Lexicaldevelop Snyd 81

You might also like

Download as pdf
Download as pdf
You are on page 1of 18
J. Child Lang. 8, 565-582. Printed in Great Britain Content and context in early lexical development* LYNN S. SNYDER University of Denver ELIZABETH BATES anp INGE BRETHERTON University of Colorado (Received 23 June 1980) ABSTRACT Investigators have targeted the onset of reference in children at varying points on the developmental continuum. Their viewpoints seem to be related to issues of the contextual flexibility, content and composition of the lexicon in comprehension and production. This study considered all of the variables in the early vocabularies of 32 children with a mean age of 1; 1.7. An intensive maternal interview was used to obtain data on the words comprehended and produced by the children and the contexts which supported their interpretation and use. The findings pointed to a relative independence between the domains of comprehension and production. Vocabulary size was related to all contextually flexible items in production while only related to the contextually flexible predicates in comprehension. Further, there was evidence for an early version of the referential style that has been reported at later stages of development. INTRODUCTION For most parents the infant’s first word marks a clear and important transition that changes the child’s status as a social partner. For child language researchers, this transition is just as important; but it is far less clear how or even when it happens. In particular, it has been difficult to determine when the child tryly understands reference, or the idea that things have names. “Research on first words has focused on three issues: \ (1) CONTEXTUAL FLEXIBILITY, or changes in the way words are com- prehended and used across a variety of situations; [*] This research was supported by The National Science Foundation Grant no. BNs76- 17624, and a grant from The Spencer Foundation, to the second and third authors. ‘Address for correspondence: Lynn S. Snyder, Department of Speech Pathology and Audiology, University of Denver, Denver, Co. 80208. 565 CHILD LANGUAGE (2) CONTENT AND composition of the lexicon, particularly the pro- portion of predicates or state and action words, and general nominals or names for things at different points in the one-word stage; and (3) COMPREHENSION VERSUS PRODUCTION, with emphasis on differences in timing, rate and content of these two domains. Depending on a given researcher’s emphases regarding these three issues, the onset of reference has been fixed at three different developmental moments in the one-word stage: (1) 9-10 months of age, when children first produce conventional sounds and show clear comprehension of at least a few words and phrases; (2) 12-13 months, when most children produce sound sequences that seem. to function like names from an adult point of view; and (3) 16-18 months, when there is a sudden surge in vocabulary growth and an increased use of single words to convey combinatorial or sentential meanings. ‘The researcher or parent who ignores the first issue would probably accept the ten-month boundary as ‘the’ transition point. A number of investigators have described the vocal schemes used by infants at this age level (Bates, Camaioni & Volterra 1975, Bruner 1975, Carter 1975, Dore 1975, Greenfield & Smith 1976, Harding & Golinkoff 1977, Piaget 1962, Sugarman-Bell 1978). Vocal conventions at ten months include all-purpose request sounds (e.g. mmmmm or nana or even mama to ask for anything), sounds used when showing or pointing to things, e.g. dat, and words used in highly stereotyped fashion within games or routines, e.g. bam said while knocking over blocks, or hitty used in a game of throwing the toy cat out of the crib. While many of these productions are indeed words in the adult lexicon, they are not used outside the ritualized context of the game or single, particularized instances of object or event. This rigid and contextually narrow use by infants suggests that such words are in some sense ‘non-referential’ (Bates, Benigni, Bretherton, Camaioni & Volterra 1979) or ‘pure’ performatives (Greenfield & Smith 1976). In short, the child has words but he has not yet extended their use to make reference to classes of objects and events. Around 13.1 qualitative changes have been observed in the way the same yocal schemes are used. The itty label may differentiate from the cat-throwing game t6 serve as an identifier for cats in a variety of settings; the sound bam may be used prior to or after a variety of banging or throwing acts, serving as a comment on the associated action. Changes like these are described by Volterra, Bates, Benigni, Bretherton & Camaioni (1979) in a longitudinal study of 25 infants from 0; 9 to 1; 1. By the