Cell proliferation has long been recognized as a basic component of multistage carcinogenesis. Based largely on the finding that certain nongenotoxic chemical carcinogens induce cell proliferation in the same organ that develops tumors after long-term exposure, some suggest that the increased rates of cell division account for the carcinogenicity of these chemicals. This paper examines relationships between chemically induced liver toxicity, cell proliferation, and liver carcinogenesis; major factors include consistency, transient vs. sustained dose-response correspondence, and scientific plausibility. For a presumed mechanism to be valid, a sustained proliferative response is critical, largely because transient increases in hepatocyte proliferation are not sufficient to induce cancer or promote liver tumor development. A consistent association between liver toxicity and carcinogenicity has not been established. Our evaluation of studies on purported relationships between chemically induced cell proliferation and liver carcinogenesis shows: 1) that inconsistencies in sex and species specificity exist, 2) that a large percentage of proliferative responses are transient, 3) that inconsistencies in response to various hepatic peroxisome proliferators are common, and 4) that dose-response and duration relationships have not been sufficiently examined. Studies of proliferative responses of putative preneoplastic cells in the liver indicate that these cells divide faster than normal hepatocytes and also have higher death rates. Chemicals that induce cell proliferation in preneoplastic foci do not always provide a persistent increase in replication rates, even with continuous exposure. A selective growth advantage to preneoplastic cells in the liver may be provided either by an enhancement of the replication rates of these cells compared to the surrounding normal hepatocytes, by inhibition of cell loss, or by inhibition of the growth rate of normal cells. More work is needed to understand how chemical carcinogens and noncarcinogens affect cell division and cell loss of normal hepatocytes and of preneoplastic cells; measurements of hepatocyte proliferation alone are not sufficient to elucidate mechanisms of liver tumor development or to predict liver carcinogenesis. Because of our limited knowledge of the complex molecular changes occurring during liver cancer, it would be inappropriate and far too premature to amend scientific risk assessment procedures for nongenotoxic chemical carcinogens based on oversimplified or incompletely tested speculations.