Nature 437, 1254-1256 (27 October 2005) | doi:10.1038/nature04283; Published online 26 October 2005

Sleep is of the brain, by the brain and for the brain

J. Allan Hobson1


Sleep is a widespread biological phenomenon, and its scientific study is proceeding at multiple levels at the same time. Marked progress is being made in answering three fundamental questions: what is sleep, what are its mechanisms and what are its functions? The most salient answers to these questions have resulted from applying new techniques from basic and applied neuroscience research. The study of sleep is also shedding light on our understanding of consciousness, which undergoes alteration in parallel with sleep-induced changes in the brain.

For decades it was assumed that brain activity was greatly reduced or absent during sleep. Subjective experience of the loss of consciousness and the lack of memory of mental activity during sleep appeared to support this conclusion. Even such great scientists as Charles Sherrington1 and Ivan Pavlov2 backed this idea.

This assumption was overturned when the regular cyclic alteration of rapid eye movement (REM) and non-REM (NREM) sleep phases was discovered in the 1950s and 60s (ref. 3). The discovery of REM sleep and its correlation with vivid hallucinatory dreaming was evidence that the brain was highly active during sleep4. Soon after this discovery it was noticed that sensory inputs and motor outputs were simultaneously blocked when the brain was activated during REM sleep, putting it 'off-line'. (See Figure 1)

Figure 1: Behavioural states in humans.
Figure 1 : Behavioural states in humans. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact

States of waking, NREM sleep and REM sleep have behavioural, polygraphic and psychological manifestations. In the row labelled behaviour, changes in position (detectable by time-lapse photography or video) can occur during waking and in concert with phase changes of the sleep cycle. Two different mechanisms account for sleep imobility. The first is disfacilitation (during stages I–IV of NREM sleep). The second is inhibition (during REM sleep). During dreams, we imagine that we move, but we do not. Sample tracings of three variables used to distinguish the state are shown: an electromyogram (EMG), an electroencephalogram (EEG) and and electro-oculogram (EOG). The EMG tracings are highest during waking, intermediate during NREM sleep and lowest during REM sleep. The EEG and EOG are both activated during waking and inactivated during NREMsleep. Each sample shown is approximately 20 seconds long. The three bottom rows describe other subjective and objective state variables. Modified from ref. 19.

High resolution image and legend (59K)

It was a great surprise to discover that the vigorous brain activation of REM sleep occurred at regular 90-minute intervals and occupied up to 20% of sleep. This fact alone invalidated the belief that sleep was caused by and associated with a cessation of brain activity. Other facts supported the idea that the brain was continuously active during sleep. The early cerebral blood flow studies of Kety5 and later Sokolov showed only a 20% reduction in cerebral blood flow during sleep. Because blood flow is correlated with neuronal activity it should not have been a surprise to find that almost as many neurons increased their firing rate at sleep onset as their activity decreased6. Even during NREM sleep, when consciousness may be totally obliterated, the brain remains significantly active.

The descriptive study of sleep, so richly productive during the early years of the sleep laboratory era, has been complemented in the past decade by the application of brain imaging7, 8, 9 and quantitative electro-encephalogram (EEG) mapping10. Imaging techniques showed that the regional activation of the brain is very different in the two EEG-activated states, REM sleep and waking, and both are different from the NREM phase of sleep (when the EEG shows high-voltage slow waves instead of low-voltage fast activity). Quantitative EEG studies also revealed regional differences in brain electrical activity10, 11.

These new data indicate that the brain is relatively quiescent during slow-wave sleep (when the EEG is dominated by sleep spindles and high-voltage slow waves). But it must be emphasized that such global deactivation is only relative. Although consciousness is dulled, the brain is still roughly 80% activated and thus capable of robust and elaborate information processing. Thus, the EEG spindles and slow waves represent changes in the excitability of cortical and thalamic circuitry and should be regarded not simply as 'noise', which subjective experience leads us to assume, but as signals used by the brain for its own functional purposes12.

