Much biological research is directed at using biological variables to predict traditional symptom-based psychiatric categories. In this article, the authors discuss the need for a research strategy in which biological variables actually define psychiatric groups.