beyond irrelevant actions: understanding the role of intentionality in children’s imitation of...

Journal of Experimental Child Psychology 119 (2014) 54–72

Contents lists available at ScienceDirect

Journal of Experimental ChildPsychology

journal homepage: www.elsevier .com/locate/ jecp

Beyond irrelevant actions: Understanding the roleof intentionality in children’s imitation of relevantactions

0022-0965/$ - see front matter � 2013 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.jecp.2013.10.008

⁄ Fax: +1 518 580 5319.E-mail address: [email protected]

Amy K. Gardiner ⁄

Department of Psychology, Skidmore College, Saratoga Springs, NY 12866, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 13 March 2013Revised 18 October 2013Available online 3 December 2013

Keywords:ImitationIntentionalityIrrelevant actionsPreschoolersEvolutionCausal understanding

The current research examines how 3- to 5-year-old children useintentionality to understand the causal structure of objects in anobservational learning context. Two studies are presented in whichthe intentionality of relevant actions was manipulated during toyretrieval demonstrations and contrasted with whether theseactions remained relevant or were rendered irrelevant for thechild’s turn. Of interest were whether children would imitate thefirst action when it was demonstrated intentionally but renderedirrelevant and how they would approach the first action when itwas demonstrated accidentally and remained relevant. Findingsrevealed that children did not align themselves with the demon-strator’s intentions in Study 1, when apparatuses were transparent,but did follow the demonstrator’s intentions in Study 2, whenapparatuses were opaque. This suggests that when causality of rel-evant actions is unambiguous, children use their own causal rea-soning abilities, but ambiguous causal structure prompts childrento defer to a demonstrator. It is suggested that opaque relevantactions may represent a real life parallel to irrelevant actions, theimitation of which is motivated by inherent ambiguity.

� 2013 Elsevier Inc. All rights reserved.

Introduction

Social learning is a critical part of development and remains fundamental throughout the lifespan.Research on children’s social learning abilities has revealed an early proficiency for learning by observ-

http://crossmark.crossref.org/dialog/?doi=10.1016/j.jecp.2013.10.008&domain=pdf

http://dx.doi.org/10.1016/j.jecp.2013.10.008

mailto:[email protected]

http://dx.doi.org/10.1016/j.jecp.2013.10.008

http://www.sciencedirect.com/science/journal/00220965

http://www.elsevier.com/locate/jecp

A.K. Gardiner / Journal of Experimental Child Psychology 119 (2014) 54–72 55

ing the behavior of others. Many studies have investigated preschoolers’ observational learning strat-egies within the domain of object use, focusing on the degree to which children copy observed actions.The study of preschoolers’ imitation of irrelevant actions, which are causally unnecessary for complet-ing a task, has become a focus for many researchers. Irrelevant actions were originally included indemonstrations of toy retrieval tasks to explore children’s social learning strategies (Horner & Whiten,2005). Of interest was whether children would emulate, broadly defined as copying an observed goalusing means other than those observed, or whether they would imitate, broadly defined as copying anobserved goal using the observed means. Emulation would require children to recognize that irrele-vant actions were unnecessary and omit these actions from their performance, whereas imitationwould require them to copy all observed actions, including the irrelevant ones. It was found that chil-dren did not emulate but instead chose to imitate, including the irrelevant actions, in their toy retrie-val attempts (Horner & Whiten, 2005). Since this behavior was demonstrated, children’s imitation ofirrelevant actions has captured the interest of researchers around the world, whose varied approachesare gradually piecing together an understanding of this phenomenon (Flynn, 2008; Kenward, 2012;Lyons, Damrosch, Lin, Macris, & Keil, 2011; McGuigan, Makinson, & Whiten, 2010; McGuigan, Whiten,Flynn, & Horner, 2007; Nielsen, Moore, & Mohamedally, 2012; Simpson & Riggs, 2011).

The behavior of copying irrelevant actions is usually referred to as overimitation, but this termseemingly implies that children imitate something above and beyond what they observed. Thus,the term overimitation undermines the fact that when children imitate irrelevant actions they areactually imitating very precisely, copying everything that they observe, and can be reasonably charac-terized as a misnomer. Given this, the term indiscriminate imitation, implying that children imitate allobserved actions regardless of causal relevance, is preferred and used in this article. In addition to pro-viding greater accuracy in describing imitation of irrelevant actions, using the term indiscriminateimitation allows for a category in which children actually do perform observed actions in excess ofwhat was demonstrated, a behavior that can be termed perseverative imitation (Gardiner, Greif, &Bjorklund, 2011).

Indiscriminate imitation has been demonstrated under a range of different conditions with chil-dren from industrialized nations (Horner & Whiten, 2005; Lyons et al., 2011; McGuigan et al., 2007)as well as traditional African Bushman communities (Nielsen & Tomaselli, 2010). Interestingly, theprecision with which children imitate appears to increase during very early childhood under condi-tions where adults provide realistic demonstrations in which they appear knowledgeable, draw chil-dren’s attention to the task at hand, and successfully achieve the task goal. In these circumstances,indiscriminate imitation of object use emerges reliably by 3 years of age and continues into adulthood(McGuigan et al., 2010). Studies with younger children and infants demonstrate that earlier imitationof object use under these conditions is selective, based on factors such as the physical rationality of thedemonstrator’s actions (Gergely, Bekkering, & Kiraly, 2002; Schwier, van Maanen, Carpenter, & Tom-asello, 2006) and the level of the demonstrator’s social engagement (Brugger, Lariviere, Mumme, &Bushnell, 2007; Nielsen, 2006). This suggests that during very early childhood infants and childrenmay be more likely to rely on their own independent evaluation of object causal structure, but bythe preschool age they become more likely to rely on a demonstrator to gain this knowledge.

Studying and explaining why children imitate irrelevant actions is interesting in its own right, butit is important to keep in mind that including intentionally performed irrelevant actions in demonstra-tions of object use occurs primarily in experimental settings. In real life, there should be no good rea-son for a knowledgeable individual to purposefully include irrelevant actions in a demonstration ofobject use. Thus, in their everyday learning of object use, children should almost always be observingand imitating relevant actions. Therefore, the current research does not focus on what indiscriminateimitation can tell us about why children imitate irrelevant actions. Rather, one aim of this work was tocomprehend the phenomenon of indiscriminate imitation in a way that may have practical applicationto understanding how children learn about everyday objects as they observe demonstrations that donot include superfluous actions.

With similar reasoning, Gardiner and colleagues (2011) sought to place indiscriminate imitation ina functional framework, proposing that it might reflect an evolved learning mechanism. In that study,children were provided with demonstrations of object use in which irrelevant actions were performedwith verbal markers to indicate that the actions were intentional (‘‘There!’’) or accidental (‘‘Whoops! I

56 A.K. Gardiner / Journal of Experimental Child Psychology 119 (2014) 54–72

didn’t mean to do that!’’). The authors found that 3-, 4-, and 5-year-olds imitated irrelevant actionswhen they were performed intentionally but not when they were performed accidentally. This sug-gests that preschool children use the intentions of a demonstrator to guide their understanding ofwhich actions should be performed to make objects function (Gardiner et al., 2011). Given that dem-onstrations of object use that include intentional unnecessary actions should be rare, both in our mod-ern lives and throughout human history, Gardiner and colleagues suggested that presuming allintentional actions are causally necessary is a reasonable assumption that would almost always leadto a correct understanding of how to use objects. Thus, indiscriminate imitation may represent anevolved capacity for the accurate cultural transmission of adaptive knowledge of how to use objectsand tools. This presumption would have been more efficient than the alternative—assuming that ademonstrator’s intentional actions did not contribute to object function, leaving one to determinehow to use an object independently. Such a trial-and-error learning process could be detrimental ifit turned out to be futile and resulted in the loss of critical, adaptive tool-use knowledge. Thus, imitat-ing all intentional actions would have been selected during human evolution as a means of accurateand efficient transmission of such important information (Gardiner et al., 2011).

