In an attempt to provide a common denominator for cognitive deficits observed in cerebellar patients, it has been suggested that they might be secondary to impaired control of attention, a 'dysmetria of attention', conceptually analogous to motor dysmetria. Albeit appealing and quite influential, the concept of attentional dysmetria as a consequence of cerebellar disease remains controversial. In an attempt to test this concept in a direct way, we compared the performance of patients with cerebellar disorders to that of normal controls on tasks requiring either overt or covert shifts of spatial attention. In the first experiment, visually guided saccades, i.e. overt shifts of spatial attention, were elicited. In the second experiment, covert shifts of attention were evoked by the need to discriminate the orientation of a Landolt C observed during controlled fixation and presented in the same locations as the saccade targets in the previous experiment. The allocation of attention was assessed by comparing acuity thresholds determined with and without spatial cueing. The patients exhibited dysmetric saccades as reflected by larger absolute position errors or a higher number of corrective saccades compared to controls. In contrast, the ability to shift attention covertly was unimpaired in the patients, as indicated by a robust improvement in visual acuity induced by spatial cueing which did not differ from the one observed in the controls and which was independent of the range of SOAs (stimulus onset asynchronies) tested. Finally, the individual amount of saccadic dysmetria did not correlate with the individual performance in the covert attentional paradigm. In summary, we conclude that the contributions of the cerebellum to attention are confined to overt manifestations based on goal-directed eye movements.