Population coding of distinct categories of behavior in the frontal eye field

Abstract

Brain regions frequently contribute to the control of a range of behaviors. To understand how a brain area controls multiple behaviors, we examined how the frontal eye field (FEF) encodes different eye movements by recording the activity of 1,200 neurons during smooth pursuit, pursuit suppression, and saccade tasks in two female Macaca fascicularis monkeys. Single neurons tended to respond on all tasks. In the absence of task-specific clusters, we analyzed the relationships in directional preference between tasks. The tuning curves during the pursuit and suppression tasks were strongly correlated, unlike the correlations between the pursuit and saccade tasks that were considerably weaker. To study the implications of single neuron coding, we examined the patterns of population activity on the three tasks. We identified the low-dimensional subspaces that captured the most variance in population activity during each task and quantified the extent of overlap between these spaces. The absence of overlap between the subspaces spanned by population activity on the pursuit and saccades tasks prompted an independent linear readout of these tasks. Conversely, pursuit and pursuit suppression showed substantial overlap in their population activity subspaces. This overlap emphasized the predominance of visual motion in pursuit encoding and indicated that the linear readouts accounting for a large part of the variability in the pursuit tasks cannot completely attenuate the activity during suppression. Overall, these results imply that at the population level, FEF is organized predominantly along sensory rather than motor parameters.