The role of the efficiency of novel actions in infants’ goal anticipation

Journal of Experimental Child Psychology 116 (2013) 415–427

Contents lists available at SciVerse ScienceDirect

Journal of Experimental Child Psychology journal homepage: www.elsevier.com/locate/jecp

The role of the efficiency of novel actions in infants’ goal anticipation Szilvia Biro ⇑ Centre for Child and Family Studies, Department of Cognitive Psychology, Leiden Institute for Brain and Cognition, Leiden University, 2333 AK Leiden, The Netherlands

a r t i c l e

i n f o

Article history: Available online 30 November 2012 Keywords: Infants Goal anticipation Action understanding Eye tracking Predictive eye movements Teleological inferences

a b s t r a c t In two experiments, we recorded infants’ eye movements to test whether the efficiency of the action influences infants’ ability to anticipate the outcome of an ongoing action performed by abstract figures. In Experiment 1, we found that predictive eye movements were elicited by both nonefficient and efficient actions, but anticipation of the outcome occurred much earlier in the efficient action condition. Experiment 2 was designed to test the effect of saliency of the goal and the possibility that automatic extrapolation of the movement was partly responsible for the predictive gaze shifts in Experiment 1. We found that when automatic extrapolation was prevented and the goal was not salient, infants showed predictive gaze shifts only in the efficient action condition. Taken together, our findings support the importance of teleological inferences in anticipating the goals of ongoing actions. Ó 2012 Elsevier Inc. All rights reserved.

Introduction The ability to anticipate the outcome of others’ ongoing actions allows us not only to attribute goals to observed actions but also to prepare ourselves to interact with or, if necessary, to counteract others before their actions are completed. Therefore, this capacity is undoubtedly useful, and its early emergence is crucial for survival as well as for social learning (Csibra & Gergely, 2007; Tomasello, Carpenter, Call, Behne, & Moll, 2005). Although several different mechanisms have been proposed to be responsible for the development of the ability to identify the likely outcome and goal of an action

⇑ Fax: +31 71 527 3783. E-mail address: [email protected] 0022-0965/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jecp.2012.09.011

416

S. Biro / Journal of Experimental Child Psychology 116 (2013) 415–427

(Gergely & Csibra, 2003; Hommel, Musseler, Aschersleben, & Prinz, 2001; Rizzolatti, Fogassi, & Gallese, 2001), there is no consensus about their relative importance. According to one of the theories on action interpretation, we use a teleological framework to attribute goals (Gergely & Csibra, 2003). Thus, when we observe an action, we seek to establish a relationship among three relevant aspects of current reality: the action, the outcome, and the physical constraints of the situation. The teleological stance theory argues that an observed action will be interpreted as goal-directed only if the action can be considered as the most efficient means to achieve the outcome under the current situational constraints. It is claimed that the application of this abstract interpretational schema requires neither extensive experience with actions nor the ability to perform them and that, therefore, it can be applied to interpret actions of a wide range of entities. Several studies have used the ‘‘violation of expectation paradigm’’ to test the application of the teleological framework during infancy. The looking time patterns found in these studies suggest that infants from around 6 to 9 months of age are able to rely on the efficiency1 of both human and novel nonhuman (robotic or self-propelled objects) actions to interpret the actions as goal-directed (Biro, Csibra, & Gergely, 2007; Biro, Verschoor, & Coenen 2011; Csibra, Gergely, Biro, Koós, & Brockbank, 1999; Csibra, 2008; Csibra et al., 2003; Gergely, Nádasdy, Csibra, & Biro, 1995; Hernik & Southgate, 2012; Kamewari, Kato, Kanda, Ishiguro, & Hiraki, 2005; Sodian, Schoeppner, & Metz, 2004; Southgate, Johnson, & Csibra, 2008; Verschoor & Biro, 2012; Woodward & Sommerville, 2000). Furthermore, the teleological stance theory claims that the teleological interpretational system also allows the generation of predictive inferences, including goal prediction. Thus, when we observe an ongoing incomplete action, we can rely on the assumption of efficiency to infer an outcome that can be justified under the given situational constraints. This hypothesized outcome can then be attributed as the goal of the action. Some studies suggest that from around 12 months of age, infants can make inferences about the unseen goals of incomplete actions. For example, during the observation of ‘‘chasing’’ events of abstract geometric figures or balls (Csibra et al., 2003; Southgate & Csibra, 2009; Wagner & Carey, 2005), infants were able to interpret the observed action as chasing and infer the outcome that the chaser would catch up. However, the infants were able to do so only if the observed ongoing action could be considered as an efficient means toward the hypothesized outcome given the situational constraints. Because the findings in all of these studies are based on duration of looking time measures, they might be open to alternative interpretations. Indeed, it has been argued that these measures do not necessarily reflect true predictions about the outcome; they could instead indicate retrospective inferences. Infants could, in principle, have evaluated the efficiency of the action when they were presented with the outcome during the test events rather than while they were watching the incomplete actions during the habituation/familiarization phase (Gredebäck & Melinder, 2010; Southgate, Johnson, Karoui, & Csibra, 2010). Another (more direct) measure that has recently been used to investigate adults’ and infants’ ability to anticipate the outcome of observed ongoing actions is predictive eye gaze (Cannon & Woodward, 2012; Eshuis, Coventry, & Vulchanova, 2009; Falck-Ytter, Gredebäck, & von Hofsten, 2006; Flanagan & Johansson, 2003; Gredebäck & Kochukhova, 2010; Gredebäck & Melinder, 2010; Gredebäck, Stasiewicz, Falck-Ytter, Rosander, & von Hofsten, 2009; Kanakogi & Itakura, 2011; Kochukhova & Gredebäck, 2010; Paulus et al., 2011). Falck-Ytter and colleagues (2006), for example, demonstrated that 12-month-olds, but not 6-month-olds, show predictive eye movements when they observe a human hand transferring balls to a basket; that is, they shift their gaze to the basket before the hand arrives and drops the ball into the basket. However, predictive eye movements were not found in this study when the balls moved to the basket by themselves without help from the hand. In this case, infants followed the movement, but their gaze did not arrive at the basket ahead of the moving balls. The authors argued that these findings suggest that predictive looking, and therefore goal anticipation of ongoing actions, is restricted to human actions and that infants need to be able to perform the actions themselves to anticipate the goals of similar actions by others. This argument supports the direct 1 In these studies, the evaluation of efficiency was based on various perceptual cues such as taking the shortest pathway, exerting the least effort, and taking only causally necessary steps. Therefore, it has been suggested that the efficiency of the action is applied as an abstract principle for representing actions during infancy (e.g., Csibra, Biro, Koós, & Gergely, 2003).

