To determine the usefulness of including a duration criterion in the definition of "negative" symptoms, the prognostic significance of a longitudinally obtained measure of negative symptoms was compared with a cross-sectionally obtained measure. As predicted, the presence of negative symptoms based on longitudinal observation was associated with most aspects of 18-month course in a group of "first-episode" schizophrenic individuals, whereas cross-sectional levels of negative symptoms were not. The findings suggest that negative symptoms, when operationalized as a trait-like phenomenon, help to portend a poor course of illness.