In objective terms, motion is the continuous displacement of an object with a spatial frame of reference; as such, motion is fully described by physical measurements (direction and speed). In subjective terms, however, the corresponding sensation of motion is not so easily defined or understood. Because the real-world displacement of an object is conveyed to an observer by a changing projection on the retinal surface, the velocity field that uniquely defines motion in physical terms is ambiguous with respect to the possible causes of the changing retinal image; indeed, a large number of physical displacements can generate the same stimulus sequence. As in other aspects of vision, this ambiguity, as noted, presents a fundamental problem, namely how in the face of uncertainty the brain generates quite definite percepts that usually (but not always) allow the observers to deal with the real-world cause of the retinal stimulus.

An Empirical Solution to the Aperture Problem?

One solution to this problem is evidently to accumulate experience interacting with moving objects, such that motion percepts gradually come to accord with the statistics of the possible displacements underlying the stimulus. The physical correspondences of the points along the line in any two sequential images cannot be determined directly because some points come into view, others disappear, and still others could represent deformation as the line in view 'expands' or 'contracts'. Although it is generally believed that physical correspondence is the best requirement for a stimulus to be correctly perceived as moving (both motion and stereoscopic vision have long been assumed to be 'correspondence problems'), it seems more likely that motion perception is derived empirically from the complete set of possible correspondences and differences between any sequential images.

Since the relative contribution of these correspondences and differences to the physical displacements underlying the stimulus cannot be determined by inspection of stimulus per se, the problem posed by this inevitable uncertainty can only be solved empirically by generating motion percepts based on past experience of what such stimulu have typically turned out to be.

To assess this conception of perceived motion, one can compute the probability distribution of the possible physical displacements underlying simple image sequences. In this way they could predict the percepts that subjects would be expected to see on a wholly empirical basis, and compare them to actual performance. Given the fundamental ambiguities embedded in the correspondence and differences underlying sequential images, any motion stimulus, this approach entails: 1) describing quantitatively how these correspondences and differences in the image plane are generated in case of a straight line moving in fronto-parallel plane; 2) using this information to derive a set of probability distributions of the possible real-world displacements underlying the stimulus; 3) deriving a principle for combining these probability distributions based on the statistical structure of the underlying events in the stimulus; and 4) devising a procedure for predicting motion perception based on this joint probability (see Yang, et al., 2001a).

References

Yang Z, Shimpi A, Purves D (2001) A wholly empirical explanation of perceived motion. Proc Natl Acad Sci USA 9:5252-5257.

Yang Z, Shimpi A, Purves D (2002) Perception of objects that are translating and rotating. Perception:31 925-942.