new acquiring proper nouns for familiar and unfamiliar animate...

Acquiring Proper Nouns for Familiar andUnfamiliar Animate Objects: Two-Year-Olds'Word-learning Biases

D. Geoffrey HalJHarvard University

HALL, D. GEOFKHEY, Acquiring Proper Nouns for Familiar and Unfamiliar Animate Objects:Two-Year-Olds' Word-learning Biases. CHILD DEVELOPMENT, 1991, 62, 1142-1154. In 2 studies,2-year-old children learned a novel word modeled as a proper noun (e.g., "This is Zav") for ananimate stuffed toy. Children who learned the word for a familiar object (i.e., one for which theyknew a basic-level count noun for Ae kind) interpreted the word appropriately as a proper nounreliably more often than chiidren who learned {he word for an unfamiliar object (i.e., one forwhich they did not know such a count noun). When the creature was familiar, children typicallyinterpreted the novel word as if it were a proper noun referring uniquely to the labeled individ-ual. When the animal was unfamiliar, children frequently interpreted the word as if it were acount noun referring to a kind of object. Children's spontaneous comments during the tasksprovided striking additional evidence that their interpretations of the proper noun varied withthe familiarity of the object. The results suggest that young children's sensitivity to the form classof proper nouns is affected by the familiarity of the referent object. The findings are discussed interms of interpretative biases in word learning, v

Upon hearing a new word for an object, Previous research has shown that 2-children must decide whether it is a proper year-old chiidren are sensitive to both se-noun that refers to the individual (e.g., "Gar- mantic and syntactic cues in decidingfield") or a count noun that refers to a kind whether a novel word applied to an objectof object of which the individual is a mem- is a proper noun. In a seminal study by Katz,ber (e.g., "cat"). To make this decision, chil- Baker, and Macnamara (1974), 2-year-oldsdren may rely on two cues that signal when heard a uovel word for a familiar object (i.e.,a proper noun, rather than a count noun, is an object for which they likely knew a countbeing used to pick out an object. The first of noun for the kind). The word was modeledthese cues is semantic: properties of certain either as a proper noun (e.g., "This is Zav")kinds of objects make them good candidates or as a count noun (e.g., "This is a zav"), andfor receiving proper nouns. The individual- the object was either a doll or a block. Iu theity of the members of many kinds of animate first phase of the study, children saw twoobjects, like people and pets, is important, objects (two dolls or two blocks), and heardbut the individuality of members of other one of them labeled. Children then per-kinds, such as pencils or socks, is uot. A sec- formed a series of actions with "Zav" or "aond cue is syntactic: certain linguistic mark- zav," including feeding the doll or puttingers make some words better candidates for the block in a house. Katz et al. inferred chiJ-being proper nouns. In English, proper dren's interpretation of the new word fromnouns typically are not preceded by deter- the proportion of times children chose theminers (e.g., we say "Garfield"), whereas named object when performing the variouscouut nouns may be preceded by deter- actions. A consistent use of the named toyminers (e.g., we say "a cat" or "the pet"). led Katz et al. to infer that a child had made

This research was supported by a Centennial fellowship from the Natural Sciences andEngineering Research Council of Canada to the author, and by a grant from the Harvard Univer-sity Faculty of Arts and Sciences Biomedical Fund awarded to Sandra Waxman, I am grateful tothe directors, teachers, parents, and children at the following centers for their invaluable assis-tance: Concord Children's Center, Serendipity Child Development Center, Soldier's Field DayCare, and Peabody Terrace Day Care, all of the Greater Boston area. I thank Sandra Waxmanand three anonymous reviewers for comments on this manuscript Portions of this research werepresented at the Stanford Child Language Research Forum, Stanford, CA, April 1990. Requestsfor reprints may be sent to the author, who is now at the Medical Research Council, CognitiveDevelopment Unit, 17 Gordon St., London WCIH OAH, England.

[Child Development, 1991, 62,1142-1154. © 1991 by the Society for Research in Child Development, Inc.All rights reserved. 0(K)9-3920/91/6205-0009SO!.()0]

D. Geoffrey Hall 1143

a proper noun interpretation. Randomchoosing between the toys was viewed asconsistent with a count noun interpretation.

Katz et al. (1974; with additional datareported in Macnamara, 1982) found thattheir 2-year-old girls, but not boys, used bothsemantic and syntactic cues in interpretingthe new words. Specifically, they found thatthe syntactic form of the word (the presenceor absence of a determiner) affected girls'selections for the dolls, but not for theblocks. Girls who were taught a proper nounfor a doll frequently selected the named doll(on 71%-75% of trials). When taught a countnoun for a doll, girls chose randomly be-tween the two dolls, showing a response pat-tern that is consistent with a count noun in-terpretation. Girls who learned either aproper noun or a count noun for a blockchose randomly between the two objects.Katz et al. (1974) interpreted their results asshowing that very young children know thatthe individuality of members of certainkinds (e.g., dolls) is important, and that thisindividuality is marked linguistically in aparticular way (i.e., through the use ofproper uouns).

Gelman aud Taylor (1984) replicatedthe results from Katz et al. (1974) in a studythat differed in several ways. First, theytested subjects who were, on average, 6months older than those in the Katz et al.study. Gelman and Taylor also elaboratedand standardized the test procedure in sev-eral important ways. The change most rele-vant to the present work was that Gelmanand Taylor used unfamiliar toys (i.e., toys forwhich children likely knew no count nounfor the kind). They pointed out that the chil-dren in the Katz et al. study probably alreadyhad known that the objects were called"dolls" or "blocks." By switching to objectsfor which children knew no kind term, Gel-man and Taylor wanted to eliminate the pos>sibility that children would be confused be-cause of a preference to reject a new labelfor an object with a known label.

