Abstract
Although it would not seem implausible for face recognition proficiency to reflect capacities for both perceptual differentiation of faces as well as memory for faces, rigorous, quantitative assessments of whether these capacities are correlated or independent across individuals have not been assessed. We addressed this issue in a delayed match-to-sample task. Participants viewed a computer-generated face (the sample) and after 0, 4 or 12 seconds viewed that identical image and a distractor of varying degrees of similarity, indicating which of the two matched the sample (Fig. 1a). To suppress rehearsal during the delay period, subjects judged which of a series of headshots were famous. There was a marked cost of delay on both accuracy and RTs (Fig. 2). To assess the capacity for perceptual differentiation independent of memory, subjects performed a separate match-to-sample task in which all three images (sample and test faces) were displayed simultaneously (Fig. 1b). The diagonal arrangement of the faces defeated a strategy of local feature/pixel search. Dissimilarity between the sample and the distractor image was quantified using the Gabor Jet model, a model that predicts psychophysical similarity of faces almost perfectly. For each subject a “cost of similarity” and a “cost of delay” were calculated as the slope of error rates over variation in Gabor dissimilarity for the former and the slope of error rates over the delay interval for the latter. These measures were highly reliable as evidenced by 1st vs 2nd half and odd-even correlations ranging from .73 to .92. Nonetheless, the correlation, over subjects, for the costs of perceptual differentiation of faces and face memory was essentially zero, indicating that capacity for the perceptual differentiation of faces and capacity for face memory are independent (Fig. 3).
Acknowledgement: Harold W. Dornsife Research Fund