In this paper, we present the results of a prospective semi-naturalistic study of the addition of trazodone for insomnia to a 4 week, 300mg/day venlafaxine treatment in 50 depressed inpatients. The Montgomery and Asberg depression rating scale was used as a rating instrument. The study is designated as semi-naturalistic due to the fact that, although the venlafaxine treatment regimen was strictly defined, the timing of the trazodone introduction and the dosage were determined by the clinicians. The indication was based on the persistency of insomnia despite the use of authorized sedative co-medication (zopiclone as a hypnotic, clorazepate as an anxiolytic). Among the 42 patients who completed the study, 27 did not receive trazodone (G1) while 15 did (G2). Although the two groups were not clinically different at study entry, G2 patients showed less improvement than G1 patients during venlafaxine treatment alone, both in terms of insomnia (MADRS item 4) and inner tension (MADRS item 3). After trazodone introduction, insomnia improved and the median (interquartile range) of this item in G1 and G2 patients showed no statistically significant difference on Day 28 (G1: 0 (0-1); G2: 0 (0-2)). However, inner tension did not improve and the median (interquartile range) was higher on Day 28 in G2 patients (G1: 1 (0-2); G2: 2 (1-4); P < 0.05). Thus, trazodone is probably used for patients who develop not only insomnia, but also inner tension/anxiety during venlafaxine treatment. However, it alleviates only the first symptom, not the second.