Action perception in infancy: the plasticity of 7-month-olds’ attention to grasping actions

Daum, Moritz M.; Wronski, Caroline; Harms, Annekatrin; Gredebäck, Gustaf

doi:10.1007/s00221-016-4651-3

Action perception in infancy: the plasticity of 7-month-olds’ attention to grasping actions

Research Article
Published: 19 April 2016

Volume 234, pages 2465–2478, (2016)
Cite this article

Download PDF

Access provided by Autonomous University of Puebla

Experimental Brain Research Aims and scope Submit manuscript

Action perception in infancy: the plasticity of 7-month-olds’ attention to grasping actions

Download PDF

Moritz M. Daum^1,2,
Caroline Wronski^1,2^nAff4,
Annekatrin Harms² &
…
Gustaf Gredebäck³

614 Accesses
6 Citations
1 Altmetric
Explore all metrics

Abstract

The present study investigates the plasticity of 7-month-old infants’ orienting of attention during their perception of grasping actions. Previous research has shown that when infants observe a grasping hand, they shift their attention in line with the grasping direction, which is indicated by a reliable priming effect in this direction. The mechanisms behind this priming effect are largely unknown, and it is unclear how malleable this priming effect is with respect to a brief exposure to novel action-target contingencies. In a spatial-cueing paradigm, we presented a series of training trials prior to a series of test trials. These training sequences significantly modulated infants’ attention. This suggests that action perception, when assessed through shifts of attention, is not solely based on the infants’ grasping experience but quickly adapts to context-specific observed regularities.

Learning to grasp and extract affordances: the Integrated Learning of Grasps and Affordances (ILGA) model

Article Open access 19 November 2015

Touch the table before the target: contact with an underlying surface may assist the development of precise visually controlled reach and grasp movements in human infants

Article 24 May 2018

Press to grasp: how action dynamics shape object categorization

Article 08 December 2015

Use our pre-submission checklist

Avoid common mistakes on your manuscript.

Introduction

Infants at an early stage of life already interpret observed actions as goal-directed (e.g. Woodward 2009). To a large extent, their ability to interpret the social environment is based on experiences made through everyday interactions. These experiences are acquired either through their own action performance (e.g. von Hofsten 2007) or through observed statistical regularities of actions performed by others (e.g. Wu et al. 2011). The present research examined whether infants’ perception of manual grasping actions is malleable through a brief observational experience.

Manual grasping is one of the most fundamental action skills acquired early in life because grasping provides the first opportunity to manipulate and reshape the environment (Bertenthal and Clifton 1998; Flanagan and Johansson 2002). A few days after birth, infants aim their arm movements towards interesting objects (von Hofsten 1982), and by the age of 5–7 months, grasping has become proficient enough that infants extrapolate object motions on linear paths and adapt their grasping movement accordingly (von Hofsten et al. 1998).

At around the same age, infants develop a remarkable sensitivity to grasping actions performed by others (e.g. Woodward 2009). This sensitivity is assessed using a variety of research designs and dependent measures tapping into a number of different cognitive processes involved in action perception (e.g. Gredebäck and Daum 2015). Some of these processes take place over a few hundred milliseconds (Gredebäck et al. 2010a). For example, overt shifts of attention (e.g. as found in predictive gaze shifts; Falck-Ytter et al. 2006; Gampe et al. 2015) or covert shifts of attention (Gredebäck et al. 2010b; Daum and Gredebäck 2011a) usually unfold within the first few hundred milliseconds after a directional action or gesture is observed (Gredebäck et al. 2010a). Other processes, for example as assessed via the measurement of looking times, extend over several seconds (Woodward 1998; Daum et al. 2012) .