end of the study, most of their subjects produced at least one or two words to anticipate, recognize, and identify or 566 EARLY LEXICAL DEVELOPMENT remember a class of objects and events (Volterra, et al. 1979: 178). Using this definition of reference, they conclude that the idea of naming is achieved by I; Still other investigators do not credit children with true insight into the naming game until several months later. Around 16-18 months, there is usually a sudden acceleration in vocabulary growth (Bloom 1973, Nelson 1973). This growth spurt coincides with some further changes in the way that single word utterances are used. While the 1-year-old may: point and say doggie only to label or request a dog, the 18-month-old may point to the dog’s dish and say doggie, in some way naming an association or possessive relationship between dog and dish (DeLaguna 1927, Greenfield & Smith 1976, Werner & Kaplan 1963). Dore (1978) has argued that such changes in size and use of vocabulary reflect a new and discontinuous level of understanding in the development of reference, a change he calls the NOMINAL INSIGHT. Nelson (1979) has also suggested that changes in the child’s understanding of reference lead to an interest in collecting names for things, with a resulting increase in vocabulary. Data on the content and composition of early vocabularies are also relevant to determining the origins of reference. As reviewed by Benedict (1979), Bowerman (1976) and Volterra, et al. (1979), there has been considerable controversy over the role of object names versus state and action terms or predicates at the earliest stages. Many of the early diary studies indicate that object names predominate from the first stages of meaningful speech, e.g. Leopold (1949). Other researchers have stressed the importance of action terms as the first linguistic extension of the child’s earlier sensorimotor action schemes, (e.g. Piaget 1962, Nelson 1978). Bloom (1973) has observed theré may also be a developmental shift from a predominance of FUNCTIONAL WORDS, e.g. more, no, and allgone, to a stage where SUBSTANTIVE forms or those ‘which make reference to classes of objects and events...’ (Bloom 1973: 70) are used more frequently and, thus, constitute the larger proportion of the child’s productive vocabulary. She suggests that this function/substantive shift results from non-linguistic changes in the child’s understanding of the object concept, leading to a more complete understanding of naming and reference in language. This proposal is compatible with Dore’s conclusion that the nominal insight occurs later in the one-word period. While Bloom’s observation has been supported By some findings, e.g. Ingram (1978), it has not been corroborated by others (see the reviews by Benedict 1979, Bowerman 1976 and Volterra, et al. 1979). Object names seem to dominate the vocabularies of some children from the very beginning of productive speech. There are two ways to interpret this. First, it is possible that this is an early version of the individual differences in vocabulary composition reported by Nelson (1973) for 18-month-old infants. Examining the composition of their productive vocabularies, she discerned two lexical styles: a REFERENTIAL 567 CHILD LANGUAGE style marked by a high proportion of substantive forms and an EXPRESSIVE style characterized by a heterogeneous vocabulary composed of words from different form classes. Secondly, it is also possible that those children who tend to use high proportions of nominals very early may be more generally advanced. ‘These children may have attained the nominal insight much earlier in develop- ment, during language comprehension prior to the onset of productive speech. For such children, we would expect to find a high proportion of nominals in comprehension as well as production, perhaps reflecting an earlier version of the function/substantive shift. This latter proposal forces, us to consider the third issue: the distinction between comprehension and production. The above suggestion rests upon the assumption that these domains are linked in such a way that insights gained in comprehension can be transferred directly to production. There is certainly considerable evidence that comprehension PRECEDES production throughout the one-word stage (Greenfield & Smith 1976, Oviatt 1979, Volterra, et al. 1979); decontextualization or increased flexibility in comprehension has corresponding parallels in later production (Huttenlocher 1974, Volterra, et al. 1979); and significant correlations between comprehension and production vocabularies fromo ; 9 to 1; 1 have been reported (Bates, Benigni, et al. 1979). However, other evidence suggests a surprising independence between the two domains in rate, content and decontextualization. Both Benedict (1979) and Goldin-Meadow, Seligman & Gelman (1976) observed that some children developed extensive comprehension vocabularies while producing very little. Moreover, they reported that the early comprehension vocabularies contained a relatively high proportion of action terms, while production vocabularies at,a similar point of comparison involved a larger ratio of object names. Rescorla (1980) observed that children responded appropriately to some words in comprehension which they then overextended in production. This compelling evidence for the difference and independence of comprehension and production processes would suggest that the discovery of reference may have had to take place twice: once in the understanding of words and again in productive naming. The studies reviewed thus far have offered rich and detailed descriptions of developmental sequences. Unfortunately, their small sample sizes make it difficult to determine any relationships among the different aspects of develop- ment. In the present study, we will examine first words in a sample of 32 13-month-old children, a sample large enough to investigate relationships among size, composition, and contextual flexibility within and between com- prehension and production vocabularies. 568 EARLY LEXICAL DEVELOPMENT METHOD "The data discussed here are taken from a larger longitudinal study of 32 infants at 0} 10, 131, 138 and 2; 4. This study includes a variety of observational measures, experimental tasks, and interviews with the.mothers, all designed to assess aspects of linguistic and non-linguistic symbol development. In this report, we will concentrate on the results of an intensive maternal interview which was conducted when the infants were 1; 1. It focussed its queries on the content, composition and contextual flexibility of comprehension and pro- duction vocabularies. Given these objectives, the maternal interview was chosen as a research tool. When one is concerned with identifying both the early, context-bound comprehension of tokens and later, highly decontextualized forms, the person who can observe the child across many contexts is an invaluable resource. The resulting maternal interview data will be more representative of the range of the child’s competence. On the other hand, observational samples tend to tell us more about robust, frequently occurring developments. Thus, some investi- gators describe a discrepancy between parental report and their samples of the child’s linguistic performance. The reported parental ‘overestimation’ may simply reflect the inherent differences between the types of data sampled with each methodology. Maternal report may not be suitable for some areas of research, e.g. phonetic/phonological studies of lexical and holistic forms emerging during this period. There is no substitute for observational data in studies taking such a direction. However, the thrust of the present study was the content, context and timing of lexical development at 1; 1. Con- sequently, the maternal interview was clearly the method of choice for this study. Limited validation of the interview measure against observational data is reported in Bates, Benigni et al. (1979) and Bates, Snyder, Bretherton & Volterra (1979). Subjects - Thirty-two infants and their mothers from middle-class families in the Boulder, Colorado, area participated in the study. The infants ranged in age from }; 0.14 to 131.21 with a mean age-of 1; 1.7. The sample contained 16 girls and 16 boys. The maternal interview: content The interview was based upon a roo-item word and phrase checklist, compiled from a review of diary studies in the literature, our previous research and that of other investigators, e.g. Nelson (1973). The checklist included specific lexical items within general categories, e.g. animal names, foods. These items 569 CHILD LANGUAGE were intended to cover the range df words likely to occur in the comprehension and production repertoires of 13-manth-old infants. The maternal interview: procedures Administration. An interviewer visited each mother at home and conducted an extensive 1}- to 2-hour interview of the child’s language comprehension and production. The initerviewer went over each item on the checklist, asking if the child comprehended and produced each word, eliciting examples of how the word or phrase was comprehended and produced, and systematically probing for any gestural, vocal, spatial and temporal contexts which may have accom- panied and supported the child’s competence with the item. For example, the interviewer would initially ask if the child understood and used the word ball. If the mother responded that the child could not say it but did understand it, the interviewer would follow with ‘Can you tell me what lets you know that he understands it?? The mother’s answer would be followed by probes such as the following: ‘When does this happen? Are you usually pointing to the ball when you ask for it? Does it only happen during a game, like rolling the ball? Can he get it for you anytime? Can he pick it out of a group of toys? Can he get it from another room? Does he only understand it in terms of one or two particular balls - or all that he sees?’ Similarly, if the mother responded that the child understood hot but did not use it, her answer would be followed by a number of probes. These probes would include: ‘When does this happen? Do you use a special tone of voice? How would you say it? Do you use any gestures when you tell him something’s hot? Does he understand it only in regard to the stove? How does he respond?’ This method, a RECOGNITION METHOD which used an item-by-item checklist with systematic probes, is different from the Bates et al. (1979) RECALL METHOD, which asked mothers to recall items within general categories such as animal names and then probed for anecdotal material. The interviewers transcribed the mother’s responses on a coding sheet which provided space to record anecdotes for comprehension and production and each of the contextual constraints on the range of reference for each word and phrase. The interviews were also recorded on a cassette tape recorder. All coding sheets were subsequently checked against the tape of the interview so that all responses were transcribed. Coding. The data were coded by two of the three authors (LS and EB), with the third author (IB) assisting in establishing reliabilities. The coding involved two phases : division into content categories, and ratings for contextual flexibility. The content codings were made entirely from an adult perspective, dividing all comprehension and production items into three types: common nouns, proper nouns and a heterogeneous class of predicates including action verbs, locatives, adverbials and routines like bye-bye. The categories were limited to 57° EARLY LEXICAL DEVELOPMENT these three content types by the small repertoires of 13-month-olds and our intent to perform correlational analyses on the proportions and degree, of contextual flexibility within content types. Inter-coder agreement, using this conservative standard, was 100%. At the second level of data reduction, the anecdotal information was utilized and its contextual flexibility determined. A set of rules was devised to determine how the whole phrases found within the comprehension vocabularies would be credited for number of lexical items. Similar to Benedict (1979) we took a somewhat conservative approach. Take, for example, the phrase Wash your hands. If the child only understood wash and hands in the context of that phrase, it was double classified: he was credited with one context-bound predicate and one context-restricted common noun. However, take a child who responded to Wash your hands by putting his hands under the water; to Show me your hands by putting his hands up in the air; and to Show me your nose by pointing to his nose. He would be credited with one predicate count for the Wash your hands unit, one predicate count for show, and two common noun entries for hands and nose respectively. Then, after this simple count was established, the fact that a predicate like wash appeared in one frame only or in several verbal contexts would be taken into consideration in determining degree of contextual dependence for each lexical item. The decontextualization of items was determined with a dichotomous classi- fication for purposes of analysis. All entries were classified as either cohtextually restricted (CR) or contextually flexible (CF). Two sets of rules were devised for these assignments. The set of general rules, governing broad lexical classes, appears in Table 1. One example is the CONTEXTUAL OVERRIDE RULE: any comprehension or production item that is typically accompanied by clear gestures or contextual cues is classified as contextually-restricted, e.g. obeying spit it out only if mother’s hand is below the child’s mouth. Another example is the ABSENT REFERENT RULE: an item is contextually-flexible if the child persistently searches for an absent referent in a variety of places, e.g. seeking Daddy all over the house or in several characteristic locations in response to Where's Daddy? A set of specific rules was developed to deal with typical anecdotes and contextual variations for individual lexical items, illustrative of the same principles as the general rules. For example, if the child understands the word hot in only one type of context such as the stove or food but not both, it is considered context-restricted. It also receives the same classification if the mother’s tone of voice seems to carry the message. However, if the child obgys it in more than one context without a characteristic tone of voice, it is considered. contextually-flexible. Since the three coders achieved 83% agreement for second level codings on a randomly selected set of protocols, all protocols were then independently coded by two raters, with discrepancies resolved at joint meetings. 