We studied the accuracy of nine corneal reflection pupillometers made by three manufacturers, because, at the time of investigation, we could not find any information about their accuracy. For this purpose two steel balls separated by spacer bars, were positioned to simulate nominal inter-pupillary distance (PD) measurements of 55, 60, 65 and 70 mm. Ten measurements of each 'PD' for distance vision, were made by two experienced observers, with each instrument. We found that interobserver differences in measurement were clinically insignificant most of the time, although some data sets were significantly different statistically. The most consistent and accurate measurements of PD were achieved with PD 65 mm. The average error in mean value of the measurements for all PD's, was 2.3 mm, and the average of the standard deviation in our measurements, was 0.26 mm. Significant sources of error were poor eye relief in all instruments evaluated, as well as friction in the scale adjustment mechanism in some instruments. Our findings show that while a given pupillometer may be relied upon to give consistent readings (within the limits of clinically acceptable error), its accuracy should be checked periodically, especially for very small and very large PD measurements.