Skin has been widely used in radiation carcinogenesis studies because of the accessibility and visibility of its tumours. Both rat and mouse models have proved to be sensitive, reproducible systems to study the dose and time response of cancer induction following different modes and qualities of radiation exposure. This paper discusses the variation in the shape of the low-LET dose responses from purely linear with no threshold to the highly quadratic curves with significant thresholds, although a linear response is more consistently reported following high-LET radiations. Some dose-response curves show no tendency to turnover at high doses, others show a declining incidence of skin cancer at the highest doses. Protraction or fractionation of the dose reduces the carcinogenic effect in rat skin, whilst the reported dose rate studies in mice are equivocal regarding any sparing effect. Mouse skin cancer studies, in particular, have empirically refuted the 'hot particle hypothesis'. The extensive studies of Albert and Burns highlight hair follicle damage at 300 microns depth as critical in the development of the majority of rat skin tumours. In contrast, mouse studies report a wide variety of cell types as the putative 'cells at risk' in the skin from the spectrum of epidermal and dermal tumours which are induced, and which have been found to be amenable to classification using human pathological categories. Despite these interspecies differences, it is shown that all of the experimental data for radiogenic skin cancer, when expressed per unit area of skin, fall on a relatively narrow and well defined response curve, which is approximately two orders of magnitude more sensitive than the human skin cancer dose response.