A vast amount of research relied on measuring infants’ looking time in response to action outcomes that are in agreement with, or violate, their expectations—referred to here as action evaluation. Research into infants’ action evaluation has shown that 6-month-olds expect grasping actions to be directed towards goals (Woodward 1998), even if only initial parts of the action are presented (Daum et al. 2008; Brandone and Wellman 2009). Research techniques to measure action evaluation have recently been complemented with measures that target infants’ abilities to assess future states of perceived actions—which is referred to as action prediction (Gredebäck and Daum 2015; Gredebäck and Falck-Ytter 2015). Action prediction studies analyse the latency of predictive gaze shifts and have shown that infants start to predict the goal of observed grasping actions at between 6 and 12 months of age (Falck-Ytter et al. 2006; Hunnius and Bekkering 2010; Kanakogi and Itakura 2011; Cannon and Woodward 2012; Melzer et al. 2012).

In the present study, infants’ action perception was assessed in an action priming paradigm; that is, via the shifts of their attention during the observation of an action. Action priming is a very early component during the processing of observed actions. It occurs as soon as a social agent has been identified, and results in a shift of attention in the direction of the agent’s potential goal—even without an actual goal being visible. In contrast to both action evaluation and prediction, infants have less time and information available to process the relation between agent and goal. Accordingly, while a number of action evaluation (e.g. Woodward 1998) and action prediction studies (Gredebäck et al. 2009; Krogh-Jespersen et al. 2015) have shown that infants interpret grasping actions as being related to particular goals, action priming assesses an earlier processing phase where infants process the observed grasping action as being directed towards a particular location in space, potentially to interact with an object that is located at this position. This modulation of attention in the direction of others’ actions occurs within milliseconds and even before overt responses (such as eye movements) are observable. One way of assessing these early attentional processes caused by goal-directed agents is the application of a spatial-cueing paradigm (as introduced by Posner 1980). The logic behind this paradigm is that if an observer identifies the directionality of a symbolic cue such as an arrow (as used e.g. by Posner 1980) or a more socially relevant cue such as a grasping hand (e.g. Daum and Gredebäck 2011b), she will be faster at detecting a target that appears at a location that is congruent with the cue’s direction than an object at an incongruent location (this difference between reaction times, i.e. faster reaction times in congruent than in incongruent trials, is referred to hereinafter as priming effect). In infancy, priming effects have previously been reported during the perception of gaze shifts (Hood et al. 1998; Senju et al. 2006), grasping gestures (e.g. Daum and Gredebäck 2011b), and pointing gestures (Rohlfing et al. 2012; Daum et al. 2013; Bertenthal et al. 2014).

Previous research has shown that besides their sensitivity to social cues (e.g. gaze, pointing, grasping), infants evince priming effects caused by differences in visual properties of a stimulus (Jakobsen et al. 2013). Jakobsen et al. found that infants already shift their visual attention in the direction of an arrow, resulting in a significant priming effect. However, in this case the priming effect was caused by the perceptual properties of the arrow (i.e. that the head of the arrow is perceptually more salient than its end) and not by the conceptual meaning of the arrow, which is learned much later in life. Priming studies using more socially relevant cues such as pointing or a grasping hand (Daum and Gredebäck 2011a; Bertenthal et al. 2014) showed that the priming effect is not caused by the perceptual properties of the hand but rather by social content and an early step in a long chain of action perception processes (Gredebäck and Daum 2015).

Two priming studies that are of particular relevance for the present purpose have demonstrated a priming effect for grasping actions (Daum and Gredebäck 2011a; Wronski and Daum 2014). In both studies, infants were presented with a grasping hand (cue) followed by a toy (target) that appeared either in the direction of grasping (congruent trials) or in the opposite direction (incongruent trials). When the cue consisted of a static picture of a grasping hand, children at the age of 7 months showed a reliable priming effect (Daum and Gredebäck 2011a). This priming effect was already evident in 5-month-olds when the cue was a naturally moving grasping hand (Wronski and Daum 2014). Together, the two studies demonstrate that at the age of 7 months, infants shift their attention in the direction of the observed grasping action. This suggests that at the age of 7 months, infants perceive a grasping hand as being directed towards a location in space along the extension of the grasping action. Because this action priming occurs without any familiarisation during the experimental procedure, we refer to this robust effect as infants’ prior expectation about grasping hands and to the resulting priming effect as the default priming effect that occurs when 7-month-old infants observe a grasping hand.

