Learnability of infants' center-of-gaze sequences predicts their habituation and posthabituation looking time

Our recent work, using both computer-animated and natural images, demonstrates that infants' eye movements are sequentially structured. In particular, we employ a simple recurrent network (SRN) to estimate the learnability of infants' gaze sequences. The current study directly compares this learnability metric with infants' global looking time on a trial-by-trial basis, during both an habituation display and two posthabituation test displays. The results confirmed our prediction that the relative learnability of infants' gaze sequences during the latter part of the habituation phase predicts global looking time. We also found that learnability during the later habituation trials predicts looking time to one of the two posthabituation test displays. These findings provide support for our hypothesis that infants' visual exploration is guided implicitly by the goal to generate gaze patterns that are sequentially predictable.