Cognition 157 (2016) 205–218
Contents lists available at ScienceDirect
Cognition journal homepage: www.elsevier.com/locate/COGNIT
Original Articles
Contrasting preschoolers’ verbal reasoning in an object-individuation task with young infants’ preverbal feats Horst Krist ⇑, Karoline Karl, Markus Krüger Department of Psychology, University of Greifswald, Greifswald, Germany
a r t i c l e
i n f o
Article history: Received 24 July 2013 Revised 30 August 2016 Accepted 12 September 2016
Keywords: Cognitive development Continuity vs. discontinuity Object individuation Physical reasoning Eye tracking
a b s t r a c t Young infants infer a second object if shown an object apparently moving on a discontinuous path (Aguiar & Baillargeon, 2002; Spelke, Kestenbaum, Simons, & Wein, 1995). In three experiments, we examined whether children aged 3–6 years and adults would do the same in their verbal explanations of an apparent continuity violation. Presenting participants with video clips (Exp. 1 and 3) as well as live events (Exp. 2) of a toy locomotive apparently passing through a tunnel without appearing in a large opening in the middle, we found virtually no evidence for generations of two-object explanations of the critical test event in preschoolers. Some of the younger children even denied a continuity violation at first. When participants were familiarized to two identical objects instead of just one, they were more likely to realize that a second object was involved in the test events but, unlike adults (Exp. 3), most children nonetheless adhered to a one-object interpretation. Analyzing 3- and 5-year-old children’s and adults’ eye movements (Exp. 3), we found that children’s difficulties to infer a second object from an apparent continuity violation were not caused by inappropriate looking strategies. We conclude that preschoolers’ physical reasoning about the numerical identity of objects is not continuous with the preverbal reasoning of infants. Rather than being exclusively constrained by the output of basic object-individuation processes, as in infants, it is also strongly influenced, in a top-down manner, by prior beliefs. Ó 2016 Elsevier B.V. All rights reserved.
1. Introduction 1.1. The early-competence debate Recently, the debate on the status of young infants’ cognitive capabilities has been revived by research suggesting early forms of false-belief understanding. The discrepancy between the wellestablished fact that preschoolers do not pass traditional Theoryof-Mind tasks (see Wellman, Cross, & Watson, 2001, for a metaanalysis) and recent findings suggesting much earlier competencies in nonverbal variants of these tasks (e.g., Onishi & Baillargeon, 2005; Southgate, Senju, & Csibra, 2007) has renewed researchers’ interest in the puzzle of why older children often appear so ‘‘dumb” and infants so ‘‘smart” (Keen, 2003). Traditionally, the early-competence debate was largely restricted to the domain of intuitive physics. It revolved around criticisms of Piaget’s assumption of a sensori-motor stage. Using the so-called Violation-of-Expectation (VoE) method, several ⇑ Corresponding author at: EMAU Greifswald, Institut für Psychologie, Entwicklungspsychologie und Pädagogische Psychologie, Franz-Mehring-Str. 47, D-17487 Greifswald, Germany. E-mail address:
[email protected] (H. Krist). http://dx.doi.org/10.1016/j.cognition.2016.09.008 0010-0277/Ó 2016 Elsevier B.V. All rights reserved.
research groups were able to demonstrate early forms of object permanence and physical reasoning. Since then, the picture of infants’ physical reasoning has been enriched considerably using this method (for an overview and theoretical account, see Baillargeon, Li, Ng, & Yuan, 2009). The VoE method is a derivative of the preferential-looking paradigm used in innumerable studies on early perceptual development. Usually, albeit not necessarily, after a period of familiarization or habituation, infants are presented with two stimuli, either simultaneously or sequentially, and their preferential attention to one of the stimuli is assessed, mostly operationalized as differential looking times. The rationale behind the VoE method is that infants, much like older children and adults, react with increased attention (usually measured as longer looking times with infants) if they observe something unexpected. In combination with results from additional conditions or experiments controlling for factors such as perceptual novelty, preferential attention (looking) for an impossible event is interpreted as evidence suggesting that the infant was ‘‘surprised” (i.e., that he or she experienced a violation of expectation) and hence as an indicator of intuitive knowledge. The interpretation of demonstrations of early competencies by means of the VoE method was heavily disputed in the 1990s and
206
H. Krist et al. / Cognition 157 (2016) 205–218
early 2000s (e.g., Bogartz, Shinskey, & Speaker, 1997; Thelen & Smith, 1994; cf. Baillargeon, 1999, 2004, with commentaries). Today, there appears to be wider agreement, however, that--on a ‘‘middle” level, at least--infants do represent occluded objects and perform cognitive operations on these representations (e.g., Carey, 2009; Scholl & Leslie, 1999). The traditional assumption that infants are purely sensori-motor beings without any internal representations and cognitive structure (Piaget, 1954) has become hardly tenable. What is still at stake, however, is the exact nature of infants’ physical reasoning as assessed indirectly, mostly via the VoE method, and how it relates (ontogenetically) to explicit forms of physical reasoning diagnosable with older children (e.g., Aschersleben, Henning, & Daum, 2013; Krist, 2010, 2013). 1.2. Young infants’ preverbal feat: positing the existence of occluded objects The present research is designed to shed new light on this issue by using what might be viewed as the most impressive of all early competencies as a benchmark for preschoolers’ explicit physical reasoning: namely, young infants’ ability to ‘‘infer” or ‘‘posit” the existence of occluded objects. This presumed competence is remarkable, indeed. It is closely related to the ability to posit unobserved causes, such as hidden forces or ‘‘essences”, and hence to the hallmark of theoretical thinking, which some authors consider to be uniquely human (e.g., Povinelli, 2000, 2012). But can young infants really accomplish such feats, albeit on an implicit, preverbal level? What is the evidence? Spelke and colleagues were the first to publish such evidence (Spelke & Kestenbaum, 1986; Spelke, Kestenbaum, Simons, & Wein, 1995). In a series of experiments, they habituated 4-month-old infants either to a continuous or to a discontinuous event. In the continuous event, a vertical rod was shown moving back and forth horizontally across a puppet stage disappearing and reappearing behind two vertical screens in turn. This event was termed ‘‘continuous” because it did not violate the continuity principle according to which objects move on uninterrupted paths (Carey & Spelke, 1994). In the discontinuous event, this principle was apparently violated because the rod did not appear between the two screens. Actually, a second rod was used to produce this apparent continuity violation: While one of the two identical rods approached the first screen, the other one was hidden behind the second screen; after the former had disappeared behind the first screen, the latter appeared from the second screen in the same manner and at the same delay as the single rod in the continuous event. After the infants had reached the habituation criterion or the maximum number of habituation trials had been presented, three pairs of test trials without any screens were shown in which either one or two rods were shown moving as in the continuous and the discontinuous event, respectively. When presented with the continuous event, infants dishabituated more strongly (i.e., looked longer) with the two-object than with the one-object event, and vice versa. In other words, infants tended to generalize the continuous habituation event to the one-object test event and the discontinuous habituation event to the two-object test event. This result suggests that infants made sense of the apparent continuity violation in the discontinuous event by ‘‘perceiving” a second object (Spelke, 1990). Yet, in comparison with baseline and control conditions, Spelke et al. (1995) obtained only mixed support for this assumption (cf. Aguiar & Baillargeon, 2002, Footnote 4). One could also argue that the 4-month-olds tested by Spelke et al. (1995) were only realizing by hindsight, when watching the two-object test event, that two objects must have been involved in the discontinuous habituation event (see Aguiar & Baillargeon, 2002; Baillargeon, 1994). More direct, but still preliminary, evidence supporting the claim that young infants are able to infer
the existence of an occluded object if confronted with an apparent continuity violation was reported by Baillargeon (1994). In her study, 5.5-month-olds were familiarized to a toy rabbit disappearing and reappearing behind a large screen and were then tested with a high- versus a low-window event. In the high-window event, the midsection of the screen’s upper half was removed, while, in the low-window event, the corresponding section of the screen’s lower half was cut out. Again, the rabbit disappeared and reappeared (from) behind the screen, but failed to be seen in the windows. While it was short enough to remain completely hidden by the screen in the high-window event, it should, have appeared in the low-window event, of course. Still, infants tended to look equally at both events. Interpreting this negative finding as preliminary evidence that 5.5-month-olds posited the involvement of a second, identical object (initially occluded by the screen) to explain the apparent discontinuity in the low-window event, Aguiar and Baillargeon (2002) sought for additional evidence to support their claim. Against the backdrop of the results obtained in Spelke’s lab (Spelke & Kestenbaum, 1986; Spelke et al., 1995) as well as those from their own lab (Aguiar & Baillargeon, 1999; Baillargeon, 1994; Baillargeon & DeVos, 1991), Aguiar and Baillargeon (2002) speculated that, with the low/high-window paradigm, infants should begin to posit an additional occluded object at some age between 2.5 and 4 months, but only after they are able to detect this particular continuity violation. In other words, infants should first exhibit differential looking times, indicating success in detecting the continuity violation, and then fail to do so, indicating success in explaining it by (correctly) inferring the involvement of a second object. This is exactly what they found: Three-month-old infants looked reliably longer at the low- than at the highwindow event, whereas 3.5-month-olds did not. As Aguiar and Baillargeon’s (2002) experiments constitute the reference point for the present research with preschoolers, they will be described in more detail next. In their main experiment (Exp. 1), 3- and 3.5-month-olds were presented with events resembling those of the rabbit study mentioned above (Baillargeon, 1994; see also Aguiar & Baillargeon, 1999). The infants were habituated to a toy mouse (‘‘Minnie Mouse”) moving back and forth along a track disappearing and reappearing behind a large rectangular screen. Following habituation, a high- and a low-window event were presented in two pairs of alternating trials. As in the rabbit study, the toy mouse did not appear in either case although it should have done so in the low-window event. All events were produced by using two identical toy mice one of which remained hidden behind the right edge of the screen until the other one, approaching from the left, had disappeared behind the left edge. After an appropriate delay, the former appeared from behind the right edge of the screen continuing the movement of the latter, before reversing its direction and repeating the movement sequence from right to left. Each trial ended as soon as the infant looked away (for 2 s), after having looked at the event for a specified duration, or looked at the event for the maximum time allowed (90 s). The habituation phase ended if the habituation criterion was met (50% decrease in mean looking time, relative to the first 3 trials) or 9 habituation trials were completed. As already mentioned, Aguiar and Baillargeon (2002) obtained longer looking times for the low- than the high-window event with the 3-month-old but not the 3.5-month-old infants. The negative finding with older infants was replicated in another experiment (Exp. 1A). The results from six follow-up experiments lent further support to Aguiar and Baillargeon’s claim that both younger and older infants were initially surprised that the toy mouse did not appear in the low window, but that only the older infants were able to explain the apparent continuity violation by positing a second object. In Experiment 2, it was shown that 3.5-month-olds do
