Abstract
Due to the aperture problem, the initial direction of tracking responses to a translating tilted bar is biased towards the direction orthogonal to the orientation of the bar. This directional error is largely reduced over the first 200 ms of tracking, consistent with the neural solution of the aperture problem (Pack & Born, 2001) and is fully corrected during the steady-state. We have proposed that pursuit dynamics reflects that of visual motion processing and can be modeled as a dynamical Bayesian inference of 2D target motion (Bogadhi et al., 2010). Such simple paradigm also offers a powerful way to explore interactions between sensory and predictive signals in controlling action (Montagnini et al., 2006). We conducted two experiments to investigate these interactions by transiently blanking the target at different moments of pursuit. First, a 45° or 135° tilted bar translating horizontally was blanked for four different durations (0, 200, 400 and 800 ms) during steady-state tracking. Bar orientation after reappearance changed on half of the trials. We found a marginal directional bias (compared to initial bias) when the target reappeared with no change of orientation and when target changed orientation with no transient blanking. However, there was a significant directional bias when the target reappeared with a change in orientation. Second, the target (an upright or a 45° tilted line) was blanked (duration: 200 ms) on half of the trials during the initiation phase of pursuit, starting at either 100 or 180 ms after pursuit onset. Line orientation was constant. We found a significant directional bias (i.e. equal to the initial bias) when 45° tilted line reappeared after a 200 ms blank starting at 100 ms. Such bias was marginal when the target reappeared after a 200 ms blank starting at 180 ms. These results suggest for a conditionally weighted mixing of retinal and extra retina signals in driving smooth pursuit.
EU Marie Curie ITN & CNRS.