Physiological, Computational, and Psychological Approaches to Understanding Spatial Vision

Project Summary Every time we make a saccade, the retinal image of the world changes drastically, yet the world appears stable to us. Patients with parietal cortex lesions do not have the ability view the world stably across saccades. This trans-saccadic visual stability (TSVS) problem has received renewed attention since the discovery of receptive- field (RF) remapping: Some cells in the lateral intraparietal area (LIP), frontal eye fields (FEF), and other brain areas shift their current, pre-saccadic RFs (cRFs) toward their future, post-saccadic RFs (fRFs), even before the saccadic onset (forward shift). These cells likely contribute to TSVS by predicting and comparing retinal images across saccades. However, under certain circumstances FEF neurons shift their RFs toward the saccade target (convergent shift), instead of toward fRFs (forward shift). This raises many important questions including 1) whether the forward RF shift, a leading physiological substrate for TSVS, actually exists in LIP and FEF; 2) whether these areas differ in their peri-saccadic RF dynamics; 3) whether LIP and FEF cells have both convergent and forward shifts; and 4) if so, what underlying mechanisms could explain them. Furthermore, it is widely assumed that the forward and convergent RF shifts produce, respectively, the observed forward and convergent mislocalization around the saccade onset, but the assumption has not been carefully evaluated. Here we will address these questions by simultaneously recording LIP and FEF neurons, modeling circuit mechanisms for the RF shifts, analyzing the shifts' perceptual implications, and testing new predictions. Our preliminary data suggest that on average, LIP and FEF are biased toward the forward and convergent shifts, respectively. Importantly, the convergent shift first appears in a delay period when the saccade command, and thus its corollary discharge (CD), are suppressed. This leads to our hypothesis that unlike the forward shift known to be produced by the saccade CD, the convergent shift is produced by saccade planning/attention to the target. While CD-gated lateral connections can explain the forward shift, we further hypothesize that attention- modulated center/surround connections account for the convergent shift. We will demonstrate that both connectivity patterns emerge automatically in neural networks trained on a memory-saccade task related to TSVS. Perceptually, our analysis indicates that RF shifts in one direction produce either no mislocalization or a mislocalization in the opposite direction, and that peri-saccadic mislocalization is confounded by mislocalization induced by retinal motion alone. We will test many predictions of our models including that the forward shift grows with saccadic amplitude whereas the convergent shift peaks at an intermediate distance between cells' cRFs and the target, and that convergent RF shifts during the delay period generate a divergent mislocalization. The project will help resolve a main controversy regarding RF shifts in LIP and FEF, provide a mechanistic and functional account of the two shift types, and clarify the perceptual consequences of RF shifts. Understanding TSVS may lead to therapeutic strategies for patients whose brain lesions have interfered with it.