H. Krist et al. / Cognition 157 (2016) 205–218
show prolonged looking at the low-window event if a two-object explanation is ruled out: Before each trial, the infants were shown that there was only a single mouse on the stage by lowering the screen at the start of each trial (the ‘‘magic trick” had therefore to be performed by surreptitiously inserting another mouse into the apparatus). In further experiments, Aguiar and Baillargeon (2002) found that 3.5-month-olds did not look reliably longer at the low- than at the high-window event when provided with a plausible hiding place for the second mouse but did show such preferential looking with an implausible hiding place. As a plausible hiding place, a narrow screen was revealed behind the right side of the larger screen at the start of each trial (Exp. 3), whereas a narrow screen with a window in its lower half was used as an implausible hiding place (Exp. 4). Furthermore, Aguiar and Baillargeon (2002) reported evidence suggesting that even 3-month-olds are able to make sense of the apparent continuity violation in the low-window event when shown two mice at the start of each trial (Exp. 6), but not when shown the narrow screen as a potential hiding place (Exp. 7). Only in the latter case, did 3-month-olds look reliably longer at the lowthan at the high-window event. Aguiar and Baillargeon (2002) therefore conclude: Unlike the older infants, the younger infants were thus unable to posit a second mouse to make sense of the low-window event; however, they could use the information that two mice were present to construct a satisfactory explanation for the low-window event. [Aguiar & Baillargeon, 2002, p. 321]
207
continuously or discontinuously along a track with two screens separated by a gap. In the continuous event, one of the bears disappeared behind the first screen, reappeared between the screens, and disappeared behind the second screen; in the discontinuous event, it did not appear in the gap. The bear reappearing, after an appropriate delay, from behind the second screen was either identical to the one disappearing behind the first screen or different from it. After each event, children were asked what they had seen; if necessary, they were further queried to assess their interpretation of the event. While most of the 3- and 4-year-old children and many 2-year-olds reported the involvement of two objects if different figures moved continuously or discontinuously, and all children interpreted the continuous movement of a single animal figure correctly (i.e., as involving only a single object), only one 2-year-old child (2.5%) and two 4-year-old children (5%) produced a two-object explanation for the discontinuous event involving the two identical bears. Remarkably, this interesting result has been largely ignored. One of the reasons for this might be the authors’ interpretation of their findings in orthodox Piagetian terms. They conclude that ‘‘it seems almost inconceivable that 4-month-olds understand—at more than the perceptual level—the continuity of movement” and that young infants only have ‘‘some single-scheme (vision) subconcepts of objects” (Melkman & Rabinovitch, 1998, p. 263). Such a one-sided, low-level interpretation of infant competencies does not appear to be consensual anymore (see Scholl & Leslie, 1999, for a critique of simplistic perception/cognition dichotomies). 1.4. The present research
1.3. Children’s and adults’ verbal responses to apparent continuity violations In light of these stunning findings, an obvious question is how adults would perceive and reason about similar events. Spelke et al. (1995) as well as Aguiar and Baillargeon (2002) therefore presented adults with the events used in their infant studies and asked what they perceived and how the events were accomplished, respectively. The 12 adults tested by Spelke et al. (1995, Appendix) reported seeing a single rod in the continuous event and two identical objects moving in succession in the discontinuous event. Aguiar and Baillargeon (2002, Exp. 1) presented 12 adult participants with the high- and 12 with the low-window event. After having watched one cycle of the high-window event, only one participant (correctly) assumed that two mice were involved; the other participants assumed that a single mouse was moved back and forth across the stage. With the low-window event, eight participants advanced the two-object solution, that is, they could describe how the ‘‘magic trick” was actually done. The remaining four participants came up with different explanations that did not involve a second object. Given that 3.5-month-olds are presumably able to infer the existence of a second object to explain the apparent continuity violation in the low-window event, one might wonder why, after all, 33% of the adults did not give a two-object explanation for this event. Of course, there are many possible reasons why these participants thought of other solutions, and they might have come up with the two-object solution if they had been asked for further ideas. But how about children beyond the infancy period? Melkman and Rabinovitch (1998) were the first to address this issue empirically. In a modification of Spelke and Kestenbaum’s (1986) paradigm, they presented children aged 2–4.5 years with cardboard figures that represented a bear family consisting of mama and papa bear as well as two identical little bears. The figures either moved
The present research was designed to shed more light on the hypothesis that young children’s verbal reasoning about the numerical identity of objects does not parallel infants’ nonverbal reasoning in object-individuation tasks. For this purpose, Aguiar and Baillargeon’s (2002) paradigm was adapted for children aged 3 years and older. Instead of toy mice, toy locomotives were used, but otherwise the events shown to the children were modeled as closely as possible on those presented to infants in the original studies. Toy locomotives were chosen because they represent vehicles, that is, self-propelled inanimate objects. This category appeared to be ideally suited for the present purpose because the existence of two identical objects is much more plausible with inanimate than with animate objects and because it is much more plausible that vehicles are able to start and stop (apparently) on their own than other inanimate objects. In three experiments, children aged 3–6 years were presented with a series of trials, either in a video (Exp. 1, Exp. 3) or live mode of presentation (Exp. 2). On three test trials, participants observed an apparent continuity violation that corresponded to variants of the low window event employed by Aguiar and Baillargeon (2002), first without and then, if necessary, with hints to the involvement of a second object. As in Aguiar and Baillargeon’s studies, hints were given by providing previews in the respective test trials. While the first hint was given by presenting a toy house as a potential hiding place for a second locomotive, a second locomotive standing behind the toy house was presented as the second and final hint. In the latter case, the inference of a second object was considered trivial because, during the preview phase, both locomotives were visible simultaneously. After the presentation of each test event, children were asked for possible explanations and were encouraged to come up with more ideas. In a familiarization trial, half of the children were presented with one toy locomotive (one-object condition) and the other half with two toy locomotives (two-object condition); in each case the locomotive(s) moved in a self-propelled manner.
208
H. Krist et al. / Cognition 157 (2016) 205–218
Our main prediction was that preschoolers would rarely come up with the idea that a second toy locomotive may have been involved doing the trick if they had to infer its existence in a non-trivial manner, that is, before the second locomotive was shown in the preview of the third test trial. We further expected the percentage of two-object solutions to be higher in the twoobject condition than in the one-object condition and to increase with children’s age. The condition effect was expected because the introduction of two identical objects should prime the inference of a second object in the test trials. 2. Experiment 1 In our first experiment on preschoolers’ ability to explicitly infer a second identical object if presented with an apparent continuity violation, all events were presented as video clips of a modelrailroad setting. In a first familiarization trial, children saw either a single locomotive (one-object condition) or two identical locomotives (two-object condition) moving at a constant slow speed, starting and stopping autonomously. A second familiarization trial was analogous to the habituation event used by Aguiar and Baillargeon (2002): A locomotive entered a tunnel and (apparently) reappeared on the other side (this event could also be interpreted as involving two identical locomotives, but there was no obvious reason to do so). In three test trials, the tunnel was shown with a cutout analogously to the low-window event employed by Aguiar and Baillargeon (2002). As the visible movement sequence remained the same as in the second familiarization trial, that is, no locomotive appeared in the cutout, it was not compatible with a one-object interpretation. While no clues to the involvement of a second locomotive were given in the first trial, a small house was revealed inside the tunnel as a potential hiding place for a second locomotive in a preview phase of the second test trial; only in the third and last test trial, the second locomotive was also presented in the preview (cf. Aguiar & Baillargeon, 2002, Exp. 3, 6, and 7). Children were interviewed after observing each event in a semistructured manner. The aim of the interview procedure was to ensure as much as possible that the child (1) had encoded the critical aspects of the respective event, (2) realized that there must be some kind of trick involved, and (3) tried to come up with plausible explanations. An experimental session was terminated prematurely if a child gave a two-object explanation of the test event, either spontaneously or after having been prompted for an(other) explanation by the experimenter; otherwise all three test trials were administered.
