Decoding of visually guided and interceptive saccades from area LIP of macaque monkeys

Abstract

Humans and non-human primates (NHPs) use a combination of saccades and smooth pursuit when tracking a suddenly appearing moving visual target. Previous behavioral studies have shown initial (so called interceptive) saccades to be spatially accurate. This is remarkable since it requires to extrapolate the motion direction and speed of the target and to anticipate the delays induced by the programming and execution of the interceptive saccade. The neural basis of this process is still under debate. The Lateral Intraparietal Area (area LIP) is part of the saccadic control network. Its neurons show peri-saccadic activation that is dependent on the saccade trajectory, but the spatial tuning is rather broad. We have shown before that LIP-activity allows to decode target position and amplitude of interceptive saccades when saccade and pursuit direction are co- or anti-aligned (1-D task). Here, we probed decoding in full oculomotor space (2-D task). We recorded peri-saccadic activity from 105 neurons in two macaque monkeys. In each trial, an initial step of the saccade target was followed by its movement in one of eight directions. In our data analysis, we first determined the tuning of neurons during saccades towards stationary targets and then trained a neuronal network to decode the saccade landing point from the combined activities of the neuronal population. We found that a combination of even a relatively small number of 50 tuned neurons allowed to predict the saccade end-point with an accuracy of <1°. In a second step we used the same previously trained neuronal network to predict the landing points of interceptive saccades. We found that the prediction was successful, however, it was less accurate than for the saccades towards stationary targets. We conclude that interceptive saccades in full 2-D space are represented in population activity in area LIP.