Discoveries in three distinct areas of dream research make it possible to suggest the outlines of a new neurocognitive theory of dreaming. The first relevant findings come from assessments of patients with brain injuries, which show that lesions in different areas have differential effects on dreaming and thereby imply the contours of the neural network necessary for dreaming. The second set of results comes from work with children ages 3–15 in the sleep laboratory, which reveals that only 20–30% of REM period awakenings lead to dream reports up to age 9 and that the dreams of children under age 5 are bland and static in content. The third set of findings comes from a rigorous system of content analysis, which demonstrates the repetitive nature of much dream content and that dream content in general is continuous with waking conceptions and emotional preoccupations. Based on these findings, dreaming is best understood as a developmental cognitive achievement that depends upon the maturation and maintenance of a specific network of forebrain structures. The output of this neural network for dreaming is guided by a “continuity principle” linked to current personal concerns on the one hand and a “repetition principle” rooted in past emotional preoccupations on the other.