To examine the relationship between the emotional quality of dreams, REM sleep variables and suicidal tendency in depressed individuals, 26 depressed volunteers (10 males and 16 females) were assessed with the Beck Depression Inventory (BDI) and the Hamilton Depression Rating Scale (HDRS), and underwent 3 nights of polysomnography. There was a significant negative correlation between suicidality scores and REM latency and a positive correlation between suicidality and REM percent. Suicidal subjects had a significantly shorter mean REM latency and a higher mean REM percentage than the non-suicidal subjects. As expected in normal subjects, 20 subjects had an increase in dream-like quality (DLQ) of REM reports between the first and second halves of the night. The six subjects with a negative DLQ difference also scored as suicidal. A reduction in dream-like quality of the REM content reports between the first and second halves of the night was found to be associated with suicidal tendency. The findings may indicate that these subjects fail to self-regulate mood and integrate affect into long-term memory networks during sleep. Theoretical and clinical implications of these findings in depression are discussed.