Studies of the correlation of subjective and objective sleep measures in depressed patients have produced mixed results so far. Further, they were carried out in sleep laboratories and tended to obtain one-off assessments, thus not taking into account the effect of treatment. We investigated forty (40) patients over the course of 8-week treatment of depression with either paroxetine or nefazodone. We used home polysomnography at baseline, nights 3 and 10, and week 8 of treatment, with extensive assessments of subjective sleep, the morning after each sleep recording. The patients were able to judge accurately their total sleep time and sleep onset latency, both before and during treatment. However, they were inaccurate in estimating the number of times they woke up during the night. Sleep satisfaction correlated negatively with Stage 1 sleep at baseline. Sleep quality was represented by a combination of subjective parameters measuring the ease of initiation and maintenance of sleep, and it appeared to derive from slow wave sleep and sleep continuity as seen in polysomnography. The partial discrepancy between subjective and objective measures suggests that a cognitive element is combined with the biological element to produce the sleep problems reported by depressed patients.