Abstract
Crowding - the disruptive influence of clutter on object recognition - occurs when flanker objects fall within an "interference zone" around a target and features from each object combine. These zones scale with eccentricity and show a characteristic teardrop shape, oriented radially from fixation. Precisely why crowding follows this pattern is unclear. Interestingly, saccadic eye movements show a similar teardrop-shaped landing distribution around an intended target location. To determine whether crowding and saccadic landing errors share a common positional uncertainty, we combined the two tasks by having observers make eye movements to crowded targets.
The target was a circular clock, with a central point and radius-length stroke oriented in a cardinal direction, flanked by two similar elements. After target offset, observers (n=5) made a saccade to the target centre and then indicated its stroke orientation (4AFC). We measured four (cardinal) saccade directions, at 4 and 8 degrees eccentricity, with flankers along either radial or tangential axes to fixation. From identification performance (as a function of element spacing) we derived "crowding interference zones" and from saccade landing positions we derived "saccadic-error zones".
Crowding and saccadic-error zones show a similar elliptical anisotropy, with 2-3 times more error in the radial than tangential direction. Both also scale with eccentricity, doubling between 4-8 degrees. This gives a significant correlation between the two zone-types, with saccade zones around one-half to one-third the scale of crowding. When flankers are inside the crowding zone, saccades would therefore fall on both target and flanker locations. Saccades would only isolate the target accurately once flanker separations exceed the crowding zone.
Our results demonstrate the potential linkage between the coding of saccade locations and crowding. We propose that "pointers" within the saccadic map are involved in determining the spatial range of feature pooling in peripheral vision.
Meeting abstract presented at VSS 2012