In order to explore structure-activity relationships, 4,4'-diaminoazobenzene and four structurally related azo dyes were tested for their ability to induce gene mutations in Salmonella typhimurium strains TA1535, TA100, TA1537, TA1538, and TA98. Only 4,4'-diaminoazobenzene and 4,4'-N(beta-hydroxyethylamino)azobenzene were found to be active in the two frameshift strains TA1538 and TA98. Further tests were performed in strain TA98, both in the presence and in the absence of Aroclor 1254-induced rat or hamster liver S9 preparations. The amount of S9 used per plate was 50, 100, 150 or 300 microliters, which corresponds to 10, 20, 30 or 60% of S9 in S9 mix. 4,4'-Diaminoazobenzene was found to be mutagenic, and its mutagenicity depended on the percentage of S9 in S9 mix and the type of S9 fraction used. 4,4'-N-(beta-Hydroxyethylamino)azobenzene was less mutagenic than 4,4'-diaminoazobenzene, indicating a reduction in mutagenicity associated with the beta-hydroxyalkyl substituents. The other three azo dyes [4'-methyl-4-N,N-di(beta-hydroxyethylamino) azobenzene; 4'-amino-6-methyl-4-N,N-di(beta-hydroxyethylamino)azobenzene; and 4'-N(beta-hydroxyethyl-amino)4-N,N-di(beta-hydroxyethylamino)azobe nzene] were inactive, both in the presence and in the absence of the metabolic activation system. The use of the preincubation test did not alter the observed positive or negative response of these compounds. The importance of this finding is that the non-mutagenicity or decreased mutagenicity of these four compounds is predictable on the basis of their chemical structures. These azo dyes, like the non-mutagenic members of series of monocyclic aromatic amines, contain large substituents on one or both of the amino groups of the parent compound, in this case 4,4'-diaminoazobenzene. From our earlier data and the experiments discussed in this paper, we conclude that the study of structure--activity relationships can provide useful information for the prediction and interpretation of mutagenic responses.