A key tenet of the theory of human capital is that investment in skills results in higher productivity. The previous literature has estimated the degree of investment in human capital for individuals by looking at individual wage growth as a proxy for productivity growth. In this paper, we have both wage and personal productivity data, and thus are able to measure of the increase in workers’ output with tenure as evidence of the degree of learning on the job. The data is from an autoglass company. Most of production occurs at the individual level so measures of output are clear. We find a very steep learning curve in the first eight months on the job: output is 53 percent higher after eight months than it is initially. Our data show that these output gains with tenure are not reflected in equal percentage pay gains: pay profiles are much flatter than output profiles in the first year and a half on the job. For these data, using wage profiles significantly underestimates the amount of investment compared to the gains evident in output-tenure profiles. The pattern of productivity rising more rapidly than pay reverses after two years of tenure, although our evidence on this point is less reliable. Worker selection is also important. Workers who stay longer have higher output levels and faster early learning.