Bayesian Dynamical Modeling of Fixational Eye Movements
share
Humans constantly move their eyes, even during visual fixations, where miniature (or fixational) eye movements are produced involuntarily. Fixational eye movements are composed of slow components (physiological drift and tremor) and fast microsaccades. The complex dynamics of physiological drift can be modeled qualitatively as a statistically self-avoiding random walk (SAW model, see Engbert et al., 2011). In this study, we implement a data assimilation approach for the SAW model to explain quantitative differences in experimental data obtained from high-resolution, video-based eye tracking. We present a likelihood function for the SAW model which allows us apply Bayesian parameter estimation at the level of individual human participants. Based on the model fits we find a relationship between the activation predicted by the SAW model and the occurrence of microsaccades. The latent model activation relative to microsaccade onsets and offsets using experimental data reveals evidence for a triggering mechanism for microsaccades. These findings suggest that the SAW model is capable of capturing individual differences and can serve as a tool for exploring the relationship between physiological drift and microsaccades as the two most important components of fixational eye movements. Our results contribute to the understanding of individual variability in microsaccade behaviors and the role of fixational eye movements in visual information processing.
READ FULL TEXT