57t CHILD LANGUAGE “ao = pordde Aiquizea axe anq saseq oaneuttojied and avy 3ey3 “919 ‘ou ‘24 ay] SPIOM Lag $YD = ‘919 ‘ypopop ay] spunog - “(AD se Ajissepo pue o[nz yuorezor auasqe Afdde ‘seo sry uy *3°9 03 uoRIppe Ur ae1D0sse-uupues r8y30 suIOs S90p Jo ,euupuessy, spuodsas ‘aovjd onsiiajovrey B 03 $208 ose) BUIpUBID) syoadxea piryo oY} IY? SoUAptAD JUDpUOdapUT sI ary} ssaTUN ase9 JJNI “Ie UB 9q P[NOYs pue yeaa st sia ‘payoxe 393 07 sulsas pyryo ayy pus ,Zurw0o seupuery, ‘shes Al JY ‘apn ‘s’a ayy epasiadns uo uondeosaqut juaAe yuasqe Ue oxy — yasdn $498 plrys pue ,2Xq-2kq BuIo8 sAUMUOYAL, ‘Burkes Aq axazu00 Jo yno D saseaa Jayporg ‘a[durexa Jo] “J 2q 01 soso Tensnun ouwlos Joy ox9y WOE st aI0Y,], “YD se IUNOD ‘parToxa syoo] Gasdn sjaH ‘saws ayy] esucdsax feuoKoura Jo PUP SUIOS st WoIsuaYyesdusOD 10s 2oUDprAa ApUO DU} JT ‘oy ‘pug 3,ue0 9y jr gosdn st ‘osaymAs0A9 syooy ‘spuy ay NUN, sve PY ss9]UN YD a7¥ ,03 $903, 30,30} SHOOT, VoUD}y “UOISTA Jo eur] ayerpeurumT ay} puodag Yosees yuaysisiod Jo} oUaprAe LIOTTAXA Sf axey} ssayuN UO se yunod ‘sosuodsoz sovyd yengiqey 9q pfnos Aey3 SyIT punos yeYa@ UO ‘spIOAL ONO, UT “YO se apoo — ur 8,94 woos Jey Jeo[D 3OU $,31 HOYM ,TeyOYsY 02 5908, skes Joyous “a — aud onsisojourEYp & 0} SuIoB SupuxvoUCo snongiquie st |ouaplAs 2UB JT “AO = (y20q prom ayy sivay ay uayM Woompeq oy} 6) Burs ‘woos JayouR ur Appec] Sunjaes “8-2) uasoyox Burs v 01 prom e soxjdde so spuodses pry ou JT "AD sfemye are uoNOnposd pue uorsusyorduos wy suoMMUN szerpouNy “(SuTuvaUT yp pares aay 3yBruy-o79 ‘puns ‘71s noysia auoye dn 10 umop yey os 3x9;U09 2Yt 0} Surpsosoe JAsyEyM 30 de] Yo syoB ‘sys ppryo ayy axoys dn/umop 30 ‘outeu Syf 0} asnf puodsar sySw pue unoU ayy smouy pliy> ay3 IY) BoUBpIAs juapusdaput St oxy aFoyA UNoU/yo0] “3*2) prom Sursuedusoooe oy; Jo Suruvaut ayy Lswo Apuspuodepur pjnoo-yeyp syxqquoo [eqraA Io ‘s3xajU00 oyeLadosdde ut skemye (no Sud ‘zoyeM WVeG BuNIEys “Burouep) syuado Jo uoNENTUT ‘syuauTUEduIOD0B ainyseB/au0 Ya (Gonuaye SunseNp ‘Surppiqzoy ‘Sutures ‘uorssrunrd) 20103 39e ypoads sapnpout Sty, “(paromsor ixoquos) YD se spoo ‘UMo sy UO SuruaUT ayp [Je A184} P[Roo yey a[qurIeA JeuONENsIS Jo aInysed ‘gu07 Aq poruedwioooe (Ue WOdUA suas “Kressaoau Ayqeqord, sdes Jj 10) ATTVASN JO SAVAVIV UOTONposd Jo UoIsudYyasduI0D 5] spunos aaneusojzod ang (+) asuodsaz jeuonowrg (£) guaszeyex suasqy (2) aplszaao [enyxaquoD (1) uononposd/uorsuayarduros : uonduosep [einoraeyog, ony DIDp yoiopsauv Surpor sof sans possuag *l ATAVL 572 Ls (6) Food and other multiple referent but single identical class words (e.g. crackers, milk, diapers, other disposables): (7) Classes where distinct referents are possible (toys, telephones, books): - (8) Classes where distinct referents are possible but not likely (e.g. the refrigerator): is unclassifiable. “ If M says will look for or pick out, assume CR unless it can be EXPLICITLY established that C will pick the person out of an array of two or more people. Picks out in pictures, always = CF. If it is an emotional response only = CR. Tf child can pick out from an array, or would be surprised if given something else, or goes to = CF. If only when object is in a characteristic location = CR unless in other rooms (see absent rule). If M says only ‘looks at’ or ‘picks up’ without explicit evidence that there was more than one choice available = CR. If only evidence is ‘looks for’ or ‘gets’ assume CR unless: (2) more than one referent is mentioned explicitly, or (2) some sort of plural morpheme is available (dogs, balls, several, any, always, knows-them, etc.). If explicitly stated that child knows/ applies only to ‘his own’ or ‘probably just his own’ or some other clear restriction to a single referent =. CR. If explicit restrictions, but to more than one member (his own books, etc.) then CF. As above, CF only if there is an absent object override, or if explicit plural evidence is available. Picks out in books as well as in real life always counts as plural. particular games If used in greeting, to label (especially in pictures), or in ways that indicate appropriate understanding of name without,explicit contextual restrictions = CF, If name is overgeneralized to more than * one referent (e.g. daddy to all men or both parents, and not just in particular games, grandma to all older women, dog’s name to all dogs) = CR. Labels or requests = CF. Overgeneralization = CF. Imitation only = CR. As in comp., except that repetitions and routine-based uses