Classical genetic selection was combined with site-directed mutagenesis to study bacteriophage T4 DNA polymerase 3'----5' exonuclease activity. A mutant DNA polymerase with very little (less than or equal to 1%) 3'----5' exonuclease activity was generated. In vivo, the 3'----5' exonuclease-deficient DNA polymerase produced the highest level of spontaneous mutation observed in T4, 500- to 1800-fold above that of wild type. The large reduction in 3'----5' exonuclease activity appears to be due to two amino acid substitutions: Glu-191 to Ala and Asp-324 to Gly. Protein sequence similarities have been observed between sequences in the Escherichia coli DNA polymerase I 3'----5' exonuclease domain and conserved sequences in eukaryotic, viral, and phage DNA polymerases. It has been proposed that the conserved sequences contain metal ion binding ligands that are required for 3'----5' exonuclease activity; however, we find that some proposed T4 DNA polymerase metal binding residues are not essential for 3'----5' exonuclease activity. Thus, our T4 DNA polymerase studies do not support the hypothesis by Bernad et al. [Bernad, A., Blanco, L., Lazaro, J.M., Martin, G. & Salas, M. (1989) Cell 59, 219-228] that many DNA polymerases, including T4 DNA polymerase, share an extensively conserved 3'----5' exonuclease motif. Therefore, extrapolation from E. coli DNA polymerase I sequence and structure to other DNA polymerases for which there is no structural information may not be valid.