2.1. Method 2.1.1. Participants A total of 60 children aged 3–6 years participated in this experiment; there were 13 three-year-olds (M = 42.7 months, SD = 2.72), 15 four-year-olds (M = 52.9 months, SD = 3.66), 17 five-year-olds (M = 63.9 months, SD = 3.60), and 15 six-year-olds (M = 76.1 months, SD = 2.72). They were randomly assigned to one of the two experimental conditions. Nineteen additional children were tested but had to be excluded from data analysis due to experimenter error (15 children), repeated testing (1 child), or noncompliance (3 children). All children were recruited from kindergartens located in the urban area of Greifswald, Mecklenburg-Vorpommern, Germany. They participated on a voluntary basis and with the consent of their parents. No information was collected on parents’ education, occupation, or income. All children tested were native German-speakers or possessed adequate conversational skills in the German language. 2.1.2. Stimuli Two identical electrical toy locomotives (Brio 33222), a wooden railroad track (4 cm wide), a wooden board representing a house (just large enough to occlude the toy locomotive), and a tunnel made of cardboard (15 cm high, 45 cm wide, 15 cm deep), were used to produce video clips of the familiarization and test events (see Fig. 1). In the center portion of the tunnel, there was a cutout creating a low window (13 cm high, 15 cm wide) that could be open or closed. The events were recorded using a digital camera (Panasonic HDC-SD300) and edited with Adobe Premiere Pro CS4 to create the final video sequences (1920 1080 pixels) for the two familiarization trials and three test trials (for a compilation of the original video clips, see Supplementary Material S9). All familiarization and test events were presented without any sound. In all events, the toy locomotive(s) moved approximately at a constant speed of 2 cm/s across the monitor. The scaling factor of the video presentations was 1: 3.4 (i.e., 1 cm on the screen represented 3.4 cm in the real setting). The two experimental conditions differed only with respect to the event presented on the first familiarization trial: In the twoobject condition, this event started by showing one of the toy locomotives standing on the railroad track (3 s), which extended horizontally across the screen. The other toy locomotive then entered the scene from the left and stopped some distance to the left of the first locomotive (7 s). After a short pause (6 s), the first locomotive started moving and left the scene to the right. The
Fig. 1. Familiarization and test events in Experiment 1. Arrows are added to the screen shots from the video clips to symbolize direction and visibility of movement. (F1, F2 = familiarization trials, T1 = test trial 1, T2 = preview for test trial 2, T3 = preview for test trial 3).
H. Krist et al. / Cognition 157 (2016) 205–218
timing and spatial arrangement of the movement sequence were chosen so that they matched the targeted two-object solution of the test trials (see Fig. 1: F1 Two-Object Condition). In the one-object condition, only one of the toy locomotives was shown. Like the second toy locomotive in the corresponding event of the two-object condition, it entered the scene from the left and stopped. After a short pause (5 s), it continued its movement and left the scene to the right. This familiarization event mainly served to show that the toy locomotive was self-propelled, that is, that it could start and stop on its own (see Fig. 1: F1 One-Object Condition). In the second familiarization trial, the closed tunnel covered the central portion of the railroad track. Otherwise, it was identical to the first familiarization trial (of both conditions). A locomotive again entered the scene from the left, disappeared in the tunnel, reappeared on the other side before it left the scene to the right (see Fig. 1: F2). In the first test trial, a low-window event was presented; it was identical to the second familiarization trial except that the tunnel now had a low window through which the locomotive should have been visible if passing this section. Except for the changed tunnel, the low-window event was identical to the second familiarization event, that is, no locomotive appeared in the tunnel’s lower window (see Fig. 1: T1). The second and third test trials were identical to the first one except that a brief preview preceded the low-window event. In the preview phase of the second test trial, the tunnel was briefly lifted (8 s) to reveal a toy house in front of the railroad track, inside the right section of the tunnel. The house was big enough to occlude a second locomotive behind it. During the preview phase, the front portion of the locomotive that would enter the scene was visible at the left edge of the screen (see Fig. 1: T2). On the third test trial, the preview of the previous trial was extended (by 8 s) to actually show that a second locomotive stood on the track behind the house (see Fig. 1: T3). 2.1.3. Procedure Children were tested individually by the same female experimenter in a suitable room of their kindergarten. During the experiment children were seated at a table facing the monitor (1700 , resolution: 1680 1050 pixels) of a HP Compaq 6830 s notebook (viewing distance: approx. 50 cm). The video sequences were presented in the above order using Microsoft PowerPoint. After each event, the children were asked to describe what they had seen (What have you just seen?) to make sure they had encoded the relevant information given by the video sequence and to assess their interpretation of the event. If necessary, the video was repeated (once or twice) and the experimenter pointed out where the child’s description was incomplete or incorrect (e.g., I haven’t seen anything here, if the child claimed having seen the locomotive in the tunnel window). The experimenter avoided giving any verbal cues concerning the number of locomotives involved. If children had given their description of the first test event without mentioning a second locomotive, the experimenter verified that they had encoded the event correctly, especially that no locomotive had appeared in the tunnel’s cutout. Only thereafter, they were asked how the seen event could have occurred in reality. If necessary, they were informed that a trick was used to accomplish the event and were asked the main question of how they thought the trick was done (How do you think the trick was done?). Any suggestions other than the targeted two-object explanation were moderately praised, but the child was informed that the trick was done otherwise and prompted for further ideas. On the second and third test trial, children were again asked for a description and possible explanations of the respective event. A testing session ended after the third and final test trial, or as soon as the child gave the tar-
209
geted two-object explanation. The complete sessions were taped on a voice recorder and transcribed for data analysis. 2.2. Results In our main analyses, we examined, based on the transcripts from the experimental sessions, whether and at which point during the session each child produced a two-object solution. An answer counted as a solution only if it referred to two locomotives and included an indication of where the first locomotive had stopped and where the second locomotive had started from (cf. Aguiar & Baillargeon, 2002). Inter-rater agreement regarding this measure was 93% (kappa = 0.90) for a randomly drawn subsample of 25% of all transcripts.1 Table 1 shows the absolute and relative solution frequencies as a function of age, condition, and test trial (for a graphical representation of the results, see Supplementary Material S1). It is apparent that solution frequencies were low even in the two-object condition, but that they increased with age, v2 (N = 60) = 18.45, p = 0.007. (All contingency analyses reported in this paper were performed as two-tailed Fisher’s exact tests using SPSS, Version 20; cf. Mehta, 1994.) Contrary to our prediction, the presentation of two locomotives in the first familiarization trial did not significantly help children to find the solution, v2 (N = 60) = 5.54, p = 0.12. As Table 1 shows, however, 6-year-olds performed much better in the two-object than in the one-object condition. Analyses performed for each age group separately confirmed that the condition effect was significant for the 6-year-olds, v2 (N = 15) = 10.49, p = 0.004, but not for the younger age groups (ps > 0.19). It is remarkable how few of the children could produce a nontrivial two-object solution even after having seen two locomotives simultaneously in the first familiarization trial (two-object condition). We therefore inspected the transcripts for cases in which children did mention a second locomotive, before it was revealed in the preview of the third trial, but did not produce the targeted solution. It turned out that there were only two such cases: One 4-year-old tested in the one-object condition and one 3-year-old tested in the two-object condition referred to two locomotives on test trials 1 and 2, respectively, without being able to give the complete two-object explanation.2 Melkman and Rabinovitch (1998) reported that many of their younger children denied the continuity violation (i.e., they claimed that the object had appeared between the two screens). In the present study, there were also three 3-year-olds who initially claimed that they had seen a locomotive moving in the tunnel window. Among the 15 children who had to be excluded from the main analysis due to experimenter error occurring later in the experimental session, there were three further 3-year-olds and one 4-year-old child who exhibited this striking behavior. 2.3. Discussion Our prediction that 3- to 6-year-old children would exhibit severe difficulties in inferring a second object when confronted with an apparent continuity violation was clearly verified. Before ever being presented with two identical objects simultaneously (i.e., in the first or second test trial of the one-object condition), only 1 The second rater was blind as to the hypotheses tested; the same held for Experiments 2 and 3. 2 In response to the experimenter’s question of how the trick was done, the 4-year said ‘‘Another locomotive” but was unable to explain how the second locomotive was involved and finally dismissed the idea. Similarly, upon being asked how it could be possible that no locomotive had appeared in the tunnel window, the 3-year-old spoke of ‘‘another one” but was not able to specify this statement any further.
210
H. Krist et al. / Cognition 157 (2016) 205–218
Table 1 Absolute (and relative) solution frequencies as a function of age group, condition, and test trial (Experiment1). Age
3 years 4 years 5 years 6 years
One-object condition (N = 29)
(N = 13) (N = 15) (N = 17) (N = 15)
Two-object condition (N = 31)
Trial 1 (without hints)
Trial 2 (hint: toy house)
Trial 3 (hint: 2nd object)
Not solved
Trial 1 (without hints)
Trial 2 (hint: toy house)
Trial 3 (hint: 2nd object)
Not solved
– – – –
1 (16.7%) – – –
1 1 5 4
4 6 3 4
– – – 3 (42.9%)
– 1 (12.5%) 1 (11.1%) 3 (42.9%)
– 3 (37.5%) 5 (55.6%) 1 (14.3%)
7 (100%) 4 (50%) 3 (33.3%) –
(16.7%) (14.3%) (62.5%) (50.0%)
2 out of 29 children (7%) referred to two objects in their attempted explanations and only one of them produced the targeted twoobject solution. This is remarkable for several reasons: First, care was taken to model the original infant task used by Aguiar and Baillargeon (2002) as closely as possible; second, toy locomotives were chosen as age-appropriate stimuli, to maximize the plausibility of the existence of two identical objects that are capable of moving in a self-propelled manner; third, a familiarization trial ensured that children actually saw that the toy locomotive(s) had this capability; fourth, children’s attention was deliberately drawn to the apparent discontinuity, even if they initially denied it; fifth, children saw the test event at least twice, before the second locomotive was revealed, and had the opportunity to come up with multiple ideas; and sixth, a plausible hiding place (a toy house), similar to the one successfully used by 3.5-month-old infants in one of Aguiar and Baillargeon’s (2002) looking-time studies, was introduced in the second test trial. Our prediction that children would benefit from being familiarized with two toy locomotives was only partially supported. While the main effect of condition (one vs. two objects in the first familiarization trial) did not reach statistical significance, it turned out that 6-year-olds performed significantly better in the two-object than in the one-object condition. An obvious reason why younger children were not more likely to solve the task in the two-object than in the one-object condition may be sought in the fact that only one object was presented in the second familiarization trial thereby counteracting a possible priming effect favoring twoobject interpretations of the test events. Of course, the second familiarization trial, in which a single locomotive appeared to pass through a closed tunnel was deliberately included: first, because such an event was also used in the habituation phase of the original infant study (Aguiar & Baillargeon, 2002), and second, because we wanted to examine children’s capability to infer a second object in cases where, at first glance, a single object appears to move in a discontinuous manner. Skipping the second familiarization trial in the two-object condition would have largely eliminated the latter requirement; it would have made it much more likely that children would simply stick to a two-object interpretation of the event. In agreement with this, we found in Experiment 2 that a few children tested in the two-object condition interpreted the second familiarization event in terms of two objects, obviously transferring their interpretation of the first familiarization event to the simple occlusion event presented in the second familiarization trial (see Section 3.3). Once children have zeroed in on a one-object interpretation, they appear to have difficulties to infer a second object as being involved in the discontinuous test event. For younger children this appears to be even the case if the second object is revealed in a preview of the event. This is attested by the fact that many children did not produce a two-object solution at all, that is, not even in the last test trial. An alternative interpretation for this result is that younger children lack the verbal and/or cognitive ability to verbalize or to construct a two-object solution. This interpretation will be considered in more detail below (see Sections 3.3 and 5.2).