S. Biro / Journal of Experimental Child Psychology 116 (2013) 415–427

417

matching hypothesis of action interpretation, which claims that the starting mechanism for understanding others’ goal-directed behavior is based on mapping the observed actions onto one’s own motor representation of those actions (Falck-Ytter et al., 2006; Fogassi et al., 2005; Gallese & Goldman, 1998). However, more recent studies have disputed the claim that predictive looks are exclusive to human actions because predictive gazes were elicited from both adults and infants when they watched actions without the involvement of human movement (Eshuis et al., 2009; Paulus et al., 2011). On the other hand, because the actions in these studies were performed by toy or animated cartoon animals, the ability to anticipate the likely outcome of these actions may have been partly due to prior knowledge or expectations about the actions of the actors. Two recent studies tested the effect of the efficiency of the action on the presence of predictive eye movements. In Gredebäck and Melinder’s (2010) study, 12-month-olds observed human feeding scenarios in which a feeder either moved the food to the mouth of the partner (efficient action condition) or placed the food onto the back of the hand of the partner, who then ate it from her own hand (nonefficient action condition). Although infants showed predictive looks in both conditions (i.e., they fixated on the partner’s head or hand before the food arrived), in the efficient action condition their predictive looks occurred earlier (i.e., they anticipated faster) during the third repetition of the feeding action, suggesting that the efficiency of the action has some influence on the timing of predictive looks. Furthermore, differences in pupil dilation between the two conditions also indicate that infants’ evaluated the efficiency of the observed action. However, as the authors also pointed out, because the efficient action was also more familiar than the nonefficient action, these differences in the two conditions may have reflected infants’ reaction to the familiarity with the action rather than to the effect of efficiency. In Paulus and colleagues’ (2011) study, 9-month-olds’ predictive looks were measured to test whether infants are able to anticipate the most efficient, but novel, action of an agent (an animated cow) to achieve its goal (catching up with a sheep) after the situational constraints had changed. Infants were found to shift their gaze to an area indicating that they anticipated the repetition of the familiar, previously seen action rather than a novel but, in the new situation, more efficient route. However, because the sheep walked off from the screen before the cow started to move, infants never saw the goal state being attained. Therefore, it is possible that applying the principle of efficient action to anticipate a novel action under new situational constraints toward an unseen goal state may have been too challenging for 9-month-olds. Consistent with this possibility, previous looking time studies suggest that 12-month-olds, but not 9-month-olds, are able to make predictive inferences using the teleological schema, possibly because some background processes necessary for hypothesis formation are too weak in 9-month-olds (see Csibra et al., 2003). Therefore, the range of the types of actions, as well as the role of efficiency in goal anticipation measured by predictive gaze, is still uncertain. The current study aimed to investigate whether predictive gaze shifts are present and influenced by the efficiency of ongoing novel nonhuman actions. Infants were shown animations involving geometric figures in two conditions: an efficient action condition and a nonefficient action condition. In both conditions, a blue circle could be categorized as an agent on the basis of showing self-initiated and varied movements (Biro & Leslie, 2007; Luo & Baillargeon, 2005; Tremoulet & Feldman, 2000). In the efficient action condition, infants watched the blue circle approaching and making contact with a red circle by ‘‘jumping over’’ a rectangular figure located in between them (see Fig. 1). The nonefficient action condition was equivalent to the efficient action condition except that the rectangular figure was not present. Thus, jumping could not be considered as an efficient action toward the outcome (contacting the red circle). The animations were repeated six times, but they differed slightly from each other in terms of the size of the obstacle and the height of the blue circle’s jump, thereby providing additional support for the efficient adjustment of the blue circle’s action to changing situational constraints in the efficient action condition and for equifinally2 varying movement in the nonefficient action condition (see also the ‘‘Stimuli’’ section below). 2 It has been proposed that identifying the particular goal of an agent could be based on the equifinal structure of an action, which comes from the observation that the agent’s different actions result in one and the same consequence (Heider, 1958). However, it has been shown using looking time measures that without evidence for efficiency of the action, this cue is insufficient to interpret the observed action as goal-directed (Csibra et al., 1999; Gergely et al., 1995).

418

S. Biro / Journal of Experimental Child Psychology 116 (2013) 415–427

EFFICIENT ACTION Jump

Goal

Obstacle

Approach

Start

NON-EFFICIENT ACTION Jump

Goal

Approach

Obstacle Start

Fig. 1. Illustration of the animation with the areas of interest analyzed in Experiment 1.

