The fidelity of DNA polymerase-alpha from calf thymus (9S enzyme) in copying bacteriophage phi174am16 DNA in vitro has been determined from the frequency of production of different revertants. In the self-priming reaction we were able to measure the frequencies of base pairing mismatches during the course of replication on biasing the ratios of deoxynucleoside triphosphates. The frequency of dGTP:T, dGTP:G and dATP:G mismatches were 7.6 x 10(-5), 4.4 x 10(-5) and 2.8 x 10(-5), respectively, at equal concentrations of the deoxynucleoside triphosphates. dCTP:A, dGTP:A, dCTP:T and dTTP:T mismatches were below the limit of detection (<5 x 10(-6)). A synthetic dodecamer primer with a 3' end covering the first two bases of the amber codon was used to determine the misinsertion frequency of the first nucleotide incorporated. This gave a misinsertion frequency of 1.5 x 10(-4) for the dGTP:T mismatch, which is slightly higher than that observed from the pool bias studies. Further, it showed no sensitivity to biasing the nucleotide pool, suggesting a different mechanism for the incorporation of the first nucleotide. These data do not support 'energy-relay'-like models for achieving high accuracy in eukaryotes. The observed misinsertion frequencies were corrected for mismatch repair of the heteroduplexes during the transfection experiments by parallel experiments using a mismatched primer. This was synthesized to have the same G:T mismatch as produced in the preceding experiment.