(66.7%) (85.7%) (37.5%) (50.0%)
Before any firm conclusions could be drawn from the present results, it was necessary to replicate them using a larger sample. Furthermore, the possibility that children’s performance was affected by the video mode of presentation (i.e., by a videodeficit effect; see Anderson & Hanson, 2010) had to be ruled out by presenting children with real events instead of video clips. Experiment 2 was designed to meet these requirements. 3. Experiment 2 This experiment aimed at replicating the main results from Experiment 1 using a larger sample of 3- to 6-year-old children and a live scenario instead of video clips. To eliminate acoustic cues, all events were presented in an adjacent room. Children watched the events through a window. Except for the presentation mode (live vs. video) and slight modifications of the events, the design and procedure were the same as in the first experiment. We expected to obtain further support for the claim that young children’s verbal reasoning does not parallel infants’ nonverbal reasoning in object-individuation tasks. In particular, we predicted that, with the present task, young children would almost completely fail to infer the involvement of a second object, while older children would perform significantly better. Although we did not obtain a significant condition effect regarding the percentage of two-object solutions in Experiment 1, we still predicted children to be more likely to solve the task and to mention two objects (on the first two test trials) after having been familiarized to two rather than one object. 3.1. Method 3.1.1. Participants In this experiment, 138 children aged 3–6 years participated; there were 38 three-year-olds (M = 41.6 months, SD = 3.70), 32 four-year-olds (M = 52.4 months, SD = 3.56), 35 five-year-olds (M = 63.1 months, SD = 2.43), and 33 six-year-olds (M = 77.0 months, SD = 2.92). Approximately half of the children of each age group were randomly assigned to each of the two experimental conditions. Sixteen additional children were tested but had to be excluded from data analysis due to equipment failure (4 children), experimenter error (1 child), noncompliance (5 children), taciturnity (3 children), or lack of verbal skills (3 children). All children were tested in our laboratory in Greifswald, Germany, while their caretakers (parent or nursery school teacher) waited in our play lounge. Children were recruited from kindergartens located in Greifswald, Germany, or by directly contacting parents living in the Greifswald area using a database obtained via the federal state government of Mecklenburg-Vorpommern, Germany. They participated on a voluntary basis and with the consent of their parents. No information was collected on parents’ education, occupation, or income. All children tested were native German-speakers or possessed adequate conversational skills in the German language. Children were rewarded with a small gift for their participation; and parents were reimbursed for trip expenses.
211
H. Krist et al. / Cognition 157 (2016) 205–218
4
End
3
Position 2
2
1
Position 1
Start
Fig. 2. Schematic depiction of the model-railroad setting used in Experiment 2. Track zones 1–4 were conducting and could be controlled via a potentiometer and switches. The movement sequence for the first familiarization event of the one-object condition was produced by switching on zone 1, and then zones 2, 3, and 4 to make the first locomotive go from Start to Position 1 and then from Position 1 to End. With all other events, zones 1 and 4 were activated to make the first locomotive move from Start to Position 1 and the second locomotive from Position 2 to End, respectively.
3.1.2. Stimuli Stimuli and events were identical to Experiment 1, except for the following modifications. In contrast to Experiment 1, events were presented live. Children watched the events through a oneway mirror (250 cm wide, 150 cm high) from an adjacent room. The setup for the model railroad was mounted onto a wooden board (240 cm wide, 30 cm deep, 1.5 cm thick). It was placed close to the one-way mirror and screened from the rest of the room by partitions to produce the impression of a diorama-like stage. To produce the live events, two identical black electrical toy locomotives (H0 PIKO, 15 cm wide, 3 cm deep, 5 cm high) were run on a track (230 cm long, 3 cm wide; track gauge: 1.65 cm) parallel to the one-way mirror (at a distance of 40 cm). This track was bolted on a wooden plank and was equipped with buffer stops located at each end. It contained four conducting zones that could be controlled via a potentiometer and switches to produce the desired movement sequences (see Fig. 2). Two different wooden tunnels were used: The one employed for the test trials (60 cm wide, 15 cm deep, 11 cm high) had a low window in its midsection (15 cm wide, 8 cm high). For the second familiarization trial, a closed tunnel (60 cm wide, 17 cm deep, 12 cm high) was imposed on the open tunnel. Tunnel entrances were just as big as necessary to allow the locomotives to pass through. A small wooden house (15 cm wide, 1 cm deep, 9 cm high) could be placed inside the tunnel and in front of the railroad track to occlude one of the locomotives standing behind it. All familiarization and test trials were presented live. An opaque curtain was raised and lowered between the trials. The second experimenter was invisible to the child but entered the stage temporarily to present the previews of the house and the second locomotive on the second and third test trial, respectively. Events always started and ended with a locomotive being in full view on the right and left side of the track, respectively. Otherwise, the live events highly resembled the video events presented in the first experiment. The second familiarization event, where a single locomotive appeared to go through the closed tunnel, was actually produced with two locomotives (i.e., like in the test events). 3.1.3. Procedure Children were tested individually by a female experimenter. A second experimenter presented the events in the room adjacent to the testing room. Via a microphone located in the testing room,
the second experimenter could monitor the conversation in the testing room and receive instructions by the first experimenter (to start the next event or to repeat the previous one). The maximum number of repetitions of the first test trial was set to three. In all other respects, the procedure was identical to the first experiment. 3.2. Results Based on the transcripts of each child’s verbal responses, it was assessed whether and at which point during the session the child produced the intended two-object solution. As a more liberal solution criterion, we also coded for two-object answers. Two-object answers were coded whenever a child referred to a second locomotive with or without being able to produce the complete solution. Inter-rater agreement regarding the classification of two-object solutions and two-object answers was 100% and 97% (kappa = 0.94), respectively, for a randomly drawn subsample of 25% of all transcripts. As in Experiment 1, it turned out that the generation of a non-trivial two-object solution constituted the exception rather than the rule: In the one-object condition, only 5 children (7%) solved the task before the second toy locomotive was revealed. Even in the two-object condition, there were only 14 children (21%) who were successful before the third test trial. Table 2 provides an overview over the solution frequencies on each test trial for both conditions of Experiment 2 (cf. Supplementary Material S2). It is apparent that, even in the two-object condition and even in the third test trial (i.e., when the second toy locomotive was shown in a preview), 3- and 4-year-old children were virtually unable to give a two-object solution. Overall, performance improved markedly with age, v2 (N = 138) = 118.89, p < 0.001, and two-object solutions were more frequent in the two-object than in the oneobject condition, v2 (N = 138) = 6.83, p = 0.05. From Table 2 it is apparent that the marginally significant condition effect was mainly due to the fact that children who were presented with two locomotives during familiarization (two-object condition) were more likely to give a two-object solution on the first test trial than children who only saw one locomotive during familiarization (21% vs. 7%). As in Experiment 1, children rarely mentioned a second locomotive in one of the first two test trials without also providing the
Table 2 Absolute (and relative) solution frequencies as a function of age group, condition, and test trial (Experiment 2). Age
3 years 4 years 5 years 6 years
One-object condition (N = 71)
(N = 38) (N = 32) (N = 35) (N = 33)
Two-object condition (N = 67)
Trial 1 (without hints)
Trial 2 (hint: toy house)
Trial 3 (hint: 2nd object)
Not solved
Trial 1 (without hints)
Trial 2 (hint: toy house)
Trial 3 (hint: 2nd object)
Not solved
– – – 5 (29.4%)
– – – –
– 1 (6.3%) 16 (88.9%) 12 (70.6%)
20 (100%) 15 (93.8%) 2 (11.1%) –
– 4 (25%) 4 (23.5%) 6 (37.5%)
– – – 1 (6.3%)
– 3 (18.8%) 9 (52.9%) 9 (56.3%)
18 (100%) 9 (56.3%) 4 (23.5%) –
212
H. Krist et al. / Cognition 157 (2016) 205–218
Table 3 Absolute (and relative) frequencies of two-object answers as a function of age group, condition, and test trials in Experiment 2. Age
3 years 4 years 5 years 6 years
One-object condition (N = 71)
(N = 38) (N = 32) (N = 35) (N = 33)
Two-object condition (N = 67)
Trial 1 (without hints)
Trial 2 (hint: toy house)
Trial 3 (hint: 2nd object)
Not mentioned
Trial 1 (without hints)
Trial 2 (hint: toy house)
Trial 3 (hint: 2nd object)
Not mentioned
– 1 (6.3%) – 5 (29.4%)
1 (5%) – – –
16 14 17 12
3 (15%) 1 (6.3%) 1 (5.6%) –
2 5 5 8
– – – –
12 (66.7%) 7 (43.8%) 11 (64.7%) 8 (50%)
4 (22.2%) 4 (25.0%) 1 (5.9%) –
(80%) (87.5%) (94.4%) (70.6%)
complete two-object solution. This occurred in two cases in the one-object condition (one 3-year-old and one 4-year-old) and in five cases in the two-object condition (two 3-year-olds, one 4year-old, one 5-year-old, and one 6-year-old). With one exception (a 3-year-old tested in the one-object condition), all of these instances occurred on the first test trial. Table 3 summarizes the results of Experiment 2 in terms of the frequency of two-object answers across all test trials as a function of age and condition (cf. Supplementary Material S3). These results agree well with those obtained for the two-object solutions, as far as the nontrivial test trials (1 and 2) are concerned (cf. Table 2). In addition, they indicate that the large majority of children referred to two toy locomotives at some point during the test, but usually only after the second locomotive had been shown in the preview phase of the third test trial. Again, the age effect turned out to be highly significant, v2 (N = 138) = 22.15, p = 0.002. The condition effect was even more reliable than in the corresponding analysis of the twoobject solutions, v2 (N = 138) = 14.11, p = 0.001; it was also highly significant with respect to the percentage of non-trivial two-object answers (10% vs. 30%), p = 0.005. In Experiment 1 some of the younger children insisted that they had seen a locomotive in the tunnel window. It was therefore of interest to assess whether this tendency to deny the continuity violation would also occur in the live scenario of the present experiment. In fact, 18 of the 3-year-olds (47%), 6 of the 4-year-olds (19%), 5 of the 5-year-olds (14%), and two 6-year-olds (6%) initially claimed that they had observed a locomotive moving in the tunnel window, v2 (N = 138) = 27.51, p < 0.001 (for the age effect). Ten of these children (nine 3-year-olds and one 5-year-old) even stuck to their erroneous claim after one or more repetitions of the first test trial (and despite the experimenter’s efforts to convince them otherwise). 3.3. Discussion Our attempt to replicate the major finding of Experiment 1 with a larger sample of 3- to 6-year-old children using live instead of video presentations was successful. Again, children hardly produced any two-object solutions without any hints. This was especially true for younger children and the one-object condition. Overall, children did not benefit from the live presentation of the events so that a video-deficit effect (Anderson & Hanson, 2010) can be ruled out as a cause for children’s difficulties to infer the involvement of a second object. If anything, younger children performed worse with the live than with the video scenario. Threeyear-olds even failed to solve the task on the last trial, that is, after they had seen the second locomotive in the preview phase. Children’s difficulties to come up with or to construct a twoobject solution cannot merely be attributed to their lack of verbal skills or unwillingness to answer because most children referred to two toy locomotives at some point during the experiment, albeit rarely before the second locomotive was shown on the third test trial. Although performance factors, such as insufficient verbal skills, may have contributed to children’s failure to articulate a complete two-object solution, the many cases in which young children gave two-object answers, but only after having seen the two