If the movement characteristics of a novel entity are sufficient to categorize it as a goal-directed agent and to anticipate its goal (Baron-Cohen, 1994; Leslie, 1994; Premack, 1990), then predictive gazes should be elicited in both conditions. If, on the other hand, infants can infer the goal of an ongoing action only when it can be justified in the given environment, as predicted by the teleological stance theory, then predictive gazes should be found only in the efficient action condition. Finally, if infants can only anticipate the outcome of human actions, then predictive gazes should not be found in any of the two conditions. Looking time studies suggest that goal prediction for ongoing incomplete actions based on efficiency emerges at the end of the first year (Csibra et al., 2003). Therefore, we tested a group of 13-month-olds. Experiment 1 Methods Participants A total of 30 13-month-old infants (mean age = 397.2 days, SD = 18.44, range = 368–423, 14 boys and 16 girls) participated in the experiment. An additional 6 infants also participated but were excluded due to a lack of reliable gaze data (see criteria for inclusion in the ‘‘Data analysis’’ section below). Infants were randomly assigned to either the efficient action condition or the nonefficient action condition. They were recruited through direct mail to their families and received a gift after the experiment. Parents could opt to have their travel costs reimbursed. Stimuli Infants were presented with animated movies involving geometric figures (see Fig. 1).3 Infants in the efficient action condition saw a blue circle on the right side, a larger red circle on the left side, and a black rectangular figure in the middle of the screen (Introduction scene, 300 ms). The blue circle 3

The stimuli in the two conditions were similar, but not identical, to the habituation stimuli used in Gergely et al. (1995) study.

S. Biro / Journal of Experimental Child Psychology 116 (2013) 415–427

419

started to move along a straight path, stopped, moved back to its original position, and then started to move again (Beginning scene, 3400 ms). It then took a parabolic path over the rectangular figure (Jumping scene, 1300 ms). After landing, it moved toward the red circle (Approaching scene, 1000 ms) and stayed in contact with it (Arriving scene, 600 ms). The animation lasted 6600 ms and was repeated six times. The six animations were identical except that the size of the ‘‘jumping’’ movement of the blue circle varied according to the changing height of the rectangular figure (low, medium, or high). The order of the animations was the same for all participants: low–medium–high–low–high–medium. Infants in the nonefficient action condition saw the same animations except for the fact that the rectangular figure was absent. The movement of the blue circle was the same. Thus, it varied its parabolic path in the same manner and order as in the efficient action condition. Procedure and apparatus Infants sat in their caregivers’ laps in a curtained booth facing the Tobii T120x eye-tracker with an integrated 17-inch TFT monitor. The height of the chair and the position of the monitor were adjusted to establish a good eye-tracking status (so that infants’ eyes were 60 cm away from the monitor). Using ClearView 2.7.1 software (Tobii Technology), first a 5-point infant calibration procedure was carried out showing infants a blue–white pulsating circle with sound. A shaking duck image with sound was shown as an attention-getting stimulus when infants looked away during calibration. The presentation of the animations immediately followed the calibration. Caregivers were informed about the procedure and were instructed to close their eyes, not to talk, and to try to keep their infants from moving or leaning. After the experiment, the caregivers were shown the gaze replay of their infants and were told the rationale of the study. Data analysis Five areas of interest (AOIs) were defined: Start, Jump, Obstacle (adjusted to its changing height across trials), Approach, and Goal (see Fig. 1). Gaze was measured during all six animations. Data from a given animation were included if there were a minimum of three fixations: one fixation in the Start AOI during the Beginning scene, one fixation in the Jump AOI during the Jumping scene, and one fixation in the Goal AOI that needed to follow the previous two fixations. This criterion was used to ensure that looking at the Goal AOI reflected anticipation based on seeing at least the start of the jumping movement of the circle, which is presumably necessary to evaluate the efficiency of the action. A fixation filter with a 50-pixel fixation radius and a 200-ms minimum fixation duration was used (e.g., Gredebäck, Örnkloo, & von Hofsten, 2006). A participant was included in the analysis only if data from a minimum of three animations (trials) could be obtained. Gaze timing data. We calculated the difference between the time of infants’ fixation to the Goal AOI and the arrival time of the blue circle in the Goal AOI (time difference) in each accepted trial. Thus, predictive looks (positive time difference) and reactive looks (negative time difference) could be identified. For each participant, the average of the time differences of the accepted trials was also calculated. In addition, to investigate the effect of repetition of the animations while avoiding problems with missing data and slight variations between the movies, the six trials were collapsed into two blocks (i.e., the data from the first and last three trials were separately averaged, resulting in Blocks 1 and 2, respectively).4 Furthermore, for each participant, we calculated the percentages of trials with predictive and reactive looks and the distribution of predictive looks during the Jumping and Approaching scenes. Finally, the number of fixations to the Goal AOI during the Beginning scene was also counted. Looking time data. We computed the total looking time for each accepted trial. Furthermore, we also calculated the ratios of looking times at the Goal, Jump, and Obstacle AOIs relative to the total looking time as well as the looking times at the Jump AOI during the Jumping scene. (The looking time data 4 Note that if a given trial was not accepted, then the block was calculated from the remaining accepted trials. If all three trials belonging to a block were missing, then that block was absent.

