Abstract
Crowding refers to the difficulty in identifying a target in the presence of nearby patterns. The purpose of this study was to examine whether observers use similar features and spatial frequencies of a stimulus to make perceptual decisions about the stimulus in the presence of crowding. We used a response-classification technique to map out the stimulus features that correlate with observers' response in discriminating the location of the gap (right or left) of a Landolt C target. The Landolt C (gap-size = 3.2 min) was embedded in a 100 × 100 pixel noise field made up of white Gaussian luminance noise of 15% RMS contrast. The noise field was refreshed on each trial. In separate blocks of trials, the Landolt C was presented alone, or in the presence of 4 flanking bars, at edge-to-edge separations of 0, 1 or 2 bar-widths (1 bar-width = 3.2 min). The classification images represent the correlation of each noise pixel with the observer's response, across 10000 trials for each condition. With decreased bar separations, pixels in the stimulus that correlate with observers' responses showed a progressive shift away from the physical locations of the gap in the stimulus. We computed the power spectra of each classification image to determine the spatial frequencies used by the observers. These power spectra reveal spatial-tuning functions. Unlike the power spectrum of the optimal template which did not change with bar separations, with decreased bar separations, the peak of the human observers' tuning functions dropped in their proportional power, the tuning became broader, and the peak tuning frequency shifted toward a higher frequency. At the closest bar separation, the tuning functions were essentially flat. The changes in the properties of the spatial-tuning functions derived from our classification images indicate that observers used sub-optimal bands of spatial frequencies for analysis in the presence of crowding, perhaps as a “misguided” attempt to avoid crowding.
Supported by NIH grant EY12810 (SC) and USC Zumberge Fund (BT).