Assumptions about Word Meaning: Individuation and Basic-Level ...

Assumptions about Word Meaning:Individuation and Basic-Level Kinds

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

Sandra R. WaxmanNorthwestern University

HALL, D. GEOFFREY, and WAXMAN, SANDRA R. Assumptions about Word Meaning: Individuationand Basic-Level Kinds. GHILD DEVELOPMENT, 1993,64,1550-1570. In 2 experiments, 3V2-year-oldchildren interpreted a novel count noun (e.g., "This is a murvil") applied to an unfamiliar stuffedanimal as referring to a basic-level kind, rather than to a kind that individuates its members bytype of situation (context or life-phase). For example, children made interpretations akin toPERSON (a basic-level kind) rather than PASSENGER (a context-restricted kind), and DOG (abasic-level kind) rather than PUPPY (a life-phase-restricted kind). These experiments also docu-ment the role of object familiarity (previous knowledge of a basic-level count noun for the animal)and explicit information (about the relevance of the animal's situation) in the learning of countnouns for situation-restricted kinds. We note that children readily learn the meanings of basic-level count nouns through ostensive definitions (e.g., "This is an X"), although ostensive defini-tions do not distinguish basic-level kinds from situation-restricted kinds. Therefore, we suggestthat children make an implicit assumption that a count noun applied to an unfamiliar solid objectrefers to a basic-level kind of object, and not to a kind that individuates its members by type ofsituation. We illustrate the importance of this assumption hy showing how it bears directly onindividuation, and therefore, on quantification (e.g., counting).

Young children readily learn the mean- simple ostension, they must succeed in map-ings of words through ostensive definition, ping the word applied ostensively to a solidthat is, through hearing the words used object (e.g., a dog) onto an object kind (cate-when things are pointed out (e.g., "This is gory) (e.g., DOG). Recent evidence suggestsan X"). However, an ostensive definition is that children between 2 and 4 years of ageinsufficient in helping children to select a will map a count noun onto an object kind,single correct word meaning from a vast instead of onto a unique individual (e.g.,number of possibilities. Children's success Gelman & Taylor, 1984), a material kindat acquiring meaning thus may reflect, at (e.g., Soja, Carey, & Spelke, 1991), or a sa-least in part, implicit assumptions about the lient property (e.g., Baldwin, 1989; Hall,meanings of ostensively defined words Waxman, & Hurwitz, in press; Taylor & Gel-(Carey, 1982; Markman, 1989). man, 1988).

Count nouns, words such as "chair," Indeed, if the object is unfamiliar (i.e.,"banana," and "dog," tend to be among the if the learner knows no basic-level countearliest words that children learn (e.g., Gent- noun for the kind), preschool children mayner, 1982). In order for children to learn the take any word applied to it ostensively asmeaning of a count noun (e.g., "dog") from referring to a kind of object. For example.

We thank the directors, teachers, parents, and children at Harvard Yard Day Gare Genter,Kehillith Israel Nursery School, Peabody Terrace Day Gare Genter, Serendipity Ghild Develop-ment Genter, and Soldier's Field Park Day Gare Genter, all of the Greater Boston area (Experi-ment 1), and Hilltop Nursery School, Eton Nursery School, St. Ghristopher's Nursery School, St.Leonard's Nursery School, and U.L.U. Day Nursery, all of the Greater London area (Experiment2). Experiment 1 was part of the first audjor's doctoral thesis submitted to the Department ofPsychology at Harvard University, Gambridge, MA. That research was supported by a GentennialFellowship from the NSERG of Ganada. We are grateful to Joanne Hemmingfield and MargaretWard for invaluable assistance in the Babylab at the G.D.U. We thank Roger Brown, Susan Garey,Robert Rosenthal, Bill Meniman, and three anonymous reviewers for their helpful commentson earlier versions of the manuscript. Requests for reprints may be addressed to the first authorat the M.R.G. Gognitive Development Unit, 4 Taviton Street, London, WGIH OBT, England.

[Child Development, 1993,64,1550-1570. ® 1993 by the Society for Research in Child Development, Inc.All rights reserved, 0009-3920/93/6405-0007$01.00]

Hall and Waxman 1551

Hall (1991) showed that 2-year-olds con-strued a proper name (e.g., "This is Zav")applied to an unfamiliar, but not a familiar,object as referring to a kind of object. Mark-man and Wachtel (1988; see also Soja et al.,1991) provided evidence that 3-year-oldswill interpret a mass noun (e.g., "This ispewter") applied to an unfamiliar, but not afamiliar, object as referring to a kind of ob-ject. And Hall et al. (in press) have shownthat 4-year-olds are more likely to take anadjective (e.g., "This is very feppish") ap-plied to an object as referring to a kind ofobject if the object is unfamiliar than fa-miliar.

The ability to map a word applied to anobject onto a kind of object is crucial to theability to learn count nouns, but by itself,even this ability does not constrain chil-dren's interpretation sufficiently to accountfor their success. Any unfamiliar solid objectcan be construed as a member of many dif-ferent object kinds. For example, a word ap-plied to a dog could be interpreted as refer-ring to kinds such as DOG, ANIMAL,POODLE, PAW, TAIL, or PUPPY. How dochildren know which is the appropriatekind?

Children's success at learning from os-tension suggests that the interpretative as-sumption that guides them must be morespecific. Children must have some way ofconverging on a unique kind following anostensive definition. Researchers haveshown that children do appear to convergeon a specific kind when learning a new wordfor an unfamiliar object through ostension;moreover, this developmentally privilegedkind has properties associated with the ba-sic-level kind.

Basic-level kinds have a particular psy-chological salience (see Rosch, Mervis,Gray, Johnson, & Boyes-Braem, 1976). Inthe hierarchy POODLE, DOG, ANIMAL,the basic-level kind is DOG, referred to bythe basic-level count noun "dog." Rosch etal. (1976) have shown that the basic-levelkind is the most inclusive kind in a hierarchyin which individuals possess significantnumbers of attributes in common, elicit com-mon motor programs, have similar shapes,and can be identified from an average shapeof members of the kind. While it has provendifficult to provide an independent defini-tion of the degree of shape similarity sharedby the individuals in a given basic-levelkind (for discussion, see Fodor, 1983; Gleit-man, Gleitman, Landau, & Wanner, 1987;

Markman, 1989), the basic-level constructhas provided a useful summary descriptionof the psychologically salient kind in a vari-ety of psychological tests (Rosch et al., 1976).

Is the developmentally privileged kinda basic-level kind? In two ways, the basic-level kind does appear to be the preferredinterpretation of an ostensively definedword applied to an unfamiliar object. First,the individuals in the developmentally priv-ileged object kinds, like those in basic-levelkinds, appear to share an intermediate levelof shape similarity. For example, Horton andMarkman (1980) showed children between4 and 7 years of age drawings of several unfa-miliar objects and labeled each with thesame novel count noun (e.g., "This is adanker"). Some children saw objects from akind in which members shared a commongeneral shape (e.g., salamander-like ani-mals); others saw objects from a kind inwhich the members did not (e.g., ungulate-like animals, including pig-like, cow-like,and other perceptually distinct kinds of ani-mal). Children found it easier to identifynew individuals of the kind in which themembers shared a common shape (new sala-mander-like animals) than the kind in whichno common shape was shared (new ungu-late-like animals). However, the degree ofperceptual similarity among individuals inthe developmentally privileged kind is notthe highest degree imaginable. Taylor andGelman (1988, 1989) presented evidencethat suggests that children interpret a countnoun applied ostensively to an unfamiliarobject as referring to a kind that includesindividuals from distinct subordinate-levelkinds (with perceptually distinct appear-ances) within a basic-level kind. In otherwords, the level of perceptual similarity as-sociated with the developmentally privi-leged kind, like the level associated with thebasic-level kind, is (in some admittedly un-clear sense) intermediate. (For further dis-cussion of the "shape bias" in word learning,see Baldwin, 1989; Landau, Smith, & Jones,1988; Soja et al., 1991.)

Second, the individuals in the develop-mentally privileged object kinds, like thosein basic-level kinds, are discrete whole ob-jects rather than parts of objects. That is, theindividuals are members of a kind likeDOG, rather than a kind like PAW or TAIL.For example, Markman and Wachtel (1988,Experiment 2) showed that 3-year-old chil-dren assume that a count noun applied to anunfamiliar salient part of an unfamiliar ob-ject (the trachea attached to a lung) refers to

1552 Child Development

the kind of object as a whole (LUNG). Whenthe unfamiliar salient part belongs to an ob-ject for which the basic-level count noun isalready known (the claw on a hammer), chil-dren are more likely to assume the word re-fers to the kind for the salient part (CLAW)(see also Shipley & Spelke, 1988).

Yet to conclude from the preceding evi-dence that the developmentally privilegedobject kind is a basic-level kind would bepremature. The individuals in basic-levelkinds have another property that has notbeen noticed previously in the develop-mental literature (see Macnamara, 1986). Tounderstand this property, consider that oneof the semantic principles associated withall object kinds is a principle of identity.This principle specifies the range of situa-tions over which the identity of individualsin the kind is traced. The principle of iden-tity associated with individuals in basic-level kinds like PERSON or DOG specifiesthat their identity should be traced acrossa certain range of situations (i.e., across therange of personhood or doghood). However,many kinds individuate their members interms of types of situations that are more re-stricted than those associated with basic-level kinds. Ostension does not indicatewhether the individuals in a kind are thosethat belong to a basic-level kind such asPERSON, or to a context-restricted kind,such as PASSENGER; to a basic-level kindsuch as DOG, or to a life-phase-restrictedkind, such as PUPPY. Notice that the indi-viduals in these pairs of kinds (persons andpassengers, or dogs and puppies) coincidein certain situations (i.e., when persons areriding in vehicles, or when dogs are young).The problem for the learner is to selectwhich individuals should be taken as mem-bers of the kind, given only an ostensivedefinition.

