Initial saccades to visual targets are generally not precise in either normal subjects or patients with saccadic dysmetria. Quantitative criteria have to be applied to clearly distinguish between normal and pathological saccadic dysmetria, which is often found in patients with cerebellar lesions. To establish these criteria, the accuracy of visually guided horizontal saccades (10 degrees and 20 degrees target steps) was studied in a group of 24 patients with lesions affecting the cerebellum or its connections, and compared with data from 17 normal subjects. Eye movements were recorded with infrared oculography. Saccades of normal subjects had an average gain of 0.92-0.95 depending on the stimulus condition. Centripetal saccades were significantly larger than centrifugal saccades, for 20 degrees target steps. Most patients (n = 15) had significantly larger saccadic amplitudes than normal subjects (hypermetria), at least in one direction. Saccades in the opposite direction could be either hypometric, hypermetric or normal. Two patients had hypometric saccades in both directions. For one of the patients with hypermetria, in addition, the amplitude difference between centrifugal and centripetal saccades was significantly larger than in the normal subjects. Five patients had no significant pathology of the initial (primary) saccade, but a pathological pattern of corrective saccades. Two patients had normal saccades under all conditions. The quantitative comparison with normal subjects allows the detection even of mild pathology. According to our results, a pathology can be assumed when the average gain of saccades in at least one direction is 1.0 or more, or when more than two out of 20 saccades are followed by two corrective saccades of which the last is in the direction opposite to the initial saccade (pathological pattern of corrective saccades). Target steps of 20 degrees reveal a pathological condition more often than 10 degrees target steps. The application of quantitative criteria might be useful to establish a diagnosis of pathologic saccadic dysmetria even in instances in which it is clinically not obvious.