We studied the baseline sleep electroencephalogram (EEG) variables and treatment-related sleep changes after 35-46 days of amitryptiline treatment (AMI) in a group of 18 depressed inpatients, mostly suffering from a major depressive disorder endogenous subtype (according to the Research Diagnostic Criteria, RDC), with a short rapid eye movement (REM) latency. The aim of the study was to identify potential sleep "predictors" of clinical response to AMI as well as short-term sleep changes associated with alleviation of depression. Clinical response to the drug was defined as a reduction of more than 50% of the Hamilton Rating Score for Depression (HRSD). Eleven men and 7 women, 25-68 years old, were included in the study. Their sleep was recorded at baseline and after an average of 39 +/- 4 days of AMI treatment, at a mean daily dose of 165 +/- 35 mg. The comparison of responders (n = 9) and nonresponders (n = 9) with Wilcoxon's test showed that responders (1) were more severely depressed at baseline, and (2) had less stage 4 sleep. A discriminant function using baseline HRSD, stage 4 and the number of stage shifts allowed for discrimination between responders and nonresponders with a 100% hit rate. Antidepressant treatment had, however, no differential effect on sleep parameters in the two response groups.