A thin slit-lamp beam illuminating the pupil margin produced clearly visible pupil oscillations. These oscillations were timed with a stopwatch, thus producing a measurement of the "pupil cycle time". The pupil cycle time was remarkably stable in various testing situations and repeatable within +/- 3% over extended periods of time. When the iris muscles were normally innervated and responsive, the pupil cycle time was dependent on the speed of conduction and the number and strength of optic nerve impulses. Pupil cycle time can be measured in most persons with active or inactive optic neuritis. In the few patients whose light reflex is so poor that the pupil cycle time cannot be measured, the inability to induce cycling can itself be taken as a definite abnormality of the light reflex arc. Only 5% of normal persons 12 to 50 years of age are expected to have a pupil cycle time in either eye longer than 954 msec, or a difference in pupil cycle time between the two eyes longer than 70 msec. Pupil cycle time was significantly longer in patients with optic neuritis, with a P-value less than .001. The pupil cycle time is similar to visual evoked response latency time in that it can detect and quantitate subclinical defects in optic nerve conduction time. Pupil cycle time is objective and quantitative for each eye individually. It is a fast, simple, and reliable clinic test of optic nerve function.