Mechanisms and functions of sleep

All in all, these findings support two radical ideas. One is that sleep is an actively regulated process, not simply the passive result of diminished waking. The other is that sleep should be regarded as a reorganization of neuronal activity rather than a cessation of activity. With respect to the first idea, it soon became apparent that although mammalian sleep occurs during the rest phase of the circadian rhythm, it is produced by brain processes in the hypothalamus and brainstem. It is in this context that Saper's recent description of a sleep switch is best understood and appreciated (see the review in this issue by Saper, Scammell and Lu, p. 1257).

Almost all mammals that have been studied show the NREM–REM cyclic alternation, which suggests not only a shared mechanism across species but a universal functional significance that must be far richer than the mere energy saving that the subjective experience-based theories thought adequate. The discovery of the continuous and elaborately modulated nature of sleep made it imperative to seek more active functional consequences, such as the homeostatic control of energy and the reinforcement of learning, that have recently been found. This is the context for thinking about Stickgold's descriptions of procedural learning enhancement during sleep (see the review in this issue by Stickgold, p. 1272).

The resurgence of interest in sleep and learning was sparked by robust evidence for sleep's promotion of consolidation and improvement in learned motor skill performance in terrestrial mammals. Some animals can learn despite having little and/or poor sleep, but this does does not mean that demonstrated sleep–learning links are irrelevant artefacts. The general principle could be that a species uses sleep for learning if it can afford to do so. The more vexing problem of understanding how sleep does (or does not) benefit narrative memory will be solved only by more assiduous and strategic study. Then, and only then, can we speak of sleep as beneficial to memory where memory is defined as the conscious recollection of learned material. Memory, so defined, depends upon learning but is not equivalent to it.

The variation in sleep between species and during their lifespan suggests, further, that sleep may have many functions. And those functions may not only vary but be absent in some animals. The same functions may also be achieved during periods of waking. But this flexibility is not evidence of functionlessness. In other words, to say that sleep is variable does not, of course, mean that sleep is not vital to those species that do sleep. It only means that relatively sleepless species have some other way of adapting to life's demands. Such plasticity is evidence of a pluralistic and, we assume, adaptive set of brain mechanisms and functions associated with sleep. This is the idea that inspires and underpins Siegel's phylogenetic work (see the review by Siegel, p. 1264).

The study of sleep phylogeny has a long history. It soon became clear that sleep, as we know it in higher mammals, tends to be correlated with relatively large brains and with homeothermy. The searching studies of Allison & Cichetti13 demonstrated that adaptations to diverse ecological niches also played a major role in determining the amount, temporal distribution and depth of sleep. The general rule was that large, carnivorous surface-dwelling animals, such as lions, slept long and deep whenever they were not foraging or mating. By contrast, smaller, herbivorous species such as rabbits tend to be nest dwellers and sleep relatively little. They have frequent awakenings and spend more time foraging and eating. They needed to be vigilant to defend themselves from predators. The commonsense conclusion from this work is that an animal sleeps if it can afford to.

Additional evidence of variability comes from the studies of human sleep that have uncovered a vast and complex set of sleep disorders. If one considers sleep to be an actively regulated process, it is understandable that some people get too little, others too much and others the wrong kind of sleep. The functional consequence of these disorders is similarly variable. Mahowald and Schenks's review (see p. 1279 in this issue) of these sleep disorders emphasizes some of the many possible dissociations of sleep and wake components that can afflict us.

What effect has the new science of sleep had on dream theory? Contrary to Freud's assertion that dreaming was stimulated by memory of that same day's experience (ref. 14), new data indicate that an equally large incorporation of recollected memory antedates its expression in dreams by as much as six days. While awaiting systematic replication, this finding should caution us about the uncritical acceptance of any theoretical formation about dreaming that is not based on evidence. A case in point is the finding that whatever the time lag to incorporation of some recent events in dreams, most dream content has no identifiable experiential antecedent15. Nielsen's studies pursue the implications of some of these findings (see the review in this issue, p. 1286).

