The "Bacterial Reverse Mutation Assay" is generally accepted to analyse the genotoxic capacity of single compounds or complex mixtures such as cigarette-smoke condensates. With an adapted and modified Ames assay, the mutagenicity of native cigarette mainstream whole smoke (WS) and its gas/vapour phase (GVP) was studied. The bacteria were directly exposed to the smoke in a CULTEX1 system closely connected to a smoking robot (VC10). A variety of standard tester strains (TA98, TA100, TA1535, TA1537, TA1538, TA102, WP2uvrApKM101) and descendants of TA98 (YG1021, YG1024, YG1041) and TA100 (YG1026, YG1029 and YG1042) were exposed to whole and filtered smoke of the research cigarette K2R4F to find the most sensitive strains for analysing the mutagenic activity of these test atmospheres. Mutagenicity of WS was detected by TA98, TA100 and their YG descendant strains as well as by WP2uvrApKM101 in the presence of S9 mix. The GVP induced a mutagenic signal in TA100, YG1029 and YG1042 and WP2uvrApKM101 only in the absence of S9 mix. To detect mutagenicity in WS the presence of the plasmid pKM101 is required and a frame-shift mutation is more effective than a missense mutation. To detect mutagenicity in GVP, the presence of the plasmid pKM101 and a missense mutation are required. The differentiating capacity of this modified Ames assay was demonstrated by exposing strain TA98 to WS and TA100 to the GVP of cigarettes with different tar content. The mutagenic activity of WS and the GVP increased with rising tar content of the cigarettes with two exceptions in WS. Thus, the concept of tar content alone is misleading and does not reflect the mutagenic activity of a cigarette.