Countermanding saccades in humans.

We used a countermanding paradigm to investigate the relationship between conflicting cues for controlling human saccades. Subjects made a saccade to a target appearing suddenly in the periphery; but on some trials, after a delay, a stop-signal was presented that instructed subjects to inhibit the saccade. As we increased this delay, subjects increasingly failed to inhibit the movement. From measurements of this relationship, and of saccadic latency in control trials, we estimated the average time needed to inhibit the saccade (the stop-signal reaction time or SSRT). SSRTs were similar across subjects, between 125 and 145 ms, and did not vary with target luminance. We then investigated a race model in which the target initiates a response preparation signal rising linearly with a rate varying randomly from trial to trial, and racing against a similarly rising signal initiated by the cue to inhibit the saccade. The first process to cross a trigger threshold determines whether the saccade is initiated or not. In Monte Carlo simulations, this model correctly predicted the probability of successful saccade inhibition as a function of the stop-signal delay, and also the statistical distributions of saccadic latency during trials in which a stop-signal was presented but the subject failed to inhibit the saccade. These findings provide a comparison to results previously described in the monkey, and show that a simple race model with a linear rise to threshold may underlie behavioural performance in tasks of this kind.