The effects of local DNA sequence on the proofreading efficiency of wild-type T4 DNA polymerase were examined by measuring the kinetics of removal of the fluorescent nucleotide analog 2-aminopurine deoxynucleoside monophosphate (dAPMP) from primer/templates of defined sequences. The effects of (1) interactions with the 5'-neighboring bases, (2) base pair stability, and (3) G.C content of the surrounding sequences on the pre-steady-state kinetics of dAPMP excision were measured. Rates of excision dAPMP from a primer 3'-terminus located opposite a template T (AP.T base pair) increased, over a 3-fold range, with the 5'-neighbor to AP in the order C < G < T < A. Rates of removal of dAPMP from AP.X base pairs located in the same surrounding sequence increased as AP.T < AP.A < AP.C < AP.G, which correlates with the decrease in the stabilities of these base pairs predicted by Tm measurements. A key finding was that AP was excised at a slower rate when mispaired opposite C located next to four G.C base pairs than when correctly paired opposite T next to four A.T base pairs, suggesting that exonuclease mismatch removal specificities may be enhanced to a much greater extent by instabilities of local primer termini than by specific recognition of incorrect base pairs. In polymerase-initiated reactions, biphasic reaction kinetics were observed for the excision of AP within most but not all sequence contexts. Rates of the rapid phases (30-40 s-1) were relatively insensitive to sequence context. Rapid-phase rates reflect the rate constants for exonucleolytic excision of dAPMP from melted primer termini for both correct and incorrect base pairs and were roughly comparable to rates of removal of dAPMP from single-stranded DNA (65-80 s-1). Rates of the slow phases (3-13 s-1) were dependent on sequence context; the slow phase may reflect the rate of switching from the polymerase to the exonuclease active site, or perhaps the conversion of a primer/template terminus from an annealed to a melted state in the exonuclease active site. These data, using wild-type T4 DNA polymerase and two exonuclease-deficient T4 polymerases, support a model in which exonuclease excision occurs on melted primer 3'-termini for both mismatched and correctly matched primer termini, and where specificity favoring removal of terminally mismatched base pairs is determined by the much larger fraction of melted-out primer 3'-termini for mispairs compared to that for correct pairs.