An in vitro model system including wild-type T4 DNA polymerase, the mutagenic nucleotide analogue 2-aminopurine deoxyribonucleoside triphosphate, and poly[d(A,C)] X oligo(dT) poly(dC) X oligo(dG) template-primers is used to measure the frequency of 2-aminopurine X cytosine base mispairs formed in the G X C----A X T mutational pathway. Incorporation and turnover of the analogue into DNA is dependent on the presence of cytosine on the template strand and is reduced significantly in the presence of dGTP. 2-Aminopurine X cytosine mispairs are observed to occur at a 2-3 order of magnitude greater frequency than adenine X cytosine mispairs. The frequency of inserting 2-aminopurine deoxyribonucleoside monophosphate in place of dGMP opposite template cytosine sites is about 3-6% when either strong or weak base-stacking partners are present on the primer strand. However, enzymatic proofreading of the mispair strongly depends on base-stacking partners. Greater than 85% of misinserted 2-aminopurine deoxynucleotides are excised whenever the mispairs are formed next to 5'-primer thymine sites. A 5-fold reduction in proofreading frequency occurs when the mispair is formed with 2-aminopurine deoxynucleoside monophosphate stacked adjacent to a 5'-primer guanine. The frequency of 2-aminopurine X cytosine base mispair formation in the G X C----A X T pathway is similar to that found previously in the A X T----G X C pathway (Watanabe, S. M., and Goodman, M.F. (1981) Proc. Natl. Acad. Sci. U.S.A. 78, 2864-2868). We propose a criterion for base selection by DNA polymerase to account for the unexpected similarity in base mispairing rates in the two transition pathways.