Abstract
Background: Spatiotemporal objects are those whose parts become visible only gradually through motion and dis-occlusion, and must be integrated to form perceptual units. Rather than representational invariance, the perception of spatiotemporal objects requires precise positional and velocity information about fragments in order to accurately align them across space and time. The data we present here suggests that intermediate and dorsal visual areas may be critical for their perception. Design: Spatiotemporal objects were defined by spatiotemporal boundary formation (SBF; Shipley & Kellman, 1993, 1994, 1997). An invisible object expanded and contracted on a background of randomly oriented Gabor elements. Whenever the object came into contact with an element, that element rotated in place by a random amount or was displaced in a random direction. Even though only a few elements changed between frames, the percept was of an expanding global form with clear boundaries. fMRI data were collected while subjects passively viewed expanding and contracting and squares, circles, or radial frequency (RF) patterns defined by elements that either rotated or displaced (6 conditions). RF patterns did not produce clear global forms and were used as a control condition. In separate scanning sessions, we functionally identified 20 topographically organized ROIs in early, intermediate, dorsal, and ventral visual areas. Results: An SVM classifier was able to distinguish between shape (circle/square) and no-shape (RF patterns) conditions as well as between circles and squares throughout visual cortex including areas not commonly identified in shape-representation studies including: V3A,V3B, LO1, LO2, TO1, TO2, IPS0-IPS5. Conclusions: Spatiotemporal object identity is represented not only in ventral, but also in intermediate and dorsal visual areas. Dynamic perceptual units formed from object fragments distributed in space and time may first arise in dorsal and parietal areas and then be passed on to ventral ones for recognition.
Meeting abstract presented at VSS 2015