Infants’ preference for individual agents within chasing interactions

Journal of Experimental Child Psychology 147 (2016) 53–70

Contents lists available at ScienceDirect

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

Infants’ preference for individual agents within chasing interactions Martyna Galazka, Pär Nyström Uppsala Child and Baby Lab, Department of Psychology, Uppsala University, S-751 42 Uppsala, Sweden

a r t i c l e

i n f o

Article history: Received 28 November 2015 Revised 19 February 2016

Keywords: Visual attention Social interactions Social perception Infant Eye tracking Chasing

a b s t r a c t Infants, like adults, are able to discriminate between chasing and non-chasing interactions when watching animations with simple geometric shapes. But where infants derive the necessary information for discrimination and how chasing detection influences later visual attention has been previously unexplored. Here, using eye tracking, we investigated how 5- and 12-month-old infants (N = 94) distribute their visual attention among individual members within different interactions depending on a type of interaction. Infant gaze was examined when observing animations depicting chasing and following interactions compared with animations displaying randomly moving shapes. Results demonstrate that when observing chasing and following interactions, all infants strongly preferred to attend to the agent that initiates an interaction and trails behind another. Low-level features, such as changes in agent-specific velocity profiles, could not account for this preference (Study 2). Rather, the strong preference for the agent going behind seems to be dependent on the initial goal-directed or ‘‘heat-seeking” motion of one agent toward another (Study 3). The current set of experiments suggests that, similar to adults, 5months-olds’ visual attention depends on the motion features of an individual agent within the interaction and is fine-tuned to agents that display goal-directed motion toward other agents. Ó 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

E-mail addresses: [email protected] (M. Galazka), [email protected] (P. Nyström) http://dx.doi.org/10.1016/j.jecp.2016.02.010 0022-0965/Ó 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

54

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

Introduction Causal relations are essential in comprehending an entire range of external events such as understanding the effect of one’s own manual actions or understanding an object’s motion trajectory when it collides with another. Whereas some types of causal relations, such as launch-and-collision events, involve direct physical contact, other types of interactions do not. This is typically the case for social exchanges. Correctly identifying such non-contact causal relations holds a great benefit in fostering learning about social interactions and the intentional agents that display them (Schlottmann & Surian, 1999). Previous studies have shown that infants detect information necessary for categorizing entities as intentional (Biro, Csibra, & Gergely, 2007; Johnson, 2003; Johnson, Slaughter, & Carey, 1998; Premack, 1990). However, no study to our knowledge has examined how non-contact causal relations guide attention to specific agents, informing what information is actively prioritized when observing interactions between animate agents. Limited visual displays, such as animations with multiple selfpropelled moving geometrical shapes, provide a promising way to investigate this question. Previous studies have demonstrated that when presented with such limited visual displays, adult observers attend to an object’s motion as it relates to others and perceive the object as an animate entity (Castelli, Happé, Frith, & Frith, 2000; Gergely, Nádasdy, Csibra, & Bíró, 1995; Heider & Simmel, 1944). When asked to report what they have observed, adults consistently describe the geometric shapes as anthropomorphized entities moving in an intentional, goal-directed manner—the ball wants to get, tries to catch, and runs away from another. Although there are many types of perceivable interactions, here we specifically focused on chasing interactions. Chasing interactions are, in an evolutionary sense, one of the most relevant animated percepts (Kanizsa & Vicario, 1968) and have been previously studied in both adults (Gao, Newman, & Scholl, 2009; Meyerhoff, Huff, & Schwan, 2013; Meyerhoff, Schwan, & Huff, 2014; Scholl & Gao, 2013) and infants (Frankenhuis, House, Barrett, & Johnson, 2013; Rochat, Morgan, & Carpenter, 1997). This type of interaction involves agents that have a relatively simple spatial relationship (one is ahead and one is behind), making it favorable to study because it is easy to detect, it does not require turn taking or role shifting, and the interaction can be extended in time without repeating. Although chasing interactions tend to be attention grabbing for participants of all ages, these three factors allow it to be studied with very young infants. In a typical chasing interaction, one object moves toward another in a direct path while its target speeds up to move away when the chasing object gets too close. Adult observers readily identify chasing (Meyerhoff et al., 2013, 2014) and interpret it in goal-directed terms (Dittrich & Lea, 1994; Heider & Simmel, 1944; Morris & Peng, 1994). More recently, adult research has focused on determining how chasing detection takes place. In one study, Meyerhoff and colleagues (2014, Experiment 4b) presented adults with a chasing pair of objects embedded in 10, 15, or 20 identical distractors. After running series of controls, the authors concluded that of the two interacting objects, adults are more biased toward the chasers, or the objects that move toward other objects, than toward the objects being chased. In other words, they were faster at identifying an object that approached another from among distractors than identifying an object that was being approached. Other research (Scholl & Gao, 2013) corroborated these findings, showing that for adult observers the chaser, not the chase, is more attentionally salient. Infants, much like adults, have been found to discriminate between chasing and non-chasing events (Rochat et al., 1997). One of the best examples examining early sensitivity to relational information between objects through motion is in the reactions to schematic action-and-reaction sequences. Through their work, Schlottmann, Ray, and Surian (2012) showed that the ability to detect differences in these types of non-contact causal relations emerges between 4 and 6 months of age. At 5 months infants show post-habituation recovery to changes in the spatiotemporal motion features, and by 6 months infants also become sensitive to changes in the perceived causality during simple action–reaction events as well as to the reversal of the causal structure. Similar sensitivity to changes in causality has also been demonstrated with infants toward the end of their first year (Schlottmann & Surian, 1999; Schlottmann, Surian, & Ray, 2009). Using more complex chase-and-escape sequences, 8- to 10-month-old infants were habituated to a chasing interaction

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

55

between two discs that differed in color (Rochat, Striano, & Morgan, 2004). At test, when the colored discs reversed roles, infants effectively dishabituated, signaling not only a reaction to a change in the causality but also an improvement in object identification and sensitivity to each agent’s specific role within the interaction (Bonatti, Frot, Zangl, & Mehler, 2002; Káldy & Leslie, 2003; Leslie, Xu, Tremoulet, & Scholl, 1998; Surian & Caldi, 2010). By 12 months, infants not only react to role reversal of the chasing agents but also are able to accurately calculate the chaser’s intent of getting the target while inferring its future motion patterns (Csibra, Bíró, Koós, & Gergely, 2003; Wagner & Carey, 2005). To date, global looking measures, such as those used in preferential looking and habituation studies, were sufficient in determining general preference for and sensitivity to social motions (Frankenhuis et al., 2013; Schlottmann & Surian, 1999; Schlottmann et al., 2009). However, global preference measures are limited in determining what the mechanisms are behind infant visual attention during the online observation of socially causal events. In other words, although it is important to know what types of interactions are attended, it is equally important to know how and why infants attend to certain types of interactions—where the information used for determining socially contingent interactions is derived from and where infants focus their attention in order to track the progression of the interaction. The question becomes whether adults’ attentional bias toward the chaser is present early in infancy or develops together with understanding agent-specific roles within the first year of life. To assess these questions, we considered how infants attend to individual objects during dynamic displays using an eye-tracking methodology. Investigating such specific looking patterns in very young infants constitutes the first effort in deconstructing infant attention patterns while expanding current knowledge on the mechanisms behind chasing detection and its subsequent influence on gaze behavior.

