Abstract
An object in peripheral vision is difficult to identify when it is surrounded by visual clutter, a phenomenon referred to as “crowding”. Among the many accounts that have been proposed to explain crowding, we sought to investigate two competing explanations: 1) visual features of nearby objects are averaged together in relatively early levels of the visual system, versus 2) a higher-level attentional filter lacks resolution to allow accurate selection of information that is otherwise veridically encoded. We employed a new method of adjustment paradigm that allowed us to tease apart the contributions of averaging and selection failures. For 500 ms centered at 10° in peripheral vision, a target Landolt C was displayed surrounded by a larger Landolt C flank of varying diameter, with the gaps independently and randomly oriented between 0 and 360°. An observer’s task was to rotate a “response” Landolt C, presented at fixation, to match the target. In line with the crowding literature, the standard deviation of the circular distribution fit to response errors decreased linearly with increasing target-flanker separations (Ester, Klee & Awh, 2014), reaching unflanked performance when the flanking ring diameter was approximately 0.35 times the target eccentricity. Subsequent analyses revealed that, when the difference between the orientations of the target and flanker gaps was large, response errors were well explained by observers’ difficulty in selecting between target and flanker, in line with attentional accounts of crowding. In a second experiment, observers’ task was to rotate a pair of response Landolt Cs to match the target and flank. Even when attentional selection accurately identified separate target and flanker orientations, responses were biased by the average of the two orientations. Our data thus suggest that attentional resolution is not the primary limit on visual awareness, but instead that visual features are averaged prior to attentional selection.
Meeting abstract presented at VSS 2015