We describe studies aimed at evaluating the physical factors governing the rate of 3'-end proofreading by the Klenow fragment of E. coli DNA polymerase I. Two nonpolar deoxynucleoside isosteres containing 2,4-difluorotoluene (F) and 4-methylbenzimidazole (Z), which are non-hydrogen-bonding shape mimics of thymine and adenine, respectively, are used to investigate the effects of base pair geometry and stability on the rate of this exonuclease activity. Steady-state kinetics measurements show that complementary T.A base pairs at the end of a primer-template duplex are edited 14-40-fold more slowly than mismatches. By contrast, a 3'-end T residue in a T. Z pair is edited at a rate equivalent to that of natural base mismatches despite the fact that it resembles a T.A pair in structure. Similarly, the A in an A.F pair is edited as rapidly as a mismatched pair despite its close structural mimicry of an A.T pair. Interestingly, when the base pairs are reversed and F or Z is located at the 3'-end, they are edited more slowly, possibly implicating specific interactions between the exonuclease domain and the base of the nucleotide being edited. Finally, thermal denaturation studies are carried out to investigate the relationship between editing and the ease of unwinding of the duplex. The rapid editing of bases opposite F or Z residues at the duplex terminus seems to correlate well with the stability of these base pairs when placed in a context resembling a primer-template duplex. In general, the rate of 3'-end editing appears to be governed by the rate of fraying of the DNA terminal pair, and base pair geometry appears to have little effect.