Study 1 The aim of the first study was to explore the distribution of attention among three identical shapes based on whether and how they interact. Looking times at identical self-propelled shapes were assessed as they engaged in a chasing interaction compared with two control events. In the Chase event, infants observed one shape (Agent Ahead) being chased by another (Agent Behind) while the third shape (Distractor) moved randomly in no relation to the other two shapes. In the first control event, the No Interaction event, all three shapes moved randomly and independently from each other at a constant speed. In the second control event, the Follow event, the Agent Behind slowed down after catching up to the Agent Ahead, giving the impression of the Agent Behind following, rather than chasing, the Agent Ahead. For this study, we adapted the ‘‘wavering wolf” paradigm by Gao and colleagues (2009), in which the Agent Behind switches the target of its chase (Distractor becomes Agent Ahead and vice versa) mid-presentation in the Chase and Follow events. This was done to examine whether infants could detect the interacting members based on the behavior of the Agent Behind. Distribution of looking was examined in 5- and 12-month-olds. As previously noted, by 5 months infants discriminate between chasing and non-chasing interactions with an emergent sensitivity toward spatiotemporal changes between interacting agents. The older 12-month-olds were tested to determine the consistency of looking across development. Previous research suggests that the end of the first year is marked by a more sophisticated conceptual understanding of causality and each agent’s role within simple action–reaction sequences as well as within more complex chasing interactions (Rochat et al., 2004). Here, we examined whether this type of understanding modulates how infants at that age distribute their attention. We hypothesized that the key difference in how infants attend to perceptually causal events is in how they attend to individual agents within the interactions. Adult research has shown that adults derive social information about the type of observed interaction by evaluating individual objects’ motion when determining whether the attended object is involved in a chasing interaction (Meyerhoff et al., 2014). Specifically, adults seem to be biased toward cues pertaining to the chasing agent (Meyerhoff et al., 2013; Scholl & Gao, 2013). Here, we examined whether the same attentional

56

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

bias occurs early in development, as reflected in increased looking times at the Agent Behind in a chasing interaction compared with a following interaction. By comparison, the Follow event maintains the same spatial structure as the Chase event, but in the Follow event the Agent Behind simply follows the trajectory of the Agent Ahead and thus its position in relation to the Agent Ahead has no consequence on the motion of the Agent Ahead. Rather, it is the motion path of the Agent Ahead that directs the interaction. We predicted that infants would display increased looking at the Agent Ahead in the Follow event compared with the Agent Ahead in the Chase event. Lastly, during events where the objects move independently from each other, we predicted equal distribution of attention across all three agents. Method Participants The final sample consisted of 28 12-month-old (16 female) and 24 5-month-old (12 female) infants. All participants were full-term without known neurological or developmental disabilities. The 12-month-olds were between 346 and 377 days old (M = 363 days, i.e., 11 months 28 days). All 5-month-olds were between 148 and 167 days old (M = 161 days, i.e., 5 months 11 days). An additional 5 infants (2 5-month-olds and 3 12-month-olds) were tested but excluded from the experiment due to fussiness from the beginning of the stimulus presentation (n = 2) or because of an insufficient gaze recording by the eye tracker (<20% of gaze samples) (n = 3). Participants were recruited from a list of parents who indicated interest in participating in research with their children. Infants were primarily from White and middle-class backgrounds and lived in a medium-sized European city. Parents received a gift voucher worth approximately 10 euros for participation. The study was conducted in accordance with the 1964 Declaration of Helsinki, and all children’s parents provided written consent according to the guidelines specified by the ethical committee at Uppsala University. Stimuli and apparatus Animations were randomly generated using Adobe Flash Professional CS5 and ActionScript 3 (Adobe Systems, San Jose, CA, USA). The animations depicted three identical orange discs (Hex color #ff6600) with a diameter of 40 pixels (1.26 visual degrees) moving across a black background. The trajectories of the objects were generated by making each object take the shortest possible path to a random target location on the screen. Once the object arrived at the target location, a new random location was assigned to the object. A physical constraint was applied to all objects so that they could not turn faster than 0.1 radians per frame in order to eliminate attention grabbing distinct changes in direction. In the chasing and following conditions, the Agent Behind’s target location was always the position of the Agent Ahead, effectively making the Agent Behind head toward the Agent Ahead but at different speeds. For each event type, a set of video clips was created and a clip where the objects never came in contact with each other or the boundary of the scene was selected for presentation. These video clips were then edited to create a 15-s video clip with background music using Adobe Premier Pro CS6 video editing software. The stimuli are available in the online supplementary material. Gaze was measured using a Tobii Model T120 corneal reflection eye tracker (Tobii Technology, Danderyd, Sweden), recorded at 60 Hz (accuracy = 0.4°, precision = 0.16°) and presented on a 17inch (43-cm) flat screen monitor (1024  768 pixels). All displays were presented at 30 frames per second. Procedure All participants were approximately 60 cm from the screen. The 5-month-olds were seated in a car seat that was placed on a chair, whereas the 12-month-olds were seated on a parent’s lap. Parents were instructed to refrain from commenting or otherwise interfering during the presentation unless to redirect their infants’ gaze toward the screen if the children looked away. A standard 5-point infant calibration was used and passed by all participants (Gredebäck, Johnson, & von Hofsten, 2010). Participants observed three presentations of each event type in a semi-random order, taking care that no event type was identical to the last one: Chase, Follow, and No Interaction. All three events