With these changes to the procedure,Gelmau and Taylor (1984) replicated theKatz et al. (1974) findings by showing againthat for surrogates of animate objects (nowunfamiliar stuffed animals), children (nowboth girls and boys) frequently made aproper noun interpretation of a word mod-eled as a proper noun. On 69% of trials, chil-dren selected the named object. Also consis-tent with the Katz et al. finding, childrenmade a count noun interpretation of a word

modeled as a count noun for one of thesestuffed animals; that is, they selected ran-domly between the two members of thenamed kind. For the inanimate toys, chil-dren did uot make a proper noun interpreta-tion in either syntactic condition.

The results from the Gelman and Taylor(1984) and Katz et al. (1974) studies suggestthat 2-year-old children should tend to inter-pret novel words as proper nouns when theyare applied to either familiar or unfamiliarobjects, provided the words are (1) appliedto animate objects and (2) marked syntacti-cally as proper nouns. However, other re-cent experimental results have shown thatfamiliarity infiuences children's interpreta-tions of a uovel word applied to an object.Specifically, there is evidence that childrenare more likely to respect syntactic cues tomeaning when interpreting novel words forobjects that are familiar (i.e., objects forwhich they know a count noun for the kind)than when they are unfamiliar (i.e., those forwhich they do not).

Familiarity effects in word learn-ing.—The discussion of familiarity effects inword learning can be broken down into anexamination of two biases that appear toguide children's interpretations of novelwords. The first of these is the object-kindbias, and it concerns children's interpreta-tions of novel words applied to unfamiliarobjects. Several researchers have presentedevidence that young children expect a novelword applied to an unfamiliar object to referto a kind of object, rather than to a set ofobjects having something else in common,such as a thematic link, a color, a substance,or a part (e.g., Baldwin, 1989; Landau,Smith, & Jones, 1988; Markman & Hutchiu-sou, 1984; Markman & Wachtel, 1988; Soja,1990; Soja, Carey, & Spelke, in press; Wax-man & Gelman, 1986). A typical finding fromthis research is that young children extend anovel count noun applied to an unfamiliarobject to other objects of the same kind. Animportant question raised by this work iswhether children initially hypothesize thata count noun applied to an nnifamiliar objectrefers to a particular attribute such as shape(e.g.. Landau et al., 1988) or whether theythink that the word refers to a kind of objectthat may be independent of shape (see Sojaet al., in press). Much of the evidence to dateis consistent with both possibilities.

Some of the strongest demonstrations ofthe object-kind bias have come from studiesthat show that it operates irrespective of the

1144 ChUd Development

grammatical form class of the novel wordand is, therefore, not restricted to the learn-ing of count nouns. For example, Soja et al.(in press; also Soja, 1990) present evidencethat suggests that very young word learners(2-year-olds) will override syntactic evi-dence wheu interpreting words of differingsyntactic form applied to unfamiliar objects.These authors showed that 2-year-olds' in-terpretations of novel words for unfamiliarobjects did not differ, whether the syntaxsuggested a count noun interpretation ("Thisis a blicket"), a mass noun interpretation("This is blicket"), or either such interpreta-tion ("This is my blicket"). Under all syntac-tic conditions, children interpreted thewords as referring to a kind of object.

Whereas the object-kind bias proposesthat young children prefer to interpret novelwords applied to unfamiliar objects as refer-ring to a kind of object, another proposedbias concems children's interpretations ofwords applied to objects for which a word isalready known. Argued to be a feature inher-ent in language design, the contrast princi-ple posits that different words mean differ-ent things. Clark (1983, 1987, 1988) hassuggested that this principle provides an im-portant source of constraint on children'sword meaning hypotheses during lexical ac-quisition. Expecting that no two words willhave the same meaning, children should in-terpret a new word applied to an object dif-ferently than any previously learned wordfor the object.

The contrast bias hypothesis has gener-ated some controversy (e.g., Gathercole,1987). Much of this controversy stems fromresearch showing that children are willing toapply two words to the same object or smallgroup of objects. As Clark has pointed out,however, it is impossible to determine inmany of these putative counterexampleswhether the word learner thinks the twowords denote kinds with identical or merelyoverlapping sets of members. If the denota-tions were merely overlapping, tiien contrastwould not be violated (see Clark, 1988,1990; Gathercole, 1987,1989, for interestingdialogue ou issues surrounding the contrastbias). In addition, Clark has summarized thesizable body of evidence in support of con-trast. Some of this evidence comes from

findings that children will preempt an inno-vative regularization (e.g., bringed, comed)when they subsequently learn the conven-tional form (e.g., brought, came). Other evi-dence in support of contrast has emergedfrom experiments with young word learners,and two recent examples are particularly rel-evant to the present research.

First, Taylor and Gelman (1988) havedemonstrated that 2-year-olds will interpreta novel count noun for an object differently,depending on whether the referent object isunfamiliar (i.e., has no known kind term) orfamiliar (i.e., has a known kind term). Chil-dren saw a set of four objects, including twofrom a single kind. One member of this pairwas labeled with a count noun. When thelabeled object was unfamiliar (a monster),children interpreted the vi'ord as referring toa kind of object, consistent with the object-kind bias; that is, they extended the novelcount noun to both objects of the same kind.When the object was familiar (a dog), how-ever, children seemed to interpret the wordas referring to a subordinate kind of object;in other words, they restricted application ofthe novel count noun to the named memberof the pair only.

Following up on this work, Taylor andGelnian (1989) found evidence to supporttheir claim that children had interpreted thecount noun for a familiar object as a subordi-nate kind term. These follow-up studieswere similar in method to the earlier studies,but now the authors varied the similarity oftiie pair of familiar objects of the same kind.Half the children saw a pair of similar exem-plars (two dogs, both terriers); half saw a dis-similar pair (two dogs, one a terrier, one abasset hound). The principal finding wasthat when the label was applied to a memberof the similar pair, children readily extendedthe new count noun to both objects; whenthe label was applied to a member of thedissimilar pair, children restricted the newlabel to the named object only. Because chil-dren restricted their selections to tihe namedobject when the pair was dissimilar but notwhen it was similar, the authors argued thatthe children were making a subordinatekind interpretation of the novel count noun(presumably a specific kind of dog.).'

