Computerized methods of examining the optic nerve head offer the possibility of providing consistent, reliable, and accurate measurements of the nerve head. The reliability and accuracy of such instruments must be examined to confirm this. This research was undertaken to gain an insight into the accuracy of such a system, which takes its measurements by tracking the luminance boundaries of the pallor area and the optic nerve head after an observer marks the starting points. Measurements made using the computer-tracked boundary were compared with those made using a boundary tracked by an image-processing system of greater sophistication, the human visual system. The comparison was made with images from one eye from each of 30 subjects (10 normal, 10 ocular hypertensive, and 10 glaucomatous). The measurements were made by two observers. Each observer made each set of measurements twice, and the results were analyzed with a five-factor analysis of variance. The measurements made using the computerized method did not differ significantly from those made using the manual boundary tracking for the two observers, and they were highly correlated (r = 0.99 for the area pallor-disc ratio). Significant differences (P < 0.05) were found between the two observers in both the manual and computerized boundary tracking. The computerized method, however, did not find differences between the two sets of measurements made by each observer; the manual method did show such differences. The computerized method appeared to trace the luminance boundaries successfully in the image, and it might reduce errors related to the observer marking the starting points on different occasions.(ABSTRACT TRUNCATED AT 250 WORDS)