The Eating Attitudes Test (EAT; Garner & Garfinkel, 1979) is one of the most widely used self-report eating disorder instruments. Originally developed to diagnose anorexia nervosa, it is often used in nonclinical samples where it has a high false-positive rate, which is likely due to changes in diagnostic criteria. Because the EAT has not been validated with Diagnostic and Statistical Manual of Mental Disorders (4th ed. [DSM-IV]; American Psychiatric Association, 1994) criteria, we examined its criterion validity for discriminating between nonclinical women with and without an undifferentiated DSM-IV eating disorder diagnosis. We also examined differences in mean EAT scores among eating-disordered, symptomatic, and asymptomatic participants. Results show that the EAT has an accuracy rate of at least 90% when used to differentially diagnose those with and without eating disorders and that mean EAT scores differed among eating-disordered, symptomatic, and asymptomatic participants.