Is the direction of second-order, contrast-defined motion patterns visible to standard motion-energy detectors: a model answer?

Previous psychophysical studies (e.g., Smith & Ledgeway, 1997) have provided evidence that under some conditions, the detection of a particular class of stimuli (contrast-modulated static noise) widely employed to study second-order motion processing may be inadvertently based on encoding local imbalances in luminance motion energy. In particular when static noise composed of relatively large noise elements is used, direction-identification performance at threshold may actually be mediated by the same mechanisms that respond to first-order motion, due to the presence of persistent spatial clusters of noise elements of the same polarity. However, Benton and Johnson (1997) modeled the responses of conventional motion-energy detectors to contrast-modulated static noise patterns and found no evidence of any systematic directional biases in such stimuli when the mean opponent motion energy was used to quantify performance. In the present paper we sought to resolve this discrepancy and show that the precise manner in which computational models are implemented is crucial in determining their response to contrast-modulated, second-order motion patterns. In particular we demonstrate that by considering the information encapsulated by the peak (rather than the mean) opponent motion energy and the predominantly local nature of imbalances in motion energy that can arise in contrast-modulated static noise, it is possible to readily model the patterns of empirical results found.