Male F344 rats, 8 weeks of age, were given 16 intrarectal administrations of N-methyl-N-nitrosourea (NMU) at one of three dose levels over a period of 8 weeks. Five days after the final NMU instillation, rats were placed on one of three diets: chow with gelatin beadlets, chow with beadlets containing 0.024% 13-cis-retinoic acid, or chow and beadlets with 0.006% of the trimethylmethoxy phenyl analog of retinoic acid ethylamide. Groups of 20-40 rats were killed at 22-26 weeks after the first carcinogen treatment. The number of rats with colon carcinoma and the number of tumors per rat were dose related. In addition, "blind" histopathologic evaluation of four predesignate colon locations revealed a dose-related incidence of microscopic preinvasive and invasive colon carcinomas. The feeding of diets containing these two retinoids did not significantly alter the incidence of these parameters of carcinogenesis or the mean histopathologic score at predesignated colon locations for preinvasive or invasive neoplastic lesions. Over 90% of the colon neoplasms induced were invasive tubulopapillary adenocarcinomas. The diameters of the tumors correlated significantly with degrees of invasion of the colons. Only 1 tumor (a signet ring carcinoma) metastasized to the peritoneal cavity. Only 2 of 300 rats treated with NMU had tumors at sites other than the colon.