Data from all-night EEG sleep studies were used to distinguish normal subjects, primary depressed patients, and primary insomniac patients. In part 1, we compared 41 normal subjects, 56 depressed patients, and 18 insomniacs. In a univariate comparison with normal subjects, depressed patients showed less total sleep, longer sleep latency, more early morning awake time, more intermittent awake time, less delta sleep, less sleep efficiency, and shorter rapid eye movement (REM) latencies; compared with insomniacs, depressed patients showed greater early morning awake time, shorter REM latency, greater REM index, and greater REM density. Using multivariate discriminant analysis, 82% of the sample were correctly classified by diagnosis: 100% of the normal subjects, 72% of the depressed patients, and 77% of the insomniacs. Eight variables contributed to the multivariate separation of depressed individuals from insomniacs and normals: total sleep time, total recording period, sleep efficiency, sleep latency, early morning awake time, awake time, REM time and REM%. When the discriminant functions were applied to a second group of 18 primary depressed patients, 82% were correctly classified as depressed. These results suggest that primary depressed patients and primary insomniac patients may show relatively characteristic patterns of sleep abnormality.