Extracting self-created retinal motion

Abstract

INTRODUCTION. Self movement generates retinal movement that is perceptually distinct from other movement. There are two types of models for how this distinction might be achieved. In the first, after self motion is detected, an internal estimate of the expected retinal movement is subtracted (a linear process) from retinal image movement. Remaining movement is interpretted as indicating external movement. In the second model, subjects internally compare observed visual motion with their internal representation: a non-linear ratio judgement which depends on the magnitude of the expected movement. A discriminable difference indicates external movement. These models respectively predict linear and non-linear distributions of the probability of regarding a given retinal motion as perceptually stable. METHODS. Our subjects' task was to distinguish self-produced from external visual motion during rotation around the yaw, pitch and roll axes and during translation in the x, y and z directions. They wore a helmet-mounted display whose position was monitorred by a mechanical head tracker with minimal lag. The visual display was modified in response to head movement. The ratio between head and image motion was adjusted by the subject until the display appeared earth-stationary. RESULTS. The distribution of ratios judged to be perceptually stable were fitted with a normal and a log normal distribution. For the rotation data a better fit was found using the log normal distribution suggesting that the non-linear ratio model is a better description of the underlying neural computations involved. No clear difference was found for the translation data.

Sponsorred by CRESTech and NSERC.