The work of Gardiner and colleagues (2011) establishes intentionality as a guiding factor in chil-dren’s understanding of which actions are causally necessary for object function when children learnby observing others; intentional actions are consistent with causal relevance, whereas accidental ac-tions are consistent with causal irrelevance. However, their study varied the intentionality of only ac-tions that were actually causally irrelevant. The primary aim of the current work was to go beyondirrelevant actions to investigate whether children also use intentionality to guide their understandingof causal necessity in the more realistic context of relevant actions.

In Study 1, children were presented with demonstrations in which three relevant actions were per-formed to retrieve toys from within a series of apparatuses constructed of clear plastic materials. Thefirst action was differentially marked as intentional (‘‘There!’’) or accidental (‘‘Whoops! I didn’t meanto do that!’’). For the child’s turn, the first action either remained causally relevant for toy retrieval orwas rendered irrelevant so that only the second and third actions needed to be performed. This cre-ated two conditions in which the intentionality of the demonstration was inconsistent with the causalrelevance of the actions for the child’s turn and two conditions in which intentionality and relevancewere consistent. Table 1 provides a summary of the study design.

Of primary interest were whether children would perform the first action when it was causallyirrelevant, given that it had been demonstrated intentionally, and how children would approach thefirst action when it was causally relevant but had been demonstrated accidentally. This inconsistencybetween intentionality and causality is similar to conditions created by DiYanni and Kelemen (2008),who investigated whether children would use intentionality in a tool-use context where a demonstra-tor’s intentions clashed with tool functionality. In this series of studies, an experimenter accidentallychose, and subsequently rejected, a functional tool and then intentionally chose a nonfunctional tool,which she attempted to use to crush a cookie. In contrast to the findings of Gardiner and colleagues(2011), the majority of 2- to 4-year-old children in most conditions did not follow the demonstrator’sintentions, instead selecting the functional tool even though it had been chosen accidentally and re-jected. The condition in the current study where a relevant action is performed accidentally and re-mains relevant for the child’s turn is similar to the pairing by DiYanni and Kelemen (2008) of anaccidental choice with a functional tool, whereas the condition where a relevant action is performedintentionally but rendered irrelevant for the child’s turn is similar to DiYanni and Kelemen’s pairing ofan intentional choice with a nonfunctional tool. The findings of DiYanni and Kelemen suggest thatchildren discriminated which tool was functional and privileged this functionality over the intention-ality of the demonstrator.

If children in the current study privilege causal relevance over intention, they should choose not toimitate the first action when it is rendered irrelevant even if it was intentionally demonstrated, andthey should not hesitate to imitate the first action when it remains relevant even if it was demon-strated accidentally. In contrast, if children use intentionality to guide their causal understanding ofrelevant actions, they should imitate an intentionally demonstrated action even if it is rendered irrel-evant. If an action is demonstrated accidentally but remains relevant, children will need to perform itto succeed on the task, so it is expected that they will perform the action in this condition. However, if

Table 1Research design.

Demonstration: Intentionalityof first action

Child’s turn: Relevancy offirst action

Condition Consistency of intentionalityand causality

Intentional Relevant Intentional RelevancyRetained

Consistent

Irrelevant Intentional RelevancyRemoved

Inconsistent

Accidental Relevant Accidental RelevancyRetained

Inconsistent

Irrelevant Accidental RelevancyRemoved

Consistent

No demonstration Relevant Control RelevancyRetained

n/a

Irrelevant Control RelevancyRemoved

n/a


their understanding of object function is based on the intentionality of the demonstrator, they shouldbelieve that the accidentally demonstrated action is irrelevant, which would need to be revised whenthey were presented with the task and this action is relevant. If a revision of causal understanding isnecessary, children may show increased latency to perform the first action in this condition. In twoconditions where the intentionality of the demonstration was consistent with the relevancy of the firstaction for the child’s turn, it was expected that children would follow the intentions of the demonstra-tor, imitating the first action when it is demonstrated intentionally and remained relevant and ignor-ing the first action when it is demonstrated accidentally and rendered irrelevant.

Study 1

Method

ParticipantsParticipants were 58 children (29 girls and 29 boys, age range = 37–71 months): 18 3-year-olds

(M = 41.11 months, SD = 2.56), 21 4-year-olds (M = 53.14 months, SD = 3.07), and 19 5-year-olds(M = 62.53 months, SD = 2.74). Most children were White and from middle and upper classbackgrounds.

MaterialsSix different apparatuses were used, each containing a series of transparent plastic compartments

and three moving parts. A small stuffed animal could be maneuvered through and retrieved from eachapparatus by manipulating the moving parts. The toy could be placed in either the first compartmentor the second compartment to begin the sequence of toy movement. Placing the toy in the secondcompartment rendered the first moving part irrelevant to retrieval. If the toy began in the second com-partment, the first moving part could still be fully manipulated. Apparatuses for Study 1 are picturedin Appendix A.

ProcedureAll apparatuses were kept behind a barrier with an assistant, who passed each apparatus through a

door in the barrier as it was needed. A within-participants design was used, with all children partic-ipating in four demonstration conditions and two control conditions. For the demonstrations, the toybegan in the first compartment. In the demonstration conditions, the assistant passed an apparatus tothe demonstrator, who placed it in front of herself and said, ‘‘Look at this.’’ She then pointed to the toyto highlight its position and said, ‘‘See the [animal]? I’m going to get the [animal] out.’’ She performedthe first action, followed by the second action, followed by the third action, and retrieved the toy.


During each demonstration, the intentionality of the actions was manipulated using verbal mark-ers. Whereas the second action and third action were always marked as intentional (‘‘There!’’), thefirst action was differentially marked as intentional or accidental. In two of the demonstrations thefirst action was marked as intentional, indicated by the word ‘‘There!’’, and in the other two demon-strations the first action was marked as accidental, indicated by the phrase, ‘‘Whoops! I didn’t mean todo that!’’

For the child’s turn, the demonstrator passed the apparatus back to the assistant, who reset theapparatus behind the barrier. In resetting each apparatus, the first action either remained relevant,by placing the toy in the first compartment, or was rendered irrelevant, by placing the toy in the sec-ond compartment. This varied with whether the first action was demonstrated intentionally or acci-dentally, creating four conditions (see Table 1). There was consistency between intentionality andcausality in the Intentional Relevancy Retained condition, where the first action was performed inten-tionally and remained relevant for the child’s turn, and in the Accidental Relevancy Removed condition,where the first action was performed accidentally and rendered irrelevant for the child’s turn. Therewas inconsistency between intentionality and causality in the Intentional Relevancy Removed condi-tion, where the first action was performed intentionally and rendered irrelevant for the child’s turn,and in the Accidental Relevancy Retained condition, where the first action was performed accidentallyand remained relevant for the child’s turn. After the apparatus was reset out of sight, the experimenterpointed to the toy to highlight its position (‘‘See the [animal]?’’) and then placed the apparatus in frontof the child, who was encouraged to retrieve the toy (‘‘Now it’s your turn. Can you get the [animal]out?’’).

In the two control conditions, children were presented with an apparatus and encouraged to re-trieve the toy on their own without prior demonstration. In the Control Relevancy Retained condition,the toy was placed in the first compartment and all three actions were required for retrieval. In theControl Relevancy Removed condition, the toy was placed in the second compartment and the secondand third actions were required for retrieval.

Each toy retrieval attempt ended when the child was successful or after 3 min had elapsed. Thestructure of three apparatuses (lever shuttle, stair step tube, and elevator) allowed for the toy to be-come stuck if the moving parts were manipulated out of order. If this happened, the experimenterwaited 10 s to give the child a chance to realize that the toy was irretrievable before saying, ‘‘Uhoh! Looks like the animal is stuck! Let’s go on to the next one.’’ This was an infrequent occurrence.