The distinction between basic-levelkinds and situation-restricted kinds is impor-tant because it is directly related to individu-ation, and therefore to quantification. Gupta(1980) offered the following example, whichwe modify slightly. In certain situations,such as riding on an airplane, a man couldbe seen as a member either of the kind,PERSON, or of the kind, PASSENGER.After the plane lands, the man ceases to bea passenger, though he continues to be a per-son. However, if he makes a round trip onthe airplane, the airline will count him (intheir annual records) as two passengers,though he is only one person. Thus the deci-sion to treat the man as either a PERSON or

a PASSENGER directly affects quantifica-tion. We count differently under basic-levelkinds and situation-restricted kinds.

The preceding example underscores theneed to understand whether the individualsthat fall in the developmentally privilegedkind are individuals like persons (individu-als whose identity is traced across an exten-sive range of situations) or individuals likepassengers (individuals whose identity istied to more restricted situations). Althoughit has often been claimed that children makea basic-level kind assumption in word learn-ing, no work has noted that this claim alsoembodies a claim about the tracing of iden-tity of individuals across a certain set of situ-ations. As a result, no research has actuallytested whether the developmentally privi-leged kind is a basic-level kind, with respectto this property. If the kinds that have beenidentified in the psychological literature asbasic level (e.g., PERSON) are, in fact, de-velopmentally privileged, then childrenshould prefer them to situation-restrictedkinds (e.g., PASSENGER) in a word-learn-ing task involving unfamiliar objects. In thefollowing experiments, we test whether chil-dren make this assumption about wordmeaning.

In Experiment 1, we examined whetherchildren interpret count nouns as pickingout individuals in basic-level kinds (e.g.,PERSON or DOG) or situation-restrictedkinds (e.g., PASSENGER or PUPPY). Wetaught one group of children a novel wordmodeled as a count noun for an unfamiliarstuffed animal depicted in a specific context(riding in a car), and another for an unfamil-iar stuffed animal depicted in a specific life-phase (while young). Children then tookpart in tasks designed to reveal their inter-pretation of the words. The central goal ofthese tasks was to determine whether chil-dren extended the count noun to includeother members of the basic-level kind, orwhether they restricted the count noun toanimals in the specific context (e.g., ridingin a car) or life-phase (e.g., while young) ofthe target. We predicted that children's as-sumption would be to construe the individu-als as members of basic-level kinds, not situ-ation-restricted kinds.

A secondary issue concerned the hierar-chical level of children's interpretation. Inaddition to varying in their situation (contextor life-phase) with respect to the target, the


stimulus animals also varied in subtype,marked by color of garments. (See Taylor &Celman, 1989, Experiment 3, for evidencethat garment type may sometimes be con-strued by young preschoolers as a subtypedistinction.) This manipulation allowed usto determine whether children would ex-tend a novel word across subtypes withinthe basic level. We predicted that children'sassumption would be to construe the indi-viduals as being members of basic-levelkinds, not members of a particular subtype(see Taylor & Celman, 1988, 1989). We alsoinclude out-of-kind distractor animals in thetest sets to enable us to distinguish a basic-level kind interpretation from a more super-ordinate-level kind interpretation.

A further motivation for the experimentwas to examine the circumstances underwhich children would make somethingother than a basic-level kind interpretation.We explored the role of two factors, intro-ductory cue and target familiarity. To exam-ine the role of introductory cue, we providedsome children with simple ostensive defini-tions (e.g., "This is a murvil"); we providedothers with ostension plus information di-rectly implying a situation-restricted inter-pretation (e.g., "This is a murvil because itis riding in a car"; "This is a murvil becauseit is very young"). We predicted that the useof such information would decrease chil-dren's tendency to make basic-level kind in-terpretations compared to the use of simpleostension, and lead them instead to makesituation-restricted interpretations.

We also examined the role of the famil-iarity of the target object (i.e., whether ornot children previously knew the basic-levelcount noun for the object). Recent studieshave demonstrated that children tend toavoid assuming that two words will have thesame meaning (see Clark, 1987, for theoreti-cal discussion; for experimental evidencesee Hall, 1991; Markman & Wachtel, 1988;Merriman & Bowman, 1989; Taylor & Cel-man, 1988). In this experiment, some chil-dren learned the word for an unfamiliar tar-get (e.g., a creature); others learned it for afamiliar target (e.g., a bear). We predictedthat children would be more Hkely to adoptan interpretation other than a basic-levelkind (a situation-restricted kind or a sub-type) if the target was familiar than if it wasunfamiliar.

Experiment 1METHOD

SubjectsForty-three 3V2-year-olds took part.

Three children could not complete the taskand were dropped from the study. The re-maining 40 children ranged in age from 3-6to 4-0, with a mean age of 3-9. There were20 girls. Children were from predominantlywhite middle- and upper-middle-class back-grounds. They were tested in a quiet cornerof their preschool classroom or in an adja-cent room during normal school hours. Tenchildren were assigned randomly to each offour conditions. The mean age (and standarddeviation), in months, in each condition wasas follows: Unfamiliar-Ostension, 45.6 (2.0);Familiar-Ostension, 44.7 (2.4); Unfamiliar-Information, 45.3 (1.8); Familiar-Informa-tion, 45.5 (1.7).

DesignWe used a 2 x 2 x 2 mixed design, with

introductory cue (Ostension, Information)and target familiarity (Unfamiliar, Familiar)as between-subjects factors, and trial type(Context, Life-Phase) as a within-subjectsfactor.

StimuliContext portion.—We used two sets of

six toys. The unfamiliar set had four focalanimals from a single basic-level kind (crea-tures with blue hair, large teeth, redtongues, and one green eye), an unfamiliardistractor (a round creature with yellow andblue fur), and a familiar distractor (a yellowrabbit). The familiar set included four fo-cal animals from a single basic-level kind(bears), an unfamiliar distractor (a roundcreature with yellow and blue fur), and a fa-miliar distractor (a yellow rabbit). Note thatthe distractors were the same for the unfa-miliar and familiar sets.

For the four members of the focal kinds(the one-eyed creatures and the bears), wecrossed the color of their garments (whitehats and white bow ties vs. plaid hats andplaid bow ties) with their context (sitting inthe back of orange plastic cars vs. sitting onthe ground). Thus in both familiar and unfa-miliar sets, there were two animals of thefocal kind sitting in cars, and two animals ofthe focal kind sitting on the ground (plus twodistractors sitting on the ground).' All ani-

' The cars looked a bit like trucks, but we referred to them as "cars" in the task, and childrenaccepted this label without difficulty.


mals were approximately 2 inches tall. Fig-ure 1 shows the two sets of context stimuli.

Life-phase portion.—Again, the unfa-miliar set consisted of four focal animalsfrom a single basic-level kind (pink and bluefurry creatures, with yellow horns, fangs,and a large green nose), an unfamiliar dis-tractor (a blue creature with long antennae),and a familiar distractor (a dinosaur). The fa-miliar set included four focal animals froma single basic-level kind (monkeys), an unfa-miliar distractor (a blue creature with longantennae), and a familiar distractor (a di-nosaur).

For the four members of the focal kinds(the homed creatures and the monkeys), wecrossed color of garment (blue hats and bluebow ties vs. paisley hats and paisley bowties) with life-phase ("very young" or ap-proximately 2 inches tall vs. "grown up" orapproximately 5 inches tall). Thus for both

unfamiliar and familiar sets, there were two"very young" animals of the focal kind andtwo "grown up" animals of the focal kind(plus two distractors that were larger thanthe "young" focal animals). Figure 2 showsthe two sets of life-phase stimuli.

As further props, we used two wash-cloths, one white, one blue, and an addi-tional empty orange plastic car (for use inthe context portion).

Familiarity of the Object KindsEight preschool children (who did not

participate in the experiment proper butwho came from the same population as thosewho did) took part in a pretest to assess thefamiliarity of the object kinds. We were in-terested in whether children could offer asuitable (basic-level) count noun for eachkind of object we used. The experimentertold children they would see a set of toys.He said that if they knew what kind of thing

FIG. 1.—Unfamiliar (top) and familiar (bottom) stimuli used in context portion of Experiment 1.In the experiment itself, hats and ties were either white or plaid.


FIG. 2.—Unfamiliar (top) and familiar (bottom) stimuli used in life-phase portion of Experiment1. In the experiment itself, hats and ties were either blue or paisley.

a toy was, they should say what it was, butif they didn't know the name, they shouldsay they didn't know (and that it was accept-able not to know).

Each child saw one member of eachkind used in the study. For the familiar focalstimuli (the bear and the monkey) and thefarniliar distractors (the dinosaur and therabbit), all eight children provided appro-priate count nouns (i.e., "bear," "monkey,""dinosaur," and "rabbit" or "bunny").These findings support the claim that the in-tended familiar stimuli were familiar to thechildren.

