For many American psychoanalysts the couch is the default position used with all analytic patients. This paper reviews empirical studies of the relation between the patient's position, sitting or lying, and free association. Findings are inconsistent; the patient's position may have an effect in some dyads and not in others. There have been no studies of a possible relationship between the patient's position and therapeutic outcome. Therefore, there is no empirical foundation for putting all analytic patients on the couch. We propose that selection of a position for the patient requires a careful, empathic, flexible, clinical judgment by the analyst, including consideration of the analyst's own theory, as well as of the patient's personality characteristics and diagnosis.