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

57

began in the same triangular arrangement, but as a control the middle block was presented in an inverted orientation, where the stimuli were identical but flipped vertically (Fig. 1). This was done to ensure that the starting position did not affect participants’ ability to track objects (e.g., children tracking only the highest positioned object). Chase event. At the start of the event, the Agent Ahead and Distractor discs began moving at 5 pixels per frame (ppf), whereas the Agent Behind moved at a constant speed of 7.5 ppf toward the Agent Ahead with 0° deviation from a direct path toward it in a ‘‘heat-seeking” manner. Given higher speed and the direction of the Agent Behind relative to the Agent Ahead, within 2 s the two discs were within 90 pixels from each other (calculated from their centers). Once within 90 pixels, the Agent Ahead accelerated to 15 ppf away from the Agent Behind until the distance between them increased (1 s). The Agent Ahead then decelerated to its initial speed of 5 ppf. Again, because the Agent Behind moved at a constantly higher speed than the Agent Ahead, the Agent Behind eventually re-approached the Agent Ahead (6 s after stimulus onset). Once again, the Agent Ahead accelerated to 15 ppf away from the Agent Behind. At this point (8 s after stimulus onset), after this second approach–escape, the Agent Behind switched the target of its pursuit and began moving with 0° deviation from a direct path toward the Distractor, making the Distractor disc the new Agent Ahead. Similarly to the first target, when the Agent Behind and new Agent Ahead were less than 90 pixels away, the new Agent Ahead accelerated to the speed of 15 ppf away from the Agent Behind and then returned to its initial speed and accelerated again when the Agent Behind was again less than 90 pixels away. Throughout the trial, there were four approach–escape instances—two with the original Agent Ahead and two with the new Agent Ahead. Follow event. The motion algorithms were identical to those in the Chase event with one exception: Once the Agent Behind and Agent Ahead were less than 90 pixels away from each other, both agents decelerated. That is, as in the Chasing event, the Agent Behind moved at the speed of 7.5 ppf toward the Agent Ahead at the stimulus onset, whereas the Agent Ahead moved at 5 ppf. After 2 s, the Agent Behind and Agent Ahead were less than 90 pixels away. But unlike in the Chase event, in the Follow event the Agent Behind decelerated from 7.5 to 3.5 ppf, following the path of the Agent Ahead that decelerated from 5 to 4 ppf for approximately 6 s. After 8 s from stimulus onset, the Agent Behind accelerated back to the speed of 7.5 ppf in the direct path toward the Distractor, making the Distractor the new Agent Ahead. Once within 90 pixels from the new Agent Ahead (10 s from stimulus onset), both the Agent Behind and Agent Ahead again decelerated and followed each other for the reminder of the stimulus (5 s). No Interaction event. All discs moved randomly at a constant speed of 5 ppf throughout the event. The discs moved independently from each other, at times coming within less than 90 pixels of each other, but that did not alter their velocity or trajectory. Data reduction For each event, dynamic circular areas of interest (AOIs) with a radius of 75 pixels (2.36 visual degrees) were created around each of the discs (Fig. 1). Percentage of looking time within each moving AOI was generated using TimeStudio Version 2.8 (available at Figshare: https://figshare.com/articles/ TimeStudio_v2_7/1293476/2), a Matlab program specifically designed for analyzing time series data. The program calculated percentage of looking as a fraction of time spent looking within a specified AOI divided by the trial length (15 s). To create the dependent variable, percentage of looking across trial length was recalculated as a proportion of looking at individual AOIs over total percentage for each trial. Trials with less than 10% of total looking were rejected and not included in the final analysis. The proportion of looking at each AOI was then averaged over trials across three event types (Chase, Follow, and No Interaction). The averaged proportions were used for the statistical analysis. For the sake of consistency, during the No Interaction condition the labels for each disc refer to the agents according to their starting position and their roles in the Chase event. Finally, in the Chase and Follow events, as the Agent Behind began moving toward the Distractor, the AOIs for the Agent Ahead

58

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

Fig. 1. Starting positions of the discs with transparent overlays of areas of interest for the Agent Behind, Agent Ahead, and Distractor in the upright (left) and inverted (right) orientations.

and the Distractor switched, effectively accounting for the change in the target object by the Agent Behind. Results To examine differences in the proportion of looking across event types and areas of interest in both age groups, a 2 (Orientation)  3 (Event Type)  3 (AOI Type) repeated measures analysis of variance (ANOVA) was conducted with event type (Chase, Follow, or No Interaction) and AOI type (Agent Behind, Agent Ahead, or Distractor) as within-participant variables and age group, gender, and stimulus presentation order as between-participant variables. Factors such as initial configuration of discs, stimulus presentation order, and infant’s gender yielded no significant effects (all ps > .05). The analysis also revealed no significant main effect of age, F(1, 12) = 0.77, p = .397, g2 = .06, or event type, F (2, 24) = 0.39, p = .677, g2 = .01. There was, however, a significant main effect of AOI type, F(2, 24) = 36.70, p < .001, g2 = .75. Follow-up analyses showed that in both age groups infants attended more to the Agent Behind (M = .46, SE = .01) than to the Agent Ahead (M = .30, SE = .01), MDifference = .166, SE = .02, 95% confidence interval (CI) [.11, .22], p < .001, and attended more to the Agent Ahead than to the Distractor (M = .24, SE = .02), MDifference = .21, SE = .03, 95% CI [.14, .29], p < .001. The analysis also yielded a significant event type and AOI type interaction, F(4, 48) = 11.19, p < .001, g2 = .48, denoting that the proportion of looking at AOIs differed according to the type of event that infants observed. For the 12-month-olds (Fig. 2), the Bonferroni-adjusted post hoc analyses showed that when observing the Chase event, infants preferred to look at the Agent Behind (M = .49, SE = .02) over the Agent Ahead (M = .30, SE = .01), MDifference = .19, SE = .03, 95% CI [.11, .27], p < .001, and looked more at the Agent Ahead than at the Distractor (M = .21, SE = .02), MDifference = .09, SE = .02, 95% CI [.03, .15], p = .002. Similarly, when observing the Follow event, infants preferred to look at the Agent Behind (M = .44, SE = .02) over the Agent Ahead (M = .30, SE = .02), MDifference = .14, SE = .02, 95% CI [.08, .20], p < .001, but looked equally at the Agent Ahead and the Distractor (M = .26, SE = .03), MDifference = .04, SE = .05, 95% CI [ .17, .08], p = 1.00. For the No Interaction event, there were no significant differences among the AOIs (MAgent Behind = .34, SE = .02; MAgent Ahead = .33 SE = .02; MDistractor = .33, SE = .02), F(1, 27) = 0.06, p = .815, g2 = .002. The 5-month-olds’ looking pattern (Fig. 3) differed from that of the 12-month-olds in the Chase and Follow events. During the Chase event, much like 12-month-olds, 5-month-olds looked more at the Agent Behind (M = .54, SE = .03) than at the Agent Ahead (M = .28, SE = .04), MDifference = .27, SE = .06, 95% CI [.11, .43], p = .001, but unlike 12-month-olds, they did not look more at the Agent Ahead than at the Distractor (M = .18, SE = .02), MDifference = .10, SE = .05, 95% CI [ .03, .23], p = .187. When observing the Follow event, the 5-month-olds, unlike the older infants, did not look significantly more at the Agent Behind (M = .48, SE = .03) than at the Agent Ahead (M = .32, SE = .04), MDifference = .16, SE = .06,

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

59

Fig. 2. Average proportion of looking time at the Agent Behind, Agent Ahead and Distractor areas of interest by Event type in 12 month olds (Study 1). Vertical lines indicate mean standard error. Chase: *Agent Behind, Agent Ahead p = .0001; *Agent Behind, Distractor p = .0001; *Agent Ahead, Distractor p = .002. Follow: *Agent Behind, Agent Ahead p = .0001; *Agent Behind, Distractor p = .0001.

95% CI [ .32, .003], p =.056. Like the 12-month-olds, they looked equally at the Agent Ahead and the Distractor (M = .205, SE = .030), MDifference = .136, SE = .023, 95% CI [.076, .195], p < .001. For the No Interaction event, there were no significant differences between the AOIs (MAgent Behind = .329, SE = .019; MAgent Ahead = .349, SE = .024; MDistractor = .322, SE = .031), F(1, 23) = 0.027, p = .870, g2 = .001. Finally, as a control analysis, we assessed whether the total amount of analyzable data differed among event types and whether other factors influenced the dependent variable. These analyses showed no significant differences among the percentages of analyzable data, F(2, 102) = 2.29, p = .106, g2 = .04.

