Abstract
We are currently attempting to use eye movement recordings to observe details of the time course of visual processing. Previously, we have used a correlogram analysis to quantify changes in pursuit arising from changes in color and contrast (ARVO ‘98). In this paradigm, an observer attempts to maintain fixation on a randomly moving target. Next, the smooth “de-saccaded” component of the eye velocity is cross-correlated with the stimulus velocity, and the resulting signals are averaged for many different random stimuli. This average signal may be thought of as the impulse response of the smooth pursuit velocity to stimulus motion. Curiously, this impulse response is sometimes biphasic: 150 msec after a rightward target motion, the eye may move to the right, but 50 msec later move in the opposite leftward direction. One possible explanation is that we are seeing a consequence of the biphasic temporal impulse response of the visual system (the negative lobe of the contrast response producing a “reverse-phi” effect). Another hypothesis is that this result is somehow artifactual, perhaps resulting from the saccade-cutting procedure. To help discriminate between these possibilities, and better understand the correlogram analysis, we have constructed a computational model of oculomotor tracking. The model has two inputs (sensed position and velocity) and outputs a smooth velocity and, depending on the parameter settings, saccades. In its simplest form, pursuit is a delayed and attenuated version of the stimulus velocity, and saccades are made when the position error exceeds a threshold. (A more sophisticated pursuit model will eventually be substituted, but is unlikely to influence the effect of saccade cutting.) A large variety of behaviors can be simulated, none of which result in a biphasic correlogram. We therefore conclude that the biphasic correlograms are not an artifact of the analysis but may reflect the inner workings of visual motion processing.
supported by NASA RTOP 711-51-12.