The close temporal relation in development between perception and performance of grasping actions (for a recent study of the neural correlates of this close link between action priming and grasping proficiency, see Bakker et al. 2015) has led to the assumption that an observed action is mapped onto the observer’s motor representation of the action, which allows the observer to generate predictions about the future consequences of the observed action in the same way as for one’s own actions (Jeannerod 2001; e.g. Gallese 2009). The assumption that this mechanism is already functional early in life has received support from findings indicating that grasping perception is correlated with infants’ own productive grasping performance (Daum et al. 2011; Kanakogi and Itakura 2011; Melzer et al. 2012), that infants’ motor experience is causally related to their grasping perception and action priming (Sommerville et al. 2005; Bakker et al. 2016), and that the infant motor cortex is active during the perception and production of grasping actions (Nyström 2008; Southgate et al. 2009; Elsner et al. 2013). Accordingly, infants’ lifetime experience of grasping performance is an important source for their perception of others’ grasping actions.

However, research in social-cognitive development has shown that infants’ action perception is not only based on their action experience but is supported by alternative sources as well (e.g. Hunnius and Bekkering 2014). Infants interpret actions as being goal-directed actions which they are not yet able to perform when, for example, said actions are performed with tools they cannot yet use (Hofer et al. 2005; Biro and Leslie 2007; Boyer et al. 2011; Southgate and Begus 2013), when they are performed by inanimate agents such as balls (Gergely et al. 1995), toy figures (Luo and Baillargeon 2005), or when they are performed by schematic animals (Schlottmann and Ray 2010; Daum et al. 2012). In all cases, the observed behaviour cannot be mapped onto the infants’ motor repertoire and they need to rely on alternative interpretational mechanisms. While some authors explain the early goal attribution by means of inborn principles and biases, another potential candidate for this is learning through observed statistical regularities. This acquisition of knowledge through statistical learning is involved in a variety of developing skills such as categorisation (Mareschal et al. 2002) and word learning (Estes et al. 2007). With respect to the perception of goal-directed actions, previous research suggests that, around their first birthday, infants use statistical regularities for their goal predictions (Henrichs et al. 2013) and they also revise their initial predictions based on observed statistical regularities if they observe unsuccessful actions (Brandone et al. 2014). The implementation of statistical regularities of observed actions therefore plays an important role in infants’ interpretation of observed events.

The goal of the present study was to test the plasticity of infants’ action perception through observed statistical regularities. Previous research has shown that action observation does not substantially alter infants’ subsequent action perception (Sommerville et al. 2012; Cannon et al. 2012) or their action evaluation (Sommerville and Woodward 2005). But so far, no prior research has tested the plasticity of action perception through action priming in infants. When compared to action priming, action evaluation and prediction differ in the important respect that they require not only the identification of the directionality of a cue but also a more in-depth processing of the observed action in relation to an observable goal. In contrast, action priming merely involves the rapid orientation of attention in the direction of an observed action cue without the relation to an observable goal. We therefore tested the assumption that action priming is a more flexible process than action prediction and evaluation. To do so, we used the same laboratory set-up and paradigm as the two previous studies mentioned above that assessed priming effects in grasping actions (Daum and Gredebäck 2011a; Wronski and Daum 2014). In addition, infants were presented with a brief training sequence at the beginning of the experiment. In this training phase, the target consistently appeared either at the congruent or at the incongruent location. In a subsequent test phase, the infants saw a sequence of test trials in which congruent and incongruent trials were presented in a random order and equal distribution.

