The mutant selection procedures of the well-characterized Chinese hamster ovary cell/hypoxanthine guanine phosphoribosyl transferase (CHO/HGPRT) mutation assay was modified. Soft agar (0.33%) in medium containing 6-thioguanine was used. The use of soft agar allowed the selection of 10(6) cells per 100-mm-diameter plate without any loss of mutants due to cross-feeding between HGPRT+ (wild-type) and HGPRT- (mutant) cells, as demonstrated by a reconstruction experiment with premixed populations of mutant and wild-type cells. Mutants selected u sing this soft-agar procedure were shown to have a greater than 99% reduction in [3H]hypoxanthine incorporation (as compared to wild type). This modified protocol decreased the incubator space requirement to 1/5 of that required in the original protocol, which allows one to increase the sampling size 5-fold with the same space requirement. The increase in sample size allows for a better quantitation of low mutagenic responses. The modified soft-agar protocol was applied using low doses (0-50 microgram/ml) of ethyl methanesulfonate and resulted in a well-defined dose-response relationship for the induction of mutants.