The dose-response relationships from the data described in Paper I were analysed. Among unpromoted animals, only doses sufficient to cause ulceration with subsequent promotion due to wound healing caused a rapid crop of tumours, so the dose-response curve exhibited strong upward curvature. Among promoted animals, the response of the skin to initiation appeared to have been nearly saturated by all DMBA doses tested, so that a 30-fold decrease in dose produced only a 3-fold decrease in effect. The dose-response relationship thus exhibited strong downward curvature. Among promoted animals, estimation of the risks associated with very low doses of carcinogen by linear extrapolation through the origin from the effects of larger doses (which is often assumed to be conservative) would under-estimate the true risks by 10-fold or more. Our results emphasize that whereas linear interpolation from the results of high doses may be reasonable for data on the effects of continuous treatment with non-toxic dose levels of carcinogen, it may be misleading when extrapolating, as here, from the effects of single large doses.