The observed wage gap between men and women is widely attributed to discrimination in the workplace. Yet within the context of the standard neoclassical framework, discrimination is neither a necessary nor sufficient condition. This paper presents a modified neoclassical model which supports equilibrium wage differentials and which has testable implications. The paper also surmounts a difficulty that has plagued many earlier assessments–separating prejudice from other explanations of the wage gap. By directing attention away from wages to other implied effects of discrimination, our model offers cleaner tests of the impact of prejudice in labor markets. Results of such tests are reported.