Abstract
Reflexive saccades towards the same appearing stimulus vary substantially in their reaction times (RT’s). While several cognitive processes, like target selection and decision making, contribute to such variability, none of these processes can begin before sensory detection of stimulus occurrence. Consistent with this, variability in the time at which primary visual cortex (V1) neurons first detect stimulus onset is significantly correlated with RT variability (Lee et al., 2010). We asked how trial-to-trial variability in the visual responses of the superior colliculus (SC), a structure with both visual sensitivity and direct projections to oculomotor control circuitry, behaves. We analyzed recordings of 188 SC neurons from two macaque monkeys that performed a reflexive visually-guided saccade task. After 550-800 ms of initial fixation, a 0.51-deg radius disc of 10-100% contrast appeared within the response field of a neuron; the fixation spot was extinguished simultaneously. We estimated the first-spike latency of visual responses using a similar method to (Lee et al., 2010). We also calculated spike likelihood across trials in 2-ms running time bins. Within a given contrast level, we observed a significant positive correlation between trial-to-trial first-spike latency and RT, like in V1. However, numerically, the correlations were surprisingly weaker than in the V1 data of Lee et al. (2010). We then analyzed spike likelihood in the SC visual responses. In this case, and unlike in V1, trial-to-trial variability in firing rate, again within a given contrast level, exhibited stronger correlations with RT (which were now negative: lower spike rates implying longer RT’s). These results suggest that a rate, rather than temporal, code dominates the relationship between trial-to-trial SC visual response variability and RT’s. We hypothesize that relying on a rate code, necessitating temporal integration, can reduce sensitivity to noise, which can erroneously trigger movements through the SC’s descending brainstem projections.