(11.1%) (31.3%) (29.4%) (50%)
locomotives simultaneously, is at odds with a pure performance account. Rather it supports the assumption that preschoolers tended to conceive the critical events shown in the first two test trials as involving a single object only, despite the apparent discontinuity of movement. This single-object bias was also expressed in children’s elaborations of their explanations. Most children referred to some kind of physical reason for the apparent discontinuity but many of the younger children also gave animistic or magical explanations. Typical physical explanations were that the locomotive did not appear in the tunnel window because it took another path (e.g., above the window or beneath the tunnel), that it went too fast to be seen, or that someone displaced it. Nonphysical explanations most often included that the locomotive hid away, made itself invisible, jumped/flew to the other side of the tunnel, or made itself disappear and then reappear again. The single-object bias was particularly obvious in those children who initially, and sometimes even repeatedly, denied the spatio-temporal evidence against a simple one-object interpretation, namely the fact that no locomotive had appeared in the tunnel window. Interestingly, denials of the (apparent) continuity violation were rather more frequent with the live presentation in Experiment 2 than with the video presentation in Experiment 1 and were now even found with some of the 5- and 6-year-olds. The prediction that children would be more likely to solve the task or to mention two objects after having been familiarized to two rather than one object was now confirmed. Especially on the first test trial, children were more likely to mention two objects and to solve the task in the two-object than in the one-object condition. This result is in line with a priming account of the condition effect. Indeed, there were six children tested in the two-object condition (one 4-year-old, four 5-year-olds, and one 6-year-old) who interpreted the event in the second familiarization trial (closed tunnel) in terms of a two-object solution despite the lack of an apparent continuity violation. In all but one of these cases, the children also solved the task on the first test trial. A similar, yet negative, carry-over effect could also be responsible for young preschoolers’ general difficulties with our task, especially their tendency to deny the apparent continuity violation. As a closed-tunnel event was shown in the second familiarization trial (in Experiments 1 and 2), it was appropriate, in that trial, to shift one’s gaze to the other end of the tunnel after the locomotive had disappeared. It is conceivable that (young) preschoolers would transfer this tracking strategy from familiarization to test where it was no longer appropriate; they would thus fail to fixate the tunnel window when the toy locomotive should have appeared therein (transfer-of-tracking hypothesis).3
4. Experiment 3 This experiment was mainly designed to rule out the transferof-tracking hypothesis, but it also served the purpose of establishing an adult baseline for our object-individuation task. To test the 3 We would like to thank an anonymous reviewer for drawing our attention to this possibility.
H. Krist et al. / Cognition 157 (2016) 205–218
transfer-of-tracking hypothesis, we recorded the eye movements of 3-year-olds, 5-year-olds, and adults by means of an eye tracker while they watched video clips of the events presented in Experiment 2. All participants were interviewed as in the preceding experiments to make sure that the testing situation was comparable and to check whether children’s verbal performance was in the same ballpark as in Experiments 1 and 2. An additional group of adults was tested by presenting them with the video clips on a laptop monitor without recording their eye movements. This was done to increase the sample size for the adult baseline and to test for possible effects of the eye-tracker setting on adults’ performance. As in the preceding experiments, two familiarization conditions (one-object vs. two-object condition) were employed in a between-subject design. 4.1. Method 4.1.1. Participants A total of 85 preschool children and 76 adults participated in this experiment. Among the children, there were 37 three-yearolds (M = 42.2 months, SD = 3.14) and 48 five-year-olds (M = 66.2, SD = 3.21). Most of the adult participants (91%) were psychology students partaking for course credit (M = 23.3 years, SD = 4.83). Participants were randomly assigned to one of the two experimental conditions (one-object vs. two-object condition). Twenty additional children were recruited but had to be replaced for data analysis, because of technical problems (5 children), experimenter error (6 children), or non-compliance (9 children). Four additional adults were recruited but had to be replaced, because they had partaken in a similar experiment before. All participants were tested in our laboratory at the University of Greifswald, Germany. The children’s caretakers generally waited in our play lounge, but in some cases the youngest children were accompanied by their caretakers to the laboratory. In these cases, caretakers were instructed not to interfere with the experimental proceedings. All participants participated on a voluntary basis. Additionally, children’s parents gave their consent. No information was collected on parents’ education, occupation, or income. All children tested were native German-speakers or possessed adequate conversational skills in the German language. 4.1.2. Stimuli and procedure The events were essentially the same as in Experiment 2, but were presented as video clips (see Supplementary Material S10). Eye movements were recorded with all participants using a Tobii T120 eye tracker (1700 monitor, 1280 1024 pixels), except for 36 of the adults who were presented with the video clips on a notebook monitor (Fujitsu Lifebook NH 71, 17.300 , 1920 1080 pixels). The scaling factor of the video presentations was 1: 6.6 and 1: 6.0 for the eye tracker and the notebook, respectively (i.e., 1 cm on the screen represented 6.6 cm or 6.0 cm in the real world of the setting described in Exp. 2). In all events, the toy locomotive(s) moved approximately at a constant speed of 2.9 cm/s across the monitor. The occlusion interval, that is, the time interval between the locomotive’s disappearance and its (apparent) reappearance, amounted to approximately 3 s. Participants watched the videos from a distance of approximately 60 cm. If necessary, children were seated on a height-adjustable child safety seat. The procedure followed the protocol of Experiment 2, but there was only a single experimenter, who started the presentation of the video clips by keystroke. Again, the experimenter pointed out where children’s descriptions of the familiarization and test events were incomplete or incorrect, but the video clips were never repeated. For eye tracking, the videos were presented using Tobii Studio 2.2.7 software. Eye gaze was registered at a sampling frequency of 60 Hz and analyzed using the Tobii Fixation Filter with default
213
parameters (velocity and distance threshold: 35 pixels/samples). The eye tracker was calibrated individually using the regular 5-point calibration procedure of Tobii Studio with default settings (red dots, grey background, medium speed, full screen). 4.2. Results and discussion 4.2.1. Verbal responses Two-object solutions were coded as in the preceding experiments from the transcripts of the experimental sessions. (Interrater reliability of this measure was perfect for a randomly drawn subsample of 25% of all transcripts). None of the 3-year-olds and only four of the 5-year-olds (8%), who were all tested in the twoobject condition, solved the task before the two-object preview on the third test trial (three children solved on trial 1 and one child on trial 2). On the third test trial, only one 3-year-old (3%) solved the task; and even among the 5-year-olds there were only 25 children (52%) who solved the task on the third test trial or earlier (see Table 4; cf. Supplementary Material S4). Except for 5-year-olds’ relatively low solution rate on the third test trial, the present results are in good agreement with those from Experiments 1 and 2. Again, the age effect was statistically reliable, v2 (N = 85) = 25.80, p < 0.001, but the condition effect did not reach statistical significance, v2 (N = 87) = 3.75, p = 0.31. As in Experiments 1 and 2, some of the preschoolers denied the apparent continuity violation by claiming that they had seen the locomotive in the tunnel window. Such denials were observed with six of the 3-year-olds (16%) and five of the 5-year-olds (10%). Most, but not all, of these denials (73%) occurred on the first test trial only. By contrast to the children, the large majority of the adults (77%) tested in the one-object condition solved the task on the first test trial; two more adults (5%) solved it on the second test trial, and the remaining seven adults (18%) did so on the third test trial (Table 4). All adults tested in the two-object condition solved the task on the first test trial. The effect of condition was statistically significant, v2 (N = 76) = 9.61, p = 0.004. Among those adult participants who were tested in the one-object condition, the solution rates did not significantly differ between the two settings (notebook vs. eye tracker), v2 (N = 39) = 3.15, p = 0.157. Our adult participants could thus mostly solve the object-individuation task but their performance was only at ceiling if the second locomotive was introduced during familiarization. In other words, even some of the adults failed to infer the existence of another identical object if it had not been presented previously, either during familiarization or as a hint in a preview of the test trial.4 Summarizing the verbal results, one may conclude that the 3- and 5-year-old children performed on a similar level as in the preceding experiments and much worse than adults did. Thus, an important precondition for testing the transfer-of-tracking hypothesis by analyzing participant’s eye movements was met. 4.2.2. Looking behavior To examine the transfer-of-tracking hypothesis, we analyzed the looking behavior of all children (N = 85) and those adults from whom eye movements were recorded and who were also included in the analyses of their verbal answers (N = 40). According to the transfer-of-tracking hypothesis, children should tend to look at the tunnel exit in anticipation of the locomotive’s reappearance 4 While children’s alternative explanations of why the locomotive could not be seen in the tunnel window resembled those in the previous experiments, those adults who did not solve the task immediately almost exclusively tried to explain the apparent discontinuity via some kind of optical trick or video editing. Yet, a few of the adults also gave some of the (less plausible) physical explanations typical for children (e.g., the locomotive went too fast to be seen or took another path).