Each condition occurred with each apparatus approximately equally often across participants. Theorder was counterbalanced so that the three different kinds of conditions (intentional, accidental, andcontrol) alternated, with no two conditions of the same kind ever occurring in direct sequence. Inaddition, the relevancy of the first action for the child’s turn was counterbalanced so that it alternatedbetween retained and removed across the six conditions. All sessions were videotaped.

Results

Task performanceThe majority of children retrieved the toy in all conditions. A comparison of success rates revealed

that there were significant differences between the Relevancy Retained conditions, Cochran Q,v2(58) = 9.65, p < .01. The success rate in the Control Relevancy Retained condition (69.0%) was signif-icantly lower than success rates in the Intentional (87.9%) and Accidental (86.2%) Relevancy Retainedconditions (McNemar tests, all ps < .05). The rate of success in the Control Relevancy Removed condi-tion (79.3%) was lower, but not significantly different, than success rates in the Intentional (86.2%) andAccidental (87.9%) Relevancy Removed conditions. To further investigate the effect of viewing a dem-onstration on children’s performance, the efficiency with which children completed the tasks wascompared across conditions. Efficiency was measured by summing the number of times each childmanipulated the apparatus in each condition. A manipulation was defined as an instance of contactmade with an apparatus, including touches and movements of moving parts and exploratory toucheson nonmoving parts. One analysis compared the three conditions in which relevancy of the first partwas retained for the child’s turn, and another compared the three conditions in which relevancy wasremoved. Results show that the demonstrations facilitated children’s performance. For the Relevancy


Retained conditions, children performed significantly more manipulations in the Control condition(M = 13.88, SD = 17.05) than in the Intentional (M = 5.93, SD = 4.93) and Accidental (M = 6.84,SD = 7.20) conditions, F(2,114) = 9.47, p < .001, g2 = .17 (Bonferroni post hoc tests, all ps < .01). Forthe Relevancy Removed conditions, children performed significantly more manipulations in the Con-trol condition (M = 11.40, SD = 11.55) than in the Accidental condition (M = 6.84, SD = 9.54),F(2,114) = 4.20, p < .05, g2 = .07 (Bonferroni post hoc test, p < .05). Children performed more manipu-lations in the Control condition than in the Intentional condition (M = 8.33, SD = 8.76), but the differ-ence was not significant. This pattern of findings suggests that even though most children successfullyachieved the toy retrieval goal in all conditions, they were engaged in independent exploratory learn-ing to a significantly higher degree when they were not provided with a demonstration. When a dem-onstration was available, children learned from their observations, which allowed greater efficiency intask completion.

Scoring of actionsChildren received a first action score, a second action score, and a third action score for each condition

that reflected the degree to which they manipulated the respective moving parts of each apparatus.The 0-to-4 scale from Gardiner and colleagues (2011) was used. This scale is related to the precisionwith which children’s manipulations replicated the experimenter’s actions. A score of 0 reflected mak-ing no contact with a part, a score of 1 reflected touching but not moving a part, a score of 2 reflectedmoving a part partially (less than demonstrated), a score of 3 reflected moving a part fully (exactly asdemonstrated), and a score of 4 reflected ‘‘overmoving’’ a part (in excess of the demonstration). Therewas one exception to this coding procedure. Whereas the experimenter did not reset any of the mov-ing parts after she manipulated them during the demonstration, during their attempts at toy retrievalmany of the children moved parts back to their original positions after using them to maneuver toysthrough the compartments of the apparatuses. For instance, if a trap door needed to be pulled out todrop a toy into a subsequent compartment, children would often push the trap door back in after thetoy was released before moving on to the next part. In most cases, this did not appear to be an attemptto elicit a further effect from the moving part; children just seemed to have a simple desire to reset theparts. Many children even reset the third part after retrieving the toy. Thus, resetting a moving partwas not coded as overmoving the part but was coded as fully moving a part. If children further manip-ulated a part after resetting it, this was scored as overmoving the part. Scores were assessed from vi-deo footage by two independent trained coders. The first coder then scored 30% of the second coder’svideos to determine interrater reliability. The intraclass correlation coefficient was .92.

Analysis of action scoresTo understand how the intentionality of the demonstrator and the relevancy of the actions affected

children’s performance, each action score were assessed in a separate mixed analysis of variance (AN-OVA) with age as a between-participants factor and with intentionality (i.e., of the first action duringthe demonstration: intentional, accidental, or no demonstration control) and relevancy (i.e., of the firstaction for the child’s turn: retained or removed) as within-participants factors. A preliminary analysisrevealed no effects of gender, and this factor was not included in subsequent analyses. Bonferroni testswere used for all post hoc comparisons.

Table 2 shows means and standard deviations for the first, second, and third action scores across allparticipants in each of the six conditions. Analysis of the first action score revealed a main effect ofrelevancy, F(1,55) = 60.58, p < .001, g2 = .52. Regardless of whether there was no demonstration inthe Control conditions or demonstration of the first action was intentional or accidental, childrenhad higher first action scores when relevancy was retained for their turn (M = 3.16, SD = 0.37) thanwhen relevancy was removed (M = 2.06, SD = 1.12) (see Fig. 1). There was no effect of age in analysisof the first action. Analysis of the second and third action scores revealed no significant effects.

It was expected that the first action score would reflect fully manipulating the first part in all Rel-evancy Retained conditions because this part needed to be manipulated to retrieve the toy. Therefore,this score is not informative as to whether children were using the demonstrator’s intentions in theAccidental Relevancy Retained condition. In this condition, if children had an understanding thatthe first action was irrelevant because it was performed accidentally, they might show increased

Table 2Means and standard deviations for action scores: Study 1.

Condition First action Second action Third action

M SD M SD M SD

Relevancy RetainedIntentional 3.05 0.69 3.07 0.45 2.95 0.76Accidental 3.10 0.52 2.98 0.69 3.10 0.52Control 3.33 0.60 3.02 0.85 2.72 1.28

Relevancy RemovedIntentional 2.05 1.77 3.00 0.53 3.10 0.64Accidental 1.88 1.68 2.93 0.59 2.95 0.69Control 2.24 1.73 2.74 0.83 2.93 0.99

fs

Fig. 1. First action scores for all conditions (Study 1), grouped by relevancy (retained or removed). Bars represent standarderrors.


latency to perform the first action as they revise their causal understanding of the action. Because only15 children performed the first action in all conditions, planned comparisons were conducted to com-pare the latency to perform the first action between conditions. Latency scores of zero, for childrenwho did not perform the first action, were excluded from each comparison. No significant differenceswere found (all ts < 2.05, all ps > .05).

Discussion

Study 1 examined the role of intentionality in children’s imitation of relevant actions using objectsconstructed of transparent materials. In contrast to findings from Gardiner and colleagues (2011), inwhich intentionality guided children’s imitation of causally irrelevant actions, causality predominatedover intentionality in children’s performance of relevant actions. The reversal of the relationship be-tween intentionality and relevancy suggests that the intentionality framework used to explain chil-dren’s imitation of irrelevant actions in Gardiner and colleagues’ study cannot be extended todescribe how children interpret the necessity of relevant actions.