Discussion The results of Study 1 demonstrate that when observing abstract geometrical shape chasing, both 12- and 5-month-old infants visually attend to the chaser (Agent Behind) relative to its target (Agent Ahead). When all three agents move independently from each other, infants in both age groups look equally at all three. The results from these analyses eliminated the possibility that factors within the stimulus presentation, such as the initial configuration of the objects and the order of the event presentation, influenced infants’ looking behavior. Rather, the way infants attended to the objects seemed to have been motivated by how the objects moved in relation to each other.

60

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

Fig. 3. Average proportion of looking time at the Agent Behind, Agent Ahead and Distractor areas of interest by Event type in 5 month olds (Study 1). Vertical lines indicate mean standard error. Chase: *Agent Behind, Agent Ahead p = .001; *Agent Behind, Distractor p = .0001.

An alternative interpretation for the lack of preference in the No Interaction event may have to do with the lack of low-level variations in velocity and direction changes inherent in the interactions such as chasing and following. In both chasing and following events, the Agent Behind’s initial speed is greater than that of the other two agents, and in both cases the Agent Behind exhibited distinct changes in the direction of its pursuit mid-presentation by switching the target. Similarly, the Agent Ahead exhibited variations in velocity and turning rates. Because the No Interaction event lacked all of these low-level visual variations, it is difficult to say whether the lack of preference in the No Interaction event can be accounted for by the lack of motion contingencies within interactions. This was assessed in Study 2. The current findings also show that participant characteristics such as the infant’s gender and age did not interact with the dependent variable. However. one noticeable age group difference emerged when observing the Follow event. Specifically, 12-month-olds seemed to have continued to attend to the Agent Behind over the Agent Ahead, whereas this difference was not significant (although it approached significance) in the younger group. This analysis also shows that neither of the age groups differentiated between the Distractor and the Agent Ahead in the Follow event, whereas only 5month-olds did so in the Chase event. These findings indicate that in both types of interactions, the Agent Behind is preferentially attended. Contrary to our predictions, its target (Agent Ahead), although part of an interaction, does not seem to do the same whether it is a target of the chase or whether it is leading the interaction in a following interaction. The continued preference for the Agent Behind in the chasing and following interactions in 5-month-olds is a striking finding with several plausible interpretations. One interpretation of this

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

61

preference is that looking at the Agent Behind reflects a fundamental scanning strategy that is present early in development where, independent of the type of observed interaction, infants always direct their gaze at the agent that is second in the line of sight to the direction of the movement. This could be due to factors such as occasional gaze lags (which would make the Agent Behind come into the center of gaze and attract attention), retinal optical flow (where the Agent Behind is always going toward the center of gaze and the Agent Ahead is always moving toward the periphery), and higher order processes directing attention toward the Agent Behind. Although the finding in itself is important for understanding how infants distribute their attention, the underlying factors are difficult to investigate using only the stimuli from Study 1. Another alternative explanation for the preference for the Agent Behind is that infants direct their visual attention to the agent that alters the spatial relationship between itself and other objects. That is, attention may be directed to the agent that approaches another agent by accelerating in the direct path toward another agent or that is moving in a heat-seeking manner (Gao et al., 2009). Previous findings using the wavering wolf paradigm with adults found an attentional bias toward agents currently involved in the chase but more specifically toward the chaser moments following the switch of its target (Scholl & Gao, 2013). It is possible that the current results simply reflect a similar bias in infants. In Study 1, the Agent Behind first approached the Agent Ahead in both interaction types and did so again mid-presentation when it switched the target of its pursuit. It is possible that this heat-seeking motion singled out this particular agent over the other two agents independent of the type of interaction in which it subsequently engaged. Consequently, modulating heat-seeking behavior may in turn lessen the attention bias toward the Agent Behind during interactions (this was assessed in Studies 2 and 3). Study 2 To examine whether motion contingency or low-level motion cues account for distribution of attention, infants were presented with two types of No Interaction events in which the acceleration and turning profiles mimicked those in the Chase and Follow events. If individual motion cues alone attract attention outside interaction, then a preference pattern for the shapes displaying these cues should emerge. To examine whether modulating initial heat-seeking motion (which signifies the initiation of interaction and contains salient cues for the different roles of the agents) in the Agent Behind would modulate preference for the Agent Behind (the second purpose of this study), infants were presented with Chase and Follow events in which the switch of the target by the Agent Behind was eliminated, reducing heat seeking as a cue for initiation of interaction for the Agent Behind. Because in the previous study we found the preference across interaction types for Agent Behind in the youngest group tested, Study 2 included only 5-month-old infants. Method Participants The final sample consisted of 24 5-month-olds (M = 158 days, range = 144–170) and was divided into two groups: No Interaction–Chase group (M = 161 days, range = 150–177; 6 female) and No Interaction–Follow group (M = 155 days, range = 144–170; 6 female). One additional infant was tested but excluded because the gaze recording by the eye tracker was less than 20%. None of the infants tested in Study 2 participated in Study 1. Stimuli and apparatus Stimuli were generated, edited, and presented in the same way as those in Study 1. All participants observed three event types: Chase–No Switch, Follow–No Switch, and one of two No Interaction events (No-Interaction–Chase event or No Interaction–Follow event). The groups differed by the type of the No Interaction event. For both No Interaction events, each agent’s motion trajectories were identical to those of agents in the Chase and Follow events in Study 1. Like in Study 1, the trajectories for