For the two unfamiliar focal stimuli, oneof the eight children offered the count noun"monster"; no other child provided any la-bel (i.e., they all said they didn't know whatthey were). For the unfamiliar distractors,children offered a few count nouns, butthere was no consensus among them. For

one of the unfamiliar distractors, one childoffered "dog," one provided "clown," andone said "mouse." For the other, one childoffered "bee," and another suggested "cater-pillar." The pretest thus also provided assur-ance that the intended unfamiliar toys wereessentially unfamiliar, in the sense of unfa-miliarity used here.

ProcedureThe experiment had two parts, a context

portion and a life-phase portion.

Context portion.—Children first saw allfour members of the focal kind (one-eyedcreatures or bears). The experimenter asked,"Which of these are riding in a car?" and"Which of these are out of a car?" Childrenhad to answer correctly before the experi-menter continued; all children did so.

Three of the four focal kind memberswere then moved to the experimenter's side.


leaving the target animal (a one-eyed crea-ture or a bear riding in a car) in front of thecbild. Pointing to the target, the experi-menter said, for example, "This is a murvil"(Ostension conditions) or "This is a murvilbecause it is riding in a car" (Informationconditions). Tbe experimenter then askedthe child to repeat the word (Ostension con-ditions) or botb the word and information(Information conditions). All children did asrequested. The experimenter repeated theintroductory cue at least four more timeswhile cbildren played witb the animal,which remained in the car at all times. Theexperimenter then brought back the threeother focal kind members. These four ani-mals now were arranged so that the targetanimal sitting in a car and the animal of thesame garment color sitting outside a carwere closest to the child. The two animalsof different garment color (one in a car, oneoutside) were placed directly bebind theseanimals. In addition, a distractor animal wasplaced on each side of the focal kind set.The child participated in three tasks, in thefollowing order:

1. Toy selection task.—The experi-menter asked the child to perform a seriesof 12 actions, each involving one of the sixanimals. Six of these actions were desig-nated as test trials, and on these the experi-menter asked the child to perform an actionin response to a request that included thenew count noun (e.g., "Can you tickle a mur-vil?"). The remainder were filler trials, in-cluded to reduce the likelihood that childrenwould select an animal simply out of an in-terest in handling it. On filler trials, the ex-perimenter told the child which object tohandle (e.g., "Can you pat this on thehead?"). Each of the six objects served onone filler trial. Pairs of filler trials were inter-spersed semirandomly among the test trials,under the constraint that no more than twotest trials occur in succession. Children per-formed a total of nine different kinds of ac-tion—six for the test trials, plus three for thefiller trials (one per pair). All actions werefamiliar to children. They consisted of: paton head, tickle, touch nose, wash with bluecloth, wash with white cloth, cover withblue cloth, cover with white cloth, point to,and wave to.

To decrease children's tendency to fo-cus exclusively on any one animal, the ex-perimenter replaced die animal chosen onany test trial in the following manner. If theanimal had been in the row closest tothe child, then the experimenter moved it to

the second row. If it had been in the secondrow away from the child, then it remainedin the second row. If it had been one of theflanking distractors, it remained on the flank.The movement of any animal involved a dis-tance of only several inches, so all six ani-mals were within the child's easy grasp onall trials (see Taylor & Celman, 1988, for useof a similar technique). Following filler tri-als, toys were replaced where they had beenpreviously located, because on those trialsthe experimenter, not the child, had selectedthe toy.

2. Yes-no task.—The experimenterpointed to each of the six toys and asked, forexample, "Is this a murvil?" For all "yes"answers, the experimenter asked for a justi-fication; he said, for example, "Why is this amurvil?"

3. Posttest task.—There were two ver-sions of the posttest. Children received theversion that matched their pattern of an-swers on the yes-no task. If children gave asituation-restricted (or a conservative) pat-tern of answers on the yes-no task (i.e., said"yes" only to the two toys in the context ofthe target, or only to the target itself), thenwe asked them two questions in the posttest.First, the experimenter pointed to one of thefocal kind animals sitting in a car (always thetarget, if the cbild bad made a conservativeinterpretation) and asked, for example, "Canyou make this not be a murvil?" Second, bepointed to one of the focal kind animals sit-ting outside the cars and asked, for example,"Can you make this be a murvil?" (Anothercar, identical to the two used in the task, hadbeen brought out for this purpose.) If chil-dren gave a situation-independent pattern ofanswers on the yes-no task (i.e., said "yes"to at least one toy from within, and one toyfrom outside, the context of the target), thenwe asked only the first question describedabove; we did not ask the second questionbecause these children had already agreedthat one or both of the toys outside the carswere, for example, "murvils."

Life-phase portion.—Children first sawall four members of the focal kind (hornedcreatures or monkeys). As a preliminaryquestion, the experimenter asked, "Whenyou grow up, are you going to get bigger orsmaller?" All children answered correctly(i.e., they said "bigger"). The experimenterthen pointed to the four toys and asked,"Which of these are very young?" and"Which of these are grown up?" Childrenhad to answer correctly before the experi-


menter continued. All children pointed tothe small creatures in response to the firstquestion, and to the large creatures in re-sponse to the second. This performance isconsistent with recent findings that childrenof this age know that life-phase and size arecorrelated for a wide range of animals (Ro-sengren, Celman, Kalish, 6f McCormick,1991). The questions were formulated usingadjectives to describe the life-phases be-cause children were to be taught countnouns in the experiment. These questionswere meant to encourage children to treatthe size differences as differences in life-phase rather than, for example, as differ-ences between two subordinate-level kindsof animal (e.g., STANDARD POODLE andTOY POODLE).

Tbree of the four focal kind memberswere then moved to the experimenter's side,leaving the target animal (a small—"veryyoung"—horned creature or monkey) infront of the child. Pointing to the target, theexperimenter said, for example, "This is akepid" (Ostension conditions) or "This is akepid because it is very young" (Informationconditions). Tbe procedure then continuedas in the context portion. Children took partin a toy selection task and then a yes-no task,exactly as in the context portion. There was,however, no posttest task for the life-phaseportion, because it was not possible to havechijldren change these animals by makingthem grow larger or smaller.

In all conditions, exactly half the chil-dren received the context portion before thelife-phase portion. For each subject in onecondition, the order of the trials in the toyselection task was determined randomly;tbese random orders then were copied in allconditions. The same order was used for anychild's context and life-phase portions. Theorder in which the yes-no questions wereasked was determined randomly for eachchild. For each child, two nonsense wordswere selected randomly from "murvil,""shennet," "dilkin," and "kepid." We alsocounterbalanced the choice of toy thatserved as a target on botb the context and thelife-phase portions; the toy of each garmentcolor was used half the time.

RESULTS AND DISCUSSION

Our interest first lay in examining therole of target familiarity and introductorycue in children's tendency to make a basic-level kind interpretation, that is, in their ten-dency to extend the novel count nouns to

objects of tbe focal kind (1) across situations(context or life-phases) and (2) across sub-type distinctions. We then examined chil-dren's justifications, and, finally, their post-test task answers.

Extending the Count Nounacross Situations

Toy selection task.—We examined theproportion of test trials on which childrenselected either of the two focal toys outsidethe situation (context or life-phase) of the tar-get. (No child selected either of the dis-tractor toys on any test trial.) We obtainedthis proportion by adding the proportions ofselections of the toys marked "DifferentContext (Life-Phase)" and "Both Different"in Table 1. The combined proportion re-flects the extent to which children general-ized beyond the situation of the target in ex-tending the novel word.

We entered these scores into an AN-OVA, with introductory cue (Ostension, In-formation) and target familiarity (Unfamiliar,Familiar) as between-subjects factors, andtrial type (Context, Life-Phase) as a within-subjects factor. There was a significant effectof introductory cue, F(l, 36) = 13.28, p <.001, indicating a higher proportion of differ-ent-situation selections in the Ostensionconditions (M = .43, SD = .17) than in theInformation conditions (M = .26, SD = .21).The significant effect of target familiarity,F(l, 36) = 22.20, p < .001, revealed a higherproportion of different-situation selectionsin the Unfamiliar (M = .45, SD = .10) thanin the Familiar (M = .23, SD = .23) condi-tions. The introductory cue x target famil-iarity interaction was also significant, F(l,36) = 5.76, p < .05. Tests of simple effectsestablished that the effect of target familiar-ity was significant in the Information ip <.001) but not in the Ostension (p > .10) con-ditions.

We next classified children according tothe pattern of their selections across all testtrials in the toy selection task. We classifiedchildren as having made a situation-independent interpretation if their six toy se-lections included at least one from outsidethe situation of the target. They were cred-ited with a situation-restricted interpreta-tion if their six toy selections all fell withinthe situation of the target, including at leastone that was not of the target itself. If chil-dren selected only the target on all six testtrials, we attributed to tbem a conservativeinterpretation.

