General mental ability (GMA) has been found to be the best predictor of job knowledge and job performance, and it is widely-used for personnel selection decisions. However, the use of GMA in selection is a concern for practitioners because of the large Black-White race differences associated with GMA tests. The use of GMA tests, therefore, results in adverse impact when basing decisions on predicted performance. In order to address this validity-diversity tradeoff, a more specific cognitive ability is examined – working memory (WM). Two-hundred participants (50% Black, 50% White) were given measures of GMA and WM before being presented with learning opportunities meant to teach them novel information. The participants were then instructed to complete tasks which apply this newly learned knowledge. WM was examined in terms of how much additional variance was accounted for in task knowledge and task performance after controlling for GMA. In addition, race group differences of WM were compared to those of GMA. Results indicated that WM was able to account for significant additional variance in knowledge and performance, and that this relationship have been moderated by task complexity. WM exhibited slightly smaller absolute race differences as well, but these reductions were nonsignificant. Results are discussed in terms of the possible use of WM in a selection context.