Poster group
Details of individual items:
poster
Object identity is the ability to determine whether an object is the same object as one encountered previously. In a study of 10- and 12-month-old infants, Xu and Carey (1996) found that infants do not attain a complete concept of object identity until 12 months of age. The researchers interpreted their findings to mean that 10-month-olds are not able to use information about an object's features to individuate objects. However, using different stimuli and different procedures than those used by Xu and Carey, Wilcox and Baillargeon (1998a; 1998b) have obtained positive results with much younger infants. Their earlier studies (Wilcox & Baillargeon, 1998a) suggest that 7.5-month-olds use featural information to individuate objects, while more recent findings (Wilcox & Baillargeon, 1998b) demonstrate that 4.5-month-olds also are capable of using featural information. The purpose of the present study was to investigate which spatial and physical object characteristics 9-month-old infants would use for the purposes of object identity. The specific object properties of interest were size, location and features.0D09Seventy infants, ranging in age from 8.5 months to 9.5 months, participated in this experiment. The task was a two-location violation of expectancy procedure with duration of the 'first look' as the dependent measure. There were four target stimuli: a small troll (ST), a large troll (LT), a small bear (SB) and a large bear (LB). The troll stimuli were identical in every respect except for size, as were the bear stimuli. Each trial began with the ST at Location A. Next, two flaps covered the two locations. Following this delay, the flaps were lowered to reveal either a Standard Event (the ST still at Location A) or a Change Event (the object's size, location, features--alone or in combination--had changed), and the duration of the infant's 'first look' was measured. Outcomes alternated between Standard and Change Events, with only one type of Change Event seen by each infant depending to which of the seven conditions that infant was randomly assigned.0D09A 'first look' (FL) was defined as the amount of time an infant looks at the display initially before looking away. A repeated measures ANOVA was performed on the FL data, yielding a significant Condition x Trial interaction. T-tests were performed for each condition to compare the mean duration of FLs to the Standard and Change Events. The t-tests indicated that infants looked significantly longer at the Change Event as compared with the Standard Event in the following conditions: Size Condition, Size & Feature Condition, and the Size &Location Condition.0D09The results are discussed vis a vis two default expectations which may guide infants' reasoning about object identity, the Same Location/Same Object Expectation and the Different Location/Different Object Expectation. Furthermore, the reasons for choosing an FL measure versus a cumulative looking time measure are explained.0D
poster
Aguiar and Baillargeon (1999, in press) recently examined 2.5- and3-month-olds' ability to predict whether an object should remain occludedwhen passing behind a screen with an opening. They concluded that, at 2.5months of age, infants possess only a primitive concept of occlusion andexpect an object to be hidden when behind an occluder and to be visibleotherwise. By 3 months, infants would already have progressed beyond thisinitial concept and would now expect objects to become temporarily visiblewhen passing behind an occluder with an opening extending from its lowerthough not its upper edge.The account of Aguiar and Baillargeon (1999, in press) assumes that infantsacquire contrastive condition-outcome rules: they expect a certain outcomewhen a condition is met, and a different outcome when that same condition isnot met (e.g., at 2.5 months: 'when behind, hidden; when not behind,visible'; at 3 months: 'when low opening, visible; when no low opening,hidden'). However, there is a weaker interpretation for their results.According to this alternative account, infants would acquire only singlecondition-outcome rules: they would expect a certain outcome when acondition is met, but would have no clear expectation as to what outcomeshould occur when the condition is not met (e.g., at 2.5 months: 'whenbehind, hidden; when not behind, ?'; at 3 months: 'when low opening,visible; when no low opening, ?').The present research tested between these two interpretations. Infants aged2.5 and 3 months were first familiarized with a tall cylinder moving backand forth behind a tall screen. Next, a portion of the screen was removed,and the infants saw two test events in which the cylinder consistentlyappeared in the opening created. In one of the test events shown to the2.5-month-olds, the entire midsection of the screen was removed to form twoseparate screens (two-screens event); in the other event, the two screensremained connected at the top by a narrow strip (connected-at-top event).The 3.5-month-olds saw the same connected-at-top test event, and anotherevent in which the screens were connected at the bottom by an identicalnarrow strip (connected-at-bottom event).The 2.5-month-olds looked reliably longer at the connected-at-top than atthe two-screens test event. These and control results suggest that theinfants expected the cylinder to appear between the two separate but not thetwo connected screens. The 3-month-olds looked reliably longer at theconnected-at-bottom than at the connected-at-top test event. These andcontrol data suggest that these older infants expected the cylinder toappear between the screens that were connected at the top but not at thebottom.These results provide strong support for the notion that infants acquirecontrastive condition-outcome rules when learning about occlusion (andpresumably other physical) events. In addition, the results make clear thatinfants formulate rules about the events around them and respond withprolonged looking (or 'surprise') to violations of these rules, whether ornot these violations actually involve physically impossible events. In thepresent research, 2.5-month-old infants were puzzled to see the cylinderappear between two screens that were connected at the top, and 3-month-oldinfants were surprised to see the cylinder appear between two screens thatwere connected at the bottom. In both cases, infants responded withprolonged looking to perfectly possible and indeed commonplace events.These data help address recent criticisms that infants inviolation-of-expectation tasks simply detect the oddness or novelty ofphysically impossible events never previously seen (e.g., Haith & Benson,1997).
