Most studies of correlation between logs of mutagenic potency (MP) and cancer potency (CP) have obtained relatively small but statistically significant estimates of correlation (r) and corresponding log-log slope (b, in Log[CP] = a + b Log[MP]). But for mutagenic carcinogens, multistage cancer theory predicts that b and r should be highest when MP values best estimate mutation yields per unit dose at concentrations at least as high as those observed to cause cancer in bioassay animals. To test this hypothesis, the correlation of Ames test and rodent cancer potencies was examined for a total number n of 134 chemicals reported as positive in both assays. Values of maximum significant cancer potency (CP, in [mmol/kg-day]-1) were obtained from a published carcinogenic potency database. Values of maximum mutagenic potency (MP, as revertants per mmol/L-plate) were estimated from 2,347 sets of Ames test data reported by the NTP mutagenicity testing program, supplemented by similar data newly obtained for ten heterocyclic amines. For compounds with one or more significantly positive MP estimates based on approximately linear Ames test dose-response data, linear regression of maximum values of Log(CP) on Log(MP) yielded b = 0.27 +/- 0.065 and r = 0.39 (P = 0.0001, n = 105), similar to previously reported results for relatively large n. As predicted, when MP values were additionally restricted to include only values estimated from Ames test data approximately linear at corresponding lowest-TD50 concentrations, similar regressions yielded significantly improved fits (e.g,. b and r approximately 0.6, P < 10(-7), n = 68). Implications of these findings are discussed concerning the quantitative role that mutations like those measured in the Ames test may have in explaining observed cancer-bioassay results.