Saccade variability in healthy subjects and cerebellar patients.

In a previous study we developed a model for the inter-trial variance of saccade trajectories in the rhesus macaque. The analysis of that model showed that signal-dependent noise results in different effector variabilities depending on whether the noise is propagated feedforward through the system (accumulating noise) or whether the noise originates from inside of a premotor feedback loop (feedback noise). This allowed the gain of the premotor feedback loop to be estimated directly from behavioral data. In the present study, we applied the model in healthy human subjects and in patients with chronic isolated cerebellar lesions due to ischemic stroke. Humans showed smaller noise coefficients of variation for both accumulating noise and feedback noise and smaller feedback gain than the monkeys. Despite these differences in the model parameters, the qualitative differences between the two noise types were similar in both species. Cerebellar patients showed larger inter-trial variance of saccade amplitude compared to controls, but saccade metrics and dynamics were well compensated. The parameters of the noise model did not differ significantly between groups. The variance of the saccade amplitude correlated highly (r=0.95) with the coefficient of variation of accumulating noise but not with the other model parameters. The results suggest that the cerebellum plays a role not only in premotor feedback but also in feedforward saccade control and that the latter is responsible for increased endpoint variance in cerebellar patients.