INSIGHT PROGRESS What are the memory sources of dreaming?

Investigators since Freud have appreciated that memories of the people, places, activities and emotions of daily life are reflected in dreams but are typically so fragmented that their predictability is nil. The mechanisms that translate such memories into dream images remain largely unknown. New research targeting relationships between dreaming, memory and the hippocampus is producing a new theory to explain how, why and when we dream of waking life events. The night-time production of dreams is an unexplained marvel of human existence. The function of dreams and their underlying brain mechanisms continue to be matters of intense academic debate. In this article, we consider several new avenues of research that may help to overcome long-standing obstacles to explaining the functions and mechanisms of dreaming. The research concerns how episodic memories (memories of personally lived experiences) are consolidated over time and how they may be modified during dreaming (Box 1). They are altered in such a way that their autobiographical origins are obscured even though the subjective context within which they appears is a credible simulation of reality. Memory sources are a key to dream formation Although the concept of episodic memory was unknown to Freud, he considered such memories to be instrumental in the formation of dreams 1 . He coined the term 'day-residue' to refer to elements that connected with experiences of the previous day and that he identified in all of the dreams he scrutinized. His detailed descriptions of day-residues in his patients' dreams as well as his theoretical model describing their transformation from memories into dream elements became central to psychoanalytic theory and therapy. These elaborate 'dreamwork mechanisms' remain largely untested today. When psychophysiological methods were introduced into the study of dreaming, the pursuit of memory sources continued with the aid of polysomnographic identification of rapid eye movement (REM) and non-REM (NREM) sleep stages and a variety of experimental methods for tracking these sources. Methods such as pre-sleep stimulation (for example, provocative films and virtual environments), sensory stimulation during sleep (for example, tones, shocks and odours) and subjects' post-hoc identification of memory sources of dream elements If dreaming is not reliably episodic as many nightmares are, it nonetheless does simulate reality in several striking respects. Dreams seem to take place in real, spatially coherent, environments with which the self interacts perceptually, for example, by orienting, seeking and assimilating sensory information, much as it does with the real world. The self also seems to engage realistic characters in emotional and intellectual exchanges. Semantic information and a sense of knowing are often also present. This apparent contradiction between dreaming's lack of a fully episodic structure and its portrayal of coherent virtual worlds raises several theoretical questions. Why do memories appear in dreams as fragments or partial episodes but also occasionally as complete replays? If episodic memory is inactive during REM sleep, as some researchers propose, why does REM sleep deprivation interfere with the consolidation of episodic memories in some cases but not others (see ref. 6 for a review)? Does the partial dissolution of memory episodes in dreaming imply that the basic 'units' of the episodic memory system can be identified from careful study of dream reports? Recent advances in our understanding of the memory functions of sleep (for example, see the review in this issue by Stickgold, p. 1272) are beginning to provide answers. And the study of dreaming is beginning to emerge as one of the most promising new methods for sampling the memory processes concealed beneath the veil of sleep. The hippocampus and episodic memories in dreaming A consensus among theorists is emerging that altered hippocampal function during sleep accounts, at least in part, for the absence of complete episodic memories during dreaming. Although they propose different mechanisms of hippocampal action Three lines of inquiry are, in our view, especially pertinent to this question. Each line of inquiry addresses a particular formal characteris