The role of eye movement consistency in learning to recognise faces: Computational and experimental examinations

Abstract

In face recognition, the frequency of looking at the eyes, the most diagnostic feature, predicts better performance in adults but not in children, suggesting that different factors may underlie children’s face recognition performance. Here we test the hypothesis that eye movement consistency plays an important role during early learning stages. Through computational modeling that combines a deep neural network and a hidden Markov model that learns eye movement strategies by interacting with the network, we showed that consistency instead of eye movement pattern better predicted face recognition performance during early learning stages. Similarly, in human studies, children’s consistency but not pattern of eye movements predicted face recognition performance, and their eye movement consistency was associated with executive function abilities. Thus, learning to recognize faces initially involves developing a consistent visual routine, which depends on executive function abilities. This finding has important implications for learning in both healthy and clinical populations.