By combining high fidelity in vitro DNA amplification and mutant DNA sequence separation by denaturing gradient gel electrophoresis, we are able to directly observe mutational hotspots in human genomic DNA. Our technological development has progressed through the stage of identifying mutant sequences in independently derived, 6-thioguanine-resistant human B cells. We are now analyzing uncloned, complex populations derived from several thousand 6-thioguanine-resistant cells and report preliminary data concerning the mutational spectra of benzo[a]pyrene diol epoxide and ultraviolet light in exon 3 of the hypoxanthine-guanine phosphoribosyltransferase gene. In addition, the approach appears to be general for any gene sequence for which a means to select mutants exists. The more global need to eliminate phenotypic selection is, however, our primary impetus. Our analysis leads us to conclude that no known in vitro DNA polymerase has sufficient fidelity to permit direct observation of unselected mutants. Therefore, an additional change in technology will be necessary to observe nonselected mutant DNA sequences at the low frequencies found in human tissues.