Abstract
Adaptation phenomena offer valuable insight into the mechanisms of visual coding. Here, evidence from psychophysical adaptation is reviewed which suggests that common computational principles underlie the processing of orientation and motion despite their distinct cortical loci. Adaptation has been shown to affect the perception, discrimination and detection of both orientation and direction-of-motion. Use of a double-angle representation for orientation (Gilbert & Wiesel, 1990) allows direct comparison of the angular tuning functions of these phenomena between the orientation and motion domains. For example, the tilt aftereffect (TAE) depends in a characteristic way on the angle between adapter and test (Gibson & Radner, 1937). The angular dependence of its motion analogue, the direction aftereffect (DAE) (Schrater & Simoncelli, 1998; Levinson & Sekuler, 1976), is strikingly similar. Notably, the “critical” values for the DAE are consistently around twice those for tilt, with the peak repulsive and attractive effects occurring at adapter-test angles of 30° and 150–160° for direction-of-motion as opposed to 10–20° and 75–80° for tilt. A similar relationship is shown to exist between the angular tuning functions for post-adaptation tilt and direction discrimination (Regan & Beverley, 1985; Phinney et al., 1997). This common phenomenology suggests that similar coding principles apply in the processing of tilt and motion. Specifically, it is argued that, in both domains, adaptation serves to reduce the transmission of redundant information by the visual system (Attneave, 1954; Barlow, 1990). A neurally plausible model based on this principle is proposed that makes testable predictions at the single neuron level.