Infants’ perception of actions and situational constraints: An eye-tracking study

https://doi.org/10.1016/j.jecp.2012.11.014Get rights and content

Abstract

Rational action understanding requires that infants evaluate the efficiency of a movement in achieving a goal with respect to situational constraints. In contrast, recent accounts have highlighted the impact of perceptual characteristics of the demonstrated movement or constraints to explain infants’ behavior in so-called rational imitation tasks. The current study employed eye tracking to investigate how 13- to 15-month-old infants distribute their visual attention to different aspects of an action demonstration. In three tasks (touchlight, house, and obstacle), infants watched videos in which a model performed an unusual action while she was or was not restricted by situational constraints. Infants’ overall looking to the demonstration as well as looking to four segments of the video (initial segment, constraint demonstration, action performance, and final segment) and to specific areas (constraint area of interest [AOI] and action AOI) was analyzed. Overall, infants looked longer at the demonstration in the constraint condition compared with the no-constraint condition. The condition differences occurred in the two video segments where the constraint or action was displayed and were especially profound for the constraint AOI. These findings indicate that infants processed the situational constraints. However, the pattern of condition differences varied slightly in the three tasks. In sum, the data imply that infants process perceptual characteristics of the movement or constraints and that low-level perceptual processes interact with higher level cognitive processes in infants’ action perception.

Introduction

During the past two decades, research has profoundly extended our knowledge about how infants process and represent the actions of others. For instance, it has been shown that already in their first year, infants respond differently to the same observed action depending on the situational conditions in which the action is performed (see Gergely & Csibra, 2003, for a review). This suggests that although the physical movement, as a dynamic event, may attract most attention in an action display, infants also process the “background” or “surroundings” of the movement. This may include information about obstacles that block the movement as well as about end states or effects that are obtained (Elsner, 2007). Such evidence has been found in looking time studies, where infants’ gazes are interpreted as allocation of visual attention or as cognitive processing. Evidence also comes from imitation studies, where infants’ reproduction is interpreted as being the outcome of action perception and encoding (e.g., Gergely & Csibra, 2003).

Although these methods are valuable, it may be argued that they provide a rather coarse picture of infants’ cognitive processing because they do not allow for the measuring of attention to different aspects of the display. Eye tracking allows for such measurement; therefore, this method is becoming popular in infant research (Feng, 2011, Gredebäck et al., 2010). In the current study, we tracked infants’ eye movements to action demonstrations of imitation tasks to gain detailed insight into how 13- to 15-month-old infants process others’ actions and the situational constraints under which these actions are performed. To our knowledge, this is the first study measuring eye movements in imitation tasks.

Gergely, Nádasdy, Csibra, and Bíró (1995); see also Csibra, Gergely, Bíró, Koós, & Brockbank, 1999) provided the first evidence that infants’ action processing is dependent on situational constraints. In their looking time study, 12-month-olds were habituated to films showing a small circle performing a means (jumping over an obstacle) and achieving a goal state (a spatial position next to another figure). Then, infants saw two test events in which the situational context was changed. In detail, the obstacle was no longer present. Now, infants looked longer when the circle still performed the familiar jumping action as compared with when the circle performed a novel straight movement that was more efficient to achieve the goal state in the new situational context. From this and other findings, Gergely and Csibra (2003) concluded that by 12 months, infants consider the constraints of the situation and expect an agent to perform the action that is the most efficient means to the goal. This “principle of rational action” is supposed to be a fundamental component of infant cognition; thus, it should affect infants’ perception of others’ actions in various situational contexts.

To test whether infants also apply the rational action principle when reproducing an observed action, Gergely, Bekkering, and Király (2002) presented 14-month-olds with an adult performing an unusual goal-directed action in two conditions: The model bent forward and activated a lamp with a head touch either while her hands were occupied by holding a blanket (constraint condition) or while her hands were free and lying idly next to the lamp (no-constraint condition). When allowed to act on the lamp with their hands free, infants imitated the unusual head action more often in the no-constraint condition than in the constraint condition. However, all infants also used their hands to activate the lamp. Gergely and colleagues (2002) (see also Gergely & Csibra, 2005) concluded that infants see the head touch in the constraint condition as an effective means given the situational constraints because the model cannot use her hands. However, the infants do not face the same constraint; therefore, they choose the most effective means available to them (i.e., the hand action) when reproducing the model’s action. In contrast, in the no-constraint condition, the model could use her hands, but she uses her head instead, which is not the most efficient means. Because the model’s behavior does not adhere to the rational action principle, infants are led to assume that the model must have a special intention to demonstrate the unusual movement. Therefore, they engage in “rational imitation” (Gergely et al., 2002); that is, they perform the head action more often in the no-constraint condition than in the constraint condition.

