Abstract
Recent behavioral (Peterson & Eckstein, 2012, 2013) and neuroimaging (Henriksson, Mur & Kriegeskorte, 2015; de Haas et al., 2016) evidence suggests that neurotypicals (NT) encode faces in a retinotopically-specific fashion: Subjects perform best recognizing familiar faces at their personal "preferred looking position" (Peterson et al., 2016). Here we ask three questions: 1) Do individuals have similar preferred looking positions for both faces and non-face objects? 2) Does the tight relationship between preferred face-looking position and face recognition accuracy reflect either a fundamental difference in representational capacity for faces at different retinotopic positions, or an advantage for matching the retinotopic position of a face at encoding and test? 3) Is face-fixation behavior or its relationship to performance different in developmental prosopagnosia (DP)? We assessed the specificity of individual differences in eye movements and performance for face recognition in 15 NT and 11 DP subjects. In three tasks, subjects were free to move their eyes as they recognized celebrities (ID), cars (CAR), and expressions (EXP). In a fourth task, subjects decided whether two sequential images of unfamiliar faces showed the same person while maintaining fixation at each of four face-positions (SD). We found that DPs were selectively impaired at face identification (ID, SD) but not at object (CAR) or expression (EXP) recognition. Identification deficits in DPs could not be explained by eye movements; compared to NTs, DPs showed no significant difference in the location (mean) or consistency (variance) of initial eye movements. Critically, individual differences in the locations of NT face-fixations were: 1) Uncorrelated with car-fixations (face-specific), and 2) Strongly predictive of face recognition ability across retinotopic positions (i.e., upper-face lookers performed best fixating the eyes, lower-face lookers performed best fixating the mouth), suggesting that individuals' face representations are tuned to the individual-specific retinotopic positions at which faces appear during natural, real-world behavior.
Meeting abstract presented at VSS 2017