poster
An interesting discrepancy in infants' ability to reason about occlusion events has recently beenrevealed. Whereas infants' as young as 4.5 months of age demonstrate the ability to use featuralinformation to individuate objects in occlusion events, they have difficulty judging whether thoseobjects map onto a subsequent no-occlusion display (Leslie et al., 1998; Wilcox, in press; Wilcox &Baillargeon, 1998a, 1998b; Wilcox & Schweinle, 1999; Xu & Carey, 1996). For example, in oneevent-mapping experiment (Wilcox & Baillargeon, 1998a), 9.5-month-olds were assigned to one oftwo conditions: box-ball or ball-ball. Infants saw a two-phase test event. In the initial phase of thebox-ball test event, a box emerged from behind the left edge of a wide screen and then returned. Next,a ball emerged from behind the right edge of the screen and returned. The screen was then lowered. In the final phase of the test event, infants saw the ball sitting behind the lowered screen. Infants in theball-ball condition saw the same event, except that a ball was seen to the left of the screen. The infantsin the two conditions looked equally at the one ball, as if the box-ball infants had failed to detect adiscrepancy between their representation of the initial box-ball event and the final one-ball display. It has been suggested that this is a particularly difficult task because it requires infants to (a)retrieve their representation of the initial box-ball or ball-ball event, (b) compare it to the one ballbefore them and (c) judge whether the two are consistent (Wilcox & Baillargeon, 1998a). Thisexplanation makes several assumptions about infants' reasoning processes. One of these assumptionsis that, instead of viewing the test event as one ongoing situation, and monitoring changes in the eventas it unfolds, infants view the test event as composed of two distinct parts. What led the infants to viewthe initial and final phases as separate events? There is evidence that infants organize the physicalevents that they observe into broad categories, such as occlusion, containment, or support (Baillargeon,1995). Hence, infants may have categorized the initial box-ball or ball-ball event as one of occlusion,and reasoned about the event based on their knowledge about occlusion phenomena. However, whenthe screen was removed, infants were prompted to view the next event as one of no-occlusion. There-categorization process led infants to initiate a new event representation. It was in their attempt toretrieve their representation of the box-ball event, to compare to the one-ball display, that they faltered. One prediction that this hypothesis makes is that if the initial and final phases of the test eventwere of the same event category, infants would view them as one continuous event, eliminating theneed to retrieve and compare event representations. The present research tested this hypothesis. 9.5-month-olds were tested with the box-ball and ball-ball events described above, with one exception:when the screen was lowered infants saw a second transparent screen covered with small dots standingin front of the ball. Thus, the situation was still one of occlusion (i.e., the ball sitting behind a screen),even though the ball was visible. If the presence of the second transparent screen led infants to regardthis as one continuous event rather than two distinct events, eliminating the need to retrieve andcompare events, infants should find the box-ball event surprising. The results confirmed thisprediction: the infants in the box-ball condition looked reliably longer at the one-ball display. Preliminary results of a second experiment suggest that if the transparent screen is replaced by atransparent container, negative results are obtained, as if the infants again viewed the initial and finalphases as instances of different event categories (see Hespos & Baillargeon, 1998). These results haveimportant implications for concept formation and event representation in infancy.
