the development of a linkage between count nouns and object … · 2017. 6. 15. · however,...

The Development of a Linkage between CountNouns and Object Categories: Evidence fromFifteen- to Twenty-One-Month-Old Infants

Sandra R. WaxmanNorthwestern University

D. Geoffrey HallM. R. C. Cognitive Development Unit, London

WAXMAN, SANDRA R. . and HALL. D. (;t;OFFRi:v. The Development of a Linkage between CountNouns and Object Categories: Evidence from Fifteen- to Twenty-One-Month-Old Infants. CHILDDEVELOPMENT, 1993, 64, 1224—1241. Recent research suggests that, although young cbildrenappreciate many different kinds oi conceptual relations among objects, they iocns specificallyon taxonomic relations in the context oi word learning. However, hecanse the evidence lorchildren's appreciation of tbis linkage between words and object categories has come primarilyfrom children wbo have made suhstantial linguistic and conceptual advances, it offers limitedinformation concerning the development of this linkage. In the experiments reported bere, weemploy a match-to-sample task to iocns speciircally on tbe development of an appreciation oithe linkage between words (here, connt noims) and object categories in iniants in the period justprior to and just subsequent to the naming explosion. The results demonstrate that, for 21-month-old infants, most of whom have recently entered tbe socahulary explosion (Experiment 1), andfor 16-montb-old infants, most of wbom bave yet to commence tbe vocabulary explosion (Experi-ment 2), novel nouns focus attention on taxonomic relations among objects. Tbis is importantbecause it reveals a nascent appreciation oi a linkage between words and object categoriesin infants wbo are at tbe very onset of language production. Results are interpreted within adevelopmental accovmt of infants' emerging appreciation oi a specific linkage between countnouns and object categories.

Linguistic aud conceptual development described by count norms and that proper-have each, independently, served as a focus ties of ofsjects will be marked by adjectivesof decades of productive investigation. How- (Markman & Hutchinson, 1984; Markman &ever, more recently, the nature of the rela- Wachtel, f988; Soja, Carey, & Spelke, f99f;tion between these two naturally emerging Taylor & Gelman, 1988; Waxman, f99()).capacities has become a topic of livelydebate (see, e.g., Gelman & Byrnes, 1991). However, beeause the experimental ev-Several researchers have pointed out that, idence for an appreciation of these precisealthough young children appreciate many linkages has come primarily from childrendifferent kinds of conceptual relations who have already made substantial progressamong objects, they do not attend to these in both the linguistic and conceptual arenas,relations equally when trying to determine we are left with a very limited understand-the meaning of anew word, fnstead, particu- ing of when and how an appreciation oflar types of words (e.g., nouns, adjectives) these linkages develops. Therefore, in thefocus young children's attention on particn- experiments reported here, we focus spe-lar types of meanings. For example, children cifically on the development of an apprecia-as young as 2 and 3 years of age expect that tion of the linkage between count nouns andsolid objects and object categories will be object categories in infants ranging from f5

This research was supported hy a grant to tbe iirst author from Harvard University. Tberesearch was conducted at Harvard University. We are grateful to the infants and parents wboparticipated in these studies. We are also indehted to Skylcr V'inton and Eauren DiStefano lorassistance in data collection and coding. Special tbanks to Dana Markow for ber contrihution toall phases of tbe work. We appreciate tbe comments made hy Marie Balahan, Patricia Bauer,Ellen Markman, and an anonymous reviewer on an earlier version ol this article. Requests iorreprints should he sent to Sandra R. Waxman, Department of Psycbology, Nortbwestern Univer-sity, 2029 Sberidan Rd., Evanston, IE 60208-2710.

[ChildDevelopment, 199.3, 64, 1224-1241. © 1993 liy the Si.i-ietv IDI- Hcs.-arch ui Child i:)evi-l()i)meiit. Inc..\11 rights reservetl. ()()09-.3920;'93/64()4-()<)2()S()l.()<»i

Waxman and Hall 1225

to 21 months of age—infants in the periodsjust prior to and just subsequent to a phe-nomenon known as the naming explosion orthe dawning of nominal insight (Dromi,1987; Goldfield & Reznick, 1990; McShane,1979).'

This developmental period is of specialinterest primarily because of infants' re-markable accomplishments in word learn-ing. Before this period opens, at approxi-mately 10-12 months of age, infants reach amajor developmental milestone—they be-gin to produce their first words. At this point,word learning proceeds at a steady, gradualpace. Yet several mouths later, both thepace and character of lexical acquisition ad-vance dramatically. At approximately 17—20months of age, or once iufauts have acquiredapproximately 50 words in their productivevocabularies, there is a sudden burst inlexical acquisition (Benedict, 1979; Dromi,1987; Goldfield & Reznick, 1990; Nelson,1973), Infants begin to learn new woj'ds at arate of approximately nine per day (Carey,1978). It is because most of the words ac-quired during this period are couut uouus(Gentner, 1982) that this period of rapid lexi-cal acquisition has been referred to as thenaming explosion. The naming explosiondraws to a close when iufauts begin toproduce multiword utterances, sometimearound the second year.

Our goal in the experiments reportedhere vvas to examine how infants' apprecia-tion of the noun-category linkage corre-sponds to the major developmental mile-stones occurring during this active period.One possibility is that au appreciation of thenoun-category linkage is in place from theverj' onset of lexical acquisition (Macna-mara, 1982; Pinker, 1984; Waxman, 1991), Ifthis is the case, then the tendency to inter-prei nouns (or, perhaps, words from anygrammatical category—see "General Dis-cussion") as referring to object categoriesshould be evident in infants even as theybegin to produce words. That is, it shouldbe evident iu infants prior to the naming ex-

plosion. Such a finding would suggest thatinfants' first efforts to map words to theirmeanings may be guided by an expectationthat words and object categories are linked.

Another possibility is that an apprecia-tion of the noun-category linkage emergeslater and is dependent on experience withthe language system. If this is the case, thenan appreciation of the linkage might be ab-sent during the very earliest stages of lexicalacquisition; it should emerge only after theprocess of word learning is well under way,after children have acquired a sufficientnumber of words to permit them to inducethe appropriate correlation between countnouns and object categories (Bauer & Man-dler, 1989; Landau, Smith, & Jones, 1988;Markman, 1992; Nelson, 1988; Waxman,1991). A finding like this would suggest that,with experience, children come to notice thecorrelation between count nouns and cate-gories of objects. They then may exploit thiscorrelation to learn the meaning of subse-quent words.

To evaluate empirically these two broadalternatives, researchers have begun to ex-amine the impact of introducing novel labelsto young infants on a range of behavioraltasks. For example, Baldwin and Markman(f989) have reported that 10-14-month-oldinfants devoted more attention to objectsthat had been labeled with a count nounthan to objects that had not been labeled.This finding suggests that labels help to di-rect infants' attention toward objects evenbefore the infants are themselves producinglanguage.

Others have taken this issue a step fur-ther to ascertain v^'hether labels focus in-fant's attention on categories of objects. Forexample, Waxman and her colleagues usedan object manipulation task, analogous tomore standard infant habituation proce-dures, to examine the influence of novel la-bels on infants' categorization (Markow &Waxman, 1992; Waxman & Heim, 1991). Thedata revealed that labels focus 12- and 13-

' In this article, we seek to establisb tbe effect ol novel connt novms on infants' objectcategorization. Becanse we test only connt nonns, we cannot compare the influence of variousother syntactic categories {e.g., proper nonns, adjectives) on infants' object categorization. Wetberefore cannot determine wbetber the effects dein(jnstrated bere are specific to connt nonns,to all nonns, or to novel words (e.g., adjectives) in general. Tbis important issue, which will beaddressed in fnture work, is considered in the "General Discussion." The existing literaturesuggests that 15—21-montb-oId iniants do not distingnisb tbe syntax of count nouns from massnouns or froin adjectives (Bloom, J990; Gordon, 1985; Markow & Waxman, 1992; McPberson,1991; Prasada, in press; Soja, 1992; Valian, 1986). However, 17-montb-old females may distin-guisb proper from coimnon nouns nnder certain c ircunistances (Katz et al., 1974).

1226 Child Development

month-old infants' attention on object cate-gories, (See the "General Discussion" for afuller treatment of this finding.) Other recentstudies have provided converging evidencethat by 13—15 mouths of age, novel labelsdirect infants' attention to object categories(Echols, 1991; Roberts & Jacob, ui press).

Thus, a clear consensus has begun toemerge regarding infants' appreciation ofthe relation between nouns and object cate-gories in the period prior to the onset of thenaming explosion. The accumulated evi-dence suggests that novel nouns focus in-fants' attention on objects and categories ofobjects. This is consistent with the hypothe-sis that sensitivity to a linkage betweennouns and object categories is in place, in atleast a rudimentary form, at the very begin-ning of the process of lexical acquisition.

