Two exposure protocols were used to establish complete dose-response relationships for the hepatic carcinogenicity and DNA adduction in vivo of aflatoxin B1 (AFB1) and aflatoxicol (AFL) in rainbow trout. By passive egg exposure, AFL was taken up less well than AFB1, but was more efficiently sequestered into the embryo itself, to produce an embryonic DNA binding curve that was linear with carcinogen dose and with a DNA binding index three-fold greater than AFB1. Both aflatoxins produced the same phenotypic response, predominantly mixed hepatocellular/cholangiocellular carcinoma. Tumor responses as logit [incidence] vs. In [dose] were parallel-offset, non-linear responses showing a three-fold greater carcinogenic potency for AFL at all doses examined (i.e. 3 times more AFB1 than AFL required to produce an equivalent liver tumor incidence). By molecular dosimetry analysis (logit [incidence] vs. In [DNA adducts]), the two data sets were coincident, indicating that, per DNA adduct formed in vivo in total embryonic DNA, these two aflatoxins were equally efficient in tumor initiation. By dietary fry exposure, both carcinogens produced linear DNA binding dose responses in liver, but with an AFL target organ DNA binding index only 1.14 times that of AFB1 by this exposure route. The tumor dose-response curves also did not exhibit the three-fold difference shown by embryo exposure, but were closely positioned non-linear curves. Since the DNA binding indices differed by only 14%, the resulting molecular dosimetry curves for AFL and AFB1 by dietary exposure were similar to the tumor response curves. These results indicate that differing exposure routes produced differing relative carcinogenicity estimates based on doses applied, as a result of protocol-dependent differences in AFL and AFB1 pharmacokinetic behaviors, but that potency comparisons based on molecular dose received were similar for the two protocols. By comparison with standard DNA adducts produced in vitro using the dimethyloxirane-produced 8,9-epoxides of AFB1 and AFL, we conclude that > 99% of AFL-DNA adducts produced in vivo were identical to those produced by AFB1. Thus similar molecular dosimetry responses should be expected under all exposure protocols in which the two parent carcinogens do not exhibit differing toxicities to the target organ.