking; thus they have only past and future forms. For perfective verbs, forms corresponding to those o f present (in imperfective verbs) are used to express futurity, w h i l e for imperfective verbs a separate analytic construction is used. I t involves the auxiliary b%d% ' T w i l l ' followed either by the "past p a r t i c i p l e " (see below): bqdq pisal ( M A S C ) , bed$ pisala ( F E M ) , bgdzie pisalo ( N E U T ) , bqdziemy pisali ( V I R ) , bqdziemy pisaty ( N V I R ) or by the i n f i n i ­ tive: b%d% pisac, bqdziemy pisac, w i t h no distinction o f gender. The choice o f the form (past participle vs. infinitive) is o p t i o n a l . 3

It should be noted that each verb has at least t w o stems. Stem I , called the "present tense s t e m " (here: pisz-) appears i n present forms (both " t r u e " present of imperfective verbs, and " f o r m a l " present, i n fact, future o f perfective verbs) and i n imperative forms. Stem I I , called, the " i n f i n i t i v e s t e m , " occurs i n past tense indicative forms (including the "past p a r t i c i p l e " o f the analytic future) and in conditional forms. What is more, many verbs require additional stem alterna­ tions WITHIN a given inflectional paradigm, e.g. bior-q T-take' but bierz-esz I n Polish, the analytic future is ungrammatical with perfective verbs, while in Serbo-Croatian it is a regular form. 3

TABLE 6.3 Finite Forms o f t h e Verb pisaclnapisac 'to write' Aspect Imperfective (IPFV) Indicative Present SG

PL

Perfective (PFV)

1. pisz-e 2. -esz -e 3. 1. pisz-emy 2. -ecie 3. -Q

Future

(0) SG

1. b%d -£

pisa-1-}a>

2. bqdzi-esz or 3. -e pisa-c PL

I) o

1. bedzi-emy pisa- j ! *' f . or 2. -ecie . . , pisa-c 3. b$d -q (

y

)

MASC FEM NEUT

1. na-pisz-q 2. na-pisz-esz 3. -e

VIR NVIR

1. na-pisz-emy 2. 3.

-ecie -Q

Past SG

PL

MASC FEM NEUT

1. pisa-1-la> + m 2.

+ s

3.

+0

1. pisa- j 2. 3.

J + sray

MASC FEM NEUT

1. na-pisa-i- -l a \ + m \o\ 2. + s 3. +0

VIR NVIR

VIR 2.

+ sde +0

3.

+ 5Ci>

+0

Conditional SG 1. pisa-1- J I + &y-m 2. 3. PL

1. pi'jfl-

PL

604

1. 2. 3. 1. 2. 3.

+ by-smy -scie

— pÌSZ-0

— pisz-my -eie

—

1. na-pisa-1- j a j- + ¿?y-m 2. 3.

-i

2.

Imperative SG

MASC FEM NEUT

VIR NVIR

-i -0

1. na-pisa-

smy

2. 3.

scie

1. 2. 3. 1.

— na-pisz-0 — na-pisz-my 2. -eie 3. —

6.

The A c q u i s i t i o n of Polish

605

'you-take', w i t h the alternation lb' or I : Ib'ezl w i t h i n Stem I , as opposed to Stem I I : /bra/ i n bra-c ' t a k e - I N F ' (cf. stem alternation i n the auxiliary be.d$lb%dziesz in Table 6.3). Stem selection is the major difficulty i n Polish conjugation. I n respect to stem alternations, Tokarski (1973) distinguishes among 11 groups o f verbs, some o f them further subdivided. For many o f those " r e g u l a r " groups it is not sufficient to k n o w the infinitive and the present tense form i n order to be able to predict all the alternations involved. Moreover, a number o f verbs are quite irregular. The forms w h i c h are especially interesting are those o f the past tense and o f the conditional mood. Their peculiarity consists o f their agglutinative structure (Tokarski, 1973), quite atypical for Polish. I n these forms, each category is expressed by a separate morpheme. The rule o f past tense formation can be phrased as follows: [INFINITIVE STEM + PAST + GENDER] + PERSON e.g. pisawrite

-IPAST

-aFEM

-s 2SG

The conditional is derived from the past tense form by inserting the special article by, w h i c h results i n : [INFINITIVE STEM + PAST + G E N D E R ] + [CONDITIONAL + PERSON] e.g. pisawrite

-IPAST

-0MASC

-byCOND

-m 1SG

Yet another peculiarity lies i n the fact that these forms are not fully synthetic: they are composed o f the "past p a r t i c i p l e " (PP): [ I N F I N I T I V E S T E M + P A S T + G E N D E R ] and the movable ending: PERSON i n past tense forms, and [ C O N D I T I O N A L + PERSON] i n conditional forms. The past participle forms a prosodic entity: i t is regularly stressed on the penultimate syllable, whether followed by the ending or not, and i t is an autonomous w o r d (e.g. when used w i t h the auxiliary bqdzie to form the analytic f u t u r e ) . Past participles: pisa} ( M A S C ) , pisala ( F E M ) , pisalo ( N E U T ) , pisali ( V I R ) , and pisafy ( N V I R ) coin­ cide w i t h the third person past forms, since these forms have a zero marker for PERSON. O n the other hand, past and conditional forms are not analytic either, since the movable endings are not words per se. They can either follow the participle, without altering its stress, or they are moved towards the beginning o f the clause, usually to follow the first stressed unit, e.g. Co ty napisaia-s? or Co-s ty napisala? 'What have you written?' 4

Even if the ending follows the verb, its presence does not affect the regular penultimate stress pattern of the past participle: pisdli-smy, pisal-bym, pisdia-bym, etc. 4

606

Smoczyhska

The endings which are most frequently moved to the front o f the clause are the syllabic ones, w h i c h are enclitic. These include the first and second plural o f the past tense -smy, -scie, and all the conditional endings: -by-m, -by-$, -by for singular and -by-smy, -by-scie, -by for plural. I n spoken language, Dlaczego-smy to napisali? is more natural than Dlaczego to napisali-smy? ' W h y d i d we write that?' and J a by-m to napisaia is much more frequent than J a to napisala-bym T would write that'. A l l these peculiarities result from the fact that the Polish past tense historically comes from an analytic construction (pisai jesm), and the present state is that o f a transitional f o r m , neither fully analytic nor fully synthetic. I n Serbo-Croatian the corresponding forms preserve their analytic status, e.g. ja sam pisao, ja bih pisao, w i t h regular auxiliaries. I n Russian the auxiliary disappeared completely, and the person distinction is expressed by the pronoun only, e.g. ja pisal, ty pisal, etc. I t is w o r t h noting that Polish marks person i n the endings o f all the finite forms, and this is w h y personal pronouns are not used unless special emphasis is required. English / write ox I am writing translates simply as Piszq, while Japiszq w o u l d mean Tt is I w h o am w r i t i n g ' or ' A s for me, I am w r i t i n g ' . Synthetic imperative forms exist for second singular and first and second plural, while the third person forms, as w e l l as that o f first singular, are formed periphrastically, e.g. niech on pisze 'let h i m w r i t e ' . I t should be noted that the imperative form is not considered impolite among familiar speakers, and it is used much more frequently than i n English. It has already been noted that the main difficulty o f Polish verb inflection consists i n stem selection. The choice o f endings does not present major prob­ lems. Linguists distinguish three conjugation types, based on the present tense vowel: -e, -il-y, or -a. Some linguists treat it as stem final v o w e l , while other consider it as a part o f ending, as we d i d i n Table 6.3. Except for the v o w e l , the sets o f endings used i n the three conjugation types are very similar. The excep­ tion is the first singular ending o f the present tense, where t w o alternative endings occur: or -am (pisz-g T - w r i t e ' vs. czyt-am T-read'). (see Table 6.7.) 1.2.6.3. Nonfinite Forms. Indeclinable forms comprise infinitives (e.g. pisac or napisac 'to w r i t e ' ) and t w o gerunds. The present gerund, e.g. piszqc (only I P F V ) is equivalent to the adverbial clause ' w h i l e w r i t i n g ' ; and the past gerund, e.g. napisawszy has the meaning 'having written' (only P F V ) . The past gerund is almost never used i n spoken language. Apart from these, there are verbal nouns e.g. pisanie or napisanie '(the action of) w r i t i n g ' (both P F V and I P F V ) and t w o adjectival participles. They are the active participle, e.g. piszqcy ' w r i t i n g (person)' (only I P F V ) and the passive participle, e.g. pisany or napisany 'written' (both I P F V and P F V ) . Verbal nouns and adjectival participles are regularly declined according to respective noun and adjective patterns. What has been referred to as "past p a r t i c i p l e " i n Section 1.2.6.2 is only a part o f some finite verbal forms and does not occur independently.

6.

The A c q u i s i t i o n of Polish

607

Passive participles are used i n passive voice constructions. I n spoken lan­ guage, however, these have a limited range o f use, as w o r d order flexibility is sufficient for expressing focus, and a number o f subjectless active constructions can be used for subject downgrading. Those occurring i n spoken language typ­ ically involve perfective participles, used i n order to refer to a resulting end state of some action, performed by an unspecified agent which is either unknown or evident or simply irrelevant. These cannot be called truncated passives, as full passives are practically never used. They are rather a subclass o f attributive sentences making statements about objects. 1.3.

Syntax

1.3.1. Simple Sentences. M o s t o f the basic constructions correspond to those i n other Indo-European languages. Thus I w i l l point only to some idiosyn­ cratic structures. There are a number o f subjectless constructions w i t h either special impersonal forms used, such as the reflexive verb form pisze sie. or the impersonal participle pisano 'one:wrote', or the third plural form without a subject, piszg 'they:write' ( = 'somebody writes'). Another specific structure is the existential sentence, i n which the verb bye 'to be' is used i n the sense o f 'to exist, to be present', e.g. Jest chlopiec-0 'is boyN O M ' ( = 'There is a b o y ' or 'The boy is here'), w i t h the negated version Nie ma chhpc-a 'not have b o y - G E N ' ( = 'The boy isn't here'). Specific structural patterns are those i n v o l v i n g an experiencer i n dative, e.g. Jest mi zimno 'is m e . D A T c o l d ' ( = T am c o l d ' ) , Jest mi smutno 'is m e : D A T sadly' ( = T am s a d ' ) or Leci mi krew 'is falling m e : D A T blood' ( = T am bleeding') (cf. Berman, 1985, for a description o f identical constructions i n Hebrew). The dative experiencer can also be used instead o f the nominative subject (i.e. agent) i n order to express the lack o f a definite agentive intention. Compare Sthkla mi si% filizanka 'the cup broke itself to m e ' , uttered by an involuntary agent, w i t h Sthklem filizankq T broke cup' on the one hand, and Filizanka si% sthkla 'cup itself b r o k e ' , where no agent is mentioned, on the other. 5

1.3.2. Negation. The negation system is simple. U n l i k e Russian, where anaphoric and sentential negation have t w o distinct forms (nyet and ni), i n Polish there is only one negative particle, nie. I n negated sentences nie immediately precedes the verb or the auxiliary, and i n constituent negation i t immediately precedes the constituent i n question, e.g. Czy Jan idzie do szkofy?-NIE. 'Does John go to s c h o o l ? ' - ' N o . ' , Jan NIE idzie do szkoty 'John not go to school' ( = 'John doesn't go to school'), Jan idzie NIE do szkofy 'John goes not to school (but . . . ) ' , and NIE Jan idzie do szkofy ' ( I t is) not John (who) goes to school (but . . . ) ' . 5

The copula jest can be deleted, which results in Zimno mi, Smutno mi, etc.

608

Smoczyhska

A peculiarity o f Slavic languages is that o f converting the object o f a negated verb into the genitive f o r m , e.g., Czytam ksiqzk-% 'read b o o k - A C C ( = T am reading the b o o k ' ) but Nie czytam ksiqzk-i ' N o t read b o o k - G E N ' ( = T am not reading the b o o k ' ) . The same transformation is applied to subjects o f negated existential sentences. Unlike English, Slavic languages display multiple negation, e.g. Jan nigdy z nikim nie idzie 'John never w i t h nobody not goes' ( = 'John never goes w i t h anybody'). 1.3.3. Questions. Question formation is also simple. Yes-no questions re­ quire either rising intonation alone or preposing the particle czy 'whether', e.g. Piszesz?ICzy piszesz? ' D o y o u w r i t e ? ' . Wh-questions are formed by preposing a given w h - w o r d or a prepositional phrase w i t h the w h - w o r d (the " s t r a n d i n g " o f prepositions is not allowed). N o additional transformations are needed for questions. 1.3.4. Modals. Polish verbs whose semantic content corresponds to that o f English modals should be treated as full verbs rather than auxiliaries i n view o f their syntactic properties (cf. Fisiak et a l . , 1978, for details). Therefore sentences w i t h modals are embedded constructions rather than simple sentences w i t h a complex predicate, e.g. Musz-% napisa-c list ' M u s t - I S G w r i t e - I N F letter' ( = T must write a letter') or even Mog-% musie-c wyjsc ' c a n - l S G must-INF leaveI N F ' ( = T could be obliged to leave'). I n these examples, the past or future tense forms o f the " m o d a l " could be used. Moreover, there are a number o f imperson­ al modals used i n subjectless constructions, e.g. mozna 'oneican', wolno 'one: is: alio wed: t o ' , trzeba 'it:is:necessary:to', etc. 1.3.5. Complex Sentences. A detailed analysis o f syntactic differences be­ tween Polish and English complex sentences is given i n Fisiak et al. (1978). I n the present discussion, I w i l l point only to some o f the most general charac­ teristics o f Polish syntax. The major difference is that Polish constructions are more transparent. I t w o u l d be premature to say that surface realizations o f claus­ es are closer to the deep structure than their English counterparts, since we still do not k n o w what the deep structure i n fact is like for adult speakers, much less for children. But we can say that Polish complex sentences are easy to process, since the surface form o f constituent clauses is much like that o f simple sen­ tences, and the abundance o f explicit local cues facilitates the processing o f internal relationships holding between clause constituents, w i t h clauses being clearly delineated by explicit boundary markers. This "transparency" o f Polish complex sentences is manifested i n various aspects, w h i c h are discussed below. 1.3.5.1. Clear Marking of Clause Boundaries. Except for some coordinate clauses which can be simply juxtaposed without any connective, complex sen­ tences must be clearly divided into separate clauses, not only by pauses (or

6.

The Acquisition of Polish

609

commas i n writing) but also by overt markers. This feature is a natural conse­ quence o f the freedom o f w o r d order. Constituents can be permuted w i t h i n a given clause, but the clause as a whole must be " b r a c k e t e d " i n some way to prevent relating a given constituent to the predicate o f the neighboring clause. Therefore connectives i n relative and complement clauses cannot be deleted. The English He said (that) he was there is rendered by Powiedzial, ZE tarn b y / w i t h the obligatory complementizer ze 'that'. The English The boy (whom) I met was hungry translates into Polish i n the form Chlopiec,

KTOREGO

spotkalem,

byi

giodny with the obligatory relative pronoun ktory ' w h o m ' , i n the accusative case. The main clause w h i c h follows the subordinate clause is usually preceded by the empty particle to ' i t ' ; for instance, the English When I came, you were sleeping translates as Jak przyszedlem,

TO ty spafei; and the English / / it rains,

stay at home is rendered by Jesli bqdzie padac, TO zostan$ w domu. The

Til

empty particle is not used when the main clause comes first; e.g., Zostanq w domu, jesli bqdzie padac, where the connective jesli ' i f is a sufficient boundary marker. Although not obligatory, to is widely used i n spoken language. I t occurs also w i t h center-embedded relative clauses to mark the beginning o f the second part o f the interrupted matrix clause, e.g., Ten chlopiec, KTOREGO

spotkafem, TO

byI giodny. I n this way the inserted relative is placed w i t h i n " b r a c k e t s " formed by ktory and

to.

6

1.3.5.2. Clear Marking

of Syntactic Relations.

Polish lacks certain rules

that result i n semantically opaque constructions, at the same time providing overt local cues that facilitate interpretations. This can be illustrated by the f o l l o w i n g examples. The first point concerns restrictions on infinitivization rules. Polish lacks constructions o f the "accusativus c u m i n f i n i t i v o " type. Where English uses a infinitival construction, Polish frequently requires a full clause w i t h the finite verb f o r m ,

e.g.

/

saw

him go(ing)

=

Widzialem,

ze

szedl

T:saw

that

he:was:going' or Widzialem (go), jak szedl T:saw (him) as he:was:going'. S i m i ­ larly, / asked him to go = Prosilem

(go), zeby poszedl

'Tasked (him) that

he:goes'. I f the verb can take an infinitival complement, the other complement appears i n dative f o r m , e.g. / ordered him to go = Kazalem mu pojsc

T:ordered

h e . D A T g o : I N F ' . Fisiak et al. (1978) cite the verb uczyc 'to teach' as the only exception, since i t takes both the accusative and the infinitival complements, e.g. Uczg go spiewa-c'kteach

h e : A C C s i n g - I N F ' ( = 1 teach h i m to sing).

It is worth noting that Polish lacks a periphrastic causative construction o f the makeH~VERB causatives,

type. The causative e.g. usypiaclspac

meaning is conveyed mostly by lexical

'make sleep'/'sleep', rozsmieszac

'make l a u g h ' / ' l a u g h ' , zasmucaclbyc

Ismiac si%

smutnym 'make sad'/ 'be sad', or by means

of the nonreflexive/reflexive contrast, e.g. zlamac I zlamac si% 'break'/'break The distribution of Polish to seems quite similar to that of ia in the creole Tok Pisin language, as described in Sankoff (1974), quoted by Slobin (1977). 6

Smoczyriska

610

i t s e l f as i n Chiopiec zlamalpatyk

'boy broke stick' vs. Patyk ziamai si% 'stick

broke i t s e l f . Another device used i n Polish is that o f employing complex sen­ tences w i t h subordinated clauses o f purpose, introduced b y zeby 'so:that' (nonfactive). T y p i c a l l y , the main clause verb specifies the K I N D o f action performed in order to realize the PURPOSE specified b y the subordinate clause as the intended result o f the main clause action. Therefore, more information is con­ veyed than i n English, Make it pretty or Make it hold, w h i c h do not specify the manner o f obtaining the causative results, cf. Polish counterparts o f the above sentences: Pomaluj to, zeby byto ladne 'paint i t so.that (it) be pretty' or Umocuj to, zeby si% trzymah

' f i x i t so:that itself (it) h e l d ' . O n the other hand, using a

general term for causative action as i n Zrob, zeby si% trzymah

'make so:that

itself (it) held' is not equivalent to the English periphrastic causative, since i t conveys additional information, T don't k n o w how to do i t ' or ' D o i t , whatever means y o u e m p l o y ' . The second factor involves demonstrative elements i n the matrix clause. I n the matrix clauses for embedded relatives, demonstrative pronominal anteced­ ents occur, w h i c h serve as " a u g u r i e s " o f the f o l l o w i n g subordination, e.g. Zrobiq. to WTEDY, KIEDY b$d$ miec czas T:shall:do i t T H E N W H E N T w i l l have t i m e ' , Znalaziem to TAM, GDZIE to poio'zyies

T.found i t T H E R E W H E R E you .put

i t ' , Znalaziem TO, CO chciales miec T:found T H A T W H A T you:wanted to:have'. Such pairs comprising the demonstrative antecedent and the relative pronoun provide cues for establishing correct relations between the matrix and the relative clause. The third point concerns the abundance o f local cues (endings). I n simple sentences, the large range o f gender/number/case

agreement may seem redun­

dant. Its usefulness, however, can be best seen i n complex sentences. Compare the f o l l o w i n g English sentence w i t h its Polish counterpart: The cow that the elephant kissed kicked the lion. Krow-a cow-FEM:NOM:SG

ktdr-q which-FEM:ACC:SG

slon-0 elephant-MASC:NOM:SG

pocalowal-0-0 kissed-MASC:3SG

kopnql-a-0 Iwa kicked-FEM:3SG lion-MASC:ACC:SG

Apart from casemarking o n nouns, w h i c h separates the agents i n nominative 'cow'

and 'elephant' from the patient ' l i o n ' and corresponds to the information

conveyed i n English b y w o r d order, gender agreement i n past tense verbs makes the situation clear at first glance. I t is only the cow that can be the agent o f kicking as the past-tense verb kopnqla is marked for feminine and the w o r d 'elephant' is masculine. Analogically, the c o w is precluded as an agent o f kissing, since pocahwal

is marked for masculine. Furthermore, the relative

pronoun ktory occurs i n the accusative feminine singular f o r m , ktorq. Thus i t agrees i n gender and number w i t h its antecedent ' c o w ' , while its case f o r m is

6.

The A c q u i s i t i o n of Polish

611

governed by the predicate o f the relative clause 'kissed'. I t is not difficult to guess that only the cow could be kissed and that the only possilbe agent o f k i c k i n g is the elephant. I n relative clauses, another relative pronoun the indeclinable co 'that' can be used. I n such cases a shadow pronoun appears i n an appropriate case f o r m i n order to make explicit the relation clause-antecedent, e.g. Chiopiec, 7

spotkaiem 'boy that H I M L r n e t ' , Chiopiec,

co GO

co NA NIEGO czekaiem 'boy that FOR

H I M I:was:waiting' (cf. similar constructions i n Hebrew i n Berman, 1985). Another example o f clear marking is the form o f nonfinite verbs. Infinitives are clearly marked as such, due to their specific endings ( i n English, although they are sometimes preceded by to, their form itself is homonymous w i t h finite forms). Polish counterparts o f the ambiguous -ing form are also clearly differ­ entiated,

e.g. pisanie

' w r i t i n g (books)', piszqcy

' w r i t i n g (people)',

piszqc

'(while) w r i t i n g ' w i t h additional cues provided by casemarking i n verbal nouns and case/gender/number marking i n adjectival participles. Thus, Poles do not have such problems w i t h visiting relatives as typical for English-speakers (and, especially, for linguists). The fourth issue is direct temporal reference. The use o f tenses is not re­ stricted by any formal rules (i.e. there is no "consecutio t e m p o r u m " ) ; for instance, Powiedziaia, ze

piszElist

'Sheisaid that ( S H E ) : i s : W R I T I N G letter', where

a complement clause i n the present tense is embedded i n a main clause i n thepas the past tense. I t means that she said T am w r i t i n g ' . O n the other hand, the Polish counterpart o f the correct English sentence She said that she was writing— Powiedziaia, ze PISALA—would

mean that she referred to a past event and said T

was w r i t i n g ' . S i m i l a r l y , an //"-clause w h i c h refers to a future event occurs i n future tense f o r m , e.g. Jesli bqdzie padac ' i f ( i t ) : w i l l r a i n ' , cf. English If it rains. Polish use o f tenses preserves the one-to-one correspondence between tense form and temporal characteristics o f external reality. Finally, we must consider factive/nonfactive reference. L i k e the use o f tense discussed above, mood is not formally restricted, and any proposition considered by the speaker as nonfactive is put into conditional mood. I n conditional sen­ tences, all possible combinations o f m o o d and tense can occur. I f the speaker conceives both events as nonfactive (only possible or truly hypothetical), both clauses occur i n conditional m o o d e.g. Jesli by-0 padai-o, poszedi-0-by-m teatru

' I f COND-3SG

ater:GEN' ( =

rain:PP-NEUT,

go:PP-MASC-COND-lSG

to

do the-

' I f i t - w o u l d - r a i n , I-would-go to the theater;), cf. English / / it

rained . . . , where the past tense is used according to a purely formal rule. There is a specific type o f conditional m o o d used i n nonfactive subordinate clauses introduced by zeby. I t is called " e n t a n g l e d " because o f the impossibility

I n colloquial speech the shadow pronoun can be omitted if the relation clause-antecedent is unambiguous, e.g. To krzeslo, co ja (na nim) siedziakm 'this chair that I (on it) was:sitting', as the usual relation between the chair and sitting is that of sitting on it. 7

612

Smoczynska

of detaching the conditional particle by from the proper connective ze. I n accor­ dance w i t h the general rule concerning conditionals (cf. Section 1.2.6.2), zeby must be immediately followed by a person suffix, w h i l e the remaining verbal form is that o f the "past p a r t i c i p l e , " e.g. Ona chce, ze-by-m poszl-a, 'She wants, t h a t - C O N D - l S G go:PP-FEM' ( = 'She wants me to g o ' ) or Ona chce, zeby-smy poszl-i 'She wants, t h a t - C O N D - I P L go:PP-VIR' ( = ' S h e wants us to g o ' ) . T o summarize, tenses and moods i n Polish tend to carry clear and nonarbitrary semantic information concerning temporal relations o f events and factive/nonfactive evaluation o f the propositional content w h i c h is made by the speaker. The above outline o f Polish presents some main features o f the system as a whole, w i t h more detailed discussions o f the formal features w h i c h are signifi­ cant for the f o l l o w i n g analysis o f the acquisition process. The structures that I have discussed and the examples that I have provided are those o f the normal colloquial use o f the language.

2. S o u r c e s o f E v i d e n c e Unlike some other languages discussed i n this volume, most o f the evidence concerning the acquisition o f Polish is provided by naturalistic data. Experimen­ tal studies dealing w i t h strictly grammatical questions are scarce. 2 . 1 . Naturalistic

Data

2 . 1 . 1 . Diary Studies. There exists a large corpus o f raw data collected 100 years ago (1885-1904) by the eminent Polish linguist Jan Baudouin de Courtenay, w h i c h comprises detailed diaries o f the speech o f his five c h i l d r e n . I n 1974, extracts from one o f those diaries were edited by M a r i a ChmuraKlekotowa. Those extracts (10% o f the whole diary) prove that Baudouin's data still have more than merely historical value (see Zarebina, 1976, and Savic, 1975). 8

Polish diary studies published so far are presented i n Table 6.4. The authors are mainly linguists who recorded and analyzed their o w n children's speech, concentrating on the earliest stages o f language development. Particular analyses differ w i t h respect to scope and delicacy. The monographs by Kaczmarek, Smoczyhski, and Zarebina contain the most detail. Later periods o f the gram­ matical development find a less complete documentation i n diary studies. There are t w o short works: (1) a paper by Siatkowscy (1956) w i t h some observations o f the authors' son o f age 2;6 and (2) a paper by W o j t o w i c z (1959), w h o reports "formations by analogy" i n her duaghter's speech from 4;0 to 5;0. The diary studies provide copious evidence on the acquisition o f the basic system; howev­ er, because o f the w e l l - k n o w n shortcomings o f diary studies i n general, the data are frequently difficult to interpret from a psycholinguistic point o f v i e w . 8

The manuscript (13,000 pages) is kept in the Polish National Library.

6.

T h e A c q u i s i t i o n of Polish

613

TABLE 6.4 Polish Diary Studies Author

Date

Children-Age

Rzetkowska

1908

girl 0-4;6

Rzetkowska

1909

boy 0-3;0

Brenstiern-Pfanhauser

1930

Wawrowska

1938

girl l;0-2;3 boy l;0-2;3 boy 0-2;3

Skorupka Kaczmarek

1949 1953

Smoczyñski

1955

Zarebina

1965

girl 0-2;0 girl 0-3;6 plus casual observations on three older children girl 0-3;0 boy 0-3;0

girl 0-2;3 plus casual observations on two younger children

Description vocabulary, remarks on morphology and syntax (33 pages) vocabulary, somewhat more detailed discussion of grammar, and a comparison with the older girl's development (54 pages) phonetics, morphology, syntax (56 pages) vocabulary, general remarks on grammar, psychological analyses (57 pages) general remarks (39 pages) full linguistic description (90 pages, summary in German) most detailed analysis of phonology and semantics of first words, remarks on morphology and early syntax (230 pages, summary in French and in Russian) full description of phonology, morphology and syntax, vocabulary (100 pages, summary in French)

The monograph by Chmura-Klekotowa (1971), concerned w i t h children's neologisms, is based on longitudinal data o f several children (including early K r a k ó w data, see Section 2.1.2) as w e l l as experimental data o f elicited w o r d formations. I t provides a detailed and systematic analysis o f 5000 neologisms recorded i n children's speech. T o the best o f m y knowledge, it is the most exhaustive published study o f derivational processes i n child language. 9

2.1.2. The Szuman Project: Early Kraków Data. A separate group o f naturalistic longitudinal data are those recorded i n K r a k ó w i n 1950-1960 as a part o f a research project on language and thinking directed by Stefan Szuman (Jagiellonian University). A l t h o u g h the samples were recorded by the mothers o f the subjects i n handwriting (or sometimes i n short-hand), they differ significantly from classic diary data. The observers d i d not record isolated utterances, selected for some reason or other, as was the case w i t h classic diaries. O n the contrary, the observers collected short sections o f discourse w h i c h covered the entire Chmura-Klekotowa used also data from Baudouin's children. She found significant differences between productive suffixes which they used to form neologisms and those employed by contemporary children. 9

614

Smoczyhska

chronological sequence o f utterances addressed to the child as w e l l as the child's productions. Changes i n behavioral and situational context were also observed and recorded. Every day several samples were taken, w i t h the observers trying to make the fullest account possible o f what they considered to be typical situations. Daily samples were collected for several years o f the child's life, and i n the case of children studied for the longest periods, they comprise up to about 20,000 utterances. A l l the children studied had at least one parent w i t h a higher educa­ tion degree, w i t h most mothers being psychologists. Four girls and four boys were studied through the f o l l o w i n g period o f development: (1) Inka l;0-8;0, (2) Basia l;5-8;0, (3) Kasia 1;3-3;1, and 3;9-4;0 (Kasia is M i c h a l ' s younger sister), (4) Tenia 1;5-1;7 (Tenia is Inka's younger sister), (5) Jas l;6-6;0, (6) M i c h a l 2;0-6;0, (7) M i s 2;9-5;10, and (8) Janek 4;2-7;3 ( M i s and Janek are t w o brothers, observed simultaneously). Szuman's project was carried out i n collaboration w i t h M a r i a Przetacznikowa and L i d i a Geppertowa. One part o f the project was concerned mainly w i t h vocabulary development: its input consists o f Szuman's (1955) w o r k on nouns and verbs, Przetacznikowa's (1956, 1959) studies on adjectives and adverbs, and Geppertowa's (1968) monograph on the acquisition o f prepositions and conjunc­ tions. Another part o f the project dealt w i t h syntactic development and is repre­ sented by Przetacznikowa's (1968) study o f the syntactic structure and function of child utterances. This w o r k is based on the corpus o f all declarative utterances recorded for Jas over the period 1;6 to 6;0. The data were classified according to a taxonomy used by Polish grammarians, and detailed quantitative analyses show the structural development o f the child's syntax. Qualitative analyses, however, reflect the approach o f the whole Szuman project; that is, they concentrate on the ways i n which children's thoughts are expressed through language rather than on the process o f learning the grammatical structure itself (Przetacznikowa, 1976, 1978). Although the children w h o were studied are now more than 20 years o l d , the data are still used for various kinds o f analyses, concerning psycholinguistic and psychological problems. M y o w n dissertation on early syntax (Smoczyhska, 1978a; also see papers written i n English: 1976a, 1976b, 1978b, 1980, 1981) was based on the early data o f five o f those children. The same corpora were analyzed by Przetacznikowa (1975) w i t h i n the framework o f Fillmore's case grammar. Other detailed studies on these data involve negation ( K u b i t , 1977), a functional analysis o f children's questions (Lig^za, 1979), and a number o f other works i n preparation. 2.1.3. Tape-Recorded Discourse: Late Krakow Data and Other Research. I n K r a k o w , at Jagiellonian University, new data were collected i n the 1970s w i t h i n a child language project directed by M a r i a Przetacznikowa. Tape record­ ings o f several children were made by their mothers, w h o were mainly members of the project. I n order to collect relatively continuous data, the recordings were

6.

The A c q u i s i t i o n of Polish

615

made t w o to four times a week i n the early stages. The frequency o f recording decreased w i t h age. The f o l l o w i n g children (one girl and five boys) were re­ corded: Agnieszka, W i t e k , Szymon, Przemek, and t w i n brothers, M i c h a l S. and Maciek from the beginnings o f syntax to about 3;0 ( w i t h some extra samples collected later). Grace W . Shugar at Warsaw University collected data from t w o children over the period 1;6 to 3;0. Those data were analyzed i n her study on relations between language structure and activity structure (Shugar, 1972, 1974, 1976a, 1976b, & 1978). These data also served as a basis o f the report on grammatical develop­ ment o f Polish children found i n Slobin (1973). A t A d a m M i c k i e w i c z University i n Poznah, Richard M . Weist conducted a research project which combined longitudinal and cross-sectional experimental designs. I n the longitudinal phase six children were tape-recorded twice a month for about 6 months and t w o children for shorter periods. The observations i n ­ clude detailed contextual notes w h i c h were integrated w i t h the discourse. The naturalistic observations concentrated on the period from 1;6 to 2;6. T w o o f the children began participation at 1;7, t w o at 1;9 and t w o at 2;2 (Weist, Wysocka, Witkowska-Stadnik, Buczowska, & Konieczna, 1984). 2.2. Experimental

Studies

Although many experimental studies concerning child language have been carried out i n Poland, those w h i c h are relevant for the present discussion are few. Smoczyhska (1972) investigated the control o f the singular noun declension forms i n 3-, 4-, 5-, and 6-year-old children (10 children per age group) using elicitation techniques for real and nonsense words. Olszowska-Guzik (in prepa­ ration) investigated selected aspects o f the control o f the linguistic system by 6year-old children w i t h respect to environmental differences (4 milieux x 30 children). She used various experimental procedures such as elicitation tech­ niques, grammaticality judgments, etc. Mystkowska (1970) analyzed 6-year-old children's productions (story narration on the basis o f a series o f pictures) which were then compared w i t h parallel data obtained from 11-year-olds and adults. Weist (1980, 1983a,b) used comprehension and production tests i n studies o f tense and aspect and o f w o r d order w i t h ten 2;6-year-old and ten 3;6-year-old children. A l l subjects were also tape recorded i n naturalistic situations for one 45 minute session. 2.3. Sources

of the Data Used in This

Paper

W h i l e examining the existing data concerning the acquisition o f Polish from the point o f view o f their relevance for the present paper, I discovered that only a few o f the previous studies could be directly utilized. As far as morphology is concerned, only the diary studies proved useful. However, even i n the case o f the diary studies the data are limited to the acquisition o f the basic system up to about 2;3, and the authors (mainly linguists) were interested i n the emergence o f adult

616

Smoczyhska

categories. These investigators d i d not attempt to identify the rules that the child could have formed for himself. T y p i c a l l y , several examples o f the first correct use o f a form were given followed by instances o f erroneous usage. Thus, the reader is not able to estimate the ratio o f correct/incorrect forms for a given linguistic expression. A s for syntax, i t has been investigated by the K r a k o w psychologists but, i n view o f their cognitive orientation, they centered on refer­ ential meaning. The acquisition o f the linguistic forms themselves was given only marginal attention. 10

Consequently, I undertook the task o f exploring the corpora o f the K r a k o w children. The first step consisted o f extracting all erroneous forms from the data of several children aged 2 to 6. I determined the morphological and syntactic forms which were especially difficult for Polish children. Subsequently, the corresponding instances o f correct use o f those forms were singled out i n order to establish the significance o f the errors, i.e. whether they were consistent or only sporadic. Sometimes it was necessary to go back to the period preceding the occurrence o f the first error i n order to trace the emergence o f the category itself. When no erroneous forms were attested, I tried to find out whether this was so because the child used the form correctly or because he d i d not employ it at a l l . Because o f the short time at m y disposal, I could not do this for all kinds o f errors, and I chose only those w h i c h seemed to be o f theoretical interest. Regard­ ing case for instance, errors i n totally irregular exceptional forms are less i n ­ teresting than those occurring w i t h more or less complex yet regular forms. I also gathered all utterances i n w h i c h the conditional mood appeared, as w e l l as those containing relative clauses. These t w o structures are interesting i n themselves, both from the cognitive and the strictly formal point o f v i e w . Another part o f this study consisted o f analyzing the emergence o f basic linguistic categories and forms i n the data o f three children covering the period from the very beginning t i l l the age o f 2;6. This analysis was carried out as a supplement to the diary studies i n order to obtain detailed knowledge concerning consecutive steps i n the acquisition o f the basic system. This research provides the basic data o f the present paper. The results ob­ tained by other authors were frequently confirmed, w h i c h I shall point out at relevant points. The analysis o f the raw data helped me to form a general outlook concerning the acquisition o f Polish w h i c h can be compared w i t h what is k n o w n about other Slavic languages. Additional arguments were provided to confirm my strong opinion that the acquisition o f a language (especially i n later stages) cannot be reasonably accounted for i f we try to divide it into the acquisition o f particular forms or subsystems, since the various parts o f the linguistic system are interlinked. The interrelationship is inherent and organic so that any analysis should be carried out w i t h i n the broader framework. Another important point o f caution to be made here concerns I N D I V I D U A L DIFFERENCES. The analyses I carried out made me realize that there can exist 10

This criticism refers to Gvozdev's (1949) work on Russian as well.

6.

The A c q u i s i t i o n of Polish

617

immense differences between the individual paths children take to arrive at roughly the same final result, i.e. mastery o f the full system. I t is not the case that one child acquires all forms more quickly than another, but rather that the order o f acquisition o f the correct adult forms is never the same for t w o different children. A n investigation o f these individual ways o f arriving at the full adult system w o u l d require a very thorough study. Having analyzed the data o f more than ten subjects, all that I can say about Polish children i n general is that there are linguistic forms whose acquisition is L I K E L Y to cause some difficulties for varied intervals o f time somewhere between the age o f 2 and 6. W i t h one child this interval can last 3 months, and w i t h another 3 years, and w i t h still another child the predicted difficulty may not appear at all. Some scholars assume that i f errors attested i n the data o f one child are not found i n other children, they must have gone unnoticed. The nonselective and continuous daily sampling used w i t h the K r a k o w children makes me preclude such a possibility. The absence o f a predicted difficulty is most puzzling, especially i f the source o f difficulty can be easily accounted for. That is w h y I prefer to discuss individual children rather than to make general statements. O n the other hand, the observations concerning individual variation call for great caution when generalizing the results obtained from a diary study based on the data from a single child.

3. O v e r a l l C o u r s e o f L i n g u i s t i c D e v e l o p m e n t A t the one-word stage there is no evidence for productive use o f any inflectional forms, w i t h no significant changes being observed i n this respect i n the earliest two-word combinations. Indeclinable words used i n this early period are either specific baby talk items, e.g. am ' f o o d , to eat' and onomatopoeias, e.g. hau-hau 'dog, to bark' or adult words. I n the latter case, nouns occur i n the nominative singular, or sometimes i n one o f the oblique cases as " c i t a t i o n f o r m s . " Verbs occur i n the third singular present tense f o r m or i n the imperative or the infinitive form. Zarebina's (1965) daughter, for example, began to use the imperative form at 1;0, e.g. [es'] = wez 'take' and [ X u c ' or oc] = chodz 'come o n ' , and the third singular indicative form at 1;3, e.g. [ g a ] = gra 'plays', [pxa] = pcha 'pushes', [c'ita] = czyta 'reads', etc. Zarebina reported that Hania used the imperative form to express demands and desires and used the indicative form to make statements. W h i l e these verbs have a pragmatic value, they are syntactically frozen forms. W e do not find contrasting verb forms at this stage o f develop­ ment, e.g. chodz 'come o n ! ' versus idzie 'he:is:going' or pcha 'he:pushes' versuspchaj 'push!' I n m y data verbs w i t h the imperative form occur M O S T L Y i n requests, and verbs w i t h the indicative form occur M O S T L Y i n the declarative context. However, i f children happen to use one o f these verbs i n the less frequent context, they do not change the form o f the verbs. Sometimes several forms are used i n free variation, e.g. da, daj, dam, dac, or damy-damy, ' g i v e ' . A l l these forms o f the verb 'to give' were used interchangeably by one o f the children studied. h

618

Smoczyriska

The development o f the syntactic structures o f m u l t i w o r d utterances, recorded in five children during the six months after the emergence o f first w o r d combina­ tions, was described i n Smoczyhska (1978a). I t was found that by the end o f the period studied the children had arrived at full control over most o f the basic simple sentence patterns. Assessment was made according to three criteria: (1) the ability to produce a number o f complete sentences o f a given type; (2) pattern generalization, as revealed by a relatively large variety o f predicates (Smoczyhska, 1981); and, (3) automaticity o f pattern generation, as evidenced by the lack o f replacement sequences on the one hand, and examples o f expanded patterns on the other. A n example o f the developmental analysis w i t h respect to the transitive sentence pattern is given i n Smoczyhska (1976a). As for morphological development, i t usually starts i n the third month o f the two-word stage. Regarding nouns, the initial contrast is accusative and genitive singular as opposed to nominative, w i t h vocative appearing at the same time or slightly earlier. I n verbs, the second singular o f the imperative emerges as opposed to the third singular o f the present tense. The infinitive is found a little later. The next step is the emergence o f the nominative plural and the accusative plural, followed by the instrumental singular, the locative singular ( i n locative phrases w i t h the preposition omitted) and the dative, w h i c h is typically m i x e d w i t h the genitive (see Section 4 . 3 . 2 . 1 ) . The f o l l o w i n g new verb forms appear more or less simultaneously—all o f them mostly i n the third singular: perfective future (formally identical w i t h the imperfective present tense form), past tense (mostly perfective, but also imperfective, see Section 6.1.2) and a little later, also the analytic imperfective future. Thus tense and aspect emerge very early and simultaneously. The above list o f forms characterizes the average level o f morphological development o f a 2-year-old. The sequence was established on the basis o f the data o f three children (Kasia, Jas, and Basia), and it is consistent w i t h the findings reported i n Zarebina (1965) and Kaczmarek (1953). It should be said, however, that such an overlap o f morphological and syntac­ tic development is not always observed. W i t h some children, the acquisition o f morphology is slowed d o w n , probably due to individual phonological difficul­ ties, cf. the data i n Smoczyhski (1955) and i n Baudouin de Courtenay's diary (1974). A l s o Smoczyhska (1978a) reported a contrastive example o f t w o o f her subjects (Jas and Tenia). I n the fourth month after the first w o r d combinations emerged both children reached the same level o f syntactic development. H o w e v ­ er, while Jas's morphological development was fairly advanced, Tenia still tend­ ed not to make inflectional contrasts. As far as the acquisition o f specific case endings is concerned, most o f them are used correctly from the very moment o f emergence o f a given category, as opposed to Russian, where " i n f l e c t i o n a l i m p e r i a l i s m " (Slobin, 1966) is com­ mon. This is due to the early mastery o f grammatical gender i n the singular, which determined assigning a given noun to an appropriate declensional pattern (see Section 6.1.1). Overgeneralization occurs mainly when an irregular ending

6.

The A c q u i s i t i o n of Polish

619

competes w i t h a regular one, and less frequently w i t h i n regular patterns (see Section 4 . 3 ) . In the third year o f life new tasks appear. The child learns the remaining elements o f the declensional pattern, especially plural case forms o f nouns w h i c h have not been used so far. D u r i n g this period o f development children expand various noun phrase constituents, begin to use oblique case forms o f adjectives, and master the whole declensional pattern o f adjectives. Przetacznikowa (1968) computed the percentage o f noun modifications i n all kinds o f expansions occur­ ring in simple declarative sentences produced by Jas, and she found that at­ tributes appeared i n about 20% o f expanded sentences from 1;6 to 3;0. Their number systematically increased w i t h 36% from 3 to 5 years and 4 2 % from 5 to 6 years. A t the same time, the child learns how to combine basic patterns that he had previously acquired to form various types o f complex sentences. The ratios o f different kinds o f declarative sentences recorded for Jas are presented i n Table 6.5. I t can be seen that coordination precedes subordination and occurs more frequently i n the early years. Jas started using conjunctions relatively early. W i t h "telegraphic" children, however, who omit conjunctions over a relatively long period, the dominance o f parataxis w o u l d be even stronger since mere juxtaposi­ tions must normally be interpreted as coordination. Another important develop­ ment shown i n Table 6.5 is the g r o w i n g ability to combine more than t w o clauses. The different structural patterns o f such constructions have been de­ scribed by Przetacznikowa (1968), and a detailed discussion o f complex sen­ tences can be found later i n this paper (see Sections 6.1.4 and 6.2).

TABLE 6.5 The Proportions of Declarative S e n t e n c e s of Varying Complexity in the Speech of Jas (Based on Przetacznikowa, 1968) 3

Age

l;6-2

simple sentences coordinate clauses main & subordinate more than 2 clauses b

Total N

0

2-3

3-4

4-5

5-6

94% 6% — —

78% 13% 7% 2%

64% 18% 13% 5%

52% 25% 14% 9%

46% 29% 12% 13%

100% 372

100% 2365

100% 2223

100% 1484

100% 2243

The proportions would be somewhat different if imperative sentences were included since some kinds of subordinate clauses typically occur first with requests. ^including expanded. subordinate or mixed. a

c

620

Smoczyhska

Another important task for the 2-year-old involves mastering the distinction o f person i n verbs. The first-person form emerges i n opposition to the third-person form, which was initially used for self-reference as w e l l as for the second and third person. Later on the second person is also introduced. The acquisition o f the rules o f deictic switching takes several months and is followed by the mastery of the pronominal declension pattern, as w e l l as the discourse rules o f pronominalization. A l l these processes are gradual. D u r i n g the third year o f life, many examples can be found i n the child's utterances o f some o f the required forms but not all o f them. W h i l e the conjugation patterns o f the indicative are supplemented by adding first- and second-person forms, the conditional m o o d emerges, first i n its entangled form i n subordinate clauses, and later i n main clauses w i t h some typical formal difficulties (see Section 4 . 6 . 3 ) . A l s o , passive constructions are used from the age o f 2 without explicit mention o f the agent. B y the end o f his third year o f life the child masters the syntax o f complex sentences as w e l l as discourse rules w h i c h makes i t possible for h i m to construct longer texts i n a more coherent and economical way. This textual function is realized by means o f a complex system i n v o l v i n g pronominalization, reflexivization, ellipsis, w o r d order, focal stress, etc., all these elements being related to each other. Finally, after 3 years o f age, the child learns the devices for compact­ ing information i n still more economical although less transparent f o r m , w h i c h involves infinitivization, nominalization, etc.

THE D A T A 4. T y p i c a l E r r o r s The present chapter gives a relatively complete outline o f errors that were found in naturalistic data o f 10 children aged 1;6 to 6;0. Although several kinds o f erroneous forms are reported i n earlier diary studies, the studies lack sufficient information concerning frequency o f occurrence and the ratio o f correct versus incorrect forms. I t is only when these aspects are taken into account that an exact evaluation o f the range o f erroneous forms can be made and the f o l l o w i n g kinds of errors distinguished: (1) consistent errors; (2) inconsistent erroneous usage o f forms i n their earliest appearance, resulting from unsuccessful attempts at quot­ ing a particular form or reconstructing its general shape; (3) mistakes, i.e. spo­ radic occurrences o f erroneous forms after the given rule has been learned, attested by a prevailing number o f correct uses (these errors constitute a 10% margin w h i c h remains after the 90% criterion has been met); (4) free variation o f correct and incorrect forms; and, (5) totally isolated occurrences o f erroneous forms which do not follow typical structural patterns. Another classification is related to the character o f the adult norm w i t h w h i c h the error is compared, where the norm we refer to is that o f spoken language as

T h e A c q u i s i t i o n of Polish

6.

621

used by educated native speakers. Some morphological patterns that occur i n a language are R U L E G O V E R N E D (or regular). Some o f these rules are simple and others are complex. Regardless o f the complexity o f the rule the distribution o f forms is based on some explicit grammatical criteria, for instance, a group o f words sharing certain features (criterion A ) takes the form a, whereas another group o f words w h i c h satisfies the criterion B takes the form b. O n the other hand, there are I R R E G U L A R morphological patterns w i t h no criteria for their application, for instance, a list o f words: C C . . . C w h i c h do not share any common feature takes the form c. The latter is found to be less interesting than the former, as i n such cases learning the correct form consists o f rote-memoriza­ tion o f a list o f exceptions, although the strategies that the child employs i n such cases can prove to be o f some interest, e.g. overgeneralization, looking for distributional criteria where there are none, etc. p

2

n

The distinction between the complexities o f the regular system on the one hand and the irregularities on the other has not received adequate attention, i n spite o f its crucial relevance. For instance, many authors claim that Slavic children are not able to acquire the declension system before the age o f 5 or even 7. I f however the term " s y s t e m " is taken to mean a given set o f rules excluding irregularities, then i n fact, the system itself is mastered much earlier. I n his discussion o f Russian, Slobin (1966, 1968) emphasized certain aspects that must have been unknown to those investigating the acquisition o f English. Slobin discussed morphological complexities and irregularities which the child must face and w i t h w h i c h — m u c h to the amazement o f scholars—he finally manages to cope. I n this place, however, I w o u l d like to introduce the distinction between a set o f rules and the exceptions to these rules i n order to make the acquisition o f Slavic languages more easily understood and a little bit less amazing. W e shall therefore try to classify particular types o f errors i n terms o f the relation between children's productions and the norms o f the language. The simplest k i n d o f error, very common w i t h English-speaking children, but consid­ erably less so w i t h Polish children, is that o f omitting a grammatical morpheme in an obligatory context. The opposite k i n d o f error is overmarking. Still another type consists o f using correct adult forms i n improper contexts. Sources o f difficulties can be found i n different characteristics o f the adult norm: (1) complexity o f the criteria for rule application; (2) complexity o f the rules themselves, i.e. when the rule involves a combination o f several opera­ tions; and, (3) given a simple criterion and a simple rule, a difficulty can arise from a mismatch between the system acquired so far by the child and the rule under acquisition. Consequently, the f o l l o w i n g situations can be distinguished: (1) disregard— or lack o f r e c o g n i t i o n — o f adult criteria o f form distribution w h i c h results i n overgeneralization or formation o f an idiosyncratic or simplified rule w h i c h does not exist i n the language; (2) acquiring all rules without learning the criteria o f their distribution, w h i c h results i n free variation o f the forms i n question; and, (3)

622

Smoczynska

difficulties i n the complete application o f a complex rule after the criteria o f its use have been learned, i.e. partial application o f a rule. A l l these types are attested both among early and late errors. W h i l e the early errors can be explained i n terms o f operating principles (whether cognitive or strictly linguistic), the later errors should be related to the entire system o f the language. W e must attempt to discover the discrepancy between a given rule (or the criteria o f its application) and the distinctions already acquired, rather than some isolated operating principles. A discrepancy o f this k i n d can result i n U shaped developments such as the blurring o f some well-learned distinctions due to the acquisition o f a rule w h i c h contradicts the distinctions. Moreover, it seems that there exists another type o f U-shaped development i n which the c h i l d , i n his attempt to create a m a x i m a l l y economical system, gets r i d o f an already acquired sub-rule and overextends the one that he had just learned. 4 . 1 . Omission

of Obligatory

Functors

I n view o f the general morphological structure o f Polish, only the f o l l o w i n g free grammatical morphemes are potential candidates for telegraphic reduction: prepositions, conjunctions, the copula jest, the future tense auxiliary bqdzie, two enclitic particles ( i . e . the reflexive siq. and the conditional by) and possibly also enclitic movable endings o f the plural past tense forms (i.e. -smy, -scie). I n the earliest data the children studied omitted prepositions and the reflexive particle si%. The copula jest was also deleted, however, it is not obligatory i n Polish. I n the case o f other candidates for reduction, omission could not occur in the earliest utterances because o f the lack o f appropriate obligatory contexts. Later, the children can be clearly divided into "telegraphic" versus " n o n telegraphic" types w i t h most o f the children falling i n the latter category. T w o of the children studied (Basia and Inka) persistently omitted some obligatory morphemes. W i t h the development o f syntax, new obligatory contexts emerged for other free morphemes and Basia and Inka now omitted the new set o f free morphemes. W i t h these t w o "telegraphic" children (as w i t h English-speaking children) the appearance o f the morphemes that had been omitted earlier was a gradual process taking place during the third year, for instance, at the given moment the children were using some conjunctions and omitting others, even though they were building sentences w i t h both kinds o f obligatory contexts. Typically, asyllabic prepositions (w ' i n ' , z ' w i t h ' ) were omitted for a longer time than the syllabic ones (do t o ' , na ' o n ' ) . This is due more to phonetic reduction o f an initial consonantal cluster than to processes discussed here. I n the case o f the remaining children, prepositions and the particle si% appeared before the age o f 2, and the remaining free morphemes were used from the moment o f the emergence o f appropriate contexts, and their usage satisfied the 90% criterion from the point o f emergence. 4

Let us turn to the data o f the "telegraphic" children. Establishing obligatory contexts for conjunctions and other connectives is not very easy since i n some

6.

The A c q u i s i t i o n of Polish

623

kinds o f complex sentences their use is optional. However, w i t h complement clauses, "complementizers" are obligatory, and Basia and Inka omitted them. The resulting utterances w o u l d be correct i n English but not i n Polish, e.g. Inka (2;4) Tatuspowiedzialjestem

grzeczna 'Daddy said L a m g o o d ' , cf. correct Tatus

powiedzial, ZEjestem grzeczna

'Daddy said T H A T L a m g o o d ' .

Another obvious example is the omission o f the conditional particle by, resulting i n utterances i n w h i c h the predicate occurs i n a past-tense form while they clearly refer to some hypothetical future events, e.g. Basia (2;6) refused her friend's invitation to c l i m b a h i l l and said, J a nie pojdg, ja spadlam, mamusiaplakala

moja

T not w i l l : g o , I fell, m y M o m m y c r i e d ' . Since Basia had never

fallen d o w n a h i l l before, and thus her mother could never have cried for that reason, i t is obvious that the g i r l actually meant to say, Ja Bv-m spadla,

moja

mamusia BY plakala ' i W O U L D - 1 S G fall (and) m y M o m m y W O U L D c r y ' . Omis­ sions o f the conditional by are easy to detect i n Basia's utterances as early as 2;0 because Basia had already demonstrated an understanding o f the correspondence between tense and temporal reference at an earlier phase o f development (see Section 6.1.2). Hence, we k n o w that a w o r d like plakala is not intended as a past form but rather as a past participle w i t h conditional particle by omitted (cf. by pi akala or plakalaby). A m o n g the candidates for omission listed above, t w o behave differently i n that one is seldom omitted and the other is often omitted irrespective o f whether the child demonstrates

" t e l e g r a p h i c " tendencies or not w i t h respect to other

functors. First o f a l l , the future tense auxiliary bqdzie appears very early i n all children. This can be explained both by its clear semantic function, and by the fact that unlike other free morphemes, the auxiliary b^dzie is polysyllabic and receives normal penultimate stress. A s such it can be more easily given the status o f a " n o r m a l " w o r d . W h i l e the periphrastic future is formed quite early, some investigators have observed the omission o f bye 'to be' during the emergence o f the form. I n the earliest data, I found examples such as Basia spa-c 'Basia sleepI N F ' ( = 'Basia to sleep') w h i c h could have been due to the omissions o f either the future auxiliary bgdzie or the verb chce ' w a n t ' . When considering truncated forms o f the periphrastic future one must be careful to evaluate the discourse context because some contexts provide the option o f deleting the auxiliary. Excluding elliptical examples, I found no exam­ ples o f auxiliary deletion w i t h the past participle, e.g. pisala instead o f bqdzie pisala 'she:will write (be w r i t i n g ) ' . Weist (personal communication) has found such truncated forms but very infrequently. I n contrast to Polish where the polysyllabic auxiliary is seldom omitted, Kolaric (1959) has found auxiliary deletion i n Slovenian where the auxiliary is monosyllabic. This shows that the poly- versus monosyllabic nature o f this functor is responsible for its resistance to omission. Secondly, the movable ending -smy o f the first person plural past tense is omitted both by " t e l e g r a p h i c " and "non-telegraphic" children. The use o f this

624

Smoczyhska

ending involves some quite special problems w h i c h are discussed i n Section 4.4.4. The second person ending -scie is very rare. 4.2.

Overmarking

Redundant marking appears most clearly w i t h movable endings. I n the pro­ cess o f learning acceptable positions o f an ending, frequently a transitional stage occurs, when i t is used twice. W i t h conditional forms, one k i n d o f error consists of using the particle by twice. For instance, when Basia began to use explicit marking at 2;7, she introduced the particle by which had been omitted before. She used the particle i n t w o positions, e.g. after she was told by her mother not to take her shoes off, she said, Bo co? Bo zaziebilam-BY si$ BY? 'Because what? Because I-would-get-a-cold w o u l d ? ' Such final repetition o f by is limited to the very early examples. Later on, the particle appears twice i n each o f the t w o alternate correct positions, i.e. as an enclitic and postverbally, e.g. Basia (2; 11) A moja mamusia tez BYmiaia-BYladne

wloski ' A n d m y m o m m y also w o u l d have

would pretty hair' (cf. correct by miaia or alternatively mialaby). Such errors can be found i n other children, and they are also reported for Russian i n Gvozdev (1949). Another k i n d o f error w i t h conditionals involves overmarking for person. A s w i l l be shown below (Section 4 . 6 . 3 ) , children tend to attach the person suffix to the past participle and not to the particle by. W h i l e learning the correct position of the person suffix, they frequently use it twice (once w i t h the past participle and w i t h by), e.g. i n the context where mother said, Nie pojezdzit-by-s

na

koniku? ' N o t ride:PP-COND-2SG on horse' ( = ' W o u l d n ' t y o u (like to) ride a horse?'),

and Jas

(3;0) replied, Nie pojezdzii-em

by-m

C O N D - 1 S G ' ( = I w o u l d n ' t (like to) ride.) cf. correct Nie

' N o t ride:PP-lSG pojezdzil-by-m.

