Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Infants' ability to use language as a guide to inductive inference Desjardins, Renée Nicole 1992

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1993_spring_desjardins_renee.pdf [ 2.31MB ]
Metadata
JSON: 831-1.0086094.json
JSON-LD: 831-1.0086094-ld.json
RDF/XML (Pretty): 831-1.0086094-rdf.xml
RDF/JSON: 831-1.0086094-rdf.json
Turtle: 831-1.0086094-turtle.txt
N-Triples: 831-1.0086094-rdf-ntriples.txt
Original Record: 831-1.0086094-source.json
Full Text
831-1.0086094-fulltext.txt
Citation
831-1.0086094.ris

Full Text

INFANTS' ABILITY TO USE LANGUAGE AS A GUIDE TO INDUCTIVEINFERENCEbyRENEE NICOLE DESJARDINSB.A. (Honours) Queen's University, 1985A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF ARTSinTHE FACULTY OF GRADUATE STUDIESDepartment of PsychologyWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIANovember 1992Renee Nicole DesjardinsDepartment ofThe University of British ColumbiaVancouver, CanadaIn presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)Date  DE-6 (2/88)ABSTRACTThe present studies were designed to address whether infants uselanguage to guide their inferences about novel objects. In Study 1, forty-eightinfants from two age groups (16-18 and 20-22 months) participated in a taskwhich used exploratory play as a window on inferential abilities. Infantsexplored novel toys presented in pairs: The first toy produced an interesting butnon-obvious property while the second toy was invisibly altered such that itfailed to produce the target property. In the Labeled condition both toys weremoderately similar in appearance and given a common label. In the UnlabeledControl condition moderately similar toys were talked about in a general way aswere dissimilar toys in the Baseline Control condition. The frequency of infants'attempts to elicit the target property was marginally higher in the UnlabeledControl condition compared to the Baseline Control condition. This suggeststhat infants may have, with difficulty, perceived the greater degree of similaritybetween moderately similar toys in the Unlabeled Control condition comparedto the dissimilar toys in the Baseline Control condition. When toys werelabeled, infants across the 16-22 month range attempted to reproduce the targetproperty with the second toy more frequently than in the Unlabeled Controlcondition indicating that they used the object labels to guide their inferencesabout object properties. Study 2 compared the frequency of infants' attempts toelicit the target property from the second toy when two toys were given acommon label (Same Label condition) as opposed to different labels (DifferentLabel condition). Eighteen 20- to 22-month-olds showed a marginally higherfrequency of target actions in the Same Label condition compared to theDifferent Label condition suggesting that they understood that two toys shouldshare a common label before that label can be used to guide expectationsabout object properties. Together these studies provide evidence that infantsfrom 20 months of age understand that object labels can be used as a guide toinference, and some suggestion that this ability may be present from 16 months.TABLE OF CONTENTSABSTRACT^ iiLIST OF TABLES viLIST OF FIGURES^ viiACKNOWLEDGEMENT viiiINTRODUCTION^ 1STUDY 1^ 7Method 9Subjects^ 9Materials 10Stimuli.^ 10Equipment. 10Design^ 10Labeled condition^ 10Unlabeled Control condition^ 12Baseline Control condition 12Procedure^ 13Coding 14Target Action Coding^ 14Transfer Coding 16Looks Coding^ 17Results and Discussion 17Effect of Similarity^ 19Effect of Labeling 23Transfer^ 24Frequency of Looking^ 25ivVSummary^ 26STUDY 2 26Method^ 28Subjects 28Materials^ 28Stimuli 28Equipment^ 28Design^ 30Same Label condition^ 30Unlabeled Control condition 30Different Label condition^ 30Procedure^ 31Coding 31Target Action Coding.^ 31Looks Coding ^ 33Results and Discussion 33Effect of Labeling^ 35Effect of Same vs. Different Labels^ 40Effect of Different Label vs. No Label 42Frequency of Looking^ 42Summary^ 43GENERAL DISCUSSION^ 43NOTES^ 52REFERENCES^ 53LIST OF TABLESToys, Their Properties, and Target Actions for Study 1^ 11Carrier Phrases Used in Study 1^ 15Toys, Their Properties, and Target Actions for Study 2^ 29Carrier Phrases Used in Study 2^ 32viLIST OF FIGURESThe frequency of target actions on the second toy as a function of condition(Study 1)^ 20Latency to infants' first target action on the second toy as a function of condition(Study1)^ 21The frequency of target actions on the second toy as a function of condition(Study 2)^ 36Latency to infants' first target action on the second toy as a function of condition(Study2)^ 37viiACKNOWLEDGEMENTI would like to thank the parents and infants who participated in these studies.Thanks to CheeSiah Chang, Judith Hole, and Siamac Saatchi for their helpwith coding. Also, to Brigitte Bill, Jane Irwin, Ruth Kirson, and Glynnis Tidball,thank-you for all your moral support and help during data collection. Thebiggest thank-you of all goes to Dare Baldwin for all her help along the way:intellectual guidance, emotional support and material help. And last, but notleast, thank-you to Ross McKitrick for being there for me.viiiINTRODUCTION Everyday object categories, such as dog and chair. tend to grouptogether objects which have a variety of physical properties in common. Quiteoften, however, the properties shared by category members cannot bedetermined by visual inspection alone; they occur at a deeper, non-obviouslevel. For example, one important feature of the category "computer" is a CPU(Central Processing Unit) which is not readily observable. One way in whichobject labels function is to identify members of a category which otherwisemight not be obvious from their surface appearance (e.g., a radio which lookslike a Coke can). Given knowledge that labels function in this way, hearing twoobjects given a common label may lead one to search for the non-obvioussimilarities which are indicated by the common label. There is evidence thatpreschool-aged children already use language in this way to structure theirthinking and expectations about the object world. However, at present little isknown about the development of this ability to use language for conceptualpurpose. The present research explores this developmental question.Current evidence suggests that children of preschool age alreadyappreciate that objects can share hidden properties and they are fairlysophisticated at using both perceptual and linguistic information to guide theirconjectures about such non-obvious commonalties. For example, perceptualfeatures such as wings suggest that a creature can fly even if that creature iscurrently immobile. Massey and Gelman (1988) have shown that children of 3to 4 years of age can go well beyond such simple perceptually-basedinferences. They showed children photographs of unfamiliar mammals, non-mammals and statues with animal-like forms and parts, asking them whetherthe photographed animal could move up and down a hill by itself. Childrenwere remarkably accurate on this task even when the stimulus was an12unfamiliar animal; they predicted that mammals and non-mammals could moveup a hill by themselves. At the same time they predicted that statues whichshared physically similar properties (e.g., legs, etc.) could not, indicating thattheir predictions were based on a complicated set of perceptual cues and werenot just a simple prediction from visibly perceptible parts.In many cases the cues of surface similarity and object labels mayconverge to indicate that the objects in question share deeper, non-obviousproperties. However, some categories exist for which perceptual information iseither not as useful or even misleading as an indicator of membership (e.g.,whales are mammals although they resemble fish). When these perceptualversus linguistic cues lead to opposite predictions about category membership,preschool children are capable of relying on object labels to guide theirinferences about category membership. Gelman and Markman (1986) pittedsurface similarity cues against object labels in order to determine whetherpreschool children would place more importance on the category label orperceptual similarity. Children were shown two familiar objects and taught anew fact about each one (e.g., "This squirrel eats bugs. This rabbit eats grass").Then they were asked which property applied to a third object which looked likesiag of the objects (e.g., the rabbit) but was given the same category label as theother object (e.g.,"squirrel"). For example, children were asked, "Does thissquirrel eat bugs like this squirrel or grass like this rabbit?" If children'sinferences were based on perceptual cues they should choose the propertybelonging to the similar-looking object. On the other hand, if childrenunderstand that objects with common labels have similar non-obviousproperties, then they should choose the property belonging to the object withthe same label as the target object. Gelman and Markman found that childreninferred that the target object had the same property as the object with the same3label. Hence, these children demonstrated an understanding that object labelscan be informative about non-obvious object properties. In a simplified design,Gelman and Markman (1987) replicated this result with three-year-old children.In a series of studies similar in design to the Gelman and Markman(1986; 1987) research, Davidson and Gelman (1990) also found thatpreschoolers tend to use novel labels to guide their inferences about categorymembership. However, in certain circumstances when the linguistic informationwas inconsistent with children's perceptually-based partitioning of the targetand test objects, they relied on perceptual similarity to guide their inferences. Ifchildren saw an object which looked similar to the target object but was given adifferent label and saw another object which looked different from the target butwas given the same label, children did not consistently choose the object withthe same label as having the same property as the target object. In this case,children ignored object labels and relied on similarity since the experimenter'suse of labels was inconsistent with respect to surface cues. Thus, althoughpreschool children are capable of relying on labels to guide their inferencesabout non-obvious object properties, they will not do so if labels are entirelyinconsistent with a perceptually-based nonlinguistic partitioning of the objects inquestion.As previous research demonstrated that preschoolers are able to useobject labels to guide their expectations about object properties (when labelsare not entirely inconsistent with respect to surface cues), Gelman and Coley(1990) investigated whether even younger children's inferences showed thispattern. They found that at 2 1/2 years of age, children accurately inferred anobject's property more often when the test and target objects were given thesame label than when they were not. For example, children were told about thetarget picture that it was a bird and lived in a nest. If they were then shown the4test picture, children were more likely to agree that it lived in a nest if they weretold the picture was of a bird than if neither target or test pictures were labeled.This study was replicated using temporary state adjectives (e.g., "This issleepy") instead of object labels in order to determine whether the effect wasdue merely to a bias to attend to the word associated with each picture. In thiscase, children's accuracy did not differ from chance indicating that they only useobject labels to guide their inferences and not temporary state adjectives.These studies indicate that by 2 1/2 years of age children are already usingobject labels to guide their expectations