Azide mutagenesis was investigated in Salmonella typhimurium and Escherichia coli. Azide was highly effective in inducing his+ revertants in excision-repair deficient (uvrB) derivatives of S. typhimurium hisG46 and in inducing high frequencies of 5-fluorouracil resistant mutants in excision-repair deficient (uvrA) derivatives of E. coli B/r WP2. In excision-repair plus strains, azide was only a marginal or ineffective mutagen, demonstrating that the bacterial excision-repair system could repair nearly all azide-induced damage. This observation suggests that the initial azide-induced lesion causes a major DNA helix distortion recognizable by the excision-repair endonucleases. The presence of recombination deficient (recB or recC) genes in combination with uvrA increased E. coli sensitivity to azide killing, but depressed azide mutagenicity. These results are similar to those reported for UV-induced mutagenesis with the E. coli strains and suggest that post-replication repair might be the error-prone step in the repair process. Azide mutagenesis specificity is, however, unique and different from UV, as demonstrated by inability of azide to revert the ochre try locus in E. coli WP2s. These results show that the initial azide-induced DNA damage is highly specific but different from UV-induced DNA damage. Metabolic inhibitors, similar in action to azide, did not induce mutations in S. typhimurium strain TA1530, a strain highly susceptible to azide mutagenesis, thus ruling out the possibility that azide mutagenesis was due to peroxide accumulation. A mechanism based on in vivo activation of azide to the actual mutagen is proposed.