In fact though, Ullman does not follow the computational approach in his own analysis of the correspondence process. The central tenet of his account is that matching is guided by a cost function. Following Gestalt tradition, he assumes this cost depends on velocity, yet he offers no computational justification for such a dependence. This is an unfortunate flaw in the analysis since the time and distance between matched elements almost certainly make separate contributions to any cost function. Ullman speculates that the cost function should reflect "the statistics of the universe": commonly occurring velocities are given low cost; uncommon, high cost. However, no attempt is made at deriving such a statistic, nor is ajustification given for the choice of a velocity statistic. It happens that a simple computational argument can be advanced againtpst the assumed velocity dependence: the correspondences established in any given stimulus pattern should not depend on the overall scale of the stimulus itself. An observer's viewing distance dictates image scale, and hence image velocity, but should not alter the organization of the perceived motion. Ullman's failure to differentiate time and distance contributions to cost may be due in part to the exclusive use of the two-frame paradigm in his empirical studies. While pairs of elements from the two frames may differ in spatial separation, they all have the same temporal separation. Also, by restricting his domain of analysis to two-fra e stimuli, Ullman has not had to deal with a central question of motion perception: when does motion analysis begin? Implicit in his theory is the assumption that analysis waits until the second frame arrives, however long that may take. Ull an does adhere to the computational approach in the final section of the book, on the determination of object shapes from motion. Projection geometry is used to discover the precise relationship between objects undergoing arbitrary motions, and their two-dimensional images. The fundamental computation underlying the perception of shape from motion must utilize this relationship. Taken as a whole, the importance of Ullman's work is not that it provides an analysis that is complete in many details, or that adheres strictly to the computational approach. Readers will find that some assertions of the theory are controversial, and some experiments are inconclusive (while others are quite ingenious). But Ullman's greatest contributions are the foundation and framework he constructs for the analysis of motion perception, as well as questions raised and assertions made within this framework that will motivate futurer research. As such this book is essential reading for all those interested in the visual perception of motion.