about object properties. Furthermore,they understand that object labels per se, not temporary state adjectives, can beused in this way.To date however, there is no evidence to clarify whether children underthe age of 2 1/2 use labels to guide their inferences about non-obvious objectproperties. Existing methodologies for testing inferential abilities have a strongverbal component and thus cannot be used for children under 2 years of age.Yet, there is reason to suspect that they may be able to do so. Evidence for twoprerequisite abilities for this competency can be found in younger children:comprehension of object labels and the ability to draw simple inductiveinferences.As early as 9 months of age, infants begin to show signs ofcomprehending words (Oviatt, 1980; 1985). In the next 6 to 12 months theydevelop an even larger comprehension vocabulary and a productivevocabulary of 30 or more words (Benedict, 1979; Goldin-Meadow, Seligman &Gelman, 1976; Huttenlocher, 1974, Nelson, 1988). At approximately 18months, some infants show dramatic increases in their vocabularies--the"naming explosion"--adding in some cases even as many as 10-20 newwords to their productive vocabulary in a single day (Gopnik & Meltzoff, 1986;5Nelson, 1977; 1991). Thus, it seems that before the age of two children mayhave at least the requisite linguistic competency for using language to guideinferences. Furthermore, research suggests that they also have the ability todraw simple inductive inferences.One method for determining infants' inferential abilities is by examiningtheir exploratory play. If after exploring an object with a certain non-obviousproperty, infants expect a second similar-looking object to have the sameproperty, this should be manifested in their attempts to elicit that property. Usingsuch an exploratory play methodology, Baldwin, Markman and Melartin (inpress) have demonstrated that infants as young as 9 months are able to drawinferences about non-obvious properties after only a brief exposure to a novelobject. In one condition (violated expectation), infants explored a novel objectwith a non-obvious and interesting property for 30 seconds, then they exploreda second toy similar in appearance but invisibly altered such that it failed toproduce the interesting property. Because the second toy failed to produce theproperty, infants' actions to elicit the property could be attributed to a priorexpectation. A second condition (boredom) in which infants explored twosimilar novel objects neither of which produced an interesting property wasincluded to distinguish actions which might result from boredom at havingreceived two similar toys in a row from true inferences in the violatedexpectation condition. A third condition (different) was included in which infantsexplored a novel object with a non-obvious and interesting property followed bya second toy dissimilar in appearance which failed to produce an interestingproperty. In both the boredom and different conditions infants explored noveltoys which failed to produce an interesting non-obvious property; these twoconditions provided a base-rate of the frequency with which infants producedthe target actions on a novel toy through miscellaneous exploration. Infants6made a first attempt more quickly and made more frequent subsequent attemptsto elicit the target property from the second toy in the violated expectationcondition than in either the boredom or different conditions suggesting thatinfants were able to use perceptual similarity to guide their expectations aboutthe non-obvious object properties.Altogether it appears that children under 2 years of age may well havethe necessary prerequisites of language comprehension and inferential abilitiesin order to be able to use object labels to guide their inferences. However, it ispossible that the ability to rely on object labels to guide inferences about non-obvious object properties is still beyond their capabilities. Perhaps they are notable to exploit their knowledge of object labels for conceptual purpose. Forexample, children's early words may be simple associations between a wordand an object. Nelson (1985, 1991) suggests that associative word-to-objectlinks may be characteristic in the early phases of word learning. She uses theterm "referential" to describe these early meanings in which labels make littlecontact with children's conceptual knowledge. If infants' comprehension oflabels is comprised of simple word-to-object links, we would not expect them tobe able to use object labels to guide their inferences about non-obvious objectproperties. If on the other hand, their understanding of object labels is such thatmeanings make contact with conceptual knowledge--Nelson refers to this asdenotational meaning and expects meanings to become increasinglydenotational with age--then we might expect that infants may indeed becapable of using language for inferential purposes.The following studies make use of the Baldwin, Markman & Melanin (inpress) exploratory play methodology to examine whether children under 2years of age are able to use object labels to guide their inferences about non-obvious object properties. Specifically, will hearing the same label applied to7somewhat different looking objects lead infants to search for shared non-obvious properties? If so, infants should demonstrate their expectations bytheir manner of exploring the second toy in a pair. If they expect the second toyto produce the same non-obvious property as the first toy, they should be morepersistent in attempts to elicit the target property when the two toys are given acommon label than when they are unlabeled.Infants in two age groups, 16-18 months and 20-22 months, had theopportunity to explore a number of novel toys in pairs. Infants between 20 and22 months were chosen as previous research indicates that by this age childrenhave acquired a sizable comprehension vocabulary of object labels (Nelson,1977). Furthermore, Gopnik and Meltzoff (1987; 1992) observed a relationshipbetween the onset of the naming explosion and an interest in categorizingobjects in infants of 18 months. For infants from 15 to 21 months of age,vocabulary size and performance on categorization tasks were obtained. Asignificant relationship first appeared between vocabulary size and certaincategorization tasks (e.g., basic level sorting) at 18 months. Thus, we mightexpect that using language to guide inferences might appear also during thisperiod and would thus be readily detectable in infants between 20-22 months.A group of younger infants (16- to 18-month-olds) was chosen to examinewhether there is a relationship between age and/or vocabulary size and infants'inferential abilities. Perhaps younger infants might be less adept at usinglanguage to guide inferences.STUDY 1In Study 1, infants explored novel pairs of toys in each of 3 conditions:Labeled, Unlabeled Control and Baseline Control. In each pair of toys, the firsttoy produced a novel, non-obvious property while the second toy was invisiblyaltered such that it failed to produce the target property. As in Baldwin et al. (in8press), the second toy in each pair was disabled in order to show that infants'attempts to reproduce the target property are the result of inferences and notmerely the result of a rewarding object property. Furthermore, novel toys wereused in order that infants' expectations about the second toy would be based onone previous exposure (the first toy) rather than resulting from more generalbackground knowledge.Toys in a pair were chosen to be only moderately similar in appearance(although capable of the same non-obvious property). This is in contrast to theBaldwin et al. studies in which toys were quite similar in appearance. In thepresent study, if toys were too similar in appearance, object labels would notprovide any additional cues above those given by perceptual similarity as to theunderlying non-obvious property shared by the toys.A comparison between the Labeled and Unlabeled Control conditions isof most interest in the present context. If infants understand that toys with acommon label share the same non-obvious property they should expect thesecond toy in a pair to possess the same non-obvious property as the first toy.Thus, they should quickly attempt to reproduce the the target property andpersist in their attempts. Such expectations should not necessarily arise whentoys are not labeled. Furthermore, a Baseline Control condition was includedwhich paired toys dissimilar in appearance. This provided a measure of thefrequency with which infants produce the target action on the second novel toythrough miscellaneous play. If infants' target actions on the second toy aremore frequent in the Unlabeled Control condition than in the Baseline Controlcondition, this would indicate that infants' exploration was the result of inferenceand not merely the result of miscellaneous play.The Baseline Control condition also allowed us to examine whetherinfants were likely merely to repeat on the second toy the target action which9they employed with the first toy even in the absence of perceptual similarity. Forexample, if the property of the first toy was a light activated by a pressing abutton, would infants show pressing actions on a periscope regardless of theobvious lack of similarity? If instances of such transfer are rare, this wouldsuggest that target actions in the Unlabeled Control condition were indeed theresult of inferences based on perceptual similarity and not merely theperseveration of a successful action pattern.Although exploratory play behaviour may be indicative of infants'expectations about toy properties, visual comparison of the toys may revealtheir expectations as well. Mandler (1988) reviews a number of studies inwhich pairs of objects are available for visual inspection. She notes that fromaround 8 months of age infants look back and forth between a novel object anda familiar object in a comparison process which appeared to be different from asimpler process of recognition observed at younger ages. In the present study,infants had the opportunity to look at the first toy in a pair while exploring thesecond toy; the frequency of such looks may be indicative of infants' attempts todiscern similarity between the two toys. If hearing two toys labeled increasesinfants' expectations that the two toys share the same non-obvious property thismight be revealed in more frequent looks in the Labeled condition to the first toyin order to compare the toys and their properties than in either the UnlabeledControl or Baseline Control conditions.Method SubjectsSixty-eight infants in two age groups (16- to 18-month-olds and 20- to 22-month-olds) participated. Twenty infants were eliminated, due to fussiness (9),prior familiarity with the stimulil (5), design criteria (two or more trials eliminateddue to fussiness and familiarity of stimuli) (3), and experimenter error (3),10yielding a final sample of 48 infants: 24 infants in each of the two age groups.All infants were from monolingual English-speaking families, were developingnormally and had no history of serious ear infection. Infants were recruitedthrough newspaper advertisements and postnatal information sessions atmaternity hospitals in the area.Materials Stimuli. Two toy sets were used, comprised of four pairs of toys each.Toy pairs were chosen to be novel, manipulable, interesting and visuallydistinctive from other pairs. Each toy pair consisted of two toys only moderatelysimilar in appearance, although capable of the same non-obvious property. Forexample, one pair consisted of a janitor's retractable key-ring and a plasticretractable measuring tape. Toy pairs were assigned to sets such that the pairsin a set would be maximally distinctive. See