In most models of DNA replication, Watson-Crick hydrogen bonding drives the incorporation of nucleotides into the new strand of DNA and maintains the complementarity of bases with the template strand. Studies with nonpolar analogues of thymine and adenine, however, have shown that replication is still efficient in the absence of hydrogen bonds. The replication of base pairs might also be influenced by steric exclusion, whereby inserted nucleotides need to be the correct size and shape to fit the active site against a template base. A simple steric-exclusion model may not require Watson-Crick hydrogen bonding to explain the fidelity of replication, nor should canonical purine and pyrimidine shapes be necessary for enzymatic synthesis of a base pair if each can fit into the DNA double helix without steric strain. Here we test this idea by using a pyrene nucleoside triphosphate (dPTP) in which the fluorescent 'base' is nearly as large as an entire Watson-Crick base pair. We show that the non-hydrogen-bonding dPTP is efficiently and specifically inserted by DNA polymerases opposite sites that lack DNA bases. The efficiency of this process approaches that of a natural base pair and the specificity is 10(2)-10(4)-fold. We use these properties to sequence abasic lesions in DNA, which are a common form of DNA damage in vivo. In addition to their application in identifying such genetic lesions, our results show that neither hydrogen bonds nor purine and pyrimidine structures are required to form a base pair with high efficiency and selectivity. These findings confirm that steric complementarity is an important factor in the fidelity of DNA synthesis.