Abstract Fundamental to any theory is the set of elements upon which the theory works. Hence fundamental to any theory of vision systems must be the elements of vision, the features that are selected from the stimulus flux, and that are recombined by the visual system according to a grammar or set of rules. Can one identify such elements for vision from experiments on human perception? The conclusion derived from a number of experiments suggests that the objects that are characteristically operated upon in machine vision systems cannot be identified with the elements of human form perception. As some of these complexities may be of interest and of challenge to engineers and computer scientists concerned with vision systems, a few phenomena now widely studied by psychologists and other behavioral scientists are described. A significant literature has developed around most of these phenomena, but as a matter of convenience the author describes aspects of them that he has worked on himself. In most cases the supportive literature is also cited for reference.