A remarkable feature of the carcinogenicity of inorganic arsenic (As i ) is the observation that human exposures to As i have been strongly associated with increases in skin, lung, and internal cancers,but As i does not typically cause tumors in standard laboratory animal test protocols. Considerable controversy has centered on whether there is epidemiological evidence of a ª thresholdº for the carcinogenic effects of As i , or at least of a highly nonlinear dose-response. Saturation of metabolism in the dose-range associated with tumors does not appear to be adequate to produce a major impact on the dose-response for carcinogenicity. If there is a strong nonlinearity, it results from the nature of the carcinogenic mechanism(s) of As i . However, no single hypothesis for the mechanism of As i carcinogenicity has widespread support. A biologically realisticcancer risk assessmentfor As i would require a quantitative description of the dose of active arsenic species in target tissues, the interactions between active arsenic and tissue constituents, and the manner in which these interactions result in tumor formation in multiple organs in humans, but not in experimental animals. Although As i has only infrequently been associated with tumors in animal studies, it has repeatedly been shown to act as a comutagen in vitro and as a cocarcinogen in vivo. As i is clastogenic, producing chromatid aberrations, but does not produce point mutations at single gene loci. Of particular interest, As i has been shown to inhibit repair of DNA single-strand breaks, a possible mechanism for its observed comutagenicity and cocarcinogenicity.We propose a cocarcinogenic mode of action in which As i acts primarily on intermediate cells de® cient in cell cycle control at a late stage in a preexisting carcinogenic process. This interaction enhances genomic fragility and accelerates conversion of premalignant lesions to more aggressive, clinically observable tumors. An indirect effect of As i on DNA repair is consistent with the expectation of a nonlinear dose-response rather than the linear dose-response traditionally assumed for mutagenic carcinogens. However, de® ning the exact nature of this tumor dose-response will require further experimental data on the dose-response for the cellular effects of As i . Because As i carcinogenicity is unlikely to be observed in normal experimental animals not exposed to other carcinogens,studies in animals and cell lines de® cient in cell cycle control should also be considered. Experimental studies speci® cally designed to address the key mechanistic and dose-response issues for As i carcinogenicity are critically needed to support public health policy decisions regarding current environmental exposures to As i .