There is, however, one pviblished reportthat is at odds with this emerging consensus.Bauer and Mandler (1989) used a triad taskto examine the categorization afjilities" off6—31-month-old infants. One goal of theirexperiments (Experiments f and 3) was to"assess the inffuence of novel labels oncategorical responding" (p, 161), They hy-pothesized that the sensitivity to the linkagebetween nouns and object categories devel-op.s during the second year of life, in tlieperiod subsequent to the naming explosion.To test this hypothesis, they compared per-formance in Novel Label and No f^abel con-ditions. They predicted that, prior to the on-set of the naining explosion, novel nounswould exert no influence on categorizationperformance, but that, subsequent to thenaming explosion, novel nouns would to-cus attention on taxonomic relations (p. 160,p, 166). Howe\'er, the data were not consis-tent with this prediction, fnstead, novelnouns had no apparent effect. Interestingly,infants at all ages and in both the Novel La-bel and No Label conditions demonstratedan overwhelming preference for the taxo-nomic alternatives.

Bauer and Mandler'.> i.s the only pub-lished report in which infants have failed toreveal a sensitivity to novel nouns in an ob-ject categorization task. When their result isconsidered against the backdrop of existingwork in the field, it raises the possibility tliatthere may be a discontinuity in the develop-ment of the linkage between nouns and cate-gory relations. Recall the evidence indicat-ing that labels highlight object categories for

infants ranging in age from f2 to 15 months(Baldwin & Markman, 1989; Echofs, f991;Markow & Waxman, 1992; Roberts & Jacob,in press; Waxman & Heim, 1991) aud forpreschool children, ranging in age from ap-proximately 30 months to 4 years (Markman& Hutchinson, 1984; Waxman, 1990; Wax-man & Kosowski, 1991). Bauer and Man-dler's data suggest that these labeling effectsmay be absent in the period from 16 to 31months of age. This apparent discontinuityis especially surprising because it wouldcoincide precisely with the period duringwhich Infants are in the most active phasesof word learning. Because the possibility ofa developmental discontinuity raises an im-portant theoretical issue, it warrants carefulattention,

A thorough examination of Bauer andMandler's procedure suggests a likely expla-nation for their null effect. One key aspectof their procedure obscured the opportunityto observe the influence of novel labels intheir task. In contrast to the proceduresadopted in other empirical investigations oflabeling effects in young subjects (e.g.,Markman & Hutchinson, f984; Waxman &Kosowski, f991), in Bauer and Mandler'sdesign, the experimenter systematically re-warded infants in both the No Label andNovel Laf)el conditions for all and only cate-gorical selections. This reinforcement beganin an initial training period, fn this period,the experimenter cheered and clappedwhen infants made categorical selections;she corrected infants when they selected thetfiematic alternative. After the training pe-riod, the reinforcement continued. f3)uringthe test trials, infants vvere reinforced posi-tively for making taxonomic selections butwere offered only neutral comments (e.g,,"Thank you") for making thematic selec-tions, ft is therefore not surprising tfiat theseinfants selected the taxonomic alternativeswith such great frequency, even in the ab-sence of a label. Put simply, intants' uni-formly high rate of taxonomic responses inboth conditions illustrates the power of rein-forcement. As Bauer and Pvfandler acknowl-edge, infants' highly taxonomic performanceleft little latitude for novel labels to influ-ence performance.

ft is important to note, however, thatbecause Bauer and Mandler did not assessthe inffuence of novel labels in the absenceof reinforcement, these data do not bear di-rectly on Questions concerning children'sinterpretive biases in the context of word


leariiiiig." This is because, under more natu-rahscic conditions, infants do not have thebeiit;ht of reinforcement that is perfectlycontingent on their performance (Brown,1957; Carey, 1978), Instead, in the more typ-ical word-learning scenario, infants are facedwith novel words, novel scenes (includingobjects and events), and a nearly infinite setof possible mappings l)etween the two. Weknow that infants appreciate many differentkinds of conceptual relations (e.g., taxo-noniic, tfiematic, causal relations). The keyquestion is whether infants sample freelyfrom among these various kinds of concep-tual relations in nonliuguistic tasks f)ut re-strict their focus to primarily taxonomic rela-tions in the context of word learning. Theenipnical evidence cited above suggests thatthis may indeed be the case, for subjectsyounger than 15 months and older than 30montlis of age. At issue, then, is whether andhow novel nouns infiuence children s per-formaiice in the intervening period.

] fie goal of the current experiments waslo examine the influence of introducingnovel count nouns to 15—21-nionrfi-oId in-fants under experimental conditions that ap-proximate the circtimstances encounteredby the young child in word learning. To thisend, we employed a forced-choice match-to-sample procedure to compare performancein a Novel Noun condition (in which the ex-perimenter labeled the target object with anovel count noun) with performance in aNo Word condition (in which the targetsreceived no label). Because we providedno reinforcement contingent upon the chil-dren .s selections, our data serve as a directtest of the influence of novel labels in anexperimental procedure that resembles thenatural word-learning scenario, ff labels di-rect iniants' attention specifically on ofyectcategories, then subjects should be morelikely to choose taxonomically in the pres-ence of novel labels than in their absence.

Experiment 1

f he purpose of this experiment was toexamine the influence of introducing novelcoui)t nouns during the periods jnst beforeand iust after the naming explosion. The ex-

perinieut was designed to reveal tiie impactof novel nouns on 15- and 21-month-Oid in-fants' attention to object categories at the ba-sic and superordinate levels.

METHOD

SubjectsTwenty 15-month-olds (with a mean

age of 15 moriths, 17 days, ranging from 14mouths, 17 days to 16 moriths, 29 days) and20 21-nioutfi-olds (with a mean age of 21months, 1 day, ranging from 19 months, 23days to 22 months, 17 days) participated inthe study. At each age level and within eachcondition (see below), there was an equalnumber of boys and girls. Subjects were re-cruited either via direct letters to parents orvia an advertisement in a rnonthly parents'newspaper serving the greater Boston area.Subjects vvere drawn from a popuia'cion ofpredotiiinantiy white middle- to upper-middle-class families in which English wasthe hrst faiiguage. All subjects completedthe procedure.

StimuliStimuli were 36 toy replicas of aniniate

and inanimate objects, A complete list ofthese stimuli may be found in Table f, Theobjects were arranged into 12 sets of tliieeobjects each. Each triad included a target ob-ject (e.g., a carrot), a taxonomically relatedobject (e.g., a tomato), and a tfiernatically re-lated object (e.g., a bunny rabbit). For afl f2triads, the target object and taxonomic alter-native were toys representing inanimate ob-jects (e.g., items of food, clothing, ftirnittire);the thematic alternative was a toy represent-ing an animate object (e.g., dolls, stuffed ani-mals). This design feattire was intended tohighlight the distinction between the the-matic and taxonomic relations among the ob-jects in each triad. We note, however, thatthis feature also imposed a confounding he-tween type of relation (thematic vs, taxo-nomic; and animacy (animate vs, inaraniate).A wide range of thematic relations was de-picted over the trials (e.g,, drinking, wash-ing, riding, eating). One additional designfeature bears mention. For half of the triads,the taxononricafly related object was a non-identical member of the same basic level

" Bauer and Mandier acknowieoge this tact in their di.iciission. We note bere that our ques-tion differs in emphasis and scope from tbe qnestion posed by Baner and Mandler. Tbey wereV>rimarily concerned with increasing taxonomic responding in their categorization rask (Baner,personal communication, August 1992); we are primarily interested in examining tbe effect ofintroducing novel labels under conditions closely analogous to the conditions under which wordlearning naturally occurs.


FABLE 1

'I- TRIADS USED IN EXM.KIMENIS 1

Target ra.xoiu)inii ThematicObject ^Itcnuiti'.-c Alternative

Basic-level trials:'Cup Cup Baby (lollShampoo bottle Shampoo bottle Baby dollAirplane .Airplane ("IowiiBed Bed SmmidollShirt Shirt Teddy bearCookie ('ookie ("eiokie monstCT

Superordinatc-level trials:Carrot Tomato RabbitRaincoat Pants Snoop\ dollChair 'Table Minnie Mouse dollCar Boat Adult figurineBrush Comb Raggedy .Aiui dollBottle "Sippy" cup Baby doll

•'Target object and taxoiidinic alternative i)n basic le\'el trials were not identical. 'riu->'differed from one another on a number of cbarat'teristifs, inchidinsi color. >i/.!'. cnid ]iatterninu.

category as the target; for the remaining half,the taxonomically related object was a mem-ber of the same superordinate level categ()r\'as the target.

