Abstract
Visual crowding is a phenomenon whereby the features of objects viewed peripherally are rendered less discriminable by adjacent objects in the visual field. Crowding has been well studied in humans; but little is known about crowding in animal models where neurophysiological recordings are tractable. We trained an awake monkey on an orientation discrimination task and used a psychophysical reverse-correlation paradigm to characterize visual crowding from adjacent locations in the visual field. Distractors at some locations led to errors that were consistent with the distractor orientations (indicative of substitution or averaging). At other locations, incorrect judgments were inconsistent with the distractor orientation (indicative of repulsion or induced tilt). Thus, even within a single stimulus array, stimuli at distinct locations in the visual field had opposing influences on the judgment of target features. Similar behavioral results were found in one human subject. In parallel, we asked how crowding affected the neuronal encoding of orientation information. We recorded the responses of neuronal populations in anesthetized macaque V1 to target gratings, presented in isolation or surrounded by distractor gratings. To assess crowding, we applied linear discriminant analysis to V1 population responses to pairs of targets with offset orientations. Decoders trained on responses to isolated stimuli had significantly worse performance on responses to crowded stimuli, across a wide range of task difficulties. Performance deficits were also apparent when decoders were trained on crowded responses. Shuffling the neuronal responses to remove correlations reduced decoding performance, but crowding was still evident, suggesting that it arises in part from effects on the responses of individual neurons. These results demonstrate that crowding involves a loss of sensory information at the first stages of cortical processing; however, as crowding effects in V1 were weaker than those seen behaviorally, maladaptive pooling by downstream circuits is also likely to contribute.
Meeting abstract presented at VSS 2016