420

S. Biro / Journal of Experimental Child Psychology 116 (2013) 415–427

were used to compare conditions as well as to test alternative explanations for differences between the conditions in the presence and timing of predictive gaze shifts.) Results There was no difference in the number of accepted trials, the number of overall fixations, and the mean total looking times between the two conditions, t(28) = 0.89, p = .37, t(28) = 0.86, p = .39, and t(28) = 0.44, p = .66, respectively. Gaze timing Using one-sample t tests, first the average time differences were tested against 0. We found that the time differences were significantly above 0 in both conditions: efficient action, t(15) = 5.84, p < .001; nonefficient action, t(13) = 3.93, p = .002 (see Fig. 2). This indicates that infants in both conditions showed predictive looking. To compare the two conditions, an independent-sample t test was carried out. It revealed a significant difference between the two conditions, t(29) = 2.09, p = .046, indicating that infants showed earlier gaze shifts in the efficient action condition than in the nonefficient action condition. A repeated-measures analysis of variance (ANOVA) was carried out with condition as a between-participants factor and block as within-participant variable, including only infants who had valid data in both blocks (13 in the nonefficient action condition and 14 in the efficient action condition). Although a significant condition effect, F(1, 25) = 6.05, p = .021, g2p ¼ :19, was again found, there was no interaction between condition and block. An independent-sample t test showed that infants produced predictive looks in a significantly higher percentage of trials in the efficient action condition (M = 86.25%) than in the nonefficient action condition (M = 55.71%), t(28) = 3.46, p = .002. Paired-sample t tests also revealed that there was a significantly higher percentage of trials with predictive looks than reactive looks in the efficient action condition, t(15) = 7.77, p < .001, whereas there was no difference between the two types of looks in the nonefficient action condition, t(13) = 0.73, p = .47. In the nonefficient action condition, a significantly higher percentage of predictive looks occurred during the Approaching scene (M = 80.23%) than during the Jumping scene (M = 19.76%), t(13) = 3.12, p = .008, whereas no difference was revealed in the efficient action condition, t(15) = 1.39, p = .18. Finally, during the Beginning scene, there were significantly more fixations to the Goal AOI in the efficient action condition (M = 2.68) than in the nonefficient action condition (M = 1.35), t(28) = 2.87, p = .008. Looking times Independent-sample t tests showed no difference between the two conditions in the average looking time ratios for the Goal AOI, t(28) = 0.33, p = .74, and for the Jump AOI, t(28) = 0.66, p = .51,

Fig. 2. Mean gaze arrival in the Goal AOI relative to the arrival of the blue circle in Experiment 1. Positive values imply that the gaze arrived before the circle. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

S. Biro / Journal of Experimental Child Psychology 116 (2013) 415–427

421

relative to the total looking time or in the average looking time for the Jump AOI during the Jumping scene, t(28) = 0.63, p = .63. Looking times for the Obstacle AOI were also analyzed. Even though the obstacle was not present in the nonefficient action condition, there was no significant difference between the two conditions in looking time for the Obstacle AOI, t(28) = 1.77, p = .087. More important, we found that only 1 infant in a single trial looked at the Goal AOI right after the infant looked at the Obstacle AOI. Finally, because the age range was relatively wide, the gaze timing and looking time analyses were also carried out using the age (in days) of the infants as a covariate whenever this was possible. All of the effects remained the same, suggesting that age had no effect on the fixations. Discussion of Experiment 1 We investigated whether the efficiency of an ongoing novel action influences the presence and timing of predictive gazes of 13-month-olds. We found that infants in both the efficient and nonefficient action conditions showed predictive eye movements; that is, they shifted their gaze to the outcome of the action before it was achieved. This finding suggests that infants’ predictive eye movements can be elicited by a nonhuman self-propelled agent that provides varying and changing movements. We also found that, overall, predictive gaze shifts occurred significantly earlier (i.e., infants were faster in anticipating the outcome) in the efficient action condition than in the nonefficient action condition. This finding implies that efficiency does have an influence on the anticipation of the outcome of an ongoing action. Our findings on the looking time analyses for the Jump and Obstacle AOIs exclude the possibility that the difference in the timing of predictive looks between the two conditions was due to perceptual differences between the two conditions, namely that the obstacle was only present in the efficient action condition. Thus, the obstacle in the efficient action condition did not function as a ‘‘jumping board’’ to quickly attract infants’ attention to the red circle. Furthermore, because the jumping action did not occupy infants’ attention more in the nonefficient action condition than in the efficient action condition, it cannot be argued that it was more interesting in the nonefficient action condition and, therefore, kept infants from looking ahead. It would seem that the finding that both conditions elicited predictive gaze does not support the strong claim of the teleological stance theory about the critical role of the efficiency of an ongoing action in anticipating its outcome. However, before we jump to this conclusion, two issues that may have helped to elicit predictive gaze shifts need to be considered. First, because the blue circle’s jump was followed by a straight-line approach of the red circle, it is possible that gaze shifts occurring during this final approach simply reflected automatic extrapolation of the straight-line movement rather than true goal anticipation (e.g., von Hofsten, Vishton, Spelke, Feng, & Rosander, 1998). We found that in the nonefficient action condition, more predictive looks occurred during the period when the blue circle was approaching the red circle along a straight line than during the period when the blue circle was jumping. Therefore, the automatic extrapolation of the straight-line movement may have been largely responsible for the predictive looks in the nonefficient action condition. The second issue concerns the role of the red circle in eliciting gaze shifts. We suggest that the saliency of the red circle may also have played a role in stimulating infants to look at the Goal AOI in both conditions. The presence of the red circle is also related to a theoretical question about the necessity of the presence of a well-defined visible goal object for goal anticipation. There is some indication that the type of goal does influence anticipatory gaze shifts. Gredebäck and colleagues (2009) showed that infants produced more anticipatory looks when objects were transferred into containers than when objects were simply displaced to unmarked locations. Eshuis and colleagues (2009) demonstrated that other outcome highlighting aspects, such as sound effects, can also influence the presence of predictive looks in adults. To address these issues, we conducted a second experiment with 13-month-olds in which the animation was changed in two respects, namely that the red circle was not present and we omitted the final straight-line approach. After the blue circle jumped, it landed and disappeared behind a black ‘‘wall’’ that had the same color as the ground and, thus, did not seem to be distinct from it (see Fig. 3). These modifications exclude the possibility of automatic straight-line extrapolation of the movement, and they reduce the saliency of the goal object/area considerably.

