Proceedings of the Annual Meeting of the Cognitive Science Society

Aging is associated with more nose-focused face scanning pattern and reduced pattern consistency during face recognition, and both effects are correlated with performance reduction. We investigated whether these changes are related to sensorimotor (increased saccade noise) or cognitive (declined memory of visual routines) declines by simulating both mechanisms in a Deep Neural Network + Hidden Markov Model (DNN+HMM) architecture that learns facial representations and oculomotor routines for face recognition simultaneously. Results showed that increasing saccade noise reduced scanning pattern consistency without making the pattern more eyes- or nose-focused, whereas reducing visual routine accuracy to simulate memory decline led to more nose-focus scanning pattern without affecting pattern consistency. Both effects were correlated with reduced recognition accuracy, consistent with human data. Our results thus suggested that changes in face scanning pattern and consistency are dissociable consequences of distinct mechanisms underlying aging-related decline in face recognition ability, with important implications for intervention strategies.