It was proposed by Bernad et al. [Cell 59 (1989) 219-228] and Blanco et al. [Gene 100 (1991) 27-38] that the 3'----5' exonuclease (Exo) domain of Escherichia coli DNA polymerase I (PolI) is structurally and functionally conserved among prokaryotic and eukaryotic DNA polymerases. The basis for this claim is the presence of three short peptide sequences in many DNA polymerases that resemble PolI sequences that have been shown by x-ray crystallographic and genetic engineering studies to be metal ion binding sites that are essential for PolI 3'----5' Exo activity [Derbyshire et al., Science 240 (1988) 199-201]. This claim is made even though there is little amino acid (aa) sequence similarity between PolI and many eukaryotic and viral DNA polymerases and in spite of significant differences in the amount of 3'----5' Exo activity in the DNA polymerases compared. For at least one DNA polymerase, bacteriophage T4 DNA polymerase, one of the proposed conserved Exo sequences does not appear to be important for 3'----5' Exo activity. This T4 DNA polymerase result provides a reminder that caution must be used when weak aa sequence similarities are used to predict protein structure and function.