To test the hypothesis that inaccurate DNA synthesis by mammalian DNA polymerase eta (pol eta) contributes to somatic hypermutation (SHM) of Ig genes, we measured the error specificity of mouse pol eta during synthesis of each strand of a mouse Ig kappa light chain transgene. We then compared the results to the base substitution specificity of SHM of this same gene in the mouse. The in vitro and in vivo base substitution spectra shared a number of common features. A highly significant correlation was observed for overall substitutions at A-T pairs but not for substitutions at G-C pairs. Sixteen mutational hotspots at A-T pairs observed in vivo were also found in spectra generated by mouse pol eta in vitro. The correlation was strongest for errors made by pol eta during synthesis of the non-transcribed strand, but it was also observed for synthesis of the transcribed strand. These facts, and the distribution of substitutions generated in vivo, support the hypothesis that pol eta contributes to SHM of Ig genes at A-T pairs via short patches of low fidelity DNA synthesis of both strands, but with a preference for the non-transcribed strand.