Abstract
How perceptual learning generalizes can help reveal the specific perceptual mechanisms that training improves. Performance in contrast detection can be explained by the existence of independent cardinal mechanisms sensitive to opposing and combined L and M cone contrast (L−M and L+M, respectively). We tested whether these mechanisms underlie learning in a detection task by training in the color direction preferred by one and testing in the direction preferred by another. Four subjects trained on a two temporal interval forced choice detection task. Stimuli were 5 deg, 5 cpd Gabor patterns that were displayed for 500 msec centered at an eccentricity of 10 deg. Stimulus contrast varied under the control of a staircase procedure, and subjects were given feedback about their performance on each trial. Two subjects were trained using patterns containing L−M cone contrast, and two were trained using patterns containing L+M cone contrast. Subjects participated in over 20 daily 30-minute training sessions. Before and after training, detection thresholds were measured for L−M, L+M, L, and M cone contrast patterns. We found that L−M and L+M training decreased thresholds (and thus increased sensitivity) for the trained color direction by 50% on average. Furthermore, the reductions in L−M thresholds following L−M training were, on average, 2.0 times larger than threshold reductions for untrained directions. L+M training showed less specificity. Our results suggest that training on detection of L−M patterns may cause red-green color opponent neurons to increase their sensitivity.
Supported by NIH EY11862, NSF Graduate Research Fellowship