The visual specification of events: a response to Ullman.

Abstract

Ullman (1977), in a reply to an earlier paper in this journal on object identity in apparent movement (Warren, 1977), questioned the validity of the finding that displays were perceived as a single object in motion and displays as two separate objects. He produced a wire-frame object which, when viewed from one angle, cast the outline of a square and, after a 50 rotation, cast the outline of a triangle. This demonstrated that the two figures or phases of the allegedly nontransformable square-and-triangle display (j) were in fact related through such an object. Hence, Ullman claimed, the principled distinction between ecologically transformable and nontransformable displays was untenable and the original finding simply a reflection of other factors, perhaps timing parameters. It is obvious that there are many possible objects which have both a square and a triangular projection. Each may be described by a projective group of all the perspectives on that object, each group containing a square and a triangle as members. This fact gives rise to two related questions: (1) Why is display (j) seen as two separate objects instead of one of the many possible objects with a square and a triangular projection? In other words, what makes. the display nontransformable? (2) Given a display which is seen as one object, why is it seen as that particular object instead of another of the possible objects which also contain the given projections? In his critique, Ullman has misinterpreted the general claim of the ecological approach to event perception (Shaw, Mcintyre, & Mace, 1974; Shaw & Pittenger, 1977) and the claim as it was stated in the previous article (Warren, 1977). The proposal is not that nontransformable displays themselves are ecologically unrelated, but that they will fail to specify a unified object under transformation to an observer (see p. 265, paragraph 2). Arbitrarily distant members of a projective group are not necessarily sufficient to specify a continuous object. The fact that they are mathematically related does 'not mean that they will be perceptually identified as different views of the same object. In fact, a central problem of event perception is to determine and appropriately characterize the visual information required to specify a certain event. Apparent movement displays which perceptually specify a transformaton should have a higher salience than those which underspecify it. Similarly, ecologically significant transformations should have a higher'salience than nonsignificant ones, because the perceptual system has been tuned over a period of phylogenetic and ontogenetic time to detect them. However, this does not mean that an individual could not see or learn to see nonecological or underspecified transformations, given the appropriate context, lack of context, or attentional set. Animated pumpkins do appear to turn into carriages. However, under apparent movement conditions, such transformations should be less salient, more unstable, and more ambiguous. Addressing the event perception problem, Shaw, McIntyre, and Mace (1974) have proposed that a display which specifies an object's period of symmetry, in this case the degree of rotation which brings the object back into congruence with itself, is sufficient for the perception of a continuous object and its shape. In such a situation, it has been argued that the display acts as the generator set for a complete group of perspectiv:es (Shaw & Wilson, 1977), thereby perceptually specifying the intermediate views seen under apparent movement conditions. Hence, to answer the first question regarding when two rather than one object should be seen: When the rotational symmetry period of a possible object is not given, no group of perspectives is specified, and two separate objects should be perceived. Such displays have been called ecologically nontransformable. The possible alternative, the continuous deformation of a rigid object, should have low salience due to its relative ecological implausibility. In response to the second question regarding multiple possible objects: What is specified by a transformable display is that object and only that object whose symmetry period is given. If additional information becomes available then a more complex object may be specified and perceived. By analogy, a soap bubble spanning increasingly complex structures will preserve a minimal surface area within the given constraints. In this sense, the ecological approach to event perception incorporates a of perceptual optimality, a not yet fully understood but related to the suggested principle of