Scanpath Analysis and Attentional Synchrony
Infants’ gaze behavior during all the trials suggested that they were interpreting the events taking place on-screen as social scenes. Patterns of attentional synchrony included looking from the individual character, who had just entered the scene, to the group characters and back to the individual character, all in-keeping with the expectation of social goal-related actions (i.e., for the individual character to join the group) and a call-response pattern of social engagement (i.e., that the group would respond to this goal) [
30]. This exact pattern in attentional synchrony occurred in 148 of a total 198 trials (74.7%).
A second pattern in attentional synchrony that emerged in some of the gaze data from the exclusion trials across both conditions was a tendency to look from the individual character to the group resource and back to the individual character, just before or as the individual character was approaching the group (four infants, 9%), or from the group characters to the group resource to the individual character, in any order, during this timeframe (12 infants, 34%). Another two infants scanned between the group and group resource as the individual character was approaching the group. Moreover, just subsequently to the exclusion event in the exclusion trials, eight infants (23%) made a visual connection in at least one trial between the group resource and the excluded character, as well as the group characters. A further 14 infants (40%) looked from the group characters to the group resource or vice versa during this same timeframe. Overall, these visual connections suggest that many of the infants were exploring the goal-related actions of the individual character to access the group resource through the group, while some were even anticipating a response from the group in relation to their allocation of the resource.
Finally, as a tendency for infants to avert their gaze offscreen during the exclusion trials had emerged, gaze aversion as a function of ‘offscreen’ looking times greater than 500 ms was analyzed, ruling out simple tracking error where data validity was low through manual observation of the session recordings. These periods of time were coded as an AOI in our scanpath analysis and transition matrices were computed to record the rate of infants looking ‘offscreen’ across conditions. Interestingly, it was found that infants who were watching the explicit exclusion scenes averted their gaze 1.8 times more than those in the implicit condition if they had previously been fixating on the group gatekeeper.
Proportion of Fixation Durations on AOIs
The results for fixation durations reported herein pertain to the duration of time spent fixating on a particular AOI over the total time spent fixating on all AOIs within a given scene, as expressed in proportions of looking time to a particular AOI. This approach to analysis minimizes the effect of individual differences in attention and fixation behavior, as well as eye-tracking error on outcomes and thus, in our view, was the most robust way to analyze gaze behavior with infants.
As our focus was on differentiating between viewing behavior across the conditions of explicit and implicit exclusion, the results from the exclusion trials are reported herein. In our between-subject analysis, it was found that overall, across all exclusion trials, infants in the implicit condition spent more time fixating on the excluded character over the group gatekeeper, than those in the explicit condition. Inversely, infants in the explicit condition spent a higher proportion of time fixating on the group gatekeeper than the excluded character (PExcluded–PGatekeeper) = 0.44, 95% CI (0.18, 0.71), p = 0.001. This was also the case when the analysis time frame was narrowed down, from the moment just subsequent to the exclusion event up until the end of the trial, when both conditions were otherwise identical in every way (PExcluded–PGatekeeper) = 0.15, 95% CI (0.06, 0.23), p < 0.001.
An analysis of gaze behaviors across exclusion trials was next conducted by grouping the infants together according to the form of exclusion they had seen and their eventual choice of character. It was found that in the implicit condition, infants who chose the
group gatekeeper in line with the prediction on average spent a higher proportion of time fixating on the
excluded character than those who chose the
excluded character in the subsequent preferential reaching task (P
Excluded–P
Gatekeeper) = 0.16, 95% CI [0, 0.33],
p = 0.05. However, infants who chose the
excluded character also looked more at the
excluded character than the
group gatekeeper, though not to a statistically different degree on the interaction of choice and character looked at
F(1,18) = 2.11,
p = 0.16. The lack of statistical significance may be more a reflection of the small number of data points available for gaze data analysis in each group. Mean proportions of fixations on primary AOIs are reported in
Table 1 where it can be seen that certain tendencies likely underpinned the eventual preference for one character over another, such as those in the explicit condition who chose the
group gatekeeper, who during the exclusion trials spent a higher proportion of time fixating on the
group gatekeeper than those who chose the
excluded character.
A within-subject analysis of gaze behaviors throughout exclusion trials (i.e., looking at the effect of trial number on the distribution of gaze) did not yield any statistically significant results. Gaze was distributed more or less equally across trials in both conditions.
However, when the final static image of the
excluded character and
group gatekeeper was presented on-screen (see
Figure 3), all three infants in the implicit condition who chose the
excluded agent (against the prediction) first looked at the
excluded agent, while only two of the 10 infants who chose the
group gatekeeper agent (in line with prediction) first looked at the
excluded character, when presented with this scene (
p = 0.008). This analysis comprised of any kind of visual hit on the AOI when the image appeared (i.e., saccadic or fixation,
n = 13). Moreover, the probability of first looking at the
group gatekeeper was not statistically different for those in the implicit condition who chose the
group gatekeeper and those in the explicit condition who chose the
excluded character in line with the predictions for each condition (
p = 0.87), such that the form of exclusion seen did not appear to have a direct role in predicting the first point of fixation for those who chose in line or against the prediction, but the eventual choice of character nevertheless differed across explicit and implicit conditions.
Furthermore, a significant difference was found when comparing infants from the explicit and implicit condition in the overall proportion of time they spent looking at either the
excluded character or the
group gatekeeper when presented with this final image, such that infants in the explicit condition had spent a higher proportion of time fixating on the
group gatekeeper than those in the implicit condition and inversely, those in the implicit condition fixated for longer periods of time on the
excluded character than those in the explicit condition (P
Excluded–P
Gatekeeper = 0.14), 95% CI [0, 0.30],
p = 0.05. When this was examined at the level of choice, it was found that the mean proportions of fixations, though not statistically different, were higher for those who chose against the prediction in both conditions than those whose choice was as predicted. This was also the case in the distribution of gaze between the
excluded character and the
group gatekeeper, where a higher proportion of fixations was linked to eventual choice in all the bar one group where it was roughly the same (i.e., those who chose the
excluded character in the explicit condition). Mean proportions of fixations are reported in
Table 2.