Sleep and consciousness

Perhaps the most far-reaching of the complications of modern sleep science concerns the riddle of the basis of consciousness, a theme not addressed by any of the articles in this Insight.

A moment's reflection supports the idea that consciousness is state dependent. For centuries we made judgements about sleep and the brain that were wrong because we assumed, mistakenly, that consciousness ceased at sleep onset and resumed only when we woke. The occasional recall of dreams should have ruined that theory but great minds, including that of Sigmund Freud, incorrectly assumed that dreaming only occurred during the process of awakening.

Although it is true that consciousness is dulled during deep NREM sleep, it is qualitatively altered in parallel with the reorganization of brain activity that occurs at sleep onset when dreamlike mental activity is fleeting. This takes place during the lighter stages of NREM sleep, when dreaming may be more sustained and, most markedly, during REM sleep when dreaming assumes its most florid character. Because memory is so severely affected by sleep it has been difficult to get reliable and valid descriptions of mental activity during sleep, but it is now clear that consciousness undergoes alteration in parallel with sleep changes in the brain.

The net effect of this conclusion is to strengthen the conscious state hypothesis, which asserts that consciousness changes its intensity and character in a stereotypical way as the brain changes state during the sleep-wake cycle. One approach to the scientific study of consciousness is to simultaneously track changes in the brain and changes in the mind and then to map back and forth between the two domains. How might brain activity mediate conscious experience? The direct study of subjective experience during sleep is an important part of the current flurry of excitement in the domain of consciousness research16. Subjective experience is so problematical that all but the bravest scientists17 have been discouraged. And yet it must be recognized that subjectivity cannot be studied at all unless pains are taken to overcome the pitfalls of using first-person data. When large samples are taken, and the level of analysis is coarse-grained and its focus formal, the subjective data correlate with brain data at high levels of statistical and functional significance. Two examples help to make this point.

Normal subjects were asked to report their mental experience during periods of active wake, quiet wake, sleep onset, NREM sleep and REM sleep. The reports were scored for descriptions of hallucinatory perception and of thinking, and a reciprocal relation to brain state was observed. Hallucinatory mental content is lowest during active waking and highest during REM sleep. The incidence of thinking is reciprocally highest during quiet waking and lowest during REM sleep18.

The implication of these findings is that the sleeping brain can either generate its own perceptions or it can think about them. It cannot do both at the same time. Dreaming is therefore as hallucinatory and thoughtless (or delusional) as so-called mental illness.

In the second study we tested this hypothesis. When psychotic schizophrenic patients were given the thematic apperception test (TAT), in which verbal descriptions of simple but ambiguous pictures are recorded and scored, when they were awake and asked to report their dreams, they had equally high scores on a bizarreness scale (designed to pick up cognitive discontinuity and incongruity) for both. Age- and sex-matched normal control subjects have the same amount of dream bizarreness as the patients but are much less bizarre in their wake-state projective test responses (S. Scarone, M. L. Manzone, O. Gambini and J. A. Hobson, unpublished data).

These findings support the hypothesis that REM sleep is a physiological brain state that produces a distinctive and psychosis-like mental content, whereas during normal waking such properties are suppressed. Put another way, when awake the brain is normally free of the formal aspects of dream activity. Conversely, normal dreaming is justifiably considered to be an entirely normal model of highly abnormal conditions of the human brain and mind. It is now clear that the kind of consciousness that a person experiences is a function of the state of the brain.

Sleep and dream research is a rare convergence point for the biological and psychological sciences. Ongoing work in this area promises to bridge areas of research from molecular and cell biology, through neuronal populations, to behavioural and conscious states. It may even prove helpful in solving the mind–body problem.

