To comprehensively assess the reliability of automated Pentacam (Oculus, Inc.) measurements. Flinders Eye Centre, Flinders Medical Centre, Bedford Park, South Australia, Australia. Both eyes of 35 normal volunteers were tested twice on the same day by 2 different observers. All automated values were recorded, and manual analysis of topographic maps was performed only to overrule variance in corneal thickness due to pupil decentration altering the central reference point. Repeatability was determined with Bland-Altman limits of agreement and reported as the coefficient of repeatability (COR = +/-1.96 standard deviation of differences). Relative repeatability (RR) was calculated as a percentage of the ratio of COR to the mean. Overall, repeatability was good. Corneal curvature, reported in diopters, showed good repeatability anteriorly (simulated keratometry mean COR+/-0.28D; RR=0.64%) and posteriorly (COR+/-011D; RR=1.85%). Peripheral corneal curvature was more reliable when calculated by the sagittal (axial) method (RR=1.57%) than by the tangential (meridional) method (RR=2.38%). Keratometric power deviation was less reliable (RR=16.39%). Anterior chamber measurements showed good reliability (RR=3.07%-5.68%) except for anterior chamber angle (RR=14.41%). Pupil diameter showed poor reliability (RR=25.77%). Central corneal thickness was comparable at pupil center and corneal vertex, but peripheral repeatability was much better when centered on the corneal vertex (COR+/-16.00microm; RR=2.56%) than at pupil center (COR+/-26.28microm; RR=4.23%). Pentacam corneal curvature and anterior chamber parameters were highly repeatable, but pupil measurements had poor repeatability. Peripheral pachymetry readings were affected by pupil decentration and required manual analysis using the corneal vertex as the point of reference to achieve good repeatability.