During DNA synthesis, DNA polymerases must select against ribonucleotides, present at much higher levels compared with deoxyribonucleotides. Most DNA polymerases are equipped to exclude ribonucleotides from their active site through a bulky side chain residue that can sterically block the 2'-hydroxyl group of the ribose ring. However, many nuclear replicative and repair DNA polymerases incorporate ribonucleotides into DNA, suggesting that the exclusion mechanism is not perfect. In this study, we show that the human mitochondrial DNA polymerase γ discriminates ribonucleotides efficiently but differentially based on the base identity. Whereas UTP is discriminated by 77,000-fold compared with dTTP, the discrimination drops to 1,100-fold for GTP versus dGTP. In addition, the efficiency of the enzyme was reduced 3-14-fold, depending on the identity of the incoming nucleotide, when it extended from a primer containing a 3'-terminal ribonucleotide. DNA polymerase γ is also proficient in performing single-nucleotide reverse transcription reactions from both DNA and RNA primer terminus, although its bypass efficiency is significantly diminished with increasing stretches of ribonucleotides in template DNA. Furthermore, we show that the E895A mutant enzyme is compromised in its ability to discriminate ribonucleotides, mainly due to its defects in deoxyribonucleoside triphosphate binding, and is also a poor reverse transcriptase. The potential biochemical defects of a patient harboring a disease mutation in the same amino acid (E895G) are discussed.