The numbers resulting from this classi-


TABLE 1

PROPORTION OF TOY SELECTIONS (and Standard Deviation) IN TOY SELECTION TASK IN EXPERIMENT 1

Ostension:UnfamiliarFamiliar

Information:UnfamiliarFamiliar

Ostension:UnfamiliarFamiliar

Information:UnfamiliarFamiliar

TargetToy

.27 (.09)

.52 (.33)

.28 (.18)62 (.27)

TargetToy

.28 (.16)

.60 (.32)

.42 (.21)

.67 (.26)

DifferentCarment

Color

.23 (.16)

.05 (.11)

.20 (.19)

.30 (.30)

DifferentCarment

Color

.25 (.26)

.08 (.21)

.25 (.12)

.23 (.24)

CONTEXT PORTION

DifferentContext

.27 (.09)

.40 (.29)

.37 (.26)

.07 (.12)

LIFE-PHASE PORTION

DifferentLife-Phase

.33 (.16)

.28 (.27)

.18 (.18)

.08 (.21)

BothDifferent

.23 (.18)

.03 (.11)

.15 (.17)

.02 (.05)

BothDifferent

.13 (.11)

.03 (.07)

.15 (.12)

.02 (.05)

Distractor

.00 (.00)

.00 (.00)

.00 (.00)

.00 (.00)

Distractor

.00 (.00)

.00 (.00)

.00 (.00)

.00 (.00)

fication appear in the first, second, and fifthcolumns of̂ Table 2. We used Fisher's exacttests to examine the relation between mak-ing or not making a situation-independentinterpretation and seeing an unfamiliar or afamiliar target. The results mirrored the sig-nificant interaction we obtained in the AN-OVA. The relation was not significant in theOstension conditions, in either the contextor the life-phase portion. In contrast, in theInformation conditions, significantly morechildren made situation-independent inter-pretations if the target was unfamiliar thanif it was familiar, p's < .05.

Yes-no task.—We classified children ashaving made a situation-independent inter-pretation if they said "yes" to at least onetoy from within, and one toy from outside,the situation of the target. We classifiedthem as having made a situation-restrictedinterpretation if they said "yes" only to thetwo toys within the situation of the target.We assigned children a conservative inter-pretation if they said "yes" only to the targettoy. The numbers that we obtained from thiscoding appear in parentheses in columns 1,2, and 5 of Table 2. We again conductedFisher's exact tests to examine the relationbetween making or not making a situation-independent interpretation and seeing ei-ther an unfamiliar or a familiar target. The

results replicated the significant findingsfrom the previous classification in all in-stances.

In sum, the results showed that childrenwere more likely to interpret a novel countnoun as applying across situations (1) if thetarget object was unfamiliar than if it wasfamiliar, and (2) if the word was introducedwith ostension than if it was introduced withinformation implying a situation-restrictedinterpretation. Strikingly, the effect of targetfamiliarity was greater in the Informationconditions than in the Ostension conditions.Despite the fact that the information calledfor a situation-restricted interpretation, chil-dren were more likely to rely on it if theobject was familiar than if it was unfamiliar.Children thus appeared to assume that aword applied to an unfamiliar object appliesto individuals across situations; childrenwere to some extent oblivious to informationthat was in conflict with this assumption.

Extending the Count Nounacross Subtypes

Toy selection task.—We examined theproportion of test trials on which childrenselected either toy of the focal kind outsidethe subtype of the target. (Recall that nochild selected either of the distractor toys onany test trial.) We obtained this proportion


TABLE 2

CLASSIFICATION OF CHILDREN BY PATTERN OF TOY SELECTIONS IN TOY SELECTION TASK(and Pattern of Answers on Yes-No Task) IN EXPERIMENT 1

SITUATION SUBTYPE

Independent Restricted Independent

Context portion:Ostension:

Unfamiliar 10 (10) 0 (0) 10 (10)Familiar 9 (9) 1 (0) 3 (2)

*(*)Information:


*(*)Life-Phase portion:

Ostension:Unfamiliar 10 (10) 0 (0) 8 (9)Familiar 7 (9) 3 (0) 3 (4)

*(*)Information:


* (*) * (*)

NOTE.—N = 10 per condition. Cons. = conservative pattern.* Indicated relation is significant, p < .05 by Fisher's exact test.

Restricted

0(0)7(7)

1(1)4(3)

2(1)7(5)

0(0)4(2)

CONS

0(0)0(1)

0(0)0(1)

0(0)0(1)

0(0)0(3)

by adding the proportions of selections ofthe toys marked "Different Carment Color"and "Both Different" in Table 1. This pro-portion reflects the extent to which childrengeneralized across subtypes in extendingthe novel word.

We entered these scores into an AN-OVA, with introductory cue (Ostension, In-formation) and target familiarity (Unfamiliar,Familiar) as between-subjects factors, andtrial type (Context, Life-Phase) as a within-subjects factor. There was a signiflcant effectof target familiarity, F(l, 36) = 13.28, p <.001, indicating a higher proportion of differ-ent-subtype selections in the Unfamiliar (M= .40, SD = .13) than in the Familiar (M =.19, SD = .23) conditions. The introductorycue X target familiarity interaction was alsosignificant, F(l, 36) = 4.17, p = .05. Testsof simple effects established that the effectof target familiarity was significant in the Os-tension ip < .001) but not in the Informationip > .25) conditions.

We next classified children according tothe pattern of their selections across all testtrials in the toy selection task. We classifiedchildren as having made a subtype-independent interpretation if their six toy se-lections included one or more from outsidethe subtype of the target. They were cred-ited with a subtype-restricted interpretation

if their six toy selections all fell within thesubtype of the target, including at least onethat was not of the target itself. If children.selected only the target on all six test trials,we classified them as having made a conser-vative interpretation.

The results of this coding appear in thethird, fourth, and fifth columns of Table 2.We examined the relation between makingor not making a subtype-independent inter-pretation and seeing either an unfamiliar ora familiar target, using Fisher's exact tests.The results for the context portion were con-sistent with our finding of a significant inter-action in the ANOVA; that is, more childrenmade a subtype-independent interpretationin the Unfamiliar- than in the Familiar-Ostension condition, p < .05; the relationwas not significant in the Information condi-tions. However, for the life-phase portion,more children made subtype-independentinterpretations in the Unfamiliar than in theFamiliar condition in both Ostension andInformation conditions, p's < .05.

Yes-no task.—We classified children ashaving made a subtype-independent inter-pretation if they said "yes" to at least onetoy from within, and one toy from outside,the subtype of the target. We classified themas having made a subtype-restricted inter-pretation if they said "yes" only to both toys


within the subtype of the target. We creditedthem with a conservative interpretation ifthey said "yes" only to the target animal.The numbers that emerged from this codingappear in parentheses in columns 3, 4, and5 of Table 2. Using Fisher's exact test, weagain explored the relation between makingor not making a subtype-independent inter-pretation and being in an Unfamiliar or aFamiliar condition; we obtained the samesignificant effects as we did from the previ-ous classification.

In sum, children were more likely to ex-tend the count noun across subtypes if thetarget was unfamiliar than if it was familiar.Moreover, this effect of target familiarity wasgreater in the Ostension conditions than inthe Information conditions. The findingfrom the Ostension conditions extends to3V2-year-olds a finding that Taylor and Cel-man (1988, 1989) previously obtained with2-year-olds.

Justifications from Yes-No TaskWe asked each of the 40 children for two

sets of justifications (one set for the contextportion, one for the life-phase portion),yielding 80 sets. We used these justificationsets to gain further insight into children'sinterpretations of the new words. The justi-fication sets were coded as basic-level-kind-relevant, subtype-relevant, situation-rele-vant, or other. Basic-level-kind-relevantjustification sets alluded to the objects' ap-pearance or kind, irrespective of situation orsubtype (e.g., "because it's brown andfurry"). Subtype-relevant justification setsmade direct reference to the animals' gar-ments (e.g., "because it's got a white hat").Situation-relevant justification sets alludedto the context or life-phase of the objects(e.g., "because it's riding in a car"). Theother category included ambiguous answers(e.g., "because it's a murvil"; "because itis") and failures to respond. The experi-menter coded all justification sets; a secondcoder, blind to the hypotheses, coded one-half of the sets; agreement was 90%.

Seventy-six (95%) of the sets fell intoone of the first three categories. We com-pared (1) whether children gave a basic-level-kind-relevant, a subtype-relevant, or asituation-relevant justification set, with (2)whether children gave a basic-level-kindpattern of answers (i.e., a pattern that wasboth situation-independent and subtype-independent), a subtype-restricted pattern,or a situation-restricted pattern on the yes-notask.

The correspondence between justifica-tions and yes-no answers was considerable.Of the 76 justification sets, 53 (70%) wereconsistent with the pattern of selections onthe yes-no task. However, there was a differ-ence in the correspondence in the Ostension(89%) and the Information (65%) conditions.The lower rate of correspondence in theInformation conditions reflects the fact thatin these conditions (particularly in theUnfamiliar—Information condition), chil-dren tended to give basic-level-kind pat-terns of answers on the yes-no task, but gavesituation-relevant justifications. For exam-ple, one child said "yes" to all four membersof the basic-level kind on the yes-no task.However, when asked to justify his selec-tions, he said that the two animals in the carwere murvils "because they were riding incars," and that the two animals outside thecar were murvils "because they were not ri-ding in cars." We think that these findingsreflect the fact that, in the Information condi-tions, children often interpreted the word asreferring to a basic-level kind; however,when they were asked for justifications afterthe task, they recalled the information theexperimenter had originally provided andfelt compelled to refer to it.

Posttest TaskThe posttest task provided us with fur-

ther evidence about children's interpreta-tion of the new word in the context portion.Children's performance on the posttest wasscored as either consistent or not consistentwith a situation-restricted interpretation. Tobe consistent with a situation-restricted in-terpretation, children had to take the animalout of the car in response to the first ques-tion; and if they were asked the secondquestion, they had to put the animal into the(newly introduced) car in response to it. Tobe not consistent with a situation-restrictedinterpretation, children had to answer theposttest question(s) without making any ref-erence to the car.

We were interested in the correspon-dence between (1) whether children re-sponded to the posttest question(s) in amanner consistent or not consistent witha situation-restricted interpretation, and(2) whether they had given a situation-restricted (or conservative) or a situation-independent pattern of answers on the yes-no task.

