Damaged DNA bases are removed from mammalian genomes by base excision repair (BER). Single nucleotide BER requires several enzymatic activities, including DNA polymerase and 5',2'-deoxyribose-5-phosphate lyase. Both activities are intrinsic to four human DNA polymerases whose base substitution error rate during gap-filling DNA synthesis varies by more than 10,000-fold. This suggests that BER fidelity could vary over a wide range in an enzyme dependent manner. To investigate this possibility, here we describe an assay to measure the fidelity of BER reactions reconstituted with purified enzymes. When human uracil DNA glycosylase, AP endonuclease, DNA polymerase beta, and DNA ligase 1 replace uracil opposite template A or G, base substitution error rates are <or=0.3 to <or=2.8 x 10-4. BER error rates are higher when excess incorrect dNTPs are included in the reaction or when wild type DNA polymerase beta is replaced by DNA polymerase beta variants that fill single nucleotide gaps with lower fidelity. Under these conditions, the base substitution fidelity of polymerase beta-dependent BER is 3-8-fold higher than is single nucleotide gap filling by polymerase beta alone. Thus other proteins in the BER reaction may enhance the base substitution fidelity of DNA polymerase beta during single nucleotide BER.