In order to study the structural principles governing DNA polymerase fidelity we have measured the rates of insertion of incorrect nucleotides and the rates of extension from the resulting mismatched base pairs, catalyzed by the Klenow fragment of DNA polymerase I. Using a combination of semi-quantitative and qualitative approaches, we have studied each of the 12 possible mismatches in a variety of sequence contexts. The results indicate that Klenow fragment discriminates between mismatches largely on the basis of the identity of the mismatch, with the surrounding sequence context playing a significant, but secondary, role. For purine-pyrimidine and pyrimidine-pyrimidine mispairs, the relative ease of mismatch synthesis and extension can be rationalized using a simple geometrical model, with the important criterion being the extent to which the mismatched base pair can conform to normal DNA geometry. Essentially similar conclusions have been reached in studies of other polymerases, suggesting that this aspect of mispair geometry is sensed and responded to in a similar way by all polymerases. Purine-purine mismatches form a less cohesive class, showing more variable behavior from mispair to mispair, and a greater apparent susceptibility to sequence context effects. Comparison of our data with studies of other polymerases also suggests that different polymerases respond to purine-purine mismatches in distinct and characteristic ways. An extensive analysis of each of the four purine-purine mispairs in approximately 100 different sequence contexts suggests that the reaction is influenced both by the local DNA structure and by the ability of the mismatched terminus to undergo slippage.