Fifty-one inpatients affected by a major depressive episode were divided into four groups according to mood disorder diagnosis and previous clinical history (bipolar disorder type I; bipolar disorder type II; major depressive disorder with at least three previous depressive episodes; and single depressive episode patients) and administered three consecutive total sleep deprivation (TSD) cycles. Mood changes were rated with a reduced version of the Hamilton Depression Rating Scale and with self-administered visual analogue scales. TSD caused better clinical effects in bipolar and single-episode patients; in particular, unipolar patients lacked effects in perceived mood after the first TSD and showed worse Hamilton ratings in respect to the other groups after the three TSD treatments. Discriminant function analysis could correctly classify 80% of bipolar patients, post hoc, based on TSD response. Further researches on the clinical efficacy of TSD must take into account the heterogeneity of depression and of its biological substrate.