^ Actually, the data are somewhat more complicated, because in a subsequent study (Experi-ment 3), Taylor and Gelman (1989) found that ehildren restrieted a novel count noun to thenamed object even when the two exemplars were similar (they were a male and a female polarbear). The authors ruled out the possibility that gender was being treated as a subordinate kinddistinction (Experiment 4), but they admitted some surprise at the discrepancy between thefindings of dieir Experiments 1 and 3.

A second source of experimental evi-dence consistent with contrast comes fromMarkman and Wachtel (1988), who foundthat 3-year-olds made different interpreta-tions of" novel labels depending on whetherthe labeled objects were unfamiliar (i.e., hadno known kind terms) or familiar (i.e., hadknown kind terms). These authors demon-strated that the children would overrideclear syntactic cues in order to interpret thefirst word for an unfamiliar object as refer-ring to the kind of object, again consistentwith the object-kind bias. Children whoheard "See this? It's pewter" applied to anunfamiliar object (pewter tongs) often inter-preted the word as an object kind term byextending tbe word to other tongs (not madeof pewter). Those children who learned thenew label for a familiar object (i.e., a cup),however, used the syntax appropriately tointerpret the word as referring to the kindof substance the object was made of. TheMarkman and Wachtel (1988) results are, infact, consistent with a more stringent versionof the contrast bias called the mutual exclu-sivity bias. The mutual exclusivity bias pro-poses that children will avoid allowing notonly two words with the same meaning fora single object but also two kind terms forthe same object. Because children in Mark-man and Wachtel (1988) avoided interpret-ing a second word for an object with a knownkind term as another kind term, the resultssupport the mutual exclusivity bias (see Au& Glusman, 1990, for discussion of this bias).

The preceding results suggest that thereis indeed a familiarity effect in word learn-ing. Young children expect a novel wordlearned for an unfamiliar object to refer to akind of object, even if this interpretation isnot consistent with the word's syntactic form(Soja, 1990; Soja et al., in press). When theobject is familiar (i.e., children already knowa kind term for the object), the object-kindbias no longer guides children's interpreta-tion and, according to the contrast bias, theyseek an alternative meaning for the novelword. Among other possibilities, this inter-pretation may be a different kind of object(e.g., a subordinate kind) if the word ismarked syntactically as a count noun (Taylor& Gelman, 1988, 1989), or it may be a sub-stance or property if the word is syntacticallya mass noun or an adjective (Markman &Wacbtel, 1988).

A familiarity effect in the acquisition ofproper nouns?—Because there appears tobe a iamiliarity effect in acquiring countnouns (Taylor & Gelman, 1988, 1989) andmass nouns (Markman & Wachtel, 1988),

D. Geoffrey HaJI 1145

there is reason to suspect that there is also afamiliarity effect in acquiring proper nouns.As noted earlier, however, the previous twostudies of the acquisition of proper nounshave demonstrated 2-year-olds' sensitivity toboth semantic and syntactic cues when thereferent objects are either unfamiliar (Gel-man & Taylor, 1984) or familiar (Katz et al.,1974). Unf"ortunately, comparing the resultsof the two studies directly is problematic be-cause the subjects differed in mean age andthe procedures were not identical. A directcomparison of the acquisition of propernouns for unfamiliar and familiar objects,therefore, would be revealing. Specifically,if a novel proper noun were applied to anunfamiliar animate object, children might bebiased to interpret the word as referring to akind of object even though the syntax calledfor a proper noun interpretation (accordingto the object-kind bias). If a novel propernoun were applied to a familiar animate ob-ject (i.e., one for which a count noun for thekind were known), however, children might(according to the contrast bias and also themutual exclusivity bias) assume that theword did not refer to a kind of object, andbe more likely to interpret it appropriatelyas a proper noun referring to a unique indi-vidual.

To test directly for such a familiarity ei-fect in 2-year-olds' learning of proper nouns,two studies were designed. As noted earlier,the two previous proper-count noun studiesfound children had a clear tendency to makeproper noun interpretations only when thestimuli were (surrogates of) animate objectsand when the words' syntactic frames calledfor a proper noun interpretation (e.g., "Thisis Zav"). In light of these findings, the fol-lowing studies compared the interpretationof novel proper nouns for animate stuffedobjects differing in familiarity.

Experiment 1

Method

Subjects.—Thirty children from pre-schools serving middle- and upper-middle-class families participated, 15 boys and 15girls (2-3 to 3-0; mean age 2-8). Ten childrenwere assigned to each of three conditions,with approximately equal numbers of boysand girls.

Materials.—In each condition, fourstuffed toys were used. In the Unfamiliarcondition, the toys were selected so that thechildren would know no kind label for eithermember of the pair from which the to-be-Iabeled toy would be selected. There were

1146 Child Development

two unfamiliar monster-like animals and twodistractor creatures, each approximately 4inches high. The monster-like animals camefrom a collection of toys called "CuddleeUglees," were covered with blue fur, hadblue arms and legs, and had large mouthswith teeth and a tongue sticking out. Theylooked identical except for their dress,which consisted of a hooded cape. Thesegarments were made of two distinct materi-als; one was purple and black, the other waswhite with thin lines of various colors. Thedistractor items included one familiar ani-mal (a yellow rabbit), and one unfamiliarmonster-like animal of a different kind ofGuddlee Uglee (it was pink and blue withyellow horns and a large green nose).

As an informal check that the childrenknew no label for the unfamiliar toys, six 4-year-old children saw the creatures andwere asked (1) whether they had ever seenthe things before and (2) whether they knewwhat kind of thing the objects were. None ofthe children admitted ever having seen thetoys before, and none volunteered a kindterm for the toys. These findings providedreassurance that the 2-year-olds in this studylikely knew no kind term for the objects.However, none of the 2-year-olds in thestudy was asked explicitly to label the ob-jects, for reasons noted by Gelman and Tay-lor (1984): children at this age will readilyoverextend known labels to objects of differ-ent kinds, even when the labels are inappro-priate (Clark, 1973), and a suitable interpre-tation of these overextensions is difficult tomake (Rescoria, 1980).