In case o f the analytic future w i t h the past participle, analogous examples o f person overmaking can be found, but they are significantly less frequent, e.g. Jas (2;7) W spodenkach

b^dzie-m lezal-em ' i n trousers w i l l - l S G l i e : P P - l S G ' ( = T

w i l l lie d o w n w i t h m y trousers o n ' ) , cf. correct bed$ lezai. Siatkowscy (1956) reported such errors i n their son's speech at 2;5. However, there is no informa­ tion on how consistent the error was. I n m y data, it should be treated as a mistake or a slip o f the tongue rather than a consistent error. Finally, another case o f overmarking is that o f using a sequence o f t w o conjunctions as bo zeby 'because so:that', bo dlatego 'because because' or t w o connectives i n comparisons as i n wiekszy niz od ciebie 'bigger than from y o u ' . 4.3.

Overgeneralization

This strategy is applied first o f all when an irregular ending competes w i t h a regular one, the latter being overextended to include irregular cases. Some exam­ ples are presented below i n Section 4 . 3 . 1 . Another process is the overgeneraliza-

6.

The A c q u i s i t i o n of Polish

625

tion o f one o f the t w o regular forms, i.e. when the child fails to discover relevant criteria (see Section 4 . 3 . 2 ) . 4 . 3 . 1 . Irregular

Versus

Regular

Form

4 . 3 . 1 . 1 . Genitive Singular of Masculine Nouns: the Irregular -u Replaced the Regular -a Ending.

by

The class o f masculine nouns w i t h the genitive -u suffix

is quite large. There exist some vague criteria, such as [—animate], e.g. pociqg 'train' or [ + mass n o u n ] , e.g. cukier 'sugar' or [ + loan w o r d ] , e.g. papier 'paper', but they are not adequate since not all nouns having one o f these features take the -u ending i n the genitive singular. Children show a tendency to overextend the regular -a ending. The reverse overextension is rare. I t should be noted, however, that the number o f words to w h i c h the -u ending is applied correctly constantly grows w i t h age, and the errors o f overextension o f -a quickly become infrequent. I am unable to explain the way i n w h i c h the children elaborate a rule which grammarians have been so far unable to make explicit. Yet i f not rule governed, the children w o u l d have to retain the information for each item sepa­ rately. This seems unlikely since the number o f words that take the -u ending is large. One o f the factors facilitating the acquisition o f correct forms may be the high frequency o f genitival forms as i t is the grammatical case that has the largest range o f functions. 4 . 3 . 1 . 2 . Regularization Consonant

of the Irregular

in the Nominative Singular.

Feminine Nouns Which End in a

The irregular feminine declensional

pattern is as follows i n the singular: N O M & A C C -0, G E N , D A T & L O C -yl-i, and I N S T R -q. Here the child faces a conflict situation. W h i l e the adjective and verb agreement indicates the feminine gender, the lack o f an ending suggests the masculine gender. These nouns constitute quite a large class. One o f the options chosen by children is that o f avoiding these forms and using regular diminutive forms, e.g. mysz 'mouse' becomes myszka. D i m i n u t i v e forms are also used by adults when they address children and even krew ' b l o o d ' becomes krewka i n the baby-talk register. I f the child does not avoid the base f o r m , he w i l l either treat i t as masculine and decline i t accordingly (together w i t h appropriate agreement) or else he w i l l provide i t w i t h the regular -a ending and treat as a regular feminine noun. The selection o f one or another pattern is subject to free variation. The child can say Widzialem kolej-% T:have:seen t r a i n - F E M : A C C : S G ' using the regular feminine accusative ending

and immediately afterwards

say Ten kolej

jechal-0

' t h i s : M A S C : N O M : S G train w a s : g o i n g - M A S C w i t h the demonstrative ten and past tense jechal marked for masculine. Smoczyhska (1972) found that children over 4;0 k n o w this special declen­ sional pattern, but they are not likely to apply i t when there is a clear diminutive correspondent o f a noun, e.g. mysz is replaced by overregularized mysz-a (under

626

Smoczyhska

the influence o f mysz-k-a) yet the noun kosc 'bone' is declined correctly since its diminutive kostka has a different meaning. 4.3.1.3. Irregular and Infrequent Forms. The dative singular form -u w i t h masculine nouns and the instrumental plural -mi can serve as examples o f irreg­ ular and infrequent forms. They are used for a very limited number o f words without any explicit criteria so that they are marginal w i t h respect to the over­ whelming number o f words that take the respective regular endings (-owi and -ami). Overgeneralization o f the regular pattern is most typical i n such cases. Smoczyhska (1972) d i d not find any progress i n 5- and 6-year-olds as compared w i t h 4-year-olds w i t h respect to the dative i n -u. Olszowska-Guzik (personal communication) has supplied me w i t h her data concerning the acquisition o f irregular forms i n 6-year-olds. Using elicitation techniques she tested 120 c h i l ­ dren from four different social backgrounds and found that some items o f the kind described above were incorrect i n nearly 100 percent o f responses, irrespec­ tive o f the social background o f the child. I n some other instances, social back­ ground proved to be significant. 4 . 3 . 2 . Competition

of Regular

Forms

4 . 3 . 2 . 1 . Genitive Case Instead of the Dative (and Vice Versa). According to Slobin (1973), so far no scholar has reported consistent m i x i n g o f t w o cases or using one case instead o f another. Gvozdev (1949, v o l . 2, p. 84) explicitly states that such errors never occur i n his data. According to m y Polish data, however, several children went through a transitional period o f substituting the genitive case for the dative. The overextension o f the genitive case usually follows some early correct uses o f dative forms, and it can persist for some time, e.g. 1;7 to 1;10 i n Kasia and 1;9 to 2;5 i n Basia. I n order to explain this phenomenon, several features o f the Polish declen­ sional system should be presented. First o f a l l , the dative case i n Polish has a very restricted range o f functions and all o f them referring to animate beings. Receiver, Benefactive, and Experiencer are the most typical f u n c t i o n s . I n child speech, the dative contexts appear quite early but usually only after the genitive has been acquired. Such early occurrences are primarily strictly limited to forms of the names o f family members w i t h the child himself included, and less frequently, the names o f some animals or animal toys occur. N o doubt the cause of the presently discussed generalization is the homonymy o f the genitive and the dative endings i n one class o f feminine nouns, i.e. stems ending i n a soft consonant. This set o f nouns includes frequently used nouns such as mamusia ' m o m m y ' , babcia 'granny' or the child's name, e.g. Basia and Kasia. The 11

unlike Russian, where the notion of 'movement towards' is expressed by the preposition k + DAT, in Polish it is expressed by means of do + G E N , which accounts for the absence of dative contexts for inanimate nouns. 1

6.

The A c q u i s i t i o n of Polish

627

homonymous genitive-equals-dative forms for these examples are: mamusi, babci, Basi, and Kasi. Because o f the very restricted use o f the dative, this homonymy is not counterbalanced by a sufficient number o f other forms where the genitive is not identical w i t h the dative. Some o f these counterexamples which do occur i n the data include the f o l l o w i n g : from the nominative mama ' m o m ' there is the dative mamie and the genitive mamy, from N O M tatus 'daddy' D A T tatusiowi versus G E N tatusia, from N O M mis 'teddy-bear' D A T misiowi versus G E N misia, etc. The children tend to use the genitive f o r m i n dative contexts, e.g. misia not misiowi. One o f the three children studied, Jas, d i d not reach the level o f m i x i n g the two cases. Probably his o w n name, w h i c h takes the dative form Jasiowi as different from the genitive Jasia, and which was used very frequently i n dative contexts, helped to avoid this confusion. However, there was a short period (from l ; 1 0 t o l ; l l ) when the -/ ending became generalized and was used also w i t h some nouns requiring -e or -owi. The examples were not numerous: (1) mami for dative, cf. correct mamie, versus the genitive mamy ' m o m ' , (2) babi for dative, cf. correct babie, versus the genitive baby 'grandma', and (3) pieski for dative cf. correctpieskowi, versus the genitivepieska 'doggie'. One month later the correct distribution was established. This example, however, points to the importance o f the homonymous -i ending. Z a r § b i n a (1965) quotes similar examples from her daughter's speech i n the last months o f the second year. They are followed, after the age o f 2, by several examples o f reversed substitution, where the dative is used instead o f the genitive in the sense o f possession, e.g. tatusiowi jest lekarstwo 'to:Daddy is medi­ cine'—quite unusual i n Polish but typical for Latin (e.g. Mihi est liber) or Hebrew (cf. Berman, 1985). Similar examples o f the possessive dative are given in Siatkowscy (1956), e.g. ksiqzka tatusiowi 'book to:Daddy' instead o f the correct ksiqzka tatusia 'book o f . D a d d y ' . The Siatkowscy examples were col­ lected during the period from 2;0 to 2;6. I n m y data such errors are only sporadic. The homonymy alone cannot adequately account for these substitutions since it is fairly frequent i n the whole declensional system and still never leads to such results. What seems necessary for this homonymy to result i n the m i x i n g o f the two cases is the close semantic p r o x i m i t y o f the possessive genitive and the dative cases, the only difference between the Possessor and the Receiver being that o f the stative versus dynamic notions. The cases are occasionally m i x e d i n locative phrases when a directional expression is used instead o f a stative one (or vice versa), e.g. 'to the tram' instead of ' i n the t r a m ' , or 'at the shop' instead o f 'to the shop'. Such examples occur sporadically i n m y data, and similar ones have been cited by Zar^bina (1965). 4.3.2.2. Genitive Plural -6w Instead of 0 Ending in Feminine and Neuter Nouns (Grammatical Gender Criterion). This overextension represents the only obvious instance o f " i n f l e c t i o n a l i m p e r i a l i s m " i n the acquisition o f Polish.

628

Smoczyhska

The adult rule is quite simple, i.e. i n the genitive plural masculine nouns have the ending -6w while feminine and neuter nouns have a zero ending. I n addition to this regular pattern, some masculine nouns take an irregular -yl-i e n d i n g . The majority o f the children studied used the -6w ending irrespective o f the noun gender. T y p i c a l l y , this ending appeared i n the first contexts for genitive plural (about the age o f 2) and was persistently and exclusively used for differing periods o f time, for instance, t i l l 3;0 for Kasia, 3;4 for M i c h a l , 3;9 for Jas, and 4;6 for B a s i a . This is a typical "late e r r o r " since all the remaining R E G U L A R case forms are mastered by the age o f 2;6 or before. I t is also worth noting that i n the Russian data for Zhenya, the corresponding -ov ending persisted much longer than other " i m p e r i a l i s t i c " endings, w h i c h were no longer overextended after the age o f 2;9 (Gvozdev, 1949, v o l . 2, p. 87). 12

13

There are five factors which should be taken into account as possibly respon­ sible for the "inflectional i m p e r a l i s m " o f -6w. First o f a l l , the zero ending i n feminine and neuter nouns contradicts the principle o f overt marking (see Slobin's (1973) Operating Principle E ) . This argument, however, does not pro­ vide a sufficient explanation o f the persistence o f this error. Furthermore, this argument does not explain w h y other -0 endings are N O T replaced by overt markers w i t h similar consistency, e.g. the accusative singular o f inanimate nouns or the second person imperative for the verb. Secondly, the form w i t h the zero ending is one (or even t w o ) syllable(s) shorter than the remaining forms o f the noun, e.g. nominative singular lampa ' l a m p ' has the genitive plural form lamp and the instrumental plural lampami. The significance o f this factor is supported by the finding that diminutive forms, e.g. lamp-(e)k-0 T a m p - D I M - G E N : P L ' which for phonetic reasons acquires an additional syllable w i t h the insertion o f /e/ between the stem and the -k suffix usually appear earlier than bare stem forms. Hence the child w h o uses the correct form lamp(e)k w i l l continue to say lamp-dw instead o f lamp. I t is possible that the child conceives /ek/ as a k i n d o f ending for genitive plural. W h i l e the first and second factors are concerned w i t h the form o f the rule, the third and fourth factors refer to the criteria o f rule application. T h i r d l y , i n case o f genitive plural, the criterion o f grammatical gender distinc­ tion between feminine and neuter nouns on the one hand and masculine on the other is i n t w o ways inconsistent from the point o f view o f the declension system. First, i n plural the distinction between masculine/feminine/neuter, relevant for singular, is abolished to be replaced by the v i r i l e / n o n v i r i l e gender distinction. W h i l e this is true for adult language, i n early child language this conflict o f double gender distinction does not actually occur because the v i r i l e / n o n v i r i l e The -yl-i ending occurs also with some irregular feminine nouns. This fact will not be taken into account here, as these nouns were not found to occur in the data. After this age, -6w continued to be overextended to replace the irregular -yl-i of masculine nouns, but this phenomenon is similar to those discussed in Section 4.3.1. There were also sporadic instances of its use with feminine and neuter nouns but the correct use was much more frequent. 12

13

6.

The A c q u i s i t i o n of Polish

629

distinction w i t h all its implications is acquired long after the appearance o f the genitive plural forms (see Section 6 . 3 . 1 ) . Moreover, i n singular, masculine and neuter nouns form the nonfeminine gender as opposed to the feminine, while i n the case o f genitive plural it is the neuter together w i t h the feminine that are opposed to the masculine. This dis­ proportion justifies the prediction o f special difficulties w i t h neuter nouns, w h i c h w o u l d therefore be likely to behave as masculine nouns, rather than take the same ending as feminine nouns. This prediction is confirmed i n the data o f Basia, who used -6w as the only genitive plural ending until the age o f 4;6, w h i l e later, from 4;6 to 5, she applied i t only to neuter nouns (apart from its correct use w i t h masculine nouns), w i t h feminine nouns having the correct 0 ending. Thus, the factor seems to be significant, although not from the very beginning o f the acquisition o f genitival f o r m . 1 4

The fourth factor concerns the fact that the basic system o f the plural declen­ sion as elaborated by small children seems to be much simpler than that o f adults. There is a clear tendency for one-to-one correspondence between case and end­ ing w i t h no gender distinction at a l l , i.e. nominative has t w o allomorphs -il-y, accusative equals n o m i n a t i v e , dative takes -om, instrumental -ami and lo­ cative -ach. This apparent regularity o f the system can contribute to the child's expectation that the genitive also has a unique form. 15

The fifth and final factor concerns the role o f agreement. Agreement forms can also be o f some importance for establishing gender distinctions. Table 6.6 shows the neutralization o f gender i n plural, where all the nouns except [ + virile] take the same agreement forms i n both adjective and verb. The importance o f this factor was revealed to me by a strange metalinguistic problem set forth by one o f the subjects studied, M i s (3;4): Mamusiu, piatek to on? ' M o m m y , is ptatek (petal) a he?' Piatek, jak jeden, to on? 'Piatek, when there is one, is it a he?'. Mother: Tak. ' Y e s . ' M i s : A dwa? ' A n d ( i f there are) t w o ( o f them)?' Mother: To platki. 'Then (they are) platki' M i s : A sto? ' A n d a hundred ( o f them)?' Mother: To platkow 'Then—platkow'. Rather than the form itself, the child wanted to learn the grammatical gender. M i s ' s use o f 'he' and 'she' was a technique that he employed as early as 2;11 to ask questions about gender. What is interesting here is the occurrence o f t w o separate questions, i.e. 'What is the gender of piatek i n singular?,' and then 'What is its gender i n plural?'. I t could mean that M i s constructed a hypothesis on the metalinguistic level concerning the change o f gender i n plural. He could have noticed that piatek, w h i c h i n singular behaved like a boy, undergoes a sharp transformation i n plural and begins to behave like There are, however, numerous early examples of the correct nominative plural -a with neuter nouns, which could be treated as evidence that the tripartite gender distinction can work in plural as well. However, the nominative form is somewhat more basic than those of oblique cases, and as such it can be learned more easily. Children use mostly inanimate plural objects, which are nonvirile. The use of virile objects, where accusative equals genitive, comes much later. 14

15

Smoczyhska

630

TABLE 6.6 Gender A g r e e m e n t in Adjectives and Past Tense Verbs in Singular and Plural Singular Gender MASC MASC FEM FEM NEUT '

Plural

Noun

'was'

'pretty'

Gender

chlopiec 'boy' kotek 'cat' matka 'mother' zabawka 'toy' jablko 'apple'

byl by! byla byla bylo

ladny ladny ladna ladna ladne

VIR NVIR NVIR NVIR NVIR

Nouns chlopcy kotki matki zabawki jablka

'were'

'pretty'

byli byly byly byly byly

ladni ladne ladne ladne ladne

girls, and thus it takes the adjective and verb forms proper to female human beings. I t could be then that children perceive the non-virile gender simply as a kind o f feminine which is due to their tendency to refer to natural sex differences in their metalinguistic reasoning (cf. Karmiloff-Smith, 1979). The combination o f all these factors can provide an adequate explanation o f the phenomenon described above. What I found to be most difficult to account for is the case o f three children among m y subjects who had no problems w i t h the genitive plural. W i t h t w o o f them, the distribution o f 0 and -6w was correct from the very beginning (2;0 to 2;2). The third one, Inka, seemed to avoid zero forms until 2;5 using -6w exclusively and only w i t h masculine nouns. Next, from 2;5 to 3;3 she used the 0 ending but only w i t h diminutive nouns (when this involves Id-insertion). Eventually after the age o f 3;3, she also used bare stem forms. Such a stategy o f avoiding difficulties is interesting i n itself, but it goes beyond the problem presently under discussion. When trying to account for the data from these children, I would claim that the first t w o factors can also be the very cause of the E A R L Y acquisition o f zero forms. W o r d forms which do not have any overt marker and are shorter than usual seem strange to all children. This peculiarity can be rejected by some children which results i n replacing the zero ending w i t h an overt marker. I n contrast other children w o u l d notice the zero ending just because o f the perceptual salience o f the lack o f an ending, and they w o u l d accept it resulting i n the establishment o f distributional criteria. 4.3.2.3. Accusative Singular of Masculine Nouns: Overextension of -a (the Animate I Inanimate Criterion). I n the accusative singular, masculine nouns subdivide into animate and inanimate, w i t h the former ending in -a and ac­ cusative equal to genitive and the latter having the 0-ending and accusative equal to nominative. The distribution o f forms is identical to the parallel distribution i n Russian, where it has been reported as a late acquisition by Gvozdev (1949) and Slobin (1966). Polish data are unclear in this resepct. The picture is obscured by the interference o f another factor. I n contemporary spoken Polish, there is a strong tendency to use the -a ending also w i t h inanimate nouns i n some limited

6.

The A c q u i s i t i o n of Polish

631

but not very well-defined contexts. So far, the research carried out by Polish linguists has not led to any clear conclusions concerning either the kinds o f words which are likely to take the -a ending, or typical contexts i n which it occurs. Hence, no adequate explanation o f the phenomenon has been provided. On the one hand, Polish children seem to acquire the -a versus 0 distinction very early (about 2;0), but at the same time, they use the -a ending w i t h inani­ mate nouns to a much larger extent than adults do, although some contexts prove that they k n o w the distinction. I t is difficult to state anything definite without clear information on adult usage. What seems to be the case here is the overex­ tension o f some adult rule rather than the simple overgeneralization o f the -a ending. T y p i c a l l y , child language more explicitly reflects a change that the language undergoes at a given time (cf. Slobin, 1977). 4.3.2.4. Locative Case of Nonfeminine Nouns: Overextension of the Ending -u (Morphophonological Criterion). The distribution o f the t w o alternative endings -u and -e depends on the character o f the stem final consonant. Further­ more, while adding -u does not require any transformation o f the stem, the -e ending involves softening o f the stem final consonant, e.g. the shift from [ v ] to [ v ' ] i n drzew-o vs. na drzewi-e 'tree'. W h e n there is no palatal counterpart o f the consonant, i t is replaced by another consonant according to a clearcut pattern, e.g. rower vs. na rowerze ' b i k e ' ; zeszyt vs. w zeszycie 'copybook', w i t h Ixl - Izl and HI - 161 alternations. I n some children one can find a dozen or so examples o f overextended -w, e.g. na roweru, w zeszytu, but the -u ending is never found to predominate to such an extent as to drive out the -e ending completely; cf. also examples i n W o j t o w i c z (1959). This overextension results i n the avoidance o f stem alternations. Further­ more, i n some children, the vocative ending -u undergoes an analogous overex­ tension. Russian has similar endings but their distribution differs from that o f Polish, and i n view o f differences i n the phonological system, no similar stem alternations are involved. I n fact Gvozdev (1949) finds a reverse tendency to overextend -e at the cost o f -u. 4 . 4 . Idiosyncratic

Rules

Idiosyncratic rules are somewhat different than overgeneralizations. Idiosyn­ cratic rules involve a simplified procedure invented by the child i n order to cope w i t h a diversity o f rules. This is not identical to overextending a rule that exists i n the adult language. O n the other hand, an idiosyncratic rule is usually related to one o f the existing rules, or else it can be based on a single linguistic form which serves as a model for overgeneralization. 4 . 4 . 1 . Present Tense: First-Person Form Equals Third-Person Plus - m . A m o n g present-tense forms the first to emerge is the third-person f o r m . W h e n shifting to the use o f first singular for self-reference, children must establish the

632

Smoczyñska TABLE 6.7 Present Tense Forms in S i n g u l a r Conjugation I

Person 1 2 3

pisz-e pisz-e-sz pisz-e 'write'

bior -§ bierz-e-sz bierz-e 'take'

Conjugation II (-Í-) rob-i-§ rob-i-sz rob-i 'do/make'

Conjugation III (-a-)

widz -£ widz'-i-sz widz'-i 'see'

czyt-a-m czyt-a-sz czyt-a 'read'

distribution o f the t w o endings -§ and -m. Table 6.7 presents the distribution o f these endings. M o s t o f the children studied acquired these rules quickly and made no consistent errors. One boy (Jas), however, used a simplified rule o f adding -m to the third-person form i n all verbs, e.g. pisze-m, bierze-m, robi-m and the correct czyta-m. I t is worth noting that this invention made i t possible for the child to avoid specific options related to the application o f i.e. replacing the stem vowel by -§ (as i n pisz-e Ipisz-e) versus adding the ending to i t (as i n rob-ilrob-i-e), or finally stem alternations (e.g. bierz-1 bior- widz/widz-). Jas applied the rule from the age o f 2;0 to 3;6. A slightly different rule w h i c h involves adding -em was used by the child reported i n Siatkowscy (1956) from 2;0 to 2;6. }

4.4.2. Sentence External Negation: nie + Sentence. M o s t Polish children do not have any difficulties w i t h sentence internal placement o f the particle nie. I n general, the negation system is mastered early. One child (Basia), however, displayed the strategy o f N E G + S c o m m o n l y used by English-speaking children (see B e l l u g i , 1967). This case was reported i n Smoczyhska (1978a), where i t was pointed out that Basia's overall syntactic development was much slower than that o f the other four children studied. Basia used this rule i n her early negated sentences, e.g. Nie Basia spi 'not Basia sleeps'. The context clearly indicated that such utterances involved sentence negation rather than the anaphoric nie + positive statement, w h i c h is contrary to B l o o m ' s (1970) claim. Furthermore the negations were used not only for rejection, as claimed by de Villiers and de Villiers (1979), but also i n declarative negative statements and questions, e.g. Nie tu brudno? 'not here dirty?' ( = ' I t ' s not dirty here?'). Later o n , Basia continued to place nie i n sentence-initial position but moved the verb to the second position, w h i c h results in correct sentences, e.g. Nie spi Basia ' N o t sleeps Basia'. Eventually, she arrived at sentence-internal placement o f nie, e.g. Basia nie spi 'Basia not sleeps' ( = Basia isn't sleeping'). O n the other hand, Kaczmarek (1953) gives several examples o f placing nie after the verb, e.g. (1;6) placze nie 'cries not'; (1;8) mamusia kopac bqdzie nie ' M o m m y d i g w i l l n o t ' . Similar examples occur sporadically i n Kasia's early negations.

6.

T h e A c q u i s i t i o n of Polish

633

Children learning English introduce multiple negation into their language. I f this is a general tendency for children, Polish ( w i t h its multiple negation) should be an ideal language to learn. Yet this is not the case. W h i l e American children reconstruct the multiple negation, nonexistent i n their language, Polish children do exactly the opposite. W i t h Kasia and occasionally also w i t h other children, the use o f a negative pronoun involved the deletion o f nie. This deletion resulted in literal counterparts o f English sentences o f the k i n d / will say nothing, Nic powiem, 'nothing I:will:say' while the correct Polish form is Nic nie powiem. Moreover, when the particle nie was used, the pronoun was not converted into its negative f o r m , e.g. Kasia (2;3) said, A Jezus to zawsze nie gryzie ' B u t Jesus always not bites' where nigdy 'never' should be used instead o f zawsze 'always'. Rz^tkowska (1980) reported the use o f emphatic wcale 'at a l l ' as a negative marker, e.g. (1;10) Jadziusia wcale placze 'Jadwiga at all cries'. 1 6

4.4.3. Syllabic Prepositions and Prefixes Realized as a Reduplication of the First Syllable of the Word. This individual strategy was used by m y o w n t w i n sons whose early speech was systematically recorded (see Section 2 . 1 . 2 ) . The twins began to combine words at the age o f 2;4. A t the beginning, they omitted all prepositions, using only appropriate case endings. A t the age o f 2;7 one o f them, M i c h a l S., began to produce strange forms o f nouns w i t h reduplicated first syllable, e g . ma-mamy, (s)to-tole or kup-kupkach. The context, as w e l l as case endings, clearly indicated that a prepositional phrase was intended, such as do mamy 'to M o m ' na stole 'on table' and po kupkach, used in Nie wolno chodzic po kupkach '(One is) not allowed to:walk on (cow) excrements'. Immediately after the emergence o f these forms, I tried to make the child repeat various kinds of prepositional phrases. This procedure revealed that the reduplication tech­ nique was consistently used to replace syllabic prepositions, while the asyllabic ones were continually omitted, e.g. the phrase z mamq ' w i t h M o m m y ' was produced mamq w i t h the preposition z deleted and w lesie ' i n forest' was pro­ duced as lesie. Later o n , the reduplication rule was extended to include syllabic verbal prefixes indicating perfectivity and/or modulation o f verb meaning, e.g. je-jechat meant either po-jechal ' w e n t ' , wy-jechat ' l e f t ' , or przy-jechal ' a r r i v e d ' . The same procedure began to be used by the other t w i n , Maciek, 3 months later and it persisted i n both o f them for several months. 1 7

18

It is more than probable that the rule was based on a single instance, namely the frequently used phrase do domu 'to h o m e ' , i n which the phonetic form o f the Contrary to the English some I any I no tripartite distinction, Polish has the binary opposition, e.g. ktoslnikt 'somebody/nobody'. From informal sources I have learned about three other cases of applying such a rule. Unfortu­ nately, data were not collected. I noted one case of erroneous morphological analysis, in which the initial stem syllable of the word wysoki 'high' was treated as a prefix, which resulted in so-soki (cf. the prefix wy- in wyjechal, above). 16

17

1 8

634

Smoczyhska

preposition happens to be the same as that o f the first syllable i n domu. Since the rule was applied only i n case o f syllabic morphemes, i t cannot be treated as a strictly grammatical rule, but rather as a procedure used to render an approximate prosodic shape o f the prefixed f o r m . There was no attempt to differentiate specific semantic content o f various prepositions and prefixes. A l l received exactly the same treatment. 4.4.4. First-Person Plural of the Past Tense: Total or Partial Identification of the Movable Ending with the Personal Pronoun. First o f a l l , it should be said that the first-person plural forms o f the past tense appear relatively late (usually after 2;6) and due to the lack o f appropriate context, they are rarely used. When referring to past actions carried out together w i t h other persons, young children are more likely to use a singular subject and a comitative, w h i c h is acquired relatively early. Past forms o f the first-person plural begin to be used more frequently about the age o f 5 i n the case o f only children, and somewhat earlier i f the child has siblings or attends a nursery school. The Polish enclitic and movable ending -smy, w h i c h is the unique marker for first-person plural o f the past tense i n all verbs, appears to cause specific difficul­ ties for all children. As i t happens, the phonetic shape o f -smy partially coincides w i t h that o f the respective personal pronoun my ' w e ' . As a consequence o f this coincidence, t w o kinds o f erroneous interpretation appear i n the data. The first one consists i n the identification o f my and -smy. The latter is considered as an allomorph o f the former. Children correctly use -smy i n sentences w h i c h do not contain a pronoun, e.g. Wczoraj poszli-smy 'yesterday went-1 P L ' . I n contrast, when the pronoun is present, they o m i t the ending and say; Wczoraj my poszli, cf. correct Wczoraj my poszli-smy or the more natural, Wczoraj my-smy poszli. Another interpretation is based on an erroneous analysis o f the morpheme -smy into t w o morphemes, i.e. the ending -s and the pronoun my. Here again sentences without pronouns seem to be correct, but for children the form poszlismy consists o f three morphemes (PP + -s + PRO) rather than t w o (PP + -smy) as i n adult language. This can be seen i n such utterances as My poszli-s or My-s poszli. I n the latter example, the idiosyncratic -s ending is given the status o f a movable suffix. For some children, such forms as poszli (first interpretation) orposzli-s (second interpretation) appear also i n sentences w i t h no pronoun used. This indicates that children can correctly apply the pronoun ellipsis rules, but they delete the pronoun from the suffix -smy where i t is supposed to be placed. W i t h some children, examples o f overmarking can be found, e.g. My-smy poszlismy. Finally, evidence for the full control o f the adult form is found i n utterances such as My-smy poszli i n w h i c h both the pronoun and the ending are present. I n other cases, the erroneous rule can result i n apparently correct forms. I n general, the k i n d o f error presently under discussion is typically a late error. The omission o f -smy w h i c h results i n forms like My poszli can also be found i n the substandard Polish o f adults w i t h a l o w educational level. The role

6.

The A c q u i s i t i o n of Polish

635

of incorrect input is evident i n the case o f Inka, who (at some period) used correct forms when referring to actions carried out together w i t h her mother, but who used the my + P A R T I C I P L E form when reporting what she was doing w i t h the children i n the yard. It w o u l d be interesting to compare the acquisition o f the ending -smy w i t h that of -scie for second person plural which does not coincide phonetically w i t h the corresponding pronoun wy. Unfortunately, contexts involving addressing several persons are even more rare than those referring to j o i n t l y performed past actions, and the second-person plural past-tense form is extremely infrequent i n the data. The forms found i n the data occurred late and were correct except for one case o f overmarking i n W i t e k (4;5) Wy-scie przebierall-scie si$ 'you-2PL changed-2PL R E F L ' ( = ' y o u changed yourselves'). From the theoretical point o f v i e w , these errors provide evidence for the significance o f prosodic features i n establishing what is a w o r d and what is not. Such errors w o u l d never occur i f -smy were not enclitic. A contrastive test-case is provided by the present-tense first-person plural ending -my, which is phonet­ ically identical w i t h the pronoun my, e.g. czyta-my ' r e a d - I P L ' . However, i t is neither enclitic nor movable, cf. the stress pattern o f czytdmy w i t h that o f czytdlismy. Conseqeuntly, children never try to identify this ending w i t h the pronoun and use correct forms from the very beginning, w i t h no errors at a l l . 4.4.5. An Idiosyncratic Causative Construction. This is also an individual case. When he was 2 ; 1 1 , m y youngest son, Wawrzont, invented a syntactic construction w h i c h does not exist i n Polish. He used the verb karmic 'to feed' i n an overextended causative sense, e.g. Mama mnie pokarmi umyc rqczki ' M o m m y me will:feed to:wash hands'. He meant that he wanted M o m m y to wash his hands instead o f h i m doing i t by himself. Other examples involved 'feeding h i m ' to eat something although the meaning o f 'eat' is normally i n ­ cluded i n that o f 'feed', i.e. 'make eat'; 'feeding h i m ' to put on his clothes; 'feeding h i m ' to wee-wee; or even 'feeding h i m ' to draw a dog after he declared he was unable to do i t . I n all these instances, a simple transitive sentence w o u l d have been quite appropriate, e.g., ' M o m m y w i l l wash m y hands' or 'draw *a dog for m e ' . What the child intended to convey by his causative constructions was a special emphasis on his refusal to do something by himself, that he was i n fact able to. Hence these constructions had a definite pragmatic sense w h i c h was narrower than that o f the English make + V E R B structure. This pragmatic sense emerged i n response to a specific situation. I was very busy for several weeks and although I was present at home, other people had to take care o f the child. The origin o f this construction can be found i n a concrete proportion, i.e. T am eating by m y s e l f : ' M o m m y is feeding m e . ' The w o r d 'feed' became a causative auxiliary. A n especially interesting aspect o f this construction is that i t represents the 'accusativus c u m i n f i n i t i v o ' type w h i c h is practically nonexistent i n Polish. I n other words W a w r z o n t added an infinitival complement to the accusative

636

Smoczyhska

object normally governed by karmic. This construction was strongly productive for a limited period o f time, and disappeared simultaneously w i t h the disap­ pearance o f the appropriate context, that is, as soon as I finished m y w o r k . 4.5. Free

Variation

This k i n d o f error occurs when the child has acquired alternative forms w i t h ­ out learning specific criteria o f their distribution, w h i c h are either irregular or too complex or inconsistent w i t h his present system. 4 . 5 . 1 . The Choice of Present (PRES) vs. Infinitive (INF) Stem Form of the Verb. Establishing w h i c h verb stem should be used w i t h a given form takes over t w o years. F r o m 2 to 4 , the erroneous choice o f the stem is clearly the most frequent error w i t h all children. As noticed i n Zarebina (1965), the number o f such errors increases w i t h age i n proportion to the growth o f vocabulary. F r o m 4;0 on, the number o f errors diminishes, but those i n v o l v i n g especially c o m p l i ­ cated kinds o f verbs persist t i l l 6;0 and over. Considerable complexity, and also irregularity o f the adult system i n this respect has been discussed i n Section 1. W h i l e i n some verbs the phonetic realization o f the stem does not change, e.g. czyta/czytac 'read' (3SG:PRES/INF), others require complex transformations, e.g. rysujelrysowac ' d r a w ' , pierzelprac ' w a s h ' , and wezmielwziqc 'take'. According to Brenstiern-Pfanhauser (1930), the child displays a tendency towards building all verb forms from the stem that he learned first. O n the other hand, Z a r § b i n a (1965) claims that this is true only for the early stages. After the child has learned both stems, he builds the given form i n an ad hoc way using one or the other stem. Thus one can observe t w o types o f errors occurring simul­ taneously: imperfective present tense or perfective future tense forms created from infinitive stems and past tense forms created from present tense stems. For instance, the present tense form rysowa '(he):draws' is built on the infinitive stem (rysowa-c ' d r a w - I N F ' ) instead o f the correct rysuje. For the verb prac 'to:wash' we find on the one hand the past tense form pierzy-I '(he):washed' obviously based on the present tense pierz-e '(he):washes' instead o f the correct pra-l (cf. infinitive pra-c). O n the other hand, however, present tense forms happen to be based on the infinitive, e.g. pra or praje instead o f pierze. As shown by the following examples from Inka, when coping w i t h such problems, children do not avail themselves o f the correct model, and even i f they do, it is only to produce an ad hoc correct form; soon afterwards they revert to the old one. A t 2;7 Inka asked her father: Co to rysowa-sz tatusiu? ' W h a t are you drawing, Dad?' (cf. correct rysuj-esz). Father: Nie rysuj-%, tylkopisz-%. T am not drawing, but I am w r i t i n g ' . Inka: Co pisa-sz? ' W h a t are you w r i t i n g ? ' (cf. correct pisz-esz). A t 2;4 Inka's grandmother said: A babcia ci cos przywioz-I-a 'Granny has brought you something'. Inka: Co przywioz-I-a-s babciu? ' W h a t

6.

T h e A c q u i s i t i o n of Polish

637

have y o u brought, Granny?' Kolejkg przywiezia-l-a-s mi? ' A train d i d you bring me' (cf. correct przywiozl-a-s). Early examples o f such errors have been reported i n all diary studies, and those for later periods can be found i n Siatkowscy (1956) and W o j t o w i c z (1959). Gvozdev (1949) has presented a detailed discussion o f similar errors i n Russian, and Guillaume (1927) has done the same for French. The phenomenon should be investigated systematically i n order to establish w h i c h kinds o f verbs are es­ pecially difficult, and to look for some rules w h i c h could underlie this overall picture o f free variation. I n all these cases, the verb is clearly marked for tense and aspect although the given realization o f these categories does not correspond to the adult norm. F r o m the theoretical point o f v i e w , however, this phe­ nomenon—in spite o f its wide range—is o f relatively little interest, as the irregularity o f the system ( i f i t can be considered a system at all) is immense. 4.5.2. Cardinal Numeral 'Two': Feminine Form dwie Versus Nonfeminine dwa. U n l i k e the errors discussed above, the occurrence o f a free variation pattern i n this case is extremely interesting. The cardinal numeral ' t w o ' when used w i t h reference to nonvirile n o u n s has t w o distinct forms: dwie for femi­ nine nouns and dwa for masculine and neuter, where the latter form is unmarked (e.g. used while counting: jeden, dwa, trzy . . . ) . Most o f the children studied had problems establishing the correct distribution o f these forms i n spite o f the fact that it is determined by a very simple rule w h i c h is based on grammatical gender, and gender is a category w h i c h is acquired a long time before numerals come i n use. I n i t i a l l y , children use the unmarked form dwa regardless o f gender. Later o n , both forms are found to occur i n free variation. 19

For Jas, this error persisted even i n the latest recordings (6;0). His case is especially interesting. Jas was very interested i n numbers, and numerals were unusually frequent i n his s p e e c h . N o t only d i d he frequently speak about thousands and m i l l i o n s , but even his nostrils were numbered and received specif­ ic labels as 'the first one' and 'the second one'. His data provide obvious support for the claim that training does not help to acquire even such a simple rule i f it happens to be blocked for some mysterious reasons. W h e n trying to find some rule governing the distribution o f dwa/dwie, I was able to analyze more than 200 examples o f ' t w o ' + N O U N constructions recorded from 2 to 6. S t i l l , I failed to find any rule. Clearly, both forms were used i n free variation w i t h a tendency to use the same form at a given moment, e.g. at (3;5) he said dwi-e bramk-i i dwi-e domk-i ' t w o - F E M gate-FEM:PL and t w o - F E M h o u s e - M A S C : P L ' ( = ' t w o gates and t w o houses') cf. correct dwie bramki i dwa domki. After he was corrected by 20

Virile nouns involve complicated rules of numeral agreement, which will not be discussed here. T h e data were collected more than 20 years ago, and it might be of some interest to add that those early interests prevailed, since he is a statistician by now. 19

20

638

Smoczyhska

his mother ('One doesn't say dwie domki but dwa domkV), he corrected himself w i t h dwa domki i dwa bramki, w i t h the incorrect form o f ' t w o ' for bramki this time. As adults started correcting his utterances, he began analyzing the problem on the metalinguistic level. The only result o f this was " c o r r e c t i n g " adults, that is, providing erroneous forms w i t h the correct argumentation concerning gender, e.g. ' A u n t i e , you should say dwi-e numerk-i ' t w o - F E M n u m b e r - M A S C : P L ' because one says ten numerek ' t h i s : M A S C : S G n u m b e r : M A S C : S G ' . He also began to produce both forms at the same time, leaving the option to the listener, e.g. Znalazlem dw-a czy dwi-e szmatk-i ' L f o u n d t w o - M A S C or t w o - F E M ragF E M : P L ' ( = T found t w o or t w o rags'). For other children, this error d i d not persist for such a long time, but still it could be considered as a typical late error, for instance, it appears i n the Baudouin data. I am unable to provide an adequate explanation for this error although three factors can be demonstrated to be o f some importance. Firstly, among plural cardinal numerals only ' t w o ' displays such a gender distinction i n addition to the opposition o f v i r i l e / n o n v i r i l e forms; 'three' and 'four' having only the latter differentiation. I f this factor were the only one at play, one w o u l d expect the children to ignore the distinction dwa/dwie and use only dwa i n order to preserve paradigmatic relations among numerals. I n fact, this distribution does occur, but only i n the early stages. Secondly, the nonvirile nominative plural form o f adjectives ends i n -e, e.g. du'z-e klock-i, pilk-i, ciastk-a ' b i g - N V I R : N O M : P L , b l o c k - N V I R : M A S C : P L , b a l l - N V I R : F E M : P L , b i s c u i t - N V I R : N E U T : P L ' . I f the child were to treat dw-a/dwi-e as a k i n d o f adjectival form (by analogy to jeden-0/jedn-a/jedn-o 'one', which takes regular adjective declension forms), he w o u l d be expected to use the form dwie exclusively for all nonvirile nouns—a prediction that is not supported by the data. Finally, it could be hypothesized that children simply refuse to take into account the masculine/feminine/neuter distinction when using plural forms. This w o u l d be congruent w i t h one o f the explanations offered i n connection w i t h their generalizing o f the genitive plural -6w ending (see Section 4 . 3 . 2 . 2 ) Note, however, that the distribution o f ' t w o ' forms is consistent w i t h the main distinction w i t h i n the singular (feminine vs. nonfeminine), whereas that of -6w/0 is not. Yet i n the case o f Jas, the latter rule is acquired E A R L I E R than the former. That is w h y I am unable to produce any definite explanation. 21

4.5.3. Using Masculine or Feminine Forms of Adjectives and Past-Tense Verbs in Free Variation. This k i n d o f error differs considerably from those described above. The rules are neither too complex nor incompatible w i t h the system. On the contrary, they are simple and consistent, but the child simply does not know them yet. The occurrence o f these errors is limited to the earliest Children, however, are perfectly aware of the gender of the noun in singular, e.g. dw-a srub-y, jedn-a . . . drug-a . . . 'two-MASC screws-NVIR, one-FEM . . . other-FEM'. 21

6.

T h e A c q u i s i t i o n of Polish

639

instances o f using adjectives and past tense forms, e.g. tatus mii-a 'Daddy niceF E M ' , babcia poszedl-0 'Granny went-3SG:MASC\ A t the very beginning, mila or poszedl can be simply the only form o f the w o r d that the child knows. These errors disappear completely long before the age o f 2 (see Section 6.1.1). 4.6. Partial

Application

of a Rule

These errors reveal difficulties i n the complete application o f a complex rule which involves performing several operations. The problem does not consist o f learning the criteria, but rather i t concerns technical difficulties i n producing a correct form. 4 . 6 . 1 . Nouns: Correct Ending Added to an Incorrect Stem Form. Errors o f this type occur infrequently i n all children. The basic noun form is nominative singular. The child tends to build other forms from the nominative form. H o w e v ­ er, there are stem alternations i n the nominative singular, and oblique cases have a different stem. I n Polish there is a phonological rule according to w h i c h the opposition o f voiced/voiceless becomes neutralized i n w o r d final position, e.g. /%lep/ versus /%leba/ 'bread' for nominative versus genitive but /%uop/ : /%uopa/ 'peasant'. I n order to know whether the final voiceless consonant i n the nominative form alternates w i t h its voiced counterpart, the child must memorize the form that a given stem takes i n oblique cases. The phonological rule is acquired early and errors such as chlepa instead o f chleba typically occur very early. W h e n this k i n d o f error occurs at a later stage o f development, it occurs when the child has just heard the nominative form for the first time. Another rule requires that -e- be inserted between consonants i n order to avoid an unacceptable word-final cluster. I t appears typically i n diminutive masculine nouns w i t h the -k suffix, e.g. kot-k-a ' c a t - D I M - G E N ' but kot-(e)k-0 ' c a t - D I M N O M ' . Because o f the abundance o f diminutive forms, this rule is acquired very early, and children normally delete the inserted -e- when producing oblique case forms. Errors such as kotek-a appear only i n the earliest period. Other forms w i t h the -e- inserted into the stem itself prove more difficult especially i f the stem is relatively short (one or t w o syllables), e.g. the genitive o f the w o r d lew /lef/ ' l i o n ' is frequently realized as lew-a (cf. correct Iw-a) or even lef-a i f the unvoiced If I is preserved and ogien ' f i r e ' results i n ogienia (cf. correct ognia). Sometimes stem-internal consonants and/or vowels alternate following defi­ nite alternation patterns. What seems to be most difficult for the child is the case when several phonemes alternate according to distinct patterns. The f o l l o w i n g example w i l l make this clear. M i c h a i (2;4) wanted to say ' G o o d morning, eagle!' Dzien dobry, oriel, where the vocative orle constrasts w i t h the nominative orzel He began w i t h Dzien dobry, orzele I oleic I then corrected himself w i t h orzle lollel, and eventually, he gave up and used the nominative orzelloleul instead. Producing the correct version required three transformations o f the nominative stem form, i.e. lul to III, deletion o f lei, and 111 to III. The sequence o f forms

640

Smoczyhska

that he produced demonstrates that he was able to perform one transformation at a time. Although he failed to perform the third one, the fact that he reverted to the more safe nominative form (also acceptable i n the vocative function) proves that he was not satisfied w i t h the form he arrived at. W o j t o w i c z (1959) noted that errors o f preserving the nominative singular stem i n oblique cases are much more frequent than those i n v o l v i n g the reverse process, e.g. an incorrect nominative singular f o r m ojc 'father', based on the genitive singular ojc-a, (cf. correct nominative singular ojciec). 4.6.2. Preserving Case Forms in Negated Sentences. The rule o f convert­ ing the accusative object o f transitive sentences into genitive when the predicate is negated is likely to create difficulties at least for some children. Sometimes the rule is grasped at the earliest period, e.g. for M i c h a l not a single error was found for the period 2;0 to 6;0, while other children tend to preserve the case ending used i n non-negated sentences and they say Nie czytam ksiqzk-% 'not Tread bookA C C instead o f ksiqzk-i ' b o o k - G E N ' . This can be seen i n the early data for Basia, who used the genitive case correctly w i t h possessives, as w e l l as i n contexts i n v o l v i n g the preposition do ' t o ' , but failed to use it w i t h negated existential sentences w i t h niema 'there:is:no' and preserved the nominative form typical for subjects o f positive existential sentences. Similarly, after she began constructing negated transitives, the object appeared i n the accusative case. Eventually, she learned the adult rule, but she frequently failed to apply i t i n negated sentences w i t h modals. Such errors occurred also w i t h other children. 4.6.3. Conditional Forms: Wrong Placement of the Person Suffix. This error can be found to occur i n practically all the children studied for at least a short period. According to the adult rule, the person suffix for first and second person i n singular and plural must follow the conditional particle by regardless o f the place i n which this particle happens to occur, that is, whether i t follows the past participle (e.g. pisal-Bv-m), precedes the verb (e.g. J a sv-m pisat), or constitutes a part o f the connective zeby (e.g. zesr-m pisat). A widespread error concerning the use o f conditional forms is that o f placing the suffix immediately after the participle (which results i n a regular past-tense form) while leaving the particle by either to follow i t or to occur i n an enclitic position, e.g. pisat-(e)m-by or Ja by pisal-(t)m T w o u l d w r i t e ' . Here the c h i l ­ dren apply the rule: A d d by to the past-tense f o r m , w i t h the option o f m o v i n g by to an enclitic position. This contrasts w i t h the adult rule: Insert by between the participle and the person suffix so that by + PERSON forms an enclitic. This erroneous rule is probably due to the influence o f the third-person forms, i n which the past participle is identical to the past-tense form where both rules lead to the same correct result, i.e. the child's pisal + 0 + by equals the adult's pisat + by + 0 (see also Section 6.1.3).

6.

T h e A c q u i s i t i o n of Polish

641

W i t h some children this error was typical only during the early uses o f the conditional form while i n others it was more persistent. I n the case o f Jas, for instance, the correct order o f morphemes had not been acquired by the age o f 6. Early errors i n v o l v i n g the incorrect order o f morphemes and/or overmarking for person (see Section 4.2) can be accounted for i f we propose that the child builds the conditional form on the basis o f the past-tense f o r m , and not the past partici­ p l e . As to late errors, some additional factors can contribute to their per­ sistence. Jas, for instance, was more likely to use the correct form i n main clauses than i n subordinate clauses, and he was especially reluctant to attach the person suffix to the connective zeby. I t could be hypothesized that children avoid applying a verbal suffix to a morpheme w h i c h is not a verb, and the morpheme by (even though, historically, it comes from an aorist) does not have the status o f an auxiliary verb. For zeby, it can be even more difficult for the child to accept the rule w h i c h requires that a connective be conjugated. Therefore, the adult rules can be rejected because they violate the well-established distinction o f w o r d classes. 2 2

4.6.4. Self-Reference: Using Feminine Forms. This error appears fre­ quently w i t h boys when they begin using the first-person form o f verbs. I t has nothing to do w i t h gender agreement, since the latter is f i r m l y established before the occurrence o f these e r r o r s . I n m y data the f o l l o w i n g sequence can be found. After the emergence o f feminine forms o f first person i n past-tense verbs, e.g. feminine pisal-a-m instead o f masculine pisai-0-(e)m, the use o f the femi­ nine form was overextended i n t w o ways to include: (1) adjectives referring to the child himself, e.g. Jestem glodn-a T:am h u n g r y - F E M , cf. correct giodn-y h u n g r y - M A S C \ and (2) the third-person past-tense form o f verbs but only i n self-reference, e.g. Michalpisal-a, ' M i c h a e l w r o t e - 3 S G : F E M \ (cf. correctpisal -0, ' w r o t e - 3 S G : M A S C ' ) . Earlier, these forms had the correct gender marking. These overextensions quickly disappeared while the feminine first-person forms persisted for several more months, occurring i n free variation w i t h correct mas­ culine forms. Eventually, the masculine form became the only one used by the child when referring to himself. This phenomenon was sometimes accompanied 23

4

I t is not the case, however, that children are unable to single out the past participle form, since they use it very early in correct future tense forms in which the person is marked on the auxiliary by means of special endings and not added to the participle, e.g. b$d-q pisal, bgdzie-sz pisal 'will -1SG write:PP, will-2SG write.PP') or overmarking for person (e.g. bqd-e pisal-(e)m 'will-lSG write:PP-lSG') are extremely rare. But on the other hand, a strange form mogl(e)-by-m (correct mogi -by-m), produced by Jas in his effort to order morphemes correctly, is clearly based on the past tense form mogl-(e)m T-could', as revealed by the presence of phonetic (e). Gvozdev (1949) reports quite a similar phenomenon in Zhenya's speech, noting that gender agreement with nouns was mastered by the age of 2;6, while feminine past-tense forms with selfreference continued to be used till as late as 3;0. 22

23

642

Smoczyhska

by the use o f the masculine second-person form irrespective o f the sex o f the person addressed (see also Rz^tkowska, 1909, and Brenstiern-Pfanhauser, 1930). It is quite clear that when learning the rules o f person switching, the male child employs the first-person form w h i c h he hears most frequently from his mother, while for the second person he selects the masculine form used by adults when addressing h i m . For girls, unless they are brought up by their fathers, the task is more simple because boys must perform an additional operation to arrive at the correct f o r m . This error is rare w i t h boys w h o have elder male siblings, since elder brothers provide the correct model. 4.7.

Summary

The above presentation o f grammatical errors can be considered as complete, in the sense that i t gives an account o f all systematic errors found i n the data, regardless o f their duration and o f whether they were made by most children studied or only one o f them. Special attention was given to errors occurring w i t h respect to regular adult language categories, while the children's problems w i t h irregular categories were only mentioned, even i f they produced a large number of erroneous forms. I f a category is not reported here or i n the section on U shaped developments (Section 5) or delayed acquisition (Section 6.3), i t means that its acquisition is error-free and does not cause any problems to children. The comparisons o f what is difficult for children w i t h what is easily acquired allow for some generalizations on children's strategies presented i n Section 7.

5. U - S h a p e d D e v e l o p m e n t s 5.1. Possessive

and Attributive

Modifiers

I n Polish, possession is expressed by means o f a genitival construction, while attribution is realized mostly by an adjective modifying the head n o u n . I n the earliest stages o f syntactic development, t w o distinct patterns appear (attribute plus head and possessive genitive plus head) w h i c h are subsequently used as constituents o f expanded sentences (Smoczyhska, 1978a). 2 4

During the third year o f life, children learn adjectival declension, and they try to reduce the number o f rules o f noun modification by replacing genitival pos­ sessors w i t h denominal adjectives. Chmura-Klekotowa (1971) reports a wide use of such possessive adjectives, e.g. mamusin, mamusiny, or mamusiowy instead of the genitive mamusi ' M o m m y ' s ' .

Although denominal possessive adjectives exist in Polish, they are not widely used by adults, the genitival form being much more common. In Serbo-Croatian and Russian the situation is differ­ ent: possessive adjectives are used much more frequently than genitival forms. 24

6.

The A c q u i s i t i o n of Polish

643

On the other hand, the data from Kasia, w h o d i d not use such possessive adjectives, provide interesting examples o f adding other case endings to the genitive form o f the noun i n order to achieve agreement i n case w i t h the head noun. N o r m a l l y , the genitive modifier preserves its form regardless o f the case o f the head. Kasia, however, treated it as i f it were an adjective, e.g. w pokoj-u Anul-i-m ' i n r o o m - L O C A n u l a - G E N - L O C or Anul-emi kredkami ' ( w i t h ) AnulaI N S T R : P L pencils:INSTR:PL' (cf. correct wpokoju Anuli and Anuli kredkami). Chmura-Klekotowa (1971) has cited similar examples. After the children have established separate rules for both kinds o f noun modification and learned possessive pronouns, another k i n d o f regularization appears. The Polish system is inconsistent i n that the function o f possessive is performed by possessive adjectives for the first and second person (as w e l l as the reflexive) while i n the third person, genitival forms o f the personal pronoun are used, w h i c h do not agree w i t h the head. I n the following set o f examples the first three show agreement and the second three do not: Moj-a ksiqzk-a ' m y - N O M b o o k - N O M ' , moj-q ksiqzk-% ' m y - A C C b o o k - A C C , and w moj-ej ksiqzc-e ' i n m y - L O C b o o k - L O C contrast w i t h jego ksiqzk-a ' h i s - G E N b o o k - N O M ' , jego ksiqzk-g ' h i s - G E N b o o k - A C C ' , and w je-go ksiqzc-e ' i n h i s - G E N b o o k - L O C . The regularization made by children consists o f using the genitive form o f the personal pronoun o f first and second person instead o f the possessive adjective, e.g. Michal chce do MNIE lozka ' M i c h a l wants (to come) into I ' S bed' (cf. correct: do MOJEGO lozka 'into M Y bed') or Gdzie jest C/EB/E rqczka? 'where is Y O U ' S hand?' (cf. correct: TWOJA rqczka ' Y O U R hand'), cited by Rzefcowska (1908). Typically, instead o f using a specific question word czyj? 'whose?' children incorrectly employ the genitive question w o r d kogo? ' o f : w h o ? ' . 5.2. Embedded Full Similar tions w i t h infinitive). embedded infinitive.

Infinitival

Complement

Replaced

by a

Clause

to English-speaking children, the earliest k i n d o f embedded construc­ the verb ' w a n t ' is that w i t h coreferential subjects ( T want' plus W h e n the t w o subjects are not coreferential, a construction w i t h an z^&y-clause is used, equivalent to the English T want h i m ' plus

W i t h Jas, infinitival clauses emerged at 2;0, e.g. Chcem pic dzem w shiku ' L w a n t to.drink j a m i n j a r ' , and the explicit zefry-clauses appeared three months later, e.g. Mother: Co ty chciales? ' W h a t do you want?' Jas: Zeby mamusia szla do kuchni z Jasiem 'that M o m m y goes to kitchen w i t h Jas'. F r o m this time o n , constructions o f both kinds were frequently used, without errors. Surprisingly enough, w i t h i n the period o f 3;1 to 3;4, a number o f examples were recorded i n which the explicit zeby-clmsz was used w i t h coreferentiality, e.g. J a chc%, zeby mialem kotka T want that Lhave cat' (cf. correct J a chc% miec kotka T want

Smoczyhska

644

to:have cat'). Although the infinitival construction acquired a long time earlier was not completely ruled out by the zeby-clause (the errors constitute about 25% of all instances o f coreferential ' w a n t ' sentences), the proportion is important enough to infer that an attempt at reducing the variety o f rules had taken place.

6. T i m i n g o f A c q u i s i t i o n This section is concerned w i t h those linguistic categories w h i c h are mastered by Polish children either especially early or especially late. There is a difficulty w i t h what should be taken as the reference point w i t h which the Polish acquisition pattern could be compared i n order to evaluate precociousness or delay i n the mastery o f particular categories. A common option here is to take the acquisition of English as the reference point, since it is best described and widely k n o w n . This option, however, w o u l d lead us, for instance, to state that for most lan­ guages the mastery o f negative and interrogative sentences is precocious, w h i l e in fact, it is i n English that i t is delayed because o f the formal complexity o f these transformations i n that language. Therefore I w i l l l i m i t myself to presenting only those categories w h i c h seem to be mastered i n Polish significantly earlier or significantly later than i n several other languages. 6 . 1 . Precocious

Acquisition

The categories reported here are: grammatical gender, w h i c h is acquired earlier than i n Russian (Section 6.1.1); tense and aspect, both o f w h i c h are acquired simultaneously and very early as compared w i t h other languages (Sec­ tion 6.1.2, presented by Richard Weist); and hypothetical reference, w h i c h ap­ pears i n Polish children t w o years earlier than i n English-speaking subjects (6.1.3). A s for the acquisition o f complex sentences, i t takes place i n Polish either at the same time as i n other languages or slightly earlier, o r — w i t h some types o f temporal clauses—later. I n order to present the data on complex sen­ tences i n a more or less systematic w a y , I included a separate section (6.2), which follows the present section. 6 . 1 . 1 . Grammatical Gender: Declensional Patterns and Gender Agree­ ment. The acquisition o f grammatical gender has been reported to occur late in Russian children, both w i t h respect to the establishment o f noun inflectional patterns and to the control o f gender agreement i n adjectives and past-tense verbs (Gvozdev, 1949; Zakharova, 1973; Popova, 1973; Slobin, 1966, 1968, 1973). A phenomenon typical i n the acquisition o f Russian is that o f "inflectional imperi­ a l i s m . " The child seizes upon one ending for a given case and uses it for all 25

T h e present discussion is limited to the acquisition of the gender distinction in singular (mas­ culine/feminine/neuter); the plural distinction into virile/nonvirile, which is acquired late, is dis­ cussed in Section 6.3.1. 25

6.

The A c q u i s i t i o n of Polish

645

nouns irrespective o f their gender. Later o n , he selects another " i m p e r i a l i s t i c " ending to drive out the initial " i m p e r i a l i s t i c " ending. Slobin (1966, 1968) gave a detailed description o f this phenomenon, trying to establish the determinants governing the selection o f the given ending as dominant. As far as gender is concerned, he concluded that this category seems to be difficult because o f its arbitrariness. M y analysis o f the early data o f several Polish children, together w i t h careful reading o f Polish diary studies, reveals that the acquisition o f gender is pre­ cocious i n Polish children. M o s t o f the children have acquired this distinction before the age o f 2. First, when a case category emerges, several endings appear simultaneously according to the gender o f the noun; thus, there is no inflectional imperialism. Second, the occurrence o f inappropriate gender agreements is l i m ­ ited to a very short period when adjectives, past-tense forms, or pronouns begin to appear, and it should be attributed to the lack o f knowledge o f possible differentiation o f these forms rather than o f that concerning noun gender. (Using feminine forms when referring to oneself (Section 4.6.4) is quite a separate phenomenon, w h i c h consists i n treating the first person-form used by the mother as the only one i n existence.) The early acquisition o f gender i n Polish is even more puzzling i f one takes into account the close similarity o f the Polish and the Russian declensional systems. The distribution o f endings i n the singular is almost identical. The Polish declension looks as i f the Russian endings were translated into the Polish phonological system. I n both languages the criterial nominative singular endings are 0 for masculine, -a for feminine, and -ol-e for neuter. Moreover, both languages have a limited number o f masculine nouns ending i n -a w h i c h take feminine case forms and masculine agreement, and a large class o f irregular feminine nouns ending i n 0 i n the nominative singular having a special set o f case endings. T w o apparently m i n o r factors appear to cause a considerable delay i n Russian acquisition. First, the fact that the Russian unstressed o is pronounced as a; and second, the existence o f diminutive masculine forms w h i c h end i n -a and are declined like feminine nouns, e.g. mishkalmyedvyed 'bear', zaykalzayats 'hare', and, very frequently, boys' first names such as Zhenya, Sasha, Kolya. These t w o apparently minor facts seem to be the cause o f the whole confusion. W h i l e most o f the nouns ending i n -a to w h i c h Polish children are exposed are regular feminine nouns, the input to the Russian child contains different declen­ sion and agreement patterns, e.g. mama ' m o m m y ' w i t h a feminine declension and agreement, Zhenya 'boy's name' w i t h a feminine declension and masculine agreement, and ydbloko [ydblaka] 'apple' w i t h a neuter declension and a neuter agreement (-o is identical to the feminine -a when unstressed). Such an inconsis­ tent input makes i t impossible for the Russian child to discover the criteria o f grammatical gender, while Polish children can tolerate the limited amount o f inconsistency to w h i c h they are exposed, simply regularizing irregular instances according to the tripartite gender distinction (see Section 4 . 3 . 1 . 3 ) .

646

Smoczyhska

Early acquisition o f gender i n Polish shows that the child is able to learn rules which are based on totally arbitrary criteria, provided these criteria are consistent and clear enough to be discovered, and that selecting one salient (or phonetically unique) ending is only a strategy that Russian children apply when faced w i t h the impossibility o f discovering adult criteria. 6.1.2. Tense and Aspect [WEIST], Weist, Wysocka, Witkowska-Stadnik, Buczowska, and Konieczna (1984) and Weist (1983a) have conducted the most intensive investigation o f tense and aspect i n child Polish. Given that tense expresses a deictic relationship between speech time and event time, Antinucci and M i l l e r (1976), A k s u (1978), Stephany (1981), and others proposed that the initial tense morphology i n languages such as Italian, Turkish, and Greek was " d e f e c t i v e " i n its function. A c c o r d i n g to the strongest form o f this "defective tense" hypothesis (see A n t i n u c c i & M i l l e r ) , children between the ages o f ap­ proximately 1;6 and 2;6 cannot express temporal deictic relationships because they lack the necessary abstract conceptual capacity. I t was proposed that when tense morphology emerges i t codes the aspectual distinction between completed and ongoing situations rather than a relationship between speech time and event time. Hence tense morphology was supposed to be coding an aspectual distinc­ tion, not a tense distinction. The research on Polish challenges this hypothesis. The defective tense hypothesis involved three lines o f argumentation and can be viewed as having the f o l l o w i n g three components: (1) semantic—only telic verbs w i l l be inflected i n the past tense because a prior change o f state w i t h a speech-time resulting state is required; (2) syntactic—aspectual distinctions w i l l not be made independently o f tense distinctions, and present versus past mor­ phological distinctions w i l l correspond i n a one-to-one manner w i t h the situa­ tional properties " c o m p l e t e d " versus " o n g o i n g " ; and (3) temporal—children w i l l refer only to the immediate past. Evidence supporting each component o f the argument can be found i n the research o f Antinucci and M i l l e r (1976), Stephany (1981), and Szagun (1979), respectively. I n fact, none o f these components o f the defective tense argument are supported by the evolution o f tense and aspect distinctions i n child Polish—nor any other Slavic language for that matter. When considering the semantic component o f the argument, one must first draw attention to the problem o f defining subcategories o f verbs w h i c h are relevant to verbal aspect (or Aktionsart). Since defining categories o f verbs according to the situations to w h i c h they refer leads to a hopelessly circular argument, objective definitions are needed. Weist et al. used a contemporary version ( D o w t y , 1979) o f Vendler's classification schema, w h i c h was shaped for Polish w i t h the guidance o f Cochrane's (1977) w o r k on Serbo-Croatian and M i l l e r ' s (1970) w o r k on Russian. Growing out o f this framework, objective distinctions were made between state versus dynamic verbs, atelic (or activity) versus telic verbs, and achievement versus accomplishment verbs. F r o m the initial period o f tensed language, Polish children use imperfective activity verbs i n the past, producing such utterances as

T h e A c q u i s i t i o n of Polish

6.

647

the f o l l o w i n g : (1) Marta (1;7): Lecial samolot f l y : I P F V : P A S T plane' ( = 'The plane was f l y i n g ' ) , (2) Bartosz (1;8): Pfywala si$ ' s w i m : I P F V : P A S T ' , and (3) Paulina (1;11): Jadiam e a t : I P F V : P A S T \ The production o f activity verbs w i t h past-tense inflections can be found i n other research on Polish (Smoczyhska, 1978a), research on Russian (Gvozdev, 1961), and on Serbo-Croatian (Radulovic, 1975). This phenomenon is by no means unique to Slavic languages, e.g., Spanish (Eisenberg, 1982), Greek (Stephany, 1981), Japanese (Rispoli, 1981), and Finnish (Toivainen, 1980). 4

4

Regarding the syntactic component o f the defective tense hypothesis, tense and aspect morphology emerge simultaneously i n Polish. D u r i n g the acquisition of the verb system, there is an early period during w h i c h children use only frozen forms. There is some evidence (e.g. Zarebina, 1965) for a primitive distinction in mood at this point, since indicative forms are typically used to make state­ ments and imperative forms to make requests. The indicative forms are typically imperfective verbs i n the present tense and i n the third-person singular. Zarebina has shown that this third-person form is generally applied making reference to second as w e l l as third person. Typically when children are between about 1;6 and 1;9, they begin to use past-tense inflections and perfective verb forms. References to the future are made at that time or soon thereafter. Since the present perfective meaning is impossible i n Polish, the combination o f three tenses and t w o basic aspects produces five possible verb forms. Only the pe­ riphrastic future form for imperfective verbs is sometimes missing (or difficult to identify) i n the initial set o f tense/aspect combinations. These distinctions appear to emerge without errors i n Russian (Gvozdev, 1961) and i n Serbo-Croatian (Radulovic, 1975) as w e l l as i n Polish. There is no evidence W H A T S O E V E R that tense morphology codes the aspectual distinction between perfective and imper­ fective or that tense morphology emerges as a redundant appendage. When considering the temporal component o f the overall argument, the first obvious finding is that the resulting state o f a previous change o f state is not required for children to make reference to past events. A t 1;7, for example, Marta was talking to her grandmother about their recent trip to the seashore and in response to the question A co panowie robili? ' A n d what were the men doing [ I P V F ] ? ' , Marta replied iowili T h e y were catching [ I P F V ] ' ; and i n response to further questioning, rybki ' f i s h ' . Lowili . . . rybki is an imperfective accom­ plishment verb phrase and as such a clear terminal point is implied, yet the child did not need any specific retrieval cues much less the resulting catch to refer to this event. Weist et al. partitioned all o f the utterances w h i c h made reference to a prior situation into the categories " m o d e l e d " versus "spontaneous" and " i m ­ mediate" versus "moderately r e m o t e " (more than t w o turns i n the discourse). During the early period o f tensed language, more than one-fourth o f the utter­ ances o f the children were both spontaneous and moderately remote. I n summa­ ry, tense and aspect emerge simultaneously and from the initial period o f tensed utterances, Polish children produce activity verb phrases i n the past, imperfective

648

Smoczyhska

as w e l l as perfective past forms, and moderately remote references to the past. Experimentation w i t h i n the cross-sectional component o f the research design only confirms this general c o n c l u s i o n . [WEIST] 26

I n general, m y data support Weist's findings. After the earliest period o f the very emergence o f tense/aspect distinctions, no instances were found o f inap­ propriate use o f tenses. Aspectual errors do occur, but they are insignificant i n proportion to the total frequency o f verbs (Polish children are forced to make an aspectual option each time they use a verb); for instance, sometimes the analyt­ ical future form occurs w i t h a perfective verb (e.g. b%d% na-pisac T:shall PFVwrite' cf. correct P F V na-pisz% or I P F V b%d% pisac), but it should be treated as a mistake, since children frequently correct themselves after having produced such a form.

[WEIST] Considering children who are somewhat older, there are t w o sources o f data w h i c h attest to the child's control o f aspectual distinctions i n comprehension and production. Weist (1983a) used a sentence-picture matching task to evaluate the capacity o f children to comprehend tense and aspect distinc­ tions w h i c h were made either by prefixation or by suffixation. T w o groups o f ten children at age levels 2;6 and 3;6 participated i n the experiment. The problems concerning aspectual distinctions contrasted pictures o f completed versus i n ­ complete situations and sentences differing only i n the form o f the verb, e.g. zjadlljadi P F V / I P F V 'ate/ate, was eating' or zamknqilzamykai P F V / I P F V 'closed/closed, was closing'. The children i n both groups were able to com­ prehend the aspectual distinction. Regarding productive capabilities, Chmura-Klekotowa (1967, 1968) has con­ tributed important insights from her w o r k on neologisms. Polish children use prefixation and suffixation processes to invent aspectual contrasts. A t 3;4, one child invented the imperfective form zwyczaic 'to:get:used:to'. For the adult, the aspectual pair is przyzwyczaiclprzyzwyczajac siq P F V / I P F V . The child appears to have used the perfective form as the basis for a prefix deletion i n the formula­ tion o f his imperfective f o r m . I n the next example, the process o f suffixation was used by a child at 4;0 to create the form laskotnq T : w i l l : t i c k l e ' . I n contrast to the adult's aspectual pair w h i c h is based on prefixation laskotaclpolaskotac I P F V / P F V , the child invents the perfective form laskot-nq-c ' t i c k l e - P F V - I N F ' by suffixation. Gvozdev ( 1 9 6 1 , pp. 4 2 4 - 4 2 7 ) has shown that Russian children 27

[ W E I S T ] This section replaces an earlier section which Smoczyhska based on the preliminary findings of this research which were reported in Krakow (Weist, 1980). In this research on tense and aspect, the classification of verbs always included verb phrase context (see Verkuyl, 1972). [WEIST] [ W E I S T ] Slobin (personal communication) has made the interesting observation that Polish and Russian children create neologisms differently, depending on whether a tense distinction or an aspect distinction is being constructed. When making a tense distinction, children appear to be searching for a common stem (see Section 4.5.1). However, when making an aspect distinction, children try to construct two different forms. [WEIST] 26

27

6.

The A c q u i s i t i o n of Polish

649

also use prefixation and suffixation to invent aspectual pairs. The form zdravil 'congratulated' observed at 6; 10 was derived by prefix deletion from the perfec­ tive form po-zdravif. The adult's aspectual pair is pozdravif Ipozdravlyaf P F V / I P F V . A t 2; 10 the f o r m katnul ' r o l l e d ' was observed. The perfective form katnuf was constructed from the imperfective kataf, by -nu suffixation. For the adult the aspectual pair is formed by prefixation kataf Ivykataf. [WEIST] Adverbs w h i c h point to aspectual and temporal characteristics o f events also emerge very early. A m o n g the earliest adverbs used are those indicating comple­ tion vs. duration o f an action: jeszcze 'more, again, s t i l l ' is a pivot w o r d for recurrence, but it is also used to express imperfectivity (cf. also jeszcze nie 'not y e t ' ) , and the opposite term is juz 'already' (Przetacznikowa, 1956; Smoczyhska, 1978a). Another very frequent adverb is teraz ' n o w ' , which is used mostly to mark immediate futurity. I n some children, numerous temporal adverbs are used i n early speech, e.g. Kasia at 1;9 used dzis 'today', jutro ' t o m o r r o w ' , niediugo 'soon',potent 'afterwards', and even asked kiedy? ' w h e n ' questions. These time words were used appropriately i n that they referred to a kind o f temporal notion; however, their specific meanings were not yet acquired. As a matter o f fact, such notions are mastered very late: during the whole period under study, numerous examples were found o f using ' t o m o r r o w ' instead o f 'yesterday', 'afterwards' instead o f 'before', etc., whereas tense-marking on the verb was always correct. Words expressing 'timelessness' such as zawsze ' a l ­ ways' and nigdy 'never' emerge before the age o f 3 that is, much earlier than i n English-speaking children (cf. Cromer, 1968). 6.1.3. Hypothetical Reference. B r o w n (1973) states that the hypothetical use o f past-tense forms comes very late i n English-speaking children. Cromer (1968) d i d not find any reference to hypothetical events, even inappropriately expressed, i n the data o f the Harvard children until the age o f 4;6. Kuczaj and Daly (1979), w h o used eliciting techniques i n naturalistic and experimental situations, obtained a number o f hypothetical statements from younger children. They found that such early uses are more often implicit than explicit and otherinitiated than self-initiated. They refer to single hypothetical events rather than to a sequence o f such events, and reference to future hypothetical events is prior to that to past ones. Gvozdev (1949) and Slobin (1966) have pointed out that relative to the emergence o f imperative and indicative moods, the conditional mood develops somewhat late, e.g. Zhenya was observed to produce conditional clauses w i t h esli ' i f at 2;9. W h i l e this may be relatively late for Russian, i t is relatively early when compared to English. The onset o f conditional forms i n Russian occurs almost 2 years earlier than i n the English-speaking Harvard children, and i t is preceded by the emergence o f non-factive subordinate clauses introduced by stob(y) which closely correspond to the Polish ze/?y-clauses.

650

Smoczynska

I n m y data, zeby-clmses are found at about 2;0 ( w i t h some children as early as 1;9 or even 1 ; 7 ) . Causal clauses providing a hypothetical reason appear at the same time, e.g. Jas (2;1): 28

Nie Not

rzuca-m throw-1SG

mokr-e wet-NEUT

pidrk-a feather-NEUT:GEN

by-0 COND-3SG

do to

wody water-FEM:GEN

bo because

byl-o be:PP-NEUT

T don't throw a feather into the water because it w o u l d be w e t ' . Note that the future tense form (bo bqdzie mokre) w o u l d have worked as w e l l , but Jas d i d not avoid the conditional mood form i n spite o f the existing option. Later o n , the first i f -clauses are used i n indicative m o o d , and somewhat later hypothetical condi­ tional clauses appear. They are not as frequent as at the age o f 5; nevertheless, several examples can be found i n the spontaneous speech o f all children before the age o f 3. The f o l l o w i n g examples, provided w i t h detailed glosses, can serve not only as evidence for hypothetical conditionals but also as illustration o f some phenomena w h i c h appear typically from 2 to 3: 4

Agnieszka (2;0) Jak If

ty you

ci you: DAT

by-s COND-2SG

chcial-a, want:PP-FEM

to PARTICLE

by-m COND-1SG

dal-a give:PP-FEM

'If you wanted (it), I would give (it) to you.' Kasia (2;6) Ja ma-m I have-lSG byl-a be:PP-FEM

dw-a two-MASC

jedn-a one-FEM

kieszonk-i. pocket-NVIR:FEM:PL.

to PARTICLE

by-0 COND-3SG

A jak And if

by-0 COND-3SG

brakowal-o. be:missing:PP-NEUT

'I have two pockets. And if there was one, (something) would be missing.' (The neuter -o i n brakowal-o expresses the impersonal 'something w o u l d be missing'.) Note the erroneous dw-a instead o f dwi-e. The f o l l o w i n g t w o examples from Kasia 2;9 involve double hypothetical implications o f a hypothetical condition: Jak If

by-m COND-1SG

wylazl-a climb:PP-FEM

na plot-0 on fence-ACC

to PARTICLE

jeszcze even

With "telegraphic" children, conditional mood clauses can be deprived of their specific markers, i.e. by and conjunctions, the only cue to the non-factive or hypothetical reference being the inappropriate use of past tense forms (see Section-4.1). 28

6.

by-0 C0ND-3SG

mnie me:ACC

bab-a old:hag-FEM

by-0 C0ND-3SG

mnie me:ACC

porwal-a? kidnap:PP-FEM

The A c q u i s i t i o n o f Polish

651

skrzyczal-a . . . Mo'ze scold:PP-FEM Perhaps

Tf I climbed the fence, the old hag would even scold me . . . Perhaps she would kidnap me?' Jak If

by-m COND-1SG

to PARTICLE

poszi-a go:PP-FEM

bez without

mis-0 teddy:bear-MASC:NOM

wolal-0-by-0 call:PP-MASC-COND-3SG

misi-a teddy :bear-MASC: GEN

by-0 COND-3SG

ptakal-0. cry:PP-MASC.

I And

Kasi-u! Kasia-VOC

Tf I went without my teddy-bear, the teddy-bear would cry. And he would call: Kasia!' The f o l l o w i n g four examples illustrate not only the hypothetical conditionals but also typical difficulties w i t h the person suffix placement, as described i n Section 4.6.3. These difficulties appear w i t h respect to the first and second person only, but i n view o f the special rule children use, the third person forms have to be glossed differently, as compared w i t h the examples cited above, and w i t h adult use.

Kasia (2;7) Jak If

by COND

ukroil-a-m cut:PP-FEM-lSG

skaleczyi-a-m. hurt:PP-FEM-lSG

to PARTICLE

Krew-0 B lood-FEM: NOM

by COND

by sie] COND R E F L mi me: DAT

leciai-a-0. fall:PP-FEM-3SG

'If I cut (it), I would hurt myself. I would bleed.' Literally: 'Blood will fall to me'—correct in Polish! Cf. adult 3SG form: by-0 COND-3SG

leciai-a fall:PP-FEM

Note the presence of two implications. Mis (2; 10) [After his brother told him he would not let their mother enter the school, he said]: Jak If

by COND

ja I

chodzil-0-(e)m go:PP-MASC-lSG

to PARTICLE

wpuscil-0-(e)m. let:in:PP-MASC-lSG Tf I were going (to school), I would let mommy in.'

by mamusi-^ COND mommy-ACC

652

Smoczyhska

Mis (3;0) Mamusi-a Mommy-FEM:NOM byl-0-(e)m be:PP-MASC-lSG sam-0 alone-MASC

by COND w in

w in

plakai-a-0 cry:PP-FEM-3SG

spoldzieln-i. shop-LOC

I ja And I

jak if

by COND

by COND

ja I

plakaI-0-(e)m cry:PP-MASC-lSG

spoldzieln-i. shop-LOC

'Mommy would cry if I were in the shop. And I would cry in the shop alone.' Jas (2; 10) Jak If

by COND

klej-em glue-INSTR nie not

Jas-0 Jas-MASC:NOM

to PARTICLE

by COND

powalai-0-0 smear:PP-MASC-3SG byl-0-(e)m be:PP-MASC-lSG

si% REFL

klej-0 glue-NOM

a and

Jas-0. Jas-NOM

'If Jas smeared himself with glue, I would be glue and not Jas.' Note the coexistence o f t w o perspectives i n self-reference. Jas uses here both first and third person while speaking about himself. This phenomenon, typical o f the third year o f life, is also illustrated i n the next sentence, where the 1SG possessive pronoun ' m y ' is used, although the verb occurs i n 3SG.

Jas (2; 10) Jak If

by COND

Hani-a Hania-FEM

by COND

Jas Jas:MASC:NOM

moj-e my-PL

pieniqdz-e. money-PL

stoil-a stand:PP-FEM-3SG musial-0-0 must:PP-MASC-3SG

na on

wdzk-u, pram-LOC

kupi-c buy-INF

to PARTICLE

wdzek-0 pram-ACC

za with

T f Hania stood on the pram, Jas w o u l d have to buy a (new) pram w i t h m y money'. Note the incorrect stem i n stoi-i-a (= sta-I-a). I n the above example, as w e l l as i n the f o l l o w i n g one, we cannot credit Jas w i t h the adult glossing o f the 3SG conditional forms. I f I insist on that, it is mostly i n order to stress that the apparently correct surface structure must not necessarily reflect a correct under­ lying rule. Jas (2; 10) [After having been scolded for locking a neighbor in the bathroom, he explains that no harm was done.]

6.

Jak If

by COND

to PARTICLE

pan-i lady-FEM (= she) by COND

si% REFL

tez also

The A c q u i s i t i o n o f Polish

zamykal-a-0 lock:PP-FEM-3SG

sie. REFL

653

sam-a, self-FEM

boil-a-0. frighten: PP-FEM-3SG

'If she locked her up herself, she would also be frightened.' Note the incorrect stem in boi-t-a (= ba-l-a) A few more observations on conditionals are relevant here. A s seen from the above examples, the children refer not only to sequences o f hypothetical events (condition-implication), but often t w o hypothetical conditions or implications are mentioned. Sometimes the logical order is reversed, e.g. Jas (2; 10) said, Jak by bralem zastrzyki, to by by km chory i mialem katar ' I f I took injections, I w o u l d be i l l and w o u l d have a c o l d ' . One o f the children studied, M i c h a l , began asking hypothetical questions at 2 ; 1 , e.g. Jak by kto Michaiowi urwalrqczki,

to co? ' I f

somebody tore o f f M i c h a l ' s hands, then what?', and then he frequently carried out " e x p e r i m e n t s , " presenting t w o alternate versions o f events, e.g. [putting his hand close to a candle] M i c h a l (2:2): A jak bym si$ sparzyla? ' A n d i f I burned myself?' Mother: To by by to He ' I t w o u l d be bad'. M i c h a l : A jak by nie? ' A n d i f not?'. Similarly at the age o f 2;3: A jak by nie mialem glowy? ' A n d i f I had no head?' Sister: To by ci% nie bylo 'Then you w o u l d n ' t exist'. M i c h a l : A jak bym byla z glowq? ' A n d i f I were w i t h a head?' Sister: To bys byl

To jestes

'Then

you w o u l d exist. Y o u exist'. M i c h a l : A jak bym nie byia z glowq? ' A n d i f I were not w i t h a head?' Finally, at the age o f 3;1 the f o l l o w i n g complex reasoning was recorded: M i c h a l : Co by bylo, gdyby Pana Boga nie bylo? ' W h a t w o u l d happen, i f there was no God?' Mother: To by i nas nie bylo 'Then we w o u l d n ' t exist either'. M i c h a l : A jak by Pana Boga nie bylo, a smy byli? ' A n d i f G o d d i d n ' t exist and we did?' To summarize, the only overlap w i t h Kuczaj and D a l y ' s results is that refer­ ence to future hypothetical events appears before that to past hypothetical events. In all other respects, the development o f Polish children is decidedly precocious when compared w i t h their English speaking peers. There are three factors that can account for this finding. First o f a l l , the conditional particle by is a salient and a unique marker for nonfactive events (hypothetical events included). Secondly, the first distinction to be made is that between factive versus nonfactive reference, and the conditional mood is intro­ duced first i n zeby-clauses, w h i c h have an explicit structure. I n contrast, the above distinction is not marked at all i n English, and the counterparts o f Polish zeby-clauses are opaque infinitival constructions. T h i r d l y , English conditional clauses are expressed b y means o f the purely conventional use o f the past-tense form i n the subordinate clause and a would + V E R B construction i n the main clause. The parallel distinction i n Polish is realized i n terms o f transparent constructions i n w h i c h every single occurrence o f nonfactivity is clearly and consistently marked by conditional m o o d .

654

Smoczyhska

6.2. Acquisition

of Complex

Sentences

I n general, the acquisition o f complex sentences matches the universal pattern described by Clancy, Jacobsen, and Silva (1976), as w e l l as that for English presented i n L i m b e r (1973) and Bowerman (1979). A t some points, however, i t seems to be precocious. I n this section I try to present the data on complex sentences i n a more or less systematic way. Complex structures which appear before the age o f 2 involve coordination, antithesis, sequence, infinitival embed­ ded clauses, and causal adverbial clauses. W h i l e conjoined structures seem to develop i n a manner w h i c h is similar to other languages (cf. Clancy et a l . ) , the two latter ones deserve special attention. W i t h children who formulate their requests as "desires" rather than as " o r ­ ders" (Smoczyhska, 1980), the earliest embedding verb is chce ' w a n t ' , e.g. Tenia (1;6): Jablko chce jesc 'apple want to:eat', Kasia (1;7): Isc na dot chce Kasia 'to:go on d o w n wants Kasia', Kasia (1;9): Kasia chce budzic babcig 'Kasia wants to:wake:up Granny'. These sentences seem more complex than those reported for English at the same a g e . Before the age o f 2, sentences involving a non-coreferential subject o f the embedded verb are not used or used incorrectly, e.g. Kasia (1;9): Kasia nie chce kaszlesz 'Kasia not wants y o u x o u g h ' , i n w h i c h the second-singular present-tense form is used instead o f a zeby-clmse (cf. correct Kasia nie chce, zeby-s kaszlal-a 'Kasia not wants that-2SG cough:PP-FEM' ( = 'Kasia doesn't want you to cough')). 29

Another type o f early embedding involves the Polish " m o d a l s " (or better "pseudo-modals"), both personal and impersonal; e.g. Jas (1;10): Mo'zna ly'zeczkq pic herbatkq? ' O n e x a n with:spoon toidrink tea?'. These were ex­ tremely frequent w i t h Kasia, who freely employed a number o f constructions w i t h "psuedo-modals" before the age o f 2 , e.g. Kasia (1;9): Trzeba przykryc mamgnaszq . . . poduszkq 'one:should to:cover m o m m y our . . . w i t h : p i l l o w ' , or the following sequence recorded at 1;9 (note the variability o f w o r d order): Musi Kasia jesc obiadek. Musi obiadek jesc. Myc naczynie Kasia musi. Musi Kasia naczynie myc. Garnuszek musi Kasia myc. ' M u s t Kasia to:eat lunch. M u s t lunch to:eat. To:wash:up dishes Kasia must. M u s t Kasia to:wash:up dishes. Cup must Kasia to:wash:up.' These constructions also seem to appear a little earlier than i n English (cf. Limber, 1973). 3 0

The earliest adverbial clauses are those providing "reasons" for requests. I n "telegraphic" children these are realized by means o f mere juxtaposition, e.g. Basia (1;9): Babciu, zdjqc to, ba to, tu dziura 'Granny, take i t off, ugly this, here hole' [Basia wants Granny to take o f f her apron because i t is dirty and has a hole in i t . ] I n other children the conjunction bo 'because' appears, e.g. Kasia (1;8): Umyc rqczk%, bo brudna 'to:wash hand, because d i r t y ' (This utterance is quite A s in English, these are preceded by simpler forms such as 'want apple' or 'want eat'. I t should be noted, however, that Kasia, who began combining words at 1;3, manifested precocious development. 2 9

30

6.

The A c q u i s i t i o n o f Polish

655

TABLE 6.8 The Proportions of Various Kinds of S u b o r d i n a t e Clauses W i t h i n C o m p l e x Declarative Sentences Recorded f o r Jas (Based o n Przetacznikowa, 1968) Age Type of Clause Complement clauses object Adverbial clauses cause purpose condition/time Relative clauses Other Number of sentences

2-3

3-4

4-5

5-6

21%

9%

12%

14%

21% 10% 25% 12% 1%

36% 8% 26% 17% 4%

37% 4% 26% 14% 7%

33% 2% 29% 14% 8%

165

290

203

280

correct i n adult Polish!). I t is only later that &6>-clauses become attached to declarative statements. A bo-clause usually follows the main clause. Thus the order o f events mentioned is reversed. Wigc 'so'-clauses w h i c h reflect temporal sequence o f events are marginal i n both child and adult speech. As for later developments, I shall concentrate on subordinate clauses. Table 6.8 shows the quantitative development o f subordinate clauses i n declaratives over the period from 2 to 6 years. The figures w o u l d be somewhat different i f all kinds o f utterances were taken into account. Some types o f clauses are especially frequent w i t h requests, e.g. cause and purpose adverbials, or w i t h questions, e.g. conditional questions o f the k i n d

' W h y i f so-and-so, then so-and-so?'. The

development o f particular kinds o f subordinate clauses w i l l be discussed below. 6 . 2 . 1 . Complement Clauses.

Object complement clauses emerge at about 2

years. The complement-taking verbs used correspond to those listed i n L i m b e r (1973). The most frequent are verba dicendi and sentiendi. W i t h " t e l e g r a p h i c " children, sentences have an English-like structure (see Section 4 . 1 ) , but most children use a whole range o f appropriate connectives, e.g. factive ze 'that', nonfactive zeby that\jak l

questions.

31

'as', etc., as w e l l as a number o f wh-words i n indirect

Errors i n the choice o f factive versus nonfactive complementizers

are extremely rare although examples can be found, e.g. Kasia (2;9): Obiecujg, embedded questions are formally identical to direct questions since neither kind involves inversion. 3

656

Smoczyhska

zeby Michal byl grzeczny w szkole T:promise so:that M i c h a l wouldibe good at:school'. This error was obviously caused by the lack o f understanding o f the verb 'promise'. As far as the 'ask/tell' problem is concerned, Polish displays certain pecu­ liarities. The meaning o f the English ask is realized by t w o distinct verbs, i.e. zapytac 'ask:a:question' and prosic 'ask:a:favor'. However, another ambiguity arises w i t h tell since Polish has no tell/say distinction and only the verb powiedziec is used followed by ze&y-clauses i n the sense o f 'order' and factive ze or wh-clauses for 'tell/say'. Although children do not m i x the t w o meanings and use appropriate complementizers, i t is often the case that they use powiedziec 'tell/say' instead o f zapytac 'ask:a:question', e.g. Basia (4;0): Ale powiem mamie jak trzeba ciqc 'but I:shall:tell t o : M o m m y how one:should c u t ' , when zapytam T:shall:ask' was intended. I t could be explained by the semantic rela­ tionships between the t w o verbs: 'asking' constitutes a subclass w i t h i n a broader class o f ' t e l l i n g ' / ' s a y i n g ' , and the child can simply use the more general term i n the matrix clause without specifying that what he says is a question. Subject complement clauses emerge later. They occur w i t h some special constructions, e.g. zdaje mi siq, ze 'seems to:me that', snilo mi si%, ze 'dreamed to:me that' ( = T dreamed'), prawda, ze 'true that', and they are very infrequent. 6.2.2. Adverbial Clauses. Clancy et al. (1976, p . 74) stated that, " E v e n i n adult language the three concepts o f cause, time, and condition are closely related and overlap to some extent." One could argue that these concepts are i n fact aspects o f a broader category w h i c h could be called something like " r e l a ­ tionships holding between events" and that the partitioning o f the particular meanings among these "concepts" differs from one language to another. M o r e ­ over, i n the case o f Polish, another " c o n c e p t " is clearly marked, namely that o f purpose, as expressed by means o f ze&y-clauses. Indeed, the means o f expressing "relationships between events" are not limited to subordinate adverbial clauses. For instance, what Italian children express by conditional clauses w i t h senno ' i f not' is translated into English by means o f coordinate clauses w i t h or (Clancy et a l . , 1976, p . 78). I n Polish the same meaning w i l l be rendered by a zeby-clause w i t h the verb negated indicating " a purpose to a v o i d , " for instance, Nie ruszaj tego, zebys si$ nie sparzyi ' D o n ' t touch i t , so that you w o u l d n ' t burn yourself,' or by a causal &oclause w i t h the verb i n future tense, for example, Nie ruszaj tego, bo si% sparzysz ' D o n ' t touch i t because you w i l l burn y o u r s e l f , w i t h the implicit ' I f you do i t ' ; i n English one w o u l d say Don't touch it or you will burn yourself. Similarly, many o f Polish ze&y-clauses w i l l be rendered i n English or Italian by means o f a causative construction. Each language therefore makes a different categorization o f possi­ ble relationships, and the same event can be shaped differently by speakers o f different languages. I t w o u l d be interesting to investigate how these linguistic differences affect cognitive development.

6.

The A c q u i s i t i o n of Polish

657

6 . 2 . 2 . 1 . Cause. Causal adverbial clauses emerge i n Polish children before the age o f 2, and they constitute the most frequent category o f subordinate clauses throughout the whole period under study (see Table 6.9). Their func­ tions, however, undergo certain significant changes. Geppertowa (1968), who analyzed the meanings o f the conjunction bo 'because' as used by several chil­ dren, distinguished between the f o l l o w i n g usages: (1) motivational "reasons" (both objective and less frequently, subjective) used to justify actions and re­ quests; (2) statements o f objective causality; and (3) beginnings o f logical rea­ soning. Table 6.9 presents the change o f meaning o f causal adverbial clauses i n Jas, as analyzed i n Przetacznikowa (1968). I t can be seen that motivational reasons dominate until 4 at w h i c h time statements o f objective causality begin to prevail. The logical reasoning category can be found at about the age o f 3 and becomes more frequent w i t h age. In the data there are t w o examples w h i c h indicate that there can be nothing causal about ^ - c l a u s e s used as ' 'reasons" for requests, their only function being that o f strengthening the request itself, e.g. talking to her mother who was laughing at her, Kasia (2;1) said: Nie smiej si% mamusiu, bo . . . jest slisko. Nie wolno si% smiac ' D o n ' t laugh, m o m m y , because . . . it's i c y . One: is mot: al­ lowed to:laugh' and this example from M i s (2;9): Ubierz mnie, bo szyby siq stlukq. (Mother: Dlaczego?) Bo szyby si% stiukq, jakby mnie si% potem ubieraio. 'Dress me, because w i n d o w will:break' ( ' W h y ? ' ) 'Because w i n d o w will:break, i f one dressed me later'. Both were uttered when the child was so angry that he could not look for a more appropriate " r e a s o n , " and just picked up the first one &oclause he could think of. 6.2.2.2. Purpose. Clauses o f purpose are reasonably frequent i n early stages (see Table 6.8). They are also used to express "reasons," e.g. M i c h a l (2;9): Ja plujg na indyki, zeby siq obraziiy T spit on turkeys so:that they:get:of-

TABLE 6.9 Meanings of Causal Adverbial Clauses in Declarative Utterances of Jas (Based on Przetacznikowa, 1968) Age Categories

2-3

3-4

4-5

5-6

motivational reasons objective causality logical reasoning

73% 25% 2%

59% 26% 15%

33% 42% 25%

39% 39% 22%

Number of sentences

52

76

93

104

658

Smoczyhska

fended'. I n fact, both kinds o f meanings overlap, w h i c h frequently makes the children use a sequence o f conjunctions, e.g. Michat (2;3): Mamo, zapal lamp%, BO ZEBVwidziaiem, bo jest ciemno ' M o m m y , put the light on, B E C A U S E S O I T H A T I:see, because i t ' s dark'. Sometimes the purpose clause is quite superflous, e.g. Jas (3;2): Chc% tuprzyjsc, zeby tu byiem ' L w a n t to:come here so:that L a m here'. A detailed analysis o f various meanings o f zeby as used by children is given i n Geppertowa (1956). 6.2.2.3. Time I Condition. Except for hypothetical conditional sentences which are clearly marked as such (see Section 6.1.3), i t is impossible to dis­ tinguish between temporal and conditional clauses since both are introduced b y the conjunction jak 'as'. The question should be asked whether this reflects the parental tendencies to use only jak when talking to their children, since normally the spoken language provides the option for distinguishing between the t w o meanings, i.e. kiedy ' w h e n / w h i l e ' and jesli ' i f . I f the hypothesis were not confirmed, i t w o u l d be somewhat surprising i f children d i d not prefer the clear marking o f temporal sentences w i t h the connective kiedy instead o f plurifunctional jak. This expectation is supported b y the fact that kiedy is also a question word for time w h i l e jak is a question w o r d for manner ( ' h o w ' ) as w e l l as a comparative term Tike, as', a complementizer 'as', etc. Clauses introduced by jak emerge early. They usually occur i n sentence initial position w i t h the main clause introduced by the semantically empty particle to. I n Polish, the distinction ' w h e n / w h i l e ' is not marked i n the conjunction. I t is expressed through the aspectual marking o f verbs i n both clauses. The imperfective aspect o f the subordinate clause points to simultaneity o f events. The perfec­ tive verb form indicates that the event referred to i n the subordinate clause occurred first. I n the sentence Spotkaiem Zosi$ jak kupowalem chleb T:met:PFV Zosia while I:was:buying:IPFV bread' the past tense kupowalem o f the imperfective verb kupowac is used and the w o r d jak means ' w h i l e ' . However, i n the sentence Spotkaiem Zosig jak kupilem chleb T met Zosia when I:bought:PFV bread' the word jak is n o w interpreted as ' w h e n ' or 'after' because o f the perfective verb kupic w h i c h appears i n the past form kupilem. The perfective aspect o f the adverbial clause points to the fact that the action o f buying bread occurred first, and the meeting w i t h Zosia, second. I t is not marked i n the sentence whether the meeting took place at the very moment when the buying was completed ( ' w h e n ' ) or some time later ('after'). I f the clauses are reversed, w i t h the ja/c-clause coming first, the order o f mention matches the temporal order of events, and i n fact this order prevails i n children's data. Temporal sentences w i t h the main clause coming first are usually those referring to simultaneous events (where the order o f mention is irrelevant). Further research w o u l d be needed to establish w h i c h o f the t w o meanings ( ' w h e n ' vs. ' w h i l e ' ) appears earlier, nonsimultaneous temporal sentences are more frequent but a large part o f them could be better described as conditional clauses.

6.

The A c q u i s i t i o n of Polish

659

As for other temporal clauses, the meaning o f 'after' is usually rendered by aspectual marking i n the jak-clmsc and no special construction is needed. There is, however, a nominalized form o f the 'after'-clause i n v o l v i n g po 'after' plus a verbal noun i n locative case. Such constructions are not recorded i n the data, except for t w o examples i n M i c h a l (4;1) [asking his mother to give h i m her pen] Mother: Nie mog$, bo pisz% T x a n n o t because I : a m : w r i t i n g ' M i c h a l : Na chwilkg albo po pisaniu 'For a moment or after w r i t i n g ' Mother: Ale ja terazpiszq ' B u t I am w r i t i n g n o w ' M i c h a l : To po napisaniu 'Then after having:written'; and Michat (5;1): Zotnierzpo zwyciezeniu idzie 'Soldier after having:won goes'. A n incorrect use was recorded i n Wawrzont at (3;7): J a nie chcg za jajkiem odpoczywac T not want after egg to:rest' ( = T don't want to have a rest after I eat my egg') i n w h i c h the w r o n g preposition is used, i.e. the spatial za 'after' or 'behind' instead o f the temporal po 'after'. The gerundial form napisawszy 'after:having:written' was not attested i n the data. Most probably it is hardly used i n spoken language. I t occurs, however, i n "short stories" written by a 9year-old. The meaning o f 'before' cannot be rendered by ./^-clauses, and the special conjunction zanim is required w h i c h introduces a regular clause w i t h a finite verb. However, I d i d not find examples o f this structure i n any o f the data studied. The only example is that o f an inappropriate use, recorded from Wawrzont at (3;8): Kto mi wytrze tylek, zanim ja zrobi% kupk%? ' W h o w i l l wipe off m y bottom before I pass m y B . M . ? ' , where the natural order o f events referred to shows that he meant 'after' and not 'before'. This example provides negative evidence, yet the fact that the child tried to use zanim can be a signal that he has noticed the structure and w i l l soon learn i t . The notion o f simultaneity involves t w o kinds o f transformed structures w h i c h are infrequently found i n the data. One o f them is przy 'at' plus verbal noun i n the locative case found i n M i c h a l as early as (2;6): Dlaczego przy lezeniu na brzuszku nie chlupnie krwi? ' W h y at l y i n g on belly blood w o n ' t spurt (from the nose)?' and i n Wawrzont at (3;8): [sitting i n his bed and reading books] Dobrze siq czujq. przy spaniu T feel w e l l at sleeping'. Another construction involves the present gerund, and it is reported i n Rz^tkowska's daughter before the age o f (4;0): Wracajqc do domu spotkafysmy Czesig 'While:going:back home we:met Czesia'; i n W i t e k (3;7): Bos ty nie widziala rozmawiajqc z ciotkq wtedy 'Because you d i d n ' t see (it) while:talking to Auntie then', and also i n Wawrzont at (3;11): [Mother had promised h i m that she w o u l d go to sleep at the time when he is going to sleep, but she continues to read a book] Wawrzont: Spisz? ' You:sleep?' Mother: Ja tak spig czytajqc T am:sleeping while reading?' Wawrzont: Ty tak spisz czytajqc? ' Y o u are:sleeping while:reading' Mother: Tak ' Y e s ' Wawrzont: To ja b%d% nie czytajqc spai, tylko zamkn% oczka 'So I w i l l not while:reading sleep, but Tshallxlose eyes'. I t seems that the meaning o f simultaneity is more available to Polish children than that o f 'before' (constructions w i t h 'after' are not necessary since jak works equally w e l l ) .

660

Smoczyriska

6.2.3. Relative

Clauses

6 . 2 . 3 . 1 . Production. I n all the children studied the first relatives appear about 2;0, f o l l o w i n g an adult question ' W h i c h one?' or ' W h i c h k i n d o f ? ' , asked in order to obtain additional information about one o f the referents previously mentioned by the child. Children usually answer by using an isolated relative clause, or they precede i t w i t h the demonstrative antecedent 'this (one)', e.g. Kasia (2;0): Dqj mi chustkq ' G i v e me handkerchief Mother: Jakq? ' W h i c h (one)?', Kasia: co LEZY ' T H A T I S : L Y I N G ' , and M i c h a l (2;2): Daj mi tu laski moje 'Give me m y walking-sticks' Mother: Jakie laski? ' W h i c h walking-sticks?' M i c h a l : TE co SA W MOIM

KACIKU

'Those T H A T A R E I N M Y R O O M ' .

A t the same time children learn to construct full complex sentences and to use them when they suppose that some additional information is needed to establish unique reference, e.g. M i c h a l (2;3): A gdzie te naczynka co JE WYJMOWALEM? ' A n d where are those dishes T H A T T H E M I : W A S : T A K I N G : O U T ? ' and Kasia (2;5): [comments on a picture] O, gofy chlopczyk. A te co SIE BAWIA to nie sq gole. 'Oh, a naked boy. A n d those T H A T A R E : P L A Y I N G are not naked', also Kasia (2;6): [to her parents who came back home after an absence] Zdechla jedna rybka. A te co NIE ZDECHEY to pfywajq. ' D i e d one fish. A n d those T H A T N O T D I E D are:swimming'. A s can be seen from the above examples, contrary to Limber's (1973) observation concerning English-speaking children, Polish c h i l ­ dren produce subject-embedded relatives early on. Note the use o f the empty particle to at the beginning o f the second part o f the main clause. Children frequently use relatives also to explain what they mean i f they do not know the name o f an object, e.g. Jas (4;0); [explains what he wants to play w i t h ] Taki z dhigq rqczkq co MA DZIURKE, taki chc% co WCZORAJ 'Such w i t h a long handle T H A T HAS A H O L E , such I:want T H A T Y E S T E R D A Y ' . Note that such an incomplete elliptical relative as ' T H A T Y E S T E R D A Y ' instead o f ' T H A T I H A D Y E S T E R D A Y ' could be also used i n adult Polish. (To satisfy the reader's curiosity, I w i l l say that the object referred to was a strainer.) A l l the examples quoted above refer to some knowledge that the listener is supposed to have. I f a relative carries some new information about a referent, i t is embedded i n a k i n d o f " i n t r o d u c t o r y " matrix clause, e.g. Kasia (3;0): A ja jestem byk potworny co ZJADA DZIECI, a mamusia byczek mafy co NIE ZJADA D Z I E C I ' A n d I am b u l l terrible T H A T E A T S C H I L D R E N , and M o m m y b u l l l i t t l e T H A T N O T E A T S C H I L D R E N ' , Jas (2; 10): Ja mam szoferk$ co MA OKNO

T have cab

T H A T HAS W I N D O W ' , Kasia (2; 10): Ja widzialam takiego psa co GRYZIE i mnie nie ugryzl i patrzyt T saw such dog T H A T BITES and (he) me not bit and l o o k e d ' , and M i c h a l (2;2): Byisobie taki Jacus co SIE BAwih CORKA 'Once there was such Jacus T H A T P L A Y E D W I T H : D A U G H T E R ' . Moreover, although less frequently, rela­ tives o f place appear early, simultaneously w i t h the attributive type exemplified so far; e.g. the f o l l o w i n g complex construction, i n v o l v i n g a hierarchical com­ bination o f both types, M i c h a l (2;6): Czy Janek jest w kuchni GDZIE JEST STARY

6. ZEGAR

co ROBI

TiK-TAK?

T h e A c q u i s i t i o n o f Polish

661

' I s Janek i n kitchen W H E R E I S O L D C L O C K T H A T S A Y S

TICK-TOCK?'

Most children employ almost exclusively the relative pronoun co which is indeclinable, and i t requires a shadow pronoun or an antecedent copy to make explicit case relationships, although i n unambiguous contexts they can be omit­ ted. The relative pronoun który w h i c h requires case marking and a gen­ der/number agreement w i t h its antecedent occurs only sporadically. These re­ quirements seem to cause special difficulties i n constructing prepositional phrases such as ' i n w h i c h ' , 'on w h i c h ' , etc., e.g. M i c h a l (2;1): Jaki wspaniafy konf Ja go kiedys narysowatem KTÓRY SIE NA NIM JEZDZI 'What a splendid horse! I h i m once drew W H I C H RIDES O N H I M ' . Note the lack o f a case marker on KTÓRY and the shadow pronoun NA NIM, cf. correct NA KTÓRYM SIE JEZDZI ' o n w h i c h one rides'. A t 3;2, while looking for the car that he has lost, Jas said Ono byio na szafie malutkiej KTÓREJ

MAM RZECZY

NA TEJ ' I t was o n cupboard little W H I C H I

H A V E T H I N G S O N T H A T ' . I n Jas's statement there is a correct case ending and gender/number agreement (KTÓREJ), but the child could not front the preposition and introduced a shadow pronoun instead, cf. correct N A KTÓREJ MAM RZECZY ' o n which I have things'. I t seems that English forms i n v o l v i n g " s t r a n d i n g " o f prepositions w o u l d be easier to produce. A m o n g the children studied i t is only M i c h a l who mastered the correct use o f który and employed i t frequently i n later stages (also w i t h prepositions); other children avoided i t throughout the whole period studied, e.g. Jas (4;3): [telling his mother what he has seen] Mamusiu, widzialas ten autobus co w TYM AUTOBUSIE

NIE BYEO

KIEROWCA

KIEROWCA

i ten autobus C O W T Y M A U T O B U S I E N I E B Y L O

ciqgnqi drugi autobus?

Mother: Pewnie jeden siq zepsul. Jas: Nie

zepsul sie tylko nie by to w tym autobusie kierowca co CIAGNAL

DRUGI

AUTOBUS.

Jas: ' M o m m y , you:saw that bus T H A T I N T H A T B U S T H E R E W A S N O D R I V E R and that bus T H A T I N T H A T B U S T H E R E W A S N O D R I V E R was .pulling another bus?'

Mother: 'Probably one broke d o w n ' . Jas: ' I t not broke d o w n , only there was i n that bus no driver T H A T W A S : P U L L I N G A N O T H E R BUS. ' Note the numerous repeti­

tions o f the whole antecedent. The data recorded i n w r i t i n g do not provide adequate cues for the restrictive versus nonrestrictive distinction, w h i c h i n Polish is clearly marked by different intonation contours. However, i n view o f the contexts i n w h i c h the recorded relatives occur, i t is highly improbable that any o f them could be classified as a definitely nonrestrictive. 6.2.3.2. Comprehension. The large corpus o f relatives recorded from the K r a k ó w children needs a thorough analysis before the developmental tendencies can be systematized. General properties o f children's production o f relatives as described above can shed some light upon numerous comprehension studies which result i n conflicting observations (see Bowerman, 1979). I n the case o f Polish, i t can be expected that Aróry-relatives although difficult to produce, do

662

Smoczyñska

not involve difficulties i n parsing, because o f the large amount o f explicit cues they provide. The only experimental study concerning Polish relatives is that o f Zabielski (1974), w h o tested comprehension o f different kinds o f sentences w i t h 3-, 4-, 5-, and 6-year-olds (20 subjects i n each age group). For relatives he used the five following items:

1. Kazik glaszcze psa, który je kielbas%. ' K a z i k is stroking a dog that eats sausage' 2. Tata rozmawia z Jackiem, który trzyma pilk%. 'Daddy is talking to Jacek, who is holding a b a l l ' 3. Jas goni kota, który uciekl na drzewo. 'Jas is running after a cat w h i c h had climbed a tree' 4. Mama podaje jablka,

które przyniosla

na talerzu.

' M o m m y is serving apples that she had brought on the plate' 5. Kasia spotkala Jacka, które go boli glow a. 'Kasia met Jacek whose head is aching'

For each item three pictures were presented, and the children had to choose the one which corresponded to the sentence uttered by the experimenter. Unfortunately, there is no exact information on the pictures used, except for one example (item 4 ) , where the t w o " w r o n g " pictures were those o f a boy serving some apples on a plate to a woman and a boy holding an empty plate, being given some apples by a woman ( w i t h no plate). The ratios o f the correct choice for all the five items were: 65% w i t h 3-year-olds, 60% at 4 , 78% at 5, and 90% at 6. These global results indicate that the level o f comprehension as measured by Zabielski was relatively high w i t h children aged 3 - 4 , and then gradually i n creased to reach 90% criterion at 6. I n view o f these results, and expecting that Polish /tfory-clauses w o u l d be easy to understand, I arranged an experiment investigating m y son at age 3; 10, w h o freely produced co-relatives whenever required by the context. Although only one child was tested, and i n a quite informal w a y , the result obtained is surprising enough to be reported. W h e n presented w i t h four classical types o f reversible fctóry-relatives (SS, SO, OS, and O O , e.g. 'The cow that kissed the horse kicked the rabbit', etc.) and asked to repeat each o f them, he d i d not make any e r r o r s . However, when asked to act-out the same sentences, i n most cases he was unable to do so, although additional information was supplied by gender marking in past tense verbs. I n his w r o n g answers he used all kinds o f "strategies," described i n the literature and summarized by Bowerman (1979). What I found 32

33

0 n the other hand, he was quite unable to repeat utterances in which który was preceded by a preposition. Some positive answers were obtained with SS sentences, but they were not consistent enough. 3 2

33

6.

The A c q u i s i t i o n o f Polish

663

most puzzling, however, were not his scored answers but the nonscored behavior which accompanied the performance. I f he used the strategy o f main-clauseonly, he acted it out quickly and, quite satisfied, looked at me asking for approval. I n all the remaining cases, his behavior showed total confusion and apparent inability to retain the whole information given i n the sentence. He asked me to repeat the sentence several times or asked ' W h o kissed?', ' W h o kicked?', etc. after he had performed one o f the actions, and still he was unable to solve the task correctly. Clearly, what he found difficult was not sentence processing but rather information processing. A n experimental situation is o f course artificial i n some way, and significant discrepancies can occur between correct performance i n natural contexts and correct solving o f experimental tasks. I n natural conversation, relatives are un­ derstood (and produced) when they are necessary to transmit a message. I t can be of no significance to the child whether it is the cow who d i d the kissing or the rhinoceros. A more relevant fact, however, is that the acting-out task is not congruent w i t h the very function o f relative clauses, w h i c h usually serve to downgrade one clause w i t h respect to another. W h e n both clauses are equally important, and both provide new information, they are coordinated. The relative transformation reduces one clause to the function o f an attribute, w h i c h either refers to some prior knowledge or points at some observable features o f the referent. I n both cases i t serves to help the listener i n identifying the referent uniquely. Thus the matrix clause usually contains new information (to be remem­ bered or acted out), whereas the relative clause contains the information to be retrieved from memory (old information) or perceptually checked. Even i f it is new information, as i n the latter case, it is not intended as a message to be remembered, it is totally instrumental w i t h respect to identifying the referent o f its antecedent. N o w , the experimental task o f acting out B O T H actions is com­ parable w i t h testing the comprehension o f The black cow is swimming i n w h i c h the child w o u l d be scored for painting the cow black. I f the child does not pay equal attention to both clauses, i t is precisely because he knows the function o f relatives. The strategy o f acting-out the main clause only is the most intelligent answer to an inappropriate task requirement. I carried out another " e x p e r i m e n t " i n which Wawrzont (3;10) had to choose by means o f a relative clause the correct toy out o f 10 toys presented to h i m . The relative clause referred to an action w h i c h had been performed by one o f the toyanimals before the presentation o f the sentence, e.g. 'The animal w h i c h has fallen d o w n kissed the elephant' and, subsequently, 'The cow kicked the animal that kissed the elephant' or even 'The animal that the elephant kicked kissed the animal that j u m p e d over the l i o n ' . The child made some errors, but i n general his responses were correct and what is most important, there were no signals o f confusion while processing these sentences. A l l these observations point to the fact that the acting-out tasks do not check comprehension o f relatives but much more beyond that. I n view o f the existence

664

Smoczyhska

of copious naturalistic data i n Polish, and also because o f the relatively easy structure o f the relatives, i t w o u l d be profitable to carry out regular experiments w i t h Polish children i n w h i c h not only the form o f relative clauses but also their function were taken into account. 6.3.

Delayed

Acquisition

T w o kinds o f phenomena can be discussed here. One is the Tate mastery" o f forms which emerged relatively early, w h i c h is revealed i n "late errors." A n ­ other phenomenon is that o f late emergence o f a category w h i c h is present i n the child's input. Typical late errors involve difficulties w i t h all kinds o f irregular forms, the choice o f verb stem being the most frequent error o f this type (see Section 4 . 5 . 1 ) . A m o n g regular forms, most difficult categories—frequently not yet fully mastered by the age o f 3 or 4 or beyond—are the f o l l o w i n g : (1) the genitive plural o f nouns (Section 4 . 3 . 2 . 2 ) , (2) the order o f morphemes i n condi­ tional forms (Section 4 . 6 . 3 ) , (3) the first person plural o f the past tense (Section 4.4.4), and (4) the distribution o f dwaldwie ' t w o ' (Section 4 . 5 . 2 ) . I t is perhaps worth noting that the very same forms appear as innovations i n various Polish country dialects (cf. Bartnicka-D^bkowska, 1965). A s far as late emergence is concerned, apart from some kinds o f temporal clauses, w h i c h have been dis­ cussed i n Section 6.2.2.3, several categories w i l l be discussed below. 4

6 . 3 . 1 . Virile vs. Nonvirile Gender Distinction (in Plural). The distinction of virile/nonvirile gender is relevant for the following linguistic forms: (1) n o m i ­ native plural (where virile nouns have a specific and highly irregular distribution of endings), (2) accusative plural ( w i t h accusative identical to nominative for nonvirile nouns, and accusative identical to genitive for v i r i l e ) , (3) gender agree­ ment i n adjectives, pronouns and past tense verbs (see Table 6.6), and (4) forms of numerals. I have not yet w h o l l y analyzed the data concerning this distinction. A l l that can be said at this point is that virile nouns and the problems involved i n their use do not occur i n the child's speech from the very beginning. Note that potential referents for virile gender (men i n plural) are rather limited i n number, as op­ posed to all other elements o f reality w h i c h have nonvirile gender. Moreover, the use o f specific virile nouns requires broader social knowledge ( i n early stages only the plural o f the semantically " n e u t r a l " pan 'man, sir' occurs, w h i c h is typically pany instead o f panowie). I n general errors emerge after 3 or 4 years, and they occur w i t h respect to all the forms listed above. Mystkowska (1970) reports gender-agreement errors i n plural forms as typical for 6-year-olds. The most interesting category is the accusative. The rule is inconsistent w i t h that for singular where all animate masculine nouns have the accusative-equalsgenitive pattern. I t can be therefore predicted that the names o f animals o f masculine gender w i l l cause difficulties i n the accusative plural since they are

6.

The A c q u i s i t i o n of Polish

665

[ + M A S C ] and [ + A N ] but [- H U M ] , and therefore [- V I R ] . Indeed such errors are frequent, e.g. Janek (6;6) [speaking about snakes]: Zeby ich zabic 'imorder them t o : k i l l ' ( = 'to k i l l t h e m ' ; cf. correct Zeby je zabic). Here Janek uses the genitive plural ich ' t h e m ' f o l l o w i n g the accusative-equals-genitive pattern, rather than the correct nominative plural je ' t h e m ' , according to the appropriate accusative-equals-nominative for nonvirile nouns. The results obtained for Pol­ ish could be then compared to those for Russian and Serbo-Croatian. A l l three languages have the same distribution o f accusative forms i n singular w i t h clearcut differences i n plural. I n Russian all [ + A N ] nouns, irrespective o f gender, have the accusative-equals-genitive pattern, and i n Serbo-Croatian all masculine nouns, irrespective o f animacy, have a specific accusative plural ending w h i c h is distinct from both the nominative and genitive plural, whereas feminine and neuter nouns have the accusative-equals-nominative pattern. Such a contrastive analysis could perhaps elucidate more general principles o f language acquisition. 6.3.2. Concessive Clauses. These appear rather late i n the data, and not i n all children. Przetacznikowa (1968) noted some single instances i n Jas during his sixth year. I n other data they were not found except for M i s at (3;7), e.g. A zajqc jest on, choc zajqc wcale nie jest chlopakiem ' A n d a hare is a he, although hare is not a boy at a l l ' , and M i c h a l , w h o even tried to use a concessive clause at 3;0 w i t h the f o l l o w i n g result: J a nie wiem jak sig robi most, choc umiem to robic T don't k n o w how to make a bridge, although I can do i t ' , but from 3;8 M i c h a l produced correct forms, e.g. (4;1): Biedny wilczek. Chociaz on jest niedobry, to ja jego zatujg 'Poor w o l f . A l t h o u g h he is bad, I pity h i m ' ; (4;1): Ale ja widziatem jak z zabitego ktos sciqgai skorg, chociaz by I dobrym czhwiekiem. ' B u t I saw somebody skinning a k i l l e d one, although he was a good m a n ' ; (4;2): Bardzo ich lubii, choc ich nie znai. Nieznajomych milowal ' H e liked them very much, although he d i d n ' t k n o w them. He loved the strangers.'; and (5;3): Te ryby w beczce sig nie pozjadajq, choc majq dopfyw do siebie 'Those fishes i n the barrel w o n ' t eat one another, although they can reach one another'. These data show that concessives are cognitively not as difficult as was expected. The failure to observe concessives i n the data could be caused either by their l o w frequency (as compared w i t h other kinds o f adverbial clauses), or by the fact that parents do not use them when speaking to their children. Additional evidence for this claim is supplied by Zabielski's (1974) comprehension tests. Children aged 4 succeeded i n 4 0 % o f the answers consisting o f choosing the right picture out o f three presented for each o f five sentences, e.g. Piesek ucieklmimo ze bytprzywiqzany do drzewa 'The dog ran away although he was tied to a tree'. This result approximated the chance expectation. Five- and 6-year-olds d i d much better g i v i n g correct responses—in 6 2 % and 82% cases, respectively. Similar results were obtained w i t h nominalized versions o f concessives, e.g. mimo przywiqzania do drzewa 'in:spite:of being:tied to (a) tree'.

666

Smoczyhska

6.3.3. Comparison, Comparatives, and Superlatives. The earliest occur­ rences o f comparative degree involve the form wiekszy 'bigger' as an egocentric expression o f power, e.g. M i c h a l (2;1); J a jestem wiekszy od tatusia. T am bigger than Daddy' [disappointed by the reaction o f his listeners, he looks around, notices a pair o f scissors and says] Od nozyczek jestem wiekszy T am bigger than the scissors'; M i c h a l (2;4) [after he said that his cousin, Pas, is a little boy] Mother: Pas jest wiqkszy od ciebie 'Pas is bigger than y o u ' M i c h a l : Tak. Pas jest wiekszy ode mnie, ale ja jestem wiekszy od Pasia 'Yes. Pas is bigger than me, but I am bigger than Pas'; M i c h a l (3;0): [to his father, w h o is drawing] Ladniej rysujesz niz mamuma ' Y o u draw better than M o m ' Father: Mamuma tez ladnie rysuje ' M o m can also draw w e l l ' M i c h a l : Tak. Mamuma ladniej od ciebie rysuje, ale ty ladniej od mamumy 'Yes. M o m draws better than y o u , but you (draw) better than M o m ' . W h i l e the example o f 2;4 could involve an egocentric meaning 'He is bigger, but I am more important', the latter example clearly shows that the notion o f comparison itself is misunderstood, and the comparative degree is used i n an absolute sense. A l l the children studied made errors that consisted o f using the positive form in comparative constructions, e.g. Janek (4;3): Ja jestem duzy niz wujek Stas T am big than uncle Stas'; and, Janek (4;3): Mis ma malo wody niz ja ' M i s has little water than T . O n the other hand, the f o l l o w i n g utterance was recorded from M i c h a l , w h i c h made explicit the relative value o f compared terms: M i c h a l (3;9): [giving cake to his mother] J a ci to dam. To jest wieksze, a ja sam mam mniejsze. Ale tez duze ' I ' l l give you this one. I t is bigger, and I have a smaller one myself. But (it is) also b i g ' . Even so, later on he continues to use the incorrect form: M i c h a l (4;2): Pies to ghiptas od wilka ' A dog is a fool than a w o l f . Such errors persist all throughout the period studied, besides correct forms, also for the superlative. Another case is that o f the expression o f "sameness." U n l i k e English or French, studied by Karmiloff-Smith (1977), Polish has t w o separate expressions for the t w o meanings: 'same one' is expressed by ten sam 'this same'; 'same k i n d ' by taki sam 'such same'. Surprisingly, children happen to make mistakes when using these forms, although the correct use normally prevails, e.g. one o f my t w i n sons at 3;8 said, Zobacz, taki sam! ' L o o k , the-same-kind-of!' when noticing the person we had met t w o hours earlier. A n d , at 5;5 he asked, Czy pani jest soma? w h i c h literally means 'Is the lady alone?'. He obviously meant to say taka sama as he intended to ask whether the pregnant woman seen i n the shop was as pregnant as his mother was at the moment. Such isolated occurrences o f the incorrect use o f those expressions indicate that there can be someting difficult i n the cognitive distinction between the t w o kinds o f sameness. Thus Karmiloff-Smith's (1979) interpretation, w h i c h at­ tributed the errors made by French-speaking children mainly to linguistic factors, should perhaps be tested against experimental results obtained for a language like Polish w h i c h makes the distinction more clearly.

6.

The A c q u i s i t i o n of Polish

667

CONCLUSIONS 7. U n i v e r s a l O p e r a t i n g P r i n c i p l e s a n d t h e A c q u i s i t i o n of Polish The data presented i n preceding sections w i l l now be compared w i t h what have been claimed to be universal principles used by the child while processing linguistic information (Slobin, 1973). Some comparisons w i l l also be made w i t h available data on other languages. 7 . 1 . Operating of

Principle

A: ''Pay attention

to the

ends

words."

Slobin's Operating Principle A refers to receptive processes ( " p a y atten­ t i o n " ) but it is mostly based on the data on production and not on comprehen­ sion. As for Polish children, they seem to obey this principle i n production. They produce w o r d endings earlier than preposed grammatical morphemes (preposi­ tions and prefixes). I t is reasonable to assume that i f the endings are produced early, they must have been noticed before, according to the general rule that comprehension precedes production. One cannot, however, preclude the pos­ sibility that preposed grammatical morphemes are noticed equally early, but for some reason or other they are not used i n production until l a t e r . 34

I n the Polish language the modification o f a w o r d ending is the most wide­ spread pattern o f grammatical marking, w i t h inflectional endings being used much more frequently than preposed morphemes. I n this situation it is impossi­ ble to say whether the facts observed are due to a universal a priori principle, to which Polish happens to f i t , or to the child's prior experience w i t h this particular language, w h i c h has led to a similar generalization. I f we had the acquisition data of a language w h i c h uses mainly preposing, and i f children, who learn such a language, ignored the commonly-used device for a less frequent postposing device, we could regard Operating Principle A as proven. Although it does not apply directly to the beginning and ends o f W O R D S , there is some less clear-cut evidence w h i c h challenges the universal validity o f Operat­ ing Principle A . I t can be found i n the acquisitional sequence o f 14 EngUsh grammatical morphemes as established by B r o w n (1973) and de Villiers and de Villiers (1973). W h y are the PREPOSITIONS in and on among the earliest gram­ matical morphemes to appear (rank order 2 and 3)? W h y are the PREPOSED articles a and the (rank order 8) acquired earlier than the regular past ending -ed (rank order 9) and the third-person ending -s (rank order 10)? The answer to these questions can be found i n the general characteristics o f the English language, i n which preposed and free grammatical morphemes are used relatively frequently. [ W E I S T ] We are now analyzing the prefixation in the aspectual and locative systems. Our preliminary findings indicate that children from 1;6 to 2;6 develop prefixes surprisingly rapidly, given what has been said about the child's preference to pay attention to the ends of words. [WEIST] 34

668

Smoczynska

As demonstrated above, English-speaking children discover this morphological pattern relatively early i n comparison to Polish children, even i f this does not f i t the universal Operating Principle A . 7.2. Operating Principle B: "The phonological words can be systematically modified."

form

of

As was demonstrated above, children not only k n o w T H A T the forms o f words can be modified, but also H O W they are usually modified i n a particular lan­ guage. W e should try to trace back the genesis o f this knowledge by analyzing the stages w h i c h precede the productive use o f morphological markers. Once the child has gained some basic knowledge o f phonology w h i c h makes it possible for h i m to recognize sequences o f sounds, his efforts concentrate on segmentation o f the flow o f speech. This early segmentation cannot be done on the whole string o f sounds except for some rough procedures o f establishing prosodic units. What the child is most interested i n is to extract those sounds which can be paired w i t h some referents present i n the situation. This task can be more or less difficult depending on the language the child is dealing w i t h . I n some languages, the sounds for such referents as d o g ' or 'wash' do not vary depending on the syntactic context except for the fact that they are sometimes accompanied by some additional sounds w h i c h precede or follow the w o r d without affecting its phonological and prosodic shape. I n other languages, there is more or less considerable variation W I T H I N the w o r d conceived as a prosodic unit. The w o r d can be divided into a constant part (a lexeme) and another part which is subject to variation (a grammatical morpheme). I n some languages the constant part itself can include some amount o f extra variation (stem alternation, etc.). I n order to recognize the meaningful referential parts o f the flow o f speech, such as ' d o g ' or ' w a s h ' , the child must have developed stategies o f I G N O R I N G possible variation involved i n their forms (cf. Berman, 1985). Paradoxically enough, an early strategy used by Polish children can be formulated as: " I g n o r e the ends o f w o r d s , " since i t is precisely the ends o f words which are subject to variation i n most cases. Once the child is able to detect lexemes w i t h some ease, his attention can be turned to what has been ignored so far. 4

What is puzzling is that i n spite o f enormous differences between particular languages as to the amount o f morphological variation involved, the one-word stage does not seem to be especially protracted i n any particular language. I n m y opinion, i t is at the end o f this stage that every child has gained passive k n o w ­ ledge o f the most widespread patterns o f morphological variation used i n the language under acquisition. This knowledge is used i n later stages, when the child starts to make productive use o f morphology. I w o u l d like to emphasize the important role o f the F R E Q U E N C Y o f mor­ phological (or syntactic) P A T T E R N S used i n a language. This factor has not usually been taken into account as only the frequency o f particular forms or words has been investigated. T o make this point clear, let us compare mor-

6.

T h e A c q u i s i t i o n o f Polish

669

phological development i n English and i n Polish. One w o u l d expect the mor­ phology o f English, being relatively simple and i n v o l v i n g a limited number o f grammatical morphemes, to be easier to learn than the complex and highly irregular morphological system o f Polish. The facts do not confirm such an expectation, as Polish children start to use inflections productively much earlier than English-speaking children do. They do so because they are forced to notice that the most frequent pattern is W O R D = S T E M + E N D I N G . English-speaking

children frequently hear forms where W O R D = S T E M , and instances o f W O R D = S T E M + E N D I N G are relatively less frequent. O n the other hand, Turkish children, who deal w i t h a language characterized b y a highly regular agglutinative mor­ phology, use inflections productively as early as the one-word stage (Aksu-Kog & Slobin, 1985). 7.3.

Operating of words

Principle C: "Pay attention and morphemes."

to the order

There is one preliminary objection as to the formulation o f this Operating Principle, and that is that the w o r d order w i t h i n an utterance and the order o f morphemes w i t h i n a w o r d are t w o distinct things. This objection is confirmed by acquisition data from many languages. There are no instances o f incorrect order­ ing o f morphemes w i t h i n a w o r d , whereas examples o f the w r o n g w o r d order are rather numerous. This demostrates that children organize their utterances b y taking words as basic units instead o f b y ordering morphemes like beads on a string. W e cannot analyze a child's utterances i n terms o f morphemes without taking into acount the status o f particular kinds o f morphemes w i t h respect to words. There are lexemes w h i c h are equal to words and lexemes w h i c h are not autonomous words. There are grammatical morphemes which cannot occur inde­ pendently (bound) and those w h i c h can be used as words (free). Moreover, since most words are also prosodic units—since they have independent stresss— special attention must be paid to clitics, w h i c h have the status o f words without this prosodic feature. I have not tried to describe the use o f w o r d order by Polish children. The task is very hard, since the rules governing its use i n adult language are so complex that they are not sufficiently w e l l described i n linguistic literature. A l l that can be said at this point is that there are large individual differences w i t h respect to adherence to a r i g i d w o r d order. A n y w a y , the freedom o f w o r d order is much larger than i n English-speaking children, w h o seem to discover the grammatical relevance o f the word-order device very early. As for the pragmatic use o f w o r d order by Polish children, i t needs further detailed study.

[WEIST] There is no question about the fact that children can process w o r d order information, and children learning English rapidly acquire a strategy w h i c h involves the use o f word-order information to recover basic relations. I t has been argued that the basic conceptual relationships w h i c h are fundamental to sentence processing are ordered, e.g., agent-action-object-recipient (Bruner, 1975, p .

670

Smoczyhska

17). The hypothesis implies that those patterns o f nouns and verbs w h i c h corre­ spond to the ordered basic relations lend themselves to a " n a t u r a l " one-to-one mapping. I t follows from this argument that order information should precede inflectional information as a cue for basic relations. This argument has been made explicitly for a Slavic language by Radulovic (1975, p . 117): " C o o r d i n a ­ tion o f S V O w o r d order is a prerequisite to the recognition o f structural rela­ tionships w h i c h may be varied i n productive language by the use o f inflections." In Polish, w o r d order has the discourse function o f distributing given and new information. According to Szwedek (1976), the least marked sentence pattern is G I V E N I N F O R M A T I O N I N I N I T I A L POSITION A N D

N E W I N F O R M A T I O N I N F I N A L POSI­

T I O N W I T H H I G H STRESS. Syntactic and semantic functions are related to the case system as described i n Table 6.1 earlier. I f w o r d order is naturally related to ordered semantic functions, children learning Polish should be faced w i t h the predicament o f first learning to express and retrieve semantic functions w i t h word order and then later learning the discourse function o f w o r d order and at the same time having to discover the relationship between the case system and syntactic-semantic functions. Weist's (1983b) experiment on the role o f w o r d order, inflectional information, and given-new discourse patterns is relevant to this issue. Ten children at t w o age levels (2;6 and 3;6) participated i n an experiment i n which the children were required to act out sentences w i t h toys (cf. Slobin, 1982). The children were presented w i t h a sequence o f t w o sentences: a context sentence followed by a test sentence. The context sentence established the given information. The test sentences were semantically reversible. One half o f the test sentences were inflected and the other half were " u n i n f l e c t e d " (i.e. inflected w i t h a zero morpheme where the nominative equals the accusative case). The following t w o sentences were typical o f the inflected and uninflected target sentences: (1) dziewczynk-a obudzi-l-a-0 chlopc-a 'girl-NOM:FEM:SG wake:up-PAST-FEM-3SG b o y - A C C : M A S C : S G ' ( = T h e g i r l woke up the b o y ' ) and (2) autobus-0 stuknq-l-0-0 samochod-0 b u s - N O M / A C C : M A S C : S G hitP A S T - M A S C - 3 S G c a r - N O M / A C C : M A S C : S G ' ( = T h e bus hit the car' or T h e car hit the bus'). Thus, the children received the f o l l o w i n g six types o f problems: (1) given-new & S V O , (2) given-new & O V S , (3) new-given & S V O , (4) newgiven & O V S , (5) given-new & N j V N , and (6) new-given & N V N . 4

2

1

2

When the test sentences contained unambiguous inflectional information (problems 1-4), the children primarily used the strategy N O M I N A T I V E C A S E = A G E N T w i t h the f o l l o w i n g accuracy: 2;6 at 87% and 3;6 at 9 2 % . W h e n the test sentences were uninflected (problems 5 & 6 ) , the children used priority i n w o r d order, i.e. an N j = A G E N T solution, w i t h the f o l l o w i n g performance levels: 2;6 at 78% and 3;6 at 82%. The manipulation o f given and new information d i d not have a strong impact on the outcome o f the experiment. I t is possible, however, that the procedure failed to maintain contextual relevance throughout the period of acting out the target sentences w i t h toys. I n any event, the children demon-

6.

The Acquisition of Polish

671

strated the capacity to use inflectional information as w e l l as word-order infor­ mation, w i t h no indication that the word-order strategy is more primitive. I f anything, precisely the opposite is true. Children w h o are approximately 2 h years o l d and learning highly inflected languages such as Polish or Turkish (Slobin, 1982) are more accurate i n their capacity to process affixes than A m e r i ­ can or Italian children are i n their capacity to process order information. x

We have begun to analyze the longitudinal component o f the Poznah research project i n ways w h i c h w i l l shed some light on the relevance o f word-order information for children from about V/2 to 2Vi years o f age. Some o f the p r e l i m i ­ nary findings have been reported (Weist, 1982). W e are w o r k i n g w i t h the hy­ pothesis that child language involves pragmatic, semantic, and syntactic distinc­ tions beginning w i t h the one-word phase. Proportionally, pragmatic distinctions (e.g. statement versus request) and semantic distinctions (e.g. agent versus pa­ tient) are the most prominent aspects o f this early phase o f language acquisition. However, Polish children begin to show evidence for a primitive concept o f the syntactic function SUBJECT at about V/2 years o l d . A t this age, children begin to use the subject properties o f case, agreement, priority (subject before object), and animacy. Other potential subject properties are absent (see Keenan, 1976). In order to understand the relevance o f w o r d order for Polish children, we have to look at the property o f priority i n the context o f other properties o f subjecthood. The nominative case emerges first i n Polish, but we cannot argue that i t marks the notion o f subject until i t contrasts w i t h other cases. A s the case system evolves, we can be more confident that the nominative case is a productive marker. The nominative, accusative, vocative, and genitive cases are normally found before 1;6. The locative, instrumental, and dative cases evolve somewhat later, e.g. for one o f the Poznah children, Bartosz, the last t w o cases were instrumental (found at 1;8) and dative (at 1;9), and for Zarebina's (1965) Hanna, the last t w o cases were dative (at 1;8) and locative (at 1; 10). Agreement has as its point o f departure a frozen third-singular form. I n singular, the first-, second-, and third-person distinctions are found w i t h numerous varied examples during the period from 1;7 to 1;10, and w i t h the tense-aspect system emerging at a similar time, agreement i n gender i n the past and future is added. For Zarebina's Hanna, the distinction between second person and third person occurred between 1;7 and 1;9 and the distinction between second person and first was found at 1; 11. Hence, as for case and agreement, the initial signs o f these subject proper­ ties are found at IV2 years and the full system is productive i n many children before 2 years-of-age. In contrast, priority is not a salient feature o f the child's early notion o f the subject function. Table 6.10 was taken from our preliminary analysis o f the productions o f t w o children (see Weist, 1982). D u r i n g the period from 1;7 to 2;8 the most frequent utterances containing a finite verb are single w o r d utterances w i t h the subject pronoun deleted. Utterances w i t h t w o nouns and a verb are relatively infrequent. Given verbs i n the indicative mood, the set o f single noun

672

Smoczyhska TABLE 6.10 Verb Plus Noun Combinations in the Indicative M o o d Produced by T w o Polish Children in 45-Minute Caretaker-Child Interactions (from Weist, 1982) Bartosz

Wawrzon

Pattern

1;7

1;8

1;9

1;11

2;0

2;2

2;4

2;5

2;6

2;7

2;8

NOUN.NOM VERB VERB—NOUN : N O M NOUN:ACC—VERB VERB NOUNlACC

15 13 5 3

4 1 10 6

5 2 5 4

2 1 8 9

1 2 4 1

6 5 14 16

1 1 4 11

8 8 5 7

11 5 4 6

14 3 0 8

20 9 2 6

plus verb combinations during this period is sufficiently large to provide a basis for comparison, and the f o l l o w i n g noun-plus-verb combinations were dis­ tinguished: N O U N : N O M + V E R B , V E R B + N O U N : N O M , N O U N : A C C + V E R B , and V E R B + N O U N : A C C . Given the combinations w i t h a noun i n the nominative case, the noun failed to gain priority i n 4 1 % o f the sentences for Bartosz and i n 34% o f the sentences for W a w r z o n . I n spite o f conclusions to the contrary, Radulovic (1975) reported similar findings for children learning SerboCroatian. W h i l e Polish (or maybe Slavic) children develop a priority bias for nouns i n the nominative case, there is no period during w h i c h w o r d order is artificially restricted. W o r d order is N O T used as a remedial device to deal w i t h the subject function while inflectional apparatus is being acquired. On the con­ trary, the acquisition o f inflectional distinctions appear to precede the priority bias. Our preliminary w o r k indicates that Polish children have a tendency to place new information i n initial position (see also Bates, 1976). D u r i n g the period from 1;6 to 2;6, children change from a N E W - G I V E N to a dominant G I V E N N E W pattern. The nominal element w i t h the subject function is often the given information. Hence, the development o f a subject priority bias may come about coincidentally because children are learning how to distribute given and new information. [WEIST] 3 5

7.4. Operating Principle rearrangement of

D: "Avoid units."

interruption

or

The evidence given in Slobin (1973) as support for this Operating Principle is not satisfactory, since other explanations can be given for the facts cited. Inver­ sion rules are difficult because they contradict other rules w h i c h have been [WEIST] Verbs in the imperative mood occur most frequently with nouns in the accusative case (see also Smoczyhska, 1976a), and while we have found both verb-ACC and ACC-verb patterns, verb-ACC patterns are clearly dominant. [WEIST] 35

6.

The A c q u i s i t i o n of Polish

673

acquired earlier. I t is the contradiction itself which causes difficulties and not the rearrangement o f w o r d order as stated i n the Operating Principle D . The evidence quoted as support for the Universal D 2 ("whenever possible, discontinuous morphemes w i l l be reduced to, or replaced by continuous mor­ phemes") concerned w i t h discontinuous morphemes is not convincing, since many other factors interfere. For instance, i n the French formula o f negation ne . . . pas only the element pas is stressed, and i n colloquial adult language the element ne is practically absent. The fact that Slavic children omit prepositions and not the endings has to do w i t h the status and frequency o f free vs. bound morphemes as w e l l as those o f preposed vs. postposed morphemes. I n general, discontinuous morphemes are used only marginally and as such they can be reduced to one part i n order to conform w i t h other more frequent morphological patterns. Insertion o f a w o r d into an already established pattern does not seem to be difficult i n any language. There are no cases o f adding a preposition to a sentence as a whole instead o f placing it before the appropriate w o r d , nor o f adjective modifiers placed at the beginning or the end o f a sentence instead o f being placed close to the head. The only example is that o f sentence-external negation appear­ ing i n some English-speaking children (Bellugi, 1967). The account o f this phenomenon, however, is not conclusive (cf. B l o o m , 1970; de Villiers & de Villiers, 1979). It should also be noted that i n an agglutinative language such as Turkish, where long strings o f grammatical morphemes occur, the ongoing acquisition o f morphology involves the insertion o f newly acquired morphemes into the w o r d form used earlier (thus, interruption o f a linguistic unit) and this fact does not seem to create any serious difficulties. 7.5. Operating Principle E: "Underlying relations should be marked overtly

semantic and clearly."

The role o f perceptual salience, as indicated i n the Universal E l ( " A child w i l l begin to mark a semantic notion earlier i f its morphological realization is more salient perceptually (ceteris p a r i b u s ) " ) , w o u l d predict that longer endings should be acquired before shorter ones. Accordingly, Polish children should acquire such endings as the dative singular masculine -owi or the instrumental plural -ami before other, mostly monosyllabic endings, but this is not the case. I t should be noted that the formulation itself o f this universal points at the recep­ tive, perceptual properties o f some markers, whereas facts given as support for the universal concern production. I n order to explain those facts we should rather take into account the prosodic consequences o f using an ending or omitting i t . I f an ending is S Y L L A B I C , w h i c h is the case for most Polish inflections, its omission affects the general prosodic shape o f the w o r d form. O n the other hand, omitting asyllabic endings does not involve such deformations. This factor cannot be ignored when comparing morphological development i n various languages.

674

Smoczyhska

As for the Universal E2 ( " T h e r e is a preference not to mark a semantic category by 0 ( " z e r o m o r p h e m e " ) . I f a category is sometimes marked by 0 and sometimes by some overt phonological form, the latter w i l l at some stage, also replace the 0"), i t should be said that the lack o f tolerance for zero-endings is typical for those languages w h i c h have overt syllabic endings i n most w o r d forms. English-speaking children are much more tolerant i n this respect than Polish children. I f they were not, they should not only replace two deer by two deers, w h i c h is a simple regularization, but also invent the missing markers for, say, the first and second person o f verbs, w h i c h they do not. The explanation is rather simple: for Polish children zero-endings are exceptions, whereas for En­ glish-speaking children it is rather the overt marking w h i c h is exceptional. C h i l ­ dren's behavior therefore can be explained on the basis o f Operating Principle F: " A v o i d exceptions." Universal E3 ( " I f there are homonymous forms i n an inflectional system, those forms w i l l tend not to be the earliest inflections acquired by the child; i.e. the child tends to select phonologically unique forms, when available, as the first realization o f inflections") does not seem to be supported by the data. Englishspeaking children should be reluctant to use the multifunctional -s, but they are not. Likewise i n Polish, the ending w h i c h serves a number o f functions i n the noun-declension system (see Table 6.2), is among the earliest endings used by children. 7.6. Operating

Principle

F: "Avoid

exceptions."

This Operating Principle is without any doubt universal. Its universality, however, is by no means restricted to language acquisition, but is present i n every domain o f cognitive development i n w h i c h the child is trying to generalize on the basis o f common experience, while ignoring exceptional instances. On the other hand, particular universals given under the heading o f this principle concern t w o distinct things. Some o f them refer to true exceptions (i.e. instances where a rule is not based on any explicit criteria), while others are concerned w i t h subrules valid for a narrower range o f cases than that o f a superordinate rule and based on some explicit criteria (cf. the discussion o f irregularity vs. formal complexity i n Section 4 ) . Exceptions—and I mean true exceptions—are a nuisance for the child and he copes w i t h this difficulty simply by ignoring them or by overregularizing correct irregular forms w h i c h have been learned earlier as separate items. Quite a different problem is that o f dealing w i t h formally complex systems, i.e. those w h i c h involve a large number o f subrules and/or complicated hier­ archies o f subrules to be applied. The number itself o f rules involved does not seem to be especially relevant. What is important is the I N T E R N A L C O H E R E N C E o f the given (sub)system. A system is internally coherent i f its subrules do not contradict more general rules and i f some proportions are preserved throughout the whole system. For instance, an English-speaking c h i l d , w h o has already

6.

The A c q u i s i t i o n of Polish

675

learned SV order, w i l l find the inversion rules used for questions difficult since VS contradicts S V . After this rule has eventually been acquired, another d i f f i ­ culty arises w i t h embedded questions that have the surface form o f questions, except for the fact they are not inverted. In order to make this point clear, let us analyze the case o f the Egyptian Arabic plural system, w h i c h has been repeatedly quoted as a canonical example of formal complexity (Slobin, 1973; Clark & Clark, 1977) and w h i c h i n fact involves not only formal complexity but also irregularity and incoherence. I f the Arabic system has three numbers, i.e. singular, dual, and plural, this is only a sign o f greater formal complexity than that o f a language like English, for example, which has only t w o numbers. I f plural nouns accompanying numerals above 10 do not take normal plural forms but a zero ending (typical for the singular) this still is a sign o f formal complexity and—what is more—an i n ­ stance o f incoherence since a zero ending is already used for the singular. The complexity, as manifested by the existence o f t w o subrules for plural forms, is not i n itself the cause o f difficulty since a similar situation can be found i n Polish and the rules are acquired early and without any problems. (Polish nouns w i t h numerals from 2 to 4 take the regular nominative plural form while those accom­ panied by numerals above 4 take the genitive plural ending.) The source o f difficulty is therefore to be found i n the contradiction o f using a zero (singular) form to mark a noun for plural. Finally, the fact that most nouns have totally unpredictable irregular plural forms has nothing to do w i t h formal complexity but is a sign o f irregularity. T o conclude, i t seems that choosing the Arabic number system as a canonical example o f formal complexity has led to confusion rather than to clarification. The evidence that complex systems can be acquired easily, provided they are coherent enough, can be found i n Polish data on the acquisition o f singular declension, where the tripartite gender distinction—masculine/femin­ ine/neuter—is preserved i n the whole paradigm w i t h some more subtle subdivi­ sions introduced at some places. O n the other hand, the plural paradigm, which involves fewer endings (and fewer rules) than the singular, is more difficult because the proportion is destroyed. I n some cases there is no gender distinction at a l l , and i n others, a new distinction o f virile/nonvirile gender is introduced, contradicting the distinction between masculine animate vs. inanimate nouns introduced i n the accusative singular. Finally, i n the genitive, the tripartite dis­ tinction is maintained, but neuter nouns are grouped together w i t h feminines, and not w i t h masculines, as they were i n the singular. It is not only the lack o f internal coherence o f the plural noun declension which makes this subsystem difficult to learn. Another factor is what can be called "external coherence"—the relation o f a subsystem to other (possibly related) subsystems w i t h i n a language. I n this case, rules governing the distribu­ tion o f endings i n plural contradict the rules o f a previously acquired related subsystem, namely, that o f singular.

676

Smoczyhska

A n y subsystem is acquired as part o f the whole system o f a language and its acquisition cannot be analyzed i n isolation. I f the structure and function o f forms under acquisition are coherent w i t h those o f related parts o f the system, and i f the same criteria are maintained across subsystems, these forms are much easier to acquire than i n cases where contradictions and incoherences are present. I n any system o f language there are weak points, incoherences, w h i c h are usually the focus o f ongoing and future linguistic change, as reflected i n adult errors, as w e l l as i n dialectal regularization. C h i l d language is especially sensitive to these incoherences, w h i c h appear to be the main source o f late errors. C h i l d language study may help to detect and interpret them, as children's forms are more system­ atic and clear than occasional adult modifications. As far as irregularity is concerned, an important question to be answered i n further study is: What are the limits o f the child's tolerance for irregularity? What amount o f irregularity can be ignored without causing confusion and forcing the child to use "avoidance strategies" such as the inflectional imperialism strategy used by Russian children, w h o are unable to discover the regular distribution o f endings? 7.7. Operating Principle G: "The use of grammatical markers should make semantic sense." W h i l e questioning the validity o f this Operating Principle, I do not mean that the child is not sensitive to semantic criteria o f rule application. I n fact, children usually look for such criteria, since i n most cases formal distinctions can be found to correspond to semantic distinctions present i n a language. What I want to say is that i f there are no semantic criteria o f formal distinctions, even purely formal criteria can be easily captured, provided the children needs to use the given forms and the criteria are consistent enough. Whether they are semantic or formal is less important than whether they are consistent or not. Therefore Universal G 3 — " S e m a n t i c a l l y consistent rules are acquired early and without significant e r r o r s " — i s valid because i t is a narrow instance o f another universal, which says: " A N Y consistent rule is acquired early and without significant e r r o r " (provided the child needs to use forms i n v o l v i n g these rules, which is not neces­ sarily the case). This broader universal, however, is nothing else than a positive formulation o f Operating Principle F: " L o o k for consistencies" instead o f " A v o i d exceptions." As has already been said, there is nothing specifically linguistic about this operating principle. The main argument for the easy acquisition o f a formal system is presented i n Section 6 . 1 . 1 , where the acquisition o f Polish gender is discussed and compared w i t h that o f Russian. Other examples are provided by the early acquisition o f the distribution o f endings based on the phonological form o f the stem or the fact, already mentioned, that the formal rule o f using the genitive plural o f nouns w i t h numerals above four is acquired easily and without errors. I n fact, one could

6.

The A c q u i s i t i o n of Polish

677

reasonably claim that the whole acquisition o f phonology is the most convincing case o f easy learning o f purely formal rules. There is nothing semantic i n basic phonological oppositions; they can only be used for semantic purposes, but they are not semantic i n themselves.

8. L a n g u a g e A c q u i s i t i o n S t r a t e g i e s : U n i v e r s a l o r Language Specific? The analysis presented above demonstrates that children learning Polish proceed as i f they conform to some Operating Principles proposed by Slobin (1973); whereas they are reluctant to do so w i t h respect to some other Operating Principles. Whether they obey or not depends on the features o f the particular language they learn. I t can be said that these features contribute either to strengthening the validity o f a particular Operating Principle, or to weakening (or even abolishing) i t . This result contradicts Slobin's assumption that the acquisi­ tion o f devices w h i c h do not fit the universal principles w i l l be delayed. What appears i n the data is that children are more likely to reject an Operating Principle which does not f i t the particular language rather than to reject a particular linguistic device which does not fit their Operating Principles. Should we there­ fore admit that the Operating Principles are not independent o f the ongoing process o f acquisition, but that they are being tested against the data o f a particu­ lar language and revised according to the results o f this verification? I n order to answer this question we have to discuss the origin o f the Operating Principles. Slobin is not explicit as far as this problem is concerned, since he avoids discussing the nativism vs. empiricism issue i n language acquisition. However, he qualifies the Operating Principles as something which every child brings to bear on the problem o f language acquisition (1973, p. 197)—as some­ thing preexisting w i t h respect to the process o f acquisition. The question I w o u l d like to ask is whether the child actually obeys some o f the preexisting Operating Principles or does he just behave AS I F he were doing so. T o put it another w a y , are the Operating Principles actually brought to the task o f learning a language as something w h i c h existed prior to the acquisition process? I n view o f what has been said above, it seems simpler to propose that the Operating Principles the child is obeying are nothing else than the outcome o f his L I N G U I S T I C E X P E R I ­ ENCE. Hence, they are the generalizations about a particular language he has made on the basis o f prior learning rather than the effect o f testing preexisting principles against the specific data. Therefore, the claim is that all the Operating Principles are totally L A N G U A G E SPECIFIC, as they are built upon the child's experience w i t h a particular language. Some strategies are universal i n the sense that they are present i n the acquisition o f any language, e.g. " A v o i d exceptions" (or, better, " L o o k for regularities and ignore the remainder"). W e should, however, take into account the fact that every language is more regular than

678

Smoczyhska

irregular, and also that " l o o k i n g for regularities" is the only economical way to build any k i n d o f knowledge. I t is a strategy which is used by the child i n many other domains o f learning and a long time before he starts to speak. Another aspect w h i c h is not taken into account i n Slobin's procedure o f establishing language acquisition universals is the system-like nature o f human language. The child not only discovers that the language he hears is rule-gov­ erned but also that the rules are interrelated to form a system. What is more, while building his o w n language the child does not memorize a list o f unrelated rules but also builds up a system. Here too, there is nothing specifically linguistic about this, since any knowledge is systematized i n a similar way. W h e n analyz­ ing the data we are frequently tempted to forget this and to isolate one category (or one rule or one subsystem) and to relate its acquisition to one Operating Principle. However, something can be difficult for the child because it cannot be easily integrated into the already existing N E T W O R K o f rules. I n such a case, the rule can be modified to f i t the system, or its acquisition can be postponed until the child is able to rebuild his system i n order to create a place for the rule i n question. I n the first case, that o f creating an "erroneous" rule which fits the child's system but not the adult's, its genesis should be analyzed i n terms o f the system as a whole rather than be attributed to a single factor as the necessary and sufficient reason o f its formation. A s I have tried to demonstrate at many points in this paper (Section 4 ) , the formation o f such a rule seems to be due to the j o i n t action o f several factors w h i c h altogether form the necessary and sufficient reason, although none o f them taken alone can be said to be sufficient by itself. There is still another assumption o f Slobin's procedure w h i c h is not supported by the data. As demonstrated by Judith Johnston (1985), cognitive development does not seem to be the main pacesetter o f linguistic development. I t is true that the child cannot express a notion which is not cognitively available at a given moment, but the fact that he is able to grasp a notion does not necessarily involve the appearance o f its linguistic expression i n his speech. I n fact, many semantic categories are not used at the time they are cognitively available, and their emergence i n the child's speech should be attributed to other factors. We are often tempted to conceive o f language acquisition as an activity comparable to those o f a linguist w h o is trying to discover the rules o f an unknown language. This metaphor does not take into account the obvious fact that the acquisition o f language is not an activity per se (as linguists' investiga­ tions are) but it is only a marginal outcome o f the child's efforts to USE an unknown language. I f we look for pacesetters o f language development, it is therefore reasonable to expect to find them i n the domain o f language use. Rules which are acquired early must be not only cognitively available (this being the necessary although not sufficient condition) but also, and foremost, U S E F U L . Sometimes the particular language forces the child to acquire a rule early, al­ though its use is based on relatively difficult criteria. I n such cases we can think about the influence o f language on cognitive development. The rules w h i c h are

6.

The A c q u i s i t i o n of Polish

679

most useful are those w h i c h involve frequently-used structures. I n this sense the role o f frequency must be taken into account. In order to build a system the child must start by learning the most basic rules of the language. These basic rules are at the same time those w h i c h are frequently used since they are present i n simple constructions, as w e l l as i n more complex ones which are derived from them. I n his discovery procedures the child is helped by the special features o f the adult language addressed to h i m , simplified at the beginning and growing more and more complex as the child's linguistic competence develops. W e should bear i n m i n d , however, that the adult i n ­ terlocutor is shaping his utterances according to the child's presumed PASSIVE competence, which significantly outgrows what he is able to produce at a given stage. Therefore, i f we investigate the child's speech at the moment when he starts to put words together, we can assume that he knows much more about putting words together than he demonstrates i n his early t w o - w o r d sentences. Yet it is very difficult to establish his passive knowledge, as comprehension tech­ niques are difficult to use at very early stages. On the other hand, we should not assume that all that adults say is processed by the child; an error commonly made by investigators o f the input language. The child probably has his o w n selection procedures, w h i c h help h i m to pay attention to particular features i n turn w h i l e ignoring the remainder. These procedures are also very difficult to investigate. As far as early child language is concerned, we should also take into account the fact that most o f the exchanges occurring between the child and the adult refer to ongoing activities and objects present i n the situation. A t this stage, utterances cannot be ambiguous since their meanings can be easily discovered from the situation itself, and this is i n fact what both child and adult do without paying much attention to grammatical markers present i n utterances. A t this stage grammatical markers are quite superfluous since successful communication could be achieved as w e l l i f they were not used at a l l . M o s t students assume, however, that the child starts to use inflections, w o r d order, or other marking devices i n order to disambiguate his utterances. I n m y opinion, this is not so until much later, when the child starts to talk about other places and other times than those o f the speech situation. His early marking devices are introduced only because he wants his utterances to resemble the adult model, and since the w o r d as prosodic unit seems to be the main element o f his utterance organization, those markers w h i c h are parts o f words (and not autonomous morphemes) and w h i c h have a prosodic value (syllabic) seem to be the easiest to acquire. This is the only result o f crosslinguistic studies performed so far which I w o u l d call universal and specifically linguistic. Nevertheless, I do not claim this to be innate but rather resulting from the situation o f communication at early stages, which is similar for all children. To summarize, the child learning a language elaborates a set o f processing procedures, w h i c h are determined by general cognitive principles on one hand, and the nature o f the specific language on the other. These procedures change as

680

Smoczyhska

the child's linguistic system develops. N e w procedures appear to deal w i t h new problems and they are always based on the linguistic experience the child has at a given stage. Crosslinguistic study should take into account the dynamic aspects of child language development. W e should compare D E V E L O P M E N T o f various systems rather than single out particular facts to be compared without paying attention to their systematic and developmental contexts unless we actually be­ lieve i n innate linguistic capacities w h i c h specify w i t h much detail the order o f acquisition. I n m y opinion, however, there are no data to support such a c l a i m , and the efforts to account for the mysteries o f child language development by innate propensity result i n explaining ignotum per ignotum.

APPENDIX A Tips o n Polish P h o n o l o g y a n d P r o n u n c i a t i o n : Grapheme, P h o n e m e , a n d English W o r d [WEIST] Stops*

Voiced Voiceless

Bilabial

Dental

Velar

b Ibl bad p /p/ pan

d Idl dog t It/ tap

g /g/ gas k /k/ karp

Bilabial and velar stops have palatalized and nonpalatalized pairs such as /k/ in kosz 'basket' vs /k'/ in kiedy 'when'. a

Fricatives

Voiced Nonpalatal Palatal Voiceless Nonpalatal Palatal

b

Labiodental

Dental

Alveolar

w NI van wi /v7

z Izl zoo

rz or z Izl treasure zi or z Izl

f

s /s/ sod

sz /§/ sHeep si or s Iii

HI

fog

fi i ri

Velar

h or eh l\l locH hi or chi l\l

^English palatal fricatives and affricates as in treasure, SHeep, I O C H , jump, r e a c H are pro­ nounced somewhere between the Polish alveolar nonpalatal and the dental palatal counterparts.

Affricates Dental Voiced Voiceless

Nonpalatal Palatal Nonpalatal Palatal

dz /J7 faDs c

Ici caTS

Alveolar dz dzi or dz cz ci or c

/J7 jump /J/ Ici reacH Ici

T h e A c q u i s i t i o n of Polish

6.

681

Résonants

Nasal

Nonpalatal Palatal Medial (glides) Lateral (liquid) Trill c

Dental

m I ml man mi I ml I lul win

n Ini note ni or ñ Ini

Palatal

j HI young 1

III leaf r III room

c

The Polish r is "rolled."

Vowels

High Lower high Mid nonnasal Mid nasal Low

i y e e

d

Back

Front

d

Alveolar

Bilabial

HI peek

u or ó lul book

1X1 sit

lol mole

ten

Iti

Ibi

lèi

a /a/ father

The English examples only approximate the Polish vowels, e.g. IXI is between sit and set.

I n general, (1) voiced consonants become voiceless i n w o r d final position or when f o l l o w e d by a voiceless consonant, e.g. the phoneme Igl i n snieg ' s n o w ' is pronounced [ k ] , cf. [g] i n sniegu; (2) when f o l l o w e d by some consonants, nasal vowels /e/ and 161 are pronounced as the nonnasal counterparts plus [ m ] , [ n ] , or [rj] depending on the place o f articulation o f the consonant, e.g.

the 16/ i n r$ce

'hands' is pronounced [fence] w h i l e i n reka ' h a n d ' i t is pronounced [ferjka];

and

(3) the stress n o r m a l l y falls on the penultimate (prefinal) syllable o f polysyllabic words, pisze ' ( h e ) : w r i t e s ' , autobus ' b u s ' , and portmonetka

'wallet'.

REFERENCES Aksu, A. A. Aspect and modality in the child's acquisition of the Turkish past tense. Unpublished doctoral dissertation, Department of Psychology, University of California, Berkeley, 1978. Aksu-Koc, A. A . , & Slobin, D. I. The acquisition of Turkish. In D. I. Slobin (Ed.), The crosslinguistic study of child language (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Antinucci, F . , & Miller, R. How children talk about what happened. Journal of Child Language, 1976, 3, 167-189. Bartnicka-Dabkowska, B . Podstawowe wiadomosci z dialektologii polskiej (z cwiczeniami) [Basic information on Polish dialectology (with exercises)]. Warsaw: Pahstwowe Zaklady Wydawnictw Szkolnych, 1965. Bates, E . Language and context: The acquisition of pragmatics. New York: Academic Press, 1976. Baudouin de Courtenay, J. Spostrzezenia nad jgzykiem dziecka [Observations on child language]. Ed. by M. Chmura-Klekotowa. Wroclaw: Zaklad Narodowy im. Ossolihskich, 1974.

682

Smoczyrïska

Bellugi, U . The acquisition of negation. Unpublished doctoral dissertation, Harvard University, 1967. Berman, R. The acquisition of Hebrew, In D. I. Slobin (Ed.), The crosslinguistic study of child language (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Bloom, L . M. Language development: Form and function in emerging grammars. Cambridge, MA: MIT Press, 1970. Bowerman, M. The acquisition of complex sentences. In P. Fletcher & M. Garman (Eds.), Language acquisition. Cambridge: Cambridge University Press, 1979. Pp. 285-306. Brenstiern-Pfanhauser, S. Rozwôj mowy dziecka [The development of child speech]. Prace Filologiczne, 1930, 75, 273-356. Brown, R. A first language: The early stages. Cambridge, MA: Harvard University Press, 1973. Bruner, J. S. The ontogenesis of speech acts. Journal of Child Language, 1975, 2, 1-19. Chmura-Klekotowa, M. Neologizmy slowotwôrcze w mowie dzieci (I) [Morphological neologisms in the speech of children (I)]. Poradnik Jçzykowy, 1967, 10, 433-445. Chmura-Klekotowa, M. Neologizmy slowotwôrcze w mowie dzieci (II) [Morphological neologisms in the speech of children (II)]. Poradnik Jçzykowy, 1968, 1, 19-25. Chmura-Klekotowa, M. Neologizmy slowotwôrcze w mowie dzieci [Morphological neologisms in the speech of children]. Prace Filologiczne, 1971, 21, 99-235. Clancy, P., Jacobsen, T . , & Silva, M. The acquisition of conjunction: A cross-linguistic study. Papers and Reports on Child Language (Department of Linguistics, Stanford University), 1976, No. 12, 71-80. Clark, H. H . , & Clark, E . V . Psychology and language. New York: Harcourt Brace Jovanovitch, 1977. Cochrane, N. J . Verbal aspect and the semantic classification of verbs in Serbo-Croatian. Unpublished doctoral dissertation, University of Texas, 1977. Comrie, B . Aspect: An introduction to the study of verbal aspect and related problems. Cambridge: Cambridge University Press, 1976. Cromer, R. F . The development of temporal reference during the acquisition of language. Unpublished doctoral dissertation, Harvard University, 1968. de Villiers, J . G . , & de Villiers, P. A. A cross-sectional study of the acquisition of grammatical morphemes. Journal of Psycholinguistic Research, 1973, 2, 267-278. de Villiers, P. A . , & de Villiers, J . G . Form and function in the development of sentence negation. Papers and Reports on Child Language Development (Department of Linguistics, Stanford University), 1979, No. 17, 57-64. Dowty, D. Word meaning and Montague Grammar. Dordrecht (Holland): D. Reidel, 1979. Eisenberg, A. R. Language development in cultural perspective. Unpublished doctoral dissertation, University of California, Berkeley, 1982. Ferrell, J. O. The meaning of perfective aspect in Russian. Word, 1951, 7, 104-135. Fisiak, J . , Lipihska-Grzegorek, M . , & Zabrocki, T. An introductory Polish-English contrastive grammar. Warsaw: Pahstwowe Wydawnictwo Naukowe, 1978. Geppertowa, L . Rozwôj rozumienia i poslugiwania sic pojçciami stosunkôw wyrazanymi przez spôjnik "zeby" u dzieci do lat pieciu [The development of understanding and usage of notions expressed by the conjunction zeby in children up to five years old]. Przeglqd Psychologiczny, 1959, 3, 47-81. Geppertowa, L . Rozwôj rozumienia i poslugiwania sic przez dzieci pojçciami stosunkôw okreslanymi przez przyimki i spôjniki [The development of children's understanding and usage of notions of relations defined by prepositions and conjunctions]. In S. Szuman (Ed.), O rozwoju jezyka i myslenia [On the development of language and thought]. Warsaw: Pahstwowe Wydawnictwo Naukowe, 1968. Pp. 149-381. Guillaume, P. Le développement des éléments formels dans le langage de l'enfant. Journal de Psychologie Normale et Pathologique, 1927, 24, 203-229.

6.

T h e A c q u i s i t i o n o f Polish

683

Gvozdev, A. N. Formirovanie u rebenka grammaticeskogo stroja russkogo jazyka. Vols. I and II. Moscow: Izd-vo Akademii Pedagogiceskix Nauk RSFSR, 1949. Gvozdev, A. N. Voprosy izucenija detskoj reci. Moscow: Izd-vo Akademii Pedagogiceskix Nauk RSFSR, 1961. Johnston, J. R. Cognitive prerequisites: The evidence from children learning English. In D. I. Slobin (Ed.), The crosslinguistic study of child language (Vol. 2). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Kaczmarek, L . Ksztahowanie sie mowy dziecka [The formation of the child's speech]. Poznah: Towarzystwo Przyjaciol Nauk, 1953. Karmiloff-Smith, A. More about the same: Children's understanding of post-articles. Journal of ChUd Language, 1977,4, 377-394. Karmiloff-Smith, A. A functional approach to child language: A study of determiners and reference. Cambridge: Cambridge University Press, 1979. Kolaric, R. Slovenski otroski govor. Godisnjak Filozofskog Fakulteta u Novom Sadu, 1959, 4, 229¬ 258. Kubit, D. Ksztahowanie sie negacji w mowie dziecka [The formation of negation in the child's speech]. Psychologia Wychowawcza, 1977, 4, 358-377. Kuczaj, S. A . , & Daly, M. J . The development of hypothetical reference in the speech of young children. Journal of Child Language, 1979, 6, 563-579. Ligeza, M. Interpersonalna funkcja pytah dzieci w wieku l;6-6;0 [Interpersonal function of ques­ tions of children at the age of l;6-6;0]. Zeszyty Naukowe Uniwersytetu Jagiellonskiego, 599 (Prace Psychologiczno-Pedagogiczne, 31), 1979, 119-135. Limber, J. The genesis of complex sentences. In T. E . Moore (Ed.), Cognitive development and the acquisition of language. New York: Academic Press, 1973. Lotko, E . Jezyk polski z typologicznego punktu widzenia [The Polish language from a typological point of view]. Polonica, 1979, 5, 9-23. Majewicz, A. F . Understanding Aspect (1). Lingua Posnaniensis, 1982, 24, 29-62. Miller, J. E . Stative verbs in Russian. Foundations of Language, 1970, 6, 488-504. Mystkowska, H. Wlasciwosci mowy dziecka szescio-siedmioletniego [The characteristics of a sixseven-year-old child's speech]. Warsaw: Pahstwowe Zaklady Wydawnictw Szkolnych, 1970. Olszowska-Guzik, T. Poziom rozwoju struktur gramatycznych jazyka dzieci szescioletnich a srodowisko spolecznokulturowe [The level of development of grammatical structures of six-yearold children and their socio-cultural environment]. Krakow: Uniwersytet Jagiellohski, in prepara­ tion. Popova, M. I. Grammatical elements of language in the speech of pre-preschool children. In C . A. Ferguson & D. I. Slobin (Eds.), Studies of child language development. New York: Holt, Rinehart & Winston, 1973. Pp. 269-281. Przetacznikowa, M. Rozwoj i rola przyslowkow w mowie i mysleniu dziecka do lat trzech [The development and function of adverbs in the speech and thought of a child up to three years old]. Biuletyn Polskiego Towarzystwa Jgzykoznawczego, 1956, 15, 139-193. Przetacznikowa, M. Odzwierciedlenie cech przedmiotow i zjawisk w mowie dzieci w wieku przedszkolnym [Reflection of the features of objects and phenomena in the speech of preschool children]. Krakow: Uniwersytet Jagiellohski, 1959. Przetacznikowa, M. Rozwoj struktury i funkcji zdah w mowie dziecka [The development of struc­ ture and function of sentences in the child's speech]. In S. Szuman (Ed.), O rozwoju jazyka i myslenia [On the development of language and thought]. Warsaw: Pahstwowe Wydawnictwo Naukowe, 1968. Pp. 383-629. Przetacznikowa, M. Semantyczna interpretacja wczesnych stadiow rozwoju skladni u dzieci [The semantic interpretation of the early stages of the development of syntax in children]. Zeszyty Naukowe Uniwersytetu Jagiellonskiego, 389 (Prace Psychologiczno-Pedagogiczne, 23), 1975, 7-53.

684

Smoczynska

Przetacznikowa, M. Additive and contrastive relations expressed in preschool children's speech. Polish Psychological Bulletin, 1976, 7, 45-54. Przetacznikowa, M. Functions of inclusive coordinate constructions in early language development. Polish Psychological Bulletin, 1978, 9, 231-238. Radulovic, L . Acquisition of language: Studies ofDubrovnik children. Unpublished doctoral disser­ tation, University of California, Berkeley, 1975. Ridjanovic, M. A synchronic study of verbal aspect in English and Serbo-Croatian. Cambridge, MA: Slavica, 1976. Rispoli, M. The emergence of verb and adjective tense-aspect inflections in Japanese. Unpublished master's thesis, University of Pennsylvania, 1981. Rzetkowska, J . Przyczynek do badan nad rozwojem mowy dziecka [Contribution to studies of the development of child speech]. Warsaw: Vol. I, 1908; Vol. 2, 1909. Savic, S. Review of J. Baudouin de Courtenay, Spostrzezenia nad jezykiem dziecka [Observations on child language]. Journal of Child Language, 1975, 2, 326-328. Schenker, A. M. Polish declension. The Hague: Mouton, 1967. Schenker, A. M. Beginning Polish. New Haven: Yale University Press, 1973. Shugar, G . W. Relations of language structure and activity structure in the early developmental period. Unpublished doctoral dissertation, University of Warsaw, 1972. Shugar, G . W. Text-constructing with an adult: A form of child activity during early language acquisition. In G . Drachman (Ed.), Akten des 1. Salzburger Kolloquiums uber Kindersprache (Salzburg, 6-8 December 1974). Tubingen (FRG): G . Narr, 1976, 345-356. (a) Shugar, G . W. Behavior stream organization during early language acquisition. Polish Psychologi­ cal Bulletin, 1976, 7, 27-36. (b) Shugar, G . W. Text analysis as an approach to the study of early linguistic operations. In N. Waterson & C . Snow (Eds.), The development of communication. Chichester: Wiley, 1978. Pp. 227-251. Siatkowscy, E . J . W zwiazku z rozprawa Smoczyhskiego "Przyswajanie przez dziecko podstaw systemu jezkowego" [A response to Smoczyhski's dissertation "The acquisition of the funda­ mentals of the language system by a child"]. Poradnik Jgzykowy, 1956, 330-338. Skorupka, S. Obserwacje nad jgzykiem dziecka [Observations on child language]. Sprawozdania z Posiedzen Komisji J^zykowej Towarzystwa Naukowe go Warszawskiego, 1949, 3, 116-155. Slobin, D. I. The acquisition of Russian as a native language. In F . Smith & G . A. Miller (Eds.), The genesis of language: A psycholinguistic approach. Cambridge, MA: MIT Press, 1966. Pp. 129-148. Slobin, D . I . Early grammatical development in several languages, with special attention to Soviet research. Working Paper No. 11, Language-Behavior Research Laboratory, University of Cal­ ifornia, Berkeley, 1968. Slobin, D. I. Cognitive prerequisites for the development of grammar. In C . A. Ferguson & D. I. Slobin (Eds.), Studies of child language development. New York: Holt, Rinehart & Winston, 1973. Pp. 175-208. Slobin, D. I. Language change in childhood and in history. In J. Macnamara (Ed.), Language learning and thought. New York: Academic Press, 1977. Pp. 185-214. Slobin, D . I . Universal and particular in the acquisition of language. In E . Wanner & L . R. Gleitman (Eds.) Language acquisition: The state of the art. Cambridge: Cambridge University Press, 1982. Smoczynska, M. Przyswajanie form deklinacji rzeczownikowej przez dzieci w wieku przedszkolnym [The acquisition of forms of noun declension by preschool children]. Psychologia Wychowawcza, 1972, 29, 515-527. Smoczynska, M. Development of the transitive sentence pattern. In G . Drachman (Ed.), Akten des 1. Salzburger Kolloquiums uber Kindersprache (pp. 221-233). Tubingen (FRG): G . Narr, 1976. (a)

6.

T h e A c q u i s i t i o n o f Polish

685

Smoczynska, M. Early syntactic development: Pivot look and pivot grammar. Polish Psychological Bulletin, 1976, 7, 37-43. (b) Smoczynska, M. Wczesne stadia rozwoju skladni w mowie dziecka [Early stages of syntactic development in child speech]. Unpublished doctoral dissertation, Uniwersytet Jagiellohski, Krakow, 1978. (a) Smoczynska, M. Semantic intention and interpersonal function: Semantic analysis of Noun + Noun constructions. In N. Waterson & C . Snow (Eds.), The development of communication. Chichester: Wiley, 1978. Pp. 289-300. (b) Smoczynska, M. Linguistic expression of requests at the early stages of syntactic development. Paper read at the XXII International Congress of Psychology, Leipzig, July 1980. Smoczynska, M. Uniformities and individual variation in early syntactic development. Polish Psy­ chological Bulletin, 1981, 12, 3-15. Smoczyhski, P. Przyswajanie przez dziecko podstaw systemu jqzykowego [The acquisition of the fundamentals of language system by a child]. Lodz: Lödzkie Towarzystwo Naukowe, 1955. Stephany, U . Verbal grammar in Modern Greek early child language. In P. S. Dale & D. Ingram (Eds), Child language: An international perspective. Baltimore: University Park Press, 1981. Szagun, G . The development of spontaneous reference to past and future: A crossslinguistic tudy. Institut für Psychologie, Technische Universität Berlin, 1979. Szuman, S. Rozwöj tresci slownika u dzieci [The development of vocabulary content in children]. Studio. Pedagogicane, 1955, 2. Szwedek, A. Word order, sentence stress and reference in English and Polish. Edmonton (Canada): Linguistic Research, Inc., 1976. Tokarski, J . Fleksja polska [Polish flection]. Warsaw: Pahstwowe Wydawnictwo Naukowe, 1973. Toivainen, J . Inflectional affixes used by Finnish-speaking children aged 1-3 years. Helsinki: Suomalaisen Kirjallisuuden Seura, 1980. Vendler, Z. Verbs and time. In Z. Vendler, Linguistics in philosophy. Ithaca, NY: Cornell Univer­ sity Press, 1967. Pp. 97-121. Verkuyl, H . J . On the compositional nature of aspect. Dordrecht: Reidel, 1972. Wawrowska, W. Badania psychologiczne nad rozwojem mowy dziecka (od 1;0 do 2;3 lat) [Psycholological studies on the development of child spech (from 1;0 to 2;3)]. In S. Szuman (Ed.), Prace psychologiczne. Warsaw: Wydawnictwo Naukowe Towarzystwa Pedagogicznego, 1938. Weist, R. M. The acquisition of temporal deixis in Polish. Paper read at the Institute of Psychology, Uniwersytet Jagiellohski, Krakow, June 1980. Weist, R. M. Cross-linguistic perspectives on sentence processing in early patterned speech. Paper presented at the meeting: Current trends in cognitive science: Conversations in the disciplines, SUNY Cortland, March 26, 1982. Weist, R. M. Prefix versus suffix information processing in the comprehension of tense and aspect, Journal of Child Language, 1983, 10, 85-96. (a) Weist, R. M. The word order myth, Journal of Child Langauge, 1983, 10, 97-106. (b) Weist, R. M . , Wysocka, H . , Witkowska-Stadnik, K . , Buczowska, E . , & Konieczna, E . The defec­ tive tense hypothesis: On he emergence of tense and aspect in child Polish, Journal of Child Language, 1984, 11, 347-374. Wöjtowicz, J . Drobne spostrzezenia o dzialaniu analogii w mowie dziecka [Some remarks on the function of analogy in child speech]. Poradnik Jgzykowy, 1959, 8, 349-352. Zabielski, S. Rozumienie relacji semantycznych przez dzieci w wieku przedszkolnym [The under­ standing of semantic relations by preschool children]. Unpublished doctoral dissertation, Uniwer­ sytet Jagiellohski, Krakow, 1974. Zakharova, A. V . Acquisition of forms of grammatical case by preschool children. In C . A.

686

Smoczynska

Ferguson & D. I. Slobin (Eds.), Studies of child language development. New York: Holt, Rinehart & Winston, 1973. Pp. 281-284. Zarçbina, M. Ksztahowanie sie systemu jçzykowego dziecka [The formation of the language system of a child]. Wroclaw: Zaklad Narodowy im. Ossolihskich, 1965. Zarçbina, M. Review of J. Baudouin de Courtenay, Spostrzezenia nad jçzykiem dziecka [Observa­ tions on child language]. Polish Psychological Bulletin, 1976, 7, 61-63.

The Acquisition of Romance, with Special Reference to French

Eve

V. Clark

Stanford

University

Contents Introduction 688 Descriptive Sketch of French, with Notes on Other Romance Languages Sources of Evidence 697 Overall Course of Development 699 The Data 702 Typical Errors 702 Error-Free Acquisition 720 Timing of Acquisition 721 Organization and Change During Development 726 Lexical Structure and Word-Formation 734 The Setting of Language Acquisition 743 Cognitive Pacesetting in Language Development 743 Language Input 751 Individual Differences 754 Conclusions 756 Some Theoretical Implications 756 Coda 762 Some Suggestions for Further Study 762

688

. . . D'abord (les enfants) ont, pour ainsi dire, une grammaire de leur âge, dont la syntaxe a des règles plus générales que la nôtre; et si Von y faisait bien attention Von serait étonné de l'exactitude avec laquelle ils suivent certaines analogies . . J —J.-J. Rousseau, Emile (1762)

*"At first (children) have, so to speak, a grammar suited to their age, a grammar whose syntax contains rules more general than ours; and if one pays close attention to their language, one is amazed at how well they follow certain analogies . . . " 687

688

Clark

INTRODUCTION The first part o f this chapter contains a brief grammatical sketch o f French, w i t h occasional comments on how other Romance languages diverge from this pic­ ture; a brief account o f the sources available on language acquisition by children; and a summary o f the overall course o f development i n children acquiring the major Romance languages. The remainder o f the review takes up different facets of the language acquisition process i n more detail.

1. D e s c r i p t i v e S k e t c h o f F r e n c h , w i t h N o t e s o n O t h e r Romance Languages The Romance languages are Indo-European, direct descendents o f the varieties of Latin spoken i n different parts o f the Roman Empire. Romance is c o m m o n l y divided into several types, depending on geographic location and history. French is considered part o f Gallo-Romance (which includes the various languages and dialects that evolved i n the northern and southern halves o f modern France); Italian belongs to Italo-Romance; Portuguese and Spanish to Ibero-Romance; and Rumanian to Balkan Romance (see Elcock, 1960; Ewert, 1953). Standard grammars for the major Romance languages are generally available in English and i n the language concerned, but the best reference grammars tend to be i n the actual language (e.g. Chevalier et a l . , 1964; Gougenheim, 1962; Grevisse, 1964; Wagner & P i n c h ó n , 1962; Wartburg & Zumthor, 1958, for French; for a list o f major grammars for each language, see M c K a y , 1979). I n the sketch that follows, I simply list some o f the main characteristics o f French and note some o f the respects i n which other Romance languages differ from the French model. M y sketch focuses on syntactic, morphological, and lexical prop­ erties, and ignores most phonological ones. I w i l l follow much the same pro­ cedure when I turn to the data on the acquisition o f these languages by children. 1.1. Word Order. The basic w o r d order i n French, and i n most M o d e r n Romance languages, is S V O , but according to Greenberg's (1963) criteria, the Romance languages are mixed rather than pure cases. I n French, for example, articles, possessive pronouns, and prepositions precede nouns but adjectival modifiers ( w i t h some subtle exceptions) follow them (see W a u g h , 1977; W i l m e t , 1981), as do relative clauses and possessive noun phrases i n possessive constructions: Mon POSS.PRO

livre book

noir black

se trouve sur la table. REFL.PRO find on the table

'My black book is on the table.'

7. Le chien the dog

que REL.PRO

The Acquisition of Romance

689

j'ai vu est a Jean. I-saw is to Jean

'The dog I saw is Jean's.' Italian, Portuguese, and Spanish follow virtually the same pattern as French i n these constructions; i n Rumanian, though, the article follows and is enclitic on the noun; possessive pronouns also follow their nouns, and the possessor noun i n possessive constructions is further marked by being i n the dative or genitive case. (Case marking has vanished, except from parts o f the pronoun systems, i n the other major modern Romance languages.) The basic w o r d order i n French changes from S V O to S O V when the direct object is pronominalized. Pronominalized indirect objects also precede the verb: Jean donne Jean gives Jean Jean

le livre the book

a Pierre. to Pierre

le OB.PRO:MASC:SG

donne a Pierre. gives to Pierre

'Jean gives it to Pierre.' Jean le Jean it

lui IO.PRO:SG

donne. gives

'Jean gives it (to) him.' Jean lui Jean to:him

donne le livre. gives the book

'Jean gives him the book.' The relative ordering o f direct and indirect objects, i n either full or pronominal form, is maintained w i t h imperative verbs, but both follow rather than precede the verb: Donnez-le give-it

a Pierre. to Pierre

'Give it to Pierre.' Donnez give

-le -it

lui. him

'Give it (to) him.' Donnez-lui give-him

le livre. the book

'Give him the book.'

690

Clark

The ordering o f direct and indirect object pronouns is further complicated when the indirect object is a first or second person pronoun since it then PRECEDES the direct object i n declarative utterances but not i n positive imperative ones: Jean Jean

me le me it

donne. gives

'Jean gives it to me.' Donnez give

-le-moi. -it-me

'Give it (to) me.' (For further discussion o f some o f complexities o f w o r d order w i t h clitic pro­ nouns, see Gaatone, 1976.) I n formal French, the basic sentential negative is discontinuous, ne . . . pas, w i t h the first element preceding the finite verb (and any direct or indirect pro­ noun objects) and the second immediately f o l l o w i n g i t : Jean Jean

ne donne pas not gives not

le livre a Pierre. the book to Pierre

'Jean doesn't give the book to Pierre.' Jean Jean

ne not

Va it-has

pas donne a Pierre. not given to Pierre

'Jean hasn't given it to Pierre.' Negative imperatives—where both direct and indirect pronominal objects pre­ cede the verb—keep the same relative ordering o f negative and pronoun forms as is found i n indicative utterances: Ne not

le it

lui (to): him

donnez give

pas. not

'Don't give it (to) him.' Ne not

me (to): me

le it

donnez give

pas. not

'Don't give it (to) me.' I n colloquial spoken French, however, the element ne is normally omitted (Gaatone, 1971). Dependent clauses i n French—relative, adverbial, and certain complement forms—also normally have S V O w o r d order. A l l o w a b l e exceptions, though, are those object relative clauses i n w h i c h a main clause noun phrase occurs as the

7.

The Acquisition of Romance

691

object o f the relative clause. I n such clauses, although the commoner order is probably O S V , the subject and verb can be inverted to yield an O V S order instead, as i n the second example: SVO:

L'homme the man

qui who: SUB J

porte carries :V

le manteau the coat:OBJ

est entré is entered

dans la maison. into the house 'The man who is carrying the coat has gone into the house.' OVS:

L'homme the man

qu' who: OBJ

a vu Jean saw:V Jean: SUB J

est entré dans la maison. is entered into the house

'The man Jean saw has gone into the house.' For object relatives w i t h a pronoun subject, the order is invariably O S V . W h e n the relative clause head is the subject o f the following verb, it is introduced by qui; otherwise, when it is the object, it is introduced by que (or qu' before a verb beginning w i t h a v o w e l ) . Another factor affecting w o r d order is the frequent reliance i n spoken French on both left- and right-dislocation. For instance, a subject noun phrase may follow an utterance w i t h a pronoun subject (right-dislocation): 2

// mange tout ce qu'il voit, he eats all that he sees,

cet homme. that man

'He eats everything he lays eyes on, that man does.' Or, having introduced the subject noun phrase, the speaker may continue w i t h a pronominalization for the same referent (left-dislocation): Cet homme,

il mange tout ce qu'il voit.

'That man there, he eats everything he lays eyes on.' I n discourse, right-dislocation tends to serve as a way o f making sure one's addressee has the correct referent i n m i n d for the subject pronoun, while leftdislocation serves more o f an emphatic function, picking out a particular referent as given, and then, having established i t , going on to comment on some aspect o f

The terms left- and right-dislocation are used here simply for descriptive simplicity, and are not intended to imply anything about the psychological processes of production or comprehension. The phenomenon of dislocation in French is combined with reliance on clitic subject pronouns (je, tu, il, etc.) that occur obligatorily with the verb. In Italian and Spanish, dislocation appears just as per­ vasive, but since the verb forms in these languages mark both person and number, there is no corresponding reliance on pronouns to fulfill those functions. 2

692

Clark

it. ( I n colloquial speech, the left-dislocated noun phrase is often followed by a nonsubordinating que.) The t w o forms just illustrated are very c o m m o n , but occur mainly w i t h subject noun phrases, less often w i t h object ones (except to " f i l l i n " one's addressee). I n Spanish and Italian, where person pronouns are not obligatorily used w i t h finite verbs, the w o r d order appears superficially to be much freer than i n French because both subject and object noun phrases can be ordered to reflect the relative importance or thematic relevance o f what's being mentioned. However, speakers normally avoid orders that w o u l d be ambiguous i n the context o f the utterance. I n general, w o r d order i n Romance is highly sensitive to the pragmatic constraints imposed by discourse and cannot be fully described without taking such factors into account. The w o r d order i n yes-no questions i n French is usually the same as i n declarative utterances, but the question is marked by a final rise i n intonation: La fille the girl

monte Vescalier? climbs the-staircase

Ts the girl going upstairs?' Yes-no questions can be marked i n addition by inversion o f subject and verb, although this is rare i n colloquial speech (Behnstedt, 1973). Inversion can take one o f t w o forms, either an inverted est-ce que 'is it that' preceding the utter­ ance, as i n : Est-ce que is-it that

la fille monte I'escalier? the girl climbs the-staircase

Ts the girl going upstairs?' or a pronominal subject and verb inverted directly: La fille, the girl

monte-t-elle climbs- -she

I'escalier? the-staircase

I n colloquial French, intonation alone is the device most c o m m o n l y used to mark yes/no questions. Both intonation and inversion, though, are required for ququestions, introduced by qui ' w h o ' , quoi ' w h a t ' , ou ' w h e r e ' , combien

'how

many, how m u c h ' , comment ' h o w ' , pour quoi ' w h y ' , quand ' w h e n ' , and quel ' w h i c h ' i n French, as i n : Ou where

va-t-elle, goes she,

la fille? the girl

'Where is the girl going?'

7.

The Acquisition of Romance

693

(alongside the c o m m o n colloquial order, Ou elle va, cette fille?) Comment how

se REFL.PRO

mangent-ils les asperges? eat-they the asparagus

'How does one eat asparagus?' Quel chien which dog

regardes-tu? look:at-you

'Which dog are you looking at?' I n many qu- questions, one can use est-ce que immediately after the question w o r d , and follow it w i t h declarative order. The inversion required may be com­ plex or simple. For instance, when the subject is a noun phrase rather than a pronoun, the inversion requires the addition o f an inverted clitic subject to the verb (as i n the first t w o qu- questions above). Otherwise, w i t h pronoun subjects, the subject and verb are merely inverted so that the subject follows rather than precedes. A s i n English, questions introduced by a question w o r d (unlike yes-no questions) do not have rising intonation. 1.2. Noun and Verb Morphology. Nouns i n French are marked for definiteness, gender (masculine or feminine), and number, by means o f their accom­ panying articles: definite PL SG MASC FEM

le la

les les

indefinite PL SG un une

des des

Thus nouns nearly always occur w i t h a definite or an indefinite article. After negatives, the form des normally becomes de (see further Gaatone, 1971). I n spoken French, the articles are the primary source o f information about gender and number (Chevalier, 1966; M o k , 1968). This is because (a) gender is not entirely predictable from the phonological shapes o f words, and (b) number is not normally marked by any modulation i n the form o f the noun. (The written changes marking plural on nouns are for the most part purely orthographic. Very few nouns have a plural form that is pronounced differently from the singular. Those that do tend to end i n -al ox-ail, e.g., cheval-chevaux, corail-coraux). Adjectives agree i n gender and number w i t h the nouns they accompany, whether i n attributive or predicative position. I n the plural, adjective agreement can provide clues to gender that are lacking from the article alone. However, where plurals refer to a mixed-gender set, all forms o f adjectives and articles go to the masculine plural. M o s t adjectives follow their nouns i n French, but a few

Clark

694

precede, e.g. petit ' s m a l l ' , bon ' g o o d ' , vieux ' o l d ' ; others may be preposed for emphasis. There is also a small group o f adjectives that have different meanings, depending on their position, e.g. sespropres mains 'his own hands' compared to ses mains propres 'his clean hands', so that the order noun-adjective is not necessarily very r i g i d i n French (see Waugh, 1977). T w o or more adjectives modifying the same noun are usually separated by et 'and'. Possession can be indicated i n t w o ways, one o f which requires agreement o f the possessive pronoun i n person w i t h the possessor, and i n number and gender w i t h the object possessed. Gender-marking is explicit i n the form o f the pos­ sessive adjective for first, second, and third person singular possessors, but not elsewhere. Person 1 2 3

SGMASC

SG:FEM

PL

SG

PL

mon ton son

ma ta sa

mes tes ses

notre votre leur

nos vos leurs

Otherwise, possession can be indicated analytically, by naming the object pos­ sessed and f o l l o w i n g it w i t h a prepositional phrase introduced by a or de w i t h a proper name, noun phrase, or a disjunctive (strong) pronoun f o r m , as i n : Le cheval de Jean est la-bas 'the horse o f Jean is over-there' ( = 'Jean's horse is over there'), or Cette voiture est a moi 'that car is to me' ( = 'That car's m i n e ' ) . The strong pronoun forms, generally used after prepositions, are moi, toi, lui (soi for impersonal or reflexive uses), nous, vous, eux 'me, y o u , h i m , us, y o u , t h e m ' . Finally the pronominal possessive forms, like the possessive adjectives, agree i n person w i t h the possessor, but i n number and gender w i t h the object possessed; they are always used w i t h articles, as i n : Cest le mien 'it:is the:MASC:SG m i n e . M A S C . S G ' , ( = ' I t ' s m i n e ' ) . The different ways o f marking possession w i l l be considered further w i t h the acquisition data. Verbs i n French mark person, number, tense and aspect, mood, and i n a few constructions gender as w e l l . Person is distinguished primarily by the person pronoun that must co-occur w i t h the verb i n the absence o f a noun phrase subject (that is, a third person form). The pronoun forms i n French are shown below w i t h the regular (first conjugation) present tense forms o f donner 'to give': Person

SG

PL

1 2 3

je donne tu donnes //, elle donne

nous donnons vous donnez ils, elles donnent

For first conjugation verbs, the 1st, 2nd, 3rd singular and 3rd plural forms are all pronounced alike (e.g. / d o n / ) ; the 1st plural verb has a distinct ending -ons (e.g. /dono/), and so does the 2nd person plural, -ez, but the latter is indistinguishable

7.

The A c q u i s i t i o n of Romance

695

in pronunciation from the infinitive form i n -er (both /done/). The accompanying pronouns therefore play a critical role i n indicating person. Number is also marked by the pronouns (and, for the 1st and 2nd persons plural, by the verb ending as w e l l ) . Although the 3rd person pronouns, singular and plural, are normally pronounced alike, the plural is distinct from the singular before a verb stem beginning w i t h a vowel since the final -s on Us or elles is then pronounced. Verbs i n French can be assigned to one o f three conjugations. The largest and most regular, w h i c h contains some 90% o f all French verbs, is the first, w i t h verb infinitives that end i n -er (like donner). This conjugation is also the one to which most new verbs, both coinages and borrowings, are assigned. The second conjugation is much smaller—some 300 verbs w i t h infinitives i n -ir, like finir 'to f i n i s h ' , shown here i n the present indicative for comparison w i t h donner: Singular:

je finis, tu finis, ilfinit

Plural:

nous finissons, vous finissez, Us finis sent

(The three singular verb forms are pronounced alike, / f i n i / , despite orthographic differences, while the plural forms contrast w i t h both the singular and w i t h each other: /finiso, finise, f i n i s / . ) The third conjugation contains the remaining verbs, all irregular i n f o r m , including some 30 i n -ir that do not follow the finir pattern, e.g. dormir 'sleep', ouvrir 'open', tenir ' h o l d ' , venir 'come'; another 30 or so i n -oir, e.g. vouloir ' w a n t ' , s'asseoir ' s i t ' , devoir 'must, ought'; and about 100 i n -re, e.g. prendre 'take', battre ' h i t ' , boire ' d r i n k ' , conduire ' d r i v e ' . Most o f the verbs i n this irregular or catch-all conjugation are very common i n everyday use, and so are more frequent than any first conjugation verbs. They are therefore very likely to be used both to and by children. Tense and aspect i n French are both marked through verb endings, but the t w o systems are intertwined rather than being separate, as i n some languages. The indicative has five commonly used forms: present, imperfect, compound past (passe compose) and t w o forms o f f u t u r e . The present picks out the actual time of utterance, while the imperfect and the compound past pick out times i n the past, prior to the time o f the utterance. The imperfect also marks incompletive aspect, and is used i n narrative for background events and descriptions. The compound past i n contrast, marks completive aspect, focusing on the result or product o f an activity, as i n this sentence where the first verb is i n the imperfect, and the next t w o i n the compound past: 3

Les gens regardaient; le Men a suivi les traces et a trouve la balle perdue. 'The people watched; the dog followed the trail and found the lost ball.' I am deliberately excluding from this sketch forms that are no longer used in modern spoken French, such as the simple past. 3

696

Clark

The compound past is constructed w i t h one o f the t w o auxiliary verbs, avoir 'to have' or être 'to be', carrying tense, plus the past participle. For talking about the future, there are t w o options: a periphrastic form constructed w i t h the verb aller 'to g o ' plus an infinitive, as i n Je vais partir demain ' I ' m going to leave t o m o r r o w ' and the simple future expressed by an inflection added to the infinitive f o r m o f the verb, as i n Je partirai demain ' I ' l l leave t o m o r r o w ' . However, the t w o forms o f the future cannot always be i n terchanged, since certain uses o f the inflectional (simple) future do not allow substitution o f the periphrastic f o r m . There are several other moods besides the indicative (the only one considered so far). The exact number, however, is a matter o f dispute and standard grammars show little agreement on this. For the present sketch, I w i l l consider the conditional, the subjunctive, and the imperative. The conditional is generally used i n main clauses for tempus irrealis, whether i n the future, present, or past. I t can be used for describing supposed facts i n general (things one's heard tell) even when they are not dependent on prior conditions, and for purpose where that is dependent on potential facts or conditions. I t can be used for the indignant rejection o f imputations, for the description o f imaginary situations (daydreaming or fiction), and for polite (attenuated) requests. (The imperfect tense can also have the latter function, as i n Je venais vous offrir . . . T was going to offer you . . . ' . ) The subjunctive, like the conditional, is used for certain kinds o f tempus irrealis. I n particular, it appears i n subordinate clauses f o l l o w i n g verbs o f wanting, ordering, forbidding, or begging; after many verbs o f feeling, and after verbs o f opinion or perception when what is to be conveyed is considered possible rather than actual by the speaker. The subjunctive is also used i n certain adverbial clauses, after conjunctions expressing purpose, and after some temporal and conditional conjunctions. I t is also used i n certain restrictive relative clauses. However, characterizing the precise domain o f the subjunctive is extremely difficult since numerous constructions where one might expect to find i t instead take the indicative (see Sandfeld, 1965). M o s t speakers o f French today make use only o f the present tense subjunctive i n colloquial speech. The imperfect tends to be used only i n written French, and even there is often avoided. The imperative is used for ordering and directing. I t occurs only i n the t w o second person forms, and is identical to the indicative (without the pertinent pronouns), e.g. for donner, the t w o forms are donne (singular) and donnez (plural). Directive utterances are often softened by using the politer conditional, as i n Vous plairait-il de recommencer? ' y o u it please + C O N D to start again,' ( = ' W o u l d you like to start over again?'). Some interrogative forms are also conventionally used as (less polite) directives, e.g. Veux-tu finir? 'want-you to finish' ( = 'Finish, w i l l you?') or Tu ne f arrêtes pas? ' Y o u aren't stopping' (='Stop it!'). Although the main properties o f verb morphology—the marking o f person, number, tense, aspect, m o o d , and gender—are similar i n all the Romance lan-

7.

The A c q u i s i t i o n of Romance

697

guages, there are also numerous differences. I n Spanish and Italian, for example, person can be identified from the verb form alone and person pronouns are therefore optional. As a result, subject pronouns tend to be used more for empha­ sis than i n French. A n d the fact that both person and number are marked i n the verb form itself makes for greater flexibility i n the w o r d orders commonly used in both Spanish and Italian. W o r d order can be varied to mark information as given and new, to mark gradients o f emphasis, and to tie a f o l l o w i n g utterance directly to whatever the preceding speaker had said. This, i n turn, suggests one might find more variability i n w o r d order, even during the early stages o f ac­ quisition, for children acquiring Italian and Spanish than for children acquiring French. 1.3. Word Formation. N e w words—that is, new forms w i t h new mean­ ings—can be constructed either by derivation or by composition i n French. M o s t derivation i n French is through the addition o f suffixes to the base or root w o r d to make adjectives, nouns, or verbs into nouns: e.g. cherche i o o k , ' [ V ] , chercheur i o o k e r , person who looks; researcher' [ N ] ; to make nouns or adjectives into verbs: e.g. boxe ' b o x i n g , ' [ N ] , boxer 'to b o x , ' [ V ] ; or to make the feminine forms o f adjectives into adverbs: e.g., douce 'quiet,' [ A D J : F E M : S G 1 , douce¬ ment ' q u i e t l y , ' [ A D V ] . Most derivational suffixes fall into the first group, for forming new nouns, but there are productive suffixes o f all three kinds, e.g. nominal -et (diminutive), -eur and -ier (both agentive); verbal -er and -ir to form nouns and adjectives into verbs (the first conjugation -er is the more productive of the t w o ) and adverbial -ment to form adverbs from adjectives. Finally, there is some reliance on zero derivation i n that the infinitive o f any verb, when preceded by an article, becomes a noun designating the pertinent activity, e.g. le monter 'the c l i m b ' , le peser 'the w e i g h i n g ' . Composition or compounding i n French is much rarer than i n the Germanic languages, but there are many idiomatic or lexicalized compounds i n modern French. Moreover, the process o f compounding seems to be getting commoner, especially i n such domains as advertising. I n N O U N + N O U N compounds* the most productive pattern, the modifier follows the head, consistent w i t h the w o r d orders elsewhere i n the language, e.g. commis-voyageur 'salesman-traveler' ( = ' t r a v e l i n g salesman') or montre-bracelet 'watch-bracelet' ( = 'braceletwatch'). The gender o f such compounds is determined by the gender o f the head noun. Derivation is favored over composition i n the other major Romance lan­ guages too, w i t h suffixation being the most usual derivational device.

2.

Sources of Evidence

The sources for this chapter range from the general observations found i n early diary studies (e.g. Egger, 1887; Perez, 1892, Compayre, 1896), to the more detailed notes kept by observers like Deville (1891) or Vinson (1915), and the broader studies carried out by Bloch ( 1 9 2 1 , 1923, 1924), Descoeudres (1922),

698

Clark

and Guillaume (1927a, 1927b). M o r e recent studies include the extensive diary kept by G r é g o i r e (1937, 1947) and Cohen's (1969) summary o f his o w n observa­ tions and his comments on the process o f language acquisition. Other published diary sources for French include A i m a r d (1975), L i g h t b o w n (1977), F r a n ç o i s et al. (1978), and Fondet (1979). These observational studies are supplemented by a large number o f experi­ mental studies o f comprehension, production, and imitation. I n many cases, these complement the observational research directly and help paint the overall picture o f acquisition i n considerable detail. I w i l l draw extensively on both observational and experimental studies i n discussing the course children follow as they acquire French as a first language. The data available on the other Romance languages appear to be rather spar­ ser: for Italian there are a few published diary sources that provide some general outlines o f early development (e.g. Ferri, 1879; Delia Valle, 1931; Frontali, 1943-44). These are supplemented by some recent analyses o f longitudinal data collected by Parisi and Antinucci (e.g. Antinucci & M i l l e r , 1976; Antinucci & Parisi, 1973; Antinucci & Volterra, 1975; Bates, 1974, 1976), and by several experimental studies o f comprehension and production (e.g. A m m o n & Slobin, 1979; Bates & Rankin, 1979; Flores d'Arcais, 1978a, b; Johnston & Slobin 1979; Slobin, 1982, Slobin & Bever, 1982). For Portuguese, there is little published research. There are some studies being carried out i n B r a z i l , though, as w e l l as i n Portugal. De Lemos (1975, 1979) analyzed the emergence, longitudinally, o f aspectual contrasts; Costa (1976) and S i m ó e s and Stoel-Gammon (1979) d i d research on inflections; Pinto (1982) worked on locative prepositions; Figueira (1977, 1979) examined several facets o f lexical development, and Simonetti (1980) looked at early uses o f articles and demonstratives. Since most o f these studies are o f spontaneous production, there is little information available on children's comprehension o f Portuguese i n the domains studied so far. For Spanish, aside from the fairly extensive observations o f G i l i Gaya (1972a), Montes Giraldo (1974, 1976), and a recent ethnographic study o f the earlier stages o f acquisition (Eisenberg, 1982), many o f the studies available are rather sketchy. Moreover, data collected i n the U . S . have mostly been collected from children who are, or are becoming, bilingual (e.g. G o n z á l e z , 1970). The observational data on production are complemented by some studies o f com­ prehension (e.g. Bermejo, 1975; Lopez Ornat, 1975; E c h e v e r r í a , 1978), but there is still relatively little published research available on Spanish as a first language. Finally, for Rumanian, there have been some observational studies o f the acquisition o f inflections, o f overall development, and o f children's dialogues (e.g. Slama-Cazacu, 1961, 1962, 1973), but I have been unable to find any detailed diaries or studies o f comprehension. A l l the sources I consulted, plus a number o f others, are listed i n the b i b l i ­ ography at the end o f this review. This list contains nearly all the studies o f

7.

The A c q u i s i t i o n of Romance

699

syntactic, morphological, and lexical development I have been able to locate, plus a few studies o f input language. Wherever possible, I have cited published sources. 3. O v e r a l l C o u r s e o f D e v e l o p m e n t A m o n g children's first words i n French, those based on adult nouns predomi­ nate. The earliest verb forms seem to be based on infinitival forms, but initially these are rare. T w o - w o r d combinations begin to appear at about 1;6 or so, and w i t h them come the first inflectional endings on verbs, usually used to mark a resultant state. Irregular verbs at this stage (and for some time to come) may appear w i t h several stems, e.g. boiv-, boir-, or buv- from boire 'to d r i n k ' , and are frequently regularized, often by being added to the first conjugation. The w o r d order i n t w o - and three-word utterances is fairly variable and, i n some children, rarely seems to match the canonical S V O order o f declaratives. This could simply be an incidental result o f the fact that the children are not yet using person pronouns, combined w i t h a tendency towards postposing subject noun phrases to produce frequent V O S and V S orders. Other children appear to stick much more closely to an S V O or SV order i n nearly all their utterances. The first adverbials—maintenant ' n o w ' , aujourd'hui 'today', id 'here'—tend to appear i n final position, but may also occur initially. Expressions like void 'here (is)' are nearly always initial. Pronouns, articles, and prepositions begin to emerge soon after the first w o r d combinations, but the linguistic subsystems they represent take several years to acquire. For example, the first pronominal forms to be picked up are often the possessives mon ' m y ' and ton ' y o u r ' , between 1;9 or so and 2;0. But for selfreference i n talking about actions or states, young children w i l l use bébé or their own names, then maybe an emphatic pronoun like moi ' m e ' , and only after that begin to make regular use o f the pronoun je T w i t h verbs. W h e n they do begin to use je, i t is often i n the combination moije. Singular pronouns tend to appear before plural ones, and the first and second person contrasts among pronouns appear before third person ones. The latter may be more difficult because they also require gender agreement to be marked. However, gender marking is typ­ ically mastered only after number agreement by French-speaking children. The definite and indefinite articles also begin to appear at this stage, but adult­ like uses may take six years or more to appear. Y o u n g children, for instance, overuse definite articles. They often treat facts as i f they were k n o w n to their addressees, tagging noun phrases w i t h definite articles, instead o f new, tagged w i t h indefinite articles. Their article use also reveals some errors o f gender— errors that suggest that the pertinent articles may not be acquired along w i t h each noun. A g a i n , children appear to master number before gender, just as they do for pronouns. Children also take a long time to w o r k out the intricacies o f article-use in negative utterances. The first preposition to appear is à 'at, t o ' . I t is used i n both locative and possessive constructions (it is sometimes hard to distinguish the t w o at this

700

Clark

stage), e.g. Nini à bout for M m à bouche ' i n N i n i ' s m o u t h ' at 1;7,22, chaise à Pierre 'Pierre's chair' at 1;8. Other prepositions like sur ' o n ' , de 'of, f r o m ' , and par ' b y ' , emerge over the next year or so, but many take longer than that for children to w o r k out their meanings and identify the other terms they contrast with. Early questions are marked by intonation alone for yes/no forms and by reduced versions o f qu- words; neither question-type appears w i t h any inversion of the subject and verb. The commonest early qu- questions are 'where' and 'what' questions, but the forms used are not always equivalent to the adult's. 'Where' questions are introduced by où 'where' or où 'est 'where is' combined w i t h a noun, as i n où portenaie? for adult Où il est, mon porte-monnaie? 'where's m y purse?' and Où 'est, ton manteau? for adult Où il est, ton manteau? ' where's your coat'. ' W h a t ' questions are introduced by zero, as i n Tu fais? ' y o u do' for adult Qu'est-ce que tu fais? 'What are you doing?'; by a reduced form o f adult Qu'est-ce que c'est que ça? 'What's that', as i n Ceça? from C'est que ça? or Que c'est?, w i t h later (ungrammatical) uses o f que alone, as i n *Qu'ifait là? 'What's he doing there?' or by the nonconventional quoi 'what for what' as i n *Quoi à main quoi? 'what i n hand, what' for Qu'est-ce que tu tiens? ' W h a t are you holding?'. Early negatives are usually marked w i t h pas, from ne . . . pas ' n o t ' , a plus 'no m o r e ' , and non ' n o ' , either preceding or f o l l o w i n g the utterance being negated. (The ne that goes w i t h pas is typically omitted entirely i n colloquial adult French.) The placing o f negatives w i t h i n utterances develops later, as do the more complex negative forms like ne . . . rien 'not . . . anything', ne . . . jamais 'not . . . ever', or ne . . . personne 'not . . . anyone'. I n the next year, children also begin to use their first subordinate clauses and complements. The first subordinate clauses tend to be introduced by conjunctions like quand ' w h e n ' and si ' i f . Coordinate clauses at this stage are either juxtaposed w i t h no conjunction or joined by et ' a n d ' , et puis 'and then', or puis 'then' alone—the latter particularly i n describing sequences o f events. Besides adverbial clauses w i t h quand and si, children begin to use the complementizer que, apparently as an all-purpose marker o f any subordinate clause, including both subject and object relative clauses. The first complements produced, as i n English, tend to be those that follow verbs like vouloir ' w a n t ' , w i t h a first person subject i n both clauses. W i t h first person subjects, the complement contains an infinitival verb form rather than a full clause w i t h an inflected verb, e.g. Je veux venir T want to come' versus Je veux que Pierre vienne T want Pierre to come'. The latter complement-type also requires the subjunctive for the verb f o l l o w i n g vouloir. Different classes o f verbs place different restrictions on the possible complement forms, and children start out w i t h a very limited repertoire o f such constructions. Later acquisitions include some o f the compound tenses for talking about recent versus remote past; the conditional and subjunctive moods—used p r i marily for tempus irrealis; complements and subordinate clauses requiring the

7.

The A c q u i s i t i o n of Romance

701

subjunctive; certain quantifiers and some functions o f articles; partitive articles (used w i t h negatives); and some types o f relative clauses. A l s o fairly late is full mastery o f the different w o r d orders required w i t h direct and indirect object clitic pronouns, i n affirmative versus negative and i n indicative versus imperative sentence forms. M a n y o f the linguistic subsystems i n French, as i n other languages, take children several years to acquire. For instance, to integrate tense and aspect w i t h temporal adverbials i n French, children must have mastered the various functions of tense and aspect for talking about time relations and for assigning information a role i n the foreground or background at different stages i n a conversation or narrative. This i n turn requires coordination o f one's perspective as speaker w i t h that o f one's addressee, i n light o f what each participant i n the conversation already knows. Each successive " l a y e r " o f forms acquired, for instance the addition o f tense and aspect markers to the earlier system o f simple adverbials like maintenant ' n o w ' and puis 'then', leads to reorganization o f the system or subsystem as a whole, and each successive reorganization may lead to new kinds o f errors i n the utterances children produce. Similar factors play a role i n the acquisition o f definite and indefinite articles, as w e l l as i n the acquisition o f other noun modifiers such as possessive adjectives or relative clauses. I n each case, mastering forms w i t h multiple functions appears to take a long time. Children may find i t difficult to identify all the functions that can be carried by a particular w o r d or construction, and even when they have identified them, they are sometimes u n w i l l i n g to rely on a single form to express multiple meanings and may temporarily construct ungrammatical forms along­ side the conventional adult form to carry some o f those meanings (e.g. Karmiloff-Smith, 1979). T w o other systems that appear to be acquired late are complex negatives (e.g. Personne n'a jamais fait ca, 'no-one has ever done that') and counterfactual conditions. Children as o l d as 12 still make errors o f form i n the latter, using conditional verb forms instead o f the imperfect indicative i n the ' i f clause, e.g. *s'il aurait for s'il avait ' i f he had . . . ' . This f o r m , however, does appear i n colloquial French. M a n y o f the constructions requiring the subjunctive mood are also acquired late. Investigation o f still other systems w i l l probably reveal even more late acquisitions both i n French and i n other Romance languages. The overall course o f acquisition appears both like and unlike that noted for such languages as English or German. The similarities can probably be attributed to the fact that cognitive development is a major determinant o f some aspects o f language acquisition, especially during the earlier stages. For instance, the se­ quence o f temporal terms and their integration w i t h tense and aspect i n the verb shows strong parallels across French, English, and German. The differences can often be attributed to differences o f formal complexity. For example, French, which marks only number and not gender i n noun plurals, or Spanish, w h i c h marks number and gender, should both be simpler than Polish,

702

Clark

which marks number, gender, and case throughout the noun system, and has three genders rather than t w o . The Romance languages, o f course, also differ from each other, for example, i n the constraints governing the placement o f clitic pronouns. This i n turn makes for different word-order rules across languages for the analogous constructions. W h i l e i n French one might say / / veut me le donner 'he wants to:me it t o : g i v e ' , i n Spanish one w o u l d say Quiere dar-me-lo 'heiwants to:give-to:me-it' ( = 'He want to give it to m e ' ) . The object pronouns in both languages are clitic on the verb 'to g i v e ' , but i n French they precede i t and i n Spanish they f o l l o w . Differences like these among the Romance lan­ guages offer interesting domains i n w h i c h to assess the effects o f formal com­ plexity. A t the same time, domains i n w h i c h there are strong similarities across the Romance languages tend to produce close similarities i n the pattern o f ac­ quisition. Such parallels are presumably due both to formal similarities i n what has to be acquired, and a common cognitive basis.

THE D A T A I n this section I first consider some o f the typical errors that have been observed in the early speech o f children acquiring Romance languages, and then summa­ rize what is k n o w n about error-free acquisition and about the t i m i n g o f acquisi­ t i o n — w h i c h constructions are typically acquired early and w h i c h late. I then take up the acquisition o f articles and o f possessives as examples o f domains i n which children organize and reorganize what they k n o w as they learn more about how particular forms are used. I conclude this section w i t h a discussion o f what children k n o w about lexical structure and word-formation, and how this k n o w l ­ edge is revealed by their attempts to extend the vocabulary they have at their disposal. 4. T y p i c a l E r r o r s What are the typical errors children produce i n the course o f acquiring French? A n d are the patterns for French also found i n the acquisition o f other Romance languages? I w i l l present the different kinds o f errors noted for French first and then provide as much comparative data as possible from other Romance languages. 4 . 1 . Overregularization. Overregularization is very c o m m o n , particularly for verbs. For children acquiring French, the preferred model for regularization appears to be the paradigm provided by first conjugation verbs (Hiriartborde, 1973), the pattern carried by some 90% o f verbs i n French. For instance, for infinitives, 2-, 3-, and 4-year-olds commonly use such forms as *rier from the 4

Starred examples (*) are utterances or examples unacceptable in the adult language either because the form is incorrect or because another expression is the one conventionally used. 4

7.

The A c q u i s i t i o n of Romance

703

stem ri- plus -er for rire 'to l a u g h ' , *buver from the stem buv- for boire 'to d r i n k ' , * batter from the stem batt- for tofre 'to h i t ' , *tiender from the stem tien- for temV 'to h o l d ' , or *éteigner from the stem éteign- for éteindre 'to turn off. For past particles, they construct forms like * coure from the stem cour- for couru ' r a n ' , * coudé from the stem cow J- for cowsw ' s e w n ' , * mette from the stem mctf- for mis ' p u t ' , * morde from the stem mord- for mordu ' b i t ' , * é t e i n d é from the stem éteind- for etemf 'turned o f f , *prendu from prend- for / ? m ' t o o k ' , * pleuve from pleuv- for 'rained', and *buvu from Z?wv- for 'drank'. M a n y irregular verbs have t w o different stem forms i n the present tense, and children may make use o f either one, e.g. boi- or buv- from boire 'to d r i n k ' , veuor voul- from vouloir 'to w a n t ' , r/en- or ten- from terar 'to h o l d ' . I n fact, the same children show considerable variation i n which stem they pick initially as the form to use i n regularizing such third conjugation verbs. The existence o f t w o or more different stems also leads to side-by-side usage by the same child o f t w o forms o f an infinitive or past participle, apparently w i t h the same meaning. For instance, one 2-year-old used both *mouri from mourir 'to die' and mort 'dead'; another used pris and *prendu ' t o o k ' i n successive utterances. Another c h i l d , aged 3;9, used *tiendre alongside tenir 'to h o l d ' , and yet another, also 3;9, used the correct ouverte alongside *ouvrie 'opened' from ouvrir (see D e v i l l e , 1891; Egger, 1887; Fondet, 1979; G r é g o i r e , 1947; Guillaume 1927b; Suppes et a l . , 1973; Vinson, 1 9 1 5 - 1 6 ) . H o w prevalent such fluctuations i n form are, and whether children regard the different forms as equivalent i n meaning has yet to be established. One possible reason for the prevalence o f regularizations based on the first conjugation pattern is the sheer frequency o f that pattern i n French. One might therefore expect that first conjugation verbs w o u l d predominate i n young children's speech, but that turns out not to be the case. Guillaume (1927b) found that although first conjugation verbs were the majority when one counted verb types used by 2- to 4-year-olds, they made up only a third o f the tokens P R O D U C E D by the children he recorded. The distributions, by conjugations, are shown i n Table 7 . 1 . Moreover, the most frequently used verbs ( w i t h only three exceptions) were all irregular third conjugation ones: être 'to be', avoir 'to have\ faire 'to d o ' , vouloir 'to w a n t ' , mettre 'to p u t ' , prendre 'to take', and voir 'to see', plus aller 'to g o , ' (irreg.). The only first conjugation verbs used w i t h any frequency were donner 'to g i v e ' , tomber 'to f a l l ' , and casser 'to break'. The effect o f frequency, then, is an indirect one. Children hear many different first conjugation verbs and appear to extract a paradigm or schema (Bybee & Slobin, 1982) based on that conjugation that they then impose on frequently used second and third conjugation verbs. The kinds o f overregularization been noted for children acquiring model o f the first conjugation (-ar comprises the majority o f Spanish

noted for French-speaking children have also Spanish. Verbs are again regularized on the infinitives i n Spanish), w h i c h , as i n French, verbs. B u t , again as i n French, many o f the

704

Clark TABLE 7.1 Percentages of Verbs f r o m Each Conjugation in French Children A g e d 2 t o 4 (Guillaume, 1927b) CONJUGATION

1 -er 2 -ir 3 remainder

TYPES

TOKENS

76 6 18

36 6 58

NOTE: Guillaume'S 3rd and 4th conjugation figures have been collapsed to conform to the clas­ sification in Grevisse (1964).

verbs used most frequently come from one o f the other conjugations. For exam­ ple, 3-year-olds form infinitives like *traigar from the stem traig- for traer 'to bring' or *pongar from the stem pong- for poner 'to p u t ' , and, from a 5-yearold: *juegar from the stem jueg- for jugar 'to play' and *cresar from the stem crec- for crecer 'to g r o w ' . Spanish-speaking children also rely on the first conjugation pattern i n forming the simple past tense, e.g. *saló for salió 'he went out' ( I N F salir), *movó for movió 'he m o v e d ' ( I N F mover), *metó for metió 'he put i n ' ( I N F meter), * perdó for perdió 'he lost' ( I N F perder), and i n forming the first person present o f common irregular verbs, e.g. *tieno from the stem tien- for tengo T have' ( I N F tener), *teno from the stem ten- for tengo ( i b i d . ) , *sabo from the stem sab- for se T k n o w ' ( I N F saber). The same regularizing tendency appears w i t h the present participle, e.g. *tengan(d)o from the stem teng- for teniendo 'holding' ( I N F tener). There also seem to be some confusions o f form between second (-er) and third (-ir) conjugation verbs. M u c h as i n French, most regularization errors for irregular verbs result from the choice o f a single stem for all the forms used, e.g. either teng- or tien- for tener. Regularization errors i n Spanish appear very early (Montes Giraldo, 1976) and are still prevalent at 5 and older (Brisk, 1974; Dato, 1971; G i l i Gaya, 1972a; Medina-Nguyen, 1978). A n d , again like French, such errors appear to be partic­ ularly frequent w i t h common irregular verbs, notably hacer 'to d o ' , poner 'to put', and tener 'to have'. The data on such regularizations i n Italian are sparser, but Frontali ( 1 9 4 3 - 4 4 ) noted that verbs were frequently overregularized by children, again on the first conjugation model (-are), the most widespread one for Italian. For example, his daughters formed first conjugation past participles like *diciato from the stem die- for adult detto 'said' ( I N F dire), *cuciato for cucito 'sewn, mended' ( I N F cucire), and *leggiato for letto 'read' ( I N F leggere). Irregular verbs were often first regularized w i t h the construction o f first conjugation past participles, e.g.

7.

The Acquisition of Romance

705

*spingiato for spinto 'pushed' ( I N F spingere). Frontali pointed out that such regularizations usually seemed to follow a period o f correct uses on the part o f younger children (see also Francescato, 1964, 1978). The same observation has been made for French and other languages (e.g. Kuczaj, 1977). Guillaume (1927b) argued that the initial correct uses were simply forms that had been picked up as whole units, and that regularization errors entered as children began to analyze the verb forms being acquired (see Bowerman, 1978). Similar overregularizations have been observed for Portuguese, both i n spon­ taneous speech (Simoes & Stoel-Gammon, 1978) and i n elicitation tasks (Costa, 1976; Mediano, 1976). Overregularizations o f nouns i n forming the plural are practically nonexistent in French. This, however, is only because the singular and plural forms o f most nouns are pronounced i n exactly the same way. (Number is indicated through the form o f the article, and i n certain cases through verb and adjective agreement.) However, when they get a chance, children acquiring French do regularize noun plurals. For instance, some children pick the singular noun stem cheval 'horse' for both singular and plural, while others pick the plural chevaux for both (e.g. Suppes, Smith & Leveille, 1973). I n Spanish, children appear to overregularize nouns so that they conform to the major paradigms as far as gender marking is concerned. For instance, nouns ending i n a consonant may have a final v o w e l added, typically -a for feminine and -o for masculine. This not only clarifies their gender but also regularizes word stress by placing it on the pre-final rather than the final syllable o f the w o r d . Thus, Montes Giraldo (1976) noted one child (2; 10) who used *unafola for una flor ' f l o w e r ' ( F E M ) and *la mujala for la mujer 'the w o m a n ' ( F E M ) . Another child (2;11) used *una mana for una mano 'hand' ( F E M ) , and another (2;6) used papelo for papel 'paper' ( M A S C ) . 4.2. Gender. The acquisition o f gender i n a language like French w o u l d appear to pose certain problems since there is no consistent semantic basis to gender assignments. Natural gender provides a basis for classifying a few nouns, but provides no clues to W H E R E i n the language gender has to be markeS, for example, i n articles, adjectives, certain participial forms, and so on. The forms of words, on the other hand, appear to offer a more consistent basis for classifica­ tion according to gender. Certain endings are typically masculine, others typ­ ically feminine. A n d , according to Tucker et al. (1968), adult speakers o f French are good at deciding on w o r d gender from the w o r d form alone. Some errors o f gender marking and agreement appear i n the early stages o f acquiring French. L i g h t b o w n (1977, p . 70), for example, noted a number o f occasions where the children she was observing used the wrong gender article w i t h a noun, as i n */a petit bouton ' t h e . F E M little button' for adult le petit bouton, *a bateau a voile ' t h e : F E M boat w i t h sails' for adult le bateau a voile 'the sailing boat', or *maman fait un maison ' m u m m y is making a : M A S C

706

Clark

house' for adult maman fait une maison ' m u m m y is making a house'. This even occurred immediately after appropriate adult uses, as i n the following exchanges: 5

(a) Mo: Alors, prends un autre couvercle. Ch: (b) Mo: Ch:

*La couvercle.

'Well, take another lid then.'

'the:FEM lid' (= le couvercle)

C'est pas la table; c'est l'assiette. C'est l'assiette. *Le 'siette.

Mo:

L'assiette.

'The plate.'

Ch:

L'assiette.

'The plate.'

'It's not the table; it's the plate.'

'It's the plate. The:MASC plate.'

Errors i n the choice o f articles provide evidence against the view that children learn each noun i n combination w i t h an appropriate (gender-marked) article. One reason they don't may be that each noun can occur w i t h at least four different forms o f the article, e.g., la chaise, une chaise, les chaises, and des chaises 'the:SG, a, the:PL, some chairs(s)'. I t w o u l d seem easier for children to start out w i t h what's invariant across these forms, namely chaise alone. A n d , as Lightbown (1977) observed, nouns i n early w o r d combinations i n French typically occur without any article. B y age 3, children appear to make few errors i n their choices o f articles. Occasional late gender errors are reported by Ervin-Tripp (1974) and Valette (1964) for second-language learners aged 5 or 6, as w e l l as by G r é g o i r e (1947). Reports o f such errors i n spontaneous usage during the early stages, however, are not as common as errors i n adjective-noun agreement, e.g. the use o f *gros porte for grosse porte ' b i g door' at 2;5. Both agreement and article errors are corrected automatically by adults i n any repetitions or expan­ sions o f what the child has said. Gender errors w i t h articles also appear during the early stages o f acquiring Spanish. Mazeika (1973) for example, observed a 2-year-old w h o , when he used articles, relied almost exclusively on the feminine singular la (53 to 2 i n the corpus analyzed). The same child also made errors on the indefinite article, e.g. *w« camisa 'a shirt' for una camisa (see also Tolbert, 1978). Brisk (1976) noted a similar over-dependence on the feminine article—for 76% o f masculine nouns in spontaneous speech; the feminine article was also the one most over-used i n children aged 6 and older i n an elicitation task she designed to study knowledge of gender. One possibility is that la, the feminine article, is more salient for children because its form is always the same. I n contrast, the masculine singular article, el, takes on the form al i n combination w i t h a 'at, t o ' and del w i t h de 'of, f r o m ' . These contractions probably make i t harder for children to discern the form el i n those contexts. (Whether French-speaking children ever choose just one o f the articles and overuse it i n this way is not clear from the data available.) 5

I am grateful to Patsy M. Lightbown for supplying these examples.

7.

T h e A c q u i s i t i o n o f Romance

707

Other gender-linked errors that remain prevalent are errors o f agreement between nouns and co-referent pronouns. I n French, children tend to rely almost exclusively on the masculine singular pronoun // ' h e ' . For instance, one finds utterances like (2;3,24) */ va nir, i va venir 'he's going to come, he's going to come' for elle va venir 'she's going to come' or (2;4,5) */ a sese des boîtes, la 'tite fille 'he looked for the boxes, the little g i r l ' for elle a cherché des boîtes, la petite fille 'she looked for the boxes, the little g i r l ' ( = 'the little g i r l looked for the boxes') (e.g., G r é g o i r e , 1947; Guillaume, 1927b). I n both examples, the pronoun fails to agree i n gender w i t h the subject noun phrase. Children typically use // i n place o f elle 'she' at least to age 3 and often beyond. They acquire the plural feminine pronoun, elles 'they', even later than the singular. Cohen (1927) noted numerous errors—uses o f Us, the masculine plural 'they,' for elles—in children up to age 9. Learning how to mark gender appropriately i n this domain is clearly complicated by having to integrate t w o distinct systems: pronouns on the one hand, where the choice o f any third person f o r m is governed by the gender o f the antecedent noun phrase (whether explicit or i m p l i c i t ) and articles on the other, w h i c h , i n French at least, are the primary indicators o f gender for their accompanying nouns. (Yet another complication is probably the existence of natural gender, assigned on the basis o f sex, for animate referents o f certain nouns.) 6

4.3. Person and Number. Person i n French is marked primarily by pronouns and only secondarily by the form o f the verb. I n French, pronunciation for most verbs does not distinguish between the three singular persons, although i t does distinguish the three plural forms. I n the other Romance languages, person and number are generally marked directly i n the verb form for each person whether singular or plural. A n d , unlike i n French, pronouns are not obligatory, for instance, i n Italian or Spanish. Typical errors o f person and number consist o f errors i n pronoun use, i n verb form, and i n agreement between subject pronouns or noun phrases and their accompanying verbs. Pronouns start to emerge around age 2. Guillaume (1927b) reported very early uses o f the third person singular // for both masculine and feminine referents, e.g. (où) il est maman chérie? 'where is he, dear m o m m y ' , where this use o f // est is probably formulaic (Fillmore, 1979; L i g h t b o w n , 1977). First person pronouns, according to G r é g o i r e (1947, p . 95), emerge around 2;0 to 2;6, w i t h je and moi T , m e : E M P H ' appearing at the same time (at 2;6 and 2;7 for one c h i l d , both together at 2;5 for the other). F r o m then o n , these pronouns are usually used together i n the form moi je.

This form is typically pronounced as [i] and could be regarded as a neutralized pronoun form, with gender unspecified, except that this is also the colloquial adult pronunciation always given to It always contrasts for adults with that given to die (phonetically [e] or [el]). 6

708

Clark

The next pronouns acquired are the second and third person—in French toi or tu and masculine // (Cohen, 1969). Although Frontali ( 1 9 4 3 - 4 4 ) reported for Italian that his children confused the first (io, T ) and second (tu, ' y o u ' ) person forms for a few months before they began to use io i n self-reference, such errors in pronoun use tend to be prevalent but poorly documented. Nonetheless, the errors that have been recorded seem to parallel exactly those observed i n other languages. The difficulty children seem to have here is w i t h the shifting nature o f pronouns like T and ' y o u ' such that T is whoever is speaking and ' y o u ' whoever is the addressee (Clark, 1978a). The next pronouns to emerge i n French are the second and third person plural, w i t h the first person plural appearing later still. L i k e adult speakers o f French, children tend to use the third person impersonal on 'one' rather than nous ' w e , us', from as early as age 3. This probably contributes to the late acquisition o f the first person plural pronoun. The emergence o f each pronoun form is typically marked by a brief period o f confusion where the new form is used inap­ propriately, presumably as children try to w o r k out where each new form belongs in the pronoun system ( G r é g o i r e , 1947). The order o f pronouns i n French is similar to the order o f emergence for verb forms i n Italian, where person is marked directly i n the verb. First comes the imperative, then the third person indicative, followed by the first person f o r m , and then, only some months later, the first person plural form. These forms emerge by age 2 or 2Vi (Frontali, 1 9 4 3 - 4 4 ) . Because the third person verb form emerges before the first person one, and because the child appears to start by making third person reference to himself (bébé, o w n name, etc.), there are frequent errors o f person agreement from the second year on. For example, Montes Giraldo (1976) noted the f o l l o w i n g i n Spanish: *Le tito caco for Me quito el saco 'he, I get r i d o f the bag' at 1 ; 1 1 , * L A loia for Me duele '(it) hurts her, me' at 2;9, and *Dónde se stento yo? fox Dónde me siento yo 'where am h e / I to sit' at 4;3. There are similar errors o f form (and hence o f reference, as i n the last Spanish example) for reflexive verbs i n French (e.g. G r é g o i r e , 1947), but, on the whole, such errors do not seem to be very prevalent. Children also make errors i n number agreement, but these too tend to be early errors. They are rare for nouns i n French because there are so few nouns where the pronunciation changes from singular to plural. A l l children need do is master the plural definite article, les, for both genders. They do make errors o f agree­ ment, though, between subject and verb, usually by combining a plural subject w i t h a singular verb, or w i t h inappropriate singular pronoun and verb combina­ tions. The former type o f error can only be detected, o f course, w i t h second and third conjugation verbs, where the singular and plural third person indicative forms are pronounced differently, as i n utterances like *Les chats, il vient for Us viennent 'the cats, he's, they're c o m i n g ' . Most such errors seem to involve a singular verb w i t h a plural subject, rarely the reverse (e.g. G r é g o i r e , 1947;

7.

The A c q u i s i t i o n o f Romance

709

LaBelle, 1976). A g a i n , such forms are common i n some varieties o f colloquial French. I n Spanish, number errors are apparently rare although nouns as w e l l as articles are marked for number ( w i t h an added -s i n most cases). Note that the overt marking o f consonant-final nouns by an -o ( M A S C ) or an -a ( F E M ) observed by Montes Giraldo (1976) makes plurals easier to form: children have simply to add a final -s to all nouns and to the feminine article (la to las). They do have to learn the form los, though, as the plural for masculine singular el 'the'. A n d an error that does occur quite commonly is the pluralization o f mass nouns. Medina-Nguyen (1978), for example, noted such instances as *tus ropas for tu ropa 'your clothes' at 4;4 and *zacates for zacate 'grass' at 3;0. Errors o f this type are common i n other languages too (e.g. Gordon, 1982), and generally seem to be the result o f overregularization, w i t h singular-form mass nouns being treated as i f they too were count nouns marked for number w i t h the regular plural ending. Overall, number errors are not as common as person errors, although they quite often appear to cause the problem when children make agreement errors. Number appears to be mastered first i n nouns and articles for all the Romance languages, and only later i n pronouns and then verbs. Number i n verbs, o f course, may be more complicated since the same inflection typically marks both person and number. A l t h o u g h person and number errors appear and then disappear fairly early, their full range has probably not been mapped for any o f the Romance languages. There has also been little analysis o f the kinds o f referential errors children make w i t h person and number. 4.4. Word Order. There seems to be considerable variation i n the w o r d order reported for young children acquiring French, and although some o f the variations appear erroneous, the data are difficult to interpret. The t w o monolingual children observed by L i g h t b o w n (1977), for example, differed i n the degree to w h i c h they observed S V or S V O w o r d order i n describing actions. Daniel (aged 1;8 at the start o f the study) observed SV order w i t h intransitive verbs over 75% o f the time, and kept to S V O order over 90% o f the time w i t h transitive verbs. (His other utterances were VS and, presumably, V O S order.) Nathalie (1;10) observed SV order w i t h intransitive verbs only 30% o f the time, so her normal order was VS at this stage, but she kept closer to the canonical S V O order w i t h transitive verbs ( w i t h S V O used about 75% o f the time). One problem, o f course, i n talking about w o r d order from the t w o - w o r d stage on is the nature o f the sequences—whether the terms fall under the same intonational contour or whether there is any pause between the intransitive verb, say, and the subject. Guillaume (1927a) noted few noncanonical orders, and where he d i d cite one (fermée fenêtre 'shut + F E M . S G w i n d o w , ' = ' w i n d o w shut'), he suggested that it was based on an adult Elle est fermée, la fenêtre ' I t ' s shut, the w i n d o w ' , w i t h right dislocation o f the subject noun phrase. Such examples make it very

710

Clark

clear how carefully one needs to take into account any intonational evidence i n assessing w o r d order i n a language like French. Bloch (1924), on the other hand, cited a number o f noncanonical orders, e.g. *Chapeau chercher 'hat look:for' ' ( I ) look for h a t ' ) , or at 2 ; 1 , *Maman let li for maman lettre lire ' M u m m y letter read' ( = ' M u m m y read letter'), w i t h the object noun preceding a transitive verb. He also cited one example o f three different orders produced i n succession by one child: the S V O Papa couper cheveux 'daddy cut hair', followed a few seconds later by: * Cheveux couper papa, couper cheveux papa ( w i t h O V S and V O S orders respectively). Such sequences suggest that the child was trying the variant orders i n a further attempt to make himself heard. Descoeudres (1922), like B l o c h and L i g h t b o w n , noted a number of variants on the normal w o r d order i n one child (2;9), e.g. V S i n *A passé tram 'went:by t r a m , ' ( ^ ' t r a m went b y ' ) and *Mord le vouvou 'bites the doggie' ( = ' t h e doggie bites'), O V S i n *Tout mangé moi ' a l l eaten m e " ( = ' m e eaten a l l ' ) , V S O i n * A vu moi le papa de Roger 'has:3SG seen me the daddy o f Roger' ( = ' I saw Roger's daddy'), and V O S i n such utterances as *Sais tout moi ' k n o w everything me' ( = T k n o w everything') or ^Chercher le jus moi ' l o o k : f o r + I N F the gravy me' ( = 'me look for the g r a v y ' ) . A l l four, o f course, might be suscepti­ ble to the explanation offered by Guillaume, namely that they are based on the order found i n adult dislocated constructions. Such variability i n w o r d order appears very common i n the speech o f children acquiring French (Leroy, 1975; Sabeau-Jouannet, 1975). What is difficult to tell is which variants should be counted as word-order errors and w h i c h not. This problem is even more acute when one considers w o r d order i n Spanish or Italian where the pragmatically governed variations are even more extensive than i n French. Once children acquire their first pronouns, they begin to use more utterances like the f o l l o w i n g , all i n self-reference (Guillaume, 1927a): *Pierre il fait 'Pierre he does,' ( = 'Pierre can do i t ) , * Pierre il sait 'Pierre he k n o w s , ' ( = 'Pierre k n o w s ' ) , Pierre il peut pas 'Pierre can not' ( = 'Pierre can do i t ' ) , * Pierre il sait 'Pierre he breaks,' ( = 'Pierre breaks'), all at 1;8, and Pierre il pleure 'Pierre he cries' ( = 'Pierre is c r y i n g ' ) at 1;9. Essentially this child is repeating the subject (Pierre) via the pronoun // 'he' w i t h left-dislocation o f the subject noun phrase. French allows both left and right dislocation, and, according to V i r b e l (1975), French-speaking 4-year-olds nearly always use such constructions i n preference to plain noun phrase subjects. I n assertions, they tend to prefer right dislocation, e.g. Il est pas mort le roi! 'he's not dead, the k i n g ' from M a t h i e u , aged 4;0, as they do i n qu- questions, e.g. Comment i va faire papa? ' h o w he is going to manage, daddy' ( = ' h o w ' s daddy going to manage?'), from Sarah, aged 4;0. I n yes/no questions, left- and right-dislocations appear to be equally c o m m o n . Overall, V i r b e l found that the usual w o r d order i n 4-year-old speech was that o f sentences w i t h a right-dislocated subject noun phrase (see also LaBelle, 1976). (It was much rarer for children to extract an object noun phrase, although that d i d occur occasionally.) The prevalence o f dislocation yields only at around age 9 to

7.

T h e A c q u i s i t i o n of Romance

711

a larger porportion o f plain subject noun phrases (Faïta, 1974). A n d i n adult colloquial spoken French, constructions w i t h extracted noun phrases are much more the norm than many grammars w o u l d suggest. The acquisition data available so far suggest that w o r d order at the t w o - w o r d stage and even later, before children make really consistent use o f articles or pronouns i n French, is very variable, especially w i t h intransitive verbs. SV and VS are both common. W i t h transitive verbs, some children appear to stick more closely to a canonical S V O order, while others prefer to move the subject noun rightwards ( V O S ) . What is not clear from the accounts given is (a) how much such variations might be attributed to partial imitation o f preceding adult utterances, or (b) whether all such sequences fall under a single intonation contour. I f not, they could simply be sequences o f single words (e.g. F ô n a g y , 1972) and thus not represent any violations o f w o r d order at a l l . Another f o r m o f word-order error appears to be due to the acquisition o f formulae—set expressions not yet analyzed into their constituent parts by the children speaking (cf. F i l l m o r e , 1976, 1979). For example, Vinson ( 1 9 1 5 - 1 6 ) reported that his son at 2;3 used il y a 'there is/are' both initially and finally. For instance, i n initial position (the appropriate one), one finds *Ana baso zadë boku (=Il-y-a oiseaux jardin beaucoup it-there-is bird(s) garden lots') for a d u l t / / y a beaucoup d'oiseaux dans le jardin 'there are lots o f birds i n the garden', or * Kakâ zadë pat ana (-Canards Jardin Plant il-y-a 'ducks garden plant it-thereis') for adult II y a de canards dans le Jardin de Plantes 'there are ducks i n the Jardin des Plantes' [a zoo]. Notice that this child also appears to have made an order error i n placing the quantifier beaucoup 'lots, many' i n the first instance cited here, but this w o u l d depend on the intonation used. 4

7

Yet another k i n d o f word-order error is common w i t h negative placement, both i n imperatives and i n indicative assertions. For instance, Guillaume (1927b, p. 214) observed several 3-year-olds use wrong orders w i t h negatives. One, for instance, used *Cache-le pas 'hide-it not' for (Ne) le cache pas '(not) i t hide not' ( = ' d o n ' t hide i t ' ) and another (3;6) used *Fais-le pas tomber 'make-it not f a l l ' for (Ne) le fais pas tomber '(not) i t make not f a l l ' ( = ' d o n ' t make h i m / i t f a l l ' ) . The object pronoun i n negative imperatives should precede the verb. Negatives like rien ' n o t h i n g ' also cause problems: for example, one child (3;2) said *Je lui ai fait rien ' I to:him have done nothing' for Je lui ai rien fait T to:him have nothing done' ( = ' I d i d n ' t do anything to h i m ' ) . The negative element, whether pas or rien, usually follows the finite verb, which i n compound tenses is the auxiliary avoir (as i n the example just cited) or être. However, it may be postponed for emphasis i n adult speech. 8

The In/ in Ana suggests this child was aiming at (//) y en a rather than // y a. Word-order errors of this type still appear in the speech of children as old as 9 and it is quite possible such forms are also used by adults, at least when speaking to young children. I would like to thank Annette Karmiloff-Smith and Patsy M. Lightbown for discussion on this point. 7

8

712

Clark

Finally, a few researchers have noted spontaneous repairs to w o r d order, even in quite young children's speech. For example, Guillaume (1927b) noted the following sequence from a child aged 2 ; 1 : * Il faut moi je l'écrive . . . * il faut moi qu'il écrive . . . *il faut moi que je l'écrive . . . il faut qu'elle écrive, Zézette T must write him/her . . . he must write . . . I must write him/her . . . she must w r i t e , Z é z e t t e ' (probably for adult: Il faut que j'écrive à Zézette T must write to Z é z e t t e ' ) and from another child age 2;5: A non pris . . . Renée . . . ça . . . Renée a pris ça ' d i d n ' t take . . . R e n é e . . . that . . . R e n é e took that'). Such spontaneous corrections suggest that the children's difficulty may sometimes lie as much w i t h getting the pieces o f an utterance straight i n production as w i t h w o r d order per se (see further Clark, 1982b; Karmiloff-Smith, 1981). Word-order patterns, however, have not been studied extensively and the prevalence o f actual errors past the t w o - or three-word stage is poorly docu­ mented. Sinclair and Bronckart (1972) looked at the interpretations 3- to 8-yearolds made o f bare nouns (without articles) and infinitive verb forms i n N V V , V N V , V V N , N N V , N V N , and V N N sequences. The younger children tended to treat the verbs as intransitive, w i t h conjoined subjects, or else as imperative. Only around age 5 or 6, w h i c h seems comparatively late, d i d they begin treating the first o f t w o nouns as a subject and the second as a direct object. These data, though, are problematic because o f the absence o f all the normal articles and verb inflections. The children i n this study may have had recourse to special strategies to cope w i t h such odd-sounding sentences. Aside from V i r b e l ' s (1975) detailed study o f the usual sentential structures used by 4-year-olds, both i n ordinary conversation and i n narratives, there has been little observational or experimental study o f w o r d order, changes i n w o r d order, or o f the problems associated w i t h inserting negative particles and their interactions w i t h other sentential elements like object pronouns. One recent exception is the series o f studies by A n n i b a l d i - V i o n (1980), w h o compared syntactic and lexical-pragmatic information i n 3- to 6-year-olds' comprehension of word order. Although the younger children showed a greater tendency to rely on pragmatic information, all o f them appeared to base their interpretations on both kinds o f information, w i t h their choices dependent on the type o f noncanonical form to be interpreted. What A n n i b a l d i - V i o n d i d not find was any consistent reliance on positional information to identify grammatical subjects or objects (e.g. first noun phrase, or the noun phrase f o l l o w i n g the verb). This suggests that even quite young children are aware that w o r d order per se, i n French, is not the only or even the best clue to identifying grammatical relations. Young children acquiring Italian also use variable w o r d orders. For example, Bates (1976) found the orders SV and V S were both used by t w o children at the two-word stage i n Italian, w i t h a statistical preference for the V S order. H o w e v ­ er, once the children began using longer utterances ( w i t h a mean length o f three morphemes), their w o r d order fluctuated from one recording session to the next, sometimes being predominantly S V O , at other times V O S (see Bates, 1976, pp.

7.

The A c q u i s i t i o n of Romance

713

188-189). W i t h increasing age, though, they made more use o f S V O . I n comprehension, though, Slobin and Bever (1982) found that young Italian children (aged 2 to 5) relied heavily on order to identify subjects and objects i n a toymoving task. The picture i n Italian for production is complicated by the complexity o f the pragmatic options available, which allow for several alternatives to an S V O order. Moreover, since the subject is inferrable from the form o f the verb, subject noun phrases and pronouns are frequently omitted. The assignment o f subject and object is often dependent on the preceding discourse, on the presence o f a clitic pronoun preceding the verb to pick out the object, as i n / / libro lo compro io 'the book i t buy I ' ( = T am buying the b o o k ' ) ; on contrastive stress; on semantic distinctions that restrict the role o f subject to only one o f the entities denoted, as in La mela mangia Giovanni 'the apple eats John' ( = 'John is eating the apple'); on number agreement i n the verb; or on person i n the verb ( w i t h first and second person assigning speaker or addressee as subject). One or more o f these cues is always available i n Italian, w h i c h may account for the relative lack o f reliance on word order per se to mark grammatical relations i n the speech o f young children (see M a c W h i n n e y & Bates, 1978). In Spanish, the commonest w o r d order i n young children's spontaneous speech appears to be S V O , but other orders also appear. For instance, E c h e v e r r í a (1978, p . 65) cited some 2-year-olds who used both O V S and V S orders, e.g. (2;3) Hatos chiches teñe la tia Pachelita 'lotsiof knick-knacks has the aunt Pachelita' ( = 'aunt Pachelita has lots o f knick-knacks'), (2;11) Ta durmiendo un pollito 'is sleeping a c h i c k ' ( = ' a chick is sleeping'). T o check on children's comprehension o f different possible w o r d orders i n Spanish, he gave 5- to 9year-olds instructions to act out w i t h toys, using the verbs perseguir ' f o l l o w ' , pegar ' h i t ' , and botar ' t h r o w ' , i n four different w o r d orders: S V O , S O V , O S V , and O V S . I n Spanish, as i n Italian, person (and hence the subject) is marked i n the verb i n all tenses. Moreover, the animate direct object for many verbs is marked by the preposition a, e.g. Veo al hombre T + s e e t o + t h e man' ( = T see the m a n ' ) , Persigue a su hermano 'he follows to his brother' ( = 'he follows his brother'). E c h e v e r r í a found that the youngest children understood the S V O and SOV orders best. A s children got older, they began to do better on O S V , and eventually on O V S too. (He d i d not report any analysis o f the errors they made.) Noncanonical w o r d orders, o f course, may be much more transparent i n a discourse setting than when presented i n isolation as instructions to move toys. Further information is needed here on how younger children (and adults) use different w o r d orders i n conversation, and how pragmatic factors affect the orders used (cf. Bates, M c N e w , M a c W h i n n e y , Devescovi, & Smith, 1982). 4.5. Pronoun Placement. The pronominal system i n French consists o f three subsystems: emphatic or tonic pronouns, on the one hand, used i n dislocated noun phrase position and i n prepositional phrases imoi, toi, lui), and clitic pronouns, on the other, used for subjects (je, tu, it) and for objects preceding the

714

Clark

verb (me, te, le, la, les, etc., for direct objects; me, te, lui, leur for indirect ones). W i t h the first and second person indirect objects ( 1 0 ) , the indirect pro­ nouns precede the direct object ( D O ) ones: Marie me le donne ' M a r i e to:me it gives' ( = ' M a r i e gives it to m e ' ) , but w i t h third person indirect objects, the order is reversed to D O before 10: Marie

le lui donne

'Maries it to:him gives'

( = ' M a r i e gives it to h i m ' ) . T w o bilingual (French-English) children studied by Connors, N u c k l e , and Greene (1981) used double clitic pronouns, but tended to avoid third person indirect objects where the order is D O - I O rather than I O - D O used everywhere else. Connors et al. followed up these observations by collect­ ing experimental data i n a production task designed to elicit such constructions. Monolingual French-speaking 4-year-olds made frequent order errors. For exam­ ple, they w o u l d say things like *Le monsieur leur le donne 'the man to:them i t gives' instead o f Le monsieur le leur donne 'the man it to:them gives' ( = ' t h e man gives it to t h e m ' ) . The children also made order errors i n an imitation task, and consistently placed the indirect object pronoun before the direct object one. I n a survey o f 4- and 8-year-olds' spontaneous speech, however, Bautier-Castaing (1977) found no cases o f double clitic pronouns being used. I n fact, the younger children rarely used even direct object clitics alone. I n answering ques­ tions, for instance, they typically repeated the full object noun phrase, as i n Qu'est-ce qu'il a fait de ses chaussures?—//

a enlevé ses chaussures

'What did

he do w i t h his shoes?—He took his shoes o f f . The older children, though, w o u l d reply w i t h / / les a enlevées

' H e took them o f f . These data suggest that

clitic object pronouns are a fairly late acquisition. Children also used tonic forms alone as subjects, where adults use the tonic pronoun i n combination w i t h a clitic subject pronoun. For example, Connors et al. reported that one o f the 4-year-olds they followed longitudinally used utter­ ances like * M o / l'a oublié

'me it has:3SG forgotten' instead of Moi, je l'ai

oublié

'me, I it h a v e : l S G forgotten' ( = T forgot i t ' ) or *Toi le sais ' y o u it k n o w ' for Toi,

tu le sais ' Y o u , y o u i t k n o w ' ( = ' Y O U k n o w i t ' ) . They also occasionally

used lui, the tonic f o r m , f o l l o w i n g the verb, i n lieu o f the clitic object, le, preceding i t , e.g. *Deuxfois,

on avait lui ' t w o times, one had i t ' pointing to a

letter T , for Deux fois, on l'avait ' t w o times, one i t had' ( = ' t w i c e we had i t ' ) or, under similar circumstances, moi for me, e.g. * L à , je vas faire moi 'there, I go:2/3SG to do m e ' for Là, je vais me dessiner

'there, I g o : l S G me to:draw

( = 'there, I ' m going to draw m y s e l f ) . They also used à elle 'to her' after the verb instead o f lui before i t , e.g. *Je dis à elle T say to her' for Je lui dis T to:her say' ( = T say to h e r ' ) . A g a i n , such forms do appear i n adult colloquial speech too. Lastly, children make mistakes when they combine clitic object pronouns w i t h imperative verbs. Guillaume (1927b) reported errors like *Cache-le

pas

'hide-it not' for (Ne) le cache pas ' ( N o t ) it hide not' ( = ' D o n ' t hide i t ' ) , presum­ ably on the model o f the positive imperative plus direct object, Cache-le

'hide-

i t ' , and *Fais-le pas tomber 'make-it not to:fall' for (Ne) le fais pas

tomber

7.

T h e A c q u i s i t i o n of Romance

715

'(Not) i t make not to:fall' ( = ' d o n ' t make it f a l l ' ) , both from 3-year-olds. Unfor­ tunately, there appear to be no data available on such constructions i n other Romance languages, where the rules for ordering clitics generally differ, at least for some constructions, from those for French. To summarize, the errors children make i n pronoun placement appear to reflect generalizations made from more widely-used patterns. The order I O - D O , for instance, holds for all first- and second-person indirect and direct object combinations. Only w i t h third-person clitics is the order reversed. A n d children make order errors i n just those constructions, using I O - D O instead o f D O - I O . They could, o f course, avoid those constructions by using full noun phrases instead o f one or both clitic pronouns. Double clitic constructions w i t h first- and second-person indirect objects, me and te (e.g. me le, me la, me les), however, are quite common i n 4-year-old speech. A n d regularization errors, assimilating third-person indirect object constructions to the first- and second-person pattern, also appear, but how frequently i n spontaneous speech has not been w e l l docu­ mented. The other order errors they make involve use o f tonic pronouns instead of clitics for direct objects, w i t h the w o r d order that w o u l d have been used w i t h a full noun phrase: S V O instead o f S O V . A n d , i n imperative constructions, they retain the V O order o f affirmatives w i t h a clitic object for negatives as w e l l , instead o f shifting to an O V order by placing the pronoun before the verb. Clitic pronouns affect w o r d order i n all the Romance languages, but the changes i n order as one goes from a noun phrase object, say, to a clitic pronoun vary w i t h different constructions i n the different languages. What children appear to rely on i n the early stages is the predominant order, so their errors consist o f relying on the orders used w i t h full noun phrases, and on the dominant orders used w i t h clitic combinations. 9

4.6. Complex Sentences. Children begin to construct their first complex sentences around age 2 to 2Vi\ they use simple juxtaposition to express cause, condition, purpose, and sequence, omitting any conjunction that might specify the intended relation more precisely (Guillaume, 1927a; see also Sechehaye, 1950). The first conjunctions to appear are coordinative, usually et 'and' or puis 'then', for l i n k i n g t w o propositions. These first emerge sometime around 2;0. The first subordinating conjunctions appear soon after, w i t h parce que 'because' the earliest, followed by si ' i f and quand ' w h e n ' . This pattern o f emergence i n production is very similar to that for English (Clark, 1970, 1973a) as w e l l as for other languages (Clancy et a l . , 1976). Another coordinating conjunction that appears around 3;0 is alors 'so, then'. Temporal conjunctions like avant que

Adult speakers of Canadian French, according to Connors et al. (1981), do apparently avoid double clitic constructions with third-person forms. 9

716

Clark

'before' and adversatives like bien que 'although' emerge rather later, usually appearing around age 4 (cf. Chambaz et a l . , 1975; G r é g o i r e , 1947). Once children begin to use coordinating and subordinating conjunctions, as well as finite complement clauses, they make a variety o f errors. Few o f these, though, have been carefully documented from point o f onset to point o f adultlike acquisition. I n relative clauses, for instance, one finds some overuse i n production o f qui ' w h o : S U B J ' to introduce both subject and object relatives; only a bit later does que ' w h o : O B J ' become established for object relatives. One reason for an initial overuse o f qui might be the many other functions que has, including its earlier general-purpose function o f marking all subordinate clauses. It serves to introduce all finite verb complements, e.g. vouloir que 'to want t o ' , dire que 'to say that'. I t forms part o f many conjunctions, e.g. avant que 'before', pour que 'so that', afin que ' i n order t o ' , and serves as a reduced form for all subordinating conjunctions on the second and subsequent uses w i t h i n the same utterance, e.g. Pendant que Pierre est sorti et que Marie n'était pas encore là, . . . ' W h i l e Pierre was out and [while] Marie hadn't yet arrived, . . . ' . I t also serves as the l i n k i n g particle i n comparative clauses, e.g. Elle est plus grande que lui 'She is bigger than h i m ' ; and it introduces exclamations, e.g. Qu'il fait beau aujourd'hui! ' W h a t a fine day it is today!'. B y choosing a single form, qui, to mark all their relative clauses, children at this stage could be deliberately avoiding one o f the multiple functions o f que until they have sorted out more o f the system. Children also confuse certain conjunctions w i t h each other, both i n production and i n comprehension. For instance, Ferreiro (1971) reported that when she asked children to imitate and then act out temporal sequence instructions i n a comprehension task, they often substituted the conjunction avant que 'before' for après que 'after', or the reverse. They made similar errors i n their spontaneous speech and i n answering questions about sequence (see also Simon et a l . , 1972-73). M a n y o f these errors appear to be semantic i n origin. Children appear to confuse the meanings o f closely allied conjunctions. These confusions i n turn may result i n errors o f syntactic form since some conjunctions (e.g. avant que, pour que, 'before, i n order t o ' ) require the subjunctive while others take the indicative. 10

M a n y conjunctions appear to be treated initially as i f they simply linked t w o clauses describing sequential events. For instance, Jakubowicz (1978) found that young French children, up to age 5 or so, seemed to have no sense o f probability or uncertainty associated w i t h their uses o f si ' i f . Rather, they simply used i t for the sequential relation i n structures o f the form ' i f p, q'. But presented w i t h utterances o f the form ' i f p, q, but i f not p, not q', they could be made aware o f

I n adult speech, avant que 'before' is followed by the subjunctive (or ought to be, according to normative grammar), while apres que 'after' is followed by the indicative. Currently, many speakers appear to be confused on this point, and often use the wrong mood for one or both conjunctions. 10

7.

The A c q u i s i t i o n of Romance

717

the uncertainty associated w i t h p. Older children seemed to use cues based on the verb forms (imperfect and conditional), and then the conjunction itself i n deciding p was uncertain. I n production, by contrast, even the youngest children (aged 3) produced conditional utterances (including counterfactuals), provided the uncertainty about the events to be described was made explicit. Where uncertainty had instead to be inferred, younger children (up to 5) used si apparently to express some k i n d o f regularity holding between the t w o events, p and q. Older children (up to age 11) used si only for hypothetical situations or ones where there was some uncertainty (see P i é r a u t - L e Bonniec, 1980a, 1980b). Children acquiring French and Spanish also make the same kinds o f mistakes as English-speaking children when they are asked to identify the subject o f complement clauses f o l l o w i n g such verbs as 'promise'. They treat the nearest noun phrase before the complement verb as its subject (Chomsky, 1969). Gold¬ b l u m (1972), for French, and E c h e v e r r í a (1978), for Spanish, found that children under 7 or 8 consistently treated promettre and prometer respectively as i f they belonged to the same class as verbs o f saying or ordering (e.g. French dire 'to say, t e l l ' , commander 'to order'; Spanish pedir 'to ask,' ordenar 'to order'). E c h e v e r r í a (1978) also looked at Spanish-speaking children's comprehension o f preguntar 'to ask' (used only for information questions) and contar 'to t e l l ' , and found that children understood how to interpret the complements o f preguntar before those o f contar, a finding apparently opposite to the relative ordering o f tell before ask i n English (see also Edelsky & M u i ñ a , 1977). Nonetheless, the patterns o f confusions—here, treating contar as i f i t meant preguntar—appear similar. I n a further study o f children's ability to interpret and then produce complement constructions w i t h the verbs dire 'to say, t e l l ' and vouloir 'to want' i n French, Streri (1979a) asked children between 3 and 6 to report on what certain puppets had said or wanted done. (The tasks were complicated by the children's having to switch point o f view as they took the roles o f or reported on different participant puppets.) The younger children tended to avoid finite complements w i t h vouloir and opted for commentaries like (3; 11) La petite fule, si elle veut pas manager, elle va pas grandir 'The little g i r l , i f she doesn't eat, she w o n ' t grow b i g ' , rather than taking the role o f the mother puppet directly: Je veux qu'elle manage T want her to eat', or indirectly: Elle veut qu' elle (sa filie) mange 'She wants her (her daughter) to eat'. Streri found that the younger children (3and 4-year-olds) d i d much better when asked what particular puppets 'said' than asked what they 'wanted'. Dire was consistently better understood than vouloir. So i t was easier for the younger children, after watching a puppet scene, to report that La maman dit (raconte) que sa filie dessine 'The mother says her daughter is drawing' than La maman veut que sa fule dessine 'The mother wants her daughter to draw' (see also Streri 1979b, 1980). The errors children made, though, were often errors o f omission rather than commission. Complement constructions and subordinate clauses generally require more investigation.

718

Clark

Finally, there has been some research on children's understanding o f the presuppositions carried by particular conjunctions such as mais 'but' and quantifiers or adverbs like seul, aussi, and même ' o n l y , too, same' (e.g. K a i l , 1979; K a i l & Weissenborn, 1980). The conjunction mais, for example, appears very early, soon after et 'and' and puis 'then' i n children's spontaneous speech (e.g. G r é g o i r e , 1937, 1947; Fondet, 1979). However, K a i l and Weissenborn (1980) found that it was still treated simply as a coordinating device to l i n k t w o clauses by 7-year-olds. A n d it wasn't until age 9 that they seemed to take full account o f the contrastive meaning o f mais i n a sentence completion task. Children may begin to use such forms, then, long before they appreciate the precise presuppositions they carry w i t h t h e m — i n the case o f mais, that there is a contrast between the t w o parts o f the statement being linked by this conjunction. Errors i n the use o f terms like mais, o f course, are notoriously hard to detect in spontaneous speech. Few, i f any, o f the diaries available provide adequate information about the context o f each utterance or about the surrounding discourse. There are very likely other errors i n children's production and comprehension o f conjunctions, for example, i n the more complex conjunctions like bien que 'although', w h i c h contradicts prior suppositions. (This conjunction is one that requires the subjunctive mood.) The general patterns o f development i n children's acquisition o f coordinate and subordinate clause constructions are still too sketchy for one to be able to draw general conclusions. Too many o f the details remain to be filled i n . Although there has been some research on the acquisition o f coordinating and subordinating conjunctions i n Italian and Spanish, again there seems to have been little analysis o f the precise errors children make. I n Italian, the first clause combinations are by juxtaposition alone, w i t h no conjunctions, much as i n French (e.g. Frontali, 1943-44; Parisi & A n t i n u c c i , 1974). The first coordinating conjunction to appear is e ' a n d ' , followed shortly by ma ' b u t ' . The earliest complements, as i n French, are infinitival i n form, after such verbs zsfare 'to d o ' , volere 'to w a n t ' , or devere 'to need'. The first subordinating conjunctions are perchè 'because' and quando ' w h e n ' , followed later by affinchè ' i n order t o ' and se ' i f (Parisi & A n t i n u c c i , 1974; Flores d'Arcais, 1978a, 1978b). These are followed by other temporal conjunctions like prima che 'before' and dopo che 'after' (Clancy, Jacobsen, & Silva, 1976). I n comprehension tasks, young c h i l dren act initially as i f order o f mention rather than the conjunction prima che or dopo che determines interpretation. So the event mentioned first is treated as the first i n sequence (Flores d'Arcais, 1978a), just it is by children acquiring other languages (Clark, 1971; Ferreiro, 1971). Only later do children w o r k out the contrasting meanings o f the conjunctions themselves. The same general patterns o f development have been noted for Spanish: the first coordinating conjunctions are y ' a n d ' , from age 2 o n , and pero 'but' from age 3 or 4. The first subordinating conjunction is que 'that' for introducing complements, as i n (3;0) Mamá dice que vamos (a la) calle ' M u m m y says we're

7.

T h e A c q u i s i t i o n of Romance

719

going out' or (3;0) Tiita no kele (=quiere) que pongo ese vestido ' A u n t y doesn't want me to put on these clothes'. Next, between age 3 and 5, according to G i l i Gaya (1972e), children begin to use porque 'because', para que ' i n order t o ' , and si ' i f . These examples appear to be typical: (5;0) Si me lo dice le pego ' I f he says that to me, I ' l l hit h i m ' , and (5;0) El perro acude pa(ra) que le echen de comer 'The dog's waiting for them to throw h i m some f o o d ' . However, their early spontaneous uses are often incorrect, and the conjunctions sometimes ap­ pear to be used simply to l i n k t w o or more clauses. This often results i n syntactic errors, for example, w i t h the verb used i n the indicative instead o f the sub­ junctive. The meanings o f some conjunctions may also be confused w i t h each other. A g a i n , this may result i n syntactic as w e l l as semantic errors. A n d still other conjunctions, like aunque 'although', are only rarely used even by 8-yearolds. Few studies have looked at children's comprehension o f conjunctions, but Gal van (1980) found that children relied on order o f mention w i t h the temporal forms antes de 'before' and después de 'after', just as children acquiring other languages do. He also found that children seemed to understand antes de rather earlier than después de. I n summary, although the various accounts mention that children make errors in using different conjunctions, relatively few examples o f actual errors have been reported, so it is difficult to tell whether there are consistent trends i n common, either across children or across the different Romance languages. One of the few domains i n w h i c h there are some production and comprehension data appears to be that o f the temporal conjunctions avant que and aprés que i n French, prima che and dopo che i n Italian, and antes de and después de i n Spanish. I n production, the errors appear typically to consist o f overuse o f one of the conjunctions (usually 'before') and some confusions; i n comprehension, the conjunctions are initially ignored—until about age 4 — i n favor o f an order o f mention strategy where children treat the first event mentioned as the first that occurred. Later, they acquire the meanings o f 'before' and 'after', and presum­ ably begin to use both conjunctions appropriately. 11

4.7. Other Typical Errors. Several other systems that typically elicit errors from children during the course o f acquisition are discussed i n more detail later in the review. For example, children have considerable difficulty w o r k i n g out the different subsystems used for marking possession, where possessive pronouns and adjectives agree i n person w i t h the possessor and i n number and gender w i t h the object possessed. Getting the person right i n the pronominal system requires that children be able to shift from first to second person, for instance, w i t h a

T h e Spanish data collected by Galvan all pertain to the infinitival construction that follows antes de and después de. There are no data available on children's comprehension or production of the conjunctions antes de que or después de que, which are followed by the subjunctive, or, with reference to events in the past, by a simple past participle. n

720

Clark

change o f speaker. A n d getting the right possessive construction requires that children take account o f w h i c h combinations o f constructions are permissible, for emphasis, say, and w h i c h are not. Children acquiring French also have some difficulty learning where to use the definite and indefinite articles. L i k e children acquiring other languages, they tend to overuse the definite. I t is not clear, therefore, that these errors are peculiar to children acquiring Romance languages. I t seems more likely that they result from children's relative lack o f skill i n keeping track o f what they and their addressees mutually k n o w about any particular topic. Errors i n the uses o f determiners i n French, whether possessives or articles, are discussed i n detail i n Section 7.

5. E r r o r - F r e e A c q u i s i t i o n There are few subsystems i n French and the other Romance languages that children acquire w i t h no errors, but there are several that appear to be mastered early, w i t h only a brief period o f erroneous use prior to mastery. Number as indicated by the singular/plural contrast, is one such subsystem: Nouns marked as singular or plural by means o f the article are produced w i t h relatively few errors i n French. One reason for this is probably that noun forms themselves do not change i n pronunciation w i t h a change i n number. I n Spanish, where number is marked both by the article and by an ending added to the noun, there appear to be more errors, particularly w i t h the forms o f nouns ending i n consonants. French-speaking children also appear to make relatively few aspectual errors. Initially, they use the compound past (at first i n the form o f the past participle alone) for results and changes o f state. This form o f the verb later comes to contrast w i t h the imperfect, used to describe " b a c k g r o u n d " events and activities w i t h limited duration (e.g. Krafft & Piaget, 1925; Bronckart, 1976). Despite such relatively early acquisitions, there appear to be few systems i n French, or elsewhere i n Romance, that are acquired entirely without error. One point to consider: Most o f the errors documented i n studies o f language acquisition are errors o f F O R M . But children also make numerous errors o f reference or C O N T E N T . For example, they may use a plural noun to pick out a singular object. But the latter errors have rarely been documented thoroughly enough for description. The emphasis i n most studies has been on the acquisition of the form rather than on accurate use o f each f o r m , given its content. Content errors are very difficult to document post facto: Transcripts without meticulous contextual notes can yield no information about such errors, although research on word-meaning acquisition suggests they are fairly prevalent, especially i n the first t w o or three years o f language acquisition. Finally, errors made during acquisition may be errors o f omission or o f commission. The former again are difficult to assess. Avoidance o f particular constructions or o f lexical items could be due to the cognitive or linguistic

7.

The A c q u i s i t i o n of Romance

721

complexity o f the distinctions being encoded. Or they could be due to the caution of the learner. Analyses o f acquisition errors have depended almost entirely on errors o f commission, i n particular on those prevalent i n the linguistic sub­ systems acquired earlier rather than later i n the acquisition process. D o children get more careful as they learn more about a particular linguistic system as a whole? Or do they simply get better at making appropriate inferences about how each new subsystem is likely to work? Analysis o f early versus late-appearing errors might yield some insight into the relative amounts o f error at different stages o f acquisition, as w e l l as any qualitative differences i n the kinds o f errors made early as opposed to late (e.g Bowerman, 1982c; K a r m i l o f f - S m i t h , 1979).

6. T i m i n g o f A c q u i s i t i o n In any language, children acquire some words and constructions before others. They also acquire some constructions without apparent error, and others w i t h many errors on the way. 6 . 1 . Precocious Acquisitions. There are no strikingly precocious acquisi­ tions i n children acquiring Romance languages, but among early acquisitions i n production are gender, number, and person. Children master phonological cues to w o r d gender i n French, for instance, by age 3;0 and w i l l assign gender, marked w i t h the appropriate articles, even to nonsense words, as long as the word forms conform to the patterns found i n the lexicon as a whole. For exam­ ple, Karmiloff-Smith (1979) found that 3-year-olds treated nonsense words w i t h typically masculine endings such as -eau or -on as masculine, and ones w i t h typically feminine endings such as -elle or -aise as feminine. But when Kar­ miloff-Smith used indefinite articles incompatible w i t h the w o r d forms, e.g. une bicron, unforsienne, children under 5 tended to change the article to match the form, e.g. le bicron, la forsienne. Older children, though, w o u l d pay more attention to the gender given by the indefinite article and w o u l d reply w i t h the equivalent definite article, as i n la bicron and le forsienne, despite the lack o f congruence between article and w o r d shape. Number is mastered early, first i n nouns and articles, and then i n pronouns. I t seems to appear slightly later i n verbs, but this may be because both number and person are marked w i t h the same device w i t h i n the verb system. Children rarely make errors o f number i n nouns. I n French, o f course, number for nouns is typically marked only i n the article, whereas the other Romance languages mark number on the noun as w e l l . This makes for a contrast between a preposed marker for number i n French versus a postposed system o f marking i n the other Romance languages where number is indicated by the addition o f inflections for plural to both articles and nouns. Agreement for number also differs across the Romance languages i n where and how systematically it has to be marked. For example, Spanish marks number

722

Clark

agreement very consistently—in the article, the noun, the adjective, and the verb. French essentially marks number i n the article and to some extent i n the verb. Most adjectives, like most nouns, do not differ i n pronunciation according to number, and so rarely mark number overtly. The extent to which agreement is marked could affect how fast children master number as a whole w i t h i n the language being acquired. French-speaking children make some errors o f number agreement, but these are not very common. They are detectable only w i t h second or third conjugation verbs where the forms for singular and plural third person differ, e.g. *Les chats, il vient for Les chats, ils viennent ( = ' t h e cats, he's coming/they're c o m i n g ' ) . Most agreement errors seem to involve a singular verb used w i t h a plural subject, seldom the reverse (e.g. G r é g o i r e , 1947). Number errors are also rare in Spanish although nouns as w e l l as articles are marked for number. I n fact, the overregularizations Spanish-speaking children make for gender—adding a final -o or -a to consonant-final nouns (Montes Giraldo, 1976)—make plurals easier to form. Children simply have to add a final -s to all nouns and to the feminine article la (for las). They do have to learn the form los, though, as the plural o f the masculine singular el. Number i n the verb is also acquired relatively early, without very many errors. I n French, o f course, for most verbs, the three singular person forms are pronounced alike (and are usually identical w i t h the third person plural form too). Children's first verb forms appear to be the singular and plural o f the imperative, e.g. donne, donnez. However, the singular imperative coincides w i t h the three singular forms o f the indicative, w h i l e the plural one coincides w i t h the infinitive i n pronunciation, for most verbs. The latter, then, can only be identified as a plural form on the basis o f second and third conjugation or irregular verbs where the infinitive and imperative forms differ, as i n second conjugation finir 'to finish' and finissez ' f i n i s h ! ' , or third conjugation prendre 'to take' and prenez 'take ( i t ) ! ' . Next to emerge seems to be the second or third person o f the indicative, e.g. the third person i n // mange 'he's eating'. A l t h o u g h the first three persons are pronounced alike for most French verbs, the second and third person forms (usually identical i n pronounciation) are distinct from the first person i n irregular verbs like aller 'to g o ' , être 'to be', and avoir 'to have'. Evidence from irregular verbs like these suggests that this f o r m is the next to emerge (e.g. Fondet, 1979; G r é g o i r e , 1947; L e r o y , 1975). The pattern o f emerging number contrasts for French is similar to those attested for Italian, Portuguese, and Spanish, where a similar sequence o f imperative forms followed by a third person indicative marks the first contrast among forms o f the verb (e.g. Della Valle, 1931, Frontali, 1943-44; S i m ò e s & StoelGammon, 1978; G i l i Gaya, 1972c). T h i r d person verb forms i n French may or may not appear w i t h an overt subject. The earliest i n verbs i n French often appear alone or w i t h a direct object when children talk about their o w n activities (e.g. B l o c h , 1921, 1923), and only occasionally w i t h a subject noun like bébé 'baby' or their o w n names. I f the actor

7.

T h e A c q u i s i t i o n o f Romance

723

is someone else, however, children appear more l i k e l y to supply a subject, typically a noun w i t h no article at all (e.g. B l o c h , 1924; Cohen, 1969; Guillaume, 1927a; L i g h t b o w n , 1977). A similar pattern appears to hold for the other Romance languages at the equivalent stages o f development (e.g. for Italian, Antinucci & Parisi, 1973; Bates, 1976; Parisi & A n t i n u c c i , 1974; for Portuguese, Scliar-Cabral, 1976; de Lemos, 1975; and for Spanish, Eisenberg, 1982; Rodriguez B r o w n , 1975). A l o n g w i t h these contrasts among verb forms come the first uses o f pronouns. F u l l mastery o f different pronoun systems, though, may take some time. Another fairly early acquisition is markers for locative relations. Children start to produce some prepositions, typically the terms for relations such as ' i n ' , ' o n ' , and 'under', from age 2 on. However, the forms produced at this age are often used erroneously, w i t h mistakes i n both form and content. Mastering all the contrasts among locatives typically takes several years (e.g., Johnston & Slobin, 1979; L e Rouzo, 1977). Children begin to modulate the forms o f words by adding different deriva­ tional suffixes from quite an early age. They use diminutive and augmentative endings on nouns (and elsewhere) from around 2;0 to 2;6 i n Spanish and Italian, and begin to use other derivational suffixes to modify or change the meaning and grammatical category o f w o r d forms at about the same age. A g a i n , though, mastering an adult-like repertoire o f possible w o r d forms for coining new lexical items takes several years. (Typical developments i n this domain are discussed i n detail i n Section 8.) Temporal distinctions marked by the different tenses also begin to emerge early. Children begin to use tense i n French from age 2 o n , and typically contrast the compound past (probably to mark results or end-states) w i t h the present. From about age 3 or so o n , they also contrast the compound past w i t h the imperfect, using the later for background and for ongoing e v e n t s . They mark future time w i t h aller + I N F I N I T I V E 'to be going t o ' , from about age 3;0 or so, and only later begin to use future tense inflections. This pattern for future marking, w i t h a periphrastic future form emerging before the inflected one, appears common to all the Romance languages. 12

The overall picture for early acquisitions is one o f systems that begin to emerge i n production relatively early, from age 2 or 3 on. But although children start on many systems early, their mastery o f most evolves fairly gradually over the course o f several years. 6.2. Late Acquisitions. A m o n g notably late acquisitions i n French are the subjunctive mood i n verbs (acquired considerably later than the imperative and indicative), certain kinds o f relative clause constructions, certain complement constructions, and subordinate clauses that require the subjunctive mood. The I n Spanish and Italian, children initially use past participial forms for results, but the next tense form acquired is the simple past—a form no longer used in spoken French. 12

724

Clark

subjunctive, now used mainly i n present tense form only i n spoken French, appears to take time to acquire both because o f the situations i n w h i c h i t is commonly used (after certain verbs and conjunctions as w e l l as for certain hypotheticals) and because o f its f o r m . Part o f its complexity is undoubtedly due to the difficulty children have i n discerning the reasons for its use (e.g. i n such arbitrary contexts as 'before' clauses but not 'after' ones), and part o f i t is probably due to the difficulty they may have i n detecting the presence o f the subjunctive as a distinct verb form. Since it coincides i n form w i t h the indicative in all first conjugation verbs, only second and third conjugation verbs provide overt evidence i n input that a non-indicative form is required i n certain construc­ tions and contexts. Remacle (1966) found that the first uses o f the subjunctive .tended to be produced i n a small set o f sentence frames that required such forms, e.g. after such expressions as //faut que ' i t is necessary that, must', je voudrais bien que T w o u l d l i k e ' , and after the conjunction pour que ' i n order t o ' . But even w i t h these favored frames, he noted that the 4- and 5-year-olds he followed used less than one-quarter (19%) o f their second and third conjugation verbs i n the subjunctive. ( A l l first conjugation verbs were necessarily omitted from this analysis.) A n d , i n a number o f other constructions that also required the subjunctive, children consistently used the indicative instead (see also Ferreiro, 1971; Streri, 1979a). The difficulty o f learning the subjunctive i n French, then, appears to be com­ pounded by the coincidence o f forms between the indicative and the subjunctive for all first conjugation verbs. Relative clauses provide children w i t h a number o f difficulties i n both produc­ tion and comprehension. I n production, for example, relative clauses that inter­ rupt the main clause are harder for children to imitate than those that don't interrupt it (e.g. K a i l , 1975a, 1975b; Deyts & Noizet, 1973). Indeed, the first relative clauses very young children produce spontaneously are nearly always i n final position, attached to the object nouns o f main clauses. K a i l (1975a) also found that relative clauses introduced by que (object relatives) were harder to imitate than those introduced by qui (subject relatives). This suggests that by age 6 (the youngest children studied by K a i l ) children have identified qui as a relative marker, but since que has so many different functions, they are less sure o f its status i n relative clauses. Moreover, w i t h delayed imitation, slightly older c h i l ­ dren often substituted qui for que i n their imitations ( K a i l , 1975b), as i f qui offered a more consistent cue to the relative nature o f the subordinate clause being imitated. (Six-year-olds—the younger children—often omitted the relativizer altogether i n this t a s k . ) Finally, one further factor that appears to make relative clauses hard to produce is noncanonical w o r d order. Object relatives 13

These data, of course, are in contrast to those from 2- and 3-year-olds who tend to use only que in early relative clauses. In these, though, que is probably being used to mark any linkage between propositions and should not be considered a relative marker at all (see Bouvier & Platone, 1976). 13

7.

The A c q u i s i t i o n o f Romance

725

w i t h O S V order are harder for children to imitate than those w i t h S V O order. Late production errors, then, are nearly always errors involving oblique cases. The comprehension data make this picture rather more complicated. For example, although subject relatives are harder to produce than object ones, they appear to be easier to understand than object relatives, especially when the latter have noncanonical order ( S é g u i & L é v e i l l é , 1977; Sheldon, 1977). Otherwise, even the youngest children tested (4-year-olds) appeared to understand all relative clause constructions pretty accurately (Ségui & L é v e i l l é , 1977). I n other words, interruptions o f the main clause d i d not seem to make for greater processing difficulty i n comprehension. The errors 4-year-olds make w i t h object relative clauses are subject to t w o interpretations. The first is that they rely on a canonical SVO strategy i n interpreting all relative clauses because they have not yet discovered that a local cue (the f o r m o f the relativizer) can indicate case and may, as a result, require a change from canonical ( S V O ) order to O S V or O V S . A second interpretation is that they consistently treat que as i f i t was qui, so that they treat object relatives as i f they were subject relatives. W i t h noncanonical object relatives ( O V S ) , they interpret the que (marking the object) as the subject o f the relative clause and the second noun phrase (the actual subject) as its object. I n the noncanonical relatives, then, both interpretations make the same prediction: children w i l l impose S V O order on O V S sequences. The difficulties 4-year-olds have i n understanding the cues to oblique case (use o f que rather than qui) i n object relatives are paralleled by the difficulties much older children have i n producing this type o f relative clause. Local cues to case such as those given by the alternations i n the relativizer are rare i n French, as is O V S order. Combined, they cause children difficulty both i n understanding and i n producing certain kinds o f relative clause. These studies have documented errors children make when called on to i m i tate or understand isolated relative clauses. But there has been little research on the functions o f relative clauses or how children use them spontaneously. For example, what forms o f relative clause do young children use to distinguish one object from another on the basis o f history, characteristics, or spatial location? Some preliminary elicitation studies by Bouvier and Platone (1976) suggest that children can produce some relative clauses for such purposes by age 4 , but even at age 6 produce many erroneous forms. One child age 5;6, for instance, used relatives like la voiture que le monsieur arrête 'the car that the man is stopping' alongside more problematic utterances like * la voiture que le monsieur met une roue 'the car that the man is putting a wheel' for la voiture à laquelle le monsieur met une roue 'the car the man is putting a wheel o n ' ; from a 5-year-old using où 'where' as a relative marker, both the appropriate le camion où on met d'Vessence 'the truck someone's putting some gas i n ' , and the problematic form */a voiture où Vgendarme l'arrête 'the car where the-policeman it-stops' for la voiture que le gendarme arrête 'the car that the policeman stops'. These elicitation data, where the relative clauses appear w i t h a determiner function, suggest

726

Clark

that even 6-year-olds still have a long way to go i n learning how to use relative clauses i n French. Relative clauses are also a late acquisition i n Spanish. Ferreiro, OtheninGirard, Chipman, and Sinclair (1976) found that 10-year-olds still made errors i n the comprehension o f object relatives. I n production, relative clauses begin to appear (as they do w i t h French-speaking children) at around age 3. The first ones used are subject relatives. Other relative clause types emerge over the course o f the next three or four years ( G i l i Gaya, 1972e). However, there has been little detailed study o f the precise forms produced, and no analysis o f the uses to w h i c h young children first put them. One note o f caution should be added here: M a n y of the comprehension studies have focused on relative clauses that may be both bizarre and difficult out o f a normal conversational context. As a result, there are extensive discrepancies from one report to another. For example, although both French and Spanish children begin to produce relative clauses spontaneously from age 3 or so onwards, they still have difficulty understanding certain forms as late as age 10. The formal difficulties that make for late acquisition appear to be oblique case marking and noncanonical w o r d orders i n some relative clause types. The functional difficulties that contribute to late acquisition appear to involve children's discovery o f when and where to use relative clauses appropri­ ately. However, aside from Bouvier and Platone (1976), no one has made any serious study o f the functions involved. I n summary, I have outlined some o f the difficulties children have w i t h t w o late acquisitions, the subjunctive mood and relative clause constructions. M o r e detailed investigation o f other constructions such as object clitic pronouns or the conditional mood w i l l add to what is k n o w n about late acquisitions and help uncover i n more detail the sources o f difficulty children face i n acquiring French as their first language.

7. O r g a n i z a t i o n a n d C h a n g e D u r i n g D e v e l o p m e n t When children first acquire new words, they use them at first i n only a few contexts. For example, tall may be applied only to trees (Carey, 1978). These limited uses are often correct. However, as children find out more about the meanings o f related words and extend each w o r d i n their repertoire, they begin to make errors where they made none before. This pattern o f acquisition—correct but limited use followed by errors i n use—has been documented, for example, by Bowerman (1978, 1982b, 1982c) i n her account o f various late-occurring errors i n English-speaking children's uses o f verbs. This pattern o f development appears to be a fairly general one since children allow many different forms to go unanalyzed for some time (e.g. V i h m a n , 1982; Johnson, 1980, 1981). There are at least t w o common types o f late-occurring errors i n children's speech. The first is errors that appear to result from children's attempts to

7.

The A c q u i s i t i o n o f Romance

727

maintain one-to-one pairings o f meanings and forms w i t h i n their language. The second is errors that result from their overmarking o f certain (possibly n e w l y discovered) semantic distinctions. The first k i n d o f error may arise either w i t h the construction o f additional— and usually ungrammatical—forms i n order to preserve a one-meaning-to-oneform approach, e.g. la même vache 'the same c o w ' for 'same one' versus *la même de vache 'the same o f c o w ' — f o u n d i n colloquial French also—for 'same k i n d ' ; or from a specialization o f different grammatical forms to convey newly acquired contrasts i n meaning, e.g. use o f the contracted de + A R T I C L E for partitive meaning, as i n J'ai du pain ' I ' v e got some bread' and *Y'a da neige, w i t h da i n lieu o f de la 'there's some snow', i n contrast to uncontracted de + A R T I C L E for possession, as i n *le chapeau de le monsieur 'the man's hat' w i t h de le instead o f du, alongside le soulier de la dame 'the woman's s h o e ' .

14

The second k i n d o f error that one finds f o l l o w i n g previously errorless use is overmarking. Children typically overmark by indicating the same semantic distinction i n more than one way on a single lexical item, as i n English *unthaw, or by marking it at several different points i n the utterance, as i n *mon mien de chapeau à moi ' M Y M I N E o f hat to M E ' ' M Y hat' or La fille a poussé puis aussi le garçon

a repoussé

then A L S O the boy RE-pushed A G A I N the S A M E

Why

un chien et

encore le même chien 'the g i r l pushed a dog and dog'.

do children begin to make such errors after earlier, apparently correct,

usage? One possible answer is that such errors are themselves indicators o f children's continuing organization and reorganization w i t h i n various subsystems of their language. For example, i t may w e l l be the realization that one f o r m is carrying t w o or even more meanings that leads children to construct additional forms. This i n turn allows children both to hold on to a one-meaning-to-one-form system and to consolidate their o w n grasp o f the different meanings they are discovering—both necessary preludes to their adopting the pertinent adult system

w i t h its more complex pairings o f meanings and forms. Overmarking, then, w o u l d be another symptom o f adult-like organization i n

the process o f emerging. Children draw on all their resources to make a particular meaning salient by indicating i t i n more than one way i n particular words, phrases, or utterances. I n overmarking, children could also be trying to avoid confusions among meanings they have just begun to realize are related to each other. For example, Bowerman (1978) documented the onset o f errors i n uses o f the English verbs put, give, make, and let, as w e l l as o f put, take, and bring, by her daughters. The children had previously used these verbs quite correctly, but

Note that contracted forms are required only when de precedes either the masculine singular article, le, or the plural article, les, for the forms du and des, respectively, whether partitive or possessive in meaning. French-speaking children as old as 5 still do not consistently contract de le to du, de les to des, a le to au, or a les to aux (LaBelle, 1976; Bautier-Castaing, 1977)). 14

728

Clark

perhaps i n a quasi-formulaic way w i t h restricted meanings. As they began to use these verbs i n a wider range o f contexts, they started to make errors o f substitu­ tion and to use inappropriate combinations o f noun phrases w i t h each verb. Their errors suggested they had begun to recognize how the different meanings i n question were related to each other—and this led to momentary confusions and errors. These later-appearing errors are particularly important for the information they offer about w h i c h meanings children are trying to keep distinct, and how they organize various subsystems i n the language being acquired. I w i l l illustrate this phenomenon i n t w o domains where children make later-occurring errors: w i t h articles and w i t h possessive constructions. 7 . 1 . Definite and Indefinite Articles. Observational studies o f French (and the other Romance languages) suggest that children begin to use articles around age 2;0 to 2;6 (e.g. G r é g o i r e , 1937, 1947; F r a n ç o i s et a l . , 1978). However, the earliest uses are often indeterminate phonetically between a definite and indefi­ nite form, and articles i n general are often omitted up to age 3. Several researchers have looked experimentally at children's article use i n order to establish more precisely what they k n o w about specific and nonspecific reference. For example, Bresson, Bouvier, Dannequin, Depreux, Hardy, and Platone (1970) studied 4- and 5-year-olds' ability to use the definite and indefi­ nite articles i n a limited set o f contexts containing one or more instances o f objects from the same class. Their children all used the indefinite appropriately in naming things, e.g. C'est un mouton ' I t ' s a sheep'. But when the indefinite article was called for i n referring to one or more o f a set o f like objects, they consistently used the definite article instead. For example, shown a number o f sheep, followed by the experimenter removing some o f them and asking Qui est parti? ' w h o went away?', 4-year-olds w o u l d reply w i t h utterances like *Les moutons sont partis, les moutons sont restés 'the sheep went away, the sheep stayed' i n lieu o f Des moutons sont partis, des moutons sont restés 'some sheep went away, some sheep stayed'. Four-year-olds used the indefinite un ' a : M A S C : S G ' and des 'some:PL' only 50% o f the time i n such contexts, and 5year-olds d i d even worse, using the indefinite only 3 1 % o f the time (see also Bresson, 1974; Bouvier & Platone, 1976). French children, then, overuse the definite article, at least up to age 6. This finding is compatible w i t h Warden's (1976) observations for English over a rather wider range o f elicitation contexts. He found that before age 5, children's referring expressions were mostly definite. Between 5 and 9, there was a general decrease i n inappropriate uses o f the definite and some inconsisten­ cy o f use overall. Warden concluded that it was only from age 9 that children showed full mastery o f the indefinite article for nonspecific references like those needed i n answering the question Qui est parti?. Overuse o f the definite article

7.

