This study examined whether a measure of learning potential could predict work skill acquisition in schizophrenia beyond the prediction offered by a single administration assessment. Fifty-seven outpatients with schizophrenia or schizoaffective disorder completed a test-train-test version of the Wisconsin Card Sorting Test as a measure of their learning potential. The outpatients were randomly assigned to training by errorless learning or conventional instruction on two work skills (index card filing and toilet tank assembly). Work skills were assessed both immediately and 3 months after training. Generally, patients with high learning potential performed better on the work skill tasks. Learning potential explained an additional 15 percent of variance beyond single administration assessment in participants' accuracy immediately after work skill training and an additional 13 percent of variance in participants' accuracy 3 months after training. These findings indicate that measures of learning potential contribute to the prediction of work skill acquisition, going beyond the predictive power of single administration assessments.