Hospitalized manic-depressive patients were studied using ethological techniques. Within the sample three subgroups were created: a) manic patients whose psychiatric condition improved during the course of hospitalization (N = 3); b) manic patients who showed little or no improvement (N = 3); c) depressed patients (N = 2). Time-sampling procedures were employed to monitor the occurrence of a wide variety of behaviors in various parts of the hospital. Data were analyzed in terms of behavioral diversity, frequency, constancy, and profiles of behavior. Findings included: a) distinctive patterns of behavior characteristic of manic and depressed patients: manics had higher frequencies for most behavioral categories; b) characteristic patterns of behavior for patients who improved as compared to patients who showed no improvement: manic-improved patients showed a marked decrease in diversity as hospitalization progressed. Theoretical and clinical implications are discussed in relation to the kinds of data typically generated from ethological investigations.