A cost-efficient mechanism of face learning.

Most people can recognise large numbers of faces and take this skill for granted. But how do we actually do it and what visual information do we rely on? Why are we struggling when learning new faces, and why are there people who can’t recognise their own relatives? We developed a theory to address these questions; it assumes facial representations are formed in a cost-efficient manner, to spare us from the cognitive burden of remembering facial details, which unavoidably lead to errors. We propose that different information is used in individual faces, regardless of familiarity, based on stability. Features that remain stable over encounters receive more representational weight. For cost-efficiency purposes, coarse information is privileged over details, but representations may refine over time. Poor recognisers would over-rely on stability and lack the ability to use facial information in a flexible manner. We will review experimental evidence that support these assumptions—both with unfamiliar and famous faces—and discuss broader implications of the theory for the field.

Biography: Christel Devue is a Research Fellow at Victoria University of Wellington, New Zealand. She has a PhD in Psychological Science from the University of Liege, Belgium, where she studies self-face recognition with Serge Brédart. She did a post-doc on visual attention with Jan Theeuwes at the Vrije Universiteit Amsterdam and a post-doc on cognitive control with Gina Grimshaw at Victoria. She is currently establishing a research program aimed at understanding human strengths and weaknesses with facial recognition and associated individual differences.