A gel fidelity assay, previously used in the analysis of DNA polymerases having no associated 3' to 5' exonuclease activity, has been generalized for use with polymerases that contain exonucleolytic proofreading. The main purpose of this study was the development of a general analysis, using a standard Markov model, to convert experimentally observed DNA primer gel bands arising from insertion and proofreading of right and wrong deoxyribonucleotides, into nucleotide incorporation velocities and, most importantly, fidelities. The model has been applied primarily to an analysis of polymerase kinetics and fidelity in the presence of a next correct rescue dNTP, but the model can be conveniently modified to investigate other experimental designs. In the presence of rescue dNTP, direct competition occurs between excision or extension of a mismatch. At concentrations of rescue dNTP sufficient to suppress the gel band intensity at the mismatch target site, nucleotide incorporation and misincorporation rates can be obtained from the ratios of gel band intensities 3' (downstream) and 5' (upstream) to the target site, measured as a function dNTP concentration for "wrong" and "right" dNTP substrates. The polymerase misincorporation efficiency, in the presence of proofreading, is given by the ratio of wrong to right incorporation efficiencies, Vmax/Km, obtained from the gel band ratios. The bacteriophage T4 polymerase with a highly active 3'-exonuclease activity was used to illustrate the assay. Nucleotide misincorporation efficiencies measured at several template sites were dCMP.A approximately equal to 10(-6), dGMP.A approximately equal to 10(-5), dTMP.T approximately equal to 2 x 10(-4), and dAMP.A < 10(-7). Proofreading of the dGMP.A mispair was suppressed by about 3-fold in the presence of high concentrations of next correct "rescue" dNTP causing a concomitant reduction in the fidelity of dGMP.A to about 3 x 10(-5).