Interindustry wage differentials in wage regressions estimated for individuals have been interpreted as evidence consistent with efficiency wage models. A principal competing explanation is that these differentials are generated by differences across workers in unobserved ability. This paper tests the unobserved ability hypothesis .by incorporating test scores into standard wage regressions as error-ridden indicators of unobserved ability. The results indicate that differences in unobserved ability explain relatively little of interindustry or interoccupation wage differentials.