Abstract
Items in the periphery of a visual display are more difficult to identify when flanked by irrelevant distractors, a phenomenon referred to as visual crowding. One perspective argues that crowding results from a compulsory averaging or integration of target and distractor features (the “integration” view). For example, although subjects are typically unable to report the orientation of a crowded stimulus, they can reliably estimate the average orientation of all stimuli in the display. This implies that local orientation signals are pooled or integrated prior to reaching awareness. An alternative perspective (the “swapping” view) argues that crowding results from a loss of positional information that leads the visual system to bind targets and distractors to incorrect spatial locations. For example, when asked to report the identity of a letter in a crowded display, subjects frequently report neighboring distractors with high confidence. Here, we attempted to distinguish between integration and swapping accounts of crowding using an analytical procedure that allowed us to simultaneously quantify the strength of distractor-target integration and the frequency of distractor-target swapping. Subjects were required to report the orientation of a parafoveal target that appeared in isolation or flanked by two oriented distractors. Subjects' estimates of orientation were used to define a distribution of response errors (i.e., reported minus actual orientation) which we attempted to describe using quantitative models that encapsulate key predictions of the integration and swapping perspectives. Across three experiments, we found no evidence for distractor-target integration. Instead, the majority of subjects' response errors were distributed around the target orientation, with the remaining responses normally distributed around distractor orientations or randomly distributed across orientation space. These findings are inconsistent with an integration account of crowding, and suggest that crowding primarily results from (1) a total loss of target information, or (2) swapping of target and distractor values.
Supported by NIH-R01MH077105-01A2 to E.A.