First consider the children who made asituation-restricted (or conservative) inter-pretation on the yes-no task, that is, said


"yes" only to the two animals riding in a car(or only to the target animal riding in a car).One child did so in the Familiar—Ostensioncondition, one did so in the Unfamiliar-Information condition, and seven did so inthe Familiar-Information condition. Thechild in the Familiar-Ostension conditionanswered the posttest question in a mannerthat was not consistent with a situation-restricted interpretation, but the child in theUnfamiliar—Information condition and allseven children in the Familiar—Informationcondition responded to the posttest ques-tions in a manner consistent with a situation-restricted interpretation. These findings of-fer further evidence that the interpretationsthese children made were indeed situation-restricted kinds.

Now consider children who made a situ-ation-independent interpretation on the yes-no task, that is, said "yes" to animals of thefocal kind both in and out of cars. Amongthe 31 children who made a situation-independent interpretation, most (27) gaveanswers that were not consistent with a situ-ation-restricted interpretation. Some ofthese simply said "I don't know." Other an-swers were more explicit, but still ignoredthe car; these answers ranged from the cute(e.g., "I'd use my magic wand," "I'd get awand and wave it") to the gruesome ("I'dchop him up," "I'd shoot him"). Only fourchildren answered the question in a mannerconsistent with a situation-restricted inter-pretation; all were in the Information condi-tions. We suggest that these children hadmade a basic-level kind interpretation of theword, but when asked for explicit answersto the posttest question, recalled the infor-mation and decided to refer to it.

In summary, the results of Experiment1 provided evidence that 3V2-year-olds inter-pret a count noun applied ostensively to anunfamiliar object as referring to a basic-levelkind, and not to a situation-restricted kindor to a subtype. The experiment alsodocumented the role of target familiarity(knowledge of a basic-level count noun) andinformation in the tendency to make a situa-tion-restricted kind interpretation. But al-though the results of Experiment 1 wereclear, several concerns led us to conduct afollow-up experiment.

The flrst was the issue of children's fa-miliarity with the restricted meanings thatwe intended to convey. In the context por-tion of Experiment 1, the intended restrictedkinds were BEAR-PASSENGER or ONE-

EYED-CREATURE-PASSENGER. Otherevidence suggests that 3-year-olds do notknow the count noun "passenger" (Hall, inpress-a), making it unlikely that they knewwords for either of the more speciflc mean-ings. For the life-phase portion, the in-tended restricted meanings were BABY-MONKEY or BABY-HORNED-CREATURE.Three-year-olds likely do know the word"baby," but they also know that speciflckinds of animal have specific names (e.g.,baby dogs get called "puppies"; baby catsget called "kittens"), and it is unlikely thatthey could have known words for either ofthe specific life-phase meanings used in theexperiment. The horned creature was unfa-miliar, and it is not clear what the Englishcount noun would be for the kind BABY-MONKEY; perhaps "juvenile." Moreover,even in the event that children were affectedby knowledge of the words "passenger" or"baby," this knowledge should have af-fected them to the same degree in the Unfa-miliar and Familiar conditions, leaving ourobserved familiarity effects unexplained.However, to rule out the possibility thatchildren's interpretations were affected byfamiliarity with the intended restrictedmeanings, in Experiment 2 we used a situa-tion-restricted meaning that is not encodedby a single English count noun. We taughtchildren a word for an animal sitting in ashallow box. Unlike the context "riding in acar," which may call to mind the count noun"passenger," the context "sitting in a box"does not suggest any single English countnoun.

A second question concerned children'sinterpretation of the relation between gar-ment color and meaning in Experiment 1. Inthe flrst experiment, we used garment color(i.e., the color of hats and bow ties) as a sub-type distinction within the basic-level kinds.It is possible that children construed this asa subordinate-level kind distinction (espe-cially given that the hats and ties were .seti>nonto the toys, making them appear to be aninherent rather than a changeable feature).Taylor and Gelman (1989, Experiment 3)have made such an observation. However, itis also possible that garment color was con-strued as a situational property, not as a sub-type marker. If this was the case, then chil-dren who selected animals wearing only onecolor of garment may have been making an-other type of situation-restricted interpreta-tion (e.g., PLAID-GARMENT-WEARER).Of course, this would not run counter to ourhypothesis, given that children tended not

1562 Child Developmentto make subtype interpretations in the Unfa-miliar—Ostension condition. (They did somore often in the Familiar—Ostension condi-tions.) Nonetheless, in Experiment 2, we ex-amined children's performance when thedistinctions within the focal kind moreclosely resembled subordinate-level kinddistinctions found within natural hierar-chies. To do this, we used toys that differedin fur color within a basic-level kind. Fur orskin color often is used to mark subordinate-level kinds in natural hierarchies (e.g., redvs. silver foxes, brown vs. black bears).

Finally, because we were concernedabout details involving the introduction andmanipulation of the toys, we changed theprocedure in four ways. First, we introducedall six toys at once, to decrease children'stendency to focus only on the four focal kindmembers from within the basic-level kind.Second, we interspersed the distractorsamong the focal toys in the array, again todraw attention away from the four focal kindmembers as a group. Third, movement oftoys after a test selection did not always in-volve moving the toy to the row farther awayfrom the child; sometimes, the movementwas to a different position in the first row.We hoped that this change would block thepossibility that the child viewed the toymovement as an indication that the chosentoy should not be selected a second time.Fourth, we altered the posttest procedureslightly. The fact that we brought out a newcar exclusively for the second question ofthe posttest in Experiment 1 may have ledchildren to think that it should be usedsomehow in responding to the questions. InExperiment 2, we did not introduce any newprop in the posttest.

Experiment 2

METHOD

SubjectsForty-five children participated, none of

whom had taken part in Experiment 1. Fivechildren were dropped from the study be-cause they failed to complete the task. Theremaining 40 children ranged in age from3-6 to 4-2, with a mean age of 3-10. Therewere 26 girls. Children were from a predom-inantly white population, ranging in socio-economic status backgrounds. Exactly halfthe children were tested in nursery schools,in a quiet corner of their classroom or in anadjacent room, during normal school hours.The remaining children were tested in atesting room in a developmental laboratory.

To recruit the latter children, we first ob-tained parents' names, primarily throughhealth-care workers, hospital prenatal clin-ics, and newspaper advertisements. Parentsreceived no remuneration for bringing chil-dren to the laboratory but were paid for theirjourney to and from it. The number of chil-dren tested in schools and in the laboratorywas approximately the same in all condi-tions. Ten children were assigned randomlyto each of four conditions. The mean age(and standard deviation), in months, in eachcondition was as follows: Unfamiliar-Ostension, 45.2 (2.4); Familiar-Ostension,45.8 (1.9); Unfamiliar-Information, 46.3(2.5); Familiar-Information, 46.1 (2.4).

DesignWe used a 2 x 2 between-subjects de-

sign, with introductory cue (Ostension, In-formation) and target familiarity (Unfamiliar,Familiar) as factors.

StimuliWe used two sets of six toys. The unfa-

miliar set included four focal animals froma single basic-level kind (round furry crea-tures with large yellow noses), an unfamiliardistractor (a blue creature with long anten-nae), and a familiar distractor (a grey cat).The familiar set consisted of four focal ani-mals from a single basic-level kind (rabbits),an unfamiliar distractor (a blue creature witblong antennae), and a familiar distractor (agrey cat).

For the four members of the focal kinds(the yellow-nosed creatures and the rabbits),we crossed their fur color (blue vs. pink)with their context (sitting in a shallow card-board box vs. sitting on the ground). Thus forboth unfamiliar and familiar stimulus sets,there were two animals of the focal kind sit-ting in boxes, and two animals of the focalkind sitting on the ground (plus two dis-tractors sitting on the ground). All animalswere approximately 3 inches tall. Figure 3shows the two sets of stimuli.

As in Experiment 1, we also used awhite washcloth and a blue washcloth.

Familiarity of the Object KindsEight preschool children (who did not

participate in the experiment proper butwho came from the same population as thosewho did) took part in a pretest to assess ob-ject kind familiarity. The procedure was thesame as in Experiment 1. For the two famil-iar stimuli—the rabbit and the cat—all chil-dren provided appropriate count nouns (i.e.,"rabbit" or "bunny" and "cat"; one child la-


FIG. 3.—Unfamiliar (top) and familiar (bottom) stimuli used in Experiment 2. In the experimentitself, fur color was either pink or green.

beled the cat with the basic-level countnoun "squirrel"). These findings support theidea tbat the intended familiar stimuli werefamiliar to the children.

For the two unfamiliar stimuli—thefocal kind member and tbe distractor—children offered few count nouns. For thefocal creature, one child offered the countnoun "monster"; another offered "bat." Noother child provided a label (i.e., they allsaid they didn't know what it was). For thedistractor, one child offered "snail," and oneprovided "bug." The pretest thus also pro-vided assurance that the intended unfamiliartoys were essentially unfamiliar, in thesense of unfamiliarity used here.

ProcedureThis was exactly like the context portion

of Experiment 1, with the following

changes: (1) all six animals were presentedtogether initially, rather than the focal kindmembers first; (2) the array of animals hadthe distractors mixed in with the focal kindmembers, rather than on the flank; (3) ani-mals were moved following all test selec-tions, but sometimes to another position inthe front row and sometimes to a position inthe rear row (determined randomly); and (4)during the posttest, no new props were in-troduced; if children wished to make use ofa box in answering the second question, theywere left to use one of the two boxes alreadyon the table.