poster
There is evidence that young infants are sensitive to a discontinuity of speed and use thisinformation to draw conclusions about the number of objects involved in a n occlusion event(Putthoff & Wilcox, 1997; Wilcox & Schweinle, 1999). In one experiment, 4.5- and 7.5-month-olds were assigned to one of two conditions: experimental or control. Each infant was tested witha two-phase procedure. In the initial phase (familiarization), infants in the experimental conditionsaw an object sitting to the left of a lowered screen; the area behind the screen was empty. Next,the screen was raised and the object moved back and forth behind the screen until the end of thetrial. In the second phase (test), infants saw a similar event, except that the object reappearedimmediately to each side of the screen. Infants in the control condition saw the samefamiliarization and test events, except that a second screen stood behind the lowered screen at thebeginning of each trial, occluding the center portion of the platform. The infants in theexperimental condition looked reliably longer at the test event than infants in the controlcondition, as if they (a) believed that only one object was present in the apparatus, (b) realizedthat two objects were required to produce the immediate reappearance test event, and hence (c)responded with surprise during the test trials. Although these and other recent results (Wilcox, in press; Wilcox & Baillargeon, 1998a,1998b; Putthoff & Wilcox, 1999) suggest sophisticated reasoning about occlusion events veryearly in infancy, there is also evidence that reasoning about occlusion events is not as easy forinfants as it first appears (Wilcox, 1997). The present research investigated the development ofinfants' ability to correctly interpret a discontinuity of speed within the context of an occlusionevent. In Experiment 1, two groups of 11.5-month-olds were tested. Infants in the first groupwere tested in a procedure identical to that described above. Infants in the second group weretested with a similar procedure with one exception: they saw only the test trials (they were notgiven familiarization trials). The infants who viewed familiarization trials prior to the test trialslooked reliably longer at the experimental than control test event. In contrast, the infants who sawonly the test trials looked about equally at the two events. These results suggest that in order forthe 11.5-month-olds to correctly interpret the event, they needed prior exposure to the occlusionsituation. In Experiment 2, 13.5-month-olds were tested with a similar procedure with oneexception: they saw only the test trials (they were not given familiarization trials). In contrast tothe 11.5-month-olds, they successfully interpreted the discontinuity of speed event. Why did the 11.5-month-olds fail to demonstrate the ability to use discontinuity of speedto individuate objects when they were not first exposed to the occlusion situation? We speculatedthat the infants used the familiarization trials to categorize and represent the vent as one ofocclusion. It was not until after this was accomplished that infants were able to draw on speedinformation to correctly interpret the occlusion event see in the test trials. Research is currentlybeing conducted to test this interpretation.
poster
Previous research has shown that young infants are capable ofindividuating objects using spatiotemporal cues at a very young age(Kellman & Spelke, 1983; Spelke, 1990). However, the age at which theybegin to use featural information, such as color and texture, is lessclear. While some researchers have found this ability present as young as4.5 months of age (Needham & Baillargeon, 1997), it appears to beextremely fragile, and other research has suggested it does not ariseuntil 10-12 months of age (Xu & Carey, 1996). This developmentaldiscontinuity may reflect an increase in infants capacity for processingcomplex visual stimuli, or in the processing resources required tosegregate objects using featural properties. Alternately, thedevelopmental change might reflect the acquisition of symbolic resourcesunique to humans -- the capacity to represent object kinds, or theacquisition of linguistic labels for those kinds (Xu, Carey & Welch,1999). Comparative research can help clarify the role of language in thisdebate. If nonlinguistic primates can individuate objects using featuralinformation, and show the same developmental discontinuity in theirability to do so as humans, then the developmental trajectory witnessed inhuman infants might not be driven by the emergence of language. A recentstudy has answered part of this question, showing that adult rhesusmacaques are capable of segregating objects using their featuralproperties (Munakata et. al, 1999). The present research examines thedevelopment of this ability in monkeys by testing at several points inearly life: can infant pigtail macaques individuate unfamiliar objectsusing only property information at 4, 8, 12, and 36 weeks of age. This study used a violation of expectation looking-time procedureadapted from Munakata et. al (1999) in which subjects were presented withone unfamiliar object stacked on top of another. A hand then lifted thetop object, presenting the subjects with one of two outcomes: either thebottom object either remained stationary (possible), or lifted togetherwith the grasped object (impossible). In a second condition, subjectsviewed a single