ProcedureIufants were tested individually in a

laboratory playroom. After a brief periodduring which they became acquainted withthe laboratory and the experimenter, the in-fants were seated at a table across from theexperimenter, either in an iufant seat or inthe parent's lap. The parent, who was pres-ent throughout the session, was instructednot to talk (to either the child or the experi-menter) or to influence in any way thechild's interest in or attention to the stimuli.The procedure consisted of 12 trials, eachof which included a familiarization periodfollowed immediately by an experimentalperiod, described below. On each of the 12trials, a different object triad was presented.The trials were presented in the same fi.xedorder for each infant, beginning with a su-perordinate level trial and alternating be-tween superordinate and basic level trial.sthroughout.

Familiarization period.—Each trial be-gan with a familiarization period, duringwhich time the experimenter pre.sented thechild with the three objects for a given trial(the target object, the taxonomic object, andthe thematic choice object). The child wasencouraged to play freely with the objectsfor a total of 30 .sec. After 30 sec had elapsed,the experimenter removed all three tovs

from the child's view, placing them in herlap below the table.

Experimental period.—To begin eachtrial in the experimental period, the experi-menter secured the child's attention by call-mg the child's name. She then introducedthe target object, placing it alone on the ta-ble, directly in front of the child, but out ofthe child's reach. Next, she placed the tworemaining objects (the taxonomic and the-matic choices) on the table, to the left and tothe right of the target, but within the child'seasy reach. The choice objects were sepa-rated by approximately 10 inches. For eachchild, the left-right placements of the taxo-nomic and thematic alternatives were deter-mined randomly for each trial. Toys wereleft within the child's reach for up to 30 sec.

Children were assigned randomly to ei-ther a No Word or Novel Noiui condition.On each trial in the No Word condition, asthe experimenter introduced the target ob-ject, she pointed to it, saying, "See thisone?" A.s she placed the choice object.s si-multaneously on the table, she said, "Canyou find another one (indicating the twochoices) just like this one (indicating the tar-get)?" Children were encouraged to indicatetheir choices by touching.

In the Novel Noun condition, the in-structions were identical to those in the NoWord condition, except for one crucial dif-ference: On each trial, the experimenter la-beled the target with a nonsense noun, using


a different noun for each trial. For example,she said, "See this X (where X was the novelnoun)? Can you Bnd another X just like tfiisX':-"

Children were given no feedback con-tingent upon their selections. For childrenin both conditions, the experimenter simplysaid, "Thank you " or "Good girl/boy " afterany selection was made. Sessions lasted ap-proximately 15—20 min and were videotapedfor later transcription.

Communicative Development Inven-tory (CDI).—While the child was engagedin the experiment, the parent was asked tocomplete the Communicative DevelopmentInventory (CDI), developed by Bates audher colleagues (Bates, Bretherton, & Snyder,1988) for use with infants. Data from thisstandardized checklist provided informationconcerning each child's productive and re-ceptive vocabulary,

CodinfiFour different measures were derived

from the videotaped sessions. For all four,the videotapes were transcribed with thesound removed to insure that the coders,who were blind to the hypotheses, were alsoblind to the condition in which each childhad participated. The measures included (1)the duration of time spent playing with eachobject in the familiarization period; (2) theobject chosen first during the experimentalperiod (either the taxonomic or the thematicalternative), (3) the duration of time spentplaying with each object during the experi-inental period, and (4) the incidence of the-matic play demonstrated by the child at anytime during a trial. For the duration mea-sures (measures 1 and 3), the coder mea-sured the total accumulated time the childspent playing with each object; durationscores could range from 0 to 30 sec. For theincidence measure (measure 4), the codersimply noted any demonstrations of the-matic relations (e.g., putting a cup to thedoll's mouth) made spontaneously by thesubject at any point during each trial,

A second rater independently coded thevideotaped sessions of 12 subjects, evenlydivided among age groups and conditions,.Agreement between coders was above 95%on each of the four measures, Interrateragreement was computed in the followingmanner: for the duration measures we usedcorrelations to compare raters' judgments ofthe amount of time each child spent playingwith each object; for the object chosen firstand for the incidence measure, we com-

puted the proportion of trials on which cod-ers agreed.

The language inventories were tabu-lated to reveal the number of words that par-ents reported iu their child's productive andreceptive vocabularies.

RESULTS AND DISCUSSION

The results of the Communicative Lan-guage Inventory confirmed that the subjectsin our two age groups fell on opposite sidesof the 50-word boundary that is typicallytaken as a rough index of the onset of thenaming explosion (Benedict, 1979; Gold-field & Reznick, 1990; Nelson, 1973), The15-month-olds had a mean productive vo-cabulary of 10.9 words (ranging from 0 to 39words). In contrast, the mean productive vo-cabulary for the 21-month-olds was 86.6words (ranging from 19 to 175 words).

On the forced-choice task itself, infantsseemed to enjoy playing with the objectsand interacting with the experimenter. Theycompleted an average of 79% of the trials byselecting one of two choice objects duringthe experimesntal period. On the remainingtrials, infants either selected both, or nei-ther, of the objects. These trials are not in-cluded in subsequent analyses. The propor-tion of trials completed hy each subject wassubmitted to a three-way analysis of vari-ance, with condition (No Word vs. NovelNoun) and age (15 vs, 21 months) as be-tween-subjects factors and hierarchical level(basic vs, superordinate) as a within-subjectsfactor. This analysis revealed that 15-month-olds completed more trials (M = ,86; SD =.17) than did 21-month-oIds (M = .72; SD =.22), F(l, 36) = 6.39., p < .02. There were noother main effects or interactions.

The next set of analyses is based onthe object chosen first on each trial duringthe experimental period. The scores were theproportions of trials in which infants firstchose the taxonomically related item, com-puted separately for the sets of basic and su-perordinate level trials. A three-way analysisof variance, with condition (No Word vs.Novel Noun) and age (15 vs, 21 months) asbetween-subjects factors and hierarchicallevel (basic vs. superordinate) as a within-subjects factor revealed an interaction be-tween age and condition, F(l, 36) = 5.1, p< .05. This interaction is depicted in Figure1. Follow-up tests revealed that, at 21months, subjects in the Novel Noun condi-tion chose the taxonomically related item ona greater proportion of trials (M = .82; SD

1230 Child Development1.0

0.8-

Novel Noun

No Word

' p< 05 (tvsfo-tailed)

•*p=.O5 (two-taited)

FIG. 1.—Experiment 1. The mean proportion of taxonomic choices made by subjects as a lanctioiiof age (N = 20 of each age) and condition. * indicates comparisons to chance responding (50%).

= .19) than did their age-rnates in the NoWord condition (M = .61; SD = .15), t{36)= 1.93, p < .05, one-tailed. In contrast, at 15months, there was no apparent influence ofintroducing novel labels: the proportions oftaxonomic choices in the Novel Noun (M =.65; SD = .20) and No Word (M = .69; SD= .16) conditions did not differ significantly.

We also carried out an analysis thattreated items, rather than subjects, as a ran-dom factor. This analysis provided converg-ing evidence for the labeling effect at 21, butnot 15, months of age. For the 21-nionth-oldsubjects, when target objects were labeledwith novel nouns, they elicited a signifi-cantly higher proportion of taxonomic re-sponses (M = .81; SD = .10) than when theobjects received no labels (M = .62; SD =.25), paired t(ll) = 2.29, p < .05, one-tailed.However, for 15-month-old subjects, therewas no effect for labeling. Target objectselicited a comparable proportion of taxo-nomic responses when they were accompa-nied by novel nouns (M = .69; SD = .15)

and by no labels (M = .70. SD = .20), paired= .31. N.S.

To provide a richer depiction of peitor-mance on this task, we examined each in-fant's individual pattern ot response. Foreach infant, we determined the proportionof taxonomic objects selected over the set of12 trials. These data are preserited m Table2, broken down to show the munbers otchildren selecting taxonomic alternatives on0.00-.33; .34-.66; and .67-1.00 of their tri-als. For the purposes ot statistical analysis,we collapsed the two left-most eoluinns, en-abling us to use a Fisher exact test to com-pare the number of children in the NovelNoun and No Word conditions making ta.xo-nomic selections on .66 or tewer trials versus.67 or more trials (Rosenthal & Rosuow,1984). At 15 months, the distribution of theseindividual patterns was comparable in theNovel Noun and No Word conditions; at 21months, the individual patterns observed ineach condition were quite distinct, p < .0.5.Thus, ttie analyses liased on individual pat-


TABLE 2

NUMBER OF INDIVIDUAL SUBJECTS MAKINGTAXONOMIC SELECTIONS FALLING IN THE LOWEST,

MIDDLE, AND HIGHEST THIRDS OF THE POSSIBLERANGE

Experiment 1:L5-month-olds:

No WordNovel Noun ..