62

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

each were generated by making each agent take the shortest possible path to a random target location on the screen. But unlike in Study 1, the target location was never that of another agent. Effectively, one agent exhibited velocity and turning rates identical to those of the Agent Behind, whereas a second object exhibited velocity and turning rates identical to those of the Agent Ahead, but without any spatial relation to each other. Like in Study 1, as a final control, all three events were presented once in an inverted configuration. In all, infants observed three blocks with three event types: two in the upright configuration and one in the inverted configuration. As in Study 1, all events began in the same triangular configuration and all trials were 15 s (examples of each event are found in the supplementary material). Chase–No Switch event. Motion algorithms were identical to those in Study 1 except that the Agent Behind never switched its target but continued to chase the same disc throughout the presentation. Follow–No Switch event. This was the same as in Study 1 except that the Agent Behind never switched its target. No Interaction–Chase event. Two discs moved randomly at 5 ppf, whereas the third disc moved at a constant speed of 7.5 ppf, mimicking the velocity of an Agent Behind. After approximately 2 s, one disc accelerated to 15 ppf for approximately 1 s and again decelerated to its original velocity, simulating the velocity change of the Agent Ahead ‘‘evading” the chaser. Approximately 6 s after stimulus onset, the same agent again accelerated and decelerated, simulating the second approach–escape. Approximately 8 s after stimulus onset, the third agent accelerated and decelerated twice, simulating the new Agent Ahead from Study 1. Consequently, in terms of velocity changes, all three discs moved similarly to the discs in the Chase event of Study 1 but without the spatial relation to each other. No Interaction–Follow event. Two discs moved randomly at a constant speed of 5 ppf, whereas the third disc moved at the speed of 7.5 ppf, mimicking the velocity of an Agent Behind. After 2 s, the disc that moved at the speed of 7.5 ppf and one of the discs that moved at 5 ppf decelerated to 3.5 and 4 ppf, respectively, simulating the velocity change in the following interaction. After approximately 8 s, the disc moving at the speed of 3.5 ppf accelerated to 7.5 ppf for approximately 2 s, after which it decelerated to 3.5 ppf, whereas a third disc that until this time moved at a constant speed of 5 ppf decelerated to 4 ppf. Thus, in terms of velocity changes, all three discs moved similar to the discs in the Follow event. Results As in Study 1, the results indicated that 5-month-olds continued to prefer looking at the Agent Behind in the Chase–No Switch and Follow–No Switch events compared with the No Interaction events (Fig. 4). Similar to Study 1, a 2 (Orientation)  3 (Event Type: Chase–No Switch, Follow–No Switch, or No Interaction)  3 (AOI Type) repeated measures ANOVA with presentation order and gender as between-participant factors revealed a significant main effect of AOI type, F(2, 26) = 27.54, p < .001, g2 = .68, with infants looking more at the Agent Behind (M = .45, SE = .02) than at the Agent Ahead (M = .31, SE = .01), MDifference = .14, SE = .03, 95% CI [.04, .23], p = .004, but looking more at the Agent Ahead than at the Distractor (M = .24, SE = .01), MDifference = .08, SE = .02, 95% CI [.03, .12], p = .001. Like in Study 1, there was no significant main effect of event type, F(2, 26) = 0.84, p =.44, g2 = .06, but the analysis demonstrated a significant event by AOI type interaction, F(4, 52) = 30.94, p < .001, g2 = .70, indicating that the proportion of looking at different AOIs differed according to the type of event that infants observed. Bonferroni-adjusted post hoc analyses revealed that in the Chase–No Switch event, infants preferred to look at the Agent Behind (M = .56, SE = .02) over the Agent Ahead (M = .25, SE = .01), MDifference = .31, SE = .03, 95% CI [.22, .39], p < .001, but did not look differently at the Agent Ahead and the Distractor (M = .19, SE = .02), MDifference = .05, SE = .03, 95% CI [.26, .47], p = .151.

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

63

Fig. 4. Average proportion of looking time at the Agent Behind, Agent Ahead and Distractor areas of interest by Event type in 5 month olds (Study 2). Vertical lines indicate mean standard error. Chase–No switch: *Agent Behind, Agent Ahead p = .0001; * Agent Behind, Distractor p = .0001. Follow–No switch: *Agent Behind, Agent Ahead p = .001; *Agent Behind, Distractor p = .0001; *Agent Ahead, Distractor p = .0001.

When observing the Follow–No Switch event, infants preferred to look at the Agent Behind (M = .49, SE = .02) over the Agent Ahead (M = .35, SE = .02), MDifference = .15, SE = .03, 95% CI [.06, .24], p = .001, but looked more at the Agent Ahead than at the Distractor (M = .15, SE = .02), MDifference = .19, SE = .03, 95% CI [.12, .27], p < .001. There were no significant differences between the AOIs in the two No Interaction events, F(2, 42) = 2.09, p = .135, g2 = .09, and a comparison between the No Interaction events in Study 1 and Study 2 showed no differences in the distribution of looking between the AOIs in the two studies (for all AOIs, ps > .20). Furthermore, total percentage of looking at the No Interaction events did not differ among 12-month-olds (M = 53.00, SE = 4.87), 5-month-olds in Study 1 (M = 42.49, SE = 5.07), and 5-montholds in Study 2 (M = 43.79, SE = 3.39), F(1, 74) = 3.28, p = .074, g2 = .04. Finally, factors such as the initial configuration of the discs, stimulus presentation order, and infant’s gender revealed no significant effects (all ps > .05). Discussion The goals of this follow-up study were twofold. The first goal was to examine whether the distribution pattern observed in the Chase and Follow events and lack thereof in the No Interaction event in Study 1 was due to the changes in low-level features such as relative velocity profiles of the Agent Behind and Agent Ahead—that is, whether certain motion cues influence attention when in isolation outside the context of an interaction. In addressing this aim, findings from Study 2 suggest that velocity profiles characteristic of the chasing and following interactions in themselves do not drive attention; overall, there was no increase in the percentage of total looking between the No Interaction event in Study 1 and the two No Interaction events in Study 2. In addition, there were no differences in the proportions of looking among the AOIs within the No Interaction events. Whether objects accelerated,

64

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

decelerated (Study 2), or moved at a constant speed (Study 1) did not seem to influence how infants attended to the objects; as long as the objects moved independently from each other, infants showed no preference to any agent and looked at all three objects equally. These results stand in contrast to previous findings (Frankenhuis et al., 2013), which found infants to fixate longer at displays in which objects accelerated compared with controls. However, the inherent difference in the design of our study (one display instead of two displays) may account for the difference in the findings. It is possible that, if given a choice, infants would orient toward displays with more varied motion cues. Based on the current findings, however, we find that when presented with a single display, the preference was equal for all three agents despite low-level features. The second aim of the study was to determine whether preference for the Agent Behind is modulated by the amount of heat-seeking motion during the initiation of interaction with the Agent Ahead. The findings demonstrate that the reduction of the target switch by the Agent Behind did not influence the preference for it. Both Chase–No Switch and Follow–No Switch replicated the distribution in Study 1, and as in Study 1, compared with the chasing interaction, in the following interaction there was an increase in looking at the Agent Ahead. The results demonstrate that the preference for the Agent Behind was not modulated proportionally to the amount of heat seeking. However, it is possible that the allocation of visual attention is logistic and that just a short exposure to heat-seeking motion at the beginning of the stimuli presentation is enough to drive the overall preference for the heat seeker. Given this interpretation, infants would rapidly attribute different roles to the agents in the interactions and sustain attention to these agents throughout the presentation. Of the possible explanations from Study 1, one can be ruled out: The acceleration profiles alone do not drive the preference. Thus, two explanations remain, namely that (a) the preference reflects a fundamental scanning strategy to always pay more attention to the Agent Behind and (b) the preference is driven by certain chaser-specific motion features such as a short period of heat seeking at the start. The goal of Study 3 was to address this possibility.

