Recognition memory for Caucasian and Korean synthetic faces

Abstract

Synthetic faces were generated by digitising 37 points on individual face photographs. Using both Caucasian and Korean synthetic faces, we designed an experiment to determine whether there was any memory difference between the two face categories defined only by face geometry and not by skin colour, hair texture, etc. In each experiment, observers first studied four Caucasian and four Korean faces presented in random order. Following this, there was a 15 min rest period before the recognition experiment began. On each recognition trial, one previously studied face was paired with three distractor faces in a spatial 4AFC procedure. Importantly, the distractor faces were all chosen to be orthogonal to one another and to lie a fixed geometric distance from the studied face. Overall recognition memory was 51% correct, far above the chance value of 25%. Recognition memory for Caucasian faces (59%) was significantly greater than for Korean faces (44%), even though observers were never told that the faces comprised two ethnic categories. We conclude that recognition memory for synthetic faces exhibits implicit categorisation effects for the geometry of Caucasian and Korean faces.