Twelve patients with cerebellar dysfunction including a limb ataxia and 12 age-matched controls performed pointing movements with an arm. In one condition, the task was a simple reaction time (RT) movement directed toward a spatially defined target. The other two conditions involved choice tasks in which the amplitude and direction of movement were varied. The variables recorded were: movement latency determined by measuring the RT, duration of movement and the terminal accuracy of pointing as reflected in the movement time (MT), pointing surfaces (PS), and systematic errors. RTs and MTs were found to be significantly longer in cerebellar patients than in controls. In both groups the choice RT increased significantly as compared to simple RT, but no significant difference between patients and controls was found for the mean increase of the choice RT as compared with the mean increase for simple RT. A strong correlation between MTs and PSs was found in the controls. The cerebellar patients showed no correlation between MTs and PSs. The results are discussed in relation to the ability in cerebellar patients to program and execute voluntary movements.