214
H. Krist et al. / Cognition 157 (2016) 205–218
Table 4 Absolute (and relative) solution frequencies as a function of age group, condition, and test trial (Experiment 3). Age
3 years (N = 37) 5 years (N = 48) Adults (N = 76)
One-object condition (N = 81)
Two-object condition (N = 81)
Trial 1 (without hints)
Trial 2 (hint: toy house)
Trial 3 (hint: 2nd object)
Not solved
Trial 1 (without hints)
Trial 2 (hint: toy house)
Trial 3 (hint: 2nd object)
Not solved
– – 30 (76.9%)
– – 2 (5.1%)
1 (5.6%) 11 (45.8%) 7 (18%)
17 (94.4%) 13 (54.2%) –
– 3 (12.4%) 37 (100%)
– 1 (4.2%) –
– 10 (41.7%) –
19 (100%) 10 (41.7%) –
2nd Familiarization Trial
1st Test Trial (Without Hints)
Fig. 3. Exemplary heat maps from Experiment 3: Three-year-olds’ looking behavior in the second familiarization trial and the first test trial (one-object condition). The heat maps represent absolute looking durations for the interval during which the locomotive(s) were occluded. Direction of movement was from right to left.
in the closed-tunnel event (second familiarization trial) and to transfer this looking strategy to the open-tunnel event (first test trial), thereby skipping the tunnel window. Mere inspection of so-called heat maps generated by the eye-tracking system already revealed that the transfer-of-tracking hypothesis was invalidated by our data (see Supplementary Material S5–S8). The heat maps represent how long participants looked at which spot of the scene in the closedtunnel and the open-tunnel event. They refer to the time interval during which the locomotives were occluded (occlusion interval). For exemplary purposes, the respective heat maps from the 3year-olds are shown in Fig. 3 for the one-object condition only. The 3-year-olds exhibited virtually the same looking pattern as the entire sample: In both conditions, they accumulated the longest looking times at the tunnel exit in the closed-tunnel event and at the tunnel window in the open-tunnel event, most probably awaiting the locomotive’s reappearance. In other words, the 3-year-olds did not skip the tunnel window on the first test trial but fixated it most of the time and much longer than the tunnel exit. For additional analyses of participants’ looking behavior, three rectangular areas of interest (AOIs) were defined, covering the entrance (AOI 1), the window (AOI 2), and the exit of the tunnel (AOI 3; see Fig. 4). For the occlusion interval, we assessed which AOI participants were fixating first and how long it took them to fixate each AOI for the first time. According to the transfer-of-tracking hypothesis, on the first test trial, children should tend to fixate the tunnel exit before the tunnel window. This prediction was clearly not fulfilled. Ninetyone (83%) of those 110 participants for whom first-fixation data could be registered during the occlusion interval looked to the tunnel window first, while the remaining participants looked to the tunnel entry first; not a single participant looked to the tunnel exit first.5 Virtually the same pattern was observed in all three age groups, v2 (N = 110) = 0.54, p = 0.77, and in both conditions (p > 0.40, for each age group). 5
For this analysis, the rate of valid eye-tracking samples amounted to approximately 93% (3-year-olds: 95%; 5-year-olds: 93%; adults: 91%).
Despite these clear-cut results and the observed looking-time patterns (Fig. 3 and Supplementary Material S5–S8), it was still conceivable that children did look to the tunnel exit in anticipation of the locomotive’s reappearance, but only with a certain delay, and only after having inspected the tunnel window. We therefore analyzed how many participants fixated the tunnel exit at all during the occlusion interval. Whereas 115 out of 121 participants (95%) did so in the closed-tunnel event (after a mean delay of 0.51 s, SD = 0.61), in the open-tunnel event, only 10 out of 110 participants (9%) fixated the tunnel exit (after a mean delay of 1.77 s, SD = 0.89). Among the latter participants, there were four 3-yearolds, two 5-year-olds, and four adults.6 In sum, both objectives of the present experiment were reached. First, the transfer-of-tracking hypothesis could be clearly falsified by analyses of participants’ eye movements. Preschoolers’ difficulties with our task and their tendency to deny the apparent continuity violation can thus not be attributed to their failure to fixate the tunnel window in the test trials. Second, an adult baseline was established for both conditions of our objectindividuation task. We found that adult participants exhibited no difficulties, whatsoever, if the two-object solution was primed (two-object condition) but that about every fifth of the adults failed to solve the task if the second object had to be inferred (one-object condition, trials 1 and 2). 5. General discussion The theoretical question guiding the present research was whether, or to what extent, children’s explicit physical reasoning about the numerical identity of objects parallels young infants’ implicit reasoning processes. A related question is whether the object-individuation processes and the corresponding physical reasoning observed in infancy remain functional throughout 6 Within each age group, exactly half of those participants who fixated the tunnel exit during the occlusion interval were tested in the one-object and in the two-object condition.
H. Krist et al. / Cognition 157 (2016) 205–218
215
Fig. 4. AOIs as defined in Experiment 3.
