Observers judged the motion coherence of random-dot cinematograms. The
oretical models were developed for coherence matches between cinematog
rams constructed from different angle distributions. Evidence is prese
nted that coherence matches are made on the basis of the Shannon-Wiene
r information entropy. We show how the formal structure of information
theory may be used to predict perceived pattern goodness when the und
erlying distributions of pattern alternatives are implicit in the judg
ment task.