Previous investigations have estimated the human immunodeficiency virus type 1 (HIV) base pair substitution rate to be approximately 10(-4) to 10(-5) per round of viral replication, and HIV has been hypothesized to be more error-prone than other retroviruses. Using a single cycle reversion assay, we unexpectedly found that the reversion rates of HIV, avian leukosis virus and Moloney murine leukemia virus were the same, within statistical error. Because both the viral enzyme reverse transcriptase (RT) and cellular RNA polymerase II (RNAP) are required for viral replication, we hypothesized that the similar reversion rates actually reflect the intrinsic error rate of RNAP, which is the enzyme common to all three retroviruses in the reversion assay. To address this possibility, HIV vectors with the U3 region replaced by a reporter reversion cassette were constructed and vector supernatant produced by transient transfection. All single integrant revertant cell lines showed the identical mutations at both long terminal repeats. This indicates that either RNAP or another cellular enzyme is responsible for these reversions, or that HIV RT only makes errors during first strand synthesis. Additionally, when HIV particles were rescued from an integrated vector as opposed to being produced by transient transfection, the reversion rate was significantly lower, suggesting that one or more factors in the virus-producing cells plays a role in the fidelity of retroviral replication. These results have implications regarding the fidelity of the transgene after transient transfection production of lentiviral vector supernatants.