development. Perhaps the most remarkable and sophisticated competence ascribed to infants so far served as a starting point to address these questions: young infants capability to infer the existence of an object they have never encountered if presented with an apparent continuity violation (Aguiar & Baillargeon, 2002; Kuhlmeier, Bloom, & Wynn, 2004; Spelke et al., 1995). 5.1. Preschoolers fail to infer an occluded object explicitly By and large, we found that children aged 3–6 years are not able to perform this feat on an explicit or verbal-cognitive level. Only very few children seemed to realize that and how the involvement of a second object could explain the observation of a single object apparently moving on a discontinuous path. Our results are thus in sharp contrast to the implicit infant competencies demonstrated in the experiments by Aguiar and Baillargeon (2002) on which the present studies were modeled. They replicate previous findings reported by Melkman and Rabinovitch (1998) and provide even more compelling evidence for children’s difficulties to infer a second identical object based on the observation of an object apparently moving on a discontinuous path. Instead of the cardboard representations of animal characters used by Melkman and Rabinovitch (1998), we presented children with real toy locomotives, either in video clips (Exp. 1 and 3) or in a live scenario (Exp. 2), and adapted the stimulus events as much as possible to those employed in the original infant studies. Additionally, we made several efforts to increase the likelihood that children would realize that and how a second object was involved. Most importantly, at the start of the experimental session, half of the children were familiarized to two identical toy locomotives starting and stopping on their own (two-object condition) in exactly the same way as in the test trials but without any occluders. Therefore, as opposed to the one-object condition, these children did not have to infer the involvement of an object they had never seen before and, when confronted with an apparent continuity violation on the first test trial, could simply refer back to the movement sequence witnessed earlier. Additionally, in the second test trial, we presented children with a potential hiding place for the second object much as Aguiar and Baillargeon (2002) did in
their Experiments 3, 6, and 7. Three-year-olds did not benefit from any of these manipulations and even had difficulties constructing the two-object solution on the last test trial where the second locomotive was revealed immediately before the start of the critical event. The latter result is in particular contrast to Aguiar and Baillargeon’s (2002) finding that 3-month-old infants, who do not yet infer a second object without any hints (i.e., react with longer looking times to the apparent continuity violation), do so when given a preview of the two objects. While older children performed significantly better than younger ones in the present experiments, non-trivial task solutions (i.e., before the third test trial) were rare even among older preschoolers, especially those who were tested in the one-object condition. Even some of the adults tested in Experiment 3 failed to infer the involvement of a second object in the one-object condition, although all adults tested in the two-object condition succeeded on the first test trial. Our experimental findings thus strongly suggest that children’s explicit physical reasoning is not (tightly) constrained by nonverbal object-individuation and/or reasoning processes as attributed to young infants (for a recent review and account of object individuation in infancy, see Baillargeon et al., 2012). Theoretically, the reason for this could be that the object-individuation processes themselves change with development. However, given neurophysiological evidence for the continuity of core competencies (e.g., Libertus, Brannon, & Woldorff, 2011) and particularly the demonstration that, in adults, neurons of the ventral visual cortex respond similarly with the presentation of two objects moving sequentially as with the presentation of a single object moving discontinuously (Yi et al., 2008), this possibility does not seem to be very likely. It is much more plausible to assume that core competencies not only develop early but also remain intact throughout development and that implicit objectindividuation constitutes one of these core competencies (Carey, 2009; Spelke, 2000; Spelke & Kinzler, 2007). However, this is not to say that the particular feat under consideration, namely the ability to infer the existence of a second object from an (apparent) continuity violation, is part of these core competencies. It may be that the basic mechanisms responsible for the individuation and tracking of objects develop early in
216
H. Krist et al. / Cognition 157 (2016) 205–218
infancy and remain functional thereafter, whereas the reasoning processes operating on the outputs of the systems of core cognition undergo substantial developmental change. In fact, the results from the present research strongly suggest that preschoolers’ explicit physical reasoning does not emerge seamlessly from infants’ implicit physical reasoning. 5.2. Children’s difficulties are genuine Before discussing the claim of a developmental discontinuity in children’s physical reasoning any further, we have to rule out alternative accounts of children’s difficulties with our objectindividuation task. According to what might be termed a verbaldeficit account, one might object that our task does not assess what children think but merely what they are able to verbalize. The mere fact that even some of the adults tested in Experiment 3 were not able to infer the existence of a second toy locomotive (in the one-object condition) suggests that preschoolers’ difficulties with the present task are genuine and renders a verbal-deficit account unlikely. Further evidence against a verbal-deficit account was obtained in Experiment 2 where we found that even most of the 3-year-old children did mention a second locomotive, if it was shown in a preview of the test trial but not otherwise. Moreover, a single-object interpretation was expressed unequivocally in the elaborate explanations that many children advanced to explain how the trick was done, not to speak of the observation that some children denied the apparent discontinuity in the first place. There is also the theoretical possibility that children may not be able to verbalize what they consciously realize due to an insufficient level of explicitness of their representations. Such a distinction between representations that are conscious but not (yet) verbalizable and those that are both, conscious and verbalizable, has been advocated by Karmiloff-Smith (1992) in her theory of representational redescription (RR theory) and termed E2 and E3, respectively. This distinction could, however, not be validated empirically (but see Pine & Messer, 2003, for longitudinal evidence partially supporting RR theory in this respect) and has remained a theoretical speculation (cf. Cheung & Wong, 2011; KarmiloffSmith, 1992). In any case, it cannot account satisfactorily for the present results because the arguments already raised against a verbal-deficit account also apply to an account appealing to the E2-E3 distinction. In particular, children who elaborated a singleobject explanation (e.g., the locomotive went so fast so that one could not see it) were most likely not aware of a second object at that time. Moreover, the fact that most children did mention the second object after it had been shown in the preview of the final test trial, but not earlier, also renders the E2-E3 account unlikely. Other alternative accounts may relate to methodological issues. Although we modeled our object-individuation task closely on the one used in the corresponding infant studies (Aguiar & Baillargeon, 2002), it may be that the specific sequence of events employed was not appropriate for preschoolers. In particular, presenting the closed-tunnel event in the second familiarization trial could have influenced children’s performance negatively. By analyzing participants’ eye movements we could exclude the transfer-of-tracking hypothesis according to which children would transfer a simple looking strategy from the closed-tunnel to the open-tunnel event (s). In Experiment 3 we found, contrary to the transfer-oftracking hypothesis, that 3- and 5-year-old children adapted their looking behavior to the presence of the tunnel window in the same way as adults did. They spent most of the time during which no locomotive was visible to fixate the tunnel exit in the closedtunnel event and the tunnel window in the open-tunnel event. Another theoretical possibility is that the observation of the closed-tunnel event may have primed a one-object interpretation of the following test events. In light of the fact that we have
obtained evidence for a priming effect related to the presentation of one versus two locomotives in the first familiarization trial (one-object vs. two-object condition), at least with the live scenario used in Experiment 2, this possibility might appear likely. However, in follow-up studies in which no familiarization trials were used, we have obtained basically the same results as in the present experiments (Krist & Sieber, 2016; Krüger & Krist, 2013; also see Melkman & Rabinovitch, 1998). A priming account of children’s single-object bias is therefore also not tenable. 5.3. A processing account of preschoolers’ single-object bias Thus, there is now strong and converging evidence supporting the assumption that the observed discrepancy between infants’ successes and preschoolers’ failures regarding the objectindividuation tasks under consideration is a real phenomenon rather than a methodological artifact. How can this discrepancy be accounted for without falling back to an empiricist or Piagetian view of infants’ minds? One possibility is to think of core competencies as being encapsulated in domain-specific modules (e.g., Leslie, 1994; cf. Carey, 2009, p. 460). But even if early competencies are encapsulated and hence inaccessible to conscious thought (Fodor, 1983), their output should nonetheless be available for further cognitive processing. In this case, higher-level thought would not be constrained by core knowledge itself but by the representations delivered by the systems of core cognition. The crucial point seems to be, however, that there are virtually no top-down influences on infants’ processing of outputs from their perceptualcognitive ‘‘input analyzers” (Carey, 2009) so that it may well be that infants perform some kind of inferential processes based on these outputs. In the case of object-individuation processes, it is conceivable that such inferential processes allow young infants to ‘‘posit” a second object and even to construct a two-object ‘‘explanation”, if confronted with an apparent continuity violation (Aguiar & Baillargeon, 2002). These inferential processes would not be accomplished by the core system responsible for basic object individuation (presumably the object-file system; see, e.g., Carey & Xu, 2001; Cheries, Mitroff, Wynn, & Scholl, 2009; Scholl & Leslie, 1999) but would operate on the output of this system. In the present case, the output could be the detection of the continuity violation (i.e., one object plus ‘‘error”) but not its explanation (i.e., two objects); the involvement of another object would still have to be inferred. Two observations lend support to this assumption: First, Aguiar and Baillargeon (2002, pp. 299–303) analyzed 3.5-month-old infants’ looking behavior across the test trials and found evidence suggesting that the infants were initially surprised when the object did not appear where it should have. Second, even the adults tested in the one-object condition of our Experiment 3 hardly ever inferred the involvement of a second object instantly (see also Aguiar & Baillargeon, 2002, p. 282). Our claim is therefore that the discontinuity in the development of the ability or likelihood to infer a second object from the observation of an object apparently moving on an interrupted path is not due to the development of the mid-level processes responsible for the tracking and individuation of objects but to developmental changes in higher-order cognitive processes. Unlike young infants, older children are equipped with conceptual knowledge encoded linguistically and organized in naive theories (cf. Carey, 2009; Gopnik & Meltzoff, 1997). This rich conceptual knowledge allows children to improve their predictions and their understanding of how objects and people behave, but it may also bias their thinking because it implies that certain possibilities are given higher prior probabilities than other ones (e.g., Bonawitz, van Schijndel, Friel, & Schulz, 2012). These considerations lead us back to the striking fact that several children, especially younger ones tested with real objects
H. Krist et al. / Cognition 157 (2016) 205–218
(Exp. 2), tended to deny the apparent discontinuity of movement in the present task. Although the same phenomenon had already been reported in the literature (Melkman & Rabinovitch, 1998), we did not expect it to be as prevalent and robust as observed. In hindsight, it makes very much sense if construed as the consequence of a strong top-down influence on the interpretation of the event. Children’s prior conception of the event combined with their naive theory of object motion led them to conclude that the locomotive actually had appeared in the tunnel window rather than to acknowledge the negative evidence and try to explain it. If this interpretation is correct, the opposite phenomenon should also occur in certain situations, namely that children’s prior assumption that two objects are involved leads them to conceive of an occlusion event as involving two objects despite spatiotemporal counterevidence. This is exactly what we found in a further study on preschoolers’ interpretation of events in which objects moved into and out of view sequentially (Krüger & Krist, 2013). Most children not only failed to infer the involvement of a second object from an apparent movement discontinuity, as in the present experiments, but many failed to do so even with an apparent color change. Yet, after they had been informed about the presence of a second object, they ‘‘blindly” transferred the two-object assumption to a subsequent trial where a screen was used that was too narrow to occlude both objects (cf. Wilcox, 1999). To be sure, the observation, that young children often tend to adhere to their naive theories or biases despite empirical counterevidence, is not new (e.g., Huang, 1931; Kaiser, Proffitt, & McCloskey, 1985). Similarly, the suggestion that children think differently than infants is far from original but at the heart of traditional theories of cognitive development (e.g., Piaget, 1951; Vygotsky, 1978). Yet, we think we should take seriously both, infants’ amazing cognitive competencies as well as discontinuities in further cognitive development. Carey’s (2009) theory of conceptual development is probably the best example of such a hybrid view. According to her theory, systems of core knowledge, such as the object-file system, remain invariant across the life span, while conceptual knowledge as represented in naive theories undergoes marked developmental change. Our account of preschoolers’ difficulties with certain object-individuation tasks modeled on infant studies is in good agreement with this theoretical framework. We assume that the output of the objectindividuation system is similar in preschoolers and in young infants; what makes the difference are the prior probabilities assigned to a one-object versus a two-object interpretation and the inferential processes induced by the (unresolved) detection of a continuity violation. Further research is needed to validate or disprove the present account. Certain empirical predictions can be derived from it. One would expect for example, that preschoolers should perform much better if the output of the object-individuation system already includes a second object, that is, if the spatio-temporal properties of a discontinuous movement lead to the immediate ‘‘perception” of two objects. Ideally, one could probe the output of the object-individuation system neurophysiologically (e.g., Yi et al., 2008), but as a proxy one could also use reaction-time paradigms or even verbal reports from adult participants (for a possible dissociation between the last two measures, see Mitroff, Scholl, & Wynn, 2005). Recording children’s eye movements and looking times, as we did in Experiment 3, could also be employed for this purpose. Perhaps the most exciting prospect of our agenda is to bridge the gap between infants’ and preschoolers’ reasoning by focusing on when and how major transitions occur, ideally using longitudinal designs. Future theories of early cognitive development will then be confronted with the challenging task to account for the complex pattern of empirical findings within and across different tasks and domains that is likely to evolve from this research.
