Abstract

Correlation and gradient models predict different responses to moving sine gratings of varying contrasts. For small contrast values, correlation models predict a quadratic dependence of velocity discrimination on contrast, whereas for gradient models velocity discrimination is independent of contrast. Drifting sine gratings that optimally activated the movement detection system were used to determine the temporal frequency needed to see drifting test gratings moving faster than a comparison sine grating of the same spatial frequency. The contrast of the gratings was varied between 1 and 20%. The temporal frequency was varied between 5 and 20 cycles/s. The spatial frequency of test and comparison patterns varied from 0.5 to 8 cycles/deg in 1-octave intervals. Test and comparison patterns were presented for 200 ms with a 500-ms interval using a temporal two-alternative forced-choice procedure. For 15–20 cycles/s, where flicker and not left–right movement was seen, there was some dependence on contrast. These results were found for velocity discrimination relative to backgrounds composed of either uniform mean luminance or single- or multiple-spatial-frequency components. Velocity discrimination thresholds are in agreement with the gradient model of movement discrimination but do not support simple correlationlike models. However, by normalizing the output of the spatiotemporal models approximately, a contrast-independent measure of velocity can be derived. These results can be predicted by neural networks composed of the paired even- and odd-symmetric simple and complex cells found in magnocellular pathways.

© 1988 Optical Society of America