unfamiliar object, which, in a similar fashion, eithersplit in half or remained whole when lifted. Preliminary results show thatolder monkeys (average age 36 weeks, n3) all looked longer at theimpossible outcome in both one- and two-object conditions, consistent withpast findings from rhesus macaques. However, young monkeys (aged 5 weeks,n5) showed no consistent preference for either outcome. One possible explanation for this difference is that thislooking-time procedure might not be effective with very young macaques, inwhich case their failure would be method-related rather than task-related.To test this possibility we performed a second looking-time experiment, anadaptation of Wynns (1992) numerical transformation task, which is knownto produce robust results even in very young human infants. Subjects saw asingle object placed on a stage, then occluded by a screen, and a secondidentical object placed behind the screen. The screen was then removed toreveal either the possible outcome of two objects, or the impossibleoutcome of one object. All five 5-week-old macaque subjects looked longerat the impossible event than the possible. While clearly preliminary, these experimental results indicate thatinfant pigtail macaques, like human infants, show a developmentaldiscontinuity in their ability to individuate objects using featural cues.This suggests that the developmental shift in human infants is notdependent on emerging cognitive resources unique to our species; rather,the transition may be the result of maturational changes that are sharedby other species of primates. Additionally, this research suggests thatinfant pigtail macaques are capable of tracking the numerosity of an arrayof objects across simple transformations. This validates the effectivenessof the looking time procedure on this population, and suggests that infantmacaques have access to numerical representations comparable to those ofhuman infants tested using this same procedure. Data gathered fromadditional infant macaque populations over the next four months will fillin data for intermediate ages, and should provide more statisticallyrobust conclusions.
poster
There is now a body of research to suggest that young infantsestablish representations of individual objects utilizing spatiotemporalcriteria (e.g., that one object cannot be at two places at the same time)and property/kind criteria (e.g., that a red object is not the same as agreen object) (Spelke et al., 1995; Xu & Carey, 1996; Leslie et al., 1999).Utilizing the looking time technique, this research has shown that10-month-old infants individuate objects primarily on the basis ofspatiotemporal information, whereas by 12 months of age both property andkind information is used to individuate objects. For example, when infantssee a toy duck and a toy truck emerging from behind a screen, one at a time,and then disappearing behind the screen, they do not expect to see twoobjects when the screen is removed until they are 12 months of age. Whenyounger infants see the two toys together at the same time on stage beforebeing placed behind the screen, they will expect to see two objects uponremoval of the screen but not if they have never seen the two objectstogether at the same time. This result led Xu and Carey to conclude that theyounger infant might have a general object kind concept 'physical object'before they develop more specific kind concepts. Because the results oflooking time methods are notoriously difficult to interpret, a new method toaddress the question of the criteria infants use for object individuationwas developed (Van de Walle, Carey & Prevor, submitted, Uller, Leslie &Carey, 1998). In the present experiment, we seek to confirm Xu and Carey's(1996) findings using reaching as the dependent measure. In this study, 6410- and 12-month-old infants see only one object at a time being placed intoa box they can only reach into, they cannot see inside. This experimentstarts with a series of familiarization trials. Infants are familiarizedwith different pairs of toys, sometimes there are two, sometimes there isonly one. They are given spatiotemporal information of the existence of theobjects. They are allowed to reach into the box, retrieve the object/s andmanipulate them for 10 seconds at a time. After this familiarization phase,the experimenter then shows the infants a novel toy, allows the infants toplay with it, and then places it inside the box. The infants are thenallowed to reach for the toy they had previously seen. In the possibletrials, the infants reach into the box and retrieve the toy they saw theexperimenter place into the box. In the impossible trials, the infants reachin and retrieve a different toy from the original one they saw being placedinto the box. If the infants identify the object they saw being placed intothe box in terms of its properties, then they should continue to search forit when they reach in and retrieve something else. The dependent measuresare the number of times the infants reach into the box to retrieve theobjects and the persistence with which the infants search for the objectthey had previously seen. Results reveal that 12-month-old infants will reach more oftenand more persistently on the impossible trials than on the possible trials,while 10-month-old infants will not. These results provide further supportto Xu and Carey (1996) and previous results by Prevor et al. (submitted) andUller et al. (1998).
