There has been some work on modeling of rotational motion, but efforts to compare the models against tests of human judgments are almost nonexistent. The present work studied psychophysical performance with rotation of random dots, and found that a model proposed by Watson & Ahumada accounts for most, but not all, of the judgments. As predicted, density of the dots was not critical to effective performance, and the time needed for judgments was appropriate. An unexpected finding was that the ability of the model to judge direction of motion did not depend on knowing the location of the center of motion. The implications for neural encoding mechanisms are discussed.