Abstract
As humans scan the surrounding world, each saccade brings an area of interest from low-resolution peripheral vision into high-resolution foveal vision. To maintain perceptual stability across saccades, these pre- and post-saccadic percepts must be integrated. Humans are able to achieve trans-saccadic integration in a near-optimal manner (Ganmor, Landy, & Simoncelli, 2015; Wolf & Schütz, 2015), however it is unclear if integration can happen as soon as the information from pre- and post-saccadic stimuli becomes available, or if integration requires the longer time usually taken to plan and execute a saccade. We measured the time-course of integration both at the saccade target and at a location between the target and initial fixation, to determine how long a stimulus needs to be presented for integration to occur. Participants were presented with oriented Gabors either pre-saccadically, post-saccadically, or both. The Gabor was presented for a variable time before and/or after saccade onset, to reduce the amount of time the stimulus information was available. Participants responded whether the Gabor was tilted clockwise or counter-clockwise. Discrimination performance was calculated for stimulus presentation durations ranging from 10-100ms, to create a continuous time-course of performance for pre-saccadic, post-saccadic and integration conditions. The results show that integration occurs even when the stimulus is only presented briefly. We compared integration performance with predicted performance for different cue combination models, showing that an integration model with early noise best describes integration performance for the majority of participants. The model comparison also shows that integration benefits are not due to increased exposure duration of either pre- or post- saccadic information alone. These findings suggest that integration can occur when only very little information is available before or after a saccade. Integration also seems to be accomplished by independent channels for pre- and post-saccadic information rather than a single, spatio-topic channel.
Meeting abstract presented at VSS 2018