We show from simple theory that the number of mutants observed in the Ames assay should be approximately proportional to alpha Cn0P where alpha is the mutation rate per concentration of mutagen, C; n0 is the initial inoculum; and P is the average number of bacteria per colony in the background lawn. We tested this theory by carrying out Ames assays for the directly acting mutagen 2-nitrofulorene on TA98 as a function of initial inoculum over the range 10(4)-10(8) bacteria/plate. P was assumed to be proportional to the average volume of the background colonies and was estimated from 100X photomicrographs of the background lawn. The initial inoculum was determined by counting background colonies in the photomicrography, by dilution-plating, and by electronic particle counting. We found that the number of mutants depended weakly on n0, but the dependence could be quantitatively accounted for by the simple theory. These data and the theory explain the least some of the reported variations in quantitation of the Ames assay; show that the slopes of dose response curves depend on n0; explain the range observed for spontaneous revertants; suggest some limitations to quantifying the Ames assay and, within these limits, suggest a method for normalizing independently obtained Ames assay data.