Preoperative ultrasonic biometry is indispensible in cataract surgery in order to calculate the power of the intraocular lens to be implanted. Various authors have suggested that the accuracy of the measurements could be improved by performing biometry several times and basing the calculation of the lens on the mean value. The problem of the retest reliability of ultrasonic measurements was investigated in 99 young subjects with healthy eyes and 60 cataract patients. Biometry was performed three times consecutively in each mode using three different scanners and different coupling techniques. A comparison of the consecutive measurements showed good agreement (between r = 0.96 and r = 0.99) in regard to axial length and hence very good retest reliability. The accuracy of the repeat measurements was approximately the same with all three coupling methods used. The correlation coefficients of retest measurements of anterior chamber depth and lens thickness were much lower (r = 0.78 to r = 0.96). This may be explained by differences in handling of the probes, in the test modes, and in the physical conditions on which the measurements are based. In view of the good reproducibility of axial length measurements the authors consider one carefully performed biometry more valuable for preoperative calculation of lens power than the mean of several measurements.