Table 1 for a list of the toy pairsthat were used, their target properties and the actions required to elicit theproperties.For all three conditions, the second toy in each pair was disabled suchthat it failed to produce the target property but showed no surface evidence ofthis alteration. For example the measuring tape was glued so that pulling onthe metal tab failed to release the tape from the housing.Equipment. A time-stamped video record provided continuousinformation about the flow of events.Design Each infant explored pairs of toys in three conditions: Labeled,Unlabeled Control and Baseline Control.Labeled condition. Infants explored the first novel toy in a pair which wascapable of producing a non-obvious property. Then infants explored a11Table 1Toys. Their Properties. and Target Actions for Study 1 First Toy^Second Toya^Target Property^Target ActionSet 1:toothbrush case corkscrew^hidden stickers^remove inner portionlittle bobber^big bobber^hinge movement/clacking^lift & drop arm piece(handleless)^(long handle)round magnifier square magnifier centre emerges^hinge tabcollapsible cup periscope^expands/contracts^pull apart endsSet 2:suction cup toy suction soap pad adheres/popping sound^suction against table& liftsharpenerjanitor's key-ringsquareflashlighttelescopemeasuring tapecylindricalflashlightcontracts/expandsautomaticallyhidden chain/automatic retraction/'zippy' soundlight flashesdepress lidpull tab to extend &releasepress on buttona The second toy is disabled such that it fails to produce the target property.1 2second moderately similar novel toy which was unable to produce the non-obvious property. Both toys were given the same novel label.Unlabeled Control condition. As in the Labeled condition, infantsexplored pairs of novel toys which were moderately similar in appearance. Thefirst toy produced a non-obvious property whereas the second toy failed toproduce the non-obvious property. Toys were talked about in a general way butwere not labeled.Baseline Control condition. Infants explored a novel toy which wascapable of producing a non-obvious property. Then they were given a secondnovel toy dissimilar in appearance to the first toy which failed to produce thenon-obvious property. Toys were talked about in a general way but were notlabeled. Dissimilar pairs were generated by taking the first toy from one pairand combining it with the second toy from another pair (e.g., a flashlight waspaired with a measuring tape). Only four dissimilar pair combinations per setwere used out of a possible twelve so that the second toy in a pair appearedequally often in each of the three conditions.All possible orders of the three conditions were used. Order of conditionsand toy pairs were counterbalanced such that each toy pair appeared equallyoften in each condition (i.e., Labeled, Unlabeled Control, Baseline Control).Infants each participated in two blocks of trials. Each block consisted of one trialof each condition. Toy sets were randomly assigned to the blocks such that forhalf of the subjects toy set 1 was assigned to the first block and for the other halfof the subjects toy set 2 was assigned to the first block. Novel labels (i.e., 1;leAand =m) were roughly counterbalanced in assignment by block, toy set andtoy pair.Each block was preceded by a pair of novel identical toys given the samenovel label which were both capable of producing the same interesting and1 3non-obvious property. One pair consisted of identical magnifying glasses on ahandle which bent in the middle; the other pair consisted of binoculars withseveral pieces which moved on hinges. No data were collected from infants'exploration of these toys. These "interim" pairs were included only in order tomaintain infants' interest in the game and to prevent infants from forming aresponse set that might undercut their exploration of the toys (i.e., a notion thatthe second toy never works). The interim pair assignment was counterbalancedby block and toy set; the assignment of labels, mg and tifs, to these interim toyswas counterbalanced by interim toy pair and block.Procedure After a ten minute warm-up period, the infant was seated in an infant seatwhich was attached to a table. The experimenter knelt on the floor on the otherside of the table facing the infant. The infant first explored an interim toy pairfollowed by three toy pairs, one in each of the three conditions: Labeled,Unlabeled Control, and Baseline Control. Then infants explored anotherinterim pair before exploring three more toy pairs, one in each of the threeconditions.First, infants were given one of the toys from the interim pair. For theinterim toy pairs and for toys in the Labeled condition, the toy was labeled sixtimes in motherese (e.g., "I'm going to show you a lama! You want to see themaja? You can play with the /oral" etc.) The experimenter also demonstratedthe property of the toy before handing it over to the infant (e.g., how to make thehandle of the magnifying glass bend). Labeling did aqt occur during thedemonstration in order to help prevent infants from associating the novel labelwith the action rather than the object itself. During the 20.1 toy explorationperiod that followed, the experimenter did not talk about the objects or makeeye-contact but let infants play as they wished without interference. When the1 4za period was up, the experimenter retrieved the toy and placed it on the tabletop out of infants' reach where it remained for the duration of infants' second toyexploration. This technique was used to reduce possible memory demands,because infants were able to look at the first toy as an aid to remembering itsproperty and the target action required.For the second toy, the experimenter did ma demonstrate the toy'sproperty. For the interim pairs and toys in the Labeled condition the toy waslabeled six times before it was handed over to the infant. During the 2.Qa toyexploration period, the experimenter again refrained from talking about theobjects or interfering with infants' play. When the al period was up, theexperimenter retrieved the toy and also collected the first toy from the corner ofthe table. Both toys were put away and the whole procedure was repeated witha new pair of toys.The only difference between the Unlabeled Control and Baseline Controlconditions was the relation between the first and second toys in a pair; toyswere moderately similar in the Unlabeled Control and quite different in theBaseline Control condition. In all other respects the procedures were the same.Toys in both the Unlabeled Control and Baseline Control conditions werepresented in the same fashion as in the Labeled condition except theexperimenter talked about them with six general phrases instead of six labelingphrases. For example, the experimenter said, "I'm going to show you a newone. You want to see it? You can play with it!", etc. Please see Table 2 for theexact list of phrases used in the different conditions.CodingTarget Action Coding. Coders reported the frequency of infants' targetactions with the first and second toys in the Labeled, Unlabeled Control andBaseline Control conditions. The coding categories for each target actionTable 2Carrier Phrases Used in Study 1 Labeled^Unlabeled Control & Baseline Control15First toy:I'm going to show you a /ma!You want to see the joma?You can play with the lona.Look! a loma!You want to play with the /ma?Okay, here's the ima.Second toy:You want to see the next ,toina?I've got a =a!I'm going to show you the next ioma.You want to see this toma?Look at this Loma.A Iona!I'm going to show you a new one!You want to see it?You can play with it.Look! yes!You want to play with it?Okay, here it is.You want to see the next one?I've got it!I'm going to show you the next one.You want to see this one?Look at this one.Ah yes!16were different for each toy pair (see Table 1). The time at which each of theseactions occurred was recorded from the time-stamped video record. In this way,latency to the first target action from the time the toy was presented could becalculated.Coders were blind to experimental condition. This was accomplished byhaving a different individual record the time at which the toy was presented.Using these times, coders then covered all the television screen except for thetime-line and advanced to the beginning of the play period before removing thecover. In this way, coders did not see which toy the infant had previouslyexplored and thus did not know if it was similar to or quite different from thesecond toy. Furthermore, coding was completed with the sound off, so thatcoders could not hear whether the toy had been labeled or not.The reliability of coders' judgments was determined by having anothercoder who was also blind to experimental condition recode all conditions fromeight randomly-selected subjects, resulting in 48 afi trials altogether. Percentagreement was calculated by dividing the number of instances on which codersagreed by the total number of instances observed. Coders demonstrated 88%agreement on the frequency of target actions.Transfer Coding. Instances of transfer of target action from the first toy tothe second toy in the Baseline Control condition were coded by having thecoder observe the infant's actions on the first toy. The number and nature ofactions on the second toy that appeared similar to target actions performed onthe first toy were noted. For example, if pressing a button elicited the targetproperty on the first toy, a flashlight, an infant might also demonstrate a pressingaction on the second (dissimilar) toy, a measuring tape. The coder also ratedthe degree of similarity (1 =not very similar to 5=very similar) and confidence(1 =very uncertain to 5=very certain) that a transfer of action was observed. A17relatively liberal criterion was used in judging instances of transfer: actions ofonly moderate similarity (3 and above) about which the coder was onlymoderately confident (3 and above) were counted as instances of transfer2. Analternative explanation for what might appear to be inferential behaviour is thatinfants merely repeat the last action which produced a successful result. Inorder to give such an explanation a fair test any plausible instance ofperseverative action was included.Looks Coding. Coders, blind to condition, recorded the number of timesinfants looked at the first toy resting on the corner of the table while the secondtoy was available for exploration. To ensure that coders remained blind tocondition, the corner of the screen where the first toy could be seen wascovered. As well, coding was completed with the sound off so that coders couldnot hear whether the toy had been labeled. Coders also judged whether theinfant seemed to look at the first toy for the purpose of comparing the first andsecond toys.The reliability of coders' judgements concerning looking patterns wasdetermined by having another coder, blind to experimental condition, recode allconditions from eight randomly-selected subjects, resulting in 48 2D2 trialsaltogether. Percent agreement was calculated by dividing the number ofinstances on which coders agreed by the total number of instances observed.Coders demonstrated 82% agreement on frequency of looks and 82%agreement on judgements of "comparative" looks.Results and Discussion Two measures of inferential ability were obtained from infants'exploration of the second (disabled) toys: the number of times infants producedthe target action required to elicit the non-obvious property and latency to thefirst target action. If infants perceive the similarity between the two toys in the18Unlabeled Control condition, they should produce more target actions andshow a shorter latency to their first action in the Unlabeled Control conditionthan in the Baseline Control condition.Furthermore, if