A method of analysis of a certain class of experiments in carcinogenesis.

Over the last ten to fifteen years a number of models have been suggested to describe some of the processes underlying the phenomena of carcinogenesis (see Armitage and Doll [1960], for a review and references). Most of these models assume that cancer begins in a single cell but differ in the assumptions made about the processes affecting the individual cell (and its lineal descendants). The models also suppose that the individual cells in a tissue' behave independently and that cancer of the tissue (the diagnosed condition) occurs (or starts) when the first cell in the tissue becomes cancerous.2 The purpose of this paper is twofold: (i) to point out that these assumptions alone strongly suggest two distributional forms that the random variable of time to occurrence (diagnosis) of carcinoma in an individual tissue may take, and (ii) to develop the basic statistical methodology required to apply to experimental data one of these distributional forms which is particularly plausible.