21-month-olds;

Novel Noun ..Experiment 2:

No WordNovel Noun

Proportion of TaxonomicSelections

0-.33 .34-.66 .67-1.00

... 0

.... 1

2.... 0

.... 0

.... 0

64

63

83

45

9

7

2

terns of performance converge with thosebased on group means.

Each of these preceding analyses is con-sistent with the hypothesis that it is onlyafter the onset of the naming explosion thatnouns begin to focus infants' attention on ob-ject categories. However, there is an alterna-tive interpretation. Perhaps this particulardependent measure—the object chosenfirst—is not a sensitive enough measure forchildren as young as 15 months. This alter-native warrants consideration, especially inview of the fact that most of the publishedwork regarding infants' sensitivity to the in-troduction of novel nouns has utilized mea-sures based on the duration of time spentattending to objects (e.g., Baldwin & Mark-man, 1989; Markow & Waxman, 1992; Rob-erts & Jacob, in press; Waxman & Heim.1991).

We therefore eonducted a different anal-ysis, leased on the duration of time childrenspent playing with the taxonomic and thethematic alternatives during the experimen-tal period. We derived a proportional mea-sure by dividing (a) the amount of time spentplaying with the taxonomic alternative by ib)the total amount of time spent playing withthe choice objects (the thematic alternative,the taxonomic alternative, or both choice ob-jects). We reasoned that, if nouns focus in-fants' attention on taxonomic relations, theninfants in the Novel Noun condition shouldspend proportionally more time than thosein the No Word condition playing with thetaxonomie alternative.

The proportional data were entered intoa three-way analysis of variance, with condi-tion (No Word vs. Novel Noun) and age (15vs. 21 months) as between-subjects factorsand hierarchical level (basic vs. superordi-nate) as a within-subjects factor. The resultsof this analysis were consistent with thosereported above. The interaction betweenage X condition was again significant, F(l,36) = 5.21, p < .05. As was the case in theprevious analysis, 15-month-olds performedcomparably in the Novel Noun (M = .65;SD = .16) aud No Word conditions {M =.71; SD = .16), f(36) = .75, N.S.; only at21 months did children in the Novel Nouncondition (M = .71; SD = .13) spend pro-portionally more time than those in the NoWord condition (M = .58; SE) = .16) playingwith the taxonomic choice object, f(36) =1.7, p < .05, one-tailed. In addition, therewas a level x condition interaction, F(l, 36)= 5.44, p < .05. At the basic level, infantsin the Novel Noun condition (M = .72; SD= .17) spent a larger proportion of time play-ing with the taxonomically related alter-native than did infants in the No Wordcondition (M = .62; SD = .16); at the super-ordinate level, infants in the Novel Nouncondition (M = .64; SD == .14) spent asmaller proportion of time playing with thetaxonomically related alternative than didinfants in the No Word condition (M = .67;SD = .16). Neither of these pair-wise com-parisons reached significance.

The preceding analyses, based on theobjects chosen first by infants and the dura-tion of time infants spent playing with theobjects, offer converging results. Both setsof analyses reveal in 21-month-olds, but not15-month-olds, a tendency to link novelnouns and taxouomic relations. However,this pattern of results must be interpretedwith some caution. As can be seen iu Figure1, ehildren at both age levels and in bothconditions had a strong tendency to selectthe taxonomic alternatives. This was surpris-ing for three reasons. First, we provided noreinforcement for responding taxonomieally;second, the published literature suggeststhat, early in development, children preferthematic over taxonomic relations (Smiley &Brown, 1979); and third, we had taken careto select object triads with unambiguousthematic relations that would be familiar to15—21-month-old infants.

There are two possible explanations forthe high proportion of taxonomic selections.First, infants may have had a general prefer-ence tor the taxonomic, as opposed to the

1232 Ghild Development

thematic, choice objects. If this were thecase, then they should demonstrate a prefer-ence for the taxonomic alternatives in thefamiliarization period as well. To examinethis possibility, we submitted the durationdata from the familiarization period to athree-way analysis of variance, with condi-tion and age as between-subjects factors andhierarchical level as a within-subjects factor.There were no significant main effects or in-teractions. Averaging across ages and condi-tions, infants demonstrated no overall pref-erences for either the taxonomic (inanimate)(M = 9,54 sec; SD = 2,95) or thematic (ani-mate) (M = 8.25 sec; SD = 3,00) test object;they devoted a comparable amount of timeto each. Thus, infants' tendency to select thetaxonomic alternatives in the test trials can-not be attributed to a general preference forthe taxonomically related (or inanimate) ob-jects themselves.

We next asked whether infants' highlytaxonomic performance in the test trialscould be attributed to their familiarity withthe intended thematic relations. We wou-dercd if the infants, particularly those in theyoungest group, were indeed familiar withthe thematic relations represented in ourtriads.

To address this question, we performeda different analysis. We examined each sub-ject's behavior on each trial for spontaneousdemonstrations of thematic relations (eg,,putting a cup to a doll's mouth). The scoreswere the proportions of basic-level and su-pcrordinate-levcl trials on which auy dem-oustration(s) the thematic relations occurred.We submitted the data for the thematic dem-onstrations to a three-way ANOVA, with ageand condition as between-subjects factorsaud hierarchical level as a within-subjectsfactor. A main effect for age, F(l, 36) = 4.07,p = .05, revealed that 21-month-olds dem-onstrated thematic relations on a greater pro-portion of trials (M = .44; SD = .25) thandid 15-month-olds (M = .32; SD = .16). Amain effect for hierarchical level, F(l, 36) =16.84, p < ,001, revealed that subjects atboth ages demonstrated thematic relationsproportionately more often on supcrordinate(Af = .44; SD = .21) than on basic (M =.31; SD = .19) level trials. In addition, aninteraction between age and hierarchicallevel, F(l, 36) = 4.56, p < .05, revealed that,on basic level trials, 15- and 21-month-oldinfants were equally likely to demonstratethematic relations (M = .29; SD = ,15, IM =.34; SD = ,23, respectively), t{36) = ,85,N.S,; yet on superordinate-lcvel trials, 15-

month-olds demonstrated thematic relationson a smaller proportion of trials (M = ,35;SD = .17) than did 21-month-olds (M = ,53;SD = ,26), ((36) = 3,0, p < .01, one-tailed.This suggests that 15-month-olds were lessinclined than 21-month-olds to demonstratethematic relations spontaneously, particu-larly those thematic relations presented onthe supcrordinate trials. This finding raisesthe possibility that 15-month-olds' taxo-nomic preference on previous analyses mayhave been a consequence of their relativeunfamiliarity with some of the specific the-matic relations represented here.

Finally, there were other relevant, ifless tangible, differences between the per-formances of the older and younger infants,A thorough examination of the videotapedsessions gave the impression that the forced-choice task may have been les,s than ideallysuited for our youngest subjects. For exam-ple, although the subjects complied on mosttrials with the experimenter's request for anobject, it was not uncommon for 15-month-olds to respond automatically, giving littleapparent consideration to which of the twoobjects to offer. Other 15-month-olds dis-played a different type of behavior; theyfailed to respond to the experimenter's re-quest promptly. When they eventually didoffer an object, they seemed to be re-sponding nnorc to her outstretched handthan to her specific verbal request. Thesetypes of behaviors may have introduced suf-ficient "noise" to obscure any existing sensi-tivity to the noun-category linkage. That is,the null effect for 15-month-old subjects mayhave been a function of the performance de-mands placed on them in this task. There-fore, in Experiment 2, we introduce severalprocedural modifications to clarify the tasktor these young subjects.

Experiment 2The purpose of this experiment was to

examine fiirther the infiuence of introducingnovel labels to infants who had not yet com-menced the naming explosion. We intro-duced several modifications to the proce-dure from Experiment 1 in an effort to makethe forced-choice task more suitable for ouryoungest group of subjects.

METHOD

SubjectsTwenty-two 16-month-olds (with a

mean age of 16 months, 22 days, rangingfrom 16 months, 14 days to 17 months, 14


days) were tested; two were excluded be-cause they failed to complete the procedure.There was no significant difference betweenthe mean ages of infants in this experimentand those in the younger group of Experi-ment 1. An equal number of boys and girlsparticipated in each condition (see below).Recruitment procedures and demographicinformation were identical to those de-scribed in Experiment 1. None of the sub-jects in this experiment participated in Ex-periment 1,

StimuliThe stimuli were identical to those used

in Experiment 1.

