We evaluated repeatability and reproducibility of the Allergan Humphrey model 820 ultrasonic biometer. Sixty eyes (30 subjects) free from ophthalmological abnormality were examined on two separate occasions by two experimenters, one experienced and one inexperienced. The 95% confidence limits for repeatability lie between +0.12/-0.13 mm and reproducibility between +0.15/-0.14 mm for AC depth, between +0.11/-0.13 mm and +0.12/-0.16 mm for lens thickness, between +0.17/-0.19 mm and +0.19/-0.18 mm for right axial length and between +0.26/-0.40 mm and +0.22/-0.25 mm for left axial length. The 40% gain used to measure left axial lengths produced poor retinal echospikes, which may account for reduced accuracy compared to the right eyes using 60% gain. Ten subjects (20 eyes) returned for 3 further visits and the variability over 5 sessions was examined using a two-factor ANOVA with repeated measures on one factor. There were no significant experimenter or session effects and no interaction effect. This biometer gives clinically repeatable and reproducible results and is convenient to use.