To further explore the phenomenon of so-called rational imitation, Schwier, van Maanen, Carpenter, and Tomasello (2006) used a different task and presented 12-month-olds with the unusual action of a toy dog entering a house through the chimney. In the constraint condition the door of the house was closed, whereas in the no-constraint condition the door was open. Infants acted on the house with the door open, and many of them performed the effective action of putting the dog through the door. However, infants also imitated rationally in that they reproduced putting the dog through the chimney more often in the no-constraint condition than in the constraint condition. Schwier and colleagues took this as evidence that by 12 months, the rational action principle influences infants’ reproduction of others’ actions in various situations.

Zmyj, Daum, and Aschersleben (2009) found more reproduction of the head touch in the no-constraint condition than in a constraint condition of the touchlight task (Gergely et al., 2002) for 12-month-olds but not for 9-month-olds. However, the original constraint condition, in which the model’s hands were occupied by holding a blanket, was not effective for the 12-month-olds. Infants showed a lower number of head touch actions only in a constraint condition where the model’s hands were visibly tied to the table. Zmyj and colleagues argued that the tied hands represented an explicit and non-voluntary constraint and that 12-month-olds would be able to detect this type of constraint. In contrast, holding the blanket represented an implicit and voluntary constraint that would not be detected by 12-month-olds. This implies that situational constraints may vary in the demands that they impose on infants’ cognitive processing.

Recently, the idea of rational imitation has been challenged, and alternative accounts have been proposed to explain the differences in infants’ behavior in the two conditions of the touchlight task. Paulus et al., 2011a, Paulus et al., 2011b argued that infants do not need to perform an evaluation of the rationality of the observed action within the constraints of the situation. Instead, they assume that the two conditions differ with respect to the motor resonance that is induced by the observed movement. In detail, due to their immature balance and motor system, 14-month-olds need to put their hands next to the lamp when performing the head touch. Therefore, the head touch in the no-constraint condition, where the model’s hands are lying on the table, provides a better match to the infants’ own movement than in the constraint condition, where the model’s hands are being held in front of her chest. Because of the larger motor resonance, infants imitate the head action more often in the no-constraint condition than in the constraint condition. Hence, Paulus and colleagues assumed that infants’ imitation is based mainly on the characteristics of the movement and that the constraints have little or no importance.

Contrary to this, Beisert et al. (2012) attributed the differences in infants’ imitation to perceptual differences of the two conditions in the touchlight task. In detail, the blanket covering the model’s whole body in the constraint condition is seen as an “eye-catching outfit” that may distract infants’ attention from the head touch movement. In contrast, infants’ attention should be less distracted in the no-constraint condition when the blanket is only loosely put over the model’s shoulders. Therefore, infants would show better encoding, and hence more imitation, of the head touch in the no-constraint condition than in the constraint condition. Beisert and colleagues showed that imitation of the head touch increased when 14-month-olds were familiarized to the model’s body covered by the blanket and that imitation was decreased when eye-catching stimuli were added to the hands-free condition. These findings support the notion that infants’ imitation is affected by the perceptual characteristics of the constraints.

A crucial problem with the studies reported so far is that inferences about the involved cognitive processes need to be made on the basis of infants’ overall looking to video events or of infants’ reproduction of observed actions. As a possible solution, eye tracking would allow for the recording of gaze patterns in more detail and, thus, would reveal at which aspects of the display infants are looking and for how long (Corbetta et al., 2012, Gredebäck et al., 2010). Infants’ gaze pattern reflects the allocation of visual attention, and the duration of fixations to a target indicates the processing load provided by the stimulus (Feng, 2011). Infants’ gaze allocation limits the information input and, hence, influences the perception and the resulting representation of an observed event (Johnson, Slemmer, & Amso, 2004). Therefore, tracking eye movements while infants observe others’ actions could indicate how infants allocate their attention to, or how they process, the demonstrated movement and the situational constraints.