infants are able to use object labels to guide theirinferences about the non-obvious object properties, infants should show moretarget actions and show a shorter latency to their first target action in theLabeled condition relative to both the Unlabeled Control and Baseline Controlconditions.Only trials in which infants demonstrated the target action on the first toywere included in the analyses. Unless infants displayed at least one action onthe first toy, we could not be sure that they had noticed the toy's non-obviousproperty. If infants had not noticed the property, then they would be unable todraw inferences about a second toy's non-obvious property. Hence, only thosetrials on which infants attempted the target action with the first toy of the pairwere analyzed. This happened rarely: only six out of a total of 288 trials wereeliminated because of failure to attempt the target action on the first toy. And anadditional ten trials were eliminated because of fussiness or prior familiarity oftoys. When trials from both blocks were available, the frequency and latencyscores were averaged across the trials in separate blocks. When only one trialwas available, one score was used in place of an average score for that infant.The raw data included two extreme scores (one 4.5 standard deviationsand the other 6 standard deviations from the mean), leading us to remove theoutliers and submit the remaining data to a mixed design MANOVA with Age asa between-group factor and Condition as a within-subject factor. Furthermore, anatural log transformation of the raw data including extreme scores was carriedout and the transformed data were submitted to a mixed design MANOVA. Asthe pattern of results remained the same using the transformed data, only19results from the analyses using raw data with extreme scores removed arereported here. Non-parametric tests with the outlying scores included wereused as an additional strategy to deal with a distribution skewed by theinclusion of extreme scores.Both frequency and latency measures were included in a mixed designMANOVA with one between-group factor, Age (16-18 months and 20-22months) and one within-subject factor, Condition (Labeled, Unlabeled Control,& Baseline Control). This analysis revealed a significant main effect ofCondition, E (4, 43) = 3.28, 12 < .02, but no significant main effect due to Age3.Univariate analyses of frequency, F (2, 45) = 6.26, g<.005, and latency, F(2, 45)= 4.03, 1:1<.05, also revealed a significant main effect of Condition. Nosignificant interactions were found between Condition and Age in either themultivariate or univariate analyses. The average frequency of target actionsand the average latency to first attempt are displayed with respect to age andcondition in Figures 1 and 2.The significant multivariate and univariate condition effects lead to anumber of questions as to the locus of the effect; these questions will be takenup in turn.Effect of SimilarityThe first question of interest is whether infants noticed the greatersimilarity between the moderately similar toys in the Unlabeled Controlcondition compared to the dissimilar toys in the Baseline Control condition.Multivariate planned contrasts with one within factor, Condition, revealed amarginally significant difference between the Unlabeled Control conditioncompared to the Baseline Control condition in frequency of target actions andlatency to the first action, Hotelling's/ (45) = 1.50, ja<.06 (one-tailed). Univariatecontrasts indicated that the frequency of infants' target actions was greater in321M 16-18 months0 20-22 months20Figure 1. The frequency of target actions on the second toy as a function of condition.(Study 1)Labeled^Unlabeled^BaselineControl ControlConditionlM 16-18 monthsEl 20-22 months21Figure 2. Latency to infants' first target action on the second toy as a function ofcondition (Study 1)Labeled^Unlabeled^BaselineControl^ControlCondition22the Unlabeled Control condition compared to the Baseline Control condition, /(46) = 1.70, 12<.05 (one-tailed). As well, the latency to infants' first target actionwas shorter in the Unlabeled Control condition than in the Baseline Controlcondition, (46) = 2.13, 12<.02 (one-tailed). Wilcoxon Matched-Pairs Signed-Ranks tests confirmed that the frequency of target actions was greater in theUnlabeled Control condition compared to the Baseline Control condition, z -2.26, 2<.02 (one-tailed), and that the latency to the first target action was shorterin the Unlabeled Control condition compared to the Baseline Control condition,z = -2.10, i2<.02 (one-tailed). The greater frequency of target actions in theUnlabeled Control condition (M = 1.53 target actions, aa = 1.05) compared tothe Baseline Control condition (M = 1.22 target actions, aa = 0.99) and theshorter latency to the first target action in the Unlabeled Control condition (M =8.25 seconds, 312 = 5.42) compared to the Baseline Control condition (M =10.73 seconds, 511 = 6.31) suggest that infants noted the greater similaritybetween the moderately similar toys in the Unlabeled Control condition andthus drew inferences about the non-obvious object properties based on theperceived similarity.Because some trials had to be eliminated due to an absence of attemptsto elicit the target property with the first toy, toy set by conditioncounterbalancing was not maintained. We were concerned that the observedcondition effects might be confounded by toy effects. Thus, Conditiondifferences were also examined across toys as opposed to across subjects (i.e.,toy was treated as a random factor). Multivariate planned contrasts with onewithin factor, Condition, yielded no significant difference between the UnlabeledControl and Baseline Control conditions on frequency and latency. Univariatecontrasts were also non-significant indicating no significant difference in eitherfrequency or latency to the first target action in the Unlabeled Control condition23relative to the Baseline Control condition. This suggests that the obtainedeffects using subjects as a random factor must be interpreted with caution.Effect of Labeling The second question of interest is whether infants used the object labelsin the Labeled condition to guide their expectations about the non-obviousobject properties. To answer this question, the frequency of target actions andlatency to the first action in the Labeled condition were compared to theUnlabeled Control condition. Multivariate planned contrasts with one withinfactor, Condition, revealed a significant difference in infants' exploratory actionin the Labeled condition compared to the Unlabeled Control condition,Hotelling's t(45) = 1.82, 2<.05 (one-tailed). Univariate analyses indicated thatinfants demonstrated more target actions in the Labeled condition than in theUnlabeled Control condition, / (46) = 2.22, j2<.02 (one-tailed). However, latencyto the first target action did not differ between the two conditions. Furthermore,Wilcoxon Matched-Pairs Signed-Ranks tests indicated a marginally significantdifference between the Labeled and Unlabeled Control conditions in frequencyof target actions, z = -1.52, 2<.07 (one-tailed), and no significant difference inlatency between the two conditions. The greater frequency of target actions inthe Labeled condition CM = 2.03 target actions, BD.. 1.31) relative to theUnlabeled Control condition (M = 1.53 target actions, a_Q = 1.05) suggests thatinfants used the object labels in the Labeled condition to guide their inferencesabout non-obvious object properties. Against expectations, however, nocorresponding effect of labeling on latency to the first target actions wasobserved (M = 7.53 seconds, 5.2 = 5.43 for the Labeled condition and M = 8.25seconds, B.12 = 5.42 for the Unlabeled Control condition). Perhaps we failed toobserve a difference in latency between the two conditions because a) infants'initial attempt to elicit the target property is based on perceptual similarity alone,24whereas b) hearing the two toys labeled leads to persistence in such attemptsas reflected in the greater frequency of attempts in the Labeled conditionrelative to the Unlabeled condition.Differences between the Labeled and Unlabeled conditions were alsoexamined across toys (as opposed to subjects). Multivariate planned contrastsrevealed that when toys were presented in the Labeled as opposed to theUnlabeled Control condition, the frequency of infants' target actions was greaterand the latency to the first target action shorter, Hotelling's/ (6) = 2.62,12<.02(one-tailed). Univariate analyses indicated that infants demonstrated morefrequent target actions on the toys when toys were shown in the Labeledcondition compared to the Unlabeled Control condition, i (7) = 3.44, 12<.01 (one-tailed); however, no significant effect for latency was obtained. Thus the effectof labeling generalized across toys suggesting that the effects reported above inwhich subjects were used as a random factor can be interpreted withconfidence as indicating that labels were guiding infants' expectations aboutobject properties.TransferThe coding of the rate of transfer of target actions in the Baseline Controlcondition provided some additional information that infants were indeed makinginferences based on physical similarity in the Unlabeled Control conditionrather than merely repeating the last action which yielded an interesting result.Instances of transfer of target action were observed on only 14 of the 93 trials(15%) in the Baseline Control condition (in which toys were dissimilar inappearance) compared to on 67 of the 92 trials (73%) in the Unlabeled Controlcondition (in which toys were moderately similar). For one pair of toys in theBaseline Control condition, the collapsible cup and the measuring tape, thetransfer of target action had a clear perceptual basis although in fact the action25would not have successfully elicited the target property and the toys were, in aglobal sense, quite different in appearance. Excluding this toy pair resulted ininstances of transfer being observed on only 6 of 81 trials (7%) in the BaselineControl condition compared to on 58 of 80 trials (73%) in the Unlabeled Controlcondition. The relatively small number of trials in the Baseline Control conditionon which transfer was observed suggests that instances observed in theUnlabeled Control were indeed due to inferences based on perceptualsimilarity.Frequency of Looking Another question of interest was whether the number of looks to the firsttoy during second toy exploration was influenced by labeling. Although the totalnumber of looks did not differ between the Labeled condition and UnlabeledControl condition or Baseline Control condition, the number of looks judged tobe for the purpose of comparing the two toys did differ between conditions. Thenumber of comparative looks was greater in the Labeled condition than inUnlabeled Control and Baseline Control conditions (x2(2, N = 48) = 8.89,p<.02). Although such comparative looks were not all that frequent, infantswere almost twice as likely to make a comparative look to the first toy in theLabeled condition than in the Unlabeled Control condition or Baseline Controlcondition (33 comparative looks in 93 (35%) trials in the Labeled conditionversus 16 in 96 (17%) trials in the Unlabeled Control and 16 in 93 (17%) trialsin the Baseline Control conditions). Expectations about the common non-obvious properties shared by two similar looking toys given the same label mayhave led infants in the Labeled condition to check the first toy more often to seehow to elicit this property than in the Unlabeled Control and Baseline Controlconditions. This result lends support to the finding that hearing two toys given a26common label leads infants to actively pursue an expectation that the