Abstract
We present a system dynamics model of a simulated real-world optic-flow motion aftereffect (i.e., illusory sense of backward self-motion induced by prolonged viewing of expansive optic flow). To induce the motion aftereffect (MAE), observers viewed on a large display a simulated real-world scene in perspective view over which self-motion was simulated (simulated height above the ground plane = 5 m). The scene was composed of uniform gray terrain upon which a variable number of vertical poles were placed. The poles, whose number ranged from 36 to 576 poles per square km, served as carriers of optic flow information. We also examined terrain with texture. Observers adapted to the optic flow for a duration of either 10, 30, 120, 240 or 480 seconds. Following adaptation, observers viewed a static version of the scene (i.e., the last frame of the simulation) which served as a stationary test pattern. Because the test pattern was static, the response of low-level as well as higher level motion mechanisms likely contributed to the aftereffect. Results: aftereffect duration was approximately proportional to the square root of adaptation duration up to the longest adaptation duration tested (480 secs), at which point aftereffect duration was approximately 50 secs. Moreover, when the MAE was measured across different combinations of speed of simulated self-motion and pole density, the aftereffect was tuned to speed rather than the temporal frequency of stimulation. A system dynamics simulation of the Grunewald-Lankeet computational model (e.g., van de Grind, van der Smagt & Verstraten, 2004) showed that their model does not predict relatively long MAE durations, such as those found in the present study. We discuss how modifications to their mathematical model can be made to account for such data.