In the two other conditions (Familiarand Familiar—Explicit), toys were selectedso that the children likely knew a kind termfor both members of the pair from which theto-be-labeled toy would be selected. Thetoys were two cats and two distractors. Thecats were orange and striped, and theylooked identical except for their dress,which again consisted of hooded capes ofthe same colors that were used to dress theUnfamiliar monsters. The distractors werethe same as those in the Unfamiliar con-dition.

In addition, a shoebox was used both asa holding place for the creatures during partsof the study, and also as a prop during someof the specific activities.

Procedure.—The procedure wasadapted from Gelman and Taylor (1984).Ghildren were tested individually in a pri-vate room adjoining their classroom or in a

quiet corner of the class. There were twoparts to the experiment: a naming portion,followed by a testing poition.

The purpose of the naming portion wasto teach the children a novel proper noun fora toy. First, the box was placed in front ofthe child, and one of the four toys was re-moved and shown to the child.

In the Unfamiliar condition, the childsaw one of the two monsters and was invitedto handle the creature briefiy. The experi-menter used only the word "this" or "thisone" to refer to the object The experimenterthen labeled the monster with a proper nounin the following manner: "See? This is Zav."The child was asked to repeat the label, andtlie experimenter then repeated the label atleast five times while the child played withthe toy (e.g., "Look at Zav," "Let's put Zavon the ground here").

In the Familiar condition, one of the twocats was shown, and the experimenter thenfollowed the same procedure as in the Unfa-miliar condition.

The Familiar—Explicit condition wasthe same as the Familiar condition, with oneexception. Instead of referring to the cat as"this" or "this one," the experimenter usedthe basic-level count noun to refer to the ob-ject: "this cat," "this kitty," or "this kitty-cat." (The form of the label that the experi-menter used to refer to the object was varied,as appropriate, to agree with any label thechild might have produced spontaneously.)The difference between the Familial- andFamiliar—Explicit conditions provided away to examine the role of explicitly men-tioning the familiar basic-level count nounin children's tendency to interpret the novelword as a proper noun.

The purpose of the testing portion of thesession was to determine children's inter-pretation of the novel proper noun. To beginthe testing portion, the experimenter toldthe child that he had some other things thathe wanted the child to see. In random order,the three remaining toys were removed fromthe box and placed in front of the child. Theexperimenter then asked the child to carryout a series of 10 simple actions. Six of theseactions were designated as test trials. Onthese trials, the experimenter asked thechild to choose an object in response to arequest that included the new name (e.g.,"Can you put Zav behind your back?"). Outhe four remaining filler trials, the experi-menter told the child which object to handle


(e.g., "Can you put this behind your back?").Each of the four toys was used once as afiller item. The use of filler trials decreasedthe likelihood that children would select atoy simply to handle it, and it removed thefocus of the task away from the named toy.After the final trial, the experimenterpointed to each toy in a random order andasked the child, "Is this Zav?" (using thename the child had learned).

The novel nouns used, the order of ac-tions, and the member of the pair of objectsthat was labeled were balanced across sub-jects. The four novel labels were: "zav,""wug," "glorp," and "dax." The 10 actionswere: put behind the back, put in the box,hold in hand, give to experimenter, put ontop of head, throw in the box, turn upside-down, drop, bug, and point to. All the chil-dren knew how to carry out these simple ac-tions. Trials were designated as "tests" or"fillers" with the constraint that no morethan two consecutive questions be of thesame type. Children could easily reach allfour toys at all times during the task. Chil-dren's choices on each trial were recorded.The sessions were audiotaped for later tran-scription of children's spontaneous com-ments.

Results and DiscussionEach child received three scores to re-

fiect the number of test trials (out of a totalof six) on which the child chose (a) thenamed toy, (b) the other toy of the samekind, and (c) one of the distractor items. Ta-ble 1 presents the mean results.

The first analyses focused on the num-bers of selections of the named toy. Therewere no sex difFerences in any group interms of these numbers. A one-way ANOVAwith condition (Unfamiliar, Familiar, andFamiliar—Explicit) as a between-subjects

factor was then conducted, using the num-bers of selections of the named toy as thedependent measure. The condition effectwas significant, F(2,27) = 4.76, p < .025.Two planned contrasts were computed. Thefirst showed that children were reliablymore likely to have selected the named toyin the Familiar (M — 5.20 times out of 6, or87%) than in the Unfamiliar (M - 3.90 timesout of 6, or 65%) condition: F(l,27) - 5.56,p < .05. The second contrast revealed thatchildren selected the named toy in the Fa-miliar condition as often as in theFamiliar—Explicit condition (M — 5.50times out of 6, or 92%); the difference be-tween these conditions was not reliable.

How do the results using this depen-dent measure compare to those from earlierstudies of the acquisition of proper nouns?The results from the Unfamiliar conditionwere similar to those obtained by Gelmanand Taylor (1984) in the analogous conditionof their study; for the Familiar condition, theresults showed even more frequent propernoun interpretations than reported by Katzet al. (1974) in the relevant condition of theirexperiment. The fact that children in Katz etal. (1974) were, on average, 8 monthsyounger (and presumably less sensitive tothe syntactic cues to proper nouns) than thechildren in this study offers a possible expla-nation of this difference.

A clearer difference among conditionsemerged from a more conservative coding ofthe data. Children were classified as fallinginto one of two groups according to whetherthey selected only the named toy across alltest trials. This measure of the tendency tomake a proper noun interpretation is quitestringent. This stringency is not unrealistic,however, given that in real life proper nounsare used categorically, not probabilistically.