two toysshare common non-obvious properties.SummaryThese analyses suggest that infants perceived the similarity between themoderately similar toys in the Unlabeled Control condition and hence inferredthat the objects possessed similar non-obvious properties. However, this effectmust be interpreted with caution as the analysis using toys as a random factorfails to support this conclusion. More importantly, the analyses provide clearevidence that when infants hear two moderately similar toys given the samename, they expect the two toys to share the same non-obvious property.The present study leaves the source of this labeling effect unresolved. Itis not clear whether infants were indeed paying attention to the fact that the twotoys were given a common label. Perhaps the increased frequency of attemptsobserved in the Labeled condition relative to the Unlabeled Control conditionwas due merely to the presence of any object label. We do not know whetherinfants were sensitive to the common label, per se. A second study attempted toclarify this issue by comparing the frequency of infants' target actions and thelatency to the first action when two toys are given the same label as opposed todifferent labels.STUDY 2In this study, infants explored pairs of moderately similar novel toys inthree conditions: Same Label, Different Label and Unlabeled Control. In theSame Label condition, corresponding to the Labeled condition of Study 1,infants heard two moderately similar toys given a common label. In theUnlabeled Control condition, as in the first study, infants explored pairs ofmoderately similar toys which were not labeled but instead talked about in ageneral way. This condition was included to allow a replication of the labeling27effect of Study 1. In the Different Label condition, infants explored pairs ofmoderately similar toys given different labels (e.g., the first toy was referred to asa pad while the second toy was called a bk).As in Study 1, we expected infants to show a higher frequency of targetactions in the Same Label condition relative to the Unlabeled Control condition.The failure to find a latency difference in Study 1 between the Labeled andUnlabeled Control conditions renders predictions about latency unclear inStudy 2. Of interest is whether the same pattern will be obtained with the newtoy set.If infants are sensitive to the common label shared by the toys in theSame Label condition, then we would expect them to show more frequent targetactions in the Same Label condition relative to the Different Label condition.Again, given that no latency difference was found between the Labeledcondition and Unlabeled Control condition in Study 1, it is unclear what toexpect from the latency measure when toys are given the same versus differentlabels.Further, no clear-cut prediction arose regarding possible differencesbetween the Different Label and Unlabeled Control conditions. It may be thathearing two moderately similar toys given different labels (Different Labelcondition) would inhibit the frequency and lengthen the latency of infants'attempts to produce the target property relative to those in the UnlabeledControl condition. On the other hand, infants may not regard the different labelsas informative in any way about shared versus differing non-obvious properties.If this is the case, then no differences in frequency of target actions and latencyto the first attempt between the Different Label and Unlabeled Controlconditions would be expected.MethodSubjectsFifty-five normally developing infants in each of two age groups (16-18months and 20-22 months) were recruited through newspaper advertisementsand postnatal information sessions at maternity hospitals in the area. Nineteeninfants were eliminated, due to fussiness (12), prior familiarity with the stimuli(2), design criteria (two or more trials eliminated due to fussiness and familiarityof stimuli) (4), and experimenter error (1), yielding a final sample of 36 infants:18 infants in each of the two age groups. All infants were from monolingualEnglish-speaking families, were developing normally and had no history ofserious ear infection.Materials Stimuli. Two toy sets were used, comprised of three pairs of toys each.As in Study 1, toy pairs were chosen to be novel, manipulable, interesting andvisually distinctive from other pairs. Each toy pair consisted of two toys onlymoderately similar in appearance, although capable of the same non-obviousproperty. The toy sets were redesigned with two criteria in mind: first,replacement of toys from Study 1 which were found to be familiar to several ofthe infants thus resulting in eliminated trials, and second, selection of toys whichwould result in a high level of inference. See Table 3 for a list of the toy pairswhich were used, their target properties and the actions required to elicit theproperties.For all three conditions (as in Study 1), the second toy in each pair wasdisabled such that it failed to produce the target property but showed no surfaceevidence of this alteration.Equipment. The same equipment was used as in Study 1.28Table 3Toys. Their Properties. and Target Actions for Study 2First Toy^Second Toya^Target Property^Target ActionSet 1:29lint brushjanitor's key-ringaccordiansqueekerhidden red 'velvety'surfacehidden chain/automatic retraction/'zippy' soundhonking soundcontracts/expandsremove inner partpull tab to extend &releasesqueeze top & bottomtogethercorkscrewmeasuring tapetelescopeSet 2:wheel-shapedsuction toycollapsible cupfoldingbinocularsbarrel-shapedsuction toyperiscopefoldingbrushadheres/popping soundexpands/contractshidden parts/clacking soundpress to suction &liftpull apart sideshinge edgesa The second toy is disabled such that it fails to produce the target property.30Design Each infant explored sets of toys in three conditions: Same Label,Unlabeled Control and Different Label.Same Label condition. Infants explored the first novel toy in a pair whichwas capable of producing a non-obvious property. Then infants explored asecond novel toy which was moderately similar to the first toy. The second toywas unable to produce the non-obvious property. Both toys were given thesame novel label.Unlabeled Control condition. As in the Same Label condition, infantsexplored the first novel toy in a pair which was capable of producing a non-obvious property. Then infants explored a second novel toy which wasmoderately similar to the first toy and unable to produce the non-obviousproperty. Toys were talked about in a general way but were not labeled.Different Label condition. Infants explored the first novel toy in a pairwhich was capable of producing a non-obvious property. Then infants exploreda second novel toy which was moderately similar to the first toy and unable toproduce the non-obvious property. The first and second toys were givendifferent novel labels.All possible orders of the three conditions were used. Order of conditionsand toy pairs were counterbalanced such that each toy pair appeared equallyoften in each condition (Same Label, Unlabeled Control, Different Label). Twoblocks of trials were run. Each block consisted of one trial of each condition.Toy sets were randomly assigned to the blocks such that for half of the subjectstoy set 1 was assigned to the first block and for the other half of the subjects toyset 2 was assigned to the first block. Three novel labels were assigned to toyset 1 (A, totOma, 1211) and three to toy set 2 (mido, nânaku, fug). Labels withina set were chosen to maximize auditory distinctiveness so that in the Different31Label condition especially, the labels would stand out as being clearly different.Each label appeared equally often assigned to the first as opposed to thesecond toy in a pair and equally often in each condition (Same Label versusDifferent Label).Infants had the opportunity before the experimental trials to play with apair of familiar toys (e.g., dog and ball). No data were be collected from infants'exploration of these toys. These pairs were included only as "warm-up" toys tohelp infants become comfortable with the experimental situation. These toyswere also presented again between blocks of trials to give infants theopportunity to engage in a less demanding task (i.e, a familiar game withfamiliar objects and familiar labels).Procedure The procedure was essentially the same as in Study 1 with threeexceptions: first, the experimenter played a hide-and-seek game with the infantsusing a dog, a ball and a basket before each block of trials; second, theexperimenter continued to look at the infant during toy exploration, rather thanlook away, to minimize fussiness; and third, labeling phrases were shortened inorder to reduce the amount of time the experimenter spent talking. Please seeTable 4 for the list of carrier expressions used.CodingTarget Action Coding. The coding procedure was the same as in Study1. The reliability of coders' judgments was determined by having another coderwho was also blind to experimental condition recode all conditions for eightrandomly-selected subjects, resulting in 48 2Qa trials altogether. Percentagreement was calculated by dividing the number of instances on which codersagreed by the total number of instances observed. Coders demonstrated 90%agreement on the frequency of target actions.Table 4Carrier Phrases Used in Study 2 Same Label & Different Label^Unlabeled Control32First toy:I'm going to show you a poi.Yes, a aeri.Want to see the aeti?Look! a =0You want the ped?Here's the pad.Second toy:Guess what I'm going to show you next?A bllslaSee the IA?Look a las.Yes, a ja.You want the Ws?Here's the la!I'm going to show you a new one.Yes, uh huh.Want to see it?Look! uh huh!You want it?Here it is.Guess what I'm going to show you next?Uh huh!See this one?Look at this.Yes, uh huh.You want it?Here it is!a In the Same Label condition, this label would be the same as the one used for thefirst toy.33Looks Coding.. The coding procedure was the same as in Study 1. Thesame coder coded looking behaviour in both Studies 1 and 2. Because thecoding scheme for both studies was identical and the coder's reliability wassufficiently high, an additional measure of reliability for Study 2 looks codingwas not considered necessary.Results and Discussion As in Study 1, two measures of inferential ability were obtained frominfants' exploration of the second (disabled) toys: the number of times infantsproduced the target action required to elicit the non-obvious property andlatency to the first target action. If infants are able to use object labels to guidetheir inferences about the non-obvious object properties, we would expectinfants to show more target actions in the Same Label condition relative to theUnlabeled Control condition. However, given the results of Study 1, it is unclearwhether latency to infants' first target action will differ between the Labeledcondition and the Unlabeled Control condition.Furthermore, if infants understand that two objects must have the same label before that label can be used to guide inferences about object properties,then we would expect infants to show more target actions on the second toy inthe Same Label condition relative to the Different Label condition. Asmentioned earlier, given that in Study 1, labeling toys did not decrease thelatency to the first target action compared to when toys were unlabeled, it isunclear whether a difference in latency in the Same Label compared to theDifferent Label condition will be observed.As in Study 1, only trials in which infants demonstrated the target actionon the first toy were included in the analyses. Unless infants displayed at leastone action on the first toy, we could not be sure that they had noticed the toy'snon-obvious property. If infants had not noticed the property, then