Abstract
Crowding is a key limiting factor of form vision in the peripheral field. A prominent theory of crowding is that of inappropriate feature integration (Levi, 2008; Pelli & Tillman, 2008). However, it is not known what features are inappropriately integrated. Tjan and Dang (2005, VSS) addressed this question by flanking a letter target with different types of flankers. They found that a noise flanker obtained by phase-scrambling a letter crowded a letter target as effectively as a letter flanker. Here we repeated their experiment with gray-scale images of everyday objects in place of letters. We measured contrast threshold elevation for identifying a target object presented at 10° below fixation as a function of target-flanker spacing (center-to-center: 1°, 1.5°, 2.25°, 3.375°). At small spacing where the target and flanker stimuli overlapped, the target was made to occlude the flankers. We used four types of flankers: intact objects, anisotropic pink noise (phase-scrambled objects), isotropic pink noise (phase-scrambled and orientation-scrambled objects), and white noise. As with letters, we found that anisotropic pink noise flankers led to essentially the same threshold vs. spacing function as intact object flankers, with a maximum threshold elevation of about 0.7 log units. White noise flankers led to the least amount of crowding, with threshold elevation less than 0.2 log units. Isotropic pink noise flankers produced an intermediate amount of crowding, with a maximum threshold elevation of about 0.6 log units. Taken together, these results suggest that the features that are inappropriately integrated are those that are common to the intact and the phase-scrambled objects, thus strongly implicating narrowband Gabor-like features. Comparisons between the isotropic and anisotropic pink noise conditions further suggest that the orientations of these Gabor features are relevant for crowding.
NIH/NEI R01-EY017707, R01-EY016093.