Superior colliculus neurons mediate the dynamic characteristics of saccades

1. The locus of activity within the superior colliculus (SC) is related to the desired displacement of the eye. Current hypotheses suggest that the location of this locus of activity determines the amplitude of the saccade and that the level of activity at this locus determines eye velocity. We present evidence that suggests that, although the locus determines the amplitude of the saccade, the level of activity in the colliculus encodes dynamic motor error (the difference between desired and current eye displacement). 2. We categorized 86 neurons in the intermediate and deep layers of the superior colliculus of two rhesus monkeys by their activity in relation to the end of saccadic eye movements. In 36% of the cells (n = 31), activity was completely cut off by the end of the saccade (clipped cells). For 53% of cells (n = 46), the major burst of activity ceased by the end of the saccade, but activity continued for 30-100 ms after the end of the movement (partially clipped cells). The remaining 10% of the cells (n = 9) had no clear burst of activity (unclipped cells) but rather had activity that increased gradually before the saccade and then slowly decreased for up to 100 ms after the saccade. These categories were part of a continuum of cell types rather than discrete classes of cells. 3. We first determined whether this new categorization of cells revealed a special relation between the discharge of clipped and partially clipped cells and saccadic amplitude and peak velocity. As expected, we found a steady increase in spike count as saccadic amplitude increased up to the center of the movement field, and an increase in peak spike discharge as peak velocity increased up to a maximum radial eye velocity. Variability in the cell discharge was substantially greater than the variability of saccadic amplitude or peak velocity. We concluded that these single point or averaged measures did not reveal any new functional relationship of these cells. 4. We then examined the relationship of the temporal pattern of discharge of clipped and partially clipped cells to instantaneous changes in radial error and radial velocity. There was a monotonic decay in spike discharge with declining radial error. In contrast, there was a complex, multivalued relationship between spike discharge and radial velocity; collicular cells produced two different values of spike discharge for the same velocity, one during acceleration and the other during deceleration of the eye during a saccade. When saccadic duration and cell discharge were normalized within groups of clipped and partially clipped cells of the same amplitude range, these relationships of radial error and radial velocity to spike discharge were maintained. 5. When all amplitude groups were normalized and averaged together, the relationship of radial error to spike discharge remained very close to linear. However, the complex, multivalued relationship of radial velocity to spike discharge became much closer to single valued, because cells whose discharge either led or lagged the changes in velocity were averaged together. 6. On the basis of these observations, we propose a collicular feedback model of the saccadic system that places the SC inside the feedback loop controlling saccadic amplitude. This model can account for the dynamic relationships of collicular firing to both saccadic amplitude and velocity. It also provides insight into previous experiments showing that a change in the level of activity in the SC can change saccadic velocity. Our model shows explicitly how the superior colliculus participates in transforming the spatial locus of collicular activity into the rate of discharge needed to innervate eye muscles (the spatial-to-temporal transformation).