Information on the variation in carcinogenicity with wavelength is crucial in risk assessments for skin cancers induced by UV radiation. Until recently the wavelength (lambda) dependencies of other detrimental UV effects, such as sunburn, have been used as substitutes. Direct information on the lambda dependency can only be obtained from animal experiments. To this end we accumulated a large data set on skin tumors induced by chronic UV exposure of albino SKH:HR1 mice (14 different broadband UV sources and about 1100 mice); the data come from the Photobiology Unit of the former Skin and Cancer Hospital in Philadelphia and from the Department of Dermatology of the University of Utrecht. The lambda dependency was extracted from this data set (a statistically satisfactory description with chi 2 = 13.4, df = 7) and represented by the Skin Cancer Utrecht-Philadelphia action spectrum, i.e., a set of factors to weight the exposures at different wavelengths according to their respective effectiveness (inversely proportional to the daily exposure required for a median tumor induction time of 300 days). The fits obtained with other already available action spectra proved to be poor (chi 2 > 60, df = 11). The maximum effectiveness was found at 293 nm, and above 340 nm the effectiveness showed a shoulder at about 10(-4) of the maximum. A sensitivity analysis of the final solution for the lambda dependency showed a large margin of uncertainty above 340 nm and an information gap below 280 nm. The large variation in tumor responses in the present data set can be transformed to a coherent, common dose-response relationship by proper spectral weighting with this single action spectrum.