DNA polymerases accurately replicate DNA by incorporating mostly correct dNTPs opposite any given template base. We have identified the chemical features of purine dNTPs that human pol α uses to discriminate between right and wrong dNTPs. Removing N-3 from guanine and adenine, two high fidelity bases, significantly lowers fidelity. Analogously, adding the equivalent of N-3 to lowfidelity benzimidazole-derived bases (i.e., bases that pol α rapidly incorporates opposite all 4 natural bases) and to generate 1-deazapurines significantly increases the ability of pol α to identify the resulting 1-deazapurines as wrong. Adding the equivalent of the purine N-1 to benzimidazole or to 1-deazapurines significantly decreases the rate at which pol α polymerizes the resulting bases opposite A, C, and G, while simultaneously enhancing polymerization opposite T. Conversely, adding the equivalent of adenine's C-6 exocyclic amine (N-6) to 1-and 3-deazapurines also enhances polymerization opposite T, but does not significantly decrease polymerization opposite A, C, and G. Importantly, if the newly inserted bases lack N-1 and N-6, pol α does not efficiently polymerize the next correct dNTP, whereas if it lacks N-3 one additional nucleotide is added and then chain termination ensues. These data indicate that pol α uses two orthogonal screens to maximize its fidelity. During dNTP polymerization, it uses a combination of negative (N-1 and N-3) and positive (N-1 and † This work was supported by grants to RDK from the NIH (GM54194 and TW007372-01) and the Army Research Office (W911NF-05-1-0172), to MH from the Ministry of Education of the Czech Republic (Centre of Biomolecules and Complex Molecular Systems, LC 512), and to JE from the Deutsche Forschungsgemeinschaft (SFB 579).