Visual cortex neurons in monkeys and cats : WILSON S . GEISLER nNn DUANE

A descriptive function method was used to measure the dctection, discrimination, and identiflcation pcr-formance of a large population of single neurons recorded from within the primary visual cortcx of thc monkey and the cat, along six stimulus dimensions: contrast. spatial position, orientation, spatial irequency, temporal fiequency, and direction of motion. First. the responses of single neurons were measurcd along each stimulus dimension, using analysis intervals comparable to a normal flxation intcrval (200 ms). Second, the measured responscs of each neuron were fitted with simple descriptivc lunctions, containing a few fi'ee parameters, fbr each stimulus dimension. These functions wcre found to account lbr approximately 9O7r: of the variancc in the measured response means and response standard deviations. (A detailed analysis of the relationship between the mean and the variancc showed that the variance is proportional to the mean.) Third, the parametcrs of the best-fitting descriptive f'unctions wcrc utilized in conjunction with Bayesian (optimal) decision theory to determine the detection, discrimination, and identification pcrfbrmance for each neuron, along cach st imulus dimension. For some of the cel ls in monkey, discrimination perfbmance was comparable to behavioral performance; for most olthe cells in cat, discrimination pcrformance was better than behavioral perfbrmance. The behavioral contrast and spatial-frequency discrimination functions wcre similar in shape to the envelope of the most sensitivc cells: they werc also similar to the discrimination functions obtained by optimal pooling of the entire population ol-cells. The statistics which summarize the parameters ol'the descriptive functions were used to estimatc the response ofthe visual coftex as a wholc to a complex natural image. The analysis suggcsts that individual corlical neurons can reliably signal precise infbnnation about thc location, size, and orientation of local image features.