The disparity in findings suggests that children process relevant and irrelevant actions differentlyduring observational learning of object use. An explanation for this may be provided by the ideas ofCsibra and Gergely (2006, 2011; see also Gergely & Csibra, 2005). In their pedagogical framework,


these authors describe potential evolved cognitive mechanisms underlying the social learning of tooluse. They suggest that because complex human tools have multiple possible functions, observabletool-use behavior is cognitively opaque. In other words, a learner knows neither the background knowl-edge of a demonstrator nor the ultimate goal (i.e., desired tool function) of his or her actions. Even if agoal is described or inferred, the action sequence to achieve it remains unknown. Given cognitiveopacity, tool-use learning mechanisms would have evolved a provision to process observable tool-use behavior while still providing an accurate and efficient means of cultural transmission. Csibraand Gergely (2006) described this provision as an assumption of relevance; learners assume that dem-onstrators provide a physically rational approach to achieving a goal and, therefore, that all actionsperformed are relevant. This assumption is apparent in children’s indiscriminate imitation; irrelevantactions, known by the demonstrator to be causally unnecessary, are cognitively opaque to the learner,who applies the assumption of relevance and imitates the irrelevant actions. The findings of Gardinerand colleagues (2011) support a qualifying intentionality component for irrelevant actions; childrenassume that all intentional irrelevant actions are relevant. In addition, a corresponding assumptionof irrelevance is applied to accidental irrelevant actions.

It does not appear that children in Study 1 used the demonstrator’s intentionality to apply assump-tions of relevance or irrelevance. The predominance of actual causal relevancy in Study 1 is consistentwith the pattern of results found by DiYanni and Kelemen (2008), in which the majority of childrendiscounted the intentionality of a demonstrator in favor of functional properties when making tool-use choices. One explanation for children privileging physical causality in these studies relates to acomponent of Csibra and Gergely’s pedagogical framework; learners may bring to the observationallearning context preexisting causal knowledge, which they can apply to their understanding of howa tool or an object works, thereby constraining their perceptions of a demonstrator’s actions (Csibra& Gergely, 2006). This idea has also been asserted by other researchers (Bonawitz et al., 2011; Buchs-baum, Gopnik, Griffiths, & Shafto, 2011; Goodman, Baker, & Tenenbaum, 2009). In Study 1, the dem-onstrated actions had physical effects that were readily observable (e.g., when a support was removedfrom beneath the toy, the toy fell). These obvious physical effects appear to have made these actionscognitively transparent; how they causally contribute to making the object work was evident andunambiguous. Thus, understanding of the causal structure of the objects is based on children’s strongprior understanding of physical causation and, therefore, resistant to other elements of the demon-stration such as the intentionality of the demonstrator. Indeed, DiYanni and Kelemen (2008) providedchildren with a familiarization period in which they interacted with the materials used in the subse-quent demonstrations. During this exploration, children may have discovered through independentexperiential learning certain functional properties of the tools that allowed them to easily recognizethat the nonfunctional tool would not work to complete the demonstrated task, overwhelming mosteffects of intentionality.

If children were learning from the demonstrations but interpreting what they observed using theirown causal reasoning, the social learning strategy that children employed in Study 1 might not be bestdescribed as imitation. Rather, their performance may reflect learning through goal emulation, inwhich a learner adopts a demonstrator’s intended goal but determines an action sequence indepen-dently (see Call & Carpenter, 2002, for delineation of social learning strategies based on differentsources of information from which learners can draw).

The proposal that actions are cognitively transparent when they have observable physical effectsand, thus, are causally unambiguous suggests that irrelevant actions are cognitively opaque becauseof the causal ambiguity created by a lack of readily observable physical effects. In other words, thatirrelevant actions do not physically contribute to making an unfamiliar object function might notbe apparent to a naive learner. With an expectation of a rational demonstrator who will provide a cor-rect display of object use (Buchsbaum et al., 2011; Goodman et al., 2009; Shafto & Goodman, 2008),the learner defers to the demonstrator rather than relying on his or her own causal reasoning and ap-plies an assumption of relevance. This leads to the incorporation of irrelevant actions into the learner’sunderstanding of object causal structure. Given the rarity of irrelevant actions in real-life object-usedemonstrations, such a stance would typically result in learning all of the actions necessary to makingan object function and no unnecessary actions.


If lack of observable physical effects makes irrelevant actions cognitively opaque, this suggests thatif the physical effects of relevant actions are unobservable, they should be rendered cognitively opaqueand subject to the same processing biases as irrelevant actions are in an observational learning con-text. With this logic in mind, the physical effects of the first moving part of each apparatus wereshrouded for Study 2. This created literal opacity, which for relevant actions should translate to cog-nitive opacity and lead children to incorporate the intentionality of the demonstrator into their under-standing of which actions are necessary. The causal ambiguity of opaque relevant actions may makethem the closest real-life equivalent to irrelevant actions. Thus, Study 2 may provide valuable insightinto how our knowledge of irrelevant action imitation can be applied to children’s everyday learningof objects.

Study 2

Method

ParticipantsParticipants were 58 children (26 girls and 32 boys, age range = 36–71 months), none of whom

took part in the first study: 18 3-year-olds (M = 41.61 months, SD = 3.79), 20 4-year-olds(M = 53.90 months, SD = 3.16), and 20 5-year-olds (M = 65.35 months, SD = 3.44). Most children wereWhite and from middle and upper class backgrounds.

MaterialsThe six apparatuses from Study 1 were used, with the effect of the first action made opaque by

shrouding the first and second compartments of each apparatus with black shelf liner paper that ad-hered to the plastic. Apparatuses for Study 2 are pictured in Appendix B. Slight modifications weremade to four of the apparatuses (lever shuttle, chute, elevator, and stair step tube) to make particularmoving parts more manageable for children to manipulate. These modifications did not affect therange of motion of the moving parts or the effects they had on maneuvering or retrieving the toy.

ProcedureThe procedure from Study 1 was used with two modifications related to how the toys were pre-

sented to the children. To maintain opacity in Study 2, the demonstrator began each condition by pre-senting a toy to the child and saying, ‘‘See the [animal]? I’m going to give the [animal] to [assistant]and she’s going to put it inside the game for us.’’ The demonstrator then passed the toy behind thebarrier to the assistant, who placed the toy within an apparatus and then passed it to the demonstra-tor. The demonstrator placed it in front of herself and said, ‘‘Now the [animal] is inside of here. I’mgoing to get the [animal] out,’’ and proceeded as in Study 1. For the child’s turn in the demonstrationconditions, on receiving a reset apparatus from the assistant, the demonstrator placed it in front of thechild as she said, ‘‘Now the animal is back inside of here. It’s your turn. Can you get the [animal] out?’’,and the procedure continued as in Study 1. The Control conditions began in the same manner as theother conditions, with the demonstrator presenting a toy to the child and passing it to the assistant forplacement in an apparatus. On receiving the apparatus, the demonstrator placed it in front of the childand provided the same instruction for the child’s turn as in the other conditions. All sessions werevideotaped.

Results

Task performanceThe majority of children retrieved the toy in all conditions except the Control Relevancy Retained

condition. Comparison of success rates revealed that there were significant differences between theRelevancy Retained conditions, Cochran Q, v2(58) = 11.72, p < .01, and between the Relevancy Re-moved conditions, Cochran Q, v2(58) = 12.08, p < .01. The success rate in the Control Relevancy Re-tained condition (46.6%) was significantly lower than that in the Intentional Relevancy Retained


condition (77.6%) (McNemar test, p < .01) and lower than, but not significantly different from, that inthe Accidental Relevancy Retained condition (56.9%). The rate of success in the Control Relevancy Re-moved condition (68.4%) was significantly lower than success rates in the Intentional (89.7%) andAccidental (87.9%) Relevancy Removed conditions (McNemar tests, all ps 6 .01). As in Study 1, to fur-ther investigate the effect of viewing a demonstration on children’s performance, the efficiency withwhich children completed the tasks was compared across conditions. The approach to measurementand analysis from Study 1 was used. For the Relevancy Retained conditions, children performed sig-nificantly more manipulations in the Control condition (M = 15.72, SD = 15.31) than in the Intentionalcondition (M = 7.19, SD = 6.99), F(2,114) = 8.72, p < .001, g2 = .13 (Bonferroni post hoc test, p = .001).Children performed more manipulations in the Control condition than in the Accidental condition(M = 10.86, SD = 8.87), but the difference was not significant. For the Relevancy Removed conditions,children performed significantly more manipulations in the Control condition (M = 12.45, SD = 10.35)than in the Intentional (M = 5.57, SD = 5.49) and Accidental (M = 4.98, SD = 5.22) conditions,F(2,114) = 18.94, p < .001, g2 = .25 (Bonferroni post hoc tests, all ps < .001). As in Study 1, this patternof findings suggests that even though most children successfully achieved the toy retrieval goal inmost conditions, they were engaged in independent exploratory learning to a significantly higher de-gree when they were not provided with a demonstration. When a demonstration was available, chil-dren learned from their observations, which allowed greater efficiency in task completion.

