Pooling and segmenting motion signals

Humans are extremely sensitive to visual motion, largely because local motion signals can be integrated over a large spatial region. On the other hand, summation is often not advantageous, for example when segmenting a moving stimulus against a stationary or oppositely moving background. In this study we show that the spatial extent of motion integration is not compulsory, but is subject to voluntary attentional control. Measurements of motion coherence sensitivity with summation and search paradigms showed that human observers can combine motion signals from cued regions or patches in an optimal manner, even when the regions are quite distinct and remote from each other. Further measurements of contrast sensitivity reinforce previous studies showing that motion integration is preceded by a local analysis akin to contrast thresholding (or intrinsic uncertainty). The results were well modelled by two standard signal-detection-theory models.