Abstract
Purpose. Recent data indicate that the primate rostral superior colliculus (rSC) might provide target-position information for both pursuit and saccades. To test whether the rSC could guide a monkey's oculomotor decisions, we have now recorded from buildup neurons while monkeys chose between two stimuli, and applied ideal-observer (ROC) analysis to simulate the population activity on single trials. Methods. Monkeys tracked one of two differently colored stimuli, based upon a prior cue. For saccades, two stationary stimuli were presented in opposite hemifields at an eccentricity of 3.5o, slightly above or below the horizontal meridian. For pursuit, two moving stimuli initially appeared at similar locations as on saccade trials, but moved horizontally toward the center of the display at 15o/s. On interleaved trials, either the cued ‘target’ or the non-cued ‘distractor’ stimulus appeared within the response field of the neuron. For the ideal-observer analysis, we repeatedly (n=5000) and randomly selected one ‘target’ and one ‘distractor’ trial from each neuron and used the distributions of firing rate across the population (n=83) to compute the probability of correctly identifying the target for each ms time-point. Results. This probability increased from chance ∼150 ms after stimulus onset and reached average peak values of ∼0.7 for pursuit and ∼0.8 for saccades. Applying fixed criteria to these time-varying probabilities correctly reproduced the monkey's decisions; the predicted latencies of movements depended on the amplitudes of the criterion. Low criteria (p<0.6) predicted “false alarms” (latencies < 50 ms) and high criteria (p>0.65 or 0.78, pursuit or saccades) predicted “misses” (criterion never reached). Criteria within these limits predicted pursuit and saccade latencies near the observed values. Conclusions. The population activity of rSC buildup neurons could provide evidence of target location for deciding where and when to initiate pursuit and saccades.