like ‘how does the doggie go’ are always CR. Eee ANYNAOTIATA TVOIXTT ATUVE CHILD LANGUAGE ‘ RESULTS AND DISCUSSION In order to examine the three central issues of early lexical acquisition, proportion scores were constructed for the total number of common nouns, proper nouns . and predicates in each child’s comprehension and production vocabularies. Similarly, the proportions of contextually-restricted and contextually-flexible items within each content category and within the total vocabulary were determined. These variables were then entered into Pearson product-moment correlations with two-tailed tests of significance. We will first examine the more general descriptive findings and then narrow our focus and consider the results of the correlational studies as they relate to the issues of the contextual flexibility, content and composition of the lexicon in comprehension and production. General findings The descriptive statistics for the data in comprehension and production appear in Table 2. The widely documented asymmetry between comprehension and production also appears in our sample. The mean comprehension vocabulary was 45°28, compared with 11-21 for production. In fact, these absolute values are comparable to Benedict’s results at 1; 1. However, these levels are quite different from those in Bates, et al. (1979) who report a mean of 10-4 items in comprehension and 8-2 items in production at the same age level. It seems that the recall format of the Bates, et al. study elicited comprehension items that were particularly robust and salient for parents, while the recognition format reminded them of behaviours which they may not have considered in terms of-language comprehension. In contrast, ALL of the child’s productive items at 1; 1 may be salient and memorable. The asymmetry is further underscored by the low correlation between the two domains: +029, P < o-11 using a two-tailed test of significance. We discover some revealing aspects of this relationship if we look at the distribution of scores in the sample. Fig. 1 presents a scatterplot of z-scores for comprehension versus production in this sample. Three different patterns were obtained: high comprehension/high production, high comprehension/low production, and low comprehension/low production. The only pattern which did not occur was high production in a child who whderstood relatively little, confirming the general finding that comprehension precedes production. It also suggests that receptive * vocabulary can intrease at a surprising rate without parallel developments in expressive speech, These data, ‘then, point to a general trend towards asymmetry between comprehension and production, represented in the sample by three different profiles. Let us now examine the way in which contextual flexibility is represented in those domains. 574 EARLY LEXICAL DEVELOPMENT TABLE 2. Descriptive statistics for maternal interview and its codings* Comprehension Production Mean Range Mean Range Total 45°28 11-97 rar o-45, Usage Restricted 27°93 7-54 453 o-18 Flexible 14:59 1-48 4°84 o-22 Unclassifiable S12 0-20 193 o-16 Common nouns Total 16°93 1-64 4:00 o2r Restricted 9°59 0-27 125 o7 Flexible 493 0-33 225 on13 Proper. nouns Total 5°96 I-21 2°25 o-7 Restricted 2°43 9 068 o-3 - Flexible 2°50 o-10 0°90 o3 Predicates Total 23°31 8-35 5°03 o-20 Restricted 15°43 $729 2°59 o9 Flexible Gar 1-13 168 o-10 Slight discrepancies in totals reflect a few items which were double-classified across categories, Comprehension -300 -200 -100 0 100 200 300 400 Production Fig. 1. Scattergram of distribution of standard scores for comprehension and production. 19 575 yer 8 CHILD LANGUAGE Contextual flexibility ‘ If we look at the general results of the contextual flexibility codings, we are again struck with differences between comprehension and production. Over 60% df the comprehension items were context-restricted, compared with 48% of the production items. The correlational results suggest even stronger difference in contextual dependence. Table 3 presents correlations among proportions of contextually flexible items and the proportion of common nouns and vocabulary totals. In comprehension, the only flexibility score which yields significant results is a correlation of -+0-41 between the proportion of contextually flexible predicates and the total vogabulary. In other words, decontextualization of predicates in comprehension relates to overall vocabulary expansion while it does not relate to anything for common and proper nouns. TABLE 3. Proportions of context-flexible items with proportion of common nouns and vocabulary totals . Total Common Proper nouns Predicates flexible nouns flexible _fiexible flexible (%) (%) (%) (%) Within comprehension Total vocabulary +029 +017 008 +o-4r , P

You might also like