Eight patients suering from major depression were treated with trazodone 150 or 200 mg bid (non-responders) over 5 weeks, preceded and followed by a placebo week. Cognitive functioning was tested in the morning at the end of each treatment week. Before trazodone treatment commenced, patients' cognitive performance was compared with non-depressed controls. The depressed group showed signi®cantly slower reaction times on the memory scanning test, but did not dier from the control group on ®ve other tests including word learning, Critical Flicker/Fusion frequency (CFF), vigilance, tracking and divided attention. CFF threshold decreased signi®cantly during trazodone treatment. CFF returned to its initial baseline level at the end of the placebo withdrawal. During the trazodone treatment period, recall from long-term memory of a previously learnt word list signi®cantly improved, but at the end of the placebo withdrawal period memory performance returned to its initial baseline level. These results indicated that the subchronic in¯uence of trazodone on cognitive functions is at least of a two-dimensional nature: decreased CFF threshold can be interpreted in terms of lowered CNS arousal, whereas a higher cognitive function, memory, improved, possibly related to the relief of depressive symptoms.