Scoring of actionsUsing the scoring procedure from Study 1, for each condition children received a first action score, a

second action score, and a third action score that reflected the degree to which they manipulated therespective moving parts of each apparatus. Scores were assessed from video footage by a trained co-der. A second trained coder then scored 30% of the first coder’s videos to determine interrater reliabil-ity. The intraclass correlation coefficient was .99.

Analysis of action scoresTo understand how the intentionality of the demonstrator and the relevancy of the actions affected

children’s performance, each action score was assessed in a separate mixed ANOVA with age as a be-tween-participants factor and with intentionality (i.e., of the first action during the demonstration:intentional, accidental, or no demonstration control) and relevancy (i.e., of the first action for thechild’s turn: retained or removed) as within-participants factors. A preliminary analysis revealed noeffects of gender, and this factor was not included in subsequent analyses. Bonferroni tests were usedfor all post hoc comparisons.

Table 3 shows means and standard deviations for the first, second, and third action scores in each ofthe six conditions. Analysis of first action scores (see Fig. 2) revealed main effects for intentionality,F(2,110) = 24.52, p < .001, g2 = .31, and relevancy, F(1,55) = 27.11, p < .001, g2 = .33, as well as an inter-action between intentionality and relevancy, F(2,110) = 6.45, p < .01, g2 = .11. Children had lowerscores when demonstration of the first action was accidental (M = 2.08, SD = 1.08) than when demon-stration was intentional (M = 3.05, SD = 0.46) or there was no demonstration in the Control conditions(M = 3.04, SD = 1.04). Children had lower scores when relevancy of the first action was removed fortheir turn (M = 2.45, SD = 0.68) than when relevancy was retained (M = 2.99, SD = 0.71). To analyzethe interaction between intentionality and relevancy, t tests compared first action scores betweenthe Relevancy Retained and Relevancy Removed conditions for each level of intentionality. Whendemonstration of the first action was intentional or there was no demonstration in the Control condi-tions, Relevancy Retained and Relevancy Removed scores were not significantly different (bothps > .10). When demonstration of the first action was accidental, Relevancy Retained scores(M = 2.64, SD = 1.35) were higher than Relevancy Removed scores (M = 1.50, SD = 1.51), t(57) = 4.51,p < .001, d = 0.80. Because children needed to perform the first action to retrieve the toy in the Acci-dental Relevancy Retained condition, this difference was expected.

Analysis of second action scores revealed main effects for age, F(2,55) = 5.94, p < .01, g2 = .18, inten-tionality, F(2,110) = 8.91, p < .001, g2 = .14, and relevancy, F(1,55) = 4.00, p = .05, g2 = .07, as well as aninteraction between intentionality and relevancy, F(2,110) = 6.10, p < .01, g2 = .10. The 3-year-olds(M = 2.86, SD = 0.35) had significantly lower second action scores than the 4-year-olds (M = 3.13,

Table 3Means and standard deviations for action scores: Study 2.

Condition First action Second action Third action

M SD M SD M SD

Relevancy RetainedIntentional 3.10 0.52 3.07 0.45 3.07 0.49Accidental 2.64 1.35 3.52 0.50 3.03 1.11Control 3.21 1.15 2.79 1.23 2.33 1.47

Relevancy RemovedIntentional 2.98 0.71 3.00 0.46 3.03 0.49Accidental 1.52 1.51 3.00 0.32 2.97 0.56Control 2.84 1.51 2.97 0.94 2.67 1.21

fs

Fig. 2. First action scores for all conditions (Study 2), grouped by relevancy (retained or removed). Bars represent standarderrors.


SD = 0.28) and 5-year-olds (M = 3.16, SD = 0.24). In a pattern opposite that of first action scores, chil-dren had higher second action scores when demonstration of the first action was accidental (M = 3.26,SD = 0.27) than when demonstration was intentional (M = 3.03, SD = 0.32) or there was no demonstra-tion in the Control conditions (M = 2.88, SD = 0.80). Children had lower second action scores when rel-evancy of the first action was removed for their turn (M = 2.98, SD = 0.35) than when relevancy wasretained (M = 3.12, SD = 0.46). To analyze the interaction between intentionality and relevancy, t testscompared second action scores between the Relevancy Retained and Relevancy Removed conditionsfor each level of intentionality. The pattern of findings for second action scores mirrored that for firstaction scores. When demonstration of the first action was intentional or there was no demonstrationin the Control conditions, Relevancy Retained and Relevancy Removed second action scores were notsignificantly different (both ps > .30). When demonstration of the first action was accidental, Rele-vancy Retained scores (M = 3.52, SD = 0.50) were higher than Relevancy Removed scores (M = 3.00,SD = 0.32), t(57) = 6.01, p < .001, d = 0.02.

Analysis of third action scores revealed a main effect of intentionality, F(2,110) = 16.49, p < .001,g2 = .23. Children had lower third action scores when there was no demonstration in the Control con-ditions (M = 2.50, SD = 0.91) than when there was a demonstration regardless of whether demonstra-tion of the first action was intentional (M = 3.05, SD = 0.32) or accidental (M = 3.00, SD = 0.62).

Returning to results for first action scores, there was a consistent pattern across the two Intentionalconditions: Regardless of whether the first action was retained or removed for their turn, when it was


demonstrated intentionally children performed it. This suggests that children were taking the demon-strator’s intentionality into account in understanding which actions should be performed to retrievethe toys from within the apparatuses, and intentional actions were considered causally meaningful.Scores in the Accidental conditions were lower than those in the Intentional conditions when col-lapsed across relevancy, but action scores are less informative for how children might be taking inten-tionality into account in the Accidental conditions. This is particularly the case in the AccidentalRelevancy Retained condition because children needed to perform the first action to retrieve the toy.

Analysis of order of manipulationTo better understand the relationship between intentionality and children’s understanding of cau-

sality, particularly in the Accidental conditions, the next set of analyses focused on the order in whichchildren manipulated the moving parts, with particular emphasis on which moving part was manip-ulated initially. If children believe that the first action is necessary, they should imitate the demonstra-tor and begin their toy retrieval attempts by initially manipulating the first moving part. In contrast, ifchildren do not believe that the first action is necessary, they should not imitate the first action andshould begin their toy retrieval attempts by manipulating the second moving part. If their understand-ing of the causal necessity of the first action is based on the demonstrator’s intentionality, with inten-tional actions understood as causally necessary and accidental actions understood as causallyunnecessary, children should initially manipulate the first part in the Intentional conditions and thesecond part in the Accidental conditions.

In the following analyses, the part children initially manipulated is considered the first moving partthat children made contact with in their interaction with an apparatus regardless of the degree towhich they manipulated this part. Mean action scores for the initially manipulated part fell primarilybetween 2.00 and 3.00, suggesting that children typically manipulated their first-chosen part to a sub-stantial degree. Cochran Q tests compared the rates of initially manipulating the first, second, andthird parts within each Intentional condition and each Accidental condition, and follow-up McNemarand exact binomial tests revealed specific differences between the rates of initial manipulation of thethree moving parts (see Fig. 3). Consistent with an understanding of causality based on demonstratorintentionality, in both Intentional conditions children were more likely to manipulate the first partthan the second or third part (both ps < .001). In addition, in the Accidental Relevancy Retained con-dition, children were more likely to manipulate the second part than the first or third part (bothps < .001). However, in the Accidental Relevancy Removed condition, children were equally likely tomanipulate the first and second parts (p = .09) and more likely to initially manipulate the second orfirst part than the third part (both ps < .05).

