The repeatability of pupil response measurements made with the Pupilscan instrument has been assessed. Two operators made 32 repeat recordings of the pupil response in each eye of eight normal subjects over four days. A total of 512 recordings were made and the results analysed by variance analysis. Room light levels were carefully controlled. The instrument measured initial pupil diameter, and after a 0.2 s light stimulus measured minimum and final diameters, maximum diameter change, constriction velocity and time to minimum diameter. For all measurements except time to minimum, there were significant differences (P less than 0.001) between the normal subjects. Day-to-day variation was also significant (P less than 0.005), but did not mask the differences between subjects. It is concluded that the average of at least four repeat measurements are made, giving error ranges similar to those observed from day-to-day changes.