The case against memory consolidation in rem sleep

BEHAVIORAL AND BRAIN SCIENCES (2000) 23, 793–1121
Printed in the United States of America
The case against memoryconsolidation in REM sleep Robert P. Vertes
Center for Complex Systems, Florida Atlantic University,
vertes@walt.ccs.fau.edu
Kathleen E. Eastman
Department of Psychology, Northern Arizona University, Flagstaff, AZ 86011
k.eastman@nau.edu
Abstract: We present evidence disputing the hypothesis that memories are processed or consolidated in REM sleep. A review of REM
deprivation (REMD) studies in animals shows these reports to be about equally divided in showing that REMD does, or does not, dis-
rupt learning/memory. The studies supporting a relationship between REM sleep and memory have been strongly criticized for the con-
founding effects of very stressful REM deprivation techniques. The three major classes of antidepressant drugs, monoamine oxidase in-
hibitors (MAOIs), tricyclic antidepressants (TCAs), and selective serotonin reuptake inhibitors (SSRIs), profoundly suppress REM sleep.
The MAOIs virtually abolish REM sleep, and the TCAs and SSRIs have been shown to produce immediate (40–85%) and sustained
(30–50%) reductions in REM sleep. Despite marked suppression of REM sleep, these classes of antidepressants on the whole do not
disrupt learning/memory. There have been a few reports of patients who have survived bilateral lesions of the pons with few lingering
complications. Although these lesions essentially abolished REM sleep, the patients reportedly led normal lives. Recent functional imag-
ing studies in humans have revealed patterns of brain activity in REM sleep that are consistent with dream processes but not with mem-
ory consolidation. We propose that the primary function of REM sleep is to provide periodic endogenous stimulation to the brain which
serