(1982) - interaction between particles and word order in the

Upload: farah-aman-khan

Post on 06-Apr-2018

214 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/3/2019 (1982) - Interaction Between Particles and Word Order in The

    1/15

    Devebpmcntal ~ ~ y c ~ y Capyri thi l9U2 b) the American Psychologiul Asuxiiiion. ln c1982. V d . 18. NO. . ( 2 -16 . 00 2- 1649/b2/ I1101-00621W,75

    Interaction-Between Particles and Word Orderin the Comprehension and Production of Simple Sentencesin Japanese ChildrenKenji H a k u t aYale University

    English is a language t hat relies extensively on word order to signal gramm aticalroles and meaning of sentences. Although studies have heavily emphasized therole of word order for children learning English, there is little information re-garding children learning languages that rely less on word order and more oninflections and particles. Such dat a would be essential in formulating a language-universal theory of language acquisition. This article reports four experimentstappin g compreh ension, production, and im itation of simple Sentences in Japanesechildren between 2 an d 6 years of age. Japanese children must learn a languagethat has a dom inant subject-object-verb order yet that allows flexibility i n wordorder due to postposed particles that signal grammatical role. Results acrossdifferent tasks suggest that children learning Japanese show neither a strongreliance on word order nor on particles alone. Rather, they possess a bias for amatching between particles and th e position in the sentence where they appear. c

    A theory of language development mustattem pt to outline the sorts of biases that th echild either brings to the language-learningsituation or develops in the course of expo-sure to the particular language. This articlespeaks to two related dimensions alongwhich different languages can vary: the basicorder of the language, and the extent towhich i t allows variability from that order.By looking at the patterns of children learn-ing languages that differ along such dimen-sions, we can begin to develop a theory ofth e specific biases that children bring to thetask of language acquisition.English-speaking children have been ex-tensively studied with respect to their com-

    This mearch was supported by the Peter B. Liv-ingston Fellowship Fund of the Hanard Medical school,National Science Foundation Grants BNS73-09150 andGSOC73-09150 to Roger Brown, and a dissertationfund from the Department of Psychology and SocialRelations. Hanard University.I would like to thank Roger Brown, Jill de Villicn,Peter de Villicn, Haj R o u , Helen Tagcr-Fluberg.andS t m Pinker for helpful comments;my mother. EmikoHakuta, for usistana in collecting the data and forpncra l wisdom; nd K. Yokoyma, M. ami, Y. Toi.H. k , and other staff of the daycare anter wherethin research was d u c t e d .Rc qua ts for reprints should k ent to Kenji HaLuta,hpartment of Psychology, Yde University, B ox 1 I AYale Station, New Haven. Connecticut 06520.

    62

    prehension of reversible active and passivesentences (e.g., Bever, 1970; Chapman, 1977;de Villiers & de Villiers, 1973; Fraser, Bel-lugi, & Brown, 1963; Maratsos, 1974) . Re-versible sentences refer to those sentences i nwhich the meaning underlying the sentencecannot be determined on the ba sis of the lex-ical items alone. Thus to correctly compre-hend the se ntence, The frog kissed the fly,one needs to have an understanding of thegram matical and semantic roles signified, inthe case of English, through word order.There are two gen eral findings that e mergefrom the studies on English-speaking chil-dren. First, active sentences are far easierto comprehend than passives. Active sen-tences appear to be mastered roughly by ag e3 or younger, whereas it is not until age 5or so that ch ildren begin comprehending th epassive correctly, and even then with com-plications well into the school years (Baldie,1976; Beilin, 1975; Horgan, 1978; Sinclair,Sinclair, & Dc Marcellus, 1971) . Second,children at a certain point in development,roughly around age 4, systematically mis-interpret passive sentences, choosing the firstnoun of the sentence as the agent. For ex-ample, The frog was kissed by the turtleis interpreted as the frog kissing the turtle.At this point in development, childrens per-

  • 8/3/2019 (1982) - Interaction Between Particles and Word Order in The

    2/15

    PARTICLES AN D W ORD ORDER 63formance on reversible passives drops belowchance level (Bever, 1970; Maratsos, 1974).Presumably, this period of systematic re-versals of passives is the result of the childovergeneralizing the statistically predomi-nant surface order of agent-action-patientt ha t exists canonically in the language.Bever (1970) suggested that the child fol-lows a strategy based on word order: AnyNoun-Verb-Noun sequence within a poten-tial internal unit in the surface structure cor-responds to [agent-action-patient] (p. 248).This research tests the generalizability ofthese results to Japanese and more generallyinvestigates the role of word order in a lan-guage where it does not play a critical func-tion in signaling grammatical relations.Unlike English, where word order deter-mines the grammatical role of nouns in asentence, Japanese signals grammatical rolethrough postpositional particles. Thus, al-though the predominant, canonical order ofa sentence is subject-object-verb (SO V),that order is free in the sense tha t word orderis not essential for assigning grammaticalrole. The major constraint on word order isthat the main verb must be sentence-final(Kuno, 1973), b u t even this constraint canbe violated through dislocation of a consti-t u e n t to a position after the verb. With re-gard to the relative frequencies of th e twoorders in the language, Kuno estimated thatthe ratio of SOV to object-subject-verb(OS V) orders appearing in newspaper arti-cles is 17:l. Although there are no compa-rable data for adult speech addressed to chil-dren, the above estimate suggests that SOVis the predominant order. .Linguists working on Japanese have notexplicitly worked out the full implicationsfor meaning of the word order change, butin general it is considered an optional rulethat does not affect the propositional mean-ing, although it clearly affects the intendedfocus, or topic, of the sentence. Thus, anOSV order sentence tends to be used whenthe object is being highlighted by the speak er.McCawley (197 6) formulated the word or-der change in terms of Rosss (1967)scrambling rule for Latin, which simplyinterchanges the position of noun phrases.One important function of the scramblingrule in Japanese, as McCawley has pointed

    Table 1Examples of Sentences C reated FromCombittarionof Tw o b e l s of Voice and Tw oLevels of Word OrderVoia/word

    order E xmpl e~

    Active/SOV AGENT-gB P A T I E N T O BlT-BctiVC.Active/OSV P A T I E N T 4 AGENT-gB BIT-active.PusiVe/SOV PATIENT-&3 AGENT-IIi BIT-passive.Pauive/osv AGENT-ni PATIENT-gB BIT-passive.Nore. SO V - rubjcct-object-verb; OSV = object-subjtct-vcrb.out, is to make multiple center-embeddedsentences comprehensible.Since the SOV and the OSV orders canbe used for both the active and passivvoices, there are four basic simple sentenceforms in Japanese: SOV/active, OSV/ac-tive, SOV/passive, and OSV/passive. Table1 lists examples for each of these four sentence forms. For expository convenience,have chosen the two nouns in the sentenceto be agent and patient, correspondingto their semantic role within the sentenceThe reader might imagine all of the sentences as referring to interactions betweenan insurance agent and his hospitalizedclient. Referring to the first example in Taequivalent to the English sentence, Theagent bit the patient. The -ga marking onAGENT indicates that it is the subject of th esentence, and the -0 marking on PATIENindicates that it is the object of the sentenceThe verb affix indicates that the sentence iin the active voice. Sentence 2, PATIENT-AGENT-ga BIT-active, is identical in everyway to the first sentence, except that theposition of the subject and the objechave been switched. Sente nce 3, PATIENT-gform. Notice now that the -ga markingwhich signals the gram matica l subject of th esentence, is on the patient. The object opassive sentences is marked by -ni. Th e verbaffix indicates passive voice. Once againwhen the subject and object are switchedit results in the fourth sentence, AGENT-n

    What sorts of predictions might be mad

    ble l , AGENT-@ PATIENT-0 BIT-active, i

    AGENT-ni BIT-PaSSiVC, is the SOV/paSSiVe

    PATIENT-ga BlT-paSSiVe.

  • 8/3/2019 (1982) - Interaction Between Particles and Word Order in The

    3/15

    64 KENJl HAKUTAabout Japanese childrens comprehension ofthese four sentence forms? Since SOV/ac-tives (agent-patient-action), being the ca-nonical simple sentence, are the most fre-quen t in th e languag e, one possibility is thatchildren will form a generalization based onthe semantic sequence inherent in these sen-tences. This would result in the Japaneseversion of the strategy proposed by Bever( 1 970) for English. It would take the form:Any noun-noun-verb sequence in the sur-face stru cture corresponds to agent-patient-action. If this were the case, we should ex-pect Sentences 1 and 4, which have theagent-patient-verb sequence, to be comp re-hended better than Sentences 2 an d 3, whichhave the patient-agent-verb sequence. Wemight also expect a period of systematic re-versals for the patient-agent-verb sentences,as has been found quite consistently for En-glish children.A second possibility is that the semanticsequence plays no role at all and that thechildren pay attention to the informationsignaled by the particles. Since actives aremore frequent than passives, we would ex-pect overall better comprehension of activesover passives; however, within the se two lev-els of voice, we would not predic t differen cesas a function of word order, since they donot differ with respect to particles.There are several studies that shed lighton these possibilities. Yamanaka (cited inMurata, 1972) used a picture-cued compre-hension procedure and tested the compre-hension of reve rsible and nonreversible ac-tive and passive sentences, but only in th eSOV order. Her results are somewhat puz-zling in that only at age 5 did her subjectsbegin performing above chance on any of thesyntactic types, including irreversible sen-tences. Her conclusion was that particlesstem to confront the Japanese child with adifficult learning task. An alternative pos-sibility is that the test was poorly designedor administered. Hayashibe (1975) foundsomewhat more positive results. Using anact-out cornprehension procedure, he lookeda t SOV nd OSV orders, but only for activesentences. His subjects were between 3 an d6 years old. Th e analysis of his data was notideal for our purposes ~ u s ce groupedhi s children according to error rates and

    showed that the group with the lowest errorrate had the highest mean age. Nevertheless,it is possible to infer from his tables tha tfrom about age 4 on, the children were re-sponding to the contrast contained in par-ticles.Four experiments are reported in this ar-ticle. I n the first experiment, children weretested on comprehension of the SO V andOSV active and passive sentences describedabove in Table 1. The second experimentattempted to look at comprehension underconditions in which th e particles were om it-ted, thus allowing us to assess the role ofword order in the absence of particles. Th ethird experiment looked at spontaneous usageof the subject- and object-marking particlesin an elicited production task. The fourthexperiment looked at childrens ability toimitate sentences in the SOV and OSV or-ders. The results consistently suggest thatJapanese children are sensitive to the cor-,relation between particles and their locationwithin the sentence, and that violation of thiscondition disrupts performance. These re-sults are compared to data from other lan-guages in the discussion.

    -

    Experiment 1Th e purpose of the present experimentwas to explicitly compare the four sentenceforms mentioned above by using an act-outprocedure.

    MerhodSubjects. Subjects were 48 children, who were di-vided evenly into four age groups:Group I , 2 yr. 3 mo.to 3 yr. 2 mo.; Group 2, 3 yr. 3 mo. to 4 yr. 2 mo.;Group 3, 4 yr. 3 mo. to 5 yr. 2 mo.; Group 4. 5 yr. 3mo. to 6 yr. 2 mo. Sex was balanced within each gr oup.For all studies rrported in this article, the children camefrom a large public day-care Ocnter in Tokyo. The secioeconom ic level of the children ranged considerably,but since eligibility for day care is b a d primarily on t is appropriate to justify the apparent arbitrary@ O U f i gof the age ranges used for this and subsequentexpcrimentr. Children in this day-care center are as-signed to classes on the basis of their age on April I .tbe beginning of the school year in Japan. For practicalremom urociated with collecting data in a large day-o r e center. the age groups I chosc were adopted as theywere rcpracntd in coch of the classes. Thus the agepoups differ slightly across experimen ts depcnding onwhen the experiment was conducted.

  • 8/3/2019 (1982) - Interaction Between Particles and Word Order in The

    4/15

    P ART ICL E S AND W ORD ORDE R 65 a. need, the majority of children, specially thow in fam-ilies with both parents working, can k escribed n~lower-middle to middle class.

    Materia ls and procedure. Thra eplications of eachof the four sentena forms resulting from the combi-nation of two levels of voice (active/possive) and twolevels of word order (SOV/OSV) ere devired. Thuathere were 12 se nt en as presented to each child. Lexicalitems were randomly assigned to the appropriate sen-tence frames. The nouns used were a l l iga tor (wad) ,gori lla (gorfra) , camel (raku da), p n d a (panda). bear(kum a) . cow (urhi ) .elephant (zo),girujJe (klrin). tiger( t o ra ) . / mg (koeru),an d turtle (komc) .The verba werekicked (ketfa in active, kerareta in passive), licked(nameta in active, namerareta in passive), Ai r (burra inactive, burarera in passive). kissed (kisu-sh lta in active,kisu-sareta in passive). and t ick led (ku ug ut tu in ac-tive, kusuguranra in passive). The animals were w-lected on the basis of the availability of toy replicas, andverbs were selected on the bask of their requiring dis-tinct actions when the child acted them out. From thefirst set of 12 sentences. a rccond set wa s crented inwhich the sa me verbs were used but the two nouns werereversed to counterbalance any effectsdue to the lexicalitems. Within each age group, an equal number of chil-dren were assigned to do each set.T h e children were tested individually in a separateroom in the day-care Center. One experimenter pre-sented the sentences and was the primary interactor withthe child while the sccond experimenter coded he childsresponse. The first experimenter was blind to the hy-potheses of the study. To familiarize the child with thematerials, each animal was introduced individually andthe child was asked to name it and encouraged to playwith it. A puppet w as then introduced. The child w astold that the game was to act out on a wpoden stagewhat the puppet said. Three simple warm-up sentences.consisting of an intransitive action and tw o irreversibleactions. were given. T h e 12 ScntenceS immediately fol-lowed in randomorder across subjects. Only the relevantanimals were placed on the stage for each sentence. Theentire procedure lasted about I S minutes.Scoring. Each correct response was scored as 1, andeach incorrect response was scored as 0.Results and Discussion

    For this and subsequent experiments inwhich two sets of sentences were crea ted andsubjects were randomly assigned to eitherset, the variable (set) was included in theinitial overall analysis of variance (ANOVA)as a between-subjects variable. When setwas not significant as a main effect or a s aninteraction with other variables, a secondANOV A was carried out pooling across set toincrease the power in testing the effects ofinterest. The ANOVA reported excluded thevariable set from the analysis except forcases in which it turned o ut significant in thefirst ANOVA.

    Table 2Means and Standard DrVl0tion.s of Voice XWord O rder InteractionWord order

    SOV osvVoice M SDI SD2 M SD, S&

    Active .819 .257 .024 369 .322 .032Passive .431 3 0 .067 .521 .307 .072Note. SOV - subject-object-vrrb; OSV - objcd-8ub-ject-verb. SD, rcfen to subject variability, and SD2n-f e n to sentence variability.

    In this experiment, overall analysis wasconducted by a three-way ANOVA, with sub-jects nested within age and the repeatedmeasures factor, sentence, nested within across of voice and word order? Th e between-subjects factor, age, was significant, F ( 3 ,1 8 ) = 9.207, p < .001. Inspectio? of themeans revealed a consistent increase withage. Age did not interact significantly withany of the repeated measures factors. Thetwo repeated measures factors, voice andword order, as well as the Voice X WordOrder interaction, were all significant. Forvoice, better performance was found for ac-tives than passives, F ( 1 , 9 ) = 36.079, p