In addition, in tbe Information condi-tions, instead of saying, for example, "Thisis a murvil because it is riding in a car," theexperimenter now said, "This is a murvil be-cause it is sitting in a box."


RESULTS AND DISCUSSION

Again, we first examined the role of in-troductory cue and familiarity in children'stendency to make a basic-level kind inter-pretation, that is, to extend the novel countnoun to members of the focal kind (1) acrosssituations (i.e., contexts), and (2) across sub-types. Then we turned to an examination ofchildren's justifications, and, finally, we con-sidered the results of the posttest.

Extending the Count Noun across ContextsToy selection task.—We examined the

proportion of test trials on which childrenselected either animal of the focal kind out-side the context of the target, as in Experi-ment 1, by adding the proportions of selec-tions of the animals marked "DifferentContext" and "Both Different" in Table 3.(In this experiment, one child selected theunfamiliar distractor toy on one test trial; ailother children made all test selections fromamong the focal kind members.)

We entered these scores into an AN-OVA, with introductory cue (Ostension, In-formation) and target familiarity (Unfamiliar,Familiar) as between-subjects factors. Theresults were consistent with those of Experi-ment 1. There was a significant effect of in-troductory cue, F(l, 36) = 14.57, p < .001,indicating a higher proportion of different-situation selections in the Ostension condi-tions (M = .52, SD = .21) than in the Infor-mation conditions (M = .27, SD = .25). Thesignificant effect of target familiarity, F(l,36) = 7.84, p < .01, revealed a higher pro-portion of different-situation selections inthe Unfamiliar (M = .48, SD = .22) than inthe Familiar (M = .30, SD = .27) conditions.The introductory cue x target familiarity in-teraction approached, but failed to reach,significance, F(l, 36) = 3.17, p = .08. Moti-vated by our findings in Experiment 1, wefollowed it up with tests of simple effects.These tests established that the effect of tar-

get familiarity was significant in the Infor-mation ip < .01) but not in the Ostension ip> .25) conditions.

We next classified children according tothe pattern of their selections across all testtrials in the toy selection task. We used thesame criteria as in Experiment 1 for attrib-uting situation-independent, situation-restricted, or conservative interpretations tochildren. The numbers appear in columns 1,2, and 5 of Table 4. As in Experiment 1, weexamined the relation between making ornot making a situation-independent inter-pretation and seeing an unfamiliar or a famil-iar target, using Fisher's exact tests. The re-sults supported the findings from the simpleeffects tests described above. The relationwas not significant in the Ostension condi-tions. In contrast, in the Information condi-tions, significantly more children made a sit-uation-independent interpretation if thetarget was unfamiliar than if it was familiar,p < .05.

Yes-no task.—Again, we relied on thesame criteria as in Experiment 1 for as-signing a child to a situation-independent,situation-restricted, or conservative inter-pretation. The numbers that emerged fromthis coding appear in parentheses in col-umns 1, 2, and 5 of Table 4. We again ex-plored the relation between making or notmaking a situation-independent interpreta-tion and being in an Unfamiliar or a Familiarcondition, using Fisher's exact tests. The re-sults replicated exactly the significant find-ings from the previous classification.

In sum, the results replicated Experi-ment 1 very closely. We found that targetfamiliarity and introductory cue exerted thesame effects on interpretation of the novelcount noun as they did in Experiment 1. Wealso found evidence that the effect of targetfamiliarity was greater in the Informationconditions than in the Ostension conditions.

TABLE 3

PROPORTION OF TOY SELECTIONS (and Standard Deviation)IN TOY SELECTION TASK IN EXPERIMENT 2

Target Different Different BothToy Fur Color Context Different Distractor

Ostension;Unfamiliar 27 (.14) .17 (.21) .30 (.17) .25 (.18) .02 (.05)Familiar 45 (.18) .07 (.09) .43 (.21) .05 (.11) .00 (.00)

Information:Unfamiliar 35 (.17) .23 (.18) .12 (.16) .30 (.15) .00 (.00)Familiar 52 (.21) .37 (.11) .07 (.12) .05 (.11) .00 (.00)


TABLE 4

CLASSIFICATION OF CHILDREN BY PATTERN OF TOY SELECTIONS IN TOY SELECTION TASK(and Pattern of Answers in Yes-No Task) IN EXPERIMENT 2

SITUATION SUBTYPE

Independent Restricted Independent Restricted CONS.

Ostension:Unfamiliar 10 (10)Familiar 9 (9)

Information:Unfamiliar 9 (9)Familiar 3 (4)

*(*)

0 (0)0 (0)

1 (1)7 (6)

9 (9)4 (4)

10 (10)10 (10)

1(1)5(5)

0(0)0(0)

0(0)1(1)

0(0)0(0)

NOTE.—N = 10 per condition. Cons. = conservative pattern.* Indicated relation is significant, p < .05 by Fisher s exact test.

The experiment thus offers a clear confirma-tion of the finding that 3-year-old childrentend to assume that a word applied to anunfamiliar object also applies to individualsof the same kind across situations, even ifthe cbildren hear information that suggeststhe contrary.

Extending the Count Nounacross Subtypes

Toy selection task.—We examined theproportion of test trials on which childrenselected either toy of the focal kind outsidethe subtype of the target, as in Experiment1, by adding the proportions of selections ofthe toys marked "Different Fur Color" and"Both Different" in Table 3. (Recall thatonly one child selected the unfamiliar dis-traetor on one trial.)

We entered these scores into an AN-OVA, with introductory cue (Ostension, In-formation) and target familiarity (Unfamiliar,Familiar) as between-subjects factors. As inExperiment 1, there was a significant effectof target familiarity, F(l, 36) = 12.58, p <.01, indicating a higher proportion of differ-ent-subtype selections in the Unfamiliarconditions (M = .48, SD = .21) than in theFamiliar conditions (M = .27, SD = .22).Unlike what we found in Experiment 1, in-troductory cue now had a significant effect,F(l, 36) = 12.58, p < .001; there was ahigher proportion of different-subtype selec-tions in the Information (M = .48, SD = .18)than in the Ostension (M = .27, SD = .24)conditions. Also unlike what we found inExperiment 1, the introductory cue x targetfamiliarity interaction was not significant,F(l, 36) = 2.44, p > .10.

We next classified children according to

the pattern of their selections across all testtrials in the toy selection task. Again, we re-lied on the same criteria as in Experiment 1for assigning children to a subtype-independent, subtype-restricted, or conser-vative interpretation. The third, fourth, andfifth columns of Table 4 show these num-bers. As in Experiment 1, we examined therelation between making or not making asubtype-independent interpretation andseeing either an unfamiliar or a familiar tar-get, using Fisher's exact test. The resultswere similar to those from the context por-tion of Experiment 1. In the Ostension con-ditions, significantly more children made asubtype-independent interpretation if thetarget was unfamiliar than if it was familiar,p < .05. In contrast, in the Information con-ditions, the relation was not significant.

Yes-no task.—The same criteria as inExperiment 1 were used to assign childrento a subtype-independent, subtype-re-stricted, or conservative interpretation. Thenumbers that emerged from this coding ap-pear in parentheses in columns, 3, 4, and 5of Table 4. We again examined the relationbetween making or not making a subtype-independent interpretation and seeing anunfamiliar or a familiar target; the results, byFisher's exact test, were the same as thosedescribed for the previous classification.

In sum, although we closely replicatedthe pattern of tendencies to extend the countnoun across situations, we found some dif-ferences (in the ANOVAs) in the pattern oftendencies to extend the word across sub-types. Recall the previous findings sug-gesting that 2-year-old children are morelikely to make a subtype-independent inter-


pretation in an Unfamiliar- than in a Famil-iar-Ostension condition (Taylor & Celman,1988,1989). We obtained this finding in bothour experiments. However, we had no pre-diction about the role of introductory cue(which had to do with the situation, not thesubtype), or about the role of the interactionbetween target familiarity and introductorycue; thus we have no account to offer of thedifferences we observed between our Ex-periments 1 and 2 in terms of these factors.

Justifications from Yes-No TaskWe asked each of the 40 children for one

set of justifications, yielding 40 sets. Again,the justification sets provide further detailsabout children's interpretation of the novelword. We coded them as basic-level-kind-relevant, subtype-relevant, situation-rele-vant, or other, analogously to what was donein Experiment 1. The experimenter codedall justification sets. A second coder, blindto the hypotheses, then coded them; agree-ment was 95%.

Thirty-three (83%) of the sets fell intoone of the first three categories. Again, wecompared (1) whether children gave a basic-level-kind-relevant, a subtype-relevant, or asituation-relevant justification set, with (2)whether children gave a basic-level kindpattern of answers (i.e., a pattern that wasboth situation-independent and subtype-independent), a subtype-restricted pattern,or a situation-restricted pattern on the yes-notask.

As in Experiment 1, the correspondencebetween yes-no answers and justificationsets was impressive. Of the 33 sets, 24 (73%)were consistent with the pattern of selec-tions on the yes-no task.^ As in Experiment1, there was a difference in this correspon-dence across the Ostension (92%) and theInformation (60%) conditions. The lowerrate of correspondence in the Informationconditions again reflects the fact that inthese conditions (especially in the Unfamil-iar-Information condition) children tendedto give basic-level-kind patterns of answerson the yes-no task, but gave situation-relevant justifications. We interpret this mis-

match as we did in Experiment 1, as re-flecting the fact that children did construethe word as referring to a basic-level kind;however, when asked for a justification afterthe task, these children recalled the informa-tion the experimenter had originally pro-vided, and they felt compelled to use it.