Study 3 The third study aimed to address whether the preference for the Agent Behind can be accounted for by the initial heat seeking that initiated the interaction. Here, one group of 5-month-old infants observed following and chasing events that began without an initiation of the interaction, that is, without one object getting close to another. We hypothesized that if the preference for the Agent Behind is due to early detection of a heat seeker, the preference should disappear with the elimination of heat seeking at the start of the following events. However, in the chasing events, it is possible that other chasing-specific cues are sufficient to detect the chaser, which would result in similar preferences as in Studies 1 and 2. Alternatively, if the preference for the chaser remains the same with the elimination of heat seeking in both interaction types, it may suggest that the preference for the chaser is reflective of a fundamental scanning strategy whereby infants always look at the agent that is second in the line of sight to another agent. Method Participants The final sample consisted of 18 5-month-old infants (M = 152 days, range = 134–168; 7 female). One additional infant was tested but excluded from the experiment due to an insufficient gaze recording by the eye tracker (<20% of gaze sample). None of the infants tested in Study 3 participated in Study 1 or 2. Stimuli and apparatus Stimuli were the same as those used in Study 2 Chase–No Switch and Follow–No Switch events with the first 2 s of the video trimmed out in order to eliminate the initial motion of the Agent Behind toward the Agent Ahead. As in Studies 1 and 2, both the Chase and Follow events were presented in

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

65

blocks of three trials, with one trial presented in an inverted configuration. The order of the presentation was counterbalanced, and the trials were 13 s in duration. Chase–No Initiation and Follow–No Initiation. The velocity characteristics of the discs were identical to those in the Chase and Follow events in Study 2, respectively. Both events began with two shapes on the lower left-hand side of the screen and the Distractor on the right-hand side. Results A 2 (Orientation)  2 (Event Type)  3 (AOI Type) repeated measures ANOVA with presentation order and gender as between-participant factors revealed no differences across gender, stimulus orientation, and the order of stimulus presentation (all ps > .05). It did indicate a main effect of AOI type, F (2, 28) = 43.86, p < .001, g2 = .76, with infants looking more at the Agent Behind (M = .55, SE = .03) than at the Agent Ahead (M = .32, SE = .03), MDifference = .24, SE = .05, 95% CI [.09, .39], p < .001, but looking more at the Agent Ahead than at the Distractor (M = .13, SE = .02), MDifference = .18, SE = .04, 95% CI [.09, .28], p < .001. There was no main effect of event type, F(1, 14) = 0.263, p = .616, g2 = .018. In addition, the analysis demonstrated a significant interaction effect between event type and AOI type, F(2, 28) = 13.95, p < .001, g2 = .49. Post hoc analyses showed that infants preferred to look more at the Agent Behind in the Chase–No Initiation event (M = .62, SE = .02) than at the Agent Ahead (M = .21, SE = .01), MDifference = .41, SE = .03, 95% CI [.34, .49], p < .001, but they did not look more at the Agent Ahead than at the Distractor (M = .17, SE = .02), MDifference = .04, SE = .03, 95% CI [–.03, 6.67], p = .357. Importantly, in the Follow–No Initiation event, infants looked equally at the Agent Behind (M = .46, SE = .03) and the Agent Ahead (M = .42, SE = .03), MDifference = .05, SE = .05, 95% CI [ .09, .18], p = 1.00, but infants continued to look more at the Agent Ahead than at the Distractor (M = .12, SE = .03), MDifference = .29, SE = .05, 95% CI [.18, .42], p < .001 (Fig. 5). The findings suggest that whereas the

Fig. 5. Average proportion of looking time at the Agent, Agent Ahead and Distractor areas of interest by Event type in 5-montholds (Study 3). Vertical lines indicate mean standard error. Chase–no initiation: *Agent Behind, Agent Ahead p = .0001; *Agent Behind, Distractor p = .0001. Follow–no initiation: *Agent Ahead, Distractor p = .0001.

66

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

chasing interaction continued to elicit preference to the chaser, the following interaction without one object initiating the interaction did not elicit the same effect. Comparison between Studies 2 and 3 To determine whether the total elimination of initial heat seeking influenced looking at the agents, a 2 (Event Type: Chase or Follow)  3 (AOI Type) repeated measures ANOVA with study (2 or 3) as a between-participant factor was performed. This analysis revealed a main effect of AOI type, F(2, 74) = 184.25, p < .001, g2 = .833, and a significant event type by AOI type interaction, F(2, 74) = 33.50, p < .001, g2 = .475. Importantly, there was a significant three-way interaction with event type, AOI type, and study, F(2, 74) = 3.86, p = .025, g2 = .095. Discussion Study 3 aimed to test whether heat-seeking motion initiating the interaction can account for the Agent Behind preference observed in Study 1 and 2. By removing the initial heat seeking, the preference for the Agent Behind over the Agent Ahead disappeared in the Follow event, where the two objects moved at the same speed and distance from each other. Correlational analysis also confirmed that the two proportions are negatively correlated, but looking at the Agent Behind and the Agent Ahead in the Follow event is at chance levels. The current findings (especially the comparison across Studies 2 and 3) suggest that infant attention over the entire trial seems to be affected by just a short period of time at the beginning of the trial (as shown in Studies 1 and 2). This change in gaze behavior speaks against the hypothesis suggesting a fundamental scanning strategy to look at the Agent Behind for all types of interactions. Rather, looking preference between individual agents appears to be modulated by different types of interactions. General discussion Recognizing causal relations between objects holds considerable adaptive value, allowing infants to gather important information from their visual environments (Adler, Inslicht, Rovee-Collier, & Gerhardstein, 1998; Colombo, McCollam, Coldren, Mitchell, & Rash, 1990; Frankenhuis & Barrett, 2013; Frankenhuis et al., 2013; Frankenhuis & Panchanathan, 2011). Prior developmental research examining infant attention to non-contact causal relations has largely focused on global preference measures, leaving current research literature in need of studies examining more specific processes such as the attention to individual objects within a single interaction. The aim of the current set of studies, therefore, was to examine online mechanisms of infant selective visual attention. In particular, we examined infants’ sensitivity to low-level motion cues and agent-specific tracking when observing self-propelled animated objects engaged in chasing and following interactions. In doing so, we used a new analytical method that combined measures of gaze behavior within a single event similar to the paradigm used with adults while examining the proportions of looking using eye tracking, an established measure of attention used with infants. Adult and infant attention to animated percepts The visual preference for specific agents that we report here is consistent with the increasing number of adult research studies on animacy perception. When observing simple geometrical shapes move across the screen, adult observers are able to interpret the movement as goal-directed while perceiving the moving shapes as intentional agents engaged in various types of social interaction (e.g., chasing, running away, stalking) (Gao & Scholl, 2011; Meyerhoff et al., 2013, 2014). Furthermore, when observing two objects chasing from among distractors, adults preferentially attend to the chasing object in particular and their ability to detect chasing is especially dependent on chaser-specific motion cues such as heat seeking and one object’s goal-directed motion toward another with 0° variation from the direct path (Gao et al., 2009).