they would34be unable to draw inferences about a second toy's non-obvious property.Hence, only those trials on which infants attempted the target action with the firsttoy of the pair were analyzed. This happened rarely: only two out of a total of216 trials were eliminated because of failure to attempt the target action on thefirst toy. And an additional thirteen trials were eliminated because of fussinessor prior familiarity of toys. When trials from both blocks were available, thefrequency and latency scores were averaged across the trials in separateblocks. When only one trial was available, one score was used in place of anaverage score for that infant.Because Study 1 revealed differences between conditions on thefrequency measure but not on the latency measure, univariate analyses werecarried out rather than multivariate analyses. The frequency data weresubmitted to a mixed design ANOVA with one between-group factor, Age (16-18months & 20-22 months) and one within-subject factor, Condition (Same Label,Different Label, Unlabeled Control). Univariate analyses of frequency revealeda significant effect of Age on frequency of target actions, F(1,34) = 9.84, p<.004.Twenty- to 22-month-old infants produced on average significantly more targetactions during the 20s play period than did 16- to 18-month-old infants (M =1.87 target actions, ap, = 1.35, and M = 1.13 target actions, BD = 0.98,respectively). No significant main effect for Condition was noted. Furthermore,no significant interactions were found between Condition and Age on frequencyof target actions. A natural log transformation of the raw data was carried out tocorrect for skewed data and the transformed data submitted to a mixed designANOVA. As the pattern of results remained the same using the transformeddata, only results from the analyses using raw data are reported here.When latency data were submitted to a mixed design ANOVA with onebetween-group factor, Age (16-18 months & 20-22 months) and one within-35subject factor, Condition (Same Label, Different Label, Unlabeled Control), nosignificant main effects or interactions were observed. Furthermore the patternof results remained the same when natural log transformed data were used.As the significant overall Age difference indicated that the frequency of16- to 18-month-olds' target actions was depressed relative to that of the 20- to22-months-olds, condition differences may have been obscured. For thisreason, separate analyses were conducted for each age group in addition toanalyses collapsed across age group.A repeated measures ANOVA with one within-factor, Condition (SameLabel, Different Label & Unlabeled Control) was carried out for the 16- to 18-month-old group revealing no significant main effect for Condition on frequencyof target actions. Furthermore, the same analysis examining the latency datarevealed no significant main effect for Condition.Likewise, repeated measures ANOVAs with one within-factor, Condition(Same Label, Different Label & Unlabeled Control) carried out for the 20- to 22-month-old group revealed no significant main effect for Condition for eitherfrequency of target actions or latency to the first action. The average frequencyof target actions and the average latency to first attempt are displayed withrespect to age and condition in Figures 3 and 4.Because a priori predictions were made about paired comparisons,comparisons between conditions were made even though there was no maineffect of Condition on either the frequency or latency measures.Effect of Labeling To determine whether the effect of labeling on infants' inferentialbehaviour found in Study 1 was replicated in Study 2, a univariate plannedcontrast with one within-factor, Condition, was carried out. A significant effect oflabeling was noted; infants demonstrated more target actions in the SameM 16-18 months0 20-22 monthsSame^Different^UnlabeledLabel Label^ControlCondition36Figure 3. The frequency of target actions on the second toy as a function of condition(Study 2).15oSameLabelDifferentLabelConditionUnlabeledControl10M 16-18 monthsI2 20-22 months37Figure 4. Latency to infants' first target action on the second toy as a function ofcondition (Study 2)38Label condition than in the Unlabeled Control condition, j(34) = 1.66,12...05(one-tailed) (M = 1.74 target actions, .52 = 1.36 in the Same Label condition, M= 1.25 target actions, BD, = 1.08 in the Unlabeled Control condition). As inStudy 1, Wilcoxon Matched-Pairs Signed-Ranks tests were carried out as well,revealing a significant difference between the Same Label and UnlabeledControl conditions in frequency of target actions, z.= -1.84, .12<.04 (one-tailed).These results replicate the labeling effect found in Study 1: in the Same Labelcondition, labeling leads infants to persist in attempts to replicate the targetproperty as reflected in the greater frequency of target actions in this conditionrelative to in the Unlabeled Control condition.Using the latency data, a univariate planned contrast with one within-factor, Condition, was carried out. Latency to the first target action did not differbetween the two conditions (M = 8.46 seconds, .E.I2 = 6.80 in the Same Labelcondition, .P4 = 8.95 seconds, 5.12 = 6.94 in the Unlabeled Control condition). AWilcoxon Matched-Pairs Signed-Ranks test also failed to show a significantdifference in latency between the two conditions. Thus, as in Study 1, nodifference in the latency to the first target action was observed between the twoconditions.Analyses Across Toys. As in Study 1, Condition differences were alsoexamined across toys as opposed to across subjects (i.e., toy was treated as arandom factor). No reliable differences in either frequency of target actions orlatency to the first target action were observed between the Same Labelcondition and Unlabeled Control condition. A Wilcoxon Matched-Pairs Signed-Ranks test using frequency data approached significance: the frequency oftarget actions was somewhat greater in the Same Label condition relative to theUnlabeled Control conditions, z = -1.36, p<.09. The failure to find significance39using toys as a random factor suggests that the obtained effects using subjectsmust be interpreted with caution.16- to 18-Month-Olds. To examine whether the labeling effect onfrequency of target actions held for each age group, univariate plannedcontrasts were carried out separately for each age group. For the 16- to 18-month-old age group, the univariate contrast did not yield a significantdifference between the Same Label and Unlabeled Control conditions infrequency of target actions (M = 1.14 target actions, 5_12 = 1.03 in the SameLabel condition, M = 1.00 target actions,^= 0.80 in the Unlabeled Controlcondition). Furthermore, a univariate planned contrast showed no significantdifference in latency between the two conditions (M = 11.04 seconds, au. . 7.05in the Same Label condition, M = 8.80 seconds, au . 7.42 in the UnlabeledControl condition). Wilcoxon Matched-Pairs Signed Ranks tests were also notsignificant for either the frequency or latency measures.20- to 22-Month-Olds. When parallel analyses to those performed for theyounger age group were carried out for the older age group, an effect oflabeling on frequency of target actions was found as well as signs of an effecton latency not revealed by the analyses collapsed across age group. Theunivariate contrast indicated that infants showed a marginally higher frequencyof target actions in the Same Label condition than in the Unlabeled Controlcondition, j(17) = 1.66, ,r2<.06 (one-tailed) (M = 2.33 target actions, 5L) = 1.40 inthe Same Label condition, M = 1.50 target actions, .aj2 = 1.27 in the UnlabeledControl condition). Furthermore, a marginally shorter latency to the first targetaction in the Same Label condition relative to the Unlabeled Control conditionwas observed, 1(17) = 1.46, 2..08 (one-tailed) (M = 5.88 seconds, 512 = 5.59 inthe Same Label condition, M = 9.09 seconds, 5_12 = 6.63 in the UnlabeledControl condition). These effects were confirmed by the non-parametric40analyses. Wilcoxon Matched-Pairs Signed Ranks tests indicated that thefrequency of target actions was indeed significantly higher in the Same Labelcondition than in the Unlabeled Control condition, z = -1.71, p<.05 (one-tailed)and that the latency to the first target action was marginally shorter in the SameLabel condition relative to the Unlabeled Control condition, z = -1.46, p=.07(one-tailed). Thus, the 20- to 22-month-old infants showed the same patternwith respect to frequency as the contrasts collapsed across age group showed:infants produced more target actions in the Same Label condition than in theUnlabeled Control condition. Furthermore, the 20- to 22-month-old infantsshowed a somewhat shorter latency to the first target action in the Same Labelcondition relative to the Unlabeled Control condition--an effect which was notrevealed by the contrasts collapsed across age group.Effect of Same vs. Different Labels The major question of interest in Study 2 was whether infantsdemonstrated a sensitivity to the fact that two toys must share a common labelbefore that label can be used to guide inferences about shared non-obviousproperties. To answer this question, univariate planned contrasts with onewithin-factor, Condition, were carried out to determine if the frequency of targetactions or the latency to the first target action differed between the Same Labelcondition and the Different Label condition. The univariate contrast usingfrequency data indicated no significant difference in frequency between the twoconditions although the means are in the predicted direction (M = 1.74 targetactions, 52 = 1.36 in the Same Label condition, M = 1.51 target actions, 52 =1.23 in the Different Label condition). Moreover, a univariate contrast usinglatency data also failed to show significant condition differences (M = 8.46seconds, a/ 6.80 in the Same Label condition, M = 9.17 seconds, 512 = 7.00in the Unlabeled Control condition). Wilcoxon Matched-Pairs Signed-Ranks41tests also failed to reveal significant differences in frequency and latencybetween the Same Label and Different Label conditions. Furthermore, whencondition differences were examined across toys, neither frequency nor latencydiffered between the Same Label and Different Label conditions. Previouslyreported analyses have demonstrated that collapsing across age groupsobscures effects, therefore, condition differences will again be examined foreach age group separately.16- to 18-Month-Olds. The same pattern of results was found for theyounger age group. The univariate contrast indicated no significant differencein frequency between the Same Label and Different Label conditions (M = 1.14target actions, 5.2 = 1.03 in the Same Label condition, M . 1.25 target actions,B.12 = 1.13 in the Different Label condition). Furthermore, latency did not differbetween the conditions either (M = 11.04 seconds, BD = 7.05 in the SameLabel condition, M = 9.50 seconds, BD = 7.33 in the Unlabeled Controlcondition).20- to 22-Month-Olds. In contrast to the analysis performed on the datacollapsed across age group, differences between the Same Label and DifferentLabel conditions were observed in both frequency of target actions and latencyto the first action for the older age group. The univariate contrast usingfrequency data revealed that infants showed a marginal increase in targetactions in the Same Label condition relative to the Different Label condition,1(17) = 1.56, p<.07 (one-tailed) (M = 2.33 target actions, BD. = 1.40 in the SameLabel condition, M = 1.78 target actions, BD = 1.31 in the Different Labelcondition). Furthermore, a univariate contrast revealed a shorter latency to thefirst action in the Same Label condition than in the Different Label condition,t(17) = 1.77, p<.05 (one-tailed) (M = 5.88 seconds, ai2 5.59 in the Same Labelcondition, M . 