In the Accidental Relevancy Removed condition, correct manipulation of the second action wouldmove the toy into the transparent third compartment, and children could then manipulate the thirdaction to retrieve the toy. Therefore, it is surprising that children were not more likely to initiallymanipulate the second part than the first part in this condition, especially given that they appearedto follow the demonstrator’s intentions and attempt to omit the first action in the Accidental Rele-vancy Retained condition. However, an examination of the order in which children experienced thetwo Accidental conditions reveals why some children began their toy retrieval attempts by manipu-lating the first part when relevancy was removed. This analysis focused on children who initiallymanipulated either the first part or the second part in the Accidental conditions. In the Accidental Rel-evancy Retained condition, children who initially manipulated the first part and children who initiallymanipulated the second part were equally likely to do so regardless of whether they experienced thiscondition before or after the Accidental Relevancy Removed condition (Fisher’s exact tests, bothps > .50). In the Accidental Relevancy Removed condition, however, children who initially manipu-lated the first part were more likely to have experienced this condition before the Accidental Rele-vancy Retained condition (p = .001), whereas children who initially manipulated the second partwere more likely to have experienced this condition after the Accidental Relevancy Retained condition(p = .02) (Fisher’s exact tests) (see Fig. 4). This order effect suggests that children were learning fromtheir experience in the Accidental Relevancy Removed condition and applying the knowledge gainedto the Accidental Relevancy Retained condition. There are several possibilities for what children werelearning addressed in the Discussion below.

Fig. 3. Percentages of children whose initial manipulation was of the first, second, and third moving parts for the Intentionaland Accidental conditions (Study 2).

Fig. 4. Percentages of children whose initial manipulation was of the first and second moving parts for the Accidentalconditions, as a function of condition order (Study 2). Black bars represent children who experienced the Accidental RelevancyRetained condition prior to the Accidental Relevancy Removed condition; gray bars represent the opposite order.


Discussion

Study 2 investigated whether children would incorporate the intentionality of a demonstrator intotheir understanding of the causality of relevant actions that had unobservable physical effects due toopaque materials. It was hypothesized that opacity would make these actions causally ambiguous inthe same way that irrelevant actions are, thereby subjecting them to the intentionality-based process-ing observed in Gardiner and colleagues’ (2011) study. In contrast to Study 1 and consistent with thefindings of Gardiner and colleagues, children imitated the first action when it was performed inten-tionally and largely omitted or attempted to omit the first action when it was performed accidentallyregardless of whether it remained relevant or was rendered irrelevant for their turn. Findings fromStudy 2 fit within the framework advanced by Gardiner and colleagues that children incorporate


the intentionality of the demonstrator into their understanding of which observed actions are causallymeaningful.

In the Intentional conditions, results are straightforwardly supportive of intentionality-based cau-sal reasoning. In the Accidental conditions, findings required deeper exploration but also revealed per-formance consistent with the interpretation that children derived their understanding of objectfunction from the demonstrator’s intentions. In the Accidental Relevancy Retained condition, mostchildren performed the first action during their interaction with the object. However, they were mostlikely to begin manipulating the object by performing the second action, attempting to bypass theaccidentally demonstrated first action. This suggests that children’s initial impression of the object’scausal structure was based on the intentionality of the demonstrator—the accidental action was per-ceived as causally unnecessary—and they revised this understanding once they realized that the sec-ond action would not result in the expected effect of maneuvering the toy into the third compartment.

Children’s performance in the Accidental Relevancy Removed condition was affected by the orderin which they experienced the two accidental conditions. A detailed assessment of this effect reveals aspectacular relationship between an understanding of causality based on intentionality and indepen-dent learning. If children experienced the Accidental Relevancy Removed condition after the Acciden-tal Relevancy Retained condition, they were likely to begin with the first part even though it had beenperformed accidentally. It appears that children applied knowledge gained in the Accidental RelevancyRetained condition to their performance in the Accidental Relevancy Removed condition. The experi-ence that most children had in the Accidental Relevancy Retained condition was following the dem-onstrator’s intentions and attempting to omit the first action, realizing a revision of causal structurewas necessary, and then manipulating the first part. This suggests that during their interaction withthe object in the Accidental Relevancy Retained condition, children became aware of the inconsistencybetween the demonstrator’s intentions and the causal structure of the apparatus. There are at leasttwo possibilities for what knowledge children gained from this realization that they then applied inthe Accidental Relevancy Removed condition. One is that children were learning that three distinct ac-tions were necessary for making the objects function, working from the top downward or from right toleft. Another is that children were learning that the demonstrator’s accidental actions should not beregarded as mistakes and were actually necessary despite the accidental performance.

The likelihood of the first possibility can be investigated by looking at children’s performance in theControl conditions. If children were generalizing knowledge about the physical structure of the appa-ratuses, it would be expected that they would be able to determine how to retrieve the toys with rel-ative ease in Control conditions that followed the Accidental Relevancy Retained condition but wouldhave greater difficulty in Control conditions that preceded the Accidental Relevancy Retained condi-tion. This would be reflected in greater degrees of exact manipulation of the first part in Control con-ditions that followed the Accidental Relevancy Retained condition, because children would beexpected to manipulate this part fully but not beyond what was necessary, and greater degrees ofmanipulation in excess of the demonstration in Control conditions that preceded the Accidental Rel-evancy Retained condition, because children would be expected to engage in considerable explorationof the apparatus. To analyze this, the rate of exact manipulation of the first part (first action score = 3)was compared with the rate of excess manipulation (first action score = 4). The data show that chil-dren were more likely to engage in exact manipulation in Control conditions that occurred beforethe Accidental Relevancy Retained condition (McNemar test, p < .001), and there was no effect ofwhere a Control condition fell in relation to the Accidental Relevancy Retained condition on children’slikelihood to engage in excess manipulation (McNemar test, p = .44). This suggests that children wereapproaching each apparatus as an individual problem-solving task rather than generalizing the struc-ture of the apparatuses across conditions.

Thus, it is likely that children were taking the knowledge that the demonstrator’s accidental actionswere not unnecessary, gained during the Accidental Relevancy Retained condition, and applying it totheir interaction with the apparatus in the Accidental Relevancy Removed condition. However, chil-dren who experienced the Accidental Relevancy Removed condition prior to the Accidental RelevancyRetained condition did not have the experience of discovering inconsistency between the demonstra-tor’s accidental performance and the necessity of the first action in the Accidental Relevancy Retainedcondition. Without this prior experience, these children aligned themselves with the demonstrator’s


intentions in the Accidental Relevancy Removed condition, omitting the first action, beginning withthe second action, continuing to the third action, and retrieving the toy.

Turning briefly to the Control conditions, children demonstrated a high degree of manipulationof the first action, with a sizable proportion of children perseverating beyond what was necessary(53.4% in the Control Relevancy Retained condition and 50.0% in the Control Relevancy Removedcondition). Given the causal ambiguity created by the opaque materials, this is an expected resultand it suggests that many children were engaging in exploration of the apparatus. This interpreta-tion is consistent with findings that children engage in more exploration when they are presentedwith confounded causal evidence than when causal relationships are disambiguated (Schulz & Bon-awitz, 2007; Schulz, Gopnik, & Glymour, 2007). Thus, children appear to have a motivation to dis-ambiguate causal structure (Schulz & Bonawitz, 2007). Given their high degree of exploration,children in the current study certainly appeared motivated to understand how to make the objectsfunction in the Control conditions. This exploration frequently led to success, with nearly half ofchildren (46.6%) retrieving the toy in the Control Relevancy Retained condition and the majority(67.2%) retrieving the toy in the Control Relevancy Removed condition. That children can discovercausal structure during their own exploratory experience with objects is consistent with previousresearch (Schulz et al., 2007).

