Interaction between visual and prior information on superior colliculus neurons

Abstract

Purpose: Saccades are guided by prior information as well as by sensory signals. Saccade-related neurons in the SC are responsive to both of these factors individually, and it is typically assumed that visual and prior information have additive effects on the activity of SC neurons. To test this idea, we have measured the sensitivity of SC neurons to prior information on a millisecond timescale.

Methods: We recorded from buildup and burst neurons (n=85) in the SC of two monkeys during a luminance discrimination task. Monkeys initially fixated a central fixation cross on a random noise background, and two boxes at 6 deg on either side of fixation demarcated possible target locations. On cued trials one box briefly reversed contrast, indicating the likely location of the upcoming target (75% certainty). After a random interval, two Gaussian-blurred discs were added, one slightly brighter than the other. The monkey was rewarded for making a saccade to the brighter disc.

Results: We constructed “cue weighting” ROC curves to assess the neuronal sensitivity to prior information. The “signal” distribution came from trials on which the target and the cue were presented in the response field; the “noise” distribution came from trials on which only the target appeared in the response field. For the population of neurons that showed higher activity with the cue (32/85), the average ROC area was significantly larger prior to stimulus onset than during the visual response. During both the baseline and visual epochs, neuronal sensitivity to prior information was correlated with the behavioral bias caused by the cue.

Conclusions: The sensitivity of SC neurons to prior information is not strictly additive with visual inputs but instead changes within a trial, especially during the visual response. The correlation of these effects with bias suggests that the interaction between visual and prior information on SC neurons reflects the current behavioral strategy of the subject.

Footnotes

 Supported by: NASA NAG 9-1465 (RJK) and NSF GRFP (DL)