TABLE 1

MEAN NUMBER OF SELECTIONS (and Standard Deviation) OF THE NAMED TOY, THE OTHER TOY OF THESAME KIND, AND THE TOYS OUTSIDE THE KIND

Named Toy

Experiment 1:Unfamiliar condition (n = 10) 3.90 (1.29)Familiar condition (n = 10) 5.20(1.32)Familiar—Explicit condition (H = 10) .5.50(1.08)

Experiment 2:Unfamiliar condition (n = 10) 3.70(1.57)Familiar condition (n = 10) 5.40(1.27)Familiar—Explicit condition (n = 10) 5.30(1.16)

Other KindMember

2.00(1.16).80 (1.32).50 (1.08)

2.00(1.49).60(1.27).50(1.08)

Outside Kind

.10 (.32)

.00 (.00)

.00 (.00)

.30 (.68)

.(X) (.00)

.20 (.63)


TABLE 2

NUMBER OF CHILDREN SELECTING I'HE NAMED TOY, IHE OFHER TOY OF IHE SAMEKIND, AND THE TOYS OUTSIDE I'HK KIND ON 0 THROUGH 6 TR[ALS IN EXPERIMEN r 1

NUMBER oi- TRIALS

6 5 4 3 2 1 {)

Unfamiliar condition (n = 10):Named toy 2 0 4 3 1 0 0Other kind memher 0 0 0 4 4 0 2Outside kind 0 0 0 0 0 1 9

Familiar condition (n = 10):Named toy 7 0 1 2 0 0 0Other kind member 0 0 0 2 1 0 7Outside kind 0 0 0 0 0 0 10

Familiar—Explicit condition (n - 10):Named toy 8 0 1 1 0 0 0Other kind member 0 0 0 1 1 0 8Outside kind 0 0 0 0 0 0 10

to pick out individuals. A person cannot besaid to be using "Garfield" as a proper nounif the person uses it some or most of the timeto refer uniquely to Garfield, but on otheroccasions uses it to refer to other cats (whoare not named "Garfield"). If these 2-year-olds had made a proper noun interpretationof the word in this study, then they shouldnever have selected anything other than thenamed toy, assuming that they could dis-criminate between toys of the same kind.The discriminability of the two objects wascontrolled to a considerable extent betweenthe familiar and unfamiliar conditionsthrough the use of similar dress.

When children were classified in thisway, two differences again were examined.First, whereas only two of the 10 childrenin the Unfamiliar condition chose only thenamed toy on al] trials, seven of the 10 chil-dren in the Familiar condition selected onlythe named toy. This difference is significantby Fisher's exact test, p — .035. Second, thenumbers of children selecting only thenamed toy across all test trials did not differreliably between the Familiar and theFamiliar—Explicit conditions; seven andeight children, respectively, selected thenamed object only. Table 2 presents thesedata."

Children's responses to the explicityes-no questions posed at the etid of the ses-sion did not reveal a clear distinction amongconditions, consistent with Gelman and Tay-lor (1984). Only two children in the Unfamil-iar condition answered these questions cor-rectly, saying "yes" only to the named toy;six children in the Familiar condition re-sponded correctly; and four did so in theFamiliar—Explicit condition. By a Fisher'sexact test, the difference between numbersin the Unfamiliar and Familiar conditionsfell short of being reliabie, p = .085. By thesame test, the difference between the Famil-iar and Familiar—Explicit conditions aisowas not reliable.

In summary, these results show thatchildren were more likely to make a propernoun interpretation of a novel proper nounapplied to an animate creature when thecreature was familiar than when it was unfa-miliar. The difference between Unfamiliarand Familiar conditions was reliable whenthe dependent measure was the number ofselections of ^ e named toy and also whenthe dependent measure was the number ofchildren who focused exclusively on thenamed toy. The results also show that explic-itly mentioning the basic-level count noundid not affect children's tendency to select

^ Additional analyses revealed that children who selected the named toy on four or moretrials out of six were exceeding chance (binomial p < .05). Application of this criterion to thesedata revealed that nine subjects in the Familiar condition selected the named toy more frequentlythan chance; six subjects did so in the Unfamiliar condition. The difference between thesenumbers is not reliable by a Fisher's exact test. Furthermore, there was no difference betweenthe Familiar and Familiar—Explicit conditions, using this same criterion. In both conditions,nine children exceeded chance.


the named familiar toy; performance in theFamiliar condition did not differ from thatin the Familiar—Explicit condition. This re-sult suggests that the familiarity of the refer-ent object per se, rather than the familiaritycoupled with explicit use of the basic-levelcount noun, accounted for the observedeffect.

While these results are clear, two con-siderations suggest the usefulness of afollow-up study. First, it would be valuableto check that the effect obtained in Experi-ment 1 held up with new stimuli. In Experi-ment 2, new stuffed creatures were used.Second, it would be informative to examinethe acquisition of proper nouns under differ-ent circumstances. It is true that the stimuliused in Experiment 1 were all animate sur-rogates and presumably good candidates forreceiving proper nouns. However, it couldbe that very young children rely on the se-mantic cue that only friends or family mem-bers (people or pets) can receive propernouns. It is thus possible that the childrenhad been reluctant to learn a proper nounfor one of the monster-like creatures becausethey failed to consider the toy as a potentialpet. To address this concern in Experiment2, a hand puppet introduced the named toyin both familiar and unfamiliar conditions as"a pet"

Experiment 2

Method

Subjects.—Thirty children from pre-schools serving middle- and upper-middle-class families participated, 15 boys and 15girls (2-4 to 3-0; mean age 2-8). Ten childrenwere assigned to each of three conditions,with approximately equal numbers of boysand girls. None had participated in the firstexperiment.

Materials.—In each condition, fourstuffed toys were used. In the Unfamiliarcondition, the toys again were selected sothat children would know no kind term foreither member of the pair from which theto-be-labeled toy was selected. There weretwo unfamiliar monster-like animals and twodistractor creatures, each approximately 4inches high. The monster-iike animais weretwo members of the kind of "Cuddlee Ug-lees" that had been used as a distractor inthe first study. They had pink bodies withblue hair, yellow horns, and bright greennoses. They looked identical except for theirdress, which consisted of hooded capes(with holes to allow the homs to stick

through) of distinct colors and materials: onewas dark brown, the other was white. Thedistractor items consisted of one familiar ani-mal (an orange cat, used in Experiment 1,but no longer dressed), and an unfamiliarmonster (one of the named monsters usedin Experiment 1, but no longer dressed).The unfamiliarity of the monsters had beenverified previously as described in Experi-ment 1.