ProcedureThe procedure differed from that used

in Experiment 1 in four ways. First, to en-courage the infants to examine the objectsmore systematically and to take note of therelations among them, the experimenterdemonstrated both the intended taxouomicand thematic relation during the familiariza-tion period on the first four trials. Second,to help focus the infants' attention on theexperimenter's specific verbal request, sherepeated her request on any trial on whichan infant failed to select a single choice ob-ject within 10 sec. Third, in an effort to re-duce the generally high rate of taxonomicselections, the experimenter asked childrento "find another one" (or "find another X")rather than to "find another one just like thisone" (or "find another X just like this X").F inally, we counterbalanced the order inwhich the trials were presented, such that(a) basic- and superordinate-Ievel trials werepresented in an alternating fashion, and (b)for exactly half of the subjects, the procedurebegan with a superordinate-level trial; forthe remaining subjects, the procedure beganwith a basic-level trial. Counterbalancingwas accomplished by randomly assigningeach of the 10 children in each condition toone of the 12 possible rows in a completeLatin square design.

Familiarization period.—The familiar-ization period was modified in the followingway: On Trials 1-4, while the experimenterpresented the three objects to the child, she

demonstrated both the thematic and the tax-onomic relations between the target and thechoice objects in a standardized fashion. Toconvey a thematic relation, the experimenterdemonstrated the thematic relation appro-priate to the objects. For example, she satthe doll in the chair. To convey a taxonomicrelation, she held the taxonomically relatedobjects next to one another. For example,she touched the two chairs together.^ Theexperimenter secured the child's attentionby saying, "Sec?" as she demonstrated eachrelation. After she had demonstrated both re-lations, the child was allowed to play freelywith all three objects for 30 sec, as had beenthe case in Experiment 1. On Trials 1 and 3,the experimenter demonstrated the thematicrelation before the taxonomic relation. OnTrials 2 and 4, she demonstrated the taxo-nomic before the thematic relation. Becausethe order of presentation for the triads wascounterbalanced, the thermatic and taxo-nomic relations were demonstrated by theexperimenter on diflierent triads across sub-jects in each condition.

For all remaining trials (Trials 5—12),the familiarization period vi'as identical tothat described in Experiment 1. The experi-ment made no demonstrations; she simplyplaced the three objects within the child'sreach for 30 sec.

Experimental period.—As in Experi-ment 1, children were assigned randomly toeither a Novel Noun or No Word condition.There were, however, two procedural modi-fications introduced during the experimentalperiod. The Hrst concerned the experi-menter's instructions to the infants. In theNovel Noun condition, the experimenterpointed to the target, and labeled it, saying."See this? This is an X," She then indicatedthe two remaining objects and asked thechild to "find another X," (In Experiment 1,she had asked children to "find another Xjust like this X.") In the No Word condition,she pointed to the target saying, "See thisone?" She then indicated the remaining ob-jects and asked the child to "find anotherone," (In Experiment 1, she had asked chil-dren to "find another onejusf like this one.")

^ We note here that, although it is a fairly straightforward matter to clearly and unamhigu-ously demonstrate a thematic relation between two objects, thi.s is not the case for taxonomicrelations. It is difflcult to indicate unambiguously a taxonomic relation between objects usingnonverbal means. We chose to simply hold the two taxonomically related objects together in aneffort to direct the infants' attention to the similarities between them. Fortunately, our goalin these demonstrations was to heighten infants' attention to the thematic, not the taxonomic,relations.


The second modification concernedthe timing of the experimenter's verbal re-quests. In this experiment, unlike Experi-ment 1, the experimenter systematicallyrepeated her specific verbal request on alltrials in which 10 sec elapsed before thechild made a clear choice (e.g., if the childoffered neither, or both, of the objects). Forexample, if a child in the No Word conditionfailed to make a clear choice within 10 sec,the experimenter would point to the choiceobjects, saying, "Can you find another one? 'In the Novel Noun condition she would ask,"Can you find another X?"

For children in both conditions, the ex-perimenter said, "Thank you" or "Good girl/boy" after all selections. Sessions were vid-eotaped for later transcription. The proce-dure lasted approximately 15—20 min.

Communicative Language Inventory(CDI).—While the child was engaged in theexperiment, the parent was asked to com-plete the CDI designed for infants.

CodingThe videotaped sessions were coded for

the same measures as in Experiment 1, Asecond rater independently coded the vid-eotaped session of six subjects, three fromeach condition. Agreement between coderswas above 90% on all measures.

RESULTS .4ND DISCUSSION

The data from this experiment supportthe hypothesis that novel nouns focus atten-tion on taxonomic relations even in infantswho have not yet commenced the namingexplosion.

The results of the Communicative Lan-guage Inventory yielded a mean productivevocabulary of 24,6 words, ranging from 0to 74 words. Although 80% of the infantsincluded in this experiment had not yetachieved the 50-word vocabulary boundarythat is typically taken as a rough index ofthe onset of the naming explosion (Benedict,1979; Goldfield & Reznick, 1990; Nelson,1973), four infants (two in each condition)did exceed this index. Therefore, in the fol-lowing analyses, we consider both the entiregroup of subjects and the subset of subieotsproducinp! fe^'er than 50 v/ords.

In this experiment, the choices made by16-month-old infants revealed a clear effectfor novel nouns. In this first set of analyses,the proportion of trials on which infants firstselected the taxonomically related alterna-tive served as the dependent measure. .A.

two-way mixed ANOVA, with condition(Novel Noun vs. No Word) as a between-subjects factor and hierarchical level (basicvs, superordinate level) as a within-subjectsfactor revealed a significant main effect forcondition, F(l, 18) = 7,27, p < .05, illus-trated in Figure 2. Infants in the Novel Nouncondition (M = ,74; SD = .14) selected thetaxonomic alternative on a greater propor-tion of trials than did their age-inates in theNo Word condition (M = ,59; SD = .11).

When we considered items, rather thansubiects, as a random factor, the same pat-tern emerged. When target objects vvere la-beled with a novel noun, they elicited moretaxonomic responses (M = .76; SD = ,21)than when no novel word was presented(M = .58; SD = .19), paired f(ll) - 2,82,p '^ ,008, one-tailed.

We also examined each infant's patternof response hy determining, for each indi-^'idual infant, the mean proportion of taxo-nomic selections made over the set of 12 tri-

Novel Noun

No Word

* p<.05 (twc-tailed)

FIG, 2.—Experiment 2. The mean proportionof taxonomic choices made by subjects (N — 20)as a function ot condition. * indicates comparisousto chance resnondinsi;


als, Ihese data are presented in Table 2,As it! Experiment 1, we compared the num-bers <>f infants in each condition makingtaxonomic selections on 0.00—,66 versus.67-1.00 of their trials. A Fisher exact testrevealed that the distribution of individualpatterns differed in the Novel Noun and NoWord conditions, p = ,03, This finding isconsistent with the argument that NovelNouns focus 15-month-o!ds' attention on ob-ject categories.

In the next analysis, we examine thepossibility that 16-month-olds' sensitivity tothe noun-category linkage derives from thefact that, in this experiment, unlike E' xperi-nient 1, two subjects in each condition hadproductive vocabularies exceeding the 50-word boundary. To determine whether in-fants prior to the onset of the naming explo-sion are indeed sensitive to the impact ofnovel nouns on categorization, we excludedthose subjects (two in each condition) withproductive vocabularies of greater than 50words. A two-way mixed ANOVA, with con-dition as a between-subjects factor and hier-archical level as a within-subjects factor mir-rored the analysis based on the entire groupof subjects. There was a significant main ef-fect for condition, F(l, 14) = 5.65, p < ,05,Infants in the Novel Noun condition made asignificantly higher proportion of taxonomicchoices (M = .75; SD = .15) than those inthe No Word condition (M = ,59; SD = ,11).This constitutes clear support for the hy-pothesis that, even before the onset of thevocabulary explosion, infants are sensitiveti) a linkage between nouns and object cate-gories. This analysis also revealed a maineffect for hierarchical level, F(l, 14) == 8,81,/" < ,05, with infants in both conditions pro-portionately more often selecting taxonomicalternatives at the basic (M = .77; SD == .12)than at the sunerordinate (M - .57- SD =.23) level.