422

S. Biro / Journal of Experimental Child Psychology 116 (2013) 415–427

EFFICIENT ACTION Jump Obstacle Start

Goal

NON-EFFICIENT ACTION Jump Obstacle

Goal

Start

Fig. 3. Illustration of the animation with the areas of interest analyzed in Experiment 2.

Experiment 2 Methods Participants A total of 36 13-month-old infants (mean age = 386.1 days, SD = 14.2, range = 368–411, 19 girls and 17 boys) participated in the experiment. An additional 12 infants also participated but were excluded due to lack of reliable gaze data. Infants were randomly assigned to either the efficient action condition or the nonefficient action condition. They were recruited through direct mail to their families and received a gift after the experiment. Parents could opt to have their travel costs reimbursed. Stimuli The stimuli were similar to those in the previous experiment. However, there were two main differences. One difference was that instead of a red circle, a black rectangular figure that looked like an elevated ground was present on the left side of the screen. The second difference was that the small blue circle did not roll to the left side of the screen but rather disappeared behind the black figure after it had descended from its jump (see Fig. 3). The durations of the scenes were as follows: Introduction, 300 ms; Beginning, 3400 ms; Jumping, 1640 ms; and Arriving (behind the black figure), 1260 ms. The total duration of one animation was 6600 ms. The order of the six animations and the difference between the efficient and nonefficient action conditions were identical to those in Experiment 1. Procedure and apparatus These were the same as those in Experiment 1. Data analyses The same criteria were used for the inclusion of participants and trials, and the same measures were calculated from the raw data, as in Experiment 1.

S. Biro / Journal of Experimental Child Psychology 116 (2013) 415–427

423

Results There was no difference between the two conditions in the number of accepted trials, t(34) = 1.36, p = .18, the mean total looking times, t(34) = 1,15, p = .25, or the number of overall fixations, t(34) = 0.12, p = .90. Gaze timing Using one-sample t tests, the average time difference was tested against 0 for each condition. We found that the average time difference was significantly above 0 in the efficient action condition, t(17) = 2.83, p = .011, but not in the nonefficient action condition, t(17) = 0.55, p = .580 (see Fig. 4). An independent-sample t test showed no significant difference between the two conditions, t(34) = 1.06, p = .29. A repeated-measures ANOVA with condition as a between-participants factor and block as a within-participant variable was also carried out with only those infants who had valid data in both blocks (16 infants in the nonefficient action condition and 18 infants in the efficient action condition). The analysis revealed no significant effect of condition, F(1, 32) = 0.76, p = .38, or interaction between block and condition, F(1, 32) = 0.47, p = .49. In addition, paired-sample t tests revealed that in the nonefficient action condition, there was a significantly higher percentage of trials with reactive looks (M = 66.39%) than with predictive looks (M = 33.61%), t(18) = 2.39, p = .028), whereas there was no difference between the two types of looks in the efficient action condition, t(18) = 0.92, p = .36. During the Beginning scene, no difference was found between the two conditions in the number of fixations in the Goal AOI, t(34) = 1.44, p = . 15, because there was only 1 infant who fixated in the Goal AOI in the nonefficient action condition and 4 infants who did so, each only once, in the efficient action condition. Looking time Independent-sample t tests showed no difference between the two conditions in the average looking time ratios for the Goal AOI, t(34) = 0.63, p = .52, and the Jump AOI, t(34) = 0.97, p = .33, relative to the total looking time, or in the average looking time for the Jump AOI during the Jumping scene, t(34) = 0.20, p = .83. An independent-sample t test showed that infants tended to look longer at the Obstacle AOI in the efficient action condition than in the nonefficient action condition, t(34) = 1.83, p = .076. Furthermore, there was no difference in the number of times infants looked at the Goal AOI right after they looked at the Obstacle AOI between the two conditions (8 infants in the efficient action condition and 6 infants in the nonefficient action condition, each only once). When possible, the gaze timing and looking time analyses were also carried out using the age (in days) of the infants as a covariate. All of the effects remained the same, suggesting that age had no effect on the fixations.

Fig. 4. Mean gaze arrival in the Goal AOI relative to the arrival of the blue circle in Experiment 2. Positive values imply that the gaze arrived before the circle. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