In the Familiar and Familiar—Explicitconditions, toys again were chosen so thatchildren likely knew a kind term for bothmembers of the pair from which the to-be-labeled toy was selected. There were twodogs and two distractors. The dogs weregrey with black spots and long brown ears,and they looked identical except for theirdress, which again consisted of hoodedcapes of the same colors that were used todress the unfamiliar monsters. The dis-tractors were the same as in the Unfamiliarcondition.

A shoebox again was used as a place tostore the creatures during parts of the studyand as a prop during some of the specificactivities.

Procedure.—The procedure was thesame as in Experiment 1 in all three condi-tions, with one exception. In all conditions,a hand puppet took the to-be-named toyfrom the box and introduced the toy to thechild. The puppet said, "I have a pet. Wouldyou like to meet my pet?" He then namedthe toy for the child exactly as in Experiment1. In the Familiar—Explicit condition, heused the basic-level count noun "dog" (or"doggie" or "puppy-dog") to refer to the to-be-named dog. (Variations in the form of thelabel were determined as in Experiment 1.)

Results and DiscussionChildren received three scores to reflect

the number of test trials (out of a total ofsix) on which they chose (a) the named toy,(b) the other member of the same objectkind, and (c) one of the distractor items. Themean results are presented in Table 1.

The first analyses were based on thenumbers of selections of the named toy. Nosex differences were found in any group interms of these numbers. A one-way ANOVAwas performed on the numbers of selectionsmade of the named toy, with condition (Un-familiar, Familiar, and Familiar—Explicit)as a between-subjects factor. The conditioneffect was reliable, F(2,27) - 5.06, p < .025.Two planned contrasts tested specific


hypotheses about the effect. The firstshowed that children were more iikeiy tohave seiected the named toy in the Familiar(M = 5.40 times out of 6, or 90%) than in theUnfamiliar (M = 3.70 times out of 6, or 62%)condition, F(l,27) = 8.04, p < .01. The sec-ond demonstrated that children were aslikely to have selected the named toy in theFamiliar condition as in the Familiar—Explicit condition (M ^ 5.30 times out of6, or 88%); there was no reliable differencebetween these conditions.

As in Experiment 1, the percentage ofnamed toy selections in the Unfamiliar con-dition was similar to, although slightly lowerthan, the one reported by Gelman and Tay-lor (1984) in the analogous condition of theirstudy. The Familiar condition results againwere somewhat higher than those obtainedby Katz et al. (1974) in the relevant conditionof their experiment. As in Experiment 1, thefact that children in this study were about 8months older, on average, than those in Katzet al. (1974) presents a possible explanationof this difference.

A more conservative coding of the datawas performed. A criterion was adoptedwhereby children were classified as havingmade a proper noun interpretation if theyselected only the named toy across all testtrials. Again, two comparisons were of inter-est. First, only two of the 10 children in theUnfamiliar condition chose only the namedtoy across a!l trials, whereas seven of the 10children in the Familiar condition selectedonly the named toy. This difference is sig-nificant by Fisher's exact test, p = .035. Sec-ond, in both the Familiar and Familiar—Explicit conditions, seven of 10 childrenselected only the named toy on all test trials.Table 3 presents these data.^

As in Experiment 1, the numbers of chil-dren in each condition giving correct an-swers to the explicit yes-no questions (askedat the end of the session) were compared.Only one child answered the questions cor-rectly in the Unfamiliar condition, whereasfive children in the Familiar condition andfive children in the Familiar—Explicit con-dition answered correctly. The differencebetween the numbers in the Unfamiliar andFamiliar conditions fell short of significance

according to a Fisher's exact test (p = .07).The results of the same test also showed noreliable difference between the Familiarand Familiar—Explicit conditions in termsof the number of correct answers given.

In this replication designed to test thehypothesis with new stimuli and to examineacquisition under different teaching circum-stances, children were more likely to make aproper noun interpretation of a proper nounapplied to an animal when the animal wasfamiliar than when it was unfamiliar. A dif-ference between Unfamiliar and Familiarconditions obtained when the dependentmeasure was the number of selections of thenamed toy and also when the dependentmeasure was the number of children whofocused exclusively on the named toy.Knowing that the creature in the Unfamiliarcondition could he labeled with the familiarsuperordinate kind term "a pet" thus did notremove children's tendency to interpret thenovel proper noun as referring to a kind ofobject. This finding suggests that the object-kind bias is an expectation for a lower-levelkind label, presumably a basic-level kind la-bel (see Rosch, Mervis, Grey, Johnson, &Boyes-Braem, 1976), for an unfamiliar ob-ject In addition, explicit mention of thebasic-level count noun did not affect chil-dren's tendency to make a proper noun in-terpretation of the novel word applied to afamiliar object. Performance in the Familiarand Familiar—Explicit conditions did notdiffer reliably.

Childbpen''8 Spontaneous Connnents

Children's spontaneous comments dur-ing the tasks provided a rich source of data.Ghildren often made comments, especiallywhen the toys were first all presented orwhen a particular toy was used on a fillertrial. These comments were transcribedfrom tapes of the play sessions, and theywere then pooled over the two studies. Thepatterns of occurrence of two types of com-ment corroborate the claims made in this ar-ticle about children's Interpretations of thenovel words.

One type, to be called a count nouncomment, involved a child's spontaneouslyand explicitly extending the novel label to

** As in Experiment 1, ehildren who selected the named toy on four or more trials out of sixexceeded chance (binomial p < .05). Nine subjects in the Familiar condition selected the namedtoy more frequently than chance; five subjects did so in the Unfamiliar condition. The differencebetween these numbers fell short of being reliable by a Fisher's exact test: p = .07. Once again,there was no differenee between the Familiar and Familiar—Explicit eonditions, using this samecriterion. In both conditions, nine children exceeded chance.