Based on the previous findings discussed above, two potential outcomes can be expected: first, if action priming is still a flexible process at the age of 7 months, a short training sequence is expected to affect infants’ attention. This effect of training should become evident, in particular, after incongruent training in which the relation between cue and target is novel, so where it is in conflict with the infants’ prior expectation about the directionality of a grasping hand. Second, and in contrast with the hypothesis formulated above, if action priming is largely rooted in the infants’ lifetime experience of grasping (which, by the age of 7 months, is about 4 months), a short sequence of incongruent trials is not expected to affect the infants’ prior expectation and the corresponding action priming. In this case, the different training sequences are not expected to modify the default priming effect.

Finally, in case a short training sequence does have an effect on the subsequent test trials, the question remains to which extent this effect remains is stable or will be affected by the additional information of the random distribution of congruent an incongruent trials during the test phase. To explore the stability of a potential training effect, we conducted an exploratory analysis of potential changes of the infants’ responses over the time course of the experimental test trials.

Experiment 1

In Experiment 1, 7-month-old infants were presented with a spatial-cueing paradigm (adapted from Daum and Gredebäck 2011a). The cue consisted of the static picture of a grasping hand followed by a target at a congruent or incongruent location. The experimental procedure included a training phase and a subsequent test phase. In the training phase, one group of infants received a series of congruent trials; another group of different infants received a series of incongruent trials. During the test phase, congruent and incongruent trials were presented in a random order and equal distribution.

Method

Participants

The final sample consisted of 48 infants aged 7 months (26 girls, 22 boys; mean age: 7 months; 3 days, range: 6;19–7;14). In addition, 26 infants aged 7 months (12 girls, 14 boys) were tested but not included in the final sample due to distress or restlessness (n = 3), technical problems (n = 1), or fewer than 6 valid training or test trials (n = 18). Contact information was obtained from public birth records. All infants were born full term (37- to 42-week gestation) and with normal birth weight (>2500 g). All parents gave informed consent before the study. The study was approved by the local ethics committee and was conducted in accordance with the Declaration of Helsinki.

Test environment, stimuli, and apparatus

The laboratory was unfurnished except for the test equipment. The stimuli were presented, and gaze was measured using a Tobii 1750 near-infrared eye tracker (Tobii Technology AB, Danderyd, Sweden) with an infant add-on (precision: 1°, accuracy: .5°, sampling rate: 50 Hz) and the software ClearView (version 2.7.1). A 9-point infant calibration was used. During calibration, a blue-and-white sphere expanded and contracted (extended diameter = 3.3°) in synchrony with a sound. Viewing distance was approximately 60 cm, and monitor diameter was 17 inch (24.8 × 20.7°). The infants were seated in a car safety seat (Maxi Cosi Cabrio) in front of the eye tracker.

The stimulus presentation followed closely the one used in Daum and Gredebäck (2011a, Experiment 1), see Fig. 1. Each trial started with one of four looming toys (horizontal and vertical dimensions: maximum 4.5°, minimum 2.3°) presented at the centre of the monitor accompanied by a sound. As soon as the infant fixated the central stimulus, a hand (cue, 5.0° × 4.6°) was presented grasping towards either the left or the right for 1000 ms. The cue was followed by a renewed presentation of the initially presented toy (now referred to as target). The target appeared at a congruent location (in the same direction as the grasping hand, see Fig. 1, upper panel) or an incongruent location (in the opposite direction to the grasping hand, see Fig. 1, lower panel). The distance from the nearest edge of the cue to the target was 9.3°. The target remained visible until the infant looked at it for approximately 1000 ms or until 5000 ms had elapsed. Then, a new trial began with another centrally presented looming stimulus.

