Directed mutagenesis has been used to construct five variants of human dihydrofolate reductase in which smaller residues are substituted for phenylalanine 34, a residue participating in the binding of substrate and methotrexate by interaction with their pteridine rings. The variant enzymes are stable and have decreased affinities for methotrexate (by factors of 2700-60000 at pH 7.65) due to a decreased rate of methotrexate association and a much larger increase in the rate constant for dissociation. However, the catalytic efficiencies of the variants are also lowered by factors of 160-5000, so that it is doubtful whether these enzymes are capable of conferring methotrexate resistance on the cells harboring them. High concentrations of dihydrofolate cause marked inhibition of all the variants, which complicates the determination of kinetic parameters. By the use of stopped-flow spectrophotometry and fluorimetry and other methods, it has been shown that, like the wild-type enzyme, the variants have a branched reaction pathway, but in contrast to the wild-type enzyme, the distribution of flux between alternate pathways is dependent on the concentration of dihydrofolate. This different branch point is a consequence of the very rapid dissociation of tetrahydrofolate from the ternary product complexes of the variant enzymes. Inhibition by dihydrofolate is due to its combination with the enzyme-NADP complex and the slow dissociation of NADP from the resulting abortive complex. When steady state kinetics for this model are simulated using the experimentally determined rate and dissociation constants for the alanine 34 variant, most steady state experimental results are closely approximated.