The specificity and efficiency of base-mispair editing by the 3'----5' exonuclease activity of phage T4 DNA polymerase has been measured using a sensitive infectivity assay. A series of oligodeoxynucleotide primer chains was synthesized chemically. These primers, when hybridized to phi X174 single-stranded DNAs containing an amber codon, result in a mispaired nucleotide at the 3'-hydroxyl end of the primer chain within the amber codon. DNA synthesis on these primer X templates without the removal of the mispaired terminal nucleotide results in the formation of heteroduplex molecules that yield viable revertants upon transfection into an amber nonsuppressor host. This method permits determination of the efficiency of editing of a mismatch to 1 in 10(6) mismatches that escape editing and allows all eight mispairs that can yield viable revertants at an amber codon to be studied. The results of experiments with primers hybridizing to phi X174 am16 and am3 codons show that the order of mispair editing by T4 DNA polymerase is Ttemplate X Gprimer less than (A X G, T X C) less than (T X T, G X A, G X G, A X C) less than A X A. The efficiency of editing depends upon the mispair, as well as the neighboring DNA sequence. Under these conditions of synthesis, the 3'----5' exonuclease activity, depending upon the mispair and DNA sequences beyond the nearest neighbors, is estimated to contribute a factor of from 2.3 X 10(3)- to greater than 10(6)-fold to the accuracy of T4 DNA polymerase.