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

67

The current findings are also consistent with developmental research. Similar to previous findings (Rochat et al., 1997), we demonstrated that when observing abstract animate objects move, 5- and 12month-old infants differentiate between events in which the objects interact and those in which objects move randomly and independently from each other. But unlike previous research, these conclusions are based on how infants attend to individual agents within each event rather than on global preference to different types of displays. Across the three studies, results demonstrate that at 5 months of age infants attend differentially between agents that are within interaction events, but they attend equally to agents that are not. We further demonstrated that the difference between looking at the interaction and non-interaction events cannot be accounted for by low-level visual or motion differences. First, all objects were visually identical. Second, when motion parameters in the non-interaction events were the same as those in interaction events, infants continued to look equally at all three agents (Study 2), allowing us to conclude that the difference between interaction events and non-interaction events has to do with the relational properties of the agents. We further found that very young infants differentiate between types of observed interactions. Across the three studies, proportional analyses demonstrated that both 5- and 12-month-olds preferentially attended to the chaser (Agent Behind) over the target of its chase (Agent Ahead) and that 5month-olds further attended more to the target (Agent Ahead) than to the Distractor that was not part of the interaction. This indicates that already by 5 months infants are sensitive to displays with goaldirected motion and specifically to the agent that initiates interaction in that manner. This was the case even though the chaser’s velocity remained constant, whereas the target demonstrated more visually varied acceleration bouts in order to successfully escape the chaser. Because this preference was a consistent feature across the three studies, it would seem that any type of goal-directed motion elicits visual attention. We found similar preference when the chaser initiated the interaction through heat seeking (Studies 1 and 2), switched the target of its pursuit (Study 1), and maintained the same target (Studies 2 and 3) and when the initiation of the interaction by the chaser was completely eliminated (Study 3). The current studies are to our knowledge the first to report an attentional bias toward the chaser within chasing interactions in an infant population. This bias is present already at 5 months of age, which is surprisingly early considering that understanding agent-specific roles and understanding intentionality have been assumed to develop during the second half of the first year (Csibra et al., 2003; Wagner & Carey, 2005). Sensitivity to goal-directed motion also seems to influence how infants attend to interactions other than chasing. When presented with a following interaction in which the two agents moved synchronously with each other rather than in an approach–escape fashion, infants continued to attend to the Agent Behind when it initiated the following interaction by heat seeking the Agent Ahead (Studies 1 and 2). Specifically, in Study 1 the proportional analyses revealed that both 12- and 5-month-old infants continued to look more at the Agent Behind than at the Agent Ahead and looked equally at the Agent Ahead and the Distractor, but these differences were significant only for the 12-month-olds.1 A similar trend was also observed in 5-month-olds, although approaching significance. Modulation of heat seeking in Study 2 was largely motivated by previous adult research suggesting that the ability to identify a chaser from among distractors may be especially evident in the wavering wolf paradigm following the wolf’s (or the chaser’s) target switch. Thus, one agent’s change in the intent of the goal-directed motion may highlight this agent’s motion. We found that for 5-montholds, modulation of the heat-seeking motion did not change the preference for the Agent Behind over the Agent Ahead in the following interaction. However, modulation for 5-month-olds was not completely ineffectual. Whereas in Study 1 the 5-month-olds did not statistically differentiate between the Agent Ahead and the Distractor, in Study 2 they seemed to do so, confirmed by the correlational analyses, which showed that they looked significantly above chance at the Agent Ahead compared with the Distractor. This indicates that eliminating the target switch gave infants a longer time to process the interaction between the Agent Behind and the Agent Ahead, whereas the Distractor remained outside the interaction and consequently was attended to the least. 1 Correlational analysis (see supplementary material) revealed that for both age groups during the Follow event, looking at the Agent Behind and Agent Ahead was negatively correlated, but only for the 5-month-olds was the proportion of looking at the Agent Behind above chance. The 12-month-olds looked at chance at the Agent Behind and Agent Ahead.

68

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

Finally, in Study 3 we found that complete elimination of the initial heat seeking by the Agent Behind changes how 5-month-olds attend to the interacting agents. We found that 5-month-olds continued to prioritize the chaser (Agent Behind) over the target (Agent Ahead) in the chasing interaction, but they did not seem to do so in the following interaction when the two agents moved synchronously together throughout the trial. Correlational analysis further confirmed that 5-month-olds tended to look at chance between the two agents in the following interaction. In all, The current findings suggest that very young infants differentiate between interactions and non-interactions based on motion cues between animate geometrical shapes. We supplement the existing literature in allowing making distinct conclusions about how infants attend to these objects within a single interaction. We found that like adults (Gao & Scholl, 2011; Gao et al., 2009), infants’ visual attention is affected by subtle changes in visual display details.2 It is still unclear whether 5-month-old infants perceive the interactions between geometrical shapes as specific kinds of interactions. We know that they are sensitive to agent-specific motion features such as goal-directed motion of one agent toward another (Frankenhuis et al., 2013). It is possible that equal attention to the objects in Study 3 in the following interaction was due to the synchrony of movement between the two agents and that they may have observed it as one entity when no other motion cue differentiates one over the other (Usher & Donnelly, 1998). Future research is needed to examine specifically how synchrony influences infant visual attention, that is, whether the same pattern of results would be evident when observing synchronous objects but not interacting objects. One technical, rather than conceptual, caveat when interpreting the data is the increased risk for falsely positive results given that, at any point and time, infants do not look at more than one AOI. To control for this, it is possible to assess the relationship between the proportions of looking at the individual agents within an interaction—specifically, whether looking at one individual AOI within an event is negatively correlated with another and which of the two relative proportions differs from chance. To examine this, bivariate correlations were performed for each of the significant post hoc tests. In cases where there were also significant correlations, we performed t-tests to assess which of the two relative proportions differs from chance. Importantly, all follow-up t-tests were in line with the ANOVA results, suggesting no bias toward false positives.3 The studies conducted here are a first effort in examining infant online gaze behavior to animate motion, but further research is necessary. For instance, by using the current analysis level, future research should examine the extent to which online visual attention relates to the formation of animated percepts in adult sense. One option would be to examine the correlation between the verbal reports of animated percepts with online gaze behavior in adults. Another option is to examine visual attention to interactions such as chasing within clinical populations such as adults and children with autism, who are found to have no difficulty in recognizing social and physical causality in schematic events but have difficulty in associating it to animate objects (Congiu, Schlottmann, & Ray, 2010; Rutherford, Pennington, & Rogers, 2006).

Conclusions Understanding how the visual system selectively derives information from motion patterns is fundamental to social cognition because it forms the foundation on which more complex social processes take place. Examining these fundamental processes during infancy holds great value because it allows for discrimination of agents toward whom developing social cognitive skills should be directed. Here, we showed that very young infants can differentially allocate visual attention between agents based on how they interact in real time, a possible precondition to social learning. 2 Correlational analysis from Study 1 revealed that slightly older infants are able to differentiate between interaction types with very few changes to the visual display. Younger infants, although not completely ignorant of the subtle changes, need more drastic differences, such as complete elimination of goal-directed movement in the following interaction (Study 3), to make the distinction. 3 For a full analysis of the bivariate correlations, see the supplementary material.

