This report compared two statistical methods of estimating tumor latency, the Weibull distribution model and the Kaplan-Meier method. Parallelism of dose-response curves of different materials and quantitative reproducibility of dermal carcinogenesis data were also examined. The Weibull method has the advantage of producing parameter estimates, even when tumor yield is low. The Kaplan-Meier method, on the other hand, is free of distribution assumptions. Overall, since the comparisons of potency are made on the basis of parameters from the same assumed distribution on the same strain of animal, the Weibull estimates are favored. A comparison of dose-response data for benzo[a]pyrene and catalytically cracked clarified oil indicated that the slopes of the two dose-response curves were significantly different. Thus the relative carcinogenic potencies of different materials vary with dose, and potency comparisons must necessarily be dose-specific. The quantitative reproducibility of dermal carcinogenesis bioassays was also assessed. The dose-response curves from the three studies of one material had significantly different slopes. Thus the results suggested that there were sources of biological variability which could contribute to experimental error.