Data from dream research laboratories is reviewed in terms of those phenomena which do and do not support Freud's hypothesis that the specific function of dreaming is to guard sleep against disruption. The content of NREM sleep is thought to be more abstract, nonvisual, and thought-like than REM sleep. The hypothesis is presented that 'conflictual, analyzable mental activity' occurs throughout both day and night and that the difference is merely the mode of representation and clinical availability. There are some recent biological studies of sleep which support some of the basic ideas of psychoanalysis. However, the author warns that biological structural theory and analytical structural theory of id, ego, and superego are not interchangeable.