In the current study, we measured eye movements of 13- to 15-month-olds while they watched action demonstrations of three tasks: the touchlight task (Gergely et al., 2002), the house task (Schwier et al., 2006), and a novel task that was designed after Gergely and colleagues (1995). In this “obstacle task,” the model made a toy bear perform an unfamiliar movement to achieve a goal state. Like the small circle in the stimuli used by Gergely and colleagues (1995), the teddy bear took a big leap to reach a cup. In the constraint condition, the leap was justified by the situational conditions because an obstacle blocked the bear’s path. In the no-constraint condition, the obstacle was not present; therefore, taking the leap was not the most efficient means to achieve the goal state. According to the rational action principle (Gergely & Csibra, 2003), infants should realize the effectiveness of the movement in the two conditions of the obstacle task, leading to longer looking to the no-constraint condition.

The infants watched videos demonstrating the actions of all three tasks either in the constraint condition or in the no-constraint condition. Our first question was whether infants’ looking to the videos differed in the two conditions. We predicted that the presence of a constraint might increase the processing load for the display because the constraint, and also the relation between the constraint and the action, need to be processed. Our second question was when exactly infants in the two conditions looked to the display. Here, we compared looking times in four segments of the video films: at the beginning, when the constraint was introduced, when the action was performed, and at the end. We expected differences between the two conditions especially during the constraint demonstration segment and the action performance segment because here is when important information for action processing is displayed. Our third question was to where exactly infants in the two conditions looked during the different segments of the video. We were especially interested in how long infants gazed at the action and at the constraint during those segments in which these aspects were relevant. Fourth, we asked whether the presence or absence of situational constraints had similar effects on infants’ looking behavior across the three tasks. If we found task-dependent processing of the constraints, this would challenge the current theoretical accounts because these try to explain infants’ behavior by general principles.

It is interesting to note that the theoretical accounts described above would make different predictions about infants’ looking in the two conditions. According to the rational action principle (Gergely & Csibra, 2003), infants should look longer in the no-constraint condition than in the constraint condition. This is because in the former condition the unfamiliar movement is not justified by the situational constraints, and looking time studies suggest that this induces higher processing load (Csibra et al., 1999, Gergely et al., 1995). In contrast, according to the perceptual distraction account (Beisert et al., 2012), infants should look longer to the constraint condition because their attention is attracted by the presence of the constraint. It may also be that the distraction does not become obvious in differences in overall looking time but only in the proportional looking times to the constraint or to the action. If so, infants in the constraint condition should look longer to the constraint area and look less to the action area than infants in the no-constraint condition. From the motor resonance account (Paulus et al., 2011a, Paulus et al., 2011b), it is difficult to derive hypotheses about infants’ looking behavior. If anything, this account allows for predictions only for the touchlight task. In the other two tasks, movements are acted out with a puppet, so it is difficult to estimate the elicited motor resonance. It may be predicted that infants either should look equally long to both conditions because they do not process the situational constraints or should look longer to the constraint condition because the head touch with the hands in front of the chest is unfamiliar to them and, thus, may require more processing. In the current study, eye tracking allowed not only for testing these different predictions of infants’ looking in the two conditions but also for evaluating whether the same explanation holds for infants’ action processing in all three tasks.

Section snippets

Participants

In total, 36 13- to 15-month-old infants (mean age = 13 months 21 days [13;21], range = 12;29–15;00, 18 girls and 18 boys) were included in the final sample (n = 18 in each condition). An additional 9 infants (3 girls and 6 boys) were excluded from analyses due to technical difficulties (n = 2) or failure to pay sustained attention to the repeated video demonstrations within a given task (n = 7). All infants were full-term and healthy. Parents were contacted by phone and signed a consent form prior to

Results

Although infants’ imitation was also assessed,

Discussion

The current study investigated infants’ looking behavior during action demonstration in so-called rational imitation tasks by means of eye tracking. The tasks were the touchlight task by Gergely and colleagues (2002), the house task by Schwier and colleagues (2006), and a novel obstacle task designed after Gergely and colleagues (1995). The recorded gaze data show that 13- to 15-month-old infants distribute their visual attention to the video displays of the demonstrated actions differently

Acknowledgments

This research was supported by Grants from the Canada Foundation for Innovation (CFI, 203229) and the Natural Sciences and Engineering Research Council of Canada (NSERC, 322013) to Petra Hauf. The data were collected at St. Francis Xavier University. We thank the parents and infants who made this work possible. Furthermore, we are grateful to Kailee MacDonald, Ali Butler, and Sara Murphy for assistance with data collection and coding.

References (17)

There are more references available in the full text version of this article.

Cited by (0)

1

Current address: Department of Psychology, Dalhousie University, Halifax, Nova Scotia, Canada B3H 4R2.

View full text