TABLE 3

NUMBER OK CHILDREN SELECTING THE NAMED TOY, THE OTHER TOY OE THE SAMEKIND, AND THE TOYS OUTSIDE THE KIND ON 0 THROUGH 6 TRIALS IN EXPERIMENT 2

6

NUMBER OF TRIALS

5 4 3 2 1

Unfamiliar condition (n = 10);Named toy 2Other kind member 0Outside kind 0

Familiar condition (n = 10):Named toy 7Other kind member 0Outside kind 0

Familiar—Explicit condition (n - 10):Named toy 7Other kind member 0Outside kind 0 0 0 0

100

200

00

220

010

20

220

000

11

321

100

01

021

020

00

028

07

10

08

0 9

the other member of the named kind, treat-ing it syntactically as a count noun (e.g., bypluralizing it or by combining it with a de-terminer). These comments were particu-larly striking because all children had heardthe word modeled only as a proper noun(e.g., "This is Zav"). Consider the f'oilowingexamples:

Girl, 2-5; Pointing to the other memberof the same kind, then the named toy, shesaid, "Two zavs! Two zavs again! Two zavs!"Later asked to "put Zav in the box," she re-sponded, "Two zavs in the box. I want to puttwo zavs in the box!"

Boy, 2-4: Pointing to the other memberof the same kind, he said, "That a dax!"

Cirl, 2-9: Pointing to the other memberof the same kind, she said, "That's anotherzav! Two zavs!"

A second type of spontaneous comment,to be called a proper noun comment, re-vealed a proper noun interpretation of thelabel. It occurred whenever a child sponta-neously and explicitly restricted the label tothe named toy, for example, when the childpointed to the unnamed kind member andasked what it was called. Some examplesfollow:

Boy, 2-11: When the other member ofthe same kind was brought out, he touchedit and asked, "Who is this name?"

Boy, 2-10: He pointed to the unnamedmember of the same kind and asked, "Whois this? Who is this?"

Girl, 2-5: Afler the labeling of the first

toy, the second toy was brought out, and sheasked, "And who's that one?"

If children had treated the novel propernouns equivalently in the Unfamiliar, Fa-miliar, and Familiar—Explicit conditions,then these two types of comment shouldhave been distributed equally in the threeconditions. However, to the extent that thefamiliarity of the referent object affectedsensitivity to the syntactic cues to propernounhood, count noun comments shouldhave been more prevalent in the Unfamiliarcondition, while proper noun commentsshould have occurred more frequently in theFamiliar and Familiar—Explicit conditions.In the Unfamiliar condition, seven childrenmade a count noun comment, whereas onlyone made a proper noun comment; in theFamiliar condition, two children made acount noun comment and five made a propernoun comment; in the Familiar—Explicitcondition, no child made a count noun com-ment, but three made a proper noun com-ment. Two Fisher's exact tests were carriedout. The first showed a reliable differencebetween the distribution of comments in theUnfamiliar and Familiar conditions, p =.035. The second test revealed no reliabledifference between the distribution of com-ments in the Familiar and Familiar—Explicit conditions, a result that is consistentwith the earlier findings of no difference be-tween children's performance in these con-ditions.

General Discu8§ion

Two studies examined 2-year-olds' in-terpretations of novel proper nouns for sur-


roeates of animate objects differing in famil-iarity. The results showed that youngchildren's interpretations of the propernouns (see Katz etal., 1974; Gelman & Tay-lor 1984) was affected by the iamiiiarity oithe referent object (see Markman & Wachtel,1988). In the first study, children were morelikely to interpret a novel proper noun ap-propriately as referring to a unique individ-ual when the word was applied to a familiaranimate creature (i.e., one for which chil-dren knew a kind label) than to an unfamiliarcreature (i.e., one for which children knewno kind label). When the creatures were un-familiar, children frequently treated thewords as if they were count nouns, despitethe proper noun syntactic cues. When theanimals were fkmlUar, children were reli-ably more likely to interpret the words asproper nouns, making appropriate use of theproper noun syntax. Furthermore, children sperformance after hearing a proper noun fora familiar object was not influenced by theexplicit menUon of the objects basic-levelcount noun. The second study replicatedthese results when the stimuli were difter-ent and when children were introduced tothe named object by a puppet who used thefan^iliar superordinate kind term "a pet.

The finding from the first study, thatchildren frequently interpreted words in-tended to be proper nouns as count nounswhen the objects were unfamiliar, is consis-tent with the claim that the first wordslearned for such objects are taken to refer toa kind of object (see Soja, 1990; Soja et al.,in press). The result from the second studyrefines this claim by suggesting that theobject-kind bias is a bias to acquire a lower-level kind term (e.g., a basic-level kindterm), because children who already knew afamiliar supeiordinate kind term ( a pet )torthe otherwise unfamiliar objects sttll inter-preted the novel proper noun as a countnoun referring to a kind of object.

The reliable difference between chil-dren's tendency to make a proper noun m-terpretation in the familiar and unlamiHaieonditions also provides support for ttieclaim that chiidren expect word meaningsnot only to contrast (e.g., Clark 1987) butalso to be mutually exclusive (Markman &Wachtel, 1988). When children knew nokind term for the kind of object, they tendedto interpret the word as referring to a kindof object. When children already knew abasic-level count noun for the kind of object(e g "cat"), they sought a different interpre-

tation of a second word learned for such ob-jects, namely, a proper noun. These resultsare consistent with previous experimentalstudies that support contrast (e.g., Taylor &Gelman, 1988) and mutual exclusivity (e.g..Markman & Wachtel, 1988).