424

S. Biro / Journal of Experimental Child Psychology 116 (2013) 415–427

Discussion of Experiment 2 In Experiment 2 we found that, overall, 13-month-olds showed predictive looks (i.e., they shifted their gaze to the outcome of the action before it was achieved) only in the efficient action condition. In contrast, in the nonefficient action condition, infants showed reactive looks (i.e., they followed the movement). This finding suggests that efficiency of the action was essential for goal anticipation in this experiment. Compared with the first experiment, Experiment 2 was modified in two important respects, namely that the final straight-line approach was omitted and the salient red circle, which acted as the goal object in Experiment 1, was replaced with a less distinct black figure behind which the blue circle disappeared. Our finding that infants no longer showed predictive gaze shifts in the nonefficient action condition in Experiment 2, thus, may have been caused by either of these two changes or by their combination. Future research should aim to disentangle their exact roles in the lack of goal anticipation. There is some evidence based on looking time measures (Csibra et al., 1999) that if only a parabolic path is taken to make contact with a salient goal object (and thus there is no straight-line approach), then infants do not interpret this action as goal-directed unless the action can be considered as efficient given the situational constraints (i.e., an obstacle is present). The efficient action condition elicited predictive gaze shifts in both Experiments 1 and 2, although the predictive looks occurred considerably later in Experiment 2. This would suggest that the modifications affected the timing of predictive looks in the efficient action condition, but we need to be careful when directly comparing the two experiments in terms of timing because the time window during which infants could produce the anticipatory gaze shifts was shorter in Experiment 2 (given that it did not contain the final approach). In any case, we can conclude from Experiment 2 that the presence of a salient goal object is not necessary to elicit predictive gaze shifts as long as the ongoing action provides evidence for efficient adjustments to the given situation. The presence of a salient, well-defined, and distinct object in the environment may facilitate hypothesis formation about the potential outcome of an ongoing action, and its attractiveness may help to anchor eye movements. Therefore, it would be interesting to test whether a completely unmarked spatial position (which would be even less well-defined than our ‘‘wall’’ in Experiment 2) could serve as a potential outcome that infants could anticipate if the efficiency of the ongoing action could, in principle, justify it. Gredebäck and colleagues’ (2009) study showed that no predictive looks were elicited when objects were simply displaced to another spatial location. However, in their study, the displacement action did not provide movement cues for efficient adjustments. Empty spaces are generally considered as difficult to represent or expect as a referent by infants (Butterworth & Jarrett, 1991; Churcher & Scaife, 1982; Csibra & Volein, 2008). A final point to make regarding the comparison of the two experiments is that neither of them revealed a ‘‘learning effect’’; that is, infants did not show earlier gaze shifts (faster anticipation) to the Goal AOI in the later trials (Block 2) than in the earlier trials (Block 1). This is somewhat surprising in that one would have expected that after several exposures to the goal attainment, infants could have remembered what the outcome was and that this would be reflected in emerging or faster predictive looks to the goal area in subsequent trials. It is, however, not an unusual finding given that other studies typically have not found a change across the trials in the timing of infants’ or adults’ anticipatory eye movements (e.g., Falck-Ytter et al., 2006; Gredebäck et al., 2009), although there are exceptions (see Gredebäck & Melinder, 2010). Note, however, that due to missing trials, the across-block comparison should be interpreted with caution in our study. We speculate that infants may have considered each trial as a somewhat new situation in our experiments due to the slight differences (changing sizes of the jump and the obstacle) between the animations and that this may have prevented them from relying on the previously seen goal in their anticipation.

Conclusions The aim of the current study was to investigate the role of the efficiency of nonhuman actions in eliciting predictive gaze shifts toward the likely outcome of the action in 13-month-old infants. The

S. Biro / Journal of Experimental Child Psychology 116 (2013) 415–427

425

findings of the two experiments together suggest that the efficiency of the action not only influences the timing of predictive looks but also may be essential for their presence. Next, we return to the three hypotheses that we put forward in the Introduction. First, because we found that predictive gaze shifts were produced in the case of novel actions of abstract geometric figures, we can reject the hypothesis that only human actions can elicit goal anticipation. Thus, the exclusivity of a direct matching mechanism for goal anticipation does not hold. Second, our finding that the nonefficient action condition did not elicit predictive gaze shifts in Experiment 2 does not support the hypothesis that agent-defining self-propelled and varied movements are sufficient for predicting the outcome of a nonhuman action. We caution, however, that Experiment 2 could not specify the role of the lack of goal saliency in the absence of predictive looks in the nonefficient action condition. Therefore, further research that clarifies whether the presence of a salient goal object can elicit predictive looks during the observation of a self-propelled, varied, but nonefficient action is necessary to secure this conclusion. Third, the finding that efficiency of the action affected the timing of predictive looks in Experiment 1 and that it was necessary for the presence of predictive looks in Experiment 2 supports the hypothesis that the application of the teleological schema is of crucial importance for the inference of the outcome of novel ongoing actions. Experiment 2 showed that infants are able to rely on the efficiency of the action to anticipate the goal even when the goal object is not salient. This ability is certainly very beneficial because potential goal states or goal objects often do not have salient features in a complex environment. Our findings are consistent with the results of previous studies that used looking time measures to test the ability of infants to infer the outcome of an incomplete nonhuman action based on the efficiency of the action (Csibra et al., 2003; Southgate & Csibra, 2009; Wagner & Carey, 2005). Moreover, because we used predictive eye movements as our measure, our results are not subject to the alternative interpretation of reflecting retrospective evaluation of the witnessed outcome because anticipatory eye movements are elicited before the outcome is achieved. Compared with other studies using predictive eye movements as a measure, our study nicely complements Gredebäck and Melinder’s (2010) study that demonstrated the influence of the efficiency of human action on goal anticipation. Unlike their study, ours avoids confounding efficiency and familiarity because the actions of geometric figures were presumably novel for the infants in our study. At first glance, our results from Experiment 2 seem to be in conflict with the finding of Paulus and colleagues’ (2011) study in which infants’ predictive looks were not found to be affected by the assumption about the efficiency of actions. One reason for the contrasting results could be the younger age of the infants (9 months) in Paulus and colleagues’ study. Another reason might be the difference in the type of inferences that the infants needed to make. We think that the two experiments are similar in the sense that the representation of the goal was demanding in both, albeit in different ways. In our Experiment 2, the agent disappeared (the circle went behind the barrier) while the nonsalient goal object remained in place. In Paulus and colleagues’ study, both the agent and the salient goal object disappeared, so the goal state was not seen as being achieved. However, the major difference between the two experiments is that in ours infants were only required to anticipate the nonsalient goal. In Paulus and colleagues’ study, on the other hand, infants needed to cope with two inferences; they needed to infer and keep in mind the unseen goal and they needed to anticipate the means. The fact that this double requirement was not present in our experiment may explain why infants in our experiment could take the efficiency of the action into account. Further research could clarify the relative contributions of age, type of inference, and goal saliency to infants’ ability to rely on the principle of efficient action for predictive inferences measured by anticipatory eye movements. For example, it could be tested whether infants are able to apply the efficiency principle in their anticipation of a novel means in case of a salient goal or whether 9-month-olds can anticipate a nonsalient goal based on the efficiency of an ongoing action. Finally, we discuss an issue that is related to the type of fixations one can collect from the goal area and the criterion that we used to consider a fixation to the goal area as anticipatory. Recall that we included only those fixations in our analysis that occurred after the start of the jumping of the circle. This criterion was used to ensure that the efficiency of the action can be evaluated in both the presence and absence of the obstacle. However, fixations to the Goal AOI were also produced before the jumping started, namely during the Beginning scene when the circle was moving back and forth.

