The REM [rapid eye movement] sleep of 23 nonpsychotic patients with primary depression was studied by means of an automated REM analyzer during a drug-free period and again during amitriptyline administration. Initial drug administration (50 mg) was associated with an immediate reduction in the number, average frequency and average size of the REM. The average REM size remained suppressed with continued drug administration while the average REM frequency showed a rebound which was responsible for a partial recovery to predrug levels of the number of REM and total REM intensity. With regard to individual REM periods, REM frequency and REM intensity were redistributed during tricyclic administration so that the 2nd REM period became more intense than the 1st REM period. This automated REM analysis technique provides an objective set of measures for characterizing discrete aspects of REM sleep during a depressive episode and for evaluating the changes in REM sleep during psychotropic trials.