Transforming sensory perceptions into motor commands: evidence from programming of eye movements

Abstract

The visual stimulus for a saccadic eye movement is encoded in place-coded maps in cerebral cortex and the dorsal superior colliculus. In contrast, the motor command for the saccade is encoded by the temporal discharge properties of ocular motoneurons and premotor burst neurons in the brain-stem reticular formation. Thus, there is need for a spatial-temporal transformation of neural signals, and recent findings suggest that the superior colliculus might contribute to this process. The ventral, output layers of the superior colliculus encode the metric of the desired saccade in polar coordinates. However, premotor neurons in the pontine and mesencephalic reticular formation are organized to generate horizontal and vertical saccades, respectively. Studies of oblique saccades in patients with slow vertical components--due to Niemann-Pick type C disease--support the interpretation that the saccadic command from the reticular formation is encoded in Cartesian coordinates. Currently, saccades are thought to be generated under local, brain-stem feedback control in which current eye displacement is continuously subtracted from desired eye displacement to compute motor error--the remaining movement required for the eye to acquire the target. If the superior colliculus is positioned in the feedback loop, then there is a need for transformation of premotor signals back into a place-coded version of motor error. Recent studies suggest that, during the saccade, this might be achieved by a wave of activity spreading rostrally, which traverses the collicular map in a direction corresponding to progressively smaller movements and finally activates a group of neurons concerned with fixation. These new hypotheses are ripe for testing by basic and clinical studies. By confronting the issue of what signal transformations are required to program visually guided saccades, new experimental approaches have emerged. Such computational approaches offer insights into how the brain controls behavior not just by measuring stimulus and response, but by asking what "currency" is being used by interacting populations of neurons at any stage in the process.