poster
Xu & Carey (1996) report that infants younger than about 10months do not take property information into account whendeciding how many objects there are in a scene. Instead, youngerinfants are said to rely solely upon spatiotemporal continuity forindividuating objects. However, Wilcox & Baillargeon (1998)challenge this conclusion by showing that there are circumstancesin which younger babies use property differences (e.g.., shape) toset up representations of distinct objects. In particular, Wilcox andBaillargeon report a simplified version of the Xu & Carey task. Intheir Expected condition, 9-month-old babies were shown an eventin which a ball moves behind a screen and then emerges on theother side of the screen in a single pass. The screen is thenremoved to reveal an empty stage. In their Unexpected condition,the ball moves behind the screen as before but what emerges is abox. Again the screen is removed to reveal an empty stage. Wilcoxand Baillargeon found that their infants looked longer in theirUnexpected condition and argued that this was because the infantshad assumed when the box emerged that it must be a differentobject and that therefore the ball must still be behind the screen.Infants were then surprised to see the empty stage. Leslie, Xu, Tremoulet & Scholl (1998) draw a distinctionbetween information that is used to establish an objectrepresentation, that is, to individuate objects, and information thatfinds its way into or is stored with the object representation. Ifproperty information finds its way into the object representation,then it will be available later for deciding whether an object is onethat has been encountered before. Such information can be used to(re)establish an object's identity. Our experiment has three goals. First, we wish to replicatethe effect reported by Wilcox & Baillargeon. Second, we want tocheck whether infants in Wilcox & Baillargeon's simplified taskreally are individuating the objects by their shape. To this end, wehave added a condition in which infants see a ball moving behindthe screen followed by a box emerging. The screen is then removedto show the expected ball. Third, we want to see if infants havestored identification information, that is, whether they expectspecifically a ball to be behind the screen or simply another object.We have thus added a fourth condition in which after the ball hasgone behind the screen and the box emerged, the screen is removedto show another box (unexpected). Data collection is on-going but we expect that our 9-month-olds will fail to identify the objects involved.ReferencesLeslie, A.M., Xu, F., Tremoulet, P., & Scholl, B. (1998). Indexing and the object concept: Developing 'what' and 'where' systems. Trends in Cognitive Sciences, 2, 10 18.Wilcox, T., & Baillargeon, R. (1998). Object individuation in infancy: The use of featural information in reasoning about occlusion events. Cognitive Psychology, 37, 97 155.Xu, F., & Carey, S. (1996). Infants' metaphysics: The case of num- erical identity. Cognitive Psychology, 30, 111 153.
