haut de page

mise à jour du 26 juin 2003
Sleep & dreaming
EF Pace-Schott et all
Cambridge ed
Phylogenetic data bearing on the REM sleep learning connection
JM Siegel
Department of Psychiatry, School of Medicine and University of California at Los Angeles, Neurobiology Research, Sepulveda VAMC, North Hills, CA 91343
Télécharger le texte ce texte au fomat PDF
The REM sleep memory consolidation hypothesis JM Siegel (PDF)


The phylogenetic data are inconsistent with the hypothesis that REM sleep duration is correlated with learning ability. Humans do not have uniquely high amounts of REM sleep. The platypus, marsupials, and other mammals not generally thought to have extraordinary learning abilities have largest amounts of REM sleep. The whales and dolphins (cetaceans) have the lowest amounts of REM sleep and may go without REM sleep for extended periods of time, despite their prodigious learning abilities.
Qu'apportent les données phylogenétiques à la notion
de lien entre sommeil paradoxal et apprentissage ?
Les données apportées par la phylogenèse ne peuvent confirmées l'hypothèse que la durée du sommeil paradoxal est corrèlée à la capacitié d'apprentissage. Les hommes ne sont pas les champions du sommeil paradoxal. L'ornithorynque, les marsupiaux, et d'autres mammifères, qui ne sont généralement pas considérés comme doués de fortes capacités d'apprentissage ont une durée de sommeil paradoxal bien supérieure à l'homme. Les baleines et les dauphins (cétacés) ont le plus court sommeil paradoxal et peuvent nager pendant de longues périodes sans avoir aucun sommeil paradoxal, malgré leurs prodigieuses capacités d'apprentissage.
The idea that REM sleep with its elaborate associated dream mentation has a role in memory consolidation is a very attractive one. It is a pity that the evidence supporting this beautiful idea is so weak. Vertes and Eastman marshal impressive evidence inconsistent with a major role of REM sleep in learning. To this I add the mammalian phylogenetic data.

Much of the evidence finit that has been advanced as supporting a positive relationship between REM sleep and learning derives from reported increases inREM sleep after learning and blockade of learning after reduction in REM sleep with deprivation procedures. As Vertes and Eastman point out, most if not all of this evidence does not withstand carefull scrutiny. Nevertheless if we pursue the logic of this approach, one would predict that animals with greater learning capacity would have greater amounts of REM sleep.

Indeed, learning and memory theorists often imply that humans have large amounts of REM sleep. In fact, humans follow the general trends within the animal kingdom. Being large animals they share the inverse relationship between size and total sleep amount.

Humans sleep less than most smaller mammals. If REM sleep is calculated as a percentage of total sleep time, humans appear to have a lot of REM sleep, though not a uniquely large amount. However, the animals with the largest amounts of REM sleep are not the primates. The animal with the most REM sleep is the duckbilled platypus, which has, depending on how the calculation is done, approximately 7-8 hours of REM sleep a day (Siegel et al. 1997; 1999). REM sleep in the platypus has some unusual features. Perhaps of most significance is the lack of the low voltage EEG that characterizes REM sleep in other adult mammals. If we put aside the platypus data, the next contenders for the REM sleep championship are the black-footed ferret and the armadillo (Marks & Shaffery 1996; Prudom & Klemm 1973; Van Twyver & Allison 1974). What intellectual attribute do these three animals have in cognition? ls it intelligence or stupidity ? Without disparaging the beauty and role of these animals in the ecosystem, they are largely instinct driven. Clearly they can learn as can all mammals, but they do not appear to be unique in their mental skills. In general, the marsupials and monotrenses have more REM sleep than the placentals.

How albout the other end of the spectrum? The mammals with the least REM sleep are thee cetaceans (whales and dolphins). Early reports in captive animals did not detect any clear episodes of REM sleep (Mukhametov et al. 1977; Mukhametov 1987; Shurley et al 1969; Oleksenko et al. 1992; Flanigan 1974). Clearly if dolphins have any REM sleep at all, they can go without it for days or weeks. A more recen study in a captive gray whale demonstrated occasional twitches during sleep (Lyamin et al. 2000). The most generous estimates of the REM sleep total in these animals would be less than 15 minutes a day. How does the learning ability of dolphins and whales, animals with the largest brains ever to exist on earth, compare with that of the platypus, ferret, and armadillo? It would he difficult to defend the notion that the latter are smarter than cetaceans. Across mammals, REM sleep time is negatively correlated with brain weight (Zeplin 1994).

Work by Jouvet-Mounier (Jouvet-Mounier et al. 197O) and a survey of the literature by Zepelin (1994) led to the conclusion that REM sleep time was correlated with immaturity at birth. Our recent findings in the platypus at the high end of the REM sleep scale and cetaceans at the low strongly support this conclusion.

The immaturity of the platypus, hatching from an egg, and remaining attached to its mother for an extended period after birth is consistent with its high level of REM sleep. The maturity at birth of the cetaceans wich can swim free of the mother and defend themselves immediately after birth is consistent with their low level of REM sleep. Neither the platypus nor the cetacean data is consistent with a relation to intellectual function or memory.

One way out of this dilemma for the learning-REM supporters is to argue that amount of REM sleep is not an informative variable; that REM sleep in the platypus may be less intense or effcient than that in the cetacean. However, this post hoc resasonning is not persuasive. There is no evidence that the very short REM sleep periods in the cetaceans are more intense or that those in the platypus are less intense. In fact the best evidence in terms of phasic event intensity argues just the reverse. The platypus has more than 6 000 phasic events during sleep/24 hours while the Gray whale has fewer than 10. As to the contention that time in REM sleep is not the important variable; this is the very basis of the claim of a relation between REM sleep and learning. The learning theorists cannot convincingly argue this point both ways.

In conclusion, the phylogenetic data provide additional evidence for the case against a key role for REM sleep in memory consolidation or intellectual function.
Why do we sleep ? Sejnowski and Destexhe
Schematic model of neurotransmitter circuits that are involved in the three states of vigilance
Hypothalamic regulation of sleep Salin-Pascual
Sommeil et respiration: influence des états de vigilance Guilleminault
The neurobiology of sleep: genetics, cellular physiology and subcortical networks Pace-Schott & Hobson
The cognitive neuroscience of sleep: neuronal systems, consciousness and learning sur le site Nature Reviews Neuroscience
Hypocrétine (orexine) et narcolepsie : dormir ou dîner E Mignot
Electrophysiological correlates of yawning KL Regehr
Sleep-wakefulness, EEG and behavioral studies of chronic cats without neocortex and striatum: the "diencephalic" cat Villablanca J, Marcus R
duree sommeil paradoxal