217
We are optimistic, however, that nature has not made the puzzle of cognitive development too complicated for human beings to solve it, eventually. Perhaps, developmentalists, like young children, have yet to overcome certain biases that prevent them from accepting evidence contradicting their beliefs. One such bias may be that cognitive development must be either continuous or discontinuous. The evidence suggests that it may be both. Acknowledgements This work was funded by DFG – Germany Grant KR-1213/3-1. Part of this research was presented at the Biennial Meeting of the Society for Research in Child Development (SRCD), March 31 - April 2, 2011, Montreal, Canada. We would like to thank Wolfgang Bartels for technical assistance as well as Caroline Atlas, Henrike Fischer, Josefine Grzesko, Julia Henke, Jonas Koch, Claudia Pasztor, Annika Nöhring, Friederike Schreiber, Michael Schulz, Lukasz Stasielowicz, Tabea Troschke, Anni Weinke, and Franziska Wussow for help in data collection, transcription and coding. We are grateful for helpful comments by two anonymous reviewers. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.cognition.2016. 09.008. References Aguiar, A., & Baillargeon, R. (1999). 2.5-month-old infants’ reasoning about when objects should and should not be occluded. Cognitive Psychology, 39, 116–157. Aguiar, A., & Baillargeon, R. (2002). Development in young infants’ reasoning about occluded objects. Cognitive Psychology, 45, 267–336. Anderson, D. R., & Hanson, K. G. (2010). From blooming, buzzing confusion to media literacy: The early development of television viewing. Developmental Review, 30, 239–255. Aschersleben, G., Henning, A., & Daum, M. M. (2013). Discontinuities in early development of the understanding of physical causality. Cognitive Development, 28, 31–40. Baillargeon, R. (1994). Physical reasoning in young infants: Seeking explanations for impossible events. British Journal of Developmental Psychology, 12, 9–33. Baillargeon, R. (1999). Young infants’ expectations about hidden objects: A reply to three challenges. Developmental Science, 2, 115–163. Baillargeon, R. (2004). Infants’ reasoning about hidden objects: Evidence for eventgeneral and event-specific expectations. Developmental Science, 7, 391–424. Baillargeon, R., & DeVos, J. (1991). Object permanence in young infants: Further evidence. Child Development, 62, 1227–1246. Baillargeon, R., Li, J., Ng, W., & Yuan, S. (2009). An account of infants’ physical reasoning. In A. Woodward & A. Needham (Eds.), Learning and the infant mind (pp. 66–116). New York: Oxford University Press. Baillargeon, R., Stavans, M., Wu, D., Gertner, Y., Setoh, P., Kittredge, A. K., & Bernard, A. (2012). Object individuation and physical reasoning in infancy: An integrative account. Language Learning & Development, 8, 4–46. Bogartz, R. S., Shinskey, J. L., & Speaker, C. J. (1997). Interpreting infant looking: The event set x event set design. Developmental Psychology, 33, 408–422. Bonawitz, E. B., van Schijndel, T. J. P., Friel, D., & Schulz, L. (2012). Children balance theories and evidence in exploration, explanation, and learning. Cognitive Psychology, 64, 215–234. Carey, S. (2009). The origins of concepts. Oxford: Oxford University Press. Carey, S., & Spelke, E. S. (1994). In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind: Domain specificity in cognition and culture (pp. 169–200). New York: Cambridge University Press. Carey, S., & Xu, F. (2001). Infants’ knowledge of objects: Beyond object files and object tracking. Cognition, 80, 179–213. Cheries, E. W., Mitroff, S. R., Wynn, K., & Scholl, B. J. (2009). Do the same principles constrain persisting object representations in infant cognition and adult peception? The cases of continuity and cohesion. In B. Hood & L. Santos (Eds.), The origins of object knowledge (pp. 107–134). Oxford: Oxford University Press. Cheung, C.-N., & Wong, W.-C. (2011). Understanding conceptual development along the implicit-explicit dimension: Looking through the lens of the representational redescription model. Child Development, 82, 2037–2052. Fodor, J. (1983). Modularity of mind. Cambridge, MA: MIT Press. Gopnik, A., & Meltzoff, A. (1997). Words, thoughts, and theories. Cambridge, MA: MIT Press. Huang, I. (1931). Children’s explanations of strange phenomena. Psychologische Forschung, 14, 63–182.
218
H. Krist et al. / Cognition 157 (2016) 205–218
Kaiser, M. K., Proffitt, D. R., & McCloskey, M. (1985). The development of beliefs about falling objects. Perception & Psychophysics, 38, 533–539. Karmiloff-Smith, A. (1992). Beyond modularity: A developmental perspective on cognitive science. Cambridge, MA: MIT Press. Keen, R. (2003). Representation of objects and events: Why do infants look so smart and toddlers so dumb? Current Directions in Psychological Science, 12, 79–83. Krist, H. (2010). Development of intuitions about support beyond infancy. Developmental Psychology, 46, 266–278. Krist, H. (2013). Development of intuitive statics: Preschoolers’ difficulties judging the stability of asymmetrically-shaped objects are not due to extraneous task demands. Zeitschrift für Entwicklungspsychologie und Pädagogische Psychologie, 45, 27–33. Krist, H., & Sieber, K. (2016). Preschoolers are reluctant to posit the existence of another identical object based on spatio-temporal evidence. Talk presented at the 58th conference of experimental psychologists (TeaP), March 21–23, 2016, Heidelberg, Germany. Krüger, M., & Krist, H. (2013). Reasoning about the unseen: Do preschoolers infer the existence of hidden objects as infants do? Poster presented at the biennial meeting of the Society for Research in Child Development (SRCD), April 18–20, 2013, Seattle, Washington, USA. Kuhlmeier, V. A., Bloom, P., & Wynn, K. (2004). Do 5-month-old infants see humans as material objects? Cognition, 94, 95–193. Leslie, A. M. (1994). ToMM, ToBY, and agency: Core architecture and domain specificity. In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind: Domain specificity in cognition and culture (pp. 119–148). New York: Cambridge University Press. Libertus, M. E., Brannon, E. M., & Woldorff, M. G. (2011). Parallels in stimulus-driven oscillatory brain responses to numerosity changes in adults and seven-monthold infants. Developmental Neurophysiology, 30, 651–667. Mehta, C. R. (1994). The exact analysis of contingency tables in medical research. Statistical Methods in Medical Research, 3, 135–156. Melkman, R., & Rabinovitch, L. (1998). Children’s perception of continuous and discontinuous movement. Developmental Psychology, 34, 258–263. Mitroff, S. R., Scholl, B. J., & Wynn, K. (2005). The relationship between object files and conscious perception. Cognition, 96, 67–92. Onishi, K. H., & Baillargeon, R. (2005). Do 15-month-old infants understand false beliefs? Science, 308, 255–258.
Piaget, J. (1951). The child’s conception of the world. London: Routledge and Kegan Paul (Original work published 1926). Piaget, J. (1954). The construction of reality in the child (M. Cook, Trans.). New York: Basic Books (Original work published 1937). Pine, K. J., & Messer, D. (2003). The development of representations as children learn about balancing. British Journal of Developmental Psychology, 21, 285–301. Povinelli, D. J. (2000). Folk physics for apes: The chimpanzee’s theory of how the world works. Oxford: Oxford University Press. Povinelli, D. J. (2012). World without weight: Perspectives on an alien mind. Oxford: Oxford University Press. Scholl, B. J., & Leslie, A. (1999). Explaining the infant’s object concept: Beyond the perception/cognition dichotomy. In E. Lepore & Z. Pylyshyn (Eds.), What is cognitive science? (pp 26–73). Oxford, UK: Blackwell. Southgate, V., Senju, A., & Csibra, G. (2007). Action anticipation through attribution of false belief by 2-year-olds. Psychological Science, 18, 587–592. Spelke, E. S. (1990). Principles of object perception. Cognitive Science, 14, 29–56. Spelke, E. S. (2000). Core knowledge. American Psychologist, 55, 1233–1243. Spelke, E. S., & Kestenbaum, R. (1986). Les origines du concept d’objet. Psychologie Française, 31, 67–72. Spelke, E. S., Kestenbaum, R., Simons, D., & Wein, D. (1995). Spatiotemporal continuity, smoothness of motion and object identity in infancy. The British Journal of Developmental Psychology, 13, 113–142. Spelke, E. S., & Kinzler, K. D. (2007). Core knowledge. Developmental Science, 10, 89–96. Thelen, E., & Smith, L. B. (1994). A dynamic systems approach to the development of cognition and action. Cambridge, MA: MIT Press. Vygotsky, L. (1978). Mind in society: The development of higher psychological processes (M. Cole, trans).. Cambridge, MA: Harvard University Press. Wellman, H. M., Cross, D., & Watson, J. (2001). Meta-analysis of theory-of-mind development: The truth about false belief. Child Development, 72, 655–684. Wilcox, T. (1999). Object individuation: Infants’ use of shape, size, pattern, and color. Cognition, 72, 125–166. Yi, D.-J., Turk-Browne, N. B., Flombaum, J. I., Kim, M.-S., Scholl, B. J., & Chun, M. M. (2008). Spatiotemporal object continuity in human ventral visual cortex. PNAS, 105, 8840–8845.