The general paradigm consisted of a training phase and a subsequent test phase. In the training phase, two conditions were presented in a between-subjects design. In the congruent training condition, the targets always appeared at the grasping-congruent location, while in the incongruent training condition, the target always appeared at the grasping-incongruent location. In the test phase, the relation between grasping direction and the location of the target appearance was no longer predictable. The order of the targets as well as the relation between the direction of the grasping hand and the location of the target was pseudo-randomised. In order to avoid adaptation effects that might potentially override the specific manipulation of the training, no more than three repetitions of cue direction, target location, or cue–target relation were allowed. In the training phase, a total of 12 trials were presented to each infant. The maximum number of test trials presented was 64. Infants were randomly assigned to one of the two training conditions and to one of four different orders in the test phase.

Procedure

The infants were tested in the laboratory at a time of day when they were likely to be alert and in a good mood. All infants were tested individually with one parent present. Each participant and his/her parents were first escorted to a reception room. For approximately 10 min, the infant was allowed to explore the room, while the research assistant described the test procedure to the parents, and one of the parents signed a consent form. Then, the infant and one parent were brought to the test room. The research assistant helped the parent to position the infant in the car seat. During stimulus presentation, the parent sat on a chair behind the infant. Parents were instructed not to interact with their children during testing. They were encouraged, however, to put both hands symmetrically close to the child if it appeared necessary to comfort the infant. Once infant and parent seemed comfortable, the research assistant left the room and the stimulus presentation started.

Data analysis

For gaze analysis, three square areas of interest (AOI) were defined on the screen covering the cue (cue AOI; horizontal and vertical dimension: 7.5°) and the two target locations (target AOIs; horizontal and vertical dimension: 4.7°). A trial was considered valid if the infant fixated the central cue for at least 200 ms (Gredebäck et al. 2006) prior to making a gaze shift to the target. The saccadic reaction time (SRT) was defined as the difference in time between the appearance of the target and the arrival of the gaze at the respective target AOI (Gredebäck et al. 2010a). Individual SRTs less than 100 ms and greater than two standard deviations of each individual mean were excluded from analysis. Infants had to provide at least 6 training trials and 6 test trials (with a minimum of three trials per type [congruent, incongruent]) to be included in the final analysis. This criterion was based on previous studies where similar criteria were used as inclusion rates and similar minimum numbers of trials were achieved (Hood et al. 1998; Daum and Gredebäck 2011a; Rohlfing et al. 2012). Since we had no particular hypothesis regarding effects of sex and order of presentation, and preliminary analyses indicated no main effects of these two factors (all ps > .11), the data were collapsed for the remaining analysis over these factors.

The main analysis was performed in three steps: (1) the number of trials performed by each infant in the respective conditions was evaluated. (2) Mean SRTs were analysed separately for training and test trials. In the training trials, the mean SRTs were analysed using a univariate analysis of variance (ANOVA) with training (congruent, incongruent) as between-subjects variable. In the test trials, the mean SRTs in congruent and incongruent trials were analysed using a 2 × 2 (training [congruent, incongruent], trial type [congruent, incongruent]) repeated-measures analysis of variance (ANOVA), with trial as a within-subjects and training as a between-subjects variable. (3) SRTs were then compared for reliable priming effects per condition. To better compare the groups of infants in the two training conditions and to compare the results of the two experiments reported in this manuscript, the proportional SRT was calculated by dividing the SRT in congruent trials by the SRT in incongruent trials. A number smaller than 1 indicate the default priming effect (i.e. shorter SRTs in congruent than in incongruent trials); a number greater than 1 indicate a reversed priming effect (i.e. longer SRTs in congruent than in incongruent trials). The proportional SRTs were analysed using a one-way ANOVA with training as a between-subjects factor. The proportional SRTs of the different conditions were further analysed discretely by separate t tests for each training condition. (4) In an exploratory analysis, potential changes of the SRTs over trials were explored using linear regression analyses. The main results of both Experiments 1 and 2 are displayed in Fig. 2; p values are reported two-tailed throughout.

