Mutations of Escherichia coli from sensitivity to nalidixic acid resistance were studied by fluctuation analysis. The mutant distributions in replicate cultures were not significantly affected either by the age of the carbon-starved preculture used for inocula or by the inoculum size. The data from 23 fluctuation tests (48 cultures each) were pooled. The mean number of mutations per culture was estimated to be 0.71 from the fraction of cultures without mutants or 0.74 and 0.77 by maximum-likelihood estimation based on the two models under consideration. When the pooled data were compared with the theoretical expectations, the fits were unsatisfactory (P < 0.005). The lack of fit was caused mainly by too high a frequency of cultures with between 17 and 32 mutants and too high a frequency of cultures with more than 128 mutants. Possible reasons for the lack of fit and its implications with respect to estimation of mutation rates from fluctuation tests are discussed.