Neuronal representations in a categorization task: sensory to motor transformation

When motor actions are triggered by sensory events, the neural networks that generate the motor commands must receive an adequate driving input, which, in general, is derived from the original sensory signals. We studied this conversion of sensory information into motor commands in a task in which monkeys categorized the speed of tactile stimuli as either high or low, reaching for one of two switches to indicate their choice. The input signal, the speed of probe movement, was varied systematically. The output of this process was an arm movement that did not depend directly on speed, but on a function of speed, its category, and the monkey had to compute it to obtain a reward. There are three essential quantities in this task: stimulus speed, speed category, and arm movement. Where and how are these variables represented in terms of neuronal activity? Are those representations independent? And, do they correlate with the monkey’s behavior? Here we investigate these questions through analytical techniques applied to extracellular recordings from three cortical structures, primary somatosensory cortex1 (S1), the supplementary motor area2 (SMA, or medial premotor cortex), and primary motor cortex3 (M1). We suggest that the primary sensory representation of the stimulus is used to compute two sets of responses that encode speed category and that participate in the generation of arm movements.