Me next sought to determine whetherthe infants' appreciation of the noun-cate-gory linkage was evident only on those trialson which the experimenter demonstratedboth the intended thematic and taxonomicrelations. We therefore compared infants'performance on the first four trials (dur-ing which the experimenter had explicitlydemonstrated both the thematic and tax-onomic relations among the objects) and thelast eight trials (during which the experi-menter provided no demonstrations). A two-way mixed ANOVA, with condition as abetween-subjects factor and demonstrationstatus (Demonstration vs. No Demonstra-

tion) as a within-subjects factor, revealedonly an effect for condition, F(l, 18) = 4.84,p < .05, echoing the condition effect de-scribed in the preceding analyses. The factthat there was neither a main effect nor anyinteraction involving the demonstration fac-tor indicates that children performed compa-rably on trials with and without demon-stration.

As in Experiment 1, v/e also examinedthe proportion of time infants spent playingwith the taxonomic choice object during theexperimental period. Infants in the NovelNoun condition devoted a mean proportionof .56 (SD = ,15) of their time to the taxo-nomic choice object; those in the No Wordcondition devoted ,48 (SD = .15) of theirtime to the taxonomic choice object. Al-though the difference between these meansis in the expected direction, it did not reachsignificance in a hierarchical level x condi-tion ANOVA.

In the next analysis, we examined in-fants' spontaneous demonstrations of the-matic relations in their play. In the No Wordand Novel Noun conditions, infants sponta-neously demonstrated thematic relations on.49 (SD = ,21) and .42 (SD = ,17) of theirtrials, respectively. A two-way ANOVA, withcondition as a between-subjects factor andhierarchical level as a within-subjects factor,revealed no significant main effects or inter-action. Further analysis revealed that thetendency to demonstrate thematic relationswas comparable on the first four trials (onwhich the experimenter demonstrated taxo-nomic and thematic relations) and on the re-maining eight trials (on which the experi-menter made no demonstrations).

Finally, we compared the percentage oftrials on which infants demonstrated the-matic play in Experiments 1 and 2. To derivethis measure, we divided the number of tri-als on which the child demonstrated a the-matic relation by the total number of trialscompleted by the subject. On average, in-fants in Experiment 2 demonstrated the-matic play on ,45 (SD = ,17) of their trials,while the 15-month-olds in Experiment 1did so on only .32 (SD = ,14) of their trials,F(l, 36) = 8,04, p < .01. This suggests thatthe procedural modifications for Experiment2 were successful in increasing the generalsalience of thematic relations for our youngsubjects.

In sum, the results of this experimentare consistent with the hypothesis that in-fants are sensitive to a linkage between


count nouns and object categories even be-fore the onset of the naming explosion.

General DiscussionThe goal of these two experiments was

to examine the development of an apprecia-tion of the linkage between count nouns andobject categories in infants on either side ofthe period known as the naming explosion.Although previous research has documentedthat novel nouns focus attention on objectsand categories of objects in infants rangingin age from 9 to 15 months (Baldwin & Mark-man, 1989; Echols, 1991; Markow & Wax-man, 1992; Roberts & Jacob, in press; Wax-man & Heim, 1991) and from 30 months to4 years (Markman & Hutchinson, 1984; Wax-man, 1990; Waxman & Kosowski, 1991), theevidence from the intervening period hasbeen scant (but see Bauer & Mandler, 1989,discussed in detail below). The results of theexperiments reported here demonstrate that,for 16-month-old infants, most of whom haveyet to commence the naming explosion (Ex-periment 2), and for 21-month-old infants,most of whom have recently entered thenaming explosion (Experiment 1), novelnouns focus attention on taxonomic relationsamong objects.

This finding is important because it re-veals a nascent appreciation of a linkage be-tween words and object categories in infantswho are at the very onset of language pro-duction. An early appreciation of such a link-age has clear advantages to the infant as aword learner: It insures that, in the contextof learning novel words, infants will payspecial attention to taxonomic, as opposedto thematic, relations among objects. More-over, the fact that an appreciation of thislinkage is evident prior to the onset of thevocabulary explosion suggests that it may beinstrumental in guiding the infants' first ef-forts at mapping words to their meanings.

The results of these experiments go be-yond the established finding that infantsform categories (Bornstein, 1984; Sugarman,1982; Younger & Cohen, 1983) to demon-strate how the introduction of novel wordsinfiuences their categorization. However,these data do not reveal whether infants'early appreciation of a linkage betweenwords and categories is part of their innateendowment or is learned on the basis of stillearlier experience with the language system.Neither do these data reveal whether the fa-cilitative effect of labels is specific to novmsor is generalizable to words from other syn-

tactic categories. Both of these issues aretreated more fully below. Nonetheless, thedata reported here do establish that an ap-preciation of the linkage between words andobject categories is available prior to the on-set of the naming explosion. These datatherefore cast serious doubt on the possibil-ity that there is a developmental discontinu-ity in infants' appreciation of this linkage.

Infants' preferences: Taxonomic or the-matic?—These data also reveal in infants asyoung as 15 months of age a surprising incli-nation to select primarily the taxonomic, asopposed to the thematic, alternative in amatch-to-sample task. This taxonomic focus,which is consistent with the data reportedby Bauer and Mandler, is particularly inter-esting in light of the well-documented find-ing that preschool children prefer thematicover taxonomic relations and that this initialthematic preference shifts toward a prefer-ence for taxonomic relations only later, dur-ing the school-aged years (Smiley & Brown,1979), The data from infants raise the inter-esting possibility that this preference forthematic relations is not yet in place at 15—21 months of age. Indeed, these data callinto question the assumption that a prefer-ence for thematic relations necessarily pre-cedes a preference for taxonomic relations(Greenfield & Scott, 1986; Scott, Serchuk, &Mundy, 1982). In future work, it will beimportant to test this assumption directly,perhaps by extending the match-to-sampleparadigm used by Smiley and Brown (1979)to include infants and toddlers.

It is quite possible that, early in devel-opment, an appreciation of thematic rela-tions among objects lags behind an apprecia-tion of similarities among objects. Noticethat to appreciate any given thematic rela-tion, one must have acquired enough realworld experience regarding the objects inquestion to know about their typical func-tions and the actions in which they can beengaged. These functions and actions mustbe learned; they are not inherent in the ob-jects themselves, but in the relations amongthem. In contrast, an early appreciation ofthe similarities among objects is likely basedon observable characteristics that are inher-ent in the objects themselves (e,g,, shape)(Baldwin, 1989; Daehler, Lonardo, & Bu-lcatko, 1979; Fenson, Cameron, & Kennedy,1988; Landau et al,, 1988), An early appreci-ation of "brute," perceptual similarity mayserve as an initial basis for early object cate-gorization and may set the stage for subse-quent, more sophisticated, systems of cate-


gorization (Gentner & Rattermann, 1991;Quine, 1960).

The role of labels in 16-21-month-olds'object categorization.—The principal focuson these experiments was not to describe in-fants' early preferences for thematic versustaxonomic relations, but to test whether andhow novel nouns influence their object cate-gorization. The fact that 16- and 21-month-old infants selected the taxonomic alterna-tives more frequently in the Novel Nouncondition than in the No Word conditionprovides evidence that nouns focus attentionon categories of objects even in infants whoare jvist beginning to produce words.

How can we reconcile this result withBauer and Mandler's (1989) data on categori-zation in infants ranging in age from 16 to31 months? In our view, the contradictoryresults reported here and by Bauer and Man-dler are most likely attributable to differ-ences in the use of reinforcement. Recallthat the latter authors reinforced their sub-jects for their selections by cheering andclapping; we offered no reinforcement. Thisis an important methodological distinctionbecause, as Bauer and Mandler acknowl-edge, in their task, the "already high levelof taxonomic responding left little latitudefor novel labels to exert an influence oncategorization" (p. 183). That is, the rein-forcement was so successful in elicitingtaxonomic responses that potential effects at-tributable to novel labels were eclipsed.

Moreover, the issue of reinforcementhas important theoretical implications forour questions concerning infants' emergingappreciation of linkages between wordmeaning and object categorization. As wehave argued, infants typically do not havethe benefit of reinforcement trials in the con-te.xt of word learning. Instead, in the typicalword-learning scenario, infants encounternew words, new objects, and manifold pos-sible mappings between the two. We there-fore omitted reinforcement in an effort to ap-proximate the interpretive problem faced byinfants in word learning. And, under thesecircumstances, infants did focus on taxo-nomic relations more in the context of hear-ing a novel noun than in the neutral, uo-word context.

We turn now to reconsider Bauer andMandler's conclusions in light of the evi-dence reported here, Bauer and Mandlerdrew two conclusions based on their sub-jects' strong preference for taxonomic selec-tions, even in the absence of a novel noun.