8.83 seconds, aD. = 6.85 in the Different Label condition). These42findings were confirmed by Wilcoxon Matched-Pairs Signed-Ranks tests, z = -1.62, p=.05 (one-tailed) for frequency, and z = -1.61, p=.05 (one-tailed) forlatency. Twenty- to 22-month-olds understand that two toys should have thesame label before that label can be used to guide inferences about non-obvious object properties as indicated by the higher frequency of target actionsand shorter latency to the first target action in the Same Label condition relativeto the Different Label condition.Effect of Different Label vs. No Label An additional contrast of interest involved the Different Label andUnlabeled Control conditions. Recall that we had no a priori expectations aboutthe frequency of target actions or latency to first action in the Different Labelcondition relative to the Unlabeled Control condition. No significant differencesbetween the two conditions in either frequency or latency were noted by theunivariate contrasts or the Wilcoxon Matched-Pairs Signed-Ranks tests. Theanalyses across toys also failed to yield significant differences for frequencyand for latency between the Different Label condition and Unlabeled Controlcondition. Contrasts carried out for each age group separately also failed toyield significant differences between the Different Label and Unlabeled Controlconditions on both frequency and latency measures.Frequency of Looking As in Study 1, we examined whether the number of looks to the first toyduring second toy exploration was influenced by labeling. The total number oflooks did not differ between the Labeled condition and Unlabeled Controlcondition or Different Label condition. And, in contrast to Study 1, the number oflooks judged to be for the purpose of comparing the two toys did not differbetween the Same Label and Unlabeled Control conditions. Moreover, noother condition differences in number of comparative looks were revealed.43Failure to observe a difference in the number of comparative looks between theSame Label condition and Unlabeled Control condition in this study may havebeen due to a methodological difference between Studies 1 and 2. In Study 1,the experimenter averted her gaze during toy exploration whereas in Study 2she looked at the infant. Perhaps having the experimenter's face to look at mayhave competed with the first toy as a possible source of information for the infantand thus inhibited the number of comparative looks towards the first toy.SummaryThese analyses partially replicate the labeling effect found in Study 1: inthe Same Label condition, labeling led infants to persist in attempts to replicatethe target property as reflected in the greater frequency of target actions in thiscondition relative to in the Unlabeled Control condition. Furthermore, for infantsof 20 to 22 months, an increased level of inferences occurred only when toyswere given a common label. No such effects of labeling or common labels wereobserved for the younger age group, perhaps because of the overall lowerfrequency of target actions in Study 2 relative to Study 1. Hence we are unableto clarify whether infants of 16 to 18 months of age are sensitive to the fact thattwo toys which share a common label share deeper, non-obvious properties.GENERAL DISCUSSIONThe results of Studies 1 and 2 provide new and interesting informationabout infants' inferential abilities. In Study 1 a comparison of the frequency oftarget actions and latency to the first action in the Unlabeled Control conditionrelative to Baseline Control condition provided evidence that infants were ableto perceive the similarity between the only moderately similar toys and, basedon that similarity, were able to draw inferences about non-obvious objectproperties. Furthermore, additional evidence provided by the Baseline Controlcondition that transfer of target actions occurred infrequently indicates that44infants' attempts to elicit the target property in the Unlabeled Control conditionwere not simply perseveration of a previously successful action. This findingreplicates and extends the findings of Baldwin, et al. (in press) by demonstratingthat infants are able draw inferences based on only moderate perceptualsimilarity, at least by 16-22 months.Furthermore, Studies 1 and 2 together suggest that infants use objectlabels in addition to perceptual similarity to guide their inferences about non-obvious object properties. However, in Study 2, when the data for each agegroup was examined separately, the labeling effect was not obtained for the 16-to 18-month-olds. In contrast, a labeling effect was noted for the 20- to 22-month-olds. Moreover, in Study 2, weak support--as indicated by marginaldifferences in frequency and in latency-- was obtained for the suggestion thatinfants, at least by 20 months of age, understand that two toys must share acommon label before that label can be used to guide expectations about objectproperties. Stronger support awaits a further replication with a toy set whichelicits a higher level of target actions, thus giving more room for conditiondifferences to occur.There were some minor methodological differences between Studies 1and 2 which may underlie the failure to replicate the labeling effect for the 16- to18-month-old group. In Study 1 before each block of toys, infants played with apair of novel identical toys given the same label which were both capable ofproducing the same interesting and non-obvious property. These "interim" pairswere included in order to maintain infants' interest in the game and to preventinfants from forming a response set (e.g., the second toy never works). Wechose to replace these pairs in Study 2 with two familiar toys (a dog and a ball)and a basket in a hide-and-seek game. We were concerned that the additionallabels required for this second study to achieve the different label condition45might be overwhelming for such young children. For this reason, we chose aless demanding but still interesting activity to maintain infants' cooperation inthe game. Perhaps, however, the interim pairs of toys in Study 1 served animportant purpose by helping infants to realize that a familiar piece of real-worldknowledge was applicable in the current game--the notion that two identicaltoys with the same name share the same non-obvious property. Furthermore,omitting these pairs may have increased the chances that infants formed aresponse set--the second toy never works--simply because the proportion ofnovel toys that did not work was now half whereas in Study 1 it was only slightlymore than one-third.One issue which awaits future resolution is the degree of similaritynecessary between two objects before infants will rely on object labels to guidetheir inferences. In these studies toys were chosen to be only moderatelysimilar in appearance; if toys were very similar in appearance object labelswould not provide any additional cues as to the underlying non-obviousproperty shared by the toys above those given by perceptual similarity.However, using toys only moderately similar in appearance may have partlymasked the effects of same and different labels. If the objects had been moresimilar in appearance, even the younger infants (16- to 18-month-olds) mighthave demonstrated an understanding that two toys sharing the same labelshare the same non-obvious properties. Perhaps, additionally, the degree ofperceptual similarity necessary before infants will rely on object labels to guidetheir inferences changes with age. It was not a goal of this study, however, tomanipulate similarity in any fine-grained way in order to determine therelationship between degree of similarity and use of object labels as a guide toinference. Further research is necessary to determine whether infants ofdifferent ages will use object labels to guide their expectations about non-46obvious object properties when perceptual similarity is even minimal ornonexistent.It might be argued that young infants would be more likely to uselanguage as a guide to inference when familiar, rather than novel, labels andcategories are used. However, novel exemplars would still be required as testtoys. In particular, a novel exemplar dissimilar in appearance would benecessary so that infants failed to recognize it as a member of the familiarcategory based on perceptual features alone. If infants immediately recognizedthe test toy as belonging to a familiar category, no additional information wouldbe provided by the label as to category membership over and above thatalready indicated by perceptual features. However, with a novel exemplardissimilar in appearance, the situation becomes one in which infants must relyon the object label, in the absence of perceptual similarity, as a guide toinference. If infants are unable to use novel labels to guide expectations whentwo novel toys are moderately similar in appearance, it seems doubtful that theywould be willing to extend a familiar label to a novel exemplar in the absence ofperceptual support. Moreover, as described earlier, Davidson and Gelman(1990) found that even preschoolers are reluctant to rely on object labels iflabels are entirely inconsistent with a perceptually-based nonlinguisticpartitioning of the test objects.The youngest infants to participate in these studies were 16 months ofage; since younger infants have not been tested, there is no definite reason tobelieve that using object labels as a guide to inference is beyond theircapability. From previous research it is evident that infants show basicinferential skills as early as 9 months of age (Baldwin, et al., in press).However, the earliest age at which there is reliable evidence of languagecomprehension is around 10 to 12 months of age (Oviatt, 1980;1985).47Therefore, it is possible that infants are capable of using object labels to guideinferences from the start of language comprehension. Evidence from otherdomains suggests that a task which does not require the infant to activelymanipulate toys might be a more sensitive window on infants' abilities. Forexample, Baillargeon and colleagues (e.g., Baillargeon, Spelke & Wasserman,1985; Baillargeon, 1987) demonstrated that infants as young as 3 1/2 monthsunderstand that an object continues to exist when hidden--several monthsbefore they show such understanding using Piagetian search tasks whichrequire infants to actively search a spatial array. Furthermore, Baillargeon(Baillargeon, DeVos & Graber, 1989) showed that even for 9-month-old infants,a task which requires the infant to perform an action, such as the AB searchtask, may not be the best indicator of their understanding of object permanence.Baillargeon, DeVos & Graber (1989) attribute 9-month-olds' failure on the ABsearch task to an inability to integrate memory information into the planning andexecution of action. They believe there may be a developmental lag betweencognition and action in which understanding precedes the ability to act onunderstanding. If there is a lag between cognition and action, then using a non-action methodology (e.g., an adaptation of the preferential-lookingmethodology) to investigate infants' use of object labels as a guide to inferencemight reveal an early emerging understanding and provide a more sensitivewindow on developmental change.These studies leave unanswered whether the observed effects oflabeling on inferential abilities are indeed language-specific. Perhaps theseeffects are merely the result of an association between any sound (in this casean object label) or, for that matter, any other associate, and the non-obviousobject properties. If labels are functioning in a simple associative manner, thenother