426

S. Biro / Journal of Experimental Child Psychology 116 (2013) 415–427

These early fixations to the Goal AOI may have been part of generic explorative behavior. On the other hand, it is possible that some of these early fixations were predictive (at least some of those that occurred after the first trial) because infants could, in principle, have anticipated the goal on the basis of having witnessed the goal state being achieved in previous trials. When we looked at fixations to the Goal AOI before the jumping started in Experiment 1, we found that more fixations occurred in the efficient action condition than in the nonefficient action condition and by many more infants. Hence, even if some of these early fixations could be considered as anticipatory looks, it is likely that these occurred more often in the efficient action condition. We could speculate that even repetition-based anticipations may have been influenced by efficiency, which would further strengthen our results. However, in Experiment 2 only a few fixations occurred in the Goal AOI before jumping started. This suggests that in Experiment 1 the saliency of the goal played an important role in eliciting these early fixations regardless of whether they were anticipatory or exploratory. Overall, we believe that the criterion we used was certainly ‘‘safe’’ in the sense that the fixation to the Goal AOI reflected true anticipation of the outcome of the action. In summary, our findings support the importance of teleological inferences for the anticipation of the goals of ongoing actions during infancy. We do not claim that teleological predictive inference is the only mechanism through which infants or adults can anticipate the likely outcome of an ongoing action. Familiarity with an action through prior observational or own action experience can allow one to anticipate its likely outcome in the absence of direct evidence for efficient adjustments of the action (Cannon & Woodward, 2012; Daum, Prinz, & Aschersleben, 2008; Falck-Ytter et al., 2006; Southgate & Csibra, 2009). Teleological inferences can, however, be particularly useful in case of novel actions and in situations where multiple potential outcomes are available, and evaluation of the efficiency of the action can help to disambiguate the likely outcome (Biro et al., 2011; Verschoor & Biro, 2012). Acknowledgment I thank the two anonymous reviewers for their valuable comments on the first version of the manuscript. References Baron-Cohen, S. (1994). How to build a baby that can read minds: Cognitive mechanisms in mindreading. Cahiers de Psychologie Cognitive [Current Psychology of Cognition], 13, 1–40. Biro, S., Csibra, G., & Gergely, G. (2007). The role of behavioral cues in understanding goal-directed actions in infancy. In C. von Hofsten & K. Rosander (Eds.). From action to cognition (Progress in Brain Research (vol. 164, pp. 303–322). Amsterdam: Elsevier. Biro, S., & Leslie, A. M. (2007). Infants’ perception of goal-directed actions. Development through cues-based bootstrapping. Developmental Science, 10, 379–398. Biro, S., Verschoor, S. A., & Coenen, L. (2011). Evidence for a unitary goal concept in 12-month-old infants. Developmental Science, 14, 1255–1260. Butterworth, G., & Jarrett, N. (1991). What minds have in common in space. Spatial mechanisms serving joint visual attention in infancy. British Journal of Developmental Psychology, 9, 55–72. Cannon, E. N., & Woodward, A. L. (2012). Infants generate goal-based action predictions. Developmental Science, 15, 292–298. Churcher, J., & Scaife, M. (1982). How infants see the point. In G. Butterworth & P. Light (Eds.), Social cognition: Studies of the development of understanding (pp. 110–136). Brighton, UK: Harvester. Csibra, G. (2008). Goal attribution to inanimate agents by 6.5-month-old infants. Cognition, 107, 705–717. Csibra, G., Biro, S., Koós, O., & Gergely, G. (2003). One year old infants use teleological representation of actions productively. Cognitive Science, 27, 111–133. Csibra, G., & Gergely, G. (2007). ‘‘Obsessed with goals’’: Functions and mechanisms of teleological interpretation of actions in humans. Acta Psychologica, 124, 60–78. Csibra, G., Gergely, G., Biro, S., Koós, O., & Brockbank, M. (1999). Goal attribution without agency cues: The perception of ‘‘pure reason’’ in infancy. Cognition, 72, 237–267. Csibra, G., & Volein, A. (2008). Infants can infer the presence of hidden objects from referential gaze information. British Journal of Developmental Psychology, 26, 1–11. Daum, M. M., Prinz, W., & Aschersleben, G. (2008). Encoding the goal of an object-directed but uncompleted reaching action in 6- and 9-month-old infants. Developmental Science, 11, 607–619. Eshuis, R., Coventry, K. R., & Vulchanova, M. (2009). Predictive eye movements are driven by goals, not by the mirror neuron system. Psychological Science, 20, 438–440. Falck-Ytter, T., Gredebäck, G., & von Hofsten, C. (2006). Infants predict other people’s action goals. Nature Neuroscience, 9, 878–879. Flanagan, J. R., & Johansson, R. S. (2003). Action plans used in action observation. Nature, 14, 769–771.

