DNA polymerases achieve high-fidelity DNA replication in part by checking the accuracy of each nucleotide that is incorporated and, if a mistake is made, the incorrect nucleotide is removed before further primer extension takes place. In order to proofread, the primer-end must be separated from the template strand and transferred from the polymerase to the exonuclease active center where the excision reaction takes place; then the trimmed primer-end is returned to the polymerase active center. Thus, proofreading requires polymerase-to-exonuclease and exonuclease-to-polymerase active site switching. We have used a fluorescence assay that uses differences in the fluorescence intensity of 2-aminopurine (2AP) to measure the rates of active site switching for the bacteriophage T4 DNA polymerase. There are three findings: (i) the rate of return of the trimmed primer-end from the exonuclease to the polymerase active center is rapid, >500 s(-1); (ii) T4 DNA polymerase can remove two incorrect nucleotides under single turnover conditions, which includes presumed exonuclease-to-polymerase and polymerase-to-exonuclease active site switching steps and (iii) proofreading reactions that initiate in the polymerase active center are not intrinsically processive.