Computational Characterization of Decision Making During Trans-saccadic Visual Perception

When viewing the environment, humans execute fast sequential saccadic eye movements that disrupt the flow of visual information, yet still preserve stability in their visual percept. At the time of the saccade subjects often experience saccadic suppression of displacement-a failure to notice changes in the location of visual objects slightly shifted during the eye movement. Although well studied, the weighing of sensory information and the extent to which this integration influences visual perceptual judgments is still poorly characterized and understood. Here we used a drift diffusion modeling (DDM) framework to systematically examine the extent to which sensory information biased perceptual judgments when detecting trans-saccadic shifts of visual targets. Healthy human participants ( N = 29, 20 female) completed a visual perception task in which a visual target was shifted following a cued saccadic eye movement (4° or 8°). Trans-saccadic target displacements occurred up to ±2.5° along the horizontal axis, and participants reported the shift direction with a button press response. We present a version of the DDM that takes the visual error as input to simulate offsets in evidence accumulation during decision-making. The model provides an explanatory framework for the integration of sensory information, capturing the temporal dynamics of evidence accumulation that explain the range of perceptual biasing and response timing effects observed across participants.

Maintaining a stable visual percept between saccadic eye movements relies on the integration of externally-generated sensory signals and internally-generated signals based on the executed movements. In this work, we provide a computational framework (a modified drift diffusion model) to quantitatively establish the relative contribution of visual error signals to perceptual judgments and decision-making behavior following saccades. We show that greater reliance on visual error leads to increased biasing effects on response accuracy and manual reaction time. These effects are characterized in the DDM by a visual error-based modulation of the drift rate, representing evidence accumulation. Collectively, the results provide novel measures that capture the influence of sensory information on decision-making and explain the inter-subject differences observed in trans-saccadic perception.