This paper draws on empirical and theoretical studies to argue that popular and professional conceptions of mental illness share specific traits with ethnic stereotypes: (1) they are exaggerated and serve to erect a qualitative boundary where none objectively exists: (2) they are maintained through selective perception, rationalization, and sanctions; (3) they help to erect the "thresholds,' i.e., the criteria, for crossing or recrossing the boundary; (4) they serve to define relations, including those of power, between groups; (5) because they perform these important cognitive and conative functions, they persist despite a flow of personnel across them and despite repeated demonstrations of their inaccuracy. They cannot be expected to change until the actual relations between groups change.