The results showing that children re-stricted their selections to the named objectin the familiar condition are indeed consis-tent with the claim that they were makinga proper noun interpretation. However, analternative interpretation, suggested by theresults of Taylor and Gelman (1989), is thatby limiting their selections to the named ani-mal only, children were demonstrating asubordinate kind interpretation (e.g., someparticular kind of cat in Experiment 1; someparticular kind of dog in Experiment 2). Thisinterpretation is unlikely. Recall the resultsfrom Taylor and Gelman (1989, Experiment1) When two similar toys (two dogs from thesame subordinate kind, differing on y inclothing) were used as stimuli, childrenreadily extended a novel label applied toone object to both objects. The authors con-cluded that this result suggested a subordi-nate level kind interpretation of the newlabel. When children restricted their se-lections to the named object, despite a highlevel of similarity between members of thepair Taylor and Gelman (Experiments 3 and4) interpreted the result as consistent with aproper noun interpretation. In the presentstudies, the StimuU were like Taylor andGelmans similar exemplars (they wereidentical, from within the same subordinatekind), and children nonetheless restrictedtheir selections to the named toy only. Thisfact suggests that children were making aproper noun, not a subordinate kind, inter-pretation of the label. Children s patterns ofobject selections, along with their spontane-ous comments, thus support the contentionthat they were more Hkely to make a propeinoun interpretation of the proper noun inthe familiar than in the unfamiliar condi-tions.

As noted earlier, mnch research on lexi-cal acquisition has examined young chil-dren's sensitivity to syntactic cues to themeaning of new words (e.g Brown, 1957;Tavlor & Gelman, 1988; Waxman, 1990;w i m a n & Kosowski, 199G). This topic ofresearch is important because syntactic m-formation provides important semantic in-formation for word learners and appears tobe exploited by word learners fiom a youngage However, these and other recent results

(e.g., Markman & Waehtel, 1988; Soja, 1990;Soja et al., in press) suggest that young chil-dren's sensitivity to syntactic cues may beaffected by the familiarity of the referent ob-ject. Up to what age the object-kind bias ex-erts such a strong infiuence on the interpre-tation of new words for unfiimiliar objects isstill unknown, as is how this bias affectsyoung children's interpretations of newwords under more naturalistic circum-stances.

These results suggest that young chil-dren may err by interpreting proper nounsas count nouns in those circumstances inwhich a proper noun is being taught for anotherwise unfamiliar object. One frequentlycited phenomenon that initially seems to of-fer support for this claim i.s young children'stendency to overextend the proper noun"Daddy" to include other men who are nottheir fathers (see Merriman, 1986, for discus-sion). However, a problem in interpretingthis observation is that young children likelyhear the word modeled both as a propernoun (e.g., "Say hi to Daddy") and also as acount noun (e.g., "He's a daddy") (Merri-man, 1986). Furthermore, other evidencesuggests that young children rarely err intheir interpretations of proper nouns (seeMaenamara, 1986). A challenging questionfbr future research is thus to specify furtherthe conditions under which young childrenwill rely on syntactic cues to make appro-priate interpretations of novel proper nounsand words from other syntactic categories.

Referenees

Au, T., & Glusman, M. (1990). The principle ofmtitnal exclusivity in word learning; To honoror not to honor? Child Development. 61,1474-1490.

Baldwin, D. (1989). Priorities in children's ex-pectations about object jabel reference:Form over color. Child Development, 60,1291-1306.

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

Clark, E. {1973). What's in a word? On the child'sacquisition of semantics in his first language.In T. E. Moore (Ed.), Cognitive developmentand the acquisition of language {pp. 65-110).New York; Academic Press.

Clark, E. (1983). Meanings and concepts. In J. Fla-vel] & E. Markman (Eds,), P. H. Mussen (Se-ries Ed.), Handbook of child psychology: Vol.3. Cognitive development (pp. 7S7-S40).New York: Wiley.

D, Geoffrey Hall 1153

Clark, E. (1987). The principle of contrast: A con-straint on language acquisition. In B. Mac-Whinney (Ed.), Mechanisms of language ac-quisition (pp. 1-33). Hillsdale, NJ: Erlbaum.

Clark, E. (1988). On the logic of contrast. Jouma/of Child Language, 15, 317-335.

Clark, E. (1990). On the pragmatics of contrast.Journal of Child Language, 17, 417-431.

Gathercole, V. (1987). The contrastive hypothesistor the acquisition of word meaning: A recon-sideration ot the theory. Journa/ of Child Lan-guage, 14,493-531.

Gathercole, V. (1989). Contrast: A semantic bias?Journal of Child Language, 16, 685-702.

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

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., & Jones, S. (1988). The im-portance of shape in early lexical leaming.Cognitive Development, 3, 299-321.

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

Macnamara, J. (1986). A border dispute. Cam-bridge, MA: MIT Press.

Markman, E,, & Hutcbinson, J. (1984). Children'ssensitivity to constraints on word meaning:Taxonomic vs. thematic relations. CognitivePsychology, 16, 1-27.

Markman, E., & Wachte], G. (1988). Children'suse of mutual exclusivity to constrain themeanings of words. Cognitive Psychology, 20,121-157.

Merriman, W. (1986). How children learn the ref-erence of concrete nouns: A critique of cur-rent hypotheses. In S. Kuczaj & M. Barrett(Eds.), The acquisition of word meaning (pp.1-38). New York; Springer.

Rescoria, L. (1980). Overextension in early lan-guage development. Journal of Child Lan-guage, 7, 321-335.

Rosch, E., Mervis, C , Cray, W., Johnson, D., &Boyes-Braem, P. (1976). Basic objects in natu-ral categories. Cognitive Psychology, 3,382-439.

Soja, N. (1990). Semantic inferences: The role ofcount/mass syntax. Papers and Reports onChild Language Development, 29, 104-111.

Soja, N., Carey, S., & Spelke, E. (in press). Onto-logical categories guide young children's in-ductions about word meaning: Object termsand substance terms. Cognition.

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


Taylor, M., & Celman, S. (1989). Incorporating Waxman, S., &c Gelman, R. (1986). Preschoolers'new words into the lexicon: Preliminary evi- use of superordinate relations in classificationdence for language hierarchies in two-year- and language. Cognitive Development, 1,old children. Child Development, 60, 625- 139-156.636- Waxman, S., & Kosowski, T. (1990). Nouns mark

Waxman, S. (1990). Linguistic biases and the es- category relations: Toddlers' and preschool-tablishment of conceptual hierarchies: Evi- ers' word learning biases. Child Develop-dence from preschool children. Cognitive De- ment, 61, 1461-1490.velopment, 5, 123-150.

new acquiring proper nouns for familiar and unfamiliar animate...

Documents