The analysis of 68 published sets of dose-incidence data for marrow failure in different species, using a double-log mortality function, indicates: (a) There is more heterogeneity, i.e. greater sums-of-squares per degree of freedom, within the data sets for mouse than for larger species (monkey, dog, sheep, goat, pig). (b) For mice the curves for acute doses are characterized by a D0 of about 100 cGy for tissue-rescuing units (or target cells), which are depleted at most to about 3 x 10(-4) at LD50. (c) Larger species are much less tolerant to target-cell depletion, the corresponding level being consistently in the range of 10(-2)-10(-3) at LD50. Also, the D0 is often lower (approximately 55 cGy), which is compatible in the dog with such a value for hemopoietic progenitor cells. (d) With larger species there is an unexpected reduction in heterogeneity when the dose rate is lower, which gives a D0 lower than expected and a higher extrapolate. It is concluded that the position and slope of the dose-incidence curves are compatible with interpretations based primarily on target-cell number and survival characteristics, modified by additional heterogeneity factors.