Slow discrimination of contrast-defined expansion patterns

We compared observers' performance in the same complex motion discrimination task using stimuli defined by luminance or by contrast. They were asked to discriminate between a centred expansion pattern, constructed from four patches of outwards motion, and a distorted expansion pattern, constructed with one patch containing inwards motion and three patches containing outwards motion. We measured performance versus duration and found that for luminance-defined patterns, observers were able to discriminate correctly between these patterns in 75% of trials when the stimulus duration was 200 ms. For contrast-defined patterns, observers required over 2 s to reach this level of performance. Observers did not require such long durations to discriminate correctly between the contrast-defined expansion patterns when the patterns contained fewer patches or when the distorted patterns contained more patches of inwards motion. This suggests that observers performed the task by searching for a patch that contained a pattern moving inwards. There was no such effect on performance with the luminance-defined patterns. These results also suggest that contrast-defined patterns are processed too slowly to provide an input to specialised optic flow detectors that guide navigation in real time. Further, the differences in performance may be due to processing delays from sequential processing of the contrast-defined local motion signals.