Abstract
Humans can make highly precise alignment judgments. There is a long standing question about how information is integrated along the length of the Vernier stimulus. In this study we introduce a novel classification image technique to explore the local contribution of each part of the stimulus to Vernier judgments. We used a two-segment Vernier stimulus in which each segment consisted of five Gabor patches (10 cpd). We introduced spatial noise by perturbing the position of the individual patches of the ‘test’ segment. The observers' task was to judge the position of the ‘test’ segment relative to the reference segment. An optimized version of the stimulus was used with binary spatial noise that allowed the number of possible noise combinations to be dramatically reduced. The segment separation was varied from abutting to widely spaced. A reliable classification image for each offset magnitude can be obtained from only 960 trials.
Human thresholds are about twice that of an ideal observer whose judgment is based on the mean of the patch locations. We found that human observers assign different weights to each patch comprising the stimulus. Intuitively the patch that is closest to the reference segment would seem to have the strongest influence. But this is not always the case. For subthreshold abutting stimuli, the second patch has a stronger influence than the first one. Observers tend to use the first four patches; ignoring the fifth. In some cases, the fifth patch even has a strong negative influence on Vernier alignment. We also find interesting contextual effects of interleaving offset stimuli on the classification images of aligned stimuli.