Attentional modulation of excitatory and inhibitory signals revealed in perceptual learning

Repeated practice in viewing a visual stimulus moving in a particular direction increases our sensitivity to the trained visual motion. Psychophysical and physiological studies have shown that the sensitivity increase is the greatest for the trained direction. The present psychophysical study demonstrates that this is not always the case. In the first experiment, subjects were divided into two groups. They participated in either coherent motion detection training or coherent motion direction discrimination training, using the same set of two stimuli which differed in motion directions by 20 degrees. The tasks differed only in instruction set. In the detection training, the subjects judged which of the two successive motion displays included coherent motion. In the discrimination training, the subjects judged whether the direction of coherent motion was the same in the two successive motion displays. In both groups, the same motion stimuli were presented with equal frequency. If the learning depended only on the visual input, the patterns of learning would be similar between the groups. Surprisingly, the results were very different between the groups. In the detection training, the sensitivity increase was greater for the trained directions than the untrained directions. However, in the discrimination training, the greatest sensitivity increase was not observed at the trained directions, but rather 5 degrees off the trained directions. This indicates a "peak repulsion" effect for perceptual learning in the discrimination task. In the second experiment in which the trained directions differed by 90 degrees, the patterns of learning became similar between the detection and discrimination groups. The greatest sensitivity increase occurred at the trained directions in both groups. A model was developed in which lateral inhibition was proposed to occur between the direction-specific mechanisms for the trained directions. The simulation of the model successfully replicated the results for the discrimination group in experiment l. When the model was simulated without lateral inhibition, the greatest sensitivity increase was found at the trained directions. These findings indicate that (1) lateral inhibition may explain the peak repulsion effect in the discrimination training in experiment 1, and (2) whether inhibition occurs depends upon the task demands.