Posttest TaskWe conducted the same posttest as in

Experiment 1 and coded children's answersin an analogous manner, as either consistentor not consistent with a situation-restrictedinterpretation. Again, we were interested inthe correspondence between children's per-formance on the posttest task and the patternof answers they gave on the yes-no task.

First consider the children who made asituation-restricted (or conservative) inter-pretation on the yes-no task, that is, said"yes" only to both animals sitting in a box(or only to the target animal sitting in a box).One child did so in the Familiar-Ostensioncondition, one did so in the Unfamiliar-Information condition, and six did so in theFamiliar-Information condition. The childin the Familiar-Ostension and the child inthe Unfamiliar-Information condition gaveanswers that were not consistent with a situ-ation-restricted interpretation; in fact, bothfailed to answer the questions (i.e., said "Idon't know"). However, all six children inthe Familiar—Information condition re-sponded to the posttest questions in a man-ner consistent with a restricted interpreta-tion. As in Experiment 1, these findings offeradditional evidence that these interpreta-tions were, for the most part, situation-restricted kinds.

Now consider children who made a situ-ation-independent interpretation on the yes-no task, that is, said "yes" to animals of thefocal kind both in and out of boxes. Again,we asked these children only the first post-test question, because they had said "yes"to the animal(s) of the focal kind outside thebox(es). Among the 32 children who madesuch an interpretation, most (26) childrengave answers that were not consistent witha restricted interpretation; some simply said

^ One child in the Familiar-Ostension condition gave a basic-level kind pattern of answerson the yes-no task (i.e., he said "yes" to all four rabbits). However, when asked for justifications,he pointed to one of the rabbits and said, "See, it has two murvils." He then pointed to each ofthe rabbit's ears. This finding shows that the child had made a salient part interpretation (i.e.,RABBIT EAR), not a basic-level kind interpretation (i.e., RABBIT). This is consistent with thefindings of Markman and Wachtel (1988; Experiment 2), and consistent with our general hypothe-ses. We did predict that target familiarity would promote fewer basic-level kind interpretations,hut we had expected the most salient alternatives would be a subtype or a situation-restrictedkind, not a part kind.


"I don't know"; other answers again rangedfrom cute (e.g., "Make some magic," "Witha magic wand") to violent (e.g., "Take itsnose off," "Cut its ears off"). Only six an-swered the question in a manner consistentwith a restricted interpretation; five of thesewere in the Information conditions. We in-terpret these answers as we did in Experi-ment 1, as reflecting the fact that childrenhad made a basic-level kind interpretationof the word during the yes-no task, but whenexplicitly asked to justify their performance,felt compelled to refer to the informationthat they had been given earlier.

In summary, the results of Experiment2 clearly replicated the major findings fromExperiment 1. We again found that the inter-pretation of a count noun applied osten-sively to an unfamiliar object tended to be abasic-level kind, and not a kind includingindividuals tied to a specific context or sub-type. The experiment also again showed thatfamiliarity (knowledge of a basic-level countnoun) and information affected the tendencyto make a situation-independent interpreta-tion of a novel count noun. Furthermore, thefact that Experiment 2 replicated Experi-ment 1 suggests that the findings of Experi-ment 1 were not due to (1) familiarity witha count noun for the situation-restrictedmeaning, (2) the fact that subtype wasmarked by garment color rather than some-thing more inherent, like fur color, or (3)certain procedural details having to do withthe presentation and manipulation of thestimuli.

General DiscussionIn two experiments, our central finding

was that 3V2-year-old children assumed thata novel count noun applied ostensively to asolid unfamiliar object referred to a basic-level kind (e.g., PERSON, DOC) rather thanto a kind that individuates its members bytype of situation (e.g., PASSENCER,PUPPY). We suggest that this finding re-flects an implicit assumption children makeabout the individuals in the developmen-tally privileged kind, namely, that theiridentity is traced across an extensive rangeof situations rather than across a more re-stricted set of situations (contexts or life-phases). Children must make this assump-tion because the individuals in basic-levelkinds and those in situation-restricted kindsare coextensive over certain ranges of situa-tions (e.g., when dogs are young, when per-sons are riding in vehicles), and so ostensive

definition, unaided, does not distinguish be-tween them.

We found that two factors weakenedchildren's tendency to make a situation-independent kind interpretation: familiarityof the target object (i.e., previous knowledgeof a basic-level count noun for tbe kind), andinformation calling for a situation-restrictedinterpretation. Children were more likely tomake a situation-independent kind interpre-tation if the target was unfamiliar than if itwas familiar, and if they learned the wordthrough ostension than if they learned itwith information implying a situation-restricted interpretation. Even if childrenheard explicit information calling for a situa-tion-restricted interpretation, they weremore likely to overlook the information if thetarget was unfamiliar than if it was familiar.The observed familiarity effects are consis-tent with several recent studies showing thatchildren have a greater tendency to interpreta word as referring to a (basic-level) objectkind if the target object is unfamiliar thanfamiliar (e.g.. Hall, 1991; Hall et al., in press;Markman & Wacbtel, 1988; Taylor & Cel-man, 1988).

Consistent with previous results, wealso found that children tended to interpreta count noun applied to an unfamiliar objectas picking out basic-level individuals ratherthan individuals within a given subtype. Wesuggest that the finding reflects another facetof the implicit assumption children makeabout the individuals in the developmen-tally privileged kind: that is, these individu-als share an intermediate level of perceptualsimilarity. In both experiments, we alsodiscovered that target familiarity affectedchildren's tendency to make a subtype-independent interpretation. This findingreplicated a finding that Taylor and Celman(1988, 1989) obtained with 2-year-olds.

The results from these experiments thusadd to recent results that suggest that thebasic-level kind is the privileged interpreta-tion of a word applied ostensively to an unfa-miliar solid object (e.g.. Hall, 1991; Hall etal., in press; Landau et al., 1988; Markman& Wachtel, 1988; Soja et al., 1991; Taylor &Celman, 1988). Once the basic-level objectkind term is acquired, children are morewilling to entertain other interpretations ofnew words. Among the possibilities are aproperty (Hall et al., in press), a materialkind (Markman & Wachtel, 1988; Prasada,1993), the individual itself (Hall, 1991), and,as these experiments show, a subordinate-

1568 Child Developmentlevel kind (see also Taylor & Gelman, 1988,1989), or a situation-restricted kind.

The distinction between basic-levelkinds and situation-restricted kinds is impor-tant because it is directly related to quanti-fication. As noted in the introduction, quan-tification (counting) with basic-level kindmembers (e.g., persons) is different thanquantification with kind members that aretied to speciflc situations (e.g., passengers)(Gupta, 1980). There are many occasions inwhich counting is done over situation-restricted individuals. Aside from airlines(which count passengers), restaurants countdiners, universities count students, and hos-pitals count patients; all of these are situa-tion-restricted individuals. Ghildren ulti-mately must learn about quantifying oversituation-restricted individuals. However,the results of the present experiments sug-gest that for young children, the privilegedway of quantifying is over individuals thatare not tied to restricted situations (for otherdiscussion of what constitutes a countableentity for young children, see Shipley &Shepperson, 1990).

The preceding point about quantifica-tional differences between basic-level kindsand situation-restricted kinds serves to high-light the distinction between basic-level(DOG) and subordinate-level (POODLE)kinds on the one hand, and basic-level(DOG) and situation-restricted kinds(PUPPY) on the other. If an individual in asubordinate-level kind (a poodle) coincideswith an individual in a basic-level kind (adog) in one situation, then it coincides inall situations. For example, the situations inwhich a certain pet (say Fifl) could be con-strued as a poodle are the same as those inwhich she could be construed as a dog. Incontrast, basic-level and situation-restrictedkinds have members that do not coincide inall situations (though they may overlap). Forexample, the situations in which Fifi couldbe construed as a puppy are not all the sameas those in which she could be taken as adog; Fifi could not he construed as a puppywhen she grows up, although she could stillbe construed as a dog. The basic-level ver-sus situation-restricted kind distinction thusdoes not reduce to the basic-level versussubordinate-level kind distinction.

To our knowledge, this is the flrst exper-imental demonstration of learners' prefer-ence for basic-level over situation-restrictedkinds in a word-learning task (for discussionof other issues related to identity and objectkinds, see Keil, 1989; Saltz & Medow, 1971).

Several important issues remain unad-dressed. First, we have examined only ani-mate kinds. We believe that our focus on ani-mate (rather than inanimate) kinds in theseexperiments provides a fair test of our hy-pothesis, because in English, situation-restricted count nouns occur frequently foranimate kinds (especially PERSON). Thereare, however, situation-restricted countnouns that refer to artifact kinds (e.g., a ja-lopy is an old car); more research will beneeded to test the hypothesis that childreninterpret words applied to inanimate objectsas basic-level kind terms, not situation-restricted ones.

Second, our test stimuli did not allow usto examine whether the situation-restrictedinterpretations were restrictions within thebasic level (e.g., BEAR-PASSENGER) orwithin a more superordinate level (e.g., AN-IMAL-PASSENGER). Our procedure lim-ited us to considering restrictions within thebasic level. This is because we did not pre-sent distractor toys (from outside the basic-level kind) situated in the context or life-phase of the target. In pilot testing, we foundthat using a larger number of toys in the testarrays made it difficult to administer a toy-selection task. However, recent results fromexperiments using a different procedure arebeginning to describe the level of abstrac-tion of the situation-restricted interpretation(Hall, in press-b). Furthermore, the fact thatvirtually no children selected either of thedistractor stimuli in our studies may havereflected children's familiarity with the rele-vant superordinate kind, ANIMAL. "Ani-mal" is presumably a familiar count noun for3V2-year-olds, and there is good evidencethat children avoid interpreting two wordsas having the same meaning (Glark, 1987). Itwill be important in future research to selectobjects from unfamiliar superordinate-levelkinds.

