Abstract

The present thesis was directed at the hypothesis that to perceive an event is to detect an invariant specific to the style of change and an invariant specific to the object undergoing the style of change--the reciprocal invariants of transformation and structure, respectively. Events were defined by seven, well-defined global mathematical transformations that produced a variety of styles of change over different objects. In Experiment 1, each transformation was applied dynamically to four different facial profiles, presented visually by computer. Observers distinguished reliably among these events on the basis of their respective transformational invariants. Experiments 2 and 3 examined the generative and abstract aspects of transformational invariants. In each, an event was presented as a sequence of static samples and examined in an interpolation task and an extrapolation task . The transformational invariants proved sufficiently generative to specify retrospectively and prospectively and sufficiently abstract to specify both event tokens and event types. Experiment 4 addressed the relative persistence of a style of change over structure. The stimulus sequence was constructed of different objects transformed the same way. Types of change and types of object structures interacted, implying that certain styles of change are more conducive to certain object structures. Experiment 5 found that, with minor qualifications, the structural invariant was reliably detectable without being affected by the particular style of change applied. Further, some styles of change, when sufficiently magnified, induced non-trivial changes in underlying structure sufficient to render successive instances of the structure non-identical. Taken together, the experiments provide support for the hypothesis that perceiving an event is essentially equivalent to detecting the two reciprocal invariants of transformation and structure that together specify the event