The fidelity of DNA polymerase-alpha-primase from calf thymus has been analyzed by measuring mutagenesis in vitro and by site-specific nucleotide misinsertion and mispair extension. Using the phi X174 am3 DNA reversion assay errors are detected at the amber3 site only when both dATP and dCTP are significantly biased during in vitro copying reactions. Analysis of these products on DNA sequencing gels reveals pause sites due to the slow extension of mispaired 3' termini. Measurements of misinsertion rates opposite template A show that the rates of dAMP or dCMP misinsertion are similar and occur 40-50 times more rapidly than dGMP misinsertion. The rate of extension from an A:C mispair is 100- and 400-fold greater than from an A:A mispair and an A:G mispair, respectively. Nucleotide misinsertions to generate all 12 possible mispairs have been measured kinetically on phi X174 DNA templates that contain either A, C, G, or T at position 587. Misinsertion frequencies range from 1/4000 to 1/10(6) depending on the mispairs generated. Extension from all 12 different mispairs was examined by starting with oligonucleotide primers that contain different 3'-terminal mispairs. Rates of extension from mispairs are 10(3) to 10(6) times slower than from correctly paired bases. Extension frequencies were purine:pyrimidine greater than pyrimidine:pyrimidine greater than purine:purine. Lack of extension of misincorporated bases suggests the involvement of exonucleolytic proofreading to enable continued DNA synthesis and to guarantee the high fidelity of eucaryotic DNA replication.