Abstract
Ensemble processing refers to the visual system's ability to process features from large collections of stimuli in a scene (i.e., an ensemble) by statistically compressing redundant information (e.g., computing the average size of a set of shapes). Typically, participants are more accurate at reporting the average feature of a set compared with any individual feature. Here, we investigated ensemble face-processing mechanisms for higher-level identity and lower-level viewpoint features. To this end, in four different experiments we examined single and average feature extraction from face ensembles that varied continuously in identity and/or viewpoint. Specifically, participants reported either the average feature of a set or that of a single face, randomly selected from an ensemble of six unfamiliar faces. Our findings indicate, first, that participants were more accurate at reporting viewpoint than identity, consistent with different levels of processing for the two types of features. Second, we found that average face reports were generally more precise than single face reports and that single face reports were biased towards the average of the set especially under increased cognitive load. Third, surprisingly, the average viewpoint of an ensemble, varying between -60 and 60 degrees relative to a frontal view, did not influence the precision of identity reports for either average or single faces. Finally, estimates of average identity and viewpoint were both precise regardless of whether they were reported simultaneously or separately, with no interference from the irrelevant feature. Thus, the present findings point to distinct levels of ensemble processing operating on viewpoint and identity information. More generally, these results argue for the existence of multiple, robust and independent mechanisms for ensemble face processing.
Meeting abstract presented at VSS 2018