Relative to the exploration shown in the Control conditions, the close imitation of the demonstra-tor’s intentional behavior shown in the four demonstration conditions highlights how adept childrenare at gathering information about objects through observation, with particular attendance to a dem-onstrator’s intentions. Furthermore, this shows that when children are given the chance to observe ademonstration of an object with causally ambiguous physical structure, they will take advantage ofthe opportunity to garner knowledge about object function from someone else rather than attemptingto determine causality through independent trial-and-error learning.

General discussion

The current work provides an enriched picture of how children use intentional understanding tolearn about novel objects in an observational context. When relevant actions had clearly visible phys-ical effects, children’s imitation indicates that they chose to forgo the demonstrator’s intentionality asreflective of object function, instead relying on their own causal reasoning to determine which actionswere necessary. In contrast, when the effects of relevant actions were unobservable, children’s imita-tion indicates an understanding of causality that relied on the demonstrator’s intentions.

The predominance of relevancy over intentionality in Study 1 indicates that children enter theobservational learning with an understanding of causal relationships that can facilitate an accurateinterpretation of the functions of relevant components of an object. This is consistent with researchon children’s causal reasoning abilities showing that children are able to derive patterns of causeand effect from novel demonstrations (Buchsbaum et al., 2011; Schulz, Hooppell, & Jenkins, 2008).However, the predominance of intentionality over relevancy in Study 2, as well as the multitude ofstudies in which children have been shown to indiscriminately imitate irrelevant actions performedon novel objects (e.g., Lyons et al., 2011; McGuigan et al., 2007, 2010; Nielsen & Tomaselli, 2010), pro-vides clear evidence that children use their own causal reasoning abilities selectively in observationallearning contexts and will often choose instead to follow the example provided by a demonstrator.

The current studies provide insight as to what prompts children to rely on a demonstrator to learnabout the function of novel objects rather than relying on themselves. Taken together, findings suggesta contingency between the degree of cognitive opacity within the observational learning context andwhether children will apply their own understanding of causal relationships or defer to a demonstra-tor. This explanation is derived partially from the ideas of Csibra and Gergely (2006, 2011; see alsoGergely & Csibra, 2005), outlined earlier, regarding cognitive opacity. Recall that a tool-use scenariocan be described as cognitively opaque when the goal and/or intended action sequence of a demon-strator is unknown to a learner. With an expectation that a demonstrator provides a physically ra-tional display of object use, a learner applies an assumption of relevance and presumes that alldemonstrated actions are necessary.


The current work suggests that in the context of learning about novel objects through observation,some actions fall into the category of cognitive opacity, whereas others fall into a corresponding cat-egory of cognitive transparency in which there is little causal ambiguity and object function is obviousto a learner. Based on the current studies, one criterion for classification of relevant actions is whetherthey have observable physical effects; relevant actions that have observable physical effects can beclassified as cognitively transparent, whereas relevant actions that have unobservable physical effectscan be classified as cognitively opaque. When actions are cognitively transparent, as with the relevantactions in Study 1, children appear to discount the demonstrator’s intentionality as a source of infor-mation about object function. Rather, given the lack of causal ambiguity, children evaluate causalstructure independently. In contrast, the difference in children’s performance between the intentionaland accidental conditions in Study 2 suggests that the opacity of the objects created a degree of causalambiguity that compelled children to rely on the demonstrator to learn about the causal structure ofthe objects. In this context, children’s understanding of object function was guided by the demonstra-tor’s intentions, with an assumption of relevance applied to intentionally performed actions and anassumption of irrelevance applied to accidentally performed actions. In sum, there is an inverse rela-tionship between cognitive opacity and children’s reliance on their own causal reasoning abilities inobservational learning contexts; cognitively transparent actions facilitate children’s independent cau-sal reasoning, whereas cognitively opaque actions lead to reliance on a demonstrator.

As suggested earlier, irrelevant actions can be classified as cognitively opaque, and this cognitiveopacity appears to be due to a lack of observable physical effects. The categorization of irrelevant ac-tions as cognitively opaque, based on lack of observable physical effects, is supported by several find-ings. First, children’s performance in Study 2, which involved relevant actions with unobservablephysical effects, was similar to children’s performance in the study by Gardiner and colleagues(2011), where irrelevant moving parts had no effect on toys placed inside transparent plastic contain-ers. In both studies, children aligned themselves with the demonstrator’s intentions, imitating inten-tionally performed actions and omitting or attempting to omit accidentally performed actions. Thissuggests that children process opaque relevant and transparent irrelevant actions similarly, withthe actions having in common a lack of observable physical effects. Second, several studies have usedobjects constructed of transparent and opaque materials to compare children’s imitation of transpar-ent and opaque irrelevant actions (Horner & Whiten, 2005; McGuigan et al., 2007, 2010), finding thatchildren imitated irrelevant actions regardless of whether they could observe if the actions had an ef-fect. This suggests that it is not whether children have the opportunity to observe the lack of physicaleffects but rather the lack of physical effects in and of itself that renders irrelevant actions cognitivelyopaque.

If this characterization of irrelevant actions is correct, children may imitate these actions becausecausality is difficult to assess and they are relying on the demonstrator to gain an understanding ofcausal structure. Given the opportunity to learn about the function of a novel object by observing ademonstration, it is advantageous for children to defer to the demonstrator to understand causalitybecause social learning is more likely to facilitate efficient and accurate acquisition of object knowl-edge than is independent causal learning. To illustrate this point, consider a real-life instance of obser-vational learning of object use in which a knowledgeable individual provides a demonstration of howan object properly functions. In this realistic scenario, intentionally performed irrelevant actionswould, in all likelihood, be nonexistent. Thus, adopting a social learning strategy and relying on thedemonstrator would lead to an accurate understanding of causality. In contrast, presuming that thedemonstrator could not provide useful information in this causally ambiguous context and that inde-pendent determination of causality was required could ultimately prove to be a futile learning strat-egy. As suggested by Gardiner and colleagues (2011), imitating all intentionally performed actionswould have been selected for during human evolution as a means of efficiently acquiring importanttool-use knowledge rather than an assumption that one should determine causality independently.The latter strategy could be maladaptive if essential tool-use knowledge is never discovered and dis-appears from the culture.

At the same time, however, choosing to rely on oneself rather than a demonstrator can be beneficialin a situation where the demonstration appears to be unreliable. An ability to discern such a situationwas shown by children who experienced the Accidental Relevancy Removed condition after the Acci-


dental Relevancy Retained condition in Study 2. In the Accidental Relevancy Retained condition, chil-dren attempted to omit the first action but then learned that this accidentally demonstrated actionwas actually necessary. They subsequently performed the first action in the Accidental Relevancy Re-moved condition, suggesting that they had recognized that the demonstrator’s accidental cues werenot a reliable source of information about causality. Thus, given the apparent unreliability of the dem-onstrator in the Accidental Relevancy Retained condition, children chose not to rely on her in the Acci-dental Relevancy Removed condition. Instead, in the Accidental Relevancy Removed condition, theyrelied on the knowledge that they had gained from their own interaction with the apparatus in theprevious accidental condition and assumed that the first action was necessary.

One aim of this work was to describe how knowledge of indiscriminate imitation might be appliedto our understanding of how children learn about novel objects in everyday contexts where irrelevantactions are rare. Findings of the current research and previous studies (e.g., Gardiner et al., 2011;Horner & Whiten, 2005) may elucidate this relationship. Specifically, experimentally artificial irrele-vant actions may represent a parallel to real-life relevant opaque actions; both present causal ambi-guity to an object learner, and children imitate both with a high degree of precision. Thus, studies inwhich children indiscriminately imitate irrelevant actions may reveal a general proclivity for relyingon a demonstrator to learn about object function when causality is ambiguous. However, the currentstudies and that of Gardiner and colleagues (2011) reveal that such reliance is affected by character-istics of the demonstrator’s behavior such as intentionality.