poster
Previous research suggests that 4.5-month-old infants can reason about widthin occlusion (e.g., Wilcox, in press) but not containment (e.g., Sitskoorn &Smitsman, 1995) events. Because these two findings were obtained usingdifferent tasks, it is difficult to determine whether this discrepancyreflects (a) the use of different tasks or (b) a genuine lag in infants'learning about width in occlusion and containment events. The presentexperiment compared 3.5-month-old infants' reasoning about width inocclusion and containment events under highly similar conditions.A second purpose of the present experiment was to examine whether younginfants could succeed at reasoning about hidden objects in a task involvingonly test trials. Previous reports of object permanence in young infantshave all involved habituation or familiarization trials; these reports havebeen criticized on the grounds that infants might have developed duringhabituation or familiarization superficial expectations that then guidedtheir responses during test (e.g., Thelen & Smith, 1994). We reasoned thatevidence that infants can reason about hidden objects in a task involvingonly test trials should help counter this criticism.The infants were assigned to a containment or an occlusion condition. Theinfants in the containment condition saw two test events. At the start ofboth events, a hand held a wide, thin rectangular object centered above awooden box (to facilitate width comparisons). After a pause, a screen wasraised to hide the box. The hand then lowered the object into the box,behind the screen. Finally, the screen was lowered to reveal the box (thebox was sufficiently tall that the infants could not determine, from theirviewpoint, whether or not the object rested inside the box). Two differentboxes were used in the two test events. Both boxes were identical in heightand depth, but differed in width: one box was wider (wide event) and one boxmuch narrower (narrow event) than the object. The infants in the occlusioncondition saw the same test events, with one exception: the bottom, back,and sides of each box were removed, leaving only the front of the box toserve as an occluder.The infants in the occlusion condition looked reliably longer at the narrowthan at the wide test event, suggesting that they realized that the objectcould be hidden behind the wide but not the narrow occluder. Thisconclusion was supported by a control condition in which a narrow object wasused that could be hidden behind either occluder; these control infantslooked about equally at the two test events.In contrast to the infants in the occlusion condition, those in thecontainment condition tended to look equally at the narrow and the wide testevents. These results confirm previous reports (e.g., Sitskoorn & Smitsman,1995) that young infants do not realize that wide objects can be hidden intowide but not narrow containers.The present results are interesting for two reasons. First, they provideadditional evidence that infants view occlusion and containment as distinctevent categories and reason and learn separately about each category (e.g.,Hespos & Baillargeon, 1999). Second, the results of the occlusion conditionindicate that 3.5-month-old infants succeed at reasoning about hiddenobjects even in tasks that involve no habituation or familiarization trials.
poster
Prior research indicates that infants aged 12.5 months and older consider the 'proportional distribution' of an asymmetrical box when judging its support: infants expect an L-shaped box to remain stable when its larger end (adequate-support display) but not its smaller end (inadequate-support display) rests on a platform. Younger infants expect any box--whether symmetrical or asymmetrical--to be stable as long as 50% or more of its bottom surface is supported.In recent experiments, infants aged 10.5 to 12 months were taught to attend to the variable 'proportional distribution' when judging the stability of an asymmetrical box. In one experiment, infants received three pairs of teaching trials before being shown the adequate- and inadequate-support test displays. In each teaching pair, the infants saw an asymmetrical box being deposited on a platform; the overlap between the box's bottom surface and the platform was always 50%, as in the test displays. In one trial, the smaller end of the box was placed on the platform and the box fell when released (box-falls event). In the other trial, the larger end of the box was placed on the platform and the box remained stable when released (box-stays event). The three pairs of teaching trials were identical except that different asymmetrical boxes were used: a box shaped like an asymmetrical 'B' on its side, a right triangle, and a staircase-shaped box. After watching the three pairs of teaching trials, the infants looked reliably longer at the inadequate- than at the adequate-support test display. These and control results suggested that infants aged 10.5 to 12 months can acquire the variable 'proportional distribution' when exposed to observations designed to highlight this variable.The present research built on these initial results. We began to systematically vary the content of the teaching observations, to determine more precisely what kind of observations are needed for learning. In the experiment described above, the infants saw the same box, with 50% of its surface resting on the platform, in both trials of each teaching pair. We speculated that having the two trials be similar in all respects except for the target variable--what proportion of the box lay on or off the platform--considerably facilitated infants' identification of the target variable. To test this hypothesis, two experiments were conducted that were identical to the experiment described above, with one exception: the two trials in each teaching pair now differed in one respect in addition to the target variable. In Experiment 1, the box used in the box-stays trial of each teaching pair differed in color from the box used in the box-falls trial. In Experiment 2, 75%, rather than 50%, of the box's bottom surface rested on the platform in the box-stays trial of each teaching pair.The results of both experiments were negative: the infants tended to look equally at the inadequate- and the adequate-support test displays. These results suggest that infants aged 10.5 to 12 months have difficulty identifying the variable 'proportional distribution' (or at least have difficulty identifying it within three teaching pairs) if each pair involves two as opposed to one noticeable difference. Presumably, infants become distracted or confused by the second difference, making it more difficult for them to zero in on the correct variable.These results (and the results of ongoing experiments focusing on changes in less central aspects of the teaching trials, such as the color of the platform) are helping to shed light on the processes involved in infants' identification of variables, and hence on the learning mechanisms that make possible their rapid acquisition of physical knowledge.