non-linguistic stimuli, such as a whistle sound or a coloured star, which48covary with non-obvious properties should also function to guide inference. Forexample, would exploring a novel toy on which the experimenter stuck a bluestar lead infants to expect a similar-looking toy which also received a blue starto share the same non-obvious property? Such controls would help clarifywhether the effects observed in this study are the result of a specific property ofobject labels or merely the result of general associative mechanisms. GivenWaxman and Balaban's (1992) evidence which suggests that 9-month-olds usewords, but not tones, to categorize objects, we might expect that 16-month-oldswould use object labels, rather than any other associate to guide inferencesabout non-obvious object properties.Even if infants restrict inferences to words rather than any associate, it isnot clear whether they would regard any common words as candidates forguiding inferences or whether they would require common object labels, per se.For example, hearing the adjective "blueish" applied to two objects might leadinfants to expect that both objects shared a common non-obvious property, suchas a retractable chain. There is evidence that by 2 1/2 years of age children willuse only object labels, not temporary state adjectives (e.g., "sleepy"), to guidetheir inferences about category membership (Gelman & Coley, 1990), but nosuch evidence yet exists for infants. Based on their research, Markow andWaxman (1992) suggest that 12-month-olds will use both nouns and adjectivesfor categorization purposes. However, it may well be that by 16 months of age,infants understand that only object labels (nouns) can be used forcategorization. Implicit in this issue are two assumptions: (1) that infants canrecognize when they have heard an object label as opposed to another kind ofword; and (2) that infants understand that the label applies to the whole objectrather than to one of its properties or its substance, for example. Theseassumptions will be taken up in turn.49For the assumption that infants can distinguish labels from other types ofwords, there is as yet no direct evidence. However, several researchers haverecently provided evidence that from 17 months of age, infants have someknowledge of syntactic structure which may allow them to differentiate labelsfrom other words (Gleitman, 1990; Hirsh-Pasek, Golinkoff, Fletcher, DeGaspeBeaubien, & Cauley, 1985; Naigles, 1990). In a preferential looking task thatHirsh-Pasek et al. devised, infants demonstrated syntactic knowledge bylooking longer at the video display matching the sentence they heard; forexample, if they heard "Big Bird is tickling Cookie Monster," infants lookedlonger at the screen depicting that event than at the one depicting CookieMonster tickling Big Bird. Using a similar procedure, Naigles demonstrated thattwo-year-olds use syntactic knowledge to map a novel action term onto thecorrect event. Based on these and other similar findings, Gleitman argues thatvery young children use syntactic knowledge to acquire new action terms.Moreover, Katz, Baker and MacNamara (1974) demonstrated that from 18months of age, female infants use syntactic cues to distinguish betweencommon nouns and proper nouns. Gelman and Taylor (1984) replicated thisfinding with two-year-olds of both sexes using a more elaborate design.Perhaps, then, children before the age of two are sensitive to the syntactic cueswhich differentiate category labels from other terms such as attributedescriptions.With regard to the second assumption, there is some evidence whichsuggests infants understand that object labels refer to the whole object ratherthan to its substance, for example. Some researchers suggest that when facedwith a novel object for which they do not know the name, infants' defaultassumption is to treat a novel label as applying to the whole object rather thanto its parts, substance or other properties (e.g., Markman, 1989; Markman &50Wachtel, 1988; Woodward, 1992). In the studies reported here, infants werefaced with exactly the situation in which this default assumption is said to beused: a novel label is applied to a novel object. Therefore, it is possible thatinfants treated the novel labels as object labels rather than attribute orsubstance terms, for example. Although this issue warrants further researchbefore it can be stated definitively that infants use object labels qua objectlabels to guide their inferences, the current methodology provides a frameworkwithin which such issues can be addressed.What would it mean to the course of development that the ability to useobject labels as a guide to inference is already available in infancy? Firstly, itwould mean that infants could rapidly acquire new concepts and thus quicklyexpand their knowledge about the object world. Without such recourse toobject labels, infants would have to explore each and every object individuallyto learn about its hidden properties unless perceptual similarity led them torecognize it as a familiar item. Thus, using object labels to guide inferencesspeeds conceptual development. Secondly, the ability to use language toguide inference would likely speed language development, as well asconceptual development, for the infant would be motivated to acquire newlabels as a means to gaining information about the world. If object labels were,for the infant, merely a means of requesting a desired object, vocabularyacquisition might proceed more slowly since there are many objects an infantmight be uninterested in possessing. An interesting possibility is that the abilityto use object labels as a guide to inference may be related developmentally tothe rapid increase in vocabulary--for example, the "naming explosion"--thatbegins in the later half of the second year of life (Benedict, 1979; Goldin-Meadow, Seligman & Gelman, 1976; Gopnik & Meltzoff, 1986, 1987, 1992;Nelson, 1977, 1985, 1991; Oviatt, 1980,1985). All in all, the early emergence ofcontact between language and conceptual knowledge provides infants withexcellent tools for quickly increasing their linguistic repertoire whilesimultaneously gaining knowledge and understanding of the world.5152NOTES1 Parents were asked to indicate whether their infant had played before with anyof the toys or with any toys similar in appearance and non-obvious property.Infants for whom toys from two or more trials were familiar were eliminated sincenovelty of toys was important in order to ensure that infants were drawinginferences from the first toy to the second toy and not drawing on previousexperiences with similar toys.2A measure of the coder's reliability was not made because it would not bepossible to derive a stable reliability estimate. There were two reasons for this:a) the infrequency of actual instances of transfer, and b) the liberal criterion foractions considered as transfer.3Although the multivariate analysis did not yield a significant main effect of Age,the univariate analysis of frequency yielded a marginally significant effect forAge, .E(1, 46) = 3.03, p<.09. However, no significant interactions with Conditionwere observed.53REFERENCESBaldwin, D.A., Markman, M.E. & Melartin, R. (in press). Infants' Ability to DrawInferences about Nonobvious Object Properties: Evidence from ExploratoryPlay. Child Development.Benedict, H. (1979). Early lexical development: Comprehension andproduction. Journal of Child Language, 5, 183-200.Davidson, N.S. & Gelman, S.A. (1990). Inductions from novel categories: Therole of language and conceptual structure. Cognitive Development Gelman, S.A. & Coley, J.D. (1990). The importance of knowing a dodo is a bird:Categories and inferences in two-year-olds. Developmental Psychology,21, 796-804.Gelman, S.A. & Markman, E.M. (1987). Young children's inductions fromnatural kinds: The role of categories and appearances. Child Development,55, 1535-1540.Gelman, S.A. & Taylor, M. (1984). How two-year-old children interpret properand common names for unfamiliar objects. Child Development, 55, 1535-1540.Gleitman, L. (1990). The structural sources of verb meanings. Language Acquisition, 1, 3-55.Goldin-Meadow, S., Seligman, M.E.P. & Gelman, R. (1976). Language in thetwo year old. Cognition, 4, 189-202.Gopnik, A. & Meltzoff, A.N. (1986). Words, plans, things, and locations:Interactions between semantic and cognitive development in the one-wordstage. In S.A. Kuczaj, Ill & M.D. Barrett (Eds.), The Development of Word meaduct (pp. 199-223). NY: Springer-Verlag.54Gopnik, A. & Meltzoff, A.N. (1987). The development of categorization in thesecond year and its relation to other cognitive and linguistic developments.Child Development, 5B,, 1523-1531.Gopnik, A. & Meltzoff, A.N. (1992). Categorization and naming: Basic-levelsorting in eighteen-month-olds and its relation to language. Child Development, fia, 1091-1103.Hirsh-Pasek, K., Golinkoff, R., Fletcher, A., DeGaspe Beaubien, F. & Cauley, K.(1985, October). In the beginning: One word speakers comprehend word order. Paper presented at Boston Language Conference, Boston.Huttenlocher, J. (1974). The origins of language comprehension. In R.L. Solso(Ed.), Theories in Cognitive Psychology. New York: Erlbaum.Katz, N., Baker, E. & Macnamara, J. (1974). What's in a name? The case ofobject names. Cognitive Psychology, :12, 63-89.Mandler, J. (1988). How to build a baby: On the development of an accessiblerepresentational system. Cognitive Development,  a, 113-136.Markman, E.M. (1989). Categorization and Naming in Children: Problems ofInduction. Cambridge, MA: MIT Press.Markman, E.M. & Watchel, G.F. (1988). Children's use of mutual exclusivity toconstrain the meanings of words. Cognitive Psychology,  22, 121-157.Markow, D.B. & Waxman, S.R. (1992, May). The Influence of Labels on 12-Month-Olds' Object Category Formation. Poster presented at theInternational Conference on Infant Studies, Miami.Massey, C.M. & Gelman, R. (1988). Preschoolers' ability to decide whether aphotographed unfamiliar object can move itself. Developmental Psychology, 24, 307-317.Naigles, L. (1990). Children use syntax to learn verb meanings. Journal ofChild Language, 11 357-374.55Nelson, K. (1985). Making Sense: The Acquisition of Shared Meaning.Orlando, Florida: Academic Press, Inc.Nelson, K. (1977). The conceptual basis for naming. In J. Macnamara (Ed.),Language Learning and Thought (pp. 117-136). New York: AcademicPress.Nelson, K. (1985). Making Sense: The Acquisition of Shared Meaning.Orlando, Florida: Academic Press, Inc.Nelson, K. (1988). Constraints on word learning? Cognitive Development,  a,221-246.Nelson, K. (1991). Concepts and Meaning in Language Development. In N.A.Krasnegor, D.M. Rumbaugh, R.L. Schiefelbusch & M. Studdert-Kennedy(Eds.), Biological and Behavioral Determinants of Language Development(1991). Hillsdale, NJ: Lawrence Erlbaum.Oviatt, S.L. (1980). The emerging ability to comprehend language: Anexperimental approach. Child Development,  51, 97-106.Oviatt, S.L. (1985). Tracing developmental change in language comprehensionbefore twelve months of age. Papers and Reports on Child LanguageDevelopment. No. 24. Stanford, CA: Stanford University.Waxman, S.R. & Balaban, M.T. (1992, May). The Influence of Words vs. Toneson 9-Month-Old Infants' Object Categorization. Poster presented at theInternational Conference on Infant Studies, Miami.Wellman, H.M. & Gelman, S.A. (1988). Children's understanding of thenonobvious. In R.J. Sternberg (Ed.), Advances in the Psychology of Human Intelligence, Vol. 4 (pp. 99-135). Hillsdale, NJ: Erlbaum.Woodward, A.L. (1992). The Role of the Whole Object Assumption in EarlyWord Learning. Unpublished doctoral dissertation, Standford University.

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0086094/manifest

Comment

Related Items