Finally, to explore the stability of the potential training effect, changes of SRTs and proportional SRTs over the course of the experimental trials were explored via two additional analyses: first, separate linear regression analyses of the SRTs in congruent trials, incongruent trials, and over all trails, independent of condition, and second, an analysis of the proportional SRTs over the course of the experimental trials. For this analysis, we divided the 64 test trials into four quartiles, calculated the mean proportional SRT for each participant, and analysed these values using a 4 × 2 (quartile [Q1, Q2, Q3, Q4] × experimental condition [congruent training, incongruent training]) ANOVA with quartile as within-subjects and experimental condition as between-subjects factor. We want to emphasise that due to the decreasing number of valid trials over the course of the test phase (see the analysis of the number of trials in “Results” section), this analysis is of a merely exploratory nature and the results need to be considered with care.

Results

Number of trials

In the congruent training condition, infants contributed an average number of 10.4 training trials (SD = 1.8, range 6–12) and 34.5 test trials (SD = 16.2, range 8–60). In the incongruent training condition, infants contributed an average number of 10.2 training trials (SD = 2.1, range 6–12) and 25.5 test trials (SD = 13.2, range 8–55). The number of training trials contributed did not differ between conditions, t(46) = .38, p = .71, but infants performed more test trials in the congruent training condition than in the incongruent training condition, t(46) = 2.1, p = .04.^{Footnote 1} A linear regression analysis showed that the number of valid data points per trial (combined for congruent and incongruent trials) decreased over the course of the experiment, both during the congruent training condition, β = −.89, t(63) = −15.34, p < .001, from M(first 5 trials) = 18.2 data points to M(last 5 trials) = 5.4 data points, and the incongruent training condition, β = −.94, t(63) = −22.24, p < .001, from M = 17.2 data points to M = 2 data points.

Mean SRTs

No between-subjects effects of training condition were found, neither in the training trials, F < 1, nor in the test trials, F < 1. Furthermore, in the test trials, the ANOVA resulted in a significant interaction between training condition (congruent, incongruent) and trial (congruent, incongruent), F(1, 46) = 5.78, p < .05, η ² = .11. After congruent training, SRTs were shorter in congruent than in incongruent trials (mean SRTs: congruent trials: M = 441.13 ms, SD = 110.60 ms; incongruent trials: M = 470.50 ms, SD = 123.33 ms), while after incongruent training the pattern was reversed and SRTs were shorter in incongruent than in congruent trials (mean SRTs: congruent trials: M = 468.62 ms, SD = 114.24 ms; incongruent trials: M = 445.68 ms, SD = 71.71 ms). This effect will be explored further in the next section.

Proportional SRTs

The one-way ANOVA on the proportional SRTs revealed a significant effect of training, F(1, 46) = 6.26, p < .05, η ² = .12. The proportional SRTs differed significantly between the two groups, which indicates that the training sequence had a significant influence on the priming effect (see Fig. 2). As expected, infants in the congruent training condition showed a default priming effect, indicated by a proportional SRT that was significantly smaller than 1 (M = .94, SD = .12), t(23) = 2.35, p < .05. In contrast, infants in the incongruent training condition showed a proportional SRT that was numerically greater but not statistically different from 1 (M = 1.05, SD = .17), t(21) = 1.42, p = .17.

Exploratory analysis of SRTs over time

We first calculated the mean SRT per trial (for both conditions combined and separately for congruent and incongruent trials) and conducted a linear regression analysis over the course of the experiments (Bonferroni corrected for multiple (n = 6) comparisons, adjusted alpha level: p = .008). In the congruent training condition, the SRTs did not change over the course of the test trials (all βs < .09, all ts < .61, all ps > .54). In the incongruent training condition, we found that SRTs overall, β = .32, t(39) = 2.11, p = .04, tended to increase over trials, but this tendency did not survive the correction for multiple comparisons. Both, SRTs in congruent trials, β = .24, t(39) = 1.55, p = .13, and incongruent trials, β = .06, t(39) = .40, p = .69, did not change over the course of the trials.

