Abstract
Many have proposed that peripheral vision operates by computing statistical summaries over local portions of the visual field, and that the loss of information associated with this process underlies the phenomenon of “crowding” (Parkes et. al. 2001; Pelli et. al. 2004; Greenwood et. al. 2009; Balas et. al. 2009; Freeman and Simoncelli, 2011). Here, we demonstrate another consequence of this hypothesis: that such statistical representation can either help or hinder visual discrimination performance depending on the observer’s task. We created synthetic texture stimuli by matching a set of higher-order statistics measured from digitized photographs (Portilla and Simoncelli, 2001). The parameters include both the marginal statistics and pairwise correlations of the responses of V1-like filters selective for different spatial frequencies, orientations, and spatial positions. Observers were asked to discriminate stimuli presented simultaneously at three degrees eccentricity windowed within circular apertures. When stimuli differed in their statistics, performance increased with increasing patch diameter. This is expected since the parameters of the model converge to different values as the patch size increases. Interestingly, when observers discriminated between different samples matched for the same parameter settings, performance decreased with patch diameter. As the statistics converge to their matching values with increasing patch diameter, subjects were no longer able to utilize the local cues that enable high performance at small patch sizes. These opposing behaviors are analogous to those found for discrimination of auditory textures as a function of temporal window duration (McDermott et al., 2013), and suggest a general processing strategy for sensory systems.
Meeting abstract presented at VSS 2015