S. Biro / Journal of Experimental Child Psychology 116 (2013) 415–427

427

Fogassi, L., Ferrari, P. F., Gesierich, B., Rozzi, S., Chersi, F., & Rizzolatti, G. (2005). Parietal lobe: From action organization to intention understanding. Science, 308, 662–667. Gallese, V., & Goldman, A. (1998). Mirror neurons and the simulation theory of mind-reading. Trends in Cognitive Science, 3, 493–501. Gergely, G., & Csibra, G. (2003). Teleological reasoning in infancy: The naive theory of rational action. Trends in Cognitive Science, 7, 287–292. Gergely, G., Nádasdy, Z., Csibra, G., & Biro, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56, 165–193. Gredebäck, G., & Kochukhova, O. (2010). Goal anticipation during action observation is influenced by synonymous action capabilities, a puzzling developmental study. Experimental Brain Research, 202, 493–497. Gredebäck, G., & Melinder, A. (2010). Infants’ understanding of everyday social interactions: A dual process account. Cognition, 114, 197–206. Gredebäck, G., Örnkloo, H., & von Hofsten, C. (2006). The development of reactive saccade latencies. Experimental Brain Research, 173, 159–164. Gredebäck, G., Stasiewicz, D., Falck-Ytter, T., Rosander, K., & von Hofsten, C. (2009). Action type and goal type modulate goal directed gaze shifts in 14-month-old infants. Developmental Psychology, 45, 1190–1194. Heider, F. (1958). The psychology of interpersonal relations. Oxford, UK: John Wiley. Hernik, M., & Southgate, V. (2012). Nine-month-old infants do not need to know what the agent prefers in order to reason about its goals: On the role of preference and persistence in infants’ goal-attribution. Developmental Science, 15, 714–722. Hommel, B., Musseler, J., Aschersleben, G., & Prinz, W. (2001). The theory of event coding (TEC): A framework for perception and action planning. Behavioral and Brain Sciences, 24, 849–878 (discussion: 878–937). Kamewari, K., Kato, M., Kanda, T., Ishiguro, H., & Hiraki, K. (2005). Six-and-a-half-month-old children positively attribute goals to human action and to humanoid–robot motion. Cognitive Development, 20, 303–320. Kanakogi, Y., & Itakura, S. (2011). Developmental correspondence between action prediction and motor ability in early infancy. Nature Communications, 2, 341. Kochukhova, O., & Gredebäck, G. (2010). Preverbal infants anticipate that food will be brought to the mouth: An eye tracking study of manual feeding and flying spoons. Child Development, 81, 1729–1738. 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. Luo, Y., & Baillargeon, R. (2005). Can a self-propelled box have a goal? Psychological reasoning in 5-month-old infants. Psychological Science, 16, 601–608. Paulus, M., Hunnius, S., van Wijngaarden, C., Vrins, S., van Rooij, I., & Bekkering, H. (2011). The role of frequency information and teleological reasoning in infants’ and adults’ action prediction. Developmental Psychology, 47, 976–983. Premack, D. (1990). The infant’s theory of self-propelled objects. Cognition, 36, 1–16. Rizzolatti, G., Fogassi, L., & Gallese, V. (2001). Neurophysiological mechanisms underlying the understanding and imitation of action. Nature Review Neuroscience, 2, 661–670. Sodian, B., Schoeppner, B., & Metz, U. (2004). Do infants apply the principle of rational action to human agents? Infant Behavior & Development, 27, 31–41. Southgate, V., & Csibra, G. (2009). Inferring the outcome of an ongoing novel action at 13 months. Developmental Psychology, 45, 1794–1798. Southgate, V., Johnson, M. H., & Csibra, G. (2008). Infants attribute goals even to biomechanically impossible actions. Cognition, 107, 1059–1069. Southgate, V., Johnson, M. H., Karoui, I., & Csibra, G. (2010). Motor system activation reveals infants’ on-line prediction of others’ goals. Psychological Science, 21, 355–359. Tomasello, M., Carpenter, M., Call, J., Behne, T., & Moll, H. (2005). Understanding and sharing intentions: The origins of cultural cognition. Behavioral and Brain Sciences, 28, 675–691. Tremoulet, P. D., & Feldman, J. (2000). Perception of animacy from the motion of a single object. Perception, 29, 943–951. Verschoor, S., & Biro, S. (2012). The primacy of means selection information over outcome selection information in infants’ goal attribution. Cognitive Science, 36, 714–725. von Hofsten, C., Vishton, P., Spelke, E. S., Feng, Q., & Rosander, K. (1998). Predictive action in infancy: Tracking and reaching for moving objects. Cognition, 67, 255–285. Wagner, L., & Carey, S. (2005). Twelve-month-old infants represent probable endings of motion events. Infancy, 7, 73–83. Woodward, A., & Sommerville, J. A. (2000). Twelve-month-old infants interpret action in context. Psychological Science, 11, 73–77.