A series of Escherichia coli mutants, constructed originally by Cupples and Miller [C.G. Cupples, J.H. Miller, A set of lacZ mutations in Escherichia coli that allow rapid detection of each of the six base substitutions, Proc. Natl. Acad. Sci. U.S.A. 86 (1989) 5345–5349], provides a unique system for quantifying base-change mutations, and the repair processes that limit their establishment, in bacteria under selective and non-selective conditions. We focussed on one strain in which a T → G replacement inactivates the lacZ gene. Reversions of this strain can occur through oxidation of G, leading to G → T transversions. We show that spontaneous reversions occurred both in lactose (selective) and glucose (non-selective) medium. The number of revertants per viable cell was much greater in medium containing lactose or both sugars than glucose alone. In glucose medium, the rate of reversion was highest below 0.6% glucose and strongly inhibited at and above that level. Evidence that reversions occurred through G → T transversions in both lactose and glucose media came from two observations: by sequence analysis of a series of revertants and by comparing the reversion rates in strains possessing and lacking the mutM gene (encoding formamidopyrimidine DNA glycosylase, FPG). However, the rate of reversion was stimulated by reducing O 2 to 1% and inhibited or delayed by increasing O 2 to 90%. In mutM − cells grown on glucose medium, the proportion of revertants increased over a 5-day period. In contrast, in mutM + cells, revertants appeared primarily during the first 2–3 days after plating; few new revertants appeared in the following days. These data imply that base excision repair initiated by FPG was less effective in the first 2 days and more effective later in stationary phase.