The A c q u i s i t i o n of Romance

729

appears to be the result o f children's failures to realize that their addressees do not necessarily k n o w all that they, as speakers, k n o w . Learning how to assess what is mutual knowledge and what is not, and, from that, being able to decide on the appropriate article to use is very complex (e.g. Clark & Marshall, 1981). Articles also seem to cause children difficulty because o f their multiple meanings that vary w i t h context—meanings that children take time to acquire. I n a series o f ingenious studies o f children's production and comprehension o f articles and other forms o f determiner, Karmiloff-Smith (1977, 1978, 1979) showed that children's initial usage depended heavily on the identification o f a single meaning or function for each term. For example, even w i t h the addition o f même 'same' and autre 'other' to the definite le and indefinite un, 3- and 4-year-olds, asked to act out a series like the following: 'The little girl pushes a duck and then the boy pushes the same d u c k ' , consistently treated an expression like le même canard 'the same duck' as i f i t picked out another instance o f the same K I N D , i.e., another duck, rather than the duck already mentioned and acted upon (Karmiloff-Smith, 1977). W i t h age, children showed a change i n interpretation from 'same k i n d ' to 'same one'. W i t h expressions like un autre canard 'another duck', most o f the children, even 3-year-olds, picked out another duck. However, i f the ducks were identical, a few o f the younger children refused to choose, apparently because they took this expression also as meaning 'one o f another k i n d ' rather than 'another one'. The transition to adult-like comprehension was accompanied slightly later by a parallel progression i n their production o f articles w i t h même or autre. Their initial uses d i d not reveal obvious errors, but as the children began to switch from the interpretation o f 'same k i n d ' to 'same one', Karmiloff-Smith noted that children frequently created new (often ungrammatical) forms to carry the additional meanings they were trying to express. For example, 4- and 5-year-olds tried to introduce into French a distinction between 'same k i n d ' and 'same one'. For 'same k i n d ' , for example, upon noticing the toys used i n the experiment, they might comment, *Moi j'ai la même de vache ' I ' v e got the same o f c o w ' , instead o f relying on the form used by younger children and adults: J'ai la même 'I've got (one) the same'. These children reserved la même for 'same one' and created a new expression for 'same k i n d ' . In another experiment, Karmiloff-Smith noted children constructing new forms o f article to distinguish the nonspecific reference function o f the indefinite, j'ai une vache ' I ' v e got a c o w ' from its numeral function, e.g. *J'ai une de vache T have one o f c o w ' ( = r v e got one c o w ' ) . These " n e w " expressions disappear again when children realize that single expressions can have several different meanings (see K a r m i l o f f - S m i t h , 1979). The youngest children's production o f articles—alone and i n combination w i t h même and autre—showed few errors o f form; i n contrast, somewhat older children used some apparently correct forms but also created ungrammatical forms to carry some o f the adult

730

Clark

functions o f articles. Finally, from around age 7 or so, children appeared to accept the multiple functions o f articles and returned to the use o f a single form w i t h several meanings. It is important to note that while these data on children's production appear to show a U-shaped curve i n development, their initial uses o f the articles i n such contexts combined w i t h même and autre d i d not i n fact reflect totally adult-like usage (see also K a i l , 1978, 1979). But this is only apparent i n light o f the comprehension data where the youngest children's preference for le même X as meaning 'same k i n d ' over 'same one', for example, reveals convincingly that what children say and what they understand do not correspond to adult usage. There has been little or no research on the acquisition o f article systems i n the other Romance languages, aside from one observational study by Simonetti (1980) o f Portuguese. Simonetti examined one child's spontaneous uses o f the different articles and demonstratives from age 1;9 to 2;4 and found that the definite article appeared to be used mainly w i t h direct objects whereas the demonstrative appeared mainly w i t h subjects. This study, though, covers too early a stage for the kinds o f errors documented by Bresson or Karmiloff-Smith for French. Given the similarities o f the functions marked by articles across different Romance languages, though, it w o u l d be surprising i f children acquiring the other languages d i d not follow similar routes i n development.

7.2. Possessive Constructions. The earliest expressions o f possession i n French vary i n form from child to child. Guillaume (1927b) reported that one o f his children used moi ' m e ' on its o w n at 1 ;4 i n the sense o f à moi or pour moi ' t o , for m e ' , and very soon used juxtaposition for the possessions o f others, e.g. taté papa for café papa 'coffee daddy'. B y 1 ; 11, this child regularly used the analytic de moi ' o f me' and de toi ' o f y o u ' , e.g. la cuiller de papa et celle de moi 'the spoon o f daddy and that o f me' ( = 'daddy's spoon and m y spoon'). B y 1;9, the child also made sporadic use o f mon and ma ' m y . M A S C ' , ' m y : F E M ' , as w e l l as of le mien ' m i n e : M A S C ' . A n d by by 1 ; 11, he added the second person possessor forms: ton, ta, and le tien ' y o u r : M A S C ' , ' y o u r : F E M ' , ' y o u r s : M A S C \ G r é g o i r e (1947) reported that initial juxtapositions were followed by reliance on the prepositions à ' t o ' and de 'of, f r o m ' i n possessive constructions, e.g. (2;4,16) La conyé (=cuiller) à bébé 'the spoon to baby' ( = 'the baby's spoon'), (2;6,22) La ke:me (=crême) de moi 'the cream o f me' ( = ' m y cream'), (2;6,23) Regarde le café de moi 'look at the coffee o f me' ( = 'look at m y coffee'), and (2;6,25) Une tasse de moi 'a cup o f m e ' ( = ' m y cup, a cup o f m i n e ' ) . I n the last three examples, w i t h de moi, adults w o u l d normally use the adjectival forms: ma crème, mon café, ma tasse (or une de mes tasses). Prepositional constructions for possession appear to be more analytic than possessive adjectives like mon ' m y : M A S C : S G ' i n that they mark explicitly the relation between object-pos-

7.

The A c q u i s i t i o n o f Romance

731

sessed and possessor. The same preference for analytic over synthetic forms also appears early i n the acquisition o f other Romance languages. I n Spanish, as i n French, children again rely initially on juxtaposition alone and then begin to mark possession w i t h the preposition a ' t o ' w i t h the possessor, as i n Pato (=zapato) a Daddy 'shoe to daddy' from a 2-year-old (Mazeika, 1973). However, the preposition de is the one usually used i n such constructions by adults. Slightly older children may continue to use juxtaposition but use articles w i t h both the object-possessed and the possessor nouns, as i n (2;9) / liito {-el librito) la yo ' t h e : M A S C b o o k + D I M the:FEM (agrees w i t h sex o f speaker) F ( = ' m y b o o k ' ) , Popita (=sopita) ya {—la) yo 'soup the I ' ( = ' m y soup'), or (3;6) La cata (=casa) le mi 'the house the me' ( = ' m y house') (Montes Giraldo, 1976). Only later do children start to use possessive adjectives instead to indicate the possessor. Acquisition o f the latter i n Spanish may be complicated by the presence o f t w o possible forms: when the possessive precedes the noun designat­ ing the object possessed, a "short f o r m " is used, as i n mi libro ' m y b o o k ' , tu casa 'your house', etc.; but when i t follows, the construction takes the form A R T I C L E + N O U N + P O S S E S S I V E ' , as i n el libro mio 'the book m y ' , la casa tuya

'the house your'. These forms are slightly more emphatic and are probably sometimes better rendered i n English as that book of mine or that house of yours. Picking out the object-possessed and the possessor i n t w o separate noun phrases linked by a preposition like ' t o ' or ' o f , appears to be a common analytic stage i n the expression o f possession. I n Rumanian, possession is marked by use o f the genitive case, but i n collo­ quial adult speech, case marking is often replaced by an analytic construction w i t h the preposition la 'at, t o ' . I n children, the first expression o f possession, according to Slama-Cazacu (1962), is simple juxtaposition. For instance, 2-yearolds use utterances like *coada calu 'tail horse' i n lieu o f the adult coada calului 'the tail o f the horse'. F r o m about age 3, children start to rely on the variety o f analytic forms. These continue i n common use up to age 6 or 7 when explicit corrections and study o f the synthetic (case-marked) forms take over i n school. Between 3 and 7, they commonly rely on la 'at, t o ' i n lieu o f the genitive case, as in (2;2) *Gura la Dorel for gura lui Dorel 'mouth to (of) D o r e l ' ( = ' D o r e l ' s m o u t h ' ) , or (2;7), i n answer to the question 'whose is it?' * L a un copil for a unui copil 'to (of) a c h i l d ' ( = 'a c h i l d ' s ' ) . F r o m age 3 o n , children come up w i t h a number o f different analytic devices they substitute for the genitive case ending. Characteristically, they use the nominative-accusative form o f the noun w i t h a preposition: la 'at, t o ' as i n the examples already cited, de ' o f , de la ' f r o m ' , despre 'about', all w i t h genitival value i n that the resultant prepositional phrases mark the possessor (Slama-Cazacu, 1962, p . 80). They also rely on analytic forms instead o f the dative case, as i n (2;5) *Dd si la Sanda for dd-i §i Sandei 'give to Sanda t o o ' . When slightly older children start to mark the dative, they appear to do so w i t h a proclitic article w h i c h precedes

732

Clark

the possessor noun, as i n (4;0) / / da sa mdnince lu cd{el, for dative ca\elului, or colloquial lu cd$elu 'he gave to eat to:it d o g ' ( = 'he fed the d o g ' ) , or (3;3) Roata lu masina, for genitive masinii or colloquial lu(i) masina 'the wheel to:it car' ( = ' t h e wheel o f the car'). Synthetic expressions for possession, then, are systematically replaced by young children who construct analytic forms instead. I n doing this, they receive some support from colloquial usage for replacements o f the dative, but the children's o w n constructions are frequently ungrammatical and do not neces­ sarily match those found i n adult speech. Rumanian children between 3 and 7, then, spell out by means o f prepositions and proclitic articles the relation be­ tween the object-possessed and the possessor. Slama-Cazacu attributed late acquisition o f the case forms to the abstractness of the relations expressed by cases. However, the delay noted for Rumanian is not typical i n other highly inflected languages. The explanation seems more likely to lie i n the existence o f a colloquial or spoken standard that relies on some analytic forms alongside the case-marked standard literary language. W i t h t w o kinds o f input for possessive constructions, children presumably first pick up those forms that are most transparent, namely the analytic ones. Since none o f the other Romance languages have retained case marking, further comparisons are difficult. However, children acquiring these languages do show a marked preference for constructions w i t h prepositions (usually 'at', ' t o ' , or ' o f ' ) to signal possession, and this from an early age. I n French, as i n the other Romance languages, there are several different subsystems to be mastered for the expression o f possession. Aside from the analytic forms w i t h a (a moi, a eux, 'to me, to them') and possessive adjectives (mon, ma, mes, etc., ' m y : M A S C , m y : F E M , m y : P L ' ) , one can also use pos­ sessive pronoun forms on their o w n . U n l i k e their English equivalents, these pronouns are used w i t h the definite article, agree i n person w i t h the possessor, and i n gender and number w i t h the object possessed (e.g. le mien ' m i n e : M A S C : S G ' , la mienne ' m i n e : F E M : S G ' , le tien ' y o u r s : M A S C : S G ' , le sien ' h i s : M A S C : S G ' ) . This subsystem, though, appears to be acquired later than the other t w o , and i t takes children some time to integrate i t w i t h prepositional forms and possessive adjectives. During its acquisition, beginning around 3;6 or 4;0, children may overmark for possession, constructing forms like * M o n mien de chapeau a moi ' m y mine o f hat to me' (Karmiloff-Smith, 1977) or identifying the possessor by using phrases like *Le tien de moi 'yours o f me' ( = ' m i n e ' ) , possibly based i n part on formulaic forms (Vinson, 1 9 1 5 - 1 6 , p . 3 3 ) . 1 5

When children were asked to identify the possessor (themselves, the experi­ menter, or a third person) o f color-coded objects heaped on a table, Depreux

Certain combinations, of course, are perfectly acceptable and commonly used, e.g. Mon livre à moi 'MY book', with the possessive pronoun (mon) combined with the prepositional à moi for emphasis. 15

7.

The A c q u i s i t i o n of Romance

733

(1977) found they made frequent mistakes w i t h the possessive pronoun forms. The youngest children (3-year-olds) avoided the possessive pronouns and relied on possessive adjectives and prepositional forms, often i n combination, as i n the question (3;0) *Cest mon papa à toi? ' i t ' s m y father to y o u ' ( = ' i s i t your father'), or identifications like (3;6) Ma voiture à les poupées ' m y car to the dolls' ( = ' t h e dolls' car'). A t this stage, they appeared to have difficulty i n shifting such pronominal forms between first and second person possessors, e.g. (3;6) *mon tien ' m y yours' ( = ' m i n e ' ) i n lieu o f le mien ' m i n e ' , or *le mien à toi 'mine to y o u ' ( = 'yours') i n lieu o f le tien 'yours'. They also had difficulty shifting to third person possessor pronouns, e.g. (3;6) *le tien au pâtisseur 'yours to the baker' ( = ' t h e baker's'). I f only the possessive adjectives and prepositional forms w i t h à are considered, the children have good mastery o f the system. What they still have to learn is how to use this new subsystem o f possessive pronouns i n conjunction w i t h the t w o they already k n o w . This is what appears to lead both to overmarking and to the shifting errors just cited. Depreux (1977) found that there was no real synchronization among the different pos­ sessive forms until about 5;6. Acquisition o f the different forms, à moi 'to m e ' , mon ' m y ' , and le mien ' m i n e ' is not enough. Children also have to learn how to use them i n different contexts. For instance, one could use mon mouchoir ' m y handkerchief where the possessive adjective serves to describe the only handkerchief i n sight, or mes voitures ' m y cars' to contrast the speaker's cars w i t h toy cars on another l o t , where the possessive serves to determine w h i c h o f t w o sets o f cars should be moved. Karmiloff-Smith (1979) found that different possessive forms were used initially w i t h descriptive functions. O n l y at age 8 or 9 d i d children begin to use them w i t h a determiner function as w e l l . As early as age 4, though, children may be aware that possessives, for instance, have some k i n d o f determiner function, but be uncertain how to express i t . This is one reason, she argued (1979, pp. 85¬ 114), w h y children overmark and produce such emphatic expressions as mes voitures à moi ' m y cars to m e ' ' M Y cars') or the ungrammatical *les miennes de voitures rouges ouvertes ' m y cars red open' ( = ' m y red convertibles'). The different functions o f possessive phrases, adjectives, and pronouns are also i n t r i ­ cately bound up w i t h other subsystems o f descripters and determiners i n the language. A n d children have to acquire not only all the forms but also each o f their functions. I n summary, children start their acquisition o f possessive forms w i t h the analytic prepositional constructions, then add the possessive adjectives. Some o f the errors o f contradictory possessor-marking (e.g. *mon papa à toi ' m y father to y o u ' ) that f o l l o w this early stage may result from children's combinations o f formulaic uses o f possessive adjectives w i t h what for them is the " r e a l " way o f indicating possession, namely the preposition à followed by a noun denoting the possessor. (Individual children clearly vary, given Guillaume's and G r é g o i r e ' s data on these t w o possessive subsystems.) Finally, having mastered the preposi-

734

Clark

tional and adjectival devices, children have to learn and integrate yet another subsystem into their set o f markings for possession: the possessive pronouns. They have to re-map the distinctions they already k n o w for the devices acquired earlier and learn both the meanings and the uses to w h i c h the new devices can be put. Elaborating each new subsystem and combining it w i t h what they already know takes time and causes errors.

8. L e x i c a l S t r u c t u r e a n d W o r d - F o r m a t i o n When children first start to acquire words their meanings often differ from the adult's (e.g. G r é g o i r e , 1937, 1947; Montes Giraldo, 1974, 1976). Words m a y b e over- or under-extended i n meaning (Clark, 1973b), be used only i n highly restricted contexts (Carey, 1978), or occur only i n formulaic utterances that have yet to be analyzed by the child (e.g. V i h m a n , 1982). Although there has been some research on lexical domains like that o f spatial terms (see section 9.1), there has been comparatively little analysis o f young children's lexical structure or o f how they b u i l d up different taxonomies o f terms. M y focus i n this section w i l l therefore be on how children exploit some o f the resources available to them for f i l l i n g gaps i n their current lexicon (Clark, 1983). Essentially, I argue that children rely heavily on what they already k n o w about the lexicon and about relations between words k n o w n to them i n finding means to talk about categories for which they still lack the conventional vocabulary. I first consider some options children pursue i n talking about actions, and then take up those they rely on i n talking about objects. I n the case o f actions, I shall focus on what children know about the relations between intransitive, transitive, and causative verbs, and i n the case o f objects, on what they k n o w about options for forming new nouns. 8.1. Doing Things with Verbs. Children begin to talk about actions rela­ tively early i n the one-word stage, but acquiring the appropriate verbs i n many instances takes a long time. Children therefore tend to f i l l i n for verbs they don't yet know by relying, for instance, on general purpose verbs like do (Clark, 1978b; Berman, 1978), on more or less iconic gestures, and on extensions o f other terms to f i l l gaps. They use intransitive verbs causatively and causative verbs intransitively, using the verbs i n question w i t h the requisite number o f arguments; they confuse the meanings o f converse verbs; they sometimes coin new verbs from nouns that denote objects involved i n the actions they wish to talk about; and they coin negative verbs for talking about the reversal o f an action. I w i l l briefly take up the evidence for each o f these observations i n turn. English-speaking children often use intransitive verbs as i f they were causative, as i n *She corned it (meaning, 'she brought i t ' ) beside the intransitive She corned for She came, or *She crossed us (meaning, 'she let us, had us cross') beside We crossed the road. Bowerman (1974, 1982b) was the first to point out that children made intransitive verbs quite freely into causatives i n English,

7.

T h e A c q u i s i t i o n of Romance

735

possibly on the model o f such pairs as walk (intransitive) and walk (causative). She argued that children's overregularizations i n this domain were evidence for their discovery o f the component C A U S E as part o f the meanings o f causative verbs. The emergence o f periphrastic causative forms at the same point i n devel­ opment, w i t h the component C A U S E made quite explicit through use o f such verbs as make or get, lends further support to this interpretation. L o r d (1979) observed that children also made causative verbs into intransitives just as freely, as i n *They don't seem to see, from a child looking for sandals. She suggested that children may make a generalization about verb meanings based on the number o f surface arguments, such that intransitives have one argument, tran­ sitives (often also causative) have t w o , and bitransitives three. They then use the appropriate schema when they need a verb o f the requisite type to convey their meanings. Children, then, presumably make use o f whatever verbs they already have i n their repertoires and stretch them by combining them w i t h differing numbers o f arguments only when needed. W h i l e there has been no systematic study o f children's uses o f verbs from this perspective i n French, examples noted by several observers show considerable resemblance to the English data. For instance, intransitive verbs get used tran­ sitively, as i n ( 1 ; 10) (Je) travaille les cailloux T am w o r k i n g the stones' ( = T am playing w i t h the stones') from a child unfamiliar w i t h the transitive sense o f travailler, (2;6,2) * Ne faut pas tomber le livre 'mustn't fall the b o o k ' or (3;2,5) * 7 « vas te mourir ' y o u ' r e going to die y o u r s e l f (Guillaume, 1927a; G r é g o i r e , 1947; Egger, 1887). Another intransitive used transitively, according to François (1978) is gigoter 'to fidget', as i n (3;6) *Je me gigote ' I ' m fidget­ i n g / w r i g g l i n g m y s e l f or (3;6) *Tu gigotes ma chaise 'you're fidgeting m y chair'. Older children continue to do this, as i n (4;6) * Siffles encore une herbe 'whistle another grass' said to an adult b l o w i n g on a blade o f grass, or (6;0) (Le loup) envole la maison 'the w o l f flies away the house' ( G r é g o i r e , 1939, M é r e s s e - P o l a e r t , 1969). 1 6

The only other Romance language for w h i c h there are data is Portuguese, where Figueira (1979, 1984) analyzed longitudinal records o f spontaneous speech and found verb uses virtually identical to those noted by Bowerman (1974) and L o r d (1979). For instance, among the intransitive to causative exam­ ples the following appear typical: sair 'to go away, leave' for tirar 'to take away' as i n (2; 11,15) * Quern saiu este esmalto do dedo, quern? ' W h o went away this polish for fingers, w h o ' ( = ' w h o took away this nail polish, eh?'); cair 'to f a l l ' for derrubar 'to drop' oxfazer cair 'to make f a l l ' as i n (3;8,15) *Nâo sei se este balanco vai te cair ' n o t : k n o w - l S G i f this swing is:going:to you f a l l ' ( = T don't know i f this swing is going to make you f a l l ' ) ; vir 'to come' for trazer 'to b r i n g ' as i n (3; 11) *A Luisa veio uma menina hoje aqui 'for Luisa I came a d o l l today here' ( = T brought a doll here today for Luisa'); morrer 'to die' for matar 'to Although this verb can be used transitively in very colloquial speech, meaning 'to seduce' or 'to do in', the child's intended meaning on this occasion had to do with falling, or rather dropping. 16

736

Clark

k i l l ' as i n (4; 8,26) *Eu vou morrer essa T m going to die this'; and dormir 'to sleep' for fazer dormir 'to make sleep' as i n (4;7,27) *Eu vou dormir ele aqui ' I ' m going to sleep h i m here'. Figueira also found causative verbs used intransitively, for instance, matar 'to k i l l ' used for morrer 'to d i e ' , as i n (4;8,25) *Apareceu uma pessôa pobre . . . com frio e . . . matou a pessôa 'a poor person appeared . . . w i t h cold and . . . the person d i e d ' , and tirar 'to take away' for sair 'to go away', as i n (4;5,1) *Ela tirou 'she took away' ( = ' s h e went away'). Overall, these data suggest that children take quite a long time to w o r k out any semantic and grammatical constraints on intransitive, transitive, and causative verb categories. I n part, o f course, they simply have to acquire enough vocabulary to realize that not all verbs allow movement from one category to another w i t h i n the language. They also have to learn that transitive and causative verbs agree i n person and number (and on occasion gender) w i t h the subject and not w i t h the direct object. For instance, Antinucci and M i l l e r (1976) found that when Italian children first began using the compound past i n Italian, at around age 2;0, they made the past participle o f the verb agree i n number and gender w i t h the noun for the object affected by the action, rather than w i t h the subject. A s the children extended their uses o f the compound past to verbs for non-change o f state activities, however, they shifted to marking agreement w i t h the subject rather than w i t h the direct object. This k i n d o f shift w o u l d suggest that these children are coming to understand that transitive and causative verbs differentiate among their surface arguments on the basis o f grammatical role such that agreement goes w i t h the subject. Their earlier strategy o f having the past participle agree w i t h the direct object, however, is interesting i n that i t suggests they might have been treating the direct object as i f it was related to the verb i n the same way the subject o f an intransitive verb is. The errors children make i n extending their options for talking about actions by adding or removing surface arguments from their current repertoire o f verbs needs much more detailed study. T o o little is k n o w n at present about the development o f lexical structure i n language acquisition. Children not only extend their verbs i n various ways. They also at times confuse them w i t h each other. For example, a common lexical error for verbs related i n meaning is confusion o f opposites or converses. I n French, children frequently confuse apprendre 'to learn' and enseigner 'to teach', often using apprendre for both meanings. They also make errors w i t h venir 'to come' and aller 'to g o ' , as i n (6;0) * ( / / dit) "Viens," alors j'ai venu avec lui 'he said " c o m e , " s o l c a m e w i t h h i m ' , or (6;0) *Ilfautpas venir à V école ' w e mustn't come to school' where the verb aller (je suis allé avec lui T went w i t h h i m ' ; aller 1 7

This child not only used the wrong deictic verb given the context of his utterance, venir for aller, but also the wrong auxiliary, avoir instead of être. Venir, like several other motion verbs, requires être rather than avoir in the compound past. 17

7.

T h e A c q u i s i t i o n o f Romance

737

à Vécole 'to go to school') was the appropriate deictic choice i n the context o f this utterance ( M é r e s s e - P o l a e r t , 1969). This particular confusion is consistent w i t h the late acquisition o f 'come' and ' g o ' i n other languages (e.g. Clark & Garnica, 1974). The confusion o f opposites i n Portuguese presents a very similar picture. Aprender 'to learn' is used for ensinar 'to teach', and vice versa; subir 'to go up' is confused w i t h llevantar 'to raise, and so on (Figueira, 1977, 1979). A n d the same picture emerges for Spanish, w i t h confusions reported between caer 'to f a l l ' and llevar 'to rise', preguntar 'to ask' and decir 'to t e l l ' , recibir 'to receive' and dar 'to g i v e ' , aprender 'to learn' and enseñar 'to teach', and venir 'to come' and ir 'to go' (Montes Giraldo, 1976; E c h e v e r r í a , 1978). A few o f these confusions may also involve different numbers o f verb arguments, as when the verb 'learn' is used i n lieu o f 'teach'. When children need to talk about a particular activity, they sometimes have recourse either to general purpose verbs or to coinage and construct new verbs for particular occasions (Clark, 1982a, 1983). The general purpose verb most used by children acquiring French is faire. Even children as o l d as 6 rely on it heavily, as i n such uses as faire des tentes for dessiner 'to make tents'; 'to draw t e n t s ' , faire de la peinture for peindre 'to do painting'; 'to paint', faire les jardins for jardiner 'to do gardens'; 'to garden \ faire un accident for provoquer un accident 'to make an accident'; 'to cause an accident' faire la circulation for régler 'to do the traffic'; 'to direct t r a f f i c \ faire la boxe for boxer 'to do b o x i n g ' ; 'to b o x ' , and even faire la marche for marcher 'to do w a l k i n g ' ; 'to w a l k ' (see M é r e s s e - P o l a e r t , 1969). The equivalent verbs i n Portuguese (fazer), Spanish (hacer), Italian (fare), and Rumanian (a face), appear to play the same role i n those languages too. 18

The other option children take is to coin a new verb just for that occasion, by forming a verb from a noun for one o f the objects involved i n the action. For example, French-speaking children form denominal verbs like *mètrer (from mètre 'ruler') for mesurer 'to measure', *pincer (from pince 'paintbrush') for peindre 'to paint'; pantoufler 'to put one's slippers o n ' from pantoufle 'slipper'; ensoldater 'to make into a soldier' from soldat 'soldier'; *pianer (from piano) for jouer du piano 'to play the p i a n o ' , and * piper (from pipe) for fumer 'to s m o k e ' . A comparison o f the categories o f verbs coined by French-speaking children w i t h those from English- and German-speaking children shows strong parallels, even though coining verbs from nouns is not as productive a device i n 19

The forms with faire are generally perfectly acceptable. However, in most instances there is a more specific verb available for the activity being talked about, a verb adults would be more likely to use on such occasions (see further Giry-Schneider, 1978). Some of these verbs, with the meanings intended by the children, have since entered the lexicon. At the time these innovations were observed, they appear not to have been considered conventional. Many of them, of course, represent perfectly legitimate innovations (see Clark, 1982a), while others are unacceptable with the meanings intended. 18

19

738

Clark

French as i t is i n the other t w o languages (Clark, 1982a). What appears critical is that children need a verb for talking about some activity, and to f i l l that gap either rely on a general purpose verb or coin one from an appropriate noun just when they need i t (Clark, 1983). Another type o f verb children often coin is one for talking about the reversal o f an action just mentioned or observed. I n French there is a very productive prefix, dé-, that serves to " u n d o " the requisite action, as i n the well-established verbs faire and défaire 'to do, undo'. Children begin to use this device early to con­ struct reversal verbs i n French. A m o n g the numerous instances that have been noted from children as young as 2;6 are *désendormir (from domir 'to sleep') for réveiller 'to wake u p ' , déchauffef (from chauffer 'to heat'), *déprocher (from approcher 'to approach') for s'éloigner 'to go away', *déprisonner (from emprisonner 'to imprison') for libérer 'to set free, let g o ' , * dé s souffler (from souffler 'to pant, b l o w ' ) for dégonfler 'to deflate', and *délumer (from allumer 'to turn o n ' ) for éteindre 'to put out, turn o f f (Clark, 1981a). Children acquiring Portuguese display a similar facility i n coining reversal verbs, this time w i t h the prefix des-. Figueira (1979), for example^ cited such instances as deszipar (from the innovative denominal zipar 'to z i p ' ) , *desabrir (from abrir 'to open') for fechar 'to shut', *desquentar (from quentar 'to w a r m ' ) for esfriar 'to c o o l ' , *desenfiar (from enfiar 'to put o n ' ) for tiror 'to take o f f ) , and destampar (from tampar 'to cover w i t h a l i d ' ) . She also noted some use o f the overtly negative prefix being used redundantly w i t h verbs that already had the requisite reversal meaning, e.g., *desvaziar for esvaziar 'to deflate' and also *desmurchar (from murchar 'to wither, w i l t ' ) , again i n place o f esvaziar. As i n the case o f French, these innovations start to appear relatively early. Although there are no pertinent data available for the other Romance languages, it seems reasonable to assume that there too it should be relatively easy for children to coin reversal verbs when they need them. The commonest setting i n which children seem to do this is one i n w h i c h they are making a direct contrast or comparison between an action and its reversal. This holds for both the French and Portuguese observations reported here. 8.2. New Words for Objects. Children form new nouns for talking about objects and events for w h i c h they lack words i n their current vocabulary. They begin to do this fairly early i n the Romance languages, and rely on a variety o f derivational suffixes depending on the k i n d o f noun they need. They coin new terms for agents and instruments (the first category I shall consider i n this section), they coin new terms for the objects that result from specific activities, and they coin new terms for talking about particular acts as events i n themselves. They also make use o f an elaborate system o f diminutive and augmentative endings, available i n most o f the Romance languages, to modify the meanings o f nouns along dimensional, affective, and evaluative lines. I n each case, children seem to master first those word-formational options that are productive i n the

7.

T h e A c q u i s i t i o n of Romance

739

language, and the early uses they make o f them are largely appropriate. Where children's uses are not appropriate, it is usually because they have coined a w o r d for some category for w h i c h there is already a w o r d available. Thus, the coinage w i t h that meaning is pre-empted by a previously established w o r d w i t h that meaning i n the language, a w o r d that therefore takes priority (see Clark & Clark, 1979; Clark, 1983). I w i l l take up the data available from each o f the four domains i n turn. Agents and instruments differ i n animacy, but otherwise appear to be close k i n i n most languages. The same suffixes can typically mark both agents and instruments, although other suffixes may be specialized for only one o f these categories. The kinship between agent and instrument is most apparent i n French w i t h the suffix -eurl-euse ( M A S C : S G , F E M : S G ) . This suffix is currently one o f the most productive agentive suffixes, and children seem to rely on it early to coin new agent and instrument nouns. A m o n g the agent nouns observed by such researchers as G r é g o i r e (1947) and A i m a r d (1975) are: (3;8) le crêmeur (from crème 'cream'; for 'eater o f c r e a m ' ) , le preneur (from prendre 'to take'; for 'the taker'), (3;11) le salisseur (from salir 'to d i r t y ' ; for 'the d i r t i e r ' ) , (4;3) *le répareur (from réparer 'to repair'; for réparateur 'repairer'), (4;6) le réfléchisseur (from réfléchir 'to t h i n k ' ; for 'the thinker'), and (5;4) Varrêteur (from arrêter 'to stop'; for 'the stopper'). 20

Earlier records o f children's innovative agent nouns suggest children 50 years ago relied more on -ier then -eur as an agentive suffix, e.g. *le poutrier (from poutre 'beam') for charpentier 'carpenter', un marronnier (from marron 'chestnut'; for 'a seller o f chestnuts'), or le cersonnier (from cerceau 'hoop'; for 'the mender o f hoops') (cf. Egger, 1887; Decroly, 1932). These observations may reflect a change i n the relative productivity o f the t w o agentive suffixes similar to the changes that appear to have taken place i n Polish since the 1890s (cf. Chmura-Klekotowa, 1970). A m o n g the innovative instrument nouns are ones like (3;6) V asticateur (from astiquer 'to p o l i s h ' ; for 'the polisher') and (4;0) *une troueuse (from trou 'hole'; for un perforateur 'a d r i l l ' ) , but children appear to use a wider range o f suffixes for instruments than they do for agents. This variety is reflected i n examples like (4;6) un croquoir (from croquer 'to crack'; for une casse-noisette 'a nutcracker'), (3;5) *un pendule (frompendre 'to hang'; for un porte-manteau 'a hanger'), (3;4) *unepesette (frompeser 'to w e i g h ' ; for une balance 'a weighing machine'), (2;11) * une fume (from fumer 'to smoke'; for une pipe 'a p i p e ' ) , or (2;10) */ (4) 10.3% (3)

Session III

Session V

Session VII

30.4% (7) 13.8% (4) 3.8% (1) 17.7% (11)

18.7% (3) 5.2% (3) 5.7% (2)

Average 24.4% 8.8% 4.6% 13.6%

(20) (10) (55) (20)

826

Ochs

Second, these Tables show that EARLY DEVELOPMENTAL MATURITY IS NOT MATCHED BY INCREASED USE OF AGENTS AND PATIENTS. L o o k i n g at each child, we can see that the use o f these constituents fluctuates from the first session to the last. I n Table 8.13, we can see than i n the final session (seession V I I ) , three of the four children actually display fewer three-constituent utterances that i n the first session. Further, i n both Tables, the youngest child i n the Table, M a t u ' u (2;1 at onset o f study), shows the greatest percentage o f utterances w i t h agents and patients expressed. What is the relation o f these results to adult language use? Children match the speech behavior o f adults i n that the expression o f all three major constituents (agent, verb, patient) is dispreferred. However, on the average, the expression o f agents and agents w i t h patients is less frequent i n the speech o f young children than i n adult speech (see Tables 8.1 and 8.5). I n this sense, children's language use does not match adult usage. This difference resembles descriptions o f chil­ dren's speech behavior i n English-speaking communities (see above references). As discussed earlier, for English-speaking children the non-expression o f these constituents reflects relative competence and counts as an error. W i t h respect to Samoan-speaking children, the meaning o f this pattern is not the same. I n the case o f English, young children are (described as) producing utterances that deviate from adult speech patterns. I n the case o f Samoan, the children are conforming to adult speech patterns. Differences between adult and child lan­ guage use are quantitative rather than qualitative. The quantitative difference I N ITSELF can not be interpreted as incompetence or error. Future research is neces­ sary to ascertain i f other pragmatic conditions enhance the expression/non-ex­ pression o f major constituents and the sensitivity o f children to these conditions.

7.

ACQUISITION

OF ERGATIVE CASE

MARKING

Having examined the acquisition o f tautala lelei and tautala leaga phonological features and word-order strategies, let us now consider the acquisition o f one morphological feature o f the language, ergative casemarking. Definitional and formal dimensions o f ergativity i n Samoan have been discussed i n Section 2 . 2 . 4 . 1 , along w i t h variation i n adult use o f the case marker. The analysis o f adult and children's use o f ergative case marking is drawn from a larger study (Ochs, 1982), to which the reader is referred. The sociolinguistic sketch o f ergative casemarking i n adult speech indicated that it is sensitive to the parameters o f social distance and sex o f speaker. As displayed i n Table 8 . 1 , the case marker is far more common i n speech to nonintimates than i n speech to intimates and more common i n the speech o f men than i n the speech o f women. The most outstanding result o f our acquisition study is that Samoan children between the ages o f 2 and 4 rarely use the ergative marker e i n their spontaneous

8.

The Acquisition of Samoan

TABLE 8.14 A g e n c y & Ergative Case Marking in Spontaneous Novel Canonical Transitives (Assertions & Y-N Q s )

Child Matu'u Iakopo Pesio Naomi Niulala Maselino

Age at Onset of Study

Total UTTS.

UTTS. w/Agents

(2;0) (2;1) (2;3) (2;10) (2;11) (3;3)

76 50 113 109 148 86

22.4% (17) 30.0% (15) 13.25% (15) 15.6% (17) 21.6% (32) 36.0% (31)

Post-Verbal Agents (Erg. Case Mark. Environment) 14.5% 12.0% 4.42% 10.1% 13.5% 33.7%

(11) (6) (5) (11) (20) (29)

827

a

Erg. Case Marking

0 0 0 .9% (1)* .7% (1) 4.6% (4)

* Partial repetition, «(from Ochs, 1982) speech. The frequency w i t h w h i c h the ergative marker appears i n canonical transitives o f five children i n our sample is displayed i n Table 8.14. We can see from this Table that the three youngest children, M a t u ' u , Iakopo, and Pesio, used absolutely no ergative case marking whatsoever. The next oldest children, N a o m i and Niulala, used the marker i n one utterance (each) only, representing . 9 % and . 7 % respectively o f the total canonical transitives and 10.1% and 13.5% respectively o f the transitives w i t h post-verbal agents i n their corpora. These extremely l o w percentages led me to examine the speech o f an older sibling, Maselino, who was not one o f the " f o c a l " children i n the study and present only intermittently throughout the recording sessions. The percent­ age o f ergative casemarking was higher i n the speech o f this c h i l d — 4 . 6 % o f the total canonical transitives and 33.7% o f these w i t h post-verbal agents. However, these figures are still extremely l o w and do not evidence that ergative casemark­ ing is part o f the productive competence o f the child. Instances o f the children's use o f ergative casemarking are presented i n exam­ ples (16) through (18) below: (16)

N a o m i , 2; 11

(Naomi hits mother, asks where her mango is:) Naomi

Mother

ikae uma magol shit finish mango 'Shit, the mango is finished.' uma finish

mago mango

a'ul my

'My mango is finished.'

828

Ochs

tae shit

tae shit

uma finish

magol mango

'Shit, shit, the mango is finished.' uma finish

mago mango

a'ul my

'My mango is finished.' uma finish

ail eat

'The eating is finished.' ai eat

e ail E R G who

'Who ate it?' (?)/

fea where 'Where?' umal finish 'Finished.' > ai eat

e oe E R G you

'You ate it.' (17) Maselino, 3;8 (Pesio, 2;7 is crying, looking at her father. Another child, Kala, has hit her, though this has not been mentioned. Her father wants her to stop crying:) Pesio: (Crying)

Father: ((soft)) (alu go

loa)l now

Paula (female caregiver): Pesiol Maselino: Kala/ hit

E R G Kala

'Kala hit her.' Elenoa: I IKalal

8.

The Acquisition of Samoan

Maselino: > fasi hit

e Kalal E R G Kala

'Kala hit her.' (18)

Maselino, 3;6

(Maselino decides to scare another child by using a common scare expression about mother absence. He turns to this child, Gike:) Maselino: >Gike!

Gike!

le DEICT

'ua 'ai PERF eat

e lei I pua'a E R G ART pig

Koel Koe* (*Gike's mother) 4

Gike! Gike! Now the pig ate Koe.'

Maselino: fiu tired

e COMPL

suel search

'They are tired of searching.' fiu tired

e COMPL

kue* search

*error

akul DEICT

'They are tired of searching around.' > le DEICT

'ua PERF

'ai eat

e le pu'al E R G A R T pig

'Now the pig ate her.'

> le ' ua DEICT PERF

'ai eat

e le povi E R G A R T cow

Gike/ Gike

'Now the cow ate her, Gike'

/ /'ua 'ai PERF eat

e le povi E R G A R T cow

'The COW ate Koe.'

a PARTICLE

Koe Koe

829

830

Ochs

W e can see from Table 1 that these results are linked to the l o w frequency o f agents expressed i n transitive declarations and yes-no interrogatives (average percentage o f agents expressed: 2 1 % (112) ) . That is, one explanation is that children simply do not often mention agents i n their spontaneous speech. This explanation w o u l d not account, however, for w h y marking is not used when agents ARE expressed and appear i n the grammatical environment that should host the ergative marker (after the verb). Y o u n g children do use constructions i n which ergative case marking is required (according to speakers' judgments o f " g o o d Samoan"), yet they do not use the casemarking. It is proposed here that THE RELATIVELY LATE ACQUISITION OF ERGATIVE CASEMARKING IS TIED TO THE SOCIOLINGUISTIC DISTRIBUTION OF THIS MARKING AMONG ADULT SAMOANS and to certain features o f the marker i n Samoan that may affect its perception. Features that enhance or constrain acquisition have been discussed by Slobin (1973, 1977, 1982). These features and their status w i t h respect to the Samoan ergative case marker are displayed i n Table 8.15. Presence o f these features, represented by the + sign, facilitates the acquisition o f the morphological marker. As can be seen from this Table, the Samoan ergative particle has six features that could delay acquisition. I n particular, i t is not postposed, stressed, obliga­ tory, tied to the noun it modifies, or applied to all pro-forms. Further, i t has pragmatic functions, such as highlighting a noun and its phrase. I t is difficult to weigh the importance o f these features, as the ergative particle i n another lan­ guage, K a l u l i (Schieffelin, 1979, 1985), has five features that could delay ac-

TABLE 8.15 S a m o a n a n d Kaluli Ergative Case M a r k i n g Features Feature: postposed : syllabic stressed obligatory tied to noun rationally ordered consistent with word order pattern non-synthetic only grammatical functions regular applied to all pro-forms no homonymous functors*

Samoan

Kaluli

+ —

+ + (uses tonal contrast, prosodic system not worked out) + n.a. + n.a. -

n.a. + n.a. + +

* "Functor" is here taken to mean "case marker."

8.

The Acquisition of Samoan

831

quisition, yet K a l u l i children acquire the particle relatively EARLY. Four out o f five perceptually distressful features o f the K a l u l i marker match those o f the Samoan marker. K a l u l i is distinguished by the fact that the particle is postposed and tied to the noun. These features could account for w h y K a l u l i children acquire the ergative marker before Samoan children. However, there are major differences i n the social norms for using the ergative case marker i n K a l u l i and Samoan. I n K a l u l i , the case marker is used across speakers and role relationships among interlocutors. I t does not vary according to social distance and sex o f speakers, as i n Samoan communities. The K a l u l i child is exposed to the marker i n his primary acquisition environment to a much greater extent than is the Samoan c h i l d . The difference i n social norms for using the ergative case marker i n Samoan and K a l u l i has, then, a profound effect on acquisition o f these t w o languages. Further, the difference affects the way i n which deletion or l o w incidence o f the marker i n children's speech should be interpreted. When a Samoan child does not use the ergative particle i n the grammatically feasible environment, that behavior is not necessarily, indeed not usually, an error. I f the child is speaking to an intimate under relatively casual conditions, the absence o f the ergative case marker is perfectly appropriate. The child has not deviated from adult norms i n a qualitative sense (as w o u l d be the case, for example, for a K a l u l i child i n similar social context). Once again, the data indicate that assessment o f errors must rely on a sociologically responsible description o f the language. T o contrast c h i l ­ dren's speech behavior w i t h typological accounts o f languages runs a heavy risk o f misunderstanding the nature o f the linguistic phenomenon under consideration and the dynamics o f its acquisition.

8. A C Q U I S I T I O N O F E R G A T I V I T Y T H R O U G H W O R D ORDER The results o f the ergative casemarking study should not be taken as evidence that young Samoan children do not express ergative distinctions i n their speech. On the contrary, ERGATIVE DISTINCTIONS ARE EXPRESSED QUITE EARLY IN"SA­ MOAN CHILD LANGUAGE. Rather than marked through morphological means, however, THE DISTINCTIONS EMERGE I N EARLY WORD-ORDER STRATEGIES. I N PARTICULAR, YOUNG CHILDREN RESERVE THE LOCATION IMMEDIATELY FOLLOW­ ING THE VERB FOR ABSOLUTIVE CONSTITUENTS INTRANSITIVE MAJOR ARGUMENTS

TRANSITIVE PATIENTS AND

BUT EXCLUDE ERGATIVE CONSTITUENTS

AGENTS—FROM THIS POSITION. ( I n this w a y , they treat patients and intransitive arguments as a single category, distinct from agents.) This pattern can be seen by comparing Table 8.9 (cf. Section 6) and Table 8.16.

Ochs

832

TABLE 8.16 Spontaneous Novel Canonical Transitives (Declaratives & Yes-No Interrogatives) Percentage o f (X)-V'erb-Patient-{X) W o r d Orders (Compared t o (X)-V'erb-Agent-iX)) Child Matu'u Iakopo Pesio Naomi Niulala

Session I

Session III

Session V

Session VII

Average

100.0% (12)

94.4% (17) 100.0% (2) 100.0% (3) 64.3% (9) 88.9% (16)

94.7% (18) 93.8% (15) 91.7% (11) 100.0% (15) 88.9% (40)

83.3% (10) 100.0% (22) 97.4% (38) 83.9% (26) 72.7% (16)

92.9% 97.5% 97.3% 86.0% 84.5%

— 100.0% (8) 95.8% (23) 87.5% (14)

The use o f post-verbal position for absolutive constituents is most striking i n the speech o f the youngest children i n this sample. I n the earliest sessions i n which V - N P transitive constructions appear, the N P is a patient 100% o f the time. That is, an agent is never expressed i n this position. A t the same time, for these children, between 96 and 100% o f the intransitive major arguments ap­ peared i n post-verbal position. For all the children i n our sample, the percentage of transitive verb-patient and intransitive verb-major argument orders is ex­ tremely high, providing strong evidence for the systematic use o f w o r d order to distinguish absolutive from ergative constituents. The preference to exclude the agent N P from the position immediately f o l l o w ­ ing the verb is evident as w e l l i n children's constructions that express BOTH patient and agent constituents. The reader is referred to Table 10, w h i c h displays the percentages and frequencies o f different w o r d order patterns o f three constitu­ ent utterances present i n the corpus. THE FIGURES IN TABLE 10 INDICATE THAT VERB-AGENT-PATIENT AND PATIENT-VERB-AGENT ORDERS ARE HIGHLY DISPREFERRED AMONG THE YOUNGEST CHILDREN IN THE STUDY. I f agent and patient are both to be expressed, these children tend to place the agent before the verb or after the patient (e.g. agent-verb-patient, verb-patient-agent). The preference for verb-agent-patient ordering increases w i t h age o f speaker, but never accounts for more than a third o f the transitive assertions and yes-no questions i n w h i c h the two major constituents appear. To summarize, ergative relations are expressed i n the early stages o f Samoan language acquisition. They are expressed through w o r d order rather than through casemarking. These results are paralleled i n the w o r k o f Goldin-Meadow (1975), who

found that the deaf children she observed used w o r d order to distinguish

causative agents from both patients and intransitive entities. These results are also consistent w i t h a number o f findings i n the child language literature that show children relying on w o r d order as an initial strategy for expressing semantic relations (Bever, 1970; B l o o m , 1970; Radulovic, 1975; but cf. Slobin, 1982).

8.

The A c q u i s i t i o n o f S a m o a n

833

9. R E L A T I N G S O C I O L I N G U I S T I C S T O L A N G U A G E ACQUISITION This paper has addressed the issue o f what constitues a n o r m , and hence, an error, i n adult and child language. T o reach an understanding o f how children attain linguistic competence, it is necessary to understand what constitutes com­ petence itself. That is, it is necessary to k n o w the range and structure o f adult language. This involves an awareness o f language i n terms o f regional dialects, social dialects, and registers. T y p i c a l l y , c h i l d language subjects are drawn from the regional and social dialect that is (best) captured i n traditional and typological grammars. But even i n such a sample, there are norms o f adult language not captured i n such grammars. The norms not captured tend to be those charac­ teristic o f registers appropriate to intimate and informal social situations, the very situations i n w h i c h young children are most exposed. H o w can we capture the relation o f child to adult language without knowledge o f these registers? The best grammar for understanding language acquisition is a sociolinguistically responsive one. B y this I mean a grammar that captures the range o f linguistic structures systematically i n use and relates those structures to the social and linguistic contexts i n w h i c h they are i n use. Such a grammar ideally w o u l d rank the linguistic and social conditions associated w i t h a particu­ lar variant (see Labov, 1972; and Cedergren & Sankoff, 1974, for a detailed discussion o f this technique). The grammar w o u l d tell us, for example, w h i c h social factor most affects the probability o f a particular variant being used (e.g. sex o f speaker, social distance between speaker and hearer, relative rank o f speaker vis-a-vis hearer, social event at hand, genre i n use, etc.). A n ideal sociolinguistic grammar w o u l d provide us w i t h sets o f linguistic structures that carry out similar communicative functions under different social conditions. I t w o u l d also specify for a particular socially significant context, those linguistic structures that regularly cooccur (Ervin-Tripp, 1972). That is, a sociolinguistic grammar should ideally specify the features that characterize and distinguish one register from another, one social dialect from another (Andersen, 1977). The latter goal is the most consuming i n that i t demands examining phonological, morphosyntactic, and discourse features as they are used i n different socially significant contexts. Such a grammar has researcher: 1. IT ALLOWS CHILD

a number o f advantages for the child

FOR A SOCIAL

CONTEXT-SENSITIVE

COMPARISON

language

OF ADULT

AND

LANGUAGE.

W i t h such a grammar, we are able to compare the language behavior o f adults in a particular social context w i t h child language i n that (or a comparable) context. W e can see, for example, the way i n w h i c h adults speak to family

834

Ochs

members i n informal situations and compare that speech to children's speech under the same conditions. Similarly, we can compare the way i n which adults speak i n a range o f more formal situations to the speech o f children i n those settings. This methodology has obvious advantages over one i n w h i c h children's speech is compared to only the careful speech elicited through linguistic inter­ views or to the speech o f a parent addressing only the child or the researcher/outsider. 2.

IT SPECIFIES QUIRED

THE SOCIAL

BY THE

STATUS

OF THE LINGUISTIC

STRUCTURE

BEING

AC­

CHILD.

A sociolinguistic description w o u l d state whether or not a structure under consideration is socially variable or invariable i n the adult linguistic repertoire. As noted by Slobin (1982), whether or not a structure is optional affects when that structure w i l l be acquired. A sociolinguistic description says more than this, however, i n that i t displays the conditions under w h i c h the structure is used and the extent to w h i c h it is used. This information provides a crucial source for predicting the point at w h i c h a young child w i l l acquire competence i n the use (or non-use) o f a structure. For example, a structure may be socially variable, but it is a distinguishing characteristic o f household vernacular; therefore, its " o p t i o n a l i t y " is less likely to inhibit its acquisition than i f i t were restricted to formal or public arenas outside the household.

3.

IT SPECIFIES PARTICULAR

FOR BOTH STRUCTURE

ADULT

AND CHILD

IN AND ACROSS

THE FREQUENCY

SOCIAL

OF USE OF A

CONTEXTS.

I n matching child and adult language, i t is important to compare frequencies o f appearance o f a particular structure i n these t w o sets o f data. A traditional or typological grammar provides information concerning linguistic rules but does not specify the extent to w h i c h that rule is applied i n actual adult language use. Matching children's frequencies against such a description runs the risk o f over­ estimating adult behavior (and underestimating children's behavior). For exam­ ple, while a grammar may specify word-order patterns o f verb, subject, and object, it may not reveal the extent to w h i c h all three constituents are expressed in adult language use. The examination o f transitive utterances i n adult Samoan, for instance, indicated that three-constituent utterances were relatively infrequent across several social contexts. The l o w frequency o f three-constituent utterances in Samoan child language must be evaluated i n this light, i.e. as reflecting an adult language n o r m (rather than violating an adult norm). Differences between children's speech and adult speech ( w i t h respect to this phenomenon) are quan­ titative, w i t h the children's speech showing lower frequencies. This i n turn leads

8.

The Acquisition of Samoan

835

to another arena o f discussion concerning the implications o f quantitative differences. 4.

IT EVALUATES THE

SPEECH

CAREGIVER

SPEECH

IN TERMS

OF REGISTRAL

VARIATION

WITHIN

COMMUNITY.

A sociolinguistic knowledge o f the language w o u l d refine substantially our understanding o f what constitutes input i n general and caregiver speech register in particular. For example, the results o f the ergativity study i n Samoan indicate that the way i n w h i c h adults speak to young children is part o f a larger register o f speech used among family members. I t also indicates that the features o f care­ giver speech are characteristic o f informal women's speech i n and out o f the family. This means that adults do not speak to a young child i n a certain manner because he/she is a child (immature) but because he/she is an intimate or because the speaker is a w o m a n speaking i n an informal setting. 5.

IT PROVIDES

SOCIAL

SOURCES

FOR ACQUISITION

STRATEGIES.

A sociological account o f language specifies not only the repertoire o f speak­ ers but VALUES attached to particular codes (dialects, registers, etc.) w i t h i n the repertoire. A good deal o f sociolinguistic research has addressed the effects o f such values on speech use (see Ferguson, 1959; B l o m & Gumperz, 1972; Hymes, 1974; Labov, 1963; for example). O f particular interest to acquisition research is the fact that social values may lead speakers to overgeneralize (or undergeneralize) particular grammatical rules. Labov's research on hypercorrection (Labov, 1966, 1972), for example, demonstrates that lower-middle-class speakers i n N e w Y o r k City w i l l use, i n their careful speech, phonological features o f the prestige dialect i n environments beyond those characteristic o f upper-middle-class speech. Similarly, certain speakers o f Samoan w i l l overextend the use o f the prestigious phonological feature IXI i n their use o f " g o o d speech" (tautala lelei). These processes are not limited to adult speech behavior. I have demonstrated that Samoan children are sensitive to the features associated w i t h highly valued tautala lelei register such that they use them i n phonological contexts not charac­ teristic o f standard tautala lelei. L i k e adults, the children are sensitive to the particular importance gven to the feature IXI as a distinctive marker o f tautala lelei i n that the bulk o f their overextensions concerns overuse o f IXI. of IXI. Information concerning the SOCIAL characteristics o f linguistic structures needs to be integrated w i t h what we already k n o w about the PHYSICAL charac­ teristics o f these structures and their grammatical environment, to assess con­ straints on acquisition and strategies that children implement over developmental

836

Ochs

t i m e . Children pay attention to the ends o f words and to the order o f words. A t the same time they acquire language i n a social w o r l d and w i l l pay attention to the social significance o f words and the social conditions for using them.

ACKNOWLEDGMENT The research on which this study is based has been supported by The National Science Foundation (Grant No. 53-482-2480) and a Senior Fellowship from The Department of Anthropology, Research School of Pacific Studies, The Australian National University.

REFERENCES Andersen, E . Learning to speak with syle: A study of the sociolinguistic skills of young children. Unpublished doctoral dissertation, Stanford University, 1977. Bever, T . The cognitive basis for linguistic structures. In J . R. Hayes (Ed.), Cognition and the development of language. New York: Wiley, 1970. Blom, J-P., & Gumperz, J . J . Social meaning in linguistic structures: Code-switching in Norway. In J. J . Gumperz & D. Hymes (Eds.), Directions in sociolinguistics: The ethnography of commu­ nication. New York: Holt, Rinehart & Winston, 1972. Bloom, L . Language development: Form and function in emerging grammars. Cambridge, MA: MIT Press, 1977. Braine, M. D. S. The ontogeny of English phrase structure: The first phase. Language, 1963,39, 1¬ 13. Brown, R. A first language: The early stages. Cambridge, MA: Harvard University Press, 1973. Brown, R . , & Gilman, A. The pronouns of power and solidarity. In T . Sebeok (Ed.), Style in language. Cambridge, MA: MIT Press, 1960. Cedergren, H . , & Sankoff, D. Variable rules: Performance as a statistical reflection of competence. Language, 1914, Chomsky, N. Deep structure, surface structure and semantic interpretation. In D. Steinberg & L . Jakobovits (Eds.), Semantics. Cambridge: Cambridge University Press, 1971. Chung, S. Case marking and grammatical relations in Polynesian. Austin: University of Texas Press, 1978. Churchward, S. A new Samoan grammar. Melbourne: Spectator Press, 1926. Comrie, B . Ergativity. In W. P. Lehmann (Ed.), Syntactic typology: Studies in the phenomenology of language. Austin: University of Texas Press, 1978. Dixon, R. M. W. Ergativity. Language, 1979, Duranti, A. Lduga and talanoaga: Structure and variation in the language of a Samoan speech event. In R. Bauman & J. Scherzer (Eds.), Working Papers in Sociolinguistics, No. 7. University of Texas, Austin, 1980. Duranti, A. The Samoan fono: A sociolinguistic study. Pacific Linguistics, Series B , No. 80, Department of Linguistics, The Australian National University, Canberra, Australia, 1981. Ervin-Tripp, S. On sociolinguistic rules: Alternation and cooccurence." In J . J . Gumperz & D. Hymes (Eds.) Directions in sociolinguistics: The ethnography of communication. New York: Holt, Rinehart & Winston, 1972. Ferguson, C . Diglossia. Word, 1959, Givon, T. (Ed.) Syntax and semantics, Vol. 12: Discourse and syntax. New York: Academic Press, 1979. Goldin-Meadow, S. The representation of semantic relations in a manual language created by deaf

8.

