If measures of rectal epithelial kinetics are to be used on a large scale, or as part of multicenter studies, the variance of repeated measures must be sufficiently small to detect the effect of clinical status and dietary, or chemoprevention interventions. Errors in measurement can be categorized as intra-reader (sigma 2I), inter-biopsy (sigma 2B), inter-laboratory (sigma 2L), and inter-reader (sigma 2R); the sum of these (sigma 2T) when compared to inter-subject variance (sigma 2S) is a measure of the precision with which cell kinetics can be estimated for a given subject. We estimated the variances of the component errors in four separate experiments, each involving 6 to 10 subjects, using the proliferating cell nuclear antigen (PCNA) technique. PCNA labeling index variances for a single laboratory study were sigma 2I = 0.02-0.07 and sigma 2B = 0.08-0.15. Inter-laboratory and inter-reader variances were sigma 2L = 0.08 and sigma 2R = 0.03. Summing the errors applicable within a laboratory gives a rounded estimate of 0.15. For inter-laboratory studies the total error (sigma 2T) is greater, estimated in our two-laboratory study as 0.3. In a separate study, the variation between biopsies (sigma 2B) could be accounted for by the variation between crypts within biopsies; an acceptable level of accuracy is achieved after counting 20-30 crypts. In contrast to the measurement errors, the variance between subjects, which includes a treatment effect of diet, was larger, estimated to be between 0.75 and 1.30.(ABSTRACT TRUNCATED AT 250 WORDS)