First, they argued persuasively that it is notexclusively in the context of word learningthat taxonomic relations become salient. Weagree entirely with this conclusion. There isno doubt that an appreciation of taxonomicrelations does not depend on verbal labels.(See Waxman, Shipley, & Shepperson, 1991,for a discussion.) Indeed, the rich litera-ture on categorization abilities in infrahu-mans (Herrnstein, 1984; Premack, 1976) andin prelinguistic infants (Sugarman, 1982;Younger & Cohen, 1983) is testimony to thefact that individuals lacking the linguisticability to produce or reliably interpret nounsare nonetheless adept at forming object cate-gories.

Notice, however, that our argument forinfants' appreciation of a linkage betweennouns and object categories is fully compati-ble with this fact. Our position does not reston an assumption that infants will fail to no-tice object categories except when they havebeen introduced to novel nouns. Indeed, wemake no specific predictions about the baserates of taxonomic responding in a neutral,no-word condition. Instead, our predictionscenter on comparisons of performance withand without novel nouns. If nouns direct in-fants' attention specifically and reliably onobject categories, then subjects should bemore likely to choose taxonomically in thepresence of novel nouns than in their ab-sence, whatever the baseline rate in a neu-tral No Word condition might be. Stated inthis way, it is clear that infants' ability toform object categories in the absence of thenovel nouns does not in itself challenge theclaim that they appreciate a linkage betweennouns and object categories. It is also clearthat infants' greater tendency to choose thetaxonomic alternative in the Novel Nounthan in the No Word condition constitutesstraightforward evidence for their apprecia-tion of a linkage between words and objectcategories.

Second, Bauer and Mandler argued thatlabels (as used by Waxman & Kosowski,1991, and by Markman & Hutchinson, 1984)and explicit reinforcement (as used by Bauer& Mandler) may accomplish the same endfor the young child: both serve as remindersof the task requirements. Although we takeno issue with this very general interpreta-tion, we do add two notes of caution. First,labeling with a noun and explicit reinforce-ment may well achieve the same behavioralends, but the means by which they achieveit are likely to be quite different, indeed.Second, although Bauer and Mandler's in-


terpretation provides an accurate descrip-tion of their own data, it is far too general toprovide a satisfactory explanation for otherfindings in this rapidly growing field of re-search. For example, the argument that bothlabels and explicit reinforcement serve to re-mind children of the task requirements can-not account for the fact that, by 2 years ofage, novel nouns "remind " children effec-tively of the task at hand, but novel adjec-tives fail to do so (Waxman, 1990; Waxman& Kosowski, 1991), The differential effectsof nouns and adjectives (vis-a-vis object cate-gorization) cannot be attributed to generalnotions of "reminding " children or keepingthem "on task. " Therefore, in what follows,we articulate a more precise account of thedevelopment of an appreciation of the link-age between words and object categories.

The development of an appreciation ofthe linkage between count nouns and objectcategories.—A central question in this rap-idly growing field of research is whether thefacilitative effect of novel labels is tied spe-cifically to language or whether it is a moregeneral function of auditory stimulation (seeBaldwin & Markman, 1989; Roberts & Ja-cob, in press; Waxman & Balaban, 1992).The answer to this question appears to de-pend crucially on the developmental statusof the subjects under consideration. In aneffort to specify more precisely the devel-opment of an appreciation of the linkagebetween count nouns and object categories,we tentatively suggest the following pro-gression.

Early in development, infants' visual at-tention is augmented by what appear to bevery general sensory and/or perceptual fac-tors rather than by specifically linguisticones. Throughout infancy, the distinctive in-tonational contours characteristic of infant-directed speech are especially effective inarousing and sustaining infants' attention(Fernald, 1992). Moreover, in the first 6months, infants' visual attention is alsoheightened when objects are presented inconjunction with moderate auditory stimula-tion (Kaplan, Fox, Schueneman, & Jenkins,1991; Mendelson & Haith, 1976; Paden,1975; Self, 1975). Initially, then, general au-ditory factors (rather than specifically lin-guistic ones) appear to intensify infants' gen-eral visual interest (rather than their interestin objects or categories of objects in par-ticular).

In the latter half of the first year, a morespecific pairing becomes evident as infants

begin to single out words from other, moregeneral sources of auditory input. By 9—12months of age, infants focus more on objectsand categories of objects in the presence ofnovel words than in their absence (Baldwin& Markman, 1989; Markow & Waxman,1992; Waxman & Balaban, 1992; Waxman &Heim, 1991). Although this focus appearsto be related to language, per se, rather thanto more general auditory factors (but seeRoberts & Jacob, in press), infants do not yetmake systematic distinctions among wordsfrom various form classes. Indeed, infants atthis point tend to interpret most words, inde-pendent of their syntactic status, as referringto objects or categories of objects, Waxmanand her colleagues (Markow & Waxman,1992; Waxman & Heim, 1991) have recentlyreported that at 12 and 13 months of age in-fants interpret words from diverse linguisticcategories (e.g,, nouns, adjectives) as refer-ring to object categories. Therefore, prior tothe onset of the naming explosion, infantsappear to assume that there is a general link-age between words (not specifically countnouns) and object categories.

The evidence suggests that by 2 yearsof age children have clearly learned to mapparticular linguistic forms onto particulartypes of meaning. (See Pinker, 1984, or Mac-namara, 1982, for an analysis of how suchmappings might be learned.) For example,2-year-old children expect that object cate-gories will be marked linguistically by countnouns (Brown, 1957; Markman & Wachtel,1988; Waxman, 1990; Waxman & Kosowski,1991), that substances will be marked bymass nouns (Prasada, in press; Soja et al.,1991), that individuals will be marked byproper nouns (Gelman & Taylor, 1984; Hall,1991; Katz, Baker, & Macnamara, 1974); andthat various properties of objects will bemarked by modifiers (Taylor & Gelman,1988; Waxman, 1990). It is interesting tonote that even at this point, when childrenare clearly capable of using syntactic infor-mation as a cue to meaning, they do not doso invariably. Instead, the tendency to usesyntactic information is affected consider-ably by the child's existing lexical and con-ceptual knowlege (Waxman et al,, 1991), Ifchildren have already acquired a count nounlabel for an object, then they are more likelyto use syntactic information in interpretingsubsequent words applied to that object.However, if they have not yet acquired acount noun for the object, they tend to relyon an earlier pattern of behavior; they tendto interpret any word applied to that object


(whether it is a count noun, a proper noun,or an adjective), as referring to an objectcategory, typically at the basic level (Hall,1991; Hall & Waxman, in press; Hall, Wax-man, & Hurwitz, in press; Markman &Wachtel, 1988).

Clearly, more work will be required totest this proposed account of the develop-ment of an appreciation of a specific linkagebetween count nouns and object categories.The existing evidence suggests that a na-scent appreciation of the linkage is in placeprior to the onset of the naming explosion,that it may serve as a general guide to lexicalacquisition, and that it will become increas-ingly specific over the course of develop-ment.

References

Baldwin, D. A. (1989). Priorities in children's ex-pectations about object label reference: Formover color. Child Development, 60, 1291—1306.

Baldwin, D. A,, & Markman, E, M, (1989). Estab-lishing word-object relations: A first step.Child Development, 60, 381-398,

Bates, E,, Bretherton, 1., & Snyder, L, (1988).From first words to grammar: Individual dif-ferences and dissociable mechanisms. NewYork: Cambridge University Press.

Bauer, P. J,, & Mandler, J. M. (1989). Taxonomiesand triads; Conceptual organization in one-to two-year-olds. Cognitive Psychology, 21,156-184.

Benedict, H, (1979). Early lexical development:Comprehension and production. Journal ofChild Language, 6, 183-200,

Bloom, P, (1990), Syntactic distinctions in childlanguage. Journal of Child Language, 17,343-355.

Bornstein, M, H, (1984), A descriptive taxonomyof psychological categories used by infants. InC. Sophian (Ed.), Origins of cognitive skills.Hillsdale, NJ: Erlbaum,

Brown, R. (1957). Linguistic determinism and thepart of speech. Journal of Abnormal and So-cial Psychology, 55, 1-5.

Carey, S. (1978). The child as word learner. InM. Halle, J, Bresnan, & G, A. Miller (Eds,),Linguistic theory and psychological reality.Cambridge, MA: MIT Press,

Daehler, M, W., Lonardo, R,, & Bukatko, D.(1979), Matching and equivalence judgmentsin very young children. Child Development,50, 170-179,

Dromi, E, (1987), Early lexical development.Cambridge: Cambridge University Press,

Echols, C. (1991). Infant's attention to objects andconsistency in linguistic and non-linguisticcontexts. Paper presented at the biennialmeeting of the Society for Research in ChildDevelopment, Seattle, April,

Fenson, L,, Cameron, M, S,, & Kennedy, M.(1988), Role of perceptual and conceptualsimilarity in category matching at age twoyears. Child Development, 59, 897-907,

Fernald, A. (1992), Human maternal vocalisationsto infants as biologically relevant signals; Anevolutionary perspective. In J, H. Barkow,L, Cosmides, & J. Tooby (Eds.), The adaptedmind: Evolutionary psychology and the gen-eration of culture. New York: Oxford Univer-sity Press.

