Abstract
Despite considerable evidence for attention-mediated changes in salience as the basis for third-order motion (Lu & Sperling, 1995, 2001), the nature of the motion mechanism responsible for its actual perception has not been established. A counterchange-sensitive, directionally selective motion detector has been proposed for this purpose. It entails the detection of oppositely signed changes in activation at pairs of spatial locations. A recently developed computational model based on the counterchange principle (Hock, Schoner & Gilroy, 2009) accounts for a wide range of phenomena for both generalized apparent motion stimuli (oppositely signed changes in contrast for two simultaneously visible surfaces) and standard apparent motion stimuli (when a surface is displaced its contrast disappears at its initial location and re-appears at its new location). An updated version of this model is presented which, in addition, accounts for the dynamical properties of apparent motion perception; i.e., it accounts for its bistability (both motion and nonmotion can be perceived for the same generalized apparent motion stimulus), the temporal persistence of these perceptual states, and the effects of adaptation. The feed forward path for the counterchange detector is composed of two biphasic subunits, one activated by decreases and the other by increases in activational input, with the motion detector's output determined by the multiplicative combination of the subunit activations (if both subunits are sufficiently excited, motion is perceived from the location of the decrease to the location of the increase in activation). Motion/nonmotion bistability is established by activation-dependent feedback from the output of the motion detector to its biphasic subunits. The temporal persistence of these states for back-and-forth motion and the temporal integration of successive motions in the same direction are accounted for by activation-dependent interactions among different directionally selective motion detectors. The model makes several novel, experimentally-testable predictions that will further inform its plausibility.