The apparent DNA site size obtained from an assay monitoring the ATPase activity of Escherichia coli recA protein (n = 3.5) differs from that determined from a direct DNA binding assay (n = 7) done under identical conditions. Investigation of this discrepancy indicates that at a DNA:protein ratio of 3.5:1, one-half of the recA protein population is less sensitive to ATPase activity inhibition by the nonhydrolyzable ATP analogue adenosine 5 -O-(3-thiotriphosphate) (ATP gamma S), suggesting that the recA protein filament is asymmetric with respect to NTP affinity. This asymmetry does not depend on the presence of ATP gamma S since the apparent Km for ATP derived from single-stranded DNA-dependent ATP hydrolysis activity is dependent on the DNA:protein ratio. Three models are proposed to account for the observed site size discrepancy and the NTP binding affinity asymmetry. They differ mainly in the intrinsic site size for each recA protein monomer and in the number of DNA-binding sites/recA molecule. Gel filtration of recA-single-stranded DNA complexes at different DNA:protein ratios complements the enzymological data and provides an additional method of distinguishing among the proposed models. The phenomenon of subunit nonequivalence within the recA protein presynaptic filament may provide a molecular basis for understanding how recA protein can discriminate between different DNA molecules during homologous pairing.