The wavelength dependence of the carcinogenicity of ultraviolet (UV) radiation needs to be known in order to assess the carcinogenic risks of various UV sources, most notably the different solar UV spectra at ground level under depleting stratospheric ozone. This wavelength dependence cannot be extracted from human data (e.g., from epidemiology); it can, however, be directly obtained from animal experiments. Precise information on the wavelength dependence, the so-called action spectrum, was not available until recently: erythemal or mutagenic action spectra have been used as substitutes. However, experimental data on skin tumors induced in hairless mice (Skh:HR1) with various polychromatic sources have been building up. Our group has found that none of the substitute action spectra yield a statistically acceptable description of our data, and we have, therefore, derived a new action spectrum, dubbed the SCUP action spectrum (SCUP stands for Skin Cancer Utrecht-Philadelphia, because the action spectrum also fits experimental data from the former Skin and Cancer Hospital in Philadelphia). The SCUP action spectrum has a maximum at 293 nm, and in the UVA region above 340 nm the relative carcinogenicity per J/m2 drops to about 10(-4) of this maximum. The effects of an ozone depletion on solar UV doses weighted with these different action spectra are compared: the erythemal and SCUP weighted dose come out as least sensitive with a 1.3% and 1.4% increase, respectively, for every 1% decrease in ozone.