Abstract
Smooth pursuit adds a velocity field to the retinal image in the eyes opposite direction. To correctly perceive the objects physical motion, the visual system must compensate for this self-induced motion. Compensation is usually assumed to involve combining the retinal signal with an extra-retinal eye velocity estimate. According to the linear model, the estimated eye velocity is added to the retinal signal: the resulting motion is perceived as the physical stimulus motion. The linear model gained wide support from studies using collinear motion, and velocities close to the pursuit-target velocity. Furthermore, the pursuit-target has typically been available as a visual referent, allowing observers to judge object-relative rather than absolute motion. Such object-relative information would yield an egocentric bias in the responses. We studied the compensation problem using stimuli moving in different directions with a wide range of speeds. Furthermore, we eliminated the visual referent by relying on residual pursuit after the pursuit-target is extinguished. We fit the linear model to our data, and obtain an extraretinal gain of 0.4, which is in agreement with the eye velocity underestimation reported in the literature. However, we find that compensation for eye movements varies dramatically as a function of retinal motion along the axis of pursuit: retinal motion is compensated when is in the eyes opposite direction, but is perceived largely uncompensated for motion on the retina in the same direction as pursuit. These results contradict the linear model, and suggest that the compensation depends not only on eye movements but also on retinal information: stimulus direction and speed. Thus, our data challenge the classical view that compensation is derived solely from extraretinal sources. We suggest instead that retinal motion is compensated by an eye movement estimate derived from a combination of extraretinal and retinal signals, the latter based on a stationarity assumption.