Celman, S. A., & Byrnes, J, P, (Eds.). (1991). Per-spectives on language and thought: Interrela-tions in development. Cambridge: Cam-bridge University Press,

Gelman, S, A., & Taylor, M. (1984), How two-year-old children interpret proper and commonnouns for unfamiliar objects. Child Develop-ment, 55, 1535-1540,

Gentner, D. (1982). Why nouns are learned beforeverbs: Linguistic relativity versus natural par-titioning. In S, Kuczaj (Ed.), Language devel-opment: Vol. 2, Language, thought, and cul-ture (pp. 301-334). Hillsdale, NJ: Erlhaum.

Gentner, D., & Rattermann, M, J, (1991). Lan-guage and the career of similarity. In S. A,Gelman & J. P. Byrnes (Eds,), Perspectives onlanguage and thought: Interrelations indevelopment (pp, 225-277). Cambridge: Cam-bridge University Press,

Goldfield, B, A,, & Reznick, J, S, (1990), Early lex-ical acquisition: Rate, content, and the vocab-ulary spurt. Journal of Child Language, 17,171-183.

Gordon, P. (1985). Evaluating the semantic cate-gories hypothesis: The case of the count/massdistinction. Cognition, 20, 209-242,

Greenfield, D. B,, & Scott, M, S. (1986). Youngchildren's preference for complementarypairs: Evidence against a shift to a taxonomicpreference. Developmental Psychology, 22,19-21.

Hall, D. G. (1991). Acquiring proper names forfamiliar and unfamiliar animate objects; Two-year-olds' word-learning biases. Child Devel-opment, 62, 1142-1154.

Hall, D. G., & Waxman, S. R, (in press). Assump-tions ahout word meaning: Individuation andbasic-level kinds. Child Development.

Hall, D, G., Waxman, S, R,, & Hurwitz, W, R, (inpress). How two- and four-year-old childreninterpret adjectives and count nouns. ChildDevelopment.

Herrnstein, R. J, (1984). Objects, categories, anddiscriminative stimuli. In H, C. Roitblat,


T, G. Bever, & H, S, Terrace (Eds,), Animalcognition: Proceedings of the Harry FrankGuggenheim Conference, June 2-4, 1982(pp, 233-261). Hillsdale, NJ; Erlbaum.

Kaplan, T., Fox, K,, Scheuneman, D., & Jenkins,L, (1991), Cross-modal facilitation of infant vi-sual fixation; Temporal and intensity effects.Infant Behavior and Development, 14,83-109,

Katz, N., Baker, E,, & Macnamara, J. (1974).What's in a name? A study of how childrenlearn common and proper names. Child De-velopment, 45, 469-473.

Landau, B,, Smith, L, B,, & Jones, S, S. (1988), Theimportance of shape in early lexical learning.Cognitive Development, 3, 299-321.

Macnamara, J. (1982). Names for things. Cam-bridge, MA: MIT Press.

Markman, E. M. (1992). Constraints on wordlearning: Speculations about their nature, ori-gins, and domain specificity. In M. R. Cunnar& M. P, Maratsos (Eds,), Minnesota sympo-sium on ehild psychology (Vol. 25). Hillsdale,,NJ: Erlbaum.

Markman, E. M., & Hutchinson, J. E. (1984).Children's sensitivity to constraints on wordmeaning: Taxonomic vs. thematic relations.Cognitive Psychology, 16, 1-27.

Markman, E. M., & Wachtel, G. A, (1988). Chil-dren's use of mutual exclusivity to constrainthe meanings of words. Cognitive Psychol-ogy,20, 121-157.

Markow, D. B,, & Waxman, S, R. (1992), The in-fluence of labels on 12-month-olds' categoryformation. Paper presented at the Eighth In-ternational Conference on Infant Studies,Miami,

McPherson, L, (1991), "A little" goes a long way;Evidence for a perceptual basis of learningfor the noun categories COUNT and MASS.Journal of Child Language, 18, 315-338.

McShane, J. (1979). The development of naming.Linguistics, 17, 879-905.

Mendelson, M. J,, & Haith, M, M. (1976). The re-lation between audition and vision in the hu-man newborn. Monographs of the Society forReseareh in Child Development, 41(4, SerialNo, 167).

Nelson, K. (1973). Some evidence for the cogni-tive primacy of categorization and its func-tional basis, Merrill-Palmer Quarterly of Be-havior and Development, 19, 21-39.

Nelson, K, (1988), Constraints on word learning?Cognitive Development, 3, 221-246,

Paden, L. (1975). The effects of variations of au-ditory stimulation (music) and interspersedstimulus procedures on visual attending be-havior in infants. Monographs of the Societyfor Researeh in Child Development, 39(158,Serial No, 158), 29-41,

Pinker, S, (1984), Language learnability and lan-guage development. Cambridge, MA; Har-vard University Press.

Prasada, S. (in press). Young children's linguisticand non-liuguistic knowledge of solid sub-stances. Cognitive Development.

Premack, D. (1976), Intelligenee in ape and man.Hillsdale, NJ; Erlbaum,

Quine, W, V. (1960). Word and object. Cambridge,MA; MIT Press.

Roberts, K., & Jacob, M. (in press). Linguistic vs,attentional infiuences on nonlinguistic cate-gorization in 15-month-old iufants. CognitiveDevelopment.

Rosenthal, R,, & Rosnow, R. (1984). Essentials ofbehavioral research: Methods and data anal-ysis. New York; McGraw-Hill.

Scott, M. S., Serchuk, R., & Mundy, P. (1982). Tax-onomic and compleuientary picture pairs;Ability in two- to five-year-olds. InternationalJournal of Behavioral Development, 5, 243—256.

Self, P. A. (1975). Control of infants' visual at-tending by auditory and interspersed stimula-tion. Monographs of the Society for Researchin Child Development, 39(2, Serial No. 158),16-28.

Smiley, S, S., & Browu, A. L. (1979). Couceptualpreference for thematic or taxonomic rela-tions; A nonmonotonic trend from preschoolto old age. Journal of Experimental ChildPsychology, 28, 249-257.

Soja, N. (1992). Inferences about the meaningsof nouns; The relationship between percep-tion and syntax. Cognitive Development, 7,29-45.

Soja, N., Carey, S., &: Spelke, E. (1991). Outologi-cal categories guide young children's induc-tions about word ineaning; Object terms andsubstance terms. Cognition, 38, 179—211.

Sugarmau, S, (1982), Developmental changes inearly representational intelligence; Evidencefrom spatial classification strategics and re-lated verbal expressions. Cognitive Psychol-ogy, 14, 410-449.

Taylor, M., & Gelman, S. A. (1988). Adjectives andnouns; Children's strategies tor learning newwords. Child Development, 59, 411-419.

Valian, V. (1986). Syntactic categories iu thespeech of young children. DevelopmentalPsychology, 22, 562-579.

Waxman, S. R. (1990), Linguistic biases and theestablishment of conceptual hierarchies; Evi-dence from preschool children. Cognitive De-velopment, 5, 123-150.

Waxman, S. R, (1991). Convergences between se-mantic and conceptual organization in thepreschool years. In S. A. Gelman & J. P.Byrnes (Eds,), Perspectives on language andthought: Interrelations in development (pp.


107-145). Cambridge; Cambridge UniversityPress.

Waxmau, S, R,, & Balaban, M, T. (1992), The in-fluence of words vs. tones on infants' c;ategori-zation. Paper presented at the Eighth Interna-tional Conference on Infant Studies, Miami,

Waxman, S, R,, & Heim, L, (1991, April), Nounshighlight category relations iu 13-month-olds,Paper presented at the annual meeting of theSociety for Research in Child Development,Seattle.

Waxman, S, R,, & Kosowski, T, D. (1991). Nounsmark category relations: Toddlers' and pre-schoolers' word-learning biases. Child Devel-opment, 61, 1461-1473,

Waxman, S, R,, Shipley, E. F,, & Shepperson, B.(1991). Establishing new subcategories; Therole of category labels and existing knowl-edge. Child Development, 62, 127-138.

Younger, B, A,, & Cohen, L. B, (1983), Infantperception of correlations among attributes.Child Development, 54, 858-867.

the development of a linkage between count nouns and object … · 2017. 6. 15. · however,...

Documents