Causal Inference for Spatial Constancy across Saccades

Abstract

Our ability to interact with the environment hinges on creating a stable visual world despite the continuous changes in retinal input. To achieve visual stability, the brain must distinguish the retinal image shifts caused by eye movements and shifts due to movements of the visual scene. This process appears not to be flawless: during saccades, we often fail to detect whether visual objects remain stable or move, which is called saccadic suppression of displacement (SSD). How does the brain evaluate the memorized information of the presaccadic scene and the actual visual feedback of the postsaccadic visual scene in the computations for visual stability? Using a SSD task, we test how participants localize the presaccadic position of the fixation target, the saccade target or a peripheral non-foveated target that was displaced parallel or orthogonal during a horizontal saccade, and subsequently viewed for three different durations. Results showed different localization errors of the three targets, depending on the viewing time of the postsaccadic stimulus and its spatial separation from the presaccadic location. We modeled the data through a Bayesian causal inference mechanism, in which at the trial level an optimal mixing of two possible strategies, integration vs. separation of the presaccadic memory and the postsaccadic sensory signals, is applied. Fits of this model generally outperformed other plausible decision strategies for producing SSD. Our findings suggest that humans exploit a Bayesian inference process with two causal structures to mediate visual stability.

Causal inference for spatial constancy across whole-body motion

“…The participant was then required to indicate the world-fixed position of the target presented before the motion. The purpose was to investigate whether in this situation of passive self-motion and uninterrupted visual input the brain solves the position updating task by combining the available memory and sensory information, on the one hand the internally-updated position of the premotion target, denoted m , and on the other hand the post-motion probe target position, denoted v , in a statistically optimal fashion, i.e., according to a causal Bayesian inference mechanism [7,9]. The ideas of this approach are now summarized informally, but more details can be found in the supplemental material.…”

“…Recently, we provided evidence for this strategy using the saccadic suppression of displacement task [8], testing how participants judge the presaccadic location of a visual object that shifted during a saccade [9]. Following the rules of Bayesian causal inference, integration was strong when predicted and actual feedback represented spatially close target locations (as if they had a common cause), but weakened with larger spatial differences, depending on the precision of these signals [9].…”

Visual landmarks calibrate auditory space across eye movements

“…Upon landing, the perceptual system has access to motor information about the magnitude of the intended movement, proprioceptive information about the movement executed, and a visual error signal indicating the difference between the actual and predicted post-saccadic visual input (Atsma et al, 2016;Collins et al, 2009;Niemeier et al, 2003;Ostendorf and Dolan, 2015). By integrating these different signals, the perceptual system can ensure both that our spatial representations remain stable across eye movements and that our eye movements remain accurate.…”

“…Due to the strong assumption of visual stability (Atsma et al, 2016;Deubel et al, 1998;Niemeier et al, 2003), related to the unlikelihood of an otherwise stationary visual landmark suddenly and abruptly changing position precisely during the saccade, the majority of this error should be attributed to an inaccuracy in the movement.…”

“…Indeed, supporting the idea that a strong assumption for visual stability acts as a prior for landmark stability, studies in which the reliability of the visual landmark was explicitly reduced (by jittering its spatial location) demonstrated that participants' localization responses became noticeably less reliant on the landmark's location (Byrne and Crawford, 2010). Similarly, studies of saccadic suppression of displacement have shown that when the displacement of the saccade target is too large (such that such a large eye error is highly unlikely) the assumption that changes are due to errors in the eye movement is broken (Atsma et al, 2016).…”