In sum, the current research goes beyond imitation of irrelevant actions to reveal how children usea demonstrator’s intentionality to learn about relevant actions and suggests that causal ambiguity ingeneral may be an important factor that children use to evaluate whether the intentionality of a dem-onstrator’s actions is reflective of causality and can be used as a guide for imitation. The potential par-allel between irrelevant actions and opaque relevant actions suggests a possible expansion of thescope of impact that studies investigating indiscriminate imitation may have on our understandingof the ways in which children learn to use objects by observing others.

Acknowledgements

The second study was supported in part by the National Science Foundation under Grant Number0820080 and the Skidmore – Union Network (SUN) Committee, and Skidmore College.

Appendix A. Apparatuses with three-step manipulation sequences (transparent): Study 1

Lever ShuttleFirst action: topvertical leverSecond action: pushstickThird action: sidedoor

Double-Decker BoxFirst action: buttonSecond action: toptrap doorThird action: bottomtrap door

ChuteFirst action: topsliding vertical trapdoorSecond action: flipdown trap doorThird action: hingeddoor

PercolatorFirst action: flip downtrap doorSecond action: slidinghorizontal trap doorThird action: latchedtrap door


Stair Step TubeFirst action: rotatinghorizontal trap doorSecond action: pushstickThird action: latchedtrap door

ElevatorFirst action: hookedseesaw leverSecond action:elevatorThird action: topsliding push paddle

Appendix B. Apparatuses with three-step manipulation sequences (opaque): Study 2

Lever ShuttleFirst action: topvertical leverSecond action: pushstickThird action: pulldown trap door

Double-Decker BoxFirst action: buttonSecond action: topsliding trap doorThird action: bottomsliding trap door

ChuteFirst action: topsliding vertical trapdoorSecond action: flipdown trap doorThird action: hingeddoor

PercolatorFirst action: flip downtrap doorSecond action: slidinghorizontal trap doorThird action: latchedtrap door

Stair Step TubeFirst action: toprotating horizontaltrap doorSecond action: pushstickThird action: latchedtrap door

ElevatorFirst action:magneticallyconnected seesaw leverSecond action: elevatorThird action: topsliding push paddle

References

Bonawitz, E., Shafto, P., Gweon, H., Goodman, N. D., Spelke, E., & Schulz, L. (2011). The double-edged sword of pedagogy:Instruction limits spontaneous exploration and discovery. Cognition, 120, 322–330.

Brugger, A., Lariviere, L., Mumme, D., & Bushnell, E. (2007). Doing the right thing: Infants’ selection of actions to imitate fromobserved event sequences. Child Development, 78, 806–824.

Buchsbaum, D., Gopnik, A., Griffiths, T. L., & Shafto, P. (2011). Children’s imitation of causal action sequences is influenced bystatistical and pedagogical evidence. Cognition, 120, 331–340.

Call, J., & Carpenter, M. (2002). Three sources of information in social learning. In K. Dautenham & C. Nehaniv (Eds.), Imitation inanimals and artifacts (pp. 211–228). Cambridge, MA: MIT Press.

Csibra, G., & Gergely, G. (2011). Natural pedagogy as evolutionary adaptation. Philosophical Transactions of the Royal Society B,366, 1149–1157.

http://refhub.elsevier.com/S0022-0965(13)00215-4/h0005











Csibra, G., & Gergely, G. (2006). Social learning and social cognition: The case for pedagogy. In Y. Munakata & M. H. Johnson(Eds.), Processes of change in brain and cognitive development: Attention and performance XXI (pp. 249–274). Oxford, UK:Oxford University Press.

DiYanni, C., & Kelemen, D. (2008). Using a bad tool with good intention: Young children’s imitation of adults’ questionablechoices. Journal of Experimental Child Psychology, 101, 241–261.

Flynn, E. (2008). Investigating children as cultural magnets: Do young children transmit redundant information along diffusionchains? Philosophical Transactions of the Royal Society B, 363, 3541–3551.

Gardiner, A. K., Greif, M. L., & Bjorklund, D. F. (2011). Guided by intention: Preschoolers’ imitation reflects inferences ofcausation. Journal of Cognition and Development, 12, 355–373.

Gergely, G., Bekkering, H., & Kiraly, I. (2002). Rational imitation in preverbal infants. Nature, 415, 755.Gergely, G., & Csibra, G. (2005). The social construction of the cultural mind: Imitative learning as a mechanism of human

pedagogy. Interaction Studies, 6, 463–481.Goodman, N. D., Baker, C. L., & Tenenbaum, J. B. (2009). Cause and intent: Social reasoning in causal learning. In N. Taatgen & H.

van Ryn (Eds.), Proceedings of the 31st annual meeting of the Cognitive Science Society (pp. 2759–2764). Amsterdam: CognitiveScience Society.

Horner, V., & Whiten, A. (2005). Causal knowledge and imitation/emulation switching in chimpanzees (Pan troglodytes) andhuman children (Homo sapiens). Animal Cognition, 8, 164–181.

Kenward, B. (2012). Over-imitating preschoolers believe unnecessary actions are normative and enforce their performance by athird party. Journal of Experimental Child Psychology, 112, 195–207.

Lyons, D. E., Damrosch, D. H., Lin, J. K., Macris, D. M., & Keil, F. C. (2011). The scope and limits of overimitation in the transmissionof artefact culture. Philosophical Transactions of the Royal Society B, 366, 1158–1167.

McGuigan, N., Makinson, J., & Whiten, A. (2010). From over-imitation to super copying: Adults imitate causally irrelevantaspects of tool use with higher fidelity than young children. British Journal of Psychology, 102, 1–18.

McGuigan, N., Whiten, A., Flynn, E., & Horner, V. (2007). Imitation of causally opaque versus causally transparent tool use by 3-and 5-year-old children. Cognitive Development, 22, 353–364.

Nielsen, M. (2006). Copying actions and copying outcomes: Social learning through the second year. Developmental Psychology,42, 555–565.

Nielsen, M., Moore, C., & Mohamedally, J. (2012). Young children overimitate in third-party contexts. Journal of ExperimentalChild Psychology, 112, 75–83.

Nielsen, M., & Tomaselli, K. (2010). Over-imitation in Kalahari Bushman children and the origins of human cultural cognition.Psychological Science, 21, 729–736.

Schulz, L. E., & Bonawitz, E. B. (2007). Serious fun: Preschoolers engage in more exploratory play when evidence is confounded.Developmental Psychology, 43, 1045–1050.

Schulz, L. E., Gopnik, A., & Glymour, C. (2007). Preschool children learn about causal structure from conditional interventions.Developmental Science, 10, 322–332.

Schulz, L. E., Hooppell, C., & Jenkins, A. C. (2008). Judicious imitation: Children differentially imitate deterministically andprobabilistically effective actions. Child Development, 79, 395–410.

Schwier, C., van Maanen, C., Carpenter, M., & Tomasello, M. (2006). Rational imitation in 12-month-old infants. Infancy, 10,303–311.

Shafto, P., & Goodman, N. (2008). Teaching games: Statistical sampling assumptions for learning in pedagogical situations. In B.C. Love, K. McRae, & V. M. Sloutsky (Eds.), Proceedings of the 30th annual conference of the Cognitive Science Society (pp. 1632–1637). Washington, DC: Cognitive Science Society.

Simpson, A., & Riggs, K. J. (2011). Three- and 4-year-olds encode modeled actions in two ways leading to immediate imitationand delayed emulation. Developmental Psychology, 47, 834–840.







































beyond irrelevant actions: understanding the role of intentionality in children’s imitation of...

Documents