Assuming that the initiation of cancer growth is due to a purely accidental fluctuation of some physicochemical condition, it is possible to derive the equation for the cancer incidence as a function of age. The final result depends on whether it is assumed that a single accidental fluctuation in a cell is sufficient to produce cancer, or that a finite number,k, of repetitions of the accidental fluctuation must occur in the same cell. In principle it is possible to determine from observed incidence curves the numberk. Actually, however, this cannot be done at present, because the difference of the theoretical curves for differentk's is rather small, and the available cancer statistics are not accurate enough to show such differences.