The Roche European-American Anticataract Trial (REACT) will assess the effect of antioxidants on progression of cataract in humans. This report evaluates the methods used in REACT. Seventy three subjects (139 eyes) with cortical (C), posterior subcapsular (P), nuclear (N) or mixed cataract were seen twice within two weeks for eye examinations, assessments of visual function, lens photographs and CCD images. The degree of cataract and nuclear color (NC) were assessed with subjective (LOCS III) and objective (computerized, CASE 2000 CCD) methods. Repeat visit values were used to calculate intraclass correlation coefficients (r1) and 95% tolerance limits (TL). A clinically significant change (CSC) was defined as one step in LOCS III. The relative power of each method to detect cataract change and sample sizes needed to achieve statistically significant results were calculated. The r1 values for visual function tests ranged from 0.76 to 0.88; if these tests of visual function were used to detect a clinically significant change in cataract severity, sample sizes of 840 to 2707 per group would be needed. The r1 values for LOCS III were 0.88 to 0.97, and sample sizes ranged from 50 to 135 per group. The r1 values for the CCD were 0.93 to 0.98, and sample sizes ranged from 1 to 42 with poorer values relating to measurement of P. We conclude that the methods used in REACT are reproducible. The analytical algorithms in the image analysis programs did not permit differentiation between C and P opacification; therefore, P cataract is best measured with LOCS III. REACT sample sizes are adequate to detect a difference of 0.2 LOCS III units/year between the mean rates of cataract progression in two groups.