Author Information Reprints and permissions information is available at



  1. Sherrington, C. Man on his Nature (Doubleday, Garden City, New York, 1995).
  2. Pavlov, I. I. Conditioned Reflexes. An Investigation of the Physiological Activity of the Cerebral Cortex (Dover, New York, 1960).
  3. Aserinsky, E. & Kleitman, N. Regularly occurring periods of eye motility and concomitant phenomena during sleep. Science 118, 273–274 (1953). | PubMed | ISI | ChemPort |
  4. Dement, W. C. & Kleitman, N. The relation of eye movements, body motility, and external stimuli to dream content. J. Exp. Psychol. 55, 543–553 (1957). | ISI |
  5. Kety, S. S. , Landau, W. M. , Freygang, W. H. , Rowland, L. P. & Sokoloff, L. The local circulation of the living brain; values in the unanesthetized and anesthetized cat. Trans. Am. Neurol. Assoc. 80, 125–129 (1955).
  6. Hobson, J. A. Dreaming: An Introduction to the Science of Sleep (Oxford Univ. Press, New York, 2002).
  7. Maquet, P. et al. Functional neuroanatomy of human rapid-eye movement sleep and dreaming. Nature 383, 163–166 (1996). | Article | PubMed | ISI | ChemPort |
  8. Nofzinger, E. A. , Mintun, M. A. , Wiseman, M. B. , Kupfer, D. J. & Moore, R. Y. Forebrain activation in REM sleep: An FDG PET study. Brain Res. 770, 192–201 (1997). | Article | PubMed | ISI | ChemPort |
  9. Braun, A. R. et al. Regional cerebral blood flow throughout the sleep-wake cycle — an (H2O)-O-15 PET study. Brain 120, 1173–1197 (1997). | Article | PubMed | ISI |
  10. Achermann, P. & Borbély, A. A. Low frequency (<1 Hz) oscillations in the human sleep electroencephalogram. Neuroscience 81, 213–222 (1997). | Article | PubMed | ISI | ChemPort |
  11. Huber, R. , Ghilardi, M. F. , Massimini, M. & Tononi, G. Local sleep and learning. Nature 430, 78–81 (2004). | Article | PubMed | ISI | ChemPort |
  12. Steriade, M. , Timofeev, I. & Grenier, F. Intracellular activity of various neocortical cell-classes during the natural wake-sleep cycle. Soc. Neurosci. Abstr. 25, 1661 (1999).
  13. Allison, T. & Cicchetti, D. V. Sleep in mammals. Ecological and constitutional correlates. Science 194, 732–734 (1976). | PubMed | ISI | ChemPort |
  14. Freud, S. An Outline of Psychoanalysis (Hogarth Press, London, 1949).
  15. Fosse, M. , Fosse, R. , Hobson, J. A. & Stickgold, R. Dreaming and episodic memory: a functional dissociation? J. Cogn Neurosci. 15, 1–9 (2003). | Article | PubMed | ISI |
  16. Libet, B. , Gleason, C. A. , Wright, E. W. and Pearl, D. K. Time of conscious intention to act in relation to onset of cerebral activity (readiness-potential). The unconscious initiation of a freely voluntary act. Brain 106, 623–642 (1983). | PubMed | ISI |
  17. Metzinger, T. On Being No One (MIT, Cambridge MA, 2003).
  18. Fosse, R. , Stickgold, R. & Hobson, J. A. Brain-mind states: reciprocal variation in thoughts and hallucinations. Psychol. Sci. 12, 30–36 (2001). | Article | PubMed | ISI | ChemPort |
  19. Hobson, J. A. & Steriade, M. In Handbook of Physiology: The Nervous System Vol. 4 (eds Mountcastle, V. and Bloom, F. E.) 701–823. (Am. Physiol. Soc., Bethesda, 1986).

Competing interests statement

The author declares no competing financial interests.

  1. Department of Psychiatry, Harvard Medical School, 74 Fenwood Road, 401 Park Drive, 2nd Floor East, Boston, Massachusetts 02115, USA.

Correspondence to: J. Allan Hobson1 Email:


These links to content published by NPG are automatically generated.


Awakening the brain

Nature News and Views (05 Sep 1996)

Psychology Insight and the sleep committee

Nature News and Views (22 Jan 2004)

See all 3 matches for News And Views