Third, our 3V2-year-old subjects werequite advanced learners. Thus, we are un-able to make any claims about the origins ofthis interpretative assumption. However, wedo know that children well below 2 years ofage expect that words applied to solid ob-jects will refer to kinds of objects (Markow& Waxman, 1992; Waxman & Hall, 1993;Waxman & Heim, 1991). Moreover, weknow that even 4-month-old babies expectthat objects will persist and maintain iden-tity across situations (Baillargeon, 1987).Thus, it is possible that the basic-level kindassumption as discussed here guides wordlearning from the outset.


It should be possible to test youngerword learners to see whether they construesolid objects as individuals in basic-levelkinds (e.g., PERSON) or context-restrictedkinds (e.g., PASSENCER). This should bepossible because such a test does not de-pend upon children's having knowledgeabout how objects change through situa-tions, only, for example, knowledge aboutobjects within and outside cars. Such knowl-edge should be available to children as soonas they know about the objects (assumingvery young children do not parse the worlddifferently than adults, thinking that a crea-ture in a car is part of the car). In contrast, itmay be more difficult to test whether veryyoung children construe a word as referringto a basic-level kind (e.g., DOC) rather thana life-phase-restricted kind (e.g., PUPPY),because children may lack sufficient knowl-edge about the specific kind of creature. Wekriow that 3-year-olds expect size differ-ences to be correlated with life-phase differ-ences for a wide range of animal kinds (Ro-sengren et al., 1991). However, we knowless about children under 3. Very youngchildren may not think that dogs and pup-pies are members of the same basic-levelkind (despite the shape similarity betweenpuppies and dogs) because children may beunaware that puppies grow up to be adultdogs. Thus, it is conceivable that very youngchildren might interpret the word "puppy,"if applied to a young dog, as referring to dogsonly when they are young, not because thechildren make a situation-restricted inter-pretation, but because PUPPY is a ("child")basic-level kind (see Mervis, 1987).

The preceding point leads us to a pointabout what we are not claiming in this pa-per. We do not claim that children interpreta word applied to a novel object as pickingout those objects across their life span. How-ever, we do claim that children interpret aword as referring to a basic-level kind, andnot a kind whose members are tied to morespecific situations. (And we accept that the"child" basic-level kind may differ from the"adult" basic-level kind; Mervis, 1987.) Formany kinds (e.g., DOC, CAT, HORSE, andothers whose perceptual form is roughly thesame across the life span), our claim amountsto a claim about identity being traced acrossthe life span. For many kinds, however, ourclaim amounts to a claim about identity be-ing traced only through restricted lifephases. For example, caterpillars grow up tobe butterflies. CATERPILLAR and BUT-TERFLY are distinct basic-level kinds, re-

ferred to in English by separate basic-levelcount nouns, "caterpillar" and "butterfly."We do not expect that children interpret aword applied to a caterpillar (assuming it tobe an unfamiliar object) as meaning CATER-PILLAR-OR-BUTTERFLY. We do expectthat they interpret it as meaning CATER-PILLAR. Interpreting a word as meaningCATERPILLAR implies a commitment totracing individuals' identity across situa-tions, but the situations are tied to the basic-level kind, CATERPILLAR. Even after peo-ple learn about the life-phase link betweencaterpillars and butterflies, we expect thattheir primary construal of a caterpillar is stillas a member of the basic-level kind, CAT-ERPILLAR. We also expect, however, thatknowledge of the life-phase link enablesthem to make statements about identityacross life phases through a kind at a higherlevel of abstraction, such as INSECT, as in"This caterpillar [in reference to one photo]and that butterfly [in reference to a secondphoto, taken at a later time] are the sameinsect"; see Macnamara (1986).

Through the two experiments reportedhere, we have attempted to elucidate the de-fault interpretation children assign to anovel word applied ostensively to an unfa-miliar solid object. We have presented evi-dence that this interpretation is a basic-levelkind. Wbile we have not been concernedwith explaining previously identified prop-erties of basic-level kinds (and we have ac-knowledged that a satisfactory explanationof these properties is lacking), we have beenconcerned with pointing out an unnoticedproperty of these kinds. This property istbat, like all object kinds, basic-level kindsinclude individuals whose identity is tracedacross a certain set of situations. Our argu-ment has been that these situations are moreextensive than those that are associated withmany situation-restricted kinds. Three-and-a-half-year-old children's interpretations ofostensively defined count nouns applied tounfamiliar objects accord with this property,providing new insight into the developmen-tally privileged word meaning.

ReferencesBaillargeon, R. (1987). Object permanence in 3V2-

and 4V2-month-old infants. DevelopmentalPsychology, 23, 655-664.

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

Carey, S. (1982). Semantic development: The


state of the art. In E. Wanner & L. Gleitman(Eds.), Language acquisition: The state of theari (pp. 347-389). Cambridge: CambridgeUniversity Press.

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.

Fodor, J. (1983). The modularity of mind. Cam-bridge, MA: MIT Press.

Gelman, S., & Taylor, M. (1984). How 2-year-oldchildren interpret proper and common namesfor unfamiliar objects. Child Development,55, 1535-1540.

Gentner, D. (1982). Why nouns are learned beforeverbs. In S. Kuczaj (Ed.), Language develop-ment (Vol. 2, pp. 301-330). Hillsdale, NJ:Erlbaum.

Gleitman, L., Gleitman, H., Landau, B., & Wan-ner, E. (1987). Where learning begins: Initialrepresentations for language learning. In F.Newmeyer (Ed.), The Cambridge linguisticssurvey (Vol. 3, pp. 150-193). Cambridge:Gambridge University Press.

Gupta, A. (1980). The logic of common nouns.New Haven, GT: Yale University Press.

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. (in press-a). How parents teach basic-level and situation-restricted count nouns.Journal of Child Language.

Hall, D. G. (in press-b). Individuation and basic-level kinds. Cognition.

Hall, D. G., Waxman, S. R., & Hurwitz, W. (inpress). How 2- and 4-year-old children learnadjectives and count nouns. Child Devel-opment.

Horton, M., & Markman, E. (1980). Develop-mental differences in the acquisition of basicand superordinate categories. Child Develop-ment, 51, 708-719.

Keil, F. (1989). Concepts, kinds, and cognitive de-velopment. Cambridge, MA: MIT Press.

Landau, B., Smith, L., (x. Jones, S. (1988). The im-portance of shape in early lexical learning.Cognitive Development, 3, 299-321.

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

Markman, E. (1989). Categorization and namingin children. Cambridge, MA: MIT Press.

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

Markow, D., & Waxman, S. (1992). The influenceof labels on 12-month-olds' category forma-

tion. Paper presented at the Eighth Interna-tional Conference on Infancy Studies, Miami,FL.

Merriman, W., & Bowman, L. (1989). The mutualexclusivity bias in children's word leaming.Monographs of the Society for Research inChild Development, 54(3-4, Serial No. 220).

Mervis, G. (1987). Ghild-basic categories andearly lexical development. In U. Neisser(Ed.), Concepts and conceptual development:Ecological and intellectual factors in catego-rization. Gambridge: Cambridge UniversityPress.

Prasada, S. (1993). Leaming names for solid sub-stances: Quantifying solid entities in terms ofportions. Cognitive Development, 8, 83-104.

Rosch, E., Mervis, G., Gray, W., Johnson, D., &Boyes-Braem, P. (1976). Basic objects in natu-ral categories. Cognitive Psychology, 8, 382-439.

Rosengren, K., Gelman, S., Kalish, G., & McCor-mick, M. (1991). As time goes by: Children'searly understanding of growth in animals.Child Development, 62, 1302-1321.

Saltz, E., & Medow, M. (1971). Concept conserva-tion in children: The dependence of beliefsystems on semantic representation. ChildDevelopment, 42, 1533-1542.

Shipley, E., & Shepperson, B. (1990). Countableentities: Developmental changes Cognition,34, 109-136.

Shipley, E., & Speike, E. (1988). Ostensive defi-nitions. Unpublished manuscript. Universityof Pennsylvania, Philadelphia.

Soja, N., Carey, S., & Speike, E. (1991). Ontologi-cal categories guide young children's induc-tions about word meaning: Object terms andsubstance terms. Cognition, 38, 179-211.

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

Taylor, M., & Gelman, S. (1989). Incorporatingnew words into the lexicon: Preliminary evi-dence for language hierarchies in 2-year-old children. Child Development, 60, 625-636.

Waxman, S., & Hall, D. G. (1993). The develop-ment of a linkage between count nouns andobject categories: Evidence from 15- to 21-month-old infants. Child Development, 64,1224-1241.

Waxman, S., & Heim, L. (1991). Nouns highlightcategory relations in 13-month-old infants.Paper presented at the biennial meetings ofthe Society for Research in Ghild Develop-ment, Seattle, WA.

Date post:	01-Feb-2017
Category:	Documents
Upload:	nguyenkhanh
View:	216 times
Download:	3 times

Assumptions about Word Meaning: Individuation and Basic-Level ...

Documents