A circuit model for transsaccadic space updating and mislocalization

Abstract

We perceive a stable, continuous world despite drastic changes of retinal images across saccades. However, while persistent objects in daily life appear stable across saccades, stimuli flashed around saccades can be grossly mislocalized. We address this puzzle with our recently proposed circuit model for perisaccadic receptive-field (RF) remapping in the lateral interparietal area (LIP) and frontal eye fields (FEF). The model uses center-excitation/surround-inhibition connections to store a relevant stimulus’ retinal location in memory as a population activity. This activity profile is updated across each saccade by directional connections gated by the corollary discharge (CD) of the saccade command. The updating is a continuous backward (against the saccade) shift of the population activity (equivalent to continuous forward remapping of the RFs), whose cumulative effect across the saccade is a subtraction of the saccade vector. The model that correctly updates persistent stimuli, and flashes well before and after saccades, produces the observed forward and backward translational mislocalization for flashes around the saccade onset and offset, respectively, because of insufficient and unnecessary cumulative updating after the saccade, caused by visual response latency and sluggish CD time course. We confirm the model prediction that for perisaccadic RFs measured with flashes before the saccades, the forward remapping magnitudes across the saccades are smaller for later flashes. Our work suggests that transsaccadic perception is stable because the presaccadic retinal position of an object is updated to match the postsaccadic (reafferent) retinal position of the same object, and that the brain uses “unaware” decoders which do not distinguish between different origins of neurons’ activities.