In previous exposure-response studies, we have documented non-linearities for some of the early effects in rat liver of diethylnitrosamine (DEN) and a near no-effect levels for initiation of promotable liver neoplasms at the lowest cumulative exposure of 0. 5 mmol/kg body weight; this in spite of formation of DNA adducts and induction of hepatocellular altered foci (HAF). To extend these investigations, in an initiation segment, young male F344 rats were administered four exposures of DEN ranging from a cumulative total of 0.25 mmol, which is half of the previously used low exposure, up to 2 mmol per kg body weight, an effective initiating exposure. These exposures were achieved by once weekly intragastric instillations of one-tenth the total exposures for up to 10 weeks. The initiation segment was followed by a 4 week recovery segment, to allow for remission of acute and subchronic effects of DEN, after which the groups were maintained on 0.06% phenobarbital in the diet for 24 weeks to promote liver tumor development in order to assess initiation. During and after initiation and at the end of recovery, selected groups were studied for several crucial effects involved in hepatocarcinogenicity. The low exposure produced a low-level of DNA ethylation at both 5 and 10 weeks of exposure, measured as O(4)-ethylthymidine, the most persistent promutagenic ethylation product. At the 5 week interval, the adduct values of the higher exposures were less than proportional to the increment of exposure, suggestive of nonlinearity. Assessment of cellular proliferation by staining for proliferating cell nuclear antigen revealed that the lowest exposure did not increase the replicating fraction of hepatocytes during the initiation (10 weeks) or recovery (4 weeks) segments, whereas in the three higher exposure groups, proliferation was increased in relation to dose and time. Preneoplastic HAF expressing glutathione S-transferase-placental-type were present at low multiplicity in control livers and their multiplicity was increased in all exposure groups by the end of exposure, at which time the increase in the high exposure group was disproportionately greater than the increment of exposure. After phenobarbital administration in the promotion segment, all exposure groups exhibited further HAF increases at 39 weeks. At the end of the promotion segment, no hepatocellular neoplasm was found in 80 controls or in 40 rats in the low exposure group. In the mid-low exposure group, which was the previously studied low exposure, only one adenoma was found, yielding a 3% incidence, while in the two higher exposure groups, 32 and 80% of rats exhibited liver neoplasms, which were increased disproportionately greater than the increments of exposure. Thus, the findings document non-linearities of early DEN effects and at the lowest cumulative dose, a no-effect level (NEL) or threshold for initiation of promotable liver neoplasms. These findings provide a conceptual basis for understanding why low-level exposures to DNA-reactive carcinogens may convey no cancer risk.