We describe the first systematic test of steric effects in the active site of a Y-family DNA polymerase, Dpo4. It has been hypothesized that low-fidelity repair polymerases in this family more readily accept damaged or mismatched base pairs because of a sterically more open active site, which might place lower geometric constraints on the incipient pair. We have tested the origin of low fidelity by use of five nonpolar thymidine analogues that vary in size by a total of 1.0 A over the series. The efficiency and fidelity of base-pair synthesis was measured by steady-state kinetics for single-nucleotide insertions. Analogues were examined both as incoming deoxynucleoside triphosphate (dNTP) derivatives and as template bases. The results showed that Dpo4 preferred to pair the thymidine shape mimics with adenine and, surprisingly, the preferred size was at the center of the range, the same optimum size as recently found for the high-fidelity Klenow fragment (Kf) of Escherichia coli DNA Pol I. However, the size preference with Dpo4 was quite small, varying by a factor of only 30-35 from most to least efficient thymidine analogue. This is in marked contrast to Kf, which showed a rigid size preference, varying by 1100-fold from best to worst. The fidelity for the non-hydrogen-bonding analogues in pairing with A over T, C, or G was much lower in Dpo4 than in the previous high-fidelity enzyme. The data establish that, unlike Kf, Dpo4 has very low steric selectivity and that steric effects alone cannot explain the fidelity (albeit low) that Dpo4 has for a correct base pair; the findings suggest that hydrogen bonds may be important in determining the fidelity of this enzyme. The results suggest that the low steric selectivity of this enzyme is the result of a conformationally flexible or loose active site that adapts with small energetic cost to different base-pair sizes (as measured by the glycosidic C1'-C1' distance), rather than a spatially large active site.