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

69

We conclude that when one agent initiates an interaction by heat seeking, infants as young as 5 months tend to frame the interaction in the same manner for both chasing and following conditions. When initial heat seeking is removed, 5-month-olds differentiate between the conditions. This suggests that the attribution of roles is processed very early in an interaction and is sustained even though other cues (e.g., the velocity profiles in the following event) suggest otherwise. In other words, ‘‘once a wolf, always a wolf” may have bearings also at the level of visual attention. To pay special attention to the chaser appears to be why detection of chasing is important. In a broader perspective, we showed that appropriate allocation of attention during motion-based interactions prompts infants to learning about social interactions in general. Thus, the current findings demonstrate attentional preconditions to social learning and that during the first months of life the visual system is already able to selectively attend to objects and agents in the environment whose motion signifies non-contact relations. Acknowledgments We thank the parents and infants for their enthusiastic cooperation. The authors also express gratitude to Gustaf Gredebäck and members of the Uppsala Child and Baby Lab for input on earlier versions of the manuscript as well as to the editor and anonymous reviewers. This research study was supported by the Swedish Research Council (VR-2011-1528). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi. org/10.1016/j.jecp.2016.02.010. References Adler, S. A., Inslicht, S., Rovee-Collier, C., & Gerhardstein, P. C. (1998). Perceptual asymmetry and memory retrieval in 3-monthold infants. Infant Behavior and Development, 21(2), 253–272. Biro, S., Csibra, G., & Gergely, G. (2007). The role of behavioral cues in understanding goal-directed actions in infancy. Progress in Brain Research, 164, 303–322. Bonatti, L., Frot, E., Zangl, R., & Mehler, J. (2002). The human first hypothesis: Identification of conspecifics and individuation of objects in the young infant. Cognitive Psychology, 44, 388–426. Castelli, F., Happé, F., Frith, U., & Frith, C. (2000). Movement and mind: A functional imaging study of perception and interpretation of complex intentional movement patterns. NeuroImage, 12, 314–325. Colombo, J., McCollam, K., Coldren, J. T., Mitchell, D. W., & Rash, S. J. (1990). Form categorization in 10-month-olds. Journal of Experimental Child Psychology, 49(2), 173–188. Congiu, S., Schlottmann, A., & Ray, E. (2010). Unimpaired perception of social and physical causality, but impaired perception of animacy in high functioning children with autism. Journal of Autism and Developmental Disorders, 40, 39–53. Csibra, G., Bíró, S., Koós, O., & Gergely, G. (2003). One-year-old infants use teleological representations of actions productively. Cognitive Science, 27, 111–133. Dittrich, W. H., & Lea, S. E. (1994). Visual perception of intentional motion. Perception, 23, 253–268. Frankenhuis, W. E., & Barrett, H. C. (2013). Design for learning: The case of chasing. In M. D. Rutherford & V. A. Kuhlmeier (Eds.), Social perception: Detection and interpretation of animacy, agency, and intention (pp. 171–195). Cambridge, MA: MIT Press. Frankenhuis, W. E., House, B., Barrett, H. C., & Johnson, S. P. (2013). Infants’ perception of chasing. Cognition, 126, 224–233. Frankenhuis, W. E., & Panchanathan, K. (2011). Individual differences in developmental plasticity may result from stochastic sampling. Perspectives on Psychological Science, 6, 336–347. Gao, T., Newman, G. E., & Scholl, B. J. (2009). The psychophysics of chasing: A case study in the perception of animacy. Cognitive Psychology, 59, 154–179. Gao, T., & Scholl, B. J. (2011). Chasing vs. stalking: Interrupting the perception of animacy. Journal of Experimental Psychology: Human Perception and Performance, 37, 669–684. Gergely, G., Nádasdy, Z., Csibra, G., & Bíró, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56, 165–193. Gredebäck, G., Johnson, S., & von Hofsten, C. (2010). Eye tracking in infancy research. Developmental Neuropsychology, 35, 1–19. Heider, F., & Simmel, M. (1944). An experimental study of apparent behavior. American Journal of Psychology, 57, 243–259. Retrieved from http://www.jstor.org/discover/10.2307/1416950?uid=3737880&uid=2&uid=4&sid=21102516677987. Johnson, S. C. (2003). Detecting agents. Philosophical Transactions of the Royal Society of London B: Biological Sciences, 358, 549–559. Johnson, S., Slaughter, V., & Carey, S. (1998). Whose gaze will infants follow? The elicitation of gaze following in 12-month-olds. Developmental Science, 1, 233–238. Káldy, Z., & Leslie, A. M. (2003). Identification of objects in 9-month-old infants: Integrating ‘‘what” and ‘‘where” information. Developmental Science, 6, 360–373.

70

M. Galazka, P. Nyström / Journal of Experimental Child Psychology 147 (2016) 53–70

Kanizsa, G., & Vicario, G. (1968). La percezione della reazione intenzionale. In G. Kanizsa & G. Vicario (Eds.), Ricerche sperimentali sulla percezione (pp. 71–126). Trieste, Italy: Università degli studi di Trieste. Leslie, A. M., Xu, F., Tremoulet, P. D., & Scholl, B. J. (1998). Indexing and the object concept: Developing ‘‘what” and ‘‘where” systems. Trends in Cognitive Sciences, 2, 10–18. Meyerhoff, H. S., Huff, M., & Schwan, S. (2013). Linking perceptual animacy to attention: Evidence from the chasing detection paradigm. Journal of Experimental Psychology: Human Perception and Performance, 39, 1003–1015. Meyerhoff, H. S., Schwan, S., & Huff, M. (2014). Perceptual animacy: Visual search for chasing objects among distractors. Journal of Experimental Psychology: Human Perception and Performance, 40, 702–717. Morris, M. W., & Peng, K. (1994). Culture and cause: American and Chinese attributions for social and physical events. Journal of Personality and Social Psychology, 67, 949–971. Premack, D. (1990). The infant’s theory of self-propelled objects. Cognition, 36, 1–16. Rochat, P., Morgan, R., & Carpenter, M. (1997). Young infants’ sensitivity to movement information specifying social causality. Cognitive Development, 12, 537–561. Rochat, P., Striano, T., & Morgan, R. (2004). Who is doing what to whom? Young infants’ developing sense of social causality in animated displays. Perception, 33, 355–369. Rutherford, M. D., Pennington, B. F., & Rogers, S. J. (2006). The perception of animacy in young children with autism. Journal of Autism and Developmental Disorders, 36, 983–992. Schlottmann, A., Ray, E. D., & Surian, L. (2012). Emerging perception of causality in action-and-reaction sequences from 4 to 6 months of age: Is it domain-specific? Journal of Experimental Child Psychology, 112, 208–230. Schlottmann, A., & Surian, L. (1999). Do 9-month-olds perceive causation-at-a-distance? Perception, 28, 1105–1113. Schlottmann, A., Surian, L., & Ray, E. (2009). Causal perception of action-and-reaction sequences in 8- to 10-month-olds. Journal of Experimental Child Psychology, 103, 87–107. Scholl, B. J., & Gao, T. (2013). Perceiving animacy and intentionality: Visual processing or higher-level judgment? In M. D. Rutherford & V. A. Kuhlmeier (Eds.), Social perception: Detection and interpretation of animacy, agency, and intention (pp. 197–229). Cambridge, MA: MIT Press. Surian, L., & Caldi, S. (2010). Infants’ individuation of agents and inert objects. Developmental Science, 13, 143–150. Usher, M., & Donnelly, N. (1998). Visual synchrony affects binding and segmentation in perception. Nature, 394, 179–182. Wagner, L., & Carey, S. (2005). 12-month-old infants represent probable endings of motion events. Infancy, 7, 73–83.