Next, we analysed whether the proportional SRTs changed over the course of the test trials. This analysis revealed no significant main effect or interaction, all Fs < 1.95, all ps > .13. Because the number of missing trials was high, in particular in the third and fourth quartile, we additionally ran the same analysis with a set of data in which we replaced the missing values with the grand mean value calculated over both experimental conditions of Experiment 1, M(grand) = 1.00 (sic!). This analysis resulted in a trend towards a main effect of experimental condition, F(1, 46) = 3.43, p = .07, η ² = .07, primarily reflecting the above-reported tendency that the proportional SRTs were larger in the incongruent training condition (Q1: M = 1.02, Q2: M = 1.03, Q3: M = 1.03; Q4: M = 1.03) than in the congruent training condition (Q1: M = .94, Q2: M = .95, Q3: M = 1.02; Q4: M = .98). No main effect of quartile was present, F < 1, nor was there an interaction between quartile and experimental condition, F < 1. In all, the difference between the two experimental conditions was stable across the course of the test trials of Experiment 1.

Discussion

Previous research has shown that with no additional training phase, 7-month-old infants shift their gaze earlier towards a target that appears in the direction of an observed grasping action (Daum and Gredebäck 2011a). The results of Experiment 1 demonstrate that this default priming effect, in which the children were assumed to show prior to the current experimental session, can be modified by a short sequence of incongruent training trials. From the observed statistical regularities during the training phase, the infants rapidly altered their prior expectation about the relation between cue and target. This modification of their expectation had a substantial impact on the priming effect, and this impact was maintained over the course of the experimental trials. In contrast, when presented with a series of congruent training trials, the infants continued to show the previously reported default priming effect (as reported in Daum and Gredebäck 2011a). We will come back to this aspect in “General discussion” section.

Two aspects of these results need further attention: the first aspect concerns the strength of the infants’ representation of the observed grasping action and the resulting expectation about the grasping direction. In Experiment 1, the stimulus material consisted of the static picture of a grasping hand. Research with adults has shown that static images of actions indeed result in premotor and motor cortex activity (e.g. Proverbio et al. 2009). However, it was also shown that dynamic actions trigger the respective motor representation more strongly (e.g. Buccino et al. 2001). Accordingly, the observation of a static hand might only have weakly triggered the infants’ motor representation of grasping. This could have resulted in a rather fragile default priming effect, which is strong enough to become evident without prior training but which can easily be modified by observed statistical regularities. The motor representation might be more strongly triggered by the observation of a more naturalistic grasping action. This stronger activation might then, in turn, hamper the rapid learning through observed statistical regularities, resulting in a smaller (or no) effect of the incongruent training on the infants’ prior expectations than in Experiment 1. In contrast, and as discussed in the introduction, learning is facilitated by the consistently observed violations of the infants’ expectation (Stahl and Feigenson 2015). In the present study, the rapid observational learning might have been facilitated by the consistently observed violations of the infants’ expectation during incongruent training. However, when presented with the picture of a static hand, the expectation about the grasping direction might be less strong when triggered by a static compared to a dynamic hand. Hence, this expectation might be more strongly violated when a more natural dynamic grasping action is observed than when only static action pictures are shown. In this case, a short incongruent training phase would enlarge the effect of the incongruent training effect as observed in Experiment 1.

The second aspect concerns the fact that in Experiment 1, the training resulted in a significant modification of the default priming effect^{Footnote 2} but did not result in a reversed priming effect. Hence, the incongruent training could just have puzzled the infant, resulting in a disappearance of the default priming effect, which could be caused by random gaze behaviour, but not in the appearance of a reversed priming effect. To replicate and extend the findings of Experiment 1, Experiment 2 tested the effect of training on infants’ perception of dynamic actions.

Experiment 2

To further test the plasticity of infants’ action perception, two groups of 7-month-old infants were presented with the same experimental paradigm as in Experiment 1 but with a dynamic grasping action serving as cue (adapted from Wronski and Daum 2014).