The Bernstein Center Freiburg

Informal Bernstein Seminar
Christian Starzynski
Institute for Physics
University Potsdam
Potsdam-Golm, Germany
Fixationsbewegungen des Auges bei bewegtem Stimulus

Thursday, April 9th, 2009
10:00 h sharp
Library, first floor
BCCN Building
Hansastr. 9a
When looking at a stationary object, your eyes perform miniature movements, which continually refresh the input to the retinal receptor systems. Without these fixational eye movements, the image rapidly fades from perception due to retinal adaption.

The analysis in this study shows that stochastic movements of a target stimulus relative to gaze position can improve perception. An experiment with eye-movement recordings was conducted. Participants were required to respond to a two-choice disrimination task as fast as possible. The discrimination stimulus performed a random walk over a static random-dot background. Different noise strengths were applied to the random walk.

A U-shaped curve of reaction time against noise-strength of the random walk was observed. A linear mixed-effects model approach was used for the stastistical analysis. As a result, the minimum of the reaction time was 84 ms shorter than its maximum. This phenomena mimics stochastic resonance, i. e., noise-enhanced performance in the visual perception, which was previously found in other human sensory systems.

A discrete random walk model is proposed which can reproduce the nonlinear function of reaction times. In this model, discrimination is a stochastic process based on elementary decision processes and accumulation of information. Finally, a method for detection of the unstable periodic orbits (UPO) is applied to the gaze positions collected in the discrimination experiment. The method transforms the gaze data such that the density function of the transformed data has peaks around fixed points.

The same transformation is applied to surrogate data, which shares the same statistical properties as the original gaze data but is created via a stochastic process. The existence of UPO's in gaze data acts as an indicator for unstable nonlinear dynamics. This leads to the conclusion that the underlying process in the generation of fixational eye-movements is not purely stochastic.
The talk is open to the public. Guests are cordially invited!