Abstract
Even during fixations the human eye is never fully at rest. Tremor, drift, and microsaccades constantly perturb gaze position, moving the visual input across the photoreceptor mosaic of the retina. While fixational drift refers to slow random movements of the eyes, microsaccades are rapid small-amplitude events that follow the kinematics of larger saccades. Statistically, the random motion of drift is positively correlated on short timescales (persistence). On longer time scales, the motion becomes antipersistent, where negatively correlated increments limit the eyes’ excursions compared to uncorrelated Brownian motion. Previous research has shown that some of the statistical characteristics of fixational drift and microsaccades are qualitatively captured by a self avoiding random walk (SAW model). Here, we implemented an advanced version of the SAW model that permits likelihood-based model inference. We used a fully Bayesian framework for parameter estimation at the level of individual observers. As the parameters of a mechanistic model represent interpretable values, the distributions themselves are informative. In a first step we use posterior predictive checks to show that the model reproduces statistical tendencies in the data. Additionally, the model is capable of representing interindividual differences. In a second step we aim to investigate the internal model states during triggering of microsaccades. This analysis provides new insights into the spatial and temporal relationship of microsaccades with fixational drift, which might contribute new mechanism for the unexplained variations of microsaccade rates across individual observers. Fixational eye movement and microsaccades are fundamental components of the visual processing stream; understanding variations in their statistics is critical for understanding active visual perception.