This paper reports findings of an experiment motivated by a dynamic labor market model that considers the problem faced by an employer in making hiring decisions of workers of different types. The question examined here is how quickly employers learn about the ability of a group of workers through observing the performance of representatives of that group in the workplace. If prior opinions are weak, the employer will quickly update any incorrect groupbased stereotypes it may have, with information from the workplace. On the other hand, if priors are heavily weighted, incorrect initial perceptions will result in persistent differences in wages. Our experimental findings are twofold. First, subjects' (employers') behavior moves quickly toward optimal choices. Second, strong priors are hard to establish. These results suggest that it would take a long time for employers to form group-based stereotypes, and that such stereotypes should go away quickly in response to signals that contradict these stereotypes.