Abstract
The visual system efficiently perceives the average in a group of similar items. For example, we can accurately derive the average expression of a crowd of faces, the average orientation of a set of lines, and the average size of a group of circles. Although ensembles are created efficiently and with minimal effort, understanding the capacity limitations of ensemble representations remains an active area of research. In the current set of experiments, we explored ensemble capacity limitations by having observers view two sets of ensembles simultaneously. Observers were instructed to attend to both ensembles, presented in a semicircle above and below the horizontal meridian, and then report the mean of the post-cued set. We conducted three experiments, varying the presented ensemble domain in each version: face/gabor, face/color, or color/gabor. Prior to the main experiment, we equated difficulty across ensemble conditions by manipulating variance in a standard QUEST procedure. In each experiment, we contrasted performance in the mixed condition, when multiple ensemble domains were presented concurrently, (e.g., face/gabor), with performance in the unmixed condition (e.g., face/face or gabor/gabor). Interestingly, the results revealed a significant benefit of mixing ensembles in the color/gabor experiment relative to the unmixed conditions. Under no circumstances did mixing ensembles reduce ensemble representation precision, which contrasts with previous work showing a perceptual cost when attending across multiple ensemble domains. However, our finding is consistent with the notion that different ensemble domains are processed independently. We may infer from these results that attending to multiple ensemble domains (mixed condition), in some instances, frees up neural resources relative to attending to multiple ensembles within a single domain (unmixed condition).
Meeting abstract presented at VSS 2016