This articie is based upon a paper which was presented at the SRP meeting on the Biological Bases of Radiation Protection Standards, October 1981. It is suggested that experimental radiation carcinogenesis data derived from animal studies will probably never provide numerical evidence of risk that is applicable to man. The uncertainties involved in any extrapolation of risk estimates from mice to men surely outweigh the uncertainties in the human epidemiological data. It is also suggested that at least in the foreseeable future animal data will not solve the perennial problem of the shape at low doses of the dose response curve for radiogenic cancer. At most the data may clarify the debate over linearity-non linearity and over the existence or otherwise of a threshold. However, the paper does suggest a very positive role for animal data in providing semi-quantitative generalizations for radiological protection concerning such variables as dose rate, radiation quality, partial body/organ exposure and in situations where the dose is received in a highly inhomogeneous fashion, e.g. the special problems of internal emitters.