T h e A c q u i s i t i o n of S a m o a n

837

children of hearing parents: A language you can't dismiss out of hand. Technical Report X X V I . University of Pennsylvania, 1975. Greenberg, J . H . (Ed.). Universals of language, 2nd ed. Cambridge, MA: MIT Press, 1966. Gumperz, J . Sociocultural knowledge in conversational inference. In M. Saville-Troike (Ed.), Georgetown University 28th Round Table on Languages and Linguistics. Washington, D . C . : Georgetown University, 1977. Hymes, D. Foundations in sociolinguistics: An ethnographic approach. Philadelphia: University of Pennsylvania Press, 1974. Irvine, J . Formality and informality in communicative events. American Anthropologist, 1979, 81, Kernan, K . The acquisition of language by Samoan children. Unpublished doctoral dissertation, University of California, Berkeley, 1969. Kernan, K . The acquisition of formal and colloquial styles of speech by Samoan Children. Anthropological Linguistics, 1974. Labov, W. The social motivation of a sound change, Word, 1963. Labov, W. The social stratification of English in New York City. Washington, D . C . : Center for Applied Linguistics, 1966. Labov, W. Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press, 1972. L i , C . (Ed.). Subject and topic. New York: Academic Press, 1976. Milner, G . B. Active, passive, or perfective in Samoan: A fresh appraisal of the problem. Journal of Polynesian Soc, 1962. Milner, G . B. Samoan Dictionary. London: Oxford University Press, 1966. Milner, G . B. It is aspect (not voice) which is marked in Samoan. Oceanic Linguistics, X I I , 1-2, 1973. Ochs, E . Ergativity and word order in Samoan child language: A sociolinguistic study. Language,

mi.

Partridge, E . Origins: A short etymological dictionary of modern English. London: Routledge & Kegan Paul, 1966. Pawley, A. Polynesian languages: A sub-grouping based on shared innovations in morphology." Journal of Polynesian Soc, 1966, 75, Pawley, A. Samoan phrase structure: The morphology syntax of a western Polynesian language Anthroplogical Linguistics, 1, 1-63. Pawley, A. The relationships of Polynesian outlier languages. Journal of Polynesian Soc., 1967, 76, Pratt, G . Pratt's grammar and dictionary of the Samoan language. Apia: Malua Printing Press, 1911. Radulovic, L . Acquisition of language: Studies ofDubrovnik children. Unpublished doctoral disser­ tation. University of California, Berkeley, 1975. Schegloff, E . , Jefferson, G . & Sacks, H. The preference for self-correction in the organization of repair in conversation. Language, 1977, 53, Schenkein, J. (Ed.). Studies in the organization of conversational interaction. New York: Academic Press, 1978. Schieffelin, B. How Kaluli children learn what to say, what to do, and how to feel: An ethnographic study of the development of communicative competence. Unpublished doctoral dissertation, Co­ lumbia University, 1979. Schieffelin, B. B . The acquisition of Kaluli. In D. I. Slobin (Ed.), The crosslinguistic study of language acquisition (Vol. 1). Hillsdale, NJ: Lawrence Erlbaum Associates, 1985. Shore, B. A Samoan theory of action: Social control and social order in a Polynesian paradox. Unpublished doctoral dissertation, University of Chicago, 1977. Shore, B. Sala'ilua: A Samoan mystery. New York: Columbia University Press, 1982. Slobin, D. I. Cognitive prerequisites for the acquisition of grammar. In C . A. Ferguson & D. I. Slobin (Eds.), Studies of child language development. New York: Holt, Rinehart & Winston, 1973.

838

Ochs

Slobin, D. I. Language change in childhood and in history. In J. Macnamara (Ed.), Language learning and thought. New York: Academic Press, 1977. Slobin, D . I . Universal and particular in the acquisition of language. In E . Wanner & L . R. Gleitman (Eds.), Language acquisition: The state of the art. Cambridge: Cambridge University Press, 1982. Tuitele, M. T . , & Kneubuhl, J . 'Upu Samoa/Samoan words. Pago Pago: Bilingual/Bicultural Education Project of American Samoa, 1978.

The Acquisition of Turkish

Ayhan A. Aksu-Kog Bogazigi

University,

Istanbul

Dan I. Slobin University

of California,

Berkeley

Contents Introduction 839 Brief Grammatical Sketch of Turkish 839 Sources of Evidence 843 Brief Summary of Overall Course of Linguistic Development in Turkish The Data 847 Typical Errors 847 Error-Free Acquisition 854 Timing of Acquisition 856 The Setting of Language Acquisition 861 Cognitive Pacesetting of Language Development 861 Linguistic Pacesetting of Cognitive Development 865 Input and Adult-Child Interaction 866 Individual Differences 869 Conclusions 869 Reorganizations in Development 869 Operating Principles 871 Suggestions for Further Study 876

845

INTRODUCTION 1. B r i e f G r a m m a t i c a l S k e t c h o f T u r k i s h Turkish is an Altaic language, exhibiting i n almost pure form the classic features of an object-verb language (Greenberg, 1966; Lehmann, 1978). The most ac­ cessible brief grammar i n English is Lewis (1953), which can be supplemented by his extensive Turkish grammar (1967). A collection o f linguistic studies o f Turkish can be found i n Slobin and Z i m m e r ( i n press). 839

840

Aksu-Koç and S l o b i n

The neutral w o r d order is S O V , w i t h concomitant features o f suffixed inflec­ tions, postpositions, and preposed demonstratives, numerals, possessives, adjec­ tives, and relative clauses. W o r d order i n simple sentences and main clauses exhibits a high degree o f variation for pragmatic purposes (Erguvanh, 1979). Morphology is agglutinating and remarkably regular, w i t h only a handful o f exceptions to general principles. V o w e l harmony operates throughout all words o f native origin and for all grammatical suffixes, w h i c h harmonize w i t h the last v o w e l o f the noun or verb stem. Suffixes f o l l o w one o f t w o main alternations: (1) a front-back alteration o f unrounded l o w vowels, el a, represented here by the phonematic u n i t i s , and (2) a front-back, rounded-unrounded alternation o f high vowels, ililiilul,

represented

by / . Compare, for example, the locative suffix -dE, realized as Izmir-de I z m i r ' and Istanbul-da Tahran-in,

' i n Istanbul', and the genitive -In, realized as

Istanbul-un,

and Atatiirk-un.

4

in

Izmir-in,

Uninterrupted v o w e l sequences are

avoided by the use o f buffer consonants for v o w e l - i n i t i a l suffixes, each such inflection carrying its o w n buffer, as, for example, the -(n)In o f the genitive, resulting i n such forms as

Ankara-nin.

I n strings o f agglutinated morphemes, each element retains its phonological and semantic identity as w e l l as its relative position i n the string. For example, consider the order o f noun suffixes: stem + (plural) + (possessive) + (case), as in el 'hand', -ler

' p l u r a l ' , -im 'first person possessive', -de

'locative'. The

following combinations are possible: el 'hand' el-im 'my hand' el-de 'in hand' el-im-de 'in my hand'

el-1er 'hands' el-ler-im 'my hands' el-ler-de 'in hands' el-ler-im-de 'in my hands'

W i t h very few exceptions, the language avoids homophonous morphs.

Each

morph is syllabic and stress is fairly evenly distributed across syllables, w i t h typical word-final stress. Nouns are case-marked for genitive, accusative, dative-directional, locative, ablative, comitative-instrumental, and deprivative ( ' w i t h o u t ' ) . The same suffix­ es are also applied to pronouns, demonstratives, question words, and derived nouns. Consider, for example, the ablative forms o f nouns (masa-DAN

'from

(the) table'), pronouns (sen-DEN

'from

t h i s ' ) , question words (nere-DEN derived nouns (yiiz-mek-TEN

' f r o m y o u ' ) , demonstratives (bun-DAN ' f r o m where', kim-DEN

' f r o m w h o m ' ) , and

' s w i m - N O M L - A B L ' = ' f r o m s w i m m i n g ' ) . There

is a variety o f denominal and d é v e r b a l derivational suffixes, as discussed i n regard to typical errors below. There is no grammatical marking o f gender. Verbal affixes mark voice, negation, modality, aspect, tense, person, and number, w i t h person and number affixes bearing much similarity w i t h nominal suffixes for the same functions. For example, consider the plural -lar and first

9.

The Acquisition of Turkish

841

singular -um i n the f o l l o w i n g verb examples, already familiar from the noun suffixes listed above: al

take al -tyor take PROG

al take

'he/she/it is taking'

'they are taking'

al take

-lyor PROG

-lyor PROG

-lar PL

-um 1SG

T am taking' Even at early ages fairly elaborated strings o f verbal affixes are produced by children. Several examples, picked at random, show the character o f Turkish verbal morphology (ages i n parentheses): (2;1) getir bring

-me -di N E G PAST

-n 2SG

'you didn't bring' (2;4) agla cry

-di PAST

-lar PL

'they cried' (3;2) duz straight

kon put

-ul PASS

-ur AORIST

-sa COND

'if one puts (it) straight' Roughly, the verb i n Turkish allows for the f o l l o w i n g series o f affixes ( w i t h i n the bounds o f semantic plausibility, o f course): stem—reflexive—reciprocal—causative— passive—potential—negative—necessitative— tense—conditional—question—person—number As this scheme shows, the particles affixed to the verb express notions o f tense, aspect, m o o d , and modality and various combinations o f such notions. I n terms of tense, there is distinct marking o f past (-di or -mis), present (-lyor or -Ir), and future (-EcEk). W i t h i n the past, a modal distinction is drawn between statements made on the basis o f direct evidence (-di) vs. indirect evidence such as inference or hearsay (-ml§) (for detailed discussion o f this distinction, see Slobin & A k s u ,

842

Aksu-Kog and Slobin

1982). I n the present there is an aspectual distinction between progressive (-lyor) and habitual (-Ir) (frequently referred to as " a o r i s t " ) . I n addition to these tense and aspect markers, there is a conditional suffix (sE), (-mElI),

a necessitative

suffix

and a collection o f optative-imperative suffixes marked for both person

and mood (e.g. 1SG -Eylm, 3SG -sin). Each o f these suffixes can be used alone w i t h a verb, followed only by person-number marking. I n addition, there is a collection o f particles that can only be interposed between the verb and the above suffixes. Four o f these " i n t e r f i x e s " modify the verb i n the f o l l o w i n g ways: passive (-//-), causative (-dlr-), reciprocal (-Is-), and reflexive (-In-). A fifth is the negative (-mE-),

w h i c h follows these operators on

the verb, or the verb stem alone, shifting stress back, onto the preceding syllable. Three o f the suffixes can be used to further modify the meanings o f verbs already containing particles listed above to express complex temporal, aspectual, and modal notions. These three suffixes are the past o f direct evidence (-all), the past o f indirect evidence (~ml§), and the conditional (-sE).

Ordering o f particles

is crucial for resultant meanings. For example, past + conditional expresses 'past conditional' (e.g. gel-di-yse conditional +

' c o m e - P A S T - C O N D ' = ' i f he came') while

past expresses counterfactual (e.g. gel-se-ydi

'come-COND-

P A S T ' = ' i f he had c o m e ' ) . Where only one resultant meaning is possible, ordering is fixed (e.g. gel-iyor-du not *

'come-PROG-PAST' = ' h e was c o m i n g ' but

gel-di-iyor).

Verbal complexes o f any length can be nominalized and treated as embedded clauses, as discussed i n examples from child speech below, i n regard to n o m i nalization errors and relative clauses. As an example o f such forms, w h i c h are fairly infrequent i n child speech, consider relative clauses formed on underlying objects, as i n the Turkish equivalent o f 'the man w h o m A l i saw'. ' A l i ' , the underlying subject i n the relative clause, surfaces i n the genitive, symbolically possessing a nominalized f o r m o f his action o f 'seeing', and this entire possessed nominal is preposed to the head noun ' m a n ' : Ali Ali

-nin gor G E N see

-dug PART

-u POSS:3

adam man

Neutral w o r d order is S O V , w i t h frequent deletion o f subject pronouns, since person and number o f subject are marked on the verb (see Eng, i n press, for a discussion o f pragmatic factors governing pronoun use and deletion; and Slobin and Talay, i n press, for a developmental study o f these issues.). The pre verbal position is one o f focus, thus a l l o w i n g for O V S to focus objects and O S V to focus subjects. A l t h o u g h the language is basically verb-final, subjects and adverbials are frequently placed after the verb i n conversation, performing pragmatic functions relating to topic continuity and turn-taking. Question words normally appear i n preverbal position.

9.

T h e A c q u i s i t i o n of Turkish

843

Sentences are conjoined either by conjunctions or by verbal suffixes, the latter device being more natural to the language. A collection o f verbal particles or " c o n v e r b s " allow for various sorts o f verb chaining to indicate temporal and causal relations. I n such series, only the last verb is finite, as i n the following example from a play monologue o f a child o f 4;4. The first verb, suffixed by-ip, indicates an action that is prior i n the series o f events: Deniz sea

-e atlay DAT jump

-ip ytiz P T L swim

-eceg FUT

-im. 1SG

'I'll jump into the sea and swim.' Numerous postpositions are also available to encode interclausal relations. I n the following example, a child o f 3;0 uses the postposition igin 'for, i n order t o ' to explain w h y she has removed her doll's clothes: Yika wash

-mak INF

igin gikarttim. for removed+lSG

T took off (doll's clothing) in order to wash (it).' Further aspects o f Turkish grammar relevant to acquisition are discussed i n more detail at various points below. 2. S o u r c e s o f E v i d e n c e The first published w o r k on Turkish child language is a report i n Turkish by Ozbaydar (1970), briefly summarizing a longitudinal observation o f one boy and one g i r l between the ages o f 1 2 - 2 4 months. A detailed longitudinal study is presented i n Ekmekgi's (1979) dissertation, presenting the results o f monthly hour-long recordings o f a girl between the ages o f 1;3 and 2;4. Ekmekci (in press) has recently published a detailed discussion o f the development o f w o r d order i n this child. I n addition to these published reports, we have had access to diary observations gathered by Belma and Sabri Ozbaydar, Dogan Cuceloglu, Ozcan B a § k a n , and N a i l § a h i n and his students. T o all o f those investigators we express our thanks. Our o w n materials, gathered by Slobin i n 1969-70 and 1972-73, and by A k s u i n 1 9 7 7 - 7 8 , include extensive speech corpora from about a dozen children, ranging i n age from 1; 10 to 5; 11. These corpora consist primarily o f child-investigator interactions, though mother-child and sibling i n ­ teractions are also included. I n 1972-73, as part o f the Berkeley Cross-Linguistic Acquisition Project, a 1

^he Berkeley Cross-Linguistic Acquisition Project was carried out with support from the William T. Grant Foundation to the Institute of Human Learning and from NIMH to the Language-

844

Aksu-Koç a n d S l o b i n

cross-sectional and micro-longitudinal sample o f 48 children was studied i n depth. W e worked w i t h groups o f six children—three boys and three girls—at each o f eight age levels, spaced at 4-month intervals between the ages o f 2;0 and 4;4. I n addition to this cross-sectional design, each child was retested 4 months later, providing one longitudinal check, and g i v i n g an overall age range o f 2;0 to 4;8. A large range o f linguistic areas was examined, including locatives, causatives, agent-patient relations, temporal expressions, comparatives, relative clauses, question comprehension, and free speech samples. The overall study, comparing Turkish, Serbo-Croatian, Italian, and English, is summarized i n Slobin (1982). The locative elicitation study is reported by Johnston and Slobin (1979); the causative study by A m m o n and Slobin (1979); the agent-patient (word-order vs. inflection) test by Slobin and Bever (1982), and Slobin (1982). Some aspects o f input are summarized i n Slobin (1975). A k s u (1975) examines the development o f the expression o f cause-effect relations. Clancy, Jacobsen, and Silva (1976) compare the acquisition o f conjunctions i n Turkish, Italian, English, and German. Slobin and Talay ( i n press) study the development o f pragmatic uses o f subject pronouns. Aksu's (1978) dissertation is a detailed examination o f aspect and modality i n the child's acquisition o f the Turkish past tense. The order o f emergence o f past tense forms for direct and indirect experience is traced longitudinally across three children between the ages o f 21 and 30 months, and experimental data from children between 3;0 and 6;4 are analyzed i n regard to issues o f aspect and modality. S a v a § i r ' s (1983) Master's Thesis explores the use o f various forms o f future expression i n three 2-year-olds from the Berkeley sample. A l l o f the data listed above were consulted i n the preparation o f this chapter. Crucial gaps exist i n several places—especially i n regard to very early stages and the emergence o f the inflectional system, and the nature o f input i n natural adultchild and child-child interaction. W e also k n o w little about the development o f discourse skills. (Iskender Sava§ir [1983] is beginning to study developmental relations between activity types and verb forms at Berkeley.) There are no

Behavior Research Laboratory, University of California, Berkeley. Computer facilities in the Berke­ ley Child Language Archive have been provided by the Sloan Foundation and NSF, allowing for continuing analysis of Turkish speech data. Ayhan Aksu, Francesco Antinucci, Thomas G . Bever, Eve V . Clark, Herbert H . Clark, Susan Ervin-Tripp, Judith R. Johnston, and Ljubica Radulovic collaborated with Dan I. Slobin in designing the investigation. Ayla Algar and Alev Alath served as testers in Istanbul and Ankara. Aksu's dissertation research received partial support from the Ameri­ can Research Institute in Turkey. The Max-Planck-Institut fur Psycholinguistik, Nijmegen, The Netherlands, provided support in the summer of 1981 for further analysis of our Turkish child speech data, as well as a collégial setting in which to work on the revised draft of this chapter. We wish to acknowledge the assistance of all of these institutions and individuals in contributing to the data summarized here. We have benefited from discussion with many colleagues and friends, and would like to thank, especially, Mary Sue Ammon, Francesco Antinucci, Judith Johnston, Brian MacWhinney, Ruth Miller, Nail §ahin, Iskender Sava§ir, Bambi Schieffelin, and Karl Zimmer.

9.

The Acquisition of Turkish

845

systematic observations o f the development o f prosody or phonology i n Turkish. There are few data on later ages, but N a i l § a h i n at M i d d l e East Technical University i n Ankara is making an inventory o f syntactic structures and lexicon of school-age children.

3. B r i e f S u m m a r y o f O v e r a l l C o u r s e o f L i n g u i s t i c D e v e l o p m e n t in T u r k i s h The inflectional system appears early, and the entire set o f noun inflections and much o f the verbal paradigm is mastered by 24 months o f age or earlier. B y this age, Turkish children inflect nouns for case (accusative, dative, ablative, pos­ sessive, instrumental) and number (plural), and verbs for tense-aspect (past result, ongoing process, intention), person, negation, and interrogation. Both noun and verb inflections are present i n the one-word stage, and there is some evidence for productive use as young as 15 months. Early words typically i n ­ clude inflections, since stems tend to be mono- or bisyllabic, and inflections are stressed suffixes. For example, bitti 'all-gone' or ' a l l done' is made up o f the verb stem bit ' f i n i s h ' and the past-tense suffix -ti; negative imperatives are composed o f the verb stem and the suffix -mE, as i n gitme ' d o n ' t g o ' ; and so forth. Such suffixes are quickly used w i t h a wide variety o f words. Evidence for productivity appears as early as age 15 months, when children fail to delete a stem-final -k before a suffix beginning w i t h a v o w e l , thus indicating that the suffix is not simply part o f a rote-learned amalgam. There are no errors i n the order o f agglutinated morphemes; however the very few available possibilities for morphological over-generalizations are followed by the child. Overall, mor­ phological errors are remarkably rare, because the extreme regularity o f the language precludes them. Morphophonological adjustments for vowel harmony and voicing assimilation are also correct at very early ages. As a result o f precocious acquisition o f grammatical morphology and lack o f overgeneralizations, Turkish child speech transcripts do not have the familiar " c h i l d language" look evidenced i n most other languages. Early utterances are not telegraphic, since the stressed, suffixed inflections are present. C h i l d utter­ ances are short and simple, but rarely ungrammatical or incomplete from the point o f view o f the adult language. Thus Turkish acquisition provides evidence that grammatically relevant notions are accessible to quite young children i f the means o f expression are sufficiently salient and analyzable. W o r d order is used flexibly for pragmatic functions, as i n the adult language. For example, children younger than 2 correctly place new information before the verb and presupposed or predictable information after the verb. A l l six orders o f subject, verb, and object are comprehended i n a reversible transitive sentence test at 24 months. Fixed w o r d order is adhered to where required i n the language (e.g. adjective + noun; noun + attribute; possessor + possessed).

846

Aksu-Kog and S l o b i n

T w o - and 3-year-olds are engaged i n mastering verb inflections for voice and modality (passive, nonwitnessed past tense, conditional, causative) and syntactic means for temporal and causal l i n k i n g o f clauses. The tense-aspect system evolves from (1) a distinction between immediately com­ pleted changes of state (-di) and durational events (-lyor), to (2) generalization of the meaning of the past-tense morpheme -di from completion to past tense, to (3) a distinction in the past between directly experienced events (-di) and events inferred from their endstates (-ml§), and finally to (4) a general past-tense distinction between witnessed (-di) and nonwitnessed (-ml§) events, including hearsay in the latter category. The passive, which is agentless in Turkish, emerges early to focus on desired changes of state in objects. For example, a child of 2;4, wanting to open a box, said: Boyle ag-d-ir 'thus open-PASS-AORIST' (='It is opened like this'). The conditional is a simple verb affix, and is used by 2-year-olds to express contingencies. For example, a child of 2;7, covering her doll's eyes, said: Karanhk ol-sa gor-mez-sin 'dark be-COND see-NEG:AOR-2SG' (= 'If it's dark you won't see'). (Counterfactual uses of the conditional are a much later acquisition.) The causative morpheme is often extended from the meaning of instigation ('make someone do something') to causation generally, with inappropriate affixa­ tion to verbs that are already inherently causative. For example, a child of 2;3 intended to use the verb kes 'cut' as a simple transitive, using the causative morpheme, -tir-, to mark transitivity; however, the form kes-tir should mean 'have someone cut'. Thus although the general notion of 'cause' is marked early on, it takes a while for children to sort out different types of causality for grammatical marking. Locative postpositions, question words, and temporal and causal clause rela­ tions emerge i n the standard crosslinguistic order, presumably based on concep­ tual development. Locatives emerge i n the order: ' i n / o n 7 ' u n d e r ' < 'beside' < 'back w i t h objects having back-front orientation' < 'front w i t h oriented objects' < 'between' < 'back withunorientedobjects' < 'front w i t h u n o r i e n t e d o b j e c t s ' . Early clausal conjoining expresses temporal and causal sequences; simultaneity and directed temporal relations ('before', 'after') develop later. 4

Late acquisitions (after age 4) are seen i n a variety o f complex constructions requiring the insertion o f nominalized verb forms o f various sorts into sentences (relative clauses and verb complements). Whereas the grammar o f simple sen­ tences and main clauses is quite transparent and easily acquired, the syntax and morphology o f subordinate clauses pose considerable difficulty to the Turkish child. Means o f combining clauses to express temporal and causal relations develop i n the f o l l o w i n g sequence: (1) U n t i l about 2;6, simple juxtaposition o f sentences predominates, without explicit grammatical markers o f connection. (2) During the next year children begin to use connectives that don't require n o m i nalizations (conjunctions and con verbs). (3) After age 4 children begin to more

9.

The Acquisition of Turkish

847

frequently use nominalizations for various subordinate clauses, but w i t h pro­ longed confusion between the various forms, and syntactic errors. Systematic data are not available beyond age 6.

THE D A T A 4. T y p i c a l E r r o r s The discussion o f typical morphological errors i n this section is briefer than comparable chapter sections on the development o f other languages, since the remarkable regularity and transparency o f Turkish morphology precludes a high rate o f error i n the early phases o f development. Where errors typically occur is in later phases, when the Turkish-speaking child encounters problems o f com­ plex syntax, as discussed i n regard to nominalization errors and errors i n dever­ bal and denominal derivation, and late acquisition o f relative clauses. 4 . 1 . Morphological

Overregularizations

The few possible morphological overgeneralizations do occur. One source o f error is failure to delete a stem-final k preceding suffixes beginning w i t h a v o w e l . For example, the corrective accusative o f tabak 'plate' is tabagi, where g is not pronounced. Ekmek^i's child practiced this accusative form at 15 months: daba, daba, daba, dabagi. Similarly, at the same age, she rendered the genitive o f bebek 'baby' as bebeki (—bebegiri). These are common errors for this class o f words, rarely heard after age 28 months. Such analogical errors (compare the correct early accusative, as, for example, Ahmedi), so abundant i n Indo-European child language, also occur where possible i n Turkish. They indicate that the child has carried out a segmentation o f root and affix, and suggest that the early correct forms are probably also productive. The common explanation is applicable to the Turkish c h i l d — n a m e l y that the child is sensitive to patterned regularities i n morphological paradigms, and applies a standard inflectional principle to all relevant members o f a class. 4 . 2 . Meaning/ess

Overmarking

of

Verbs

A t early stages o f development (below age 2;6 or so), verbs are sometimes pronounced w i t h extra, meaningless syllables between the stem and the final person-number affixes. Early o n , i t seems that the child attempts to retain some rhythmic picture o f complex verbs, incomprehendingly inserting morphemes that sound like passive and causative particles. Such errors are quite different from the errors considered below, where the added morphemes perform an interpreta¬ ble semantic function i n context. Whereas later errors o f overmarking reveal a semantically motivated analysis o f the morphological system, the early errors

848

Aksu-Kog a n d S l o b i n

seem to reveal a semantically unmotivated analysis o f words into combinable syllables—an obvious prerequisite to the discovery o f principles o f productive morphology. 4.3. Errors

in Causative

Marking

The causative morpheme can be inserted i n intransitive verbs to make them transitive, as i n dl ' d i e ' and dl-dür ' k i l l ' . W h e n this morpheme occurs w i t h a verb which is already transitive, it renders an instigative meaning as i n kir 'break' and kir-dir 'have (someone) break'. The morpheme can be doubled i n verbs o f the first type, bringing about an instigative meaning ('have (someone) k i l l ' ) . Some children apparently abstract a general causative meaning from this morpheme, using it incorrectly w i t h verbs which are already inherently causative-transitive. For example: Adult:

Kim who

kes cut

-ti PAST

onu? it+ACC

'Who cut it?' Child (2;3): Ben I

kes cut

-tir CAUS

-di -m. PAST 1 S G

'Intended meaning: T cut (it).' Literal meaning: T had (someone) cut (it).' Children also have difficulty i n determining w h i c h verbs have lexical causative forms and which verbs allow for productive causative derivation. (The latter option is much more widely represented i n the language.) For example, a girl o f 3, pointing to her hurt eye, said: *Bu this

-ra -si L O C POSS

-ni yan A C C burn

-dir CAUS

-lyor PROG

'It's making this point burn (hurt).' Here, the intransitive verb yan ' b u r n ' has been transitivized w i t h the causative particle -dir, whereas i t has a lexical causative counterpart yak 'cause to b u r n ' . (See Clark & Hecht, 1982, for children's problems i n distinguishing between conventional and productive parts o f the lexicon.) The other side o f the coin is represented by errors o f undermarking the causative, treating an intransitive verb as i f i t were lexically transitive, as shown in the following interchange, which arose when a child wanted the experimenter to remove a small plastic toy pasted to a cardboard mounting:

9. Child:

*§u that

The Acquisition of Turkish

849

-nu kalk. A C C get: up

Adult: Efendim? Excuse me? Child:

*§u that

-nu kalk A C C get:up

-sana. IMP

The child's intended meaning was ' l i f t that up' (disconnect i t ) , but he used the intransitive verb kalk 'get u p ' . The grammatical form requires the causative particle:

kal-dir-sana.

4 . 4 . Simplification

of the Negative

System

Some children younger than 2 have been observed to make various attempts to simplify the system o f negation. Verbal predicates are negated by the insertion o f the negative particle -mE- immediately after the verb root or verb + voice particles, but before the modal, tense, and person suffixes. Stress is shifted to the syllable preceding the negative particle (a clear exception to the usual pattern o f word-final stress); for example, al-di 'take-PAST:3SG'—dl-ma-di 'take-NEGP A S T : 3 S G ' . Negation o f nonverbal predicates, on the other hand, involves the uses o f lexical negatives yok 'nonexistent' and degil 'is not'. Degil is the nega­ tive for substantive (adjectival and nominal) predicates, while yok negates exis­ tential predicates. For example: su degil 'water N E G : B E ' ( = ' i t isn't water') vs. su yok 'water N E G : E X I S ' ( = 'there is no w a t e r ' ) . The negative existential, yok, is also used for negative possessives, contrasting w i t h the affirmative existential, var. Compare adam-in at-i var ' m a n - G E N horse-POSS E X I S ' ( = 'the man has a horse') and adam-in at-i yok ' m a n - G E N horse-POSS N E G : E X I S ' ( = ' t h e man doesn't have a horse'), along w i t h the contrasting substantive negative adorn at degil 'man horse N E G : B E ' ( = 'a man is not a horse'). (These examples are only relevant to timeless or nontensed statements. Both types o f nonverbal predicates, substantive and existential, appear as the negation o f the auxiliary verb ol 'be' when they take tense-aspect-modality markers and when they are embedded.) It appears that some children at first pick only one o f these lexical forms, or a phonological variant o f -mE- (mi mi), or i-ih (the sound accompanying gestural negation i n Turkish) as a universal negative marker, applying i t to all types o f predication after the model for negation o f nonverbal predicates. Thus the nega­ tion paradigm o f nonverbal predicates is overgeneralized to verbal predicates. (Note that thereby the child adheres to a pattern o f negation that is more wide­ spread among the languages o f the w o r l d , as w e l l as being consistent w i t h an early tendency i n child speech for sentence external negation.) One child is reported to have used degil as a general negative marker until age 2;3, and

850

Aksu-Koc. and S l o b i n

another to have used ih until about 2;6. Another child was observed to use mi mi as a negator until age 2;0. To cite examples from t w o o f these children: *Anne mother

otur, sit

kalk degil. get: up N E G

'Mother sit, don't get up' (= kalk-ma 'get:up-NEG'). *Yap do

-wag FUT

-an ih. 1SG N E G

T won't do (it)' (= yap-mi-yacag-im 'do-NEG-FUT-lSG'). The occasional carryover o f this early tendency is exemplified below i n the speech o f one child w h o d i d produce inflectional negation. *Ay, oh

koy put

-du PAST

Intended meaning: Adult:

-m yok. 1SG N E G T can't put it.' (= koy-ma-dim)

Buskui

ver

-eyim

mi

sana?

cookie

give

0PT:1SG

Q

2SG+DAT

'Shall I give you a cookie?' Child:

*Buskui yok. cookie NEG 'Intended meaning: 'Don't give me cookies.' (= Biiskiii ver-me) Literal meaning: 'There are no cookies.'

Although systematic longitudinal data from earlier periods are needed to understand the exact course o f such developments, i t appears that the paradigm for the negation o f nonverbal predicates is acquired earlier than that for verbal predicates. The early isolation o f lexical negatives is probably facilitated by the fact that these forms are perceptually salient, independent words w h i c h occur i n sentence final position. I n addition, their use avoids problems o f interruption o f linguistic units called for i n the insertion o f the inflectional negative -mE- w i t h i n the verb. W h e n the latter form emerges, one often hears laboriously slow and clear syllable-by-syllable enuncation o f negated verbs, w i t h unusual stress on the usually unstressed negative particle. A l t h o u g h accomplished fairly early, analy­ sis and composition o f inflected negative forms is not carried out without difficulty. 4.5. Nominalization

Errors

Errors are made i n various participial forms from age 3 on, increasing i n frequency for 4- and 5-year-olds, as occasions to use such forms become more common, given increased complexity i n the child's communicative intent. The

9.

The A c q u i s i t i o n of Turkish

851

means o f deriving such forms are complex, based on principles probably rela­ tively abstract to the c h i l d , and often resulting i n surface forms i n which a clause is conflated i n a nominalized verb. For example, statements o f fact and state­ ments o f potentiality or activity require different nominalizations, using the participle -dlk for factive clauses and -mE for potential clauses, as shown below: Ahmed Ahmet

-in GEN

ytiz swim

-dug FACT

-il POSS

T know that Ahmet swam/is swimming.' Ahmed Ahmet

-in yiiz G E N swim

-me -si POT POSS

T'm waiting for Ahmet to swim.'

-nu ACC

hiliyorum. know + l S G .

[fact]

-ni bekliyorum. A C C wait+lSG [potentiality]

Children tend to use the latter form for both types o f constructions. I t may be that the distinction between fact and potentiality is not accessible to children o f this age. I n addition, the use o f a single form for these t w o closely-related comple­ ment constructions may reflect a common tendency o f economy o f means. B u t , in either case, w h y should the latter form, the -mE participle, be preferred? The avoidance o f the -dlk form may be due to morphological complexity. Note that i n the above t w o examples the subject o f the embedded clause, Ahmet, is i n the genitive case, Ahmedin. The subject is thus treated as the " s y m b o l i c possessor" of the action attributed to h i m (literally ' A h m e t ' s fact-of-swimming I k n o w ' and 'Ahmet's potentiality-of-swimming I a w a i t ' ) . The form is therefore doubly opaque, i n that the subject is marked i n a nonstandard way, and the verb has lost its characteristic verbal morphology. However, when the subject o f the main clause is coreferential w i t h the subject o f the embedded clause the genitive is not required, allowing for the syntactically simpler forms: Ahmet Ahmet

yiiz swim

-me -si POT POSS

-ni biliyor. A C C know+3SG

Ahmet Ahmet

yiiz swim

-me -yi biliyor. POT A C C know+3SG.

'Ahmet knows how to swim.' This simpler nominal complement is acquired earlier than the t w o more complex forms presented above. When children begin to use the factive nominalization, they often tend to mention the embedded subject i n the standard nominative (unmarked) case, as i n the coreferential situation, rather than i n the required genitive. As a general principle, there is probably a tendency to l i m i t the range o f functions carried out by a given case inflection, resulting here i n the avoidance o f using the genitive—normally the case o f the possessor—to mark a subject. The

852

Aksu-Kog a n d Slobln

simpler option, thus, is to use the -mE participle w i t h embedded nominative subject for both factives and potentials, and, upon emergence o f the -dlk participle, still to avoid genitive marking o f the embedded subject. 4.6. Errors

in Deverbal

Derivation

Other participial errors occur i n the formation o f deverbal attributions— another set i n which similar forms are distinguished by fairly abstract formal criteria. The details are arcane for the general reader, but the general interpreta­ tion is that i t takes children a while to sort out nuances o f verbal aspect and a range o f surface forms w h i c h all serve to convert actions into states (e.g. forms expressing such notions as 'sleeping cat', 'sliced meat', 'fallen l e a f ) . The range of forms incorrectly chosen i n particular instances all come from a class o f verbal affixes w h i c h depict an event i n a stative or timeless mode, but the particular choice is often incorrect. Several types o f examples are given below. I n all o f these situations semantic motivation seems to take precedence over the acquisi­ tion o f more formally motivated means o f expression. Aksu (1978) has experimentally studied t w o such affixes that form deverbal attributives, the present participle suffix -En and the past participle suffix -ml§. I n describing stative representations o f actions, 3-year-olds prefer either verbal predications i n the present progressive or substantive and existential predica­ tions. Attributive descriptions w i t h -En and -ml§ show a significant increase after age 4, w i t h constructions w i t h -En reaching a comparable level o f production about six months later than -ml§. The past participle, -ml§, is used to encode resultant states, such as pi§-mi§ elma 'cook-mis apple' ( = 'cooked apple'), whereas the present participle, -En, stativizes processes, such as kos-an gocuk 'run-En c h i l d ' ( = 'running c h i l d ' / ' c h i l d who is r u n n i n g ' ) . As discussed i n Section 5.3, below, the -ml§ suffix is acquired earlier as a verbal inflection referring to present states resulting from past processes. A t the point o f acquiring the syntactic function o f forming participial adjectives, the child may begin w i t h -ml§, extending this already established form to a new function. Once this syntactic function has been established for past participials, it may be easier for the child to acquire a new form, -En, for deriving present participials. Here we may have a realization on the syntactic plane o f the principle: " N e w forms first express o l d functions, and new functions are first expressed by o l d f o r m s . " Most common errors up to age 4 are attributives derived w i t h the future participle -EcEk, the infinitive suffix -mEk, and the nominalizing suffix -mE, i n contexts where the stative -ml§ participle w o u l d be appropriate. The f o l l o w i n g examples are ungrammatical although the choice o f -mE instead o f -ml§ is appropriate on semantic grounds since both suffixes derive adjectives w i t h pas­ sive meaning. Typical responses i n describing a picture o f a cut or bitten apple were:

9. *kes cut

-il PASS

-me NOMINAL

elma apple

*mr bite

-il PASS

-ma NOMINAL

elma apple

The Acquisition of Turkish

853

In both cases the passive notion has been marked twice, i.e., redundantly. The grammatically correct description of a picture of a cut apple would be: kes cut

-il PASS

-mis PAST. PART

elma apple

Using the -mE nominal, without the passive particle -//-, results in grammatically correct but frozen forms expressing potentiality: yar split

-ma NOMINAL

seftali peach

'peach that can be split' These errors suggest some uncertainty in the range of meanings carried by the -mE nominal and the passive in situations where states are to be described with forms based on verbal stems. Difficulty with such notions is also revealed in another typical error of deriva­ tional morphology—in this case, undermarking of the passive. Children often neglect to passivize transitive verbs in forming present participial adjectives. The result is an ungrammatical active participle modifying an inanimate patient noun, as in: *isir bite

-an PRES. PART

elma apple

'apple that is biting' *yi eat

-yen PRES. PART

elma apple

'apple that is eating' Grammatical versions require the passive particle: isir bite

-il PASS

'bitten apple'

-an PRES. PART

elma apple

854

ye eat

Aksu-Kog and Slobin

-nil PASS

-en PRES.PART

elma apple

'eaten apple' These errors may be due to the taxing operations o f morphological derivation. They may also be due to an insufficient analysis o f certain predicates i n terms o f transitivity/intransitivity, as is also suggested by errors o f overcausativization discussed above. The facts that errors observed i n morphological derivation are the same as those i n inflectional derivation strengthens the second interpretation. 4 . 7 . Errors

in Denominal

Derivation

Occasional errors occur i n the formation o f denominal attributives w i t h -//, a highly productive suffix denoting 'the possession o f the object or quality i n d i ­ cated by the n o u n ' . The f o l l o w i n g example is an error because it makes no sense semantically, but on the other hand it reflects the underlying knowledge that -// derives adjectives from nouns, since the child has correctly nominalized the verb. I n attempting to describe a picture o f an apple from w h i c h a bite has been taken, the child has added -mE to the verb stem, thus nominalizing i t , followed by the adjectival derivation -II: *ye eat

-me NOMINAL

-li ADJECTIVAL

elma apple

5. E r r o r - F r e e A c q u i s i t i o n 5.1.

Morphology

W i t h the exception o f the marginal early and late errors summarized above, Turkish child speech is almost entirely free o f error. This is undoubtedly at­ tributable to the extreme regularity o f the morphological systems, resulting i n a situation i n w h i c h the language hardly provides opportunity for error. M o s t o f the agglutinative m o r p h o l o g y — n o m i n a l and verbal—is used productively at the two-word period, before the age o f 2. There are many interlocking reasons for the ease o f acquisition o f these systems. W e can think o f at least 12 factors w h i c h play a role i n facilitating acquisition, and they cannot all be pulled apart in considering any single lan­ guage. The morphemes are: (1) postposed, (2) syllabic, and (3) stressed, making them especially salient to perception and immediate memory. They are (4) oblig­ atory, rather than optional. (For example, the optionality o f the Japanese object particle delays its acquisition and its use i n sentence comprehension, i n relation to Turkish [Hakuta, 1977; Slobin, 1982].) (5) The inflections are tied to the content w o r d , noun or verb, and are not conflated w i t h other parts o f speech.

9.

The A c q u i s i t i o n o f Turkish

855

(For example, the German case system, conflated w i t h articles, is acquired much more slowly than the corresponding Slavic noun suffixes.) (6) The postposing o f inflections is consistent w i t h the verb-final typology o f Turkish, and children may be sensitive to such typological consistencies. Semantically, the Turkish particles (7) seem to follow the stem i n an order reflecting decreasing relevance (Bybee, 1985) to the inherent meaning o f the stem ( N O U N - P L U R A L - P O S ­ SESSIVE-CASE and V E R B - M O D A L I T Y - T E N S E / A S P E C T - P E R S O N / N U M ­ BER), and (8) are generally nonsynthetic i n their mapping o f functions onto form. Clairty o f semantic mapping probably facilitates acquisition. ( B y com­ parison, the fusional quality o f Indo-European inflectional morphemes probably adds to their complexity; e.g. the typical conflation o f number, gender, and case.) (9) Functionally, the morphemes express only grammatical roles, while other devices, such as contrastive w o r d orders and focusing particles, are used for pragmatic functions. ( B y comparison, Japanese particles, w h i c h express both pragmatic and syntactic functions, seem to be more difficult to master.) I n terms of distribution and diversity, the paradigms are (10) almost entirely regular (i.e. exceptionless), (11) consistently applied to all content words and pro-forms (nouns and pronouns, demonstratives, question words, nominalizations; main verbs and auxiliaries), and (12) relatively distinct (i.e. there are almost no homo­ nymous functors). Whatever the relative strengths o f all o f these factors, i t is clear that at least some o f them greatly facilitate the acquisition o f inflections i n Turkish. 5.2.

Morpheme

and Word

Order

Order rules, where they apply, are also acquired free o f error. The aggluti­ nated morphemes occur i n proper order. This seems remarkable given the range and complexity o f possible combinations i n both the nominal and verbal systems, as summarized i n the grammatical sketch above. I t is not uncommon, for exam­ ple, to find verbs i n 3-year-old speech containing particles o f negation, voice, modality, tense-aspect, and person-number—all i n the proper order. Obligatory w o r d order w i t h i n clauses is adhered to (e.g. adjective-noun, possessor-pos­ sessed, noun-predicative attribute, postposing o f postpositions). On the other hand, varying w o r d order w i t h i n sentences for pragmatic effect is also easily mastered (e.g. using pre verbal position for focus, resulting i n postposed subjects for object focusing and postposed objects for subject focusing). There appears to be a sensitivity to the role o f ordering at different linguistic levels. A t the w o r d level, where morphemes are bound by v o w e l harmony and w o r d intonation, strict ordering is adhered to. Strict ordering is also adhered to w i t h i n constituent phrases. A t the sentence level, however, ordering o f main sentence elements is free to vary i n appropriate discourse contexts, as discussed i n Section 6.1.2, below. General principles like ordering, therefore, do not apply across the board, but interact w i t h definition o f linguistic level.

856

Aksu-Koc. and Slobin

6. T i m i n g o f A c q u i s i t i o n 6.1. Early

Acquisition

6 . 1 . 1 . Inflections. The agglutinative morphology is acquired strikingly ear­ l y , parts o f it apparently productive as early as 15 months. Explanations come from the consistency o f the system, as discussed above. I n sentence comprehension experiments (Slobin & Bever, 1982), children rely on the accusative inflection, rather than w o r d order, to identify agent and patient. F r o m the youngest age group tested—2;0—children correctly acted out reversible transitive sentences i n all six orders o f subject, verb, and object. Slobin and Bever also presented children w i t h strings o f t w o nouns and a verb, w i t h no case inflection on either noun, i n the orders N N V , N V N , and V N N . B y and large, response to such strings was random, indicating that Turkish children rely on inflections rather than w o r d order for the identification o f grammatical relations. 6.1.2. Word Order. Ekmekgi ( i n press) has documented early use o f a variety o f w o r d orders i n one child during the age range 1 ; 7 - 2 ; 4 . Early control o f the functions o f w o r d order is reflected i n a number o f contrastive uses, including the following: (1) Preposed adjectives are used i n attributive expressions (e.g. soguk su 'cold water', said at 1;7 when asking for cold water), whereas postposed adjectives are used i n predicative expressions (e.g. gorba sicak 'soup hot', said at 2;0 as a complaint). (2) Indefinite or nonreferential direct objects always directly precede the verb (e.g. kalem getir 'bring (a) p e n c i l ' ) , whereas definite direct objects (marked by the accusative inflection) can also follow the verb (e.g. both kalem-i getir ' p e n c i l - A C C b r i n g " and getir kalem-i 'bring pen­ c i l - A C C = 'bring the pencil') [age 1; 10]. (3) Such examples reflect a more general tendency to place new information before the verb and presupposed or predictable information after the verb. Consider, for example, the f o l l o w i n g narrative sequence, i n w h i c h 'three sisters' are introduced as a sentence-initial topic i n the first sentence, followed by postposing o f 'sister' i n a f o l l o w i n g sentence: Ug three

kardes sister

var EXIS

-mi§. PAST: REPORT

'There were three sisters.' Bir one

-i POSS

buytig -umus big PAST: REPORT

kardes -in. sister G E N

'One of the sisters was big.' The verb can be highlighted by verb initialization and postposing o f all additional material, as i n the f o l l o w i n g (2;0):

9. Sev

-mi

love

NEG FUT

-eceg

-im

onu

1SG 3SG:ACC

The Acquisition of Turkish

857

daha.

more

T won't love her anymore.' I n reading through our o w n transcripts o f Turkish child speech we have been struck by the extreme rarity o f contextually inappropriate w o r d orders, reinforc­ ing the impression that pragmatic variation i n w o r d order is a precocious acquisi­ tion. (Support is given by discourse constraints o f adult-child conversation, as pointed out i n the discussion o f input, below.) W o r k by Slobin and Talay ( i n press) on children between 2;1 and 3;8 shows early control o f the pragmatics o f first-person pronoun placement. For example, utterances w i t h T i n normal pre verbal position are used when the subject is at issue, as i n responses to who'-questions or i n drawing a contrast w i t h the subject of a previous utterance, as w e l l as i n making a neutral report o f past action. B y contrast, postverbal T is used when the verb or object is i n focus, and the continuing subject is presupposed. Such uses o f subject pronouns are standard for adult conversation as w e l l ; however, where young children may differ from adults is i n overuse o f pronouns. This issue is taken up i n more detail below, i n the discussion o f reorganizations i n development (Section 11.4). 4

I n the Slobin and Bever (1982) sentence comprehension study reported above, there is suggestive evidence that children are sensitive to the fact that S O V is the standard w o r d order o f the language. Although most children responded ran­ domly to strings containing t w o uninflected nouns and a verb, a small number o f children d i d respond consistently to some o f these strings, picking either the first or second noun as agent. Their responses reflected the frequency o f occurrence of w o r d order types i n Turkish speech: 13 children responded consistently to N N V strings, w h i c h parallel the standard S O V order; 7 were consistent on N V N , which is the next most frequent order i n speech samples; and only 4 were consistent on V N N , w h i c h is the least frequent order. Children's imitations o f sentences i n the six orders o f S, V , and O showed both great tolerance for w o r d order variability and a degree o f sensitivity to the standard S O V order. Overall, children tended to imitate sentences correctly, regardless o f w o r d order. A l ­ though reorderings i n imitation were rare, when they d i d occur they again re­ flected a sensitivity to the frequency o f occurrence o f the order types i n speech: (1) Verb-final sentences were almost never reordered; (2) verb-medial ( N V N ) sentences were reordered less frequently than verb-initial ( V N N ) sentences, and always into verb-final order ( N N V ) ; (3) verb initial sentences were reordered most frequently—generally into N N V order, but also into N V N . Younger c h i l ­ dren reordered most frequently (from 4 6 % at age 3;0 to 1 1 % by 3;8) and made more conversions into verb-final order i n their imitations. Thus even though Turkish children vary w o r d order freely i n their speech, and are guided by inflectional cues i n sentence comprehension, they are also cognizant o f the basic verb-final character o f their language.

858

Aksu-Kog and Slobin

6.1.3. Passive. The passive is an early acquisition i n Turkish, where i t is always agentless, and is indicated by a simple and regular verbal affix, as are other parts o f the verbal system discussed above. The pragmatic function that it serves is one o f focusing on the state o f the patient, such as the English truncated passive (e.g. it got broken). The Turkish child does not have to face the complex­ ities o f forms like the full passive o f English, since variations i n w o r d order, as discussed above, allow for shifts i n focus. Sava§ir (1983) has examined uses o f the passive i n three children from the Berkeley sample i n the age range 2 ; 3 - 2 ; 4 . He notes a preponderance o f early use o f passives i n the negative, i n situations i n which the child fails to bring about a desired goal (e.g. ag-il-mi-yor 'open-PASSN E G - P R O G ' ( = it is not being opened', uttered by a child o f 2 ;4 who has not succeeded i n opening a box). Sava§ir suggests that the passive is used i n third person present negative utterances, such as this one, to attribute a nonoccurrence of an event to an object: " I t w o u l d seem that the arliest occurrences o f the passive i n the present tense are used to report those instances i n which the child's intentions or plans are inhibited due to a resistance from an object" (p. 39). He suggests that the next developmental advance involves a distinction between focus on the agent and a non-agentive focus on objects, as revealed i n alterna­ tions between first person active and third person passive utterances, such as boyle ag-ar-im 'thus open-AORIST-ISG ( = T open (it) like this')and boyle ag-ij -ir 'thus open-PASS-AORIST' ( = ' I t is open like this'). I n S a v a § i r ' s terms, children come to "use the passive voice to represent the changes that arise out o f the properties o f the object" (pp. 3 9 - 4 0 ) . (He notes, as w e l l , that the aorist, i n distinction to the present progressive, is used i n such situations w h i c h are viewed " n o n - a g e n t i v e l y . " Thus the combination o f aorist and passive allows children to talk about events per se.) 4

6.2. Late

Acquisition

Relative clauses, verb complements, and some types o f conjoined construc­ tions are strikingly late, not fully mastered until age 5 or later. These are all constructions i n w h i c h an embedded sentence is treated as a participle, as dis­ cussed above. Such clauses are relatively opaque—that is, they do not look like surface sentences, since the verbs appear i n various nominal or participial forms and casemarking on nouns is often different than i n the corresponding simple clauses. I n all o f the forms discussed below there is abundant morphological and syntactic complexity. Such constructions, w h i c h increase the distance between surface form and underlying meaning, pose special acquisitional problems to children. 6 . 2 . 1 . Causal conjunction. The order o f acquisition o f the various syntactic structures for conjoining t w o clauses i n a causal relation presents a control case, since the semantic complexity o f the underlying notions can be assumed to be constant across various forms o f conjunction. The connectives constituted o f a

9.

The A c q u i s i t i o n of Turkish

859

demonstrative pronoun i n the ablative case, on-dan ' t h a t - A B L ' ( = ' f r o m that'), or i n the genitive case followed by a postposition, on-un igin 'that-GEN for' ( = ' f o r that reason'), are quite transparent and are acquired first. These forms allow for conjoining o f independent clauses, as i n the f o l l o w i n g example from a child o f 2;8: Oyuncaklari toys+ACC

gotiirdun you+brought

de ONUN PTL

IQIN

kizdi. got:angry+3SG

'You brought the toys, THAT'S W H Y she got angry.' On the other hand, participial or infinitival nominalizations o f the verb followed by a postposition, -dig-I igin ' N O M L - G E N for' ( = 'because o f his . . . i n g ' ) , and -mEk igin ' I N F for' ( = ' i n order t o ' ) , emerge much later i n the expression o f the same causal relation. I n such constructions there is a clear distinction be­ tween main and subordinate clause, and only the main clause has a finite verb, as in the f o l l o w i n g example from a child o f 3;8, representing the earliest appearance of such constructions i n a cross-sectional sample from 2;1 to 5;0: Oyuncak toy

ol be

-dug PART

-u igin ugmaz. G E N for fly+NEG+AORIST

'Because it's a toy it doesn't fly.' (The

order o f development o f conjunction types is discussed i n Section 7.2,

below.) 6.2.2. Relative Clauses.

Difficulties i n dealing w i t h nontransparent embed­

ded clauses are most evident i n the acquisition o f relative clauses, clearly a very late acquisition for Turkish children. Slobin ( i n press) has compared the use o f relative clauses i n 57 matched English and Turkish child speech samples between the ages o f 1;0 and 4;8, extracting all o f the relative clauses spoken by the children and the adult investigators w h o interacted w i t h them i n standard settings (playing w i t h toys and discussing everyday events). I n both languages, relative clauses are quite rare, and none are found before age 2;4; however, they are twice as frequent i n English as i n T u r k i s h . Overall, there are 96 relative clauses in the English transcripts and only 42 i n Turkish (based on over 40 hours o f interaction). The same asymmetry is reflected i n the investigators' speech to the children, w i t h 40 relative clauses i n English and 22 i n Turkish. As shown i n F i g . 9 . 1 , the development o f relative clauses is much faster i n English, w i t h a major spurt around age 3;6, w h i l e the mastery o f relative clauses i n Turkish must take place later than 4;8, the oldest age i n our sample. (Similar evidence comes from experiments on the comprehension o f relative clauses, discussed i n Section 12.1.3, below. Turkish children o f 4;8 are able to extract only the main clause for acting out.)

860

Aksu-Kog and Slobin 50

E N G L S I H

T U R K S I H

2 y r .

3 y r . o l d s

o l d s

FIG. 9.1. Growth of Relative Clauses in English and Turkish Child Speech

o l d s

The difficulty i n Turkish relative clauses lies i n the deformation o f the embed­ ded clause, w h i c h loses the finite verb and normal case inflections o f a canonical main clause, as pointed out i n the grammatical sketch, above. I n addition, different participles are required for subject and nonsubject relatives (details i n Slobin, i n press). Consider, for example, the f o l l o w i n g Turkish utterances from our sample o f adult speech to children, along w i t h their English equivalents i n the glosses: Bizim our

ev house

-in onun G E N front

-e gel D A T come

-en SUBJ.REL

kedi -ye cat D A T

benziyor, resembles Tt looks like the cat that comes to the front of our house.' Kazan earn

-dig OBJ.REL

-in POSS.2SG

para money

-yla with

ne yapacan? what do+FUT+2SG

'What will you do with the money that you earn?' The Turkish sentences have nonfinite verbs, w i t h different participles for subject and object relatives. I n addition, i n the first example, normal subject-verb order is reversed: gelen kedi 'cat c o m e + S U B J . R E L ' ( = ' c a t that comes') as opposed to canonical kedi geliyor ' c o m e + P R E S cat' ('cat comes'). The equivalent En­ glish form is maximally transparent: the cat that comes, corresponding to canoncial the cat comes. The second example, an object relative, condenses subject and transitive verb into a single w o r d i n Turkish, the possessed nominalization kazandigin, i n comparison w i t h the full you earn i n English. Small wonder that relative clauses are a late acquisition i n Turkish. They are replaced i n discourse by periphrastic circumlocutions w h i c h serve to establish common reference. Typically, a referent w i l l be set up i n one clause, and then referred back to i n the following clause—something like the equivalent of: ' Y o u k n o w , I keep m y cars in a box? W e l l , that's the b o x . ' (where an English-speaking child w o u l d say something like: " T h a t ' s the box I keep m y cars i n . " ) .

9.

The Acquisition of Turkish

861

THE SETTING OF L A N G U A G E ACQUISITION 7. C o g n i t i v e P a c e s e t t i n g o f L a n g u a g e D e v e l o p m e n t 7. / . Locative

Development

The course of development of locative notions follows a standard order, presumably based on language-free conceptual development, even though the principles of locative suffixation and postpositions are acquired early. The first locative expressions are simple nominal suffixes: -E 'moving towards', -dE 'located at', -dEn 'moving away from'. These suffixes do not require encoding of specific object and locational features, expressing the simple oppositions between location and movement, and movement towards and away from a refer­ ent point. These basic notions are accessible at a fairly early stage of cognitive development. The locative postpositions are possessed names of locations, with the locative suffixes appended, e.g.: masa

-ntn

table

G E N top

ust

'on the table'

-tin

(literally:

masa

-nin

table

G E N top

ust

'off of the table'

-de

POSS L O C

-tin

'at the table's top') -den

POSS A B L (literally:

'from the table's top')

It can be seen that the nominal suffixes, originally mastered in relation to nouns (e.g. ev-de 'in the house' [literally: 'house-at'], ev-den 'from the house' ['housefrom']), can be simply extended to postpositions, which are formally nouns. These locative postpositions are easily analyzable and semantially transparent; that is, each one is the name of a familiar location (e.g. arka 'back', yan 'side', alt 'bottom', etc.). One might predict, then, on formal grounds, that the entire set would be quickly and easily mastered. However, the underlying spatial-relational notions differ in complexity and have their own developmental history. The order of emergence of locative postpositions in a production test follows a developmental history roughly sim­ ilar to child speech in Indo-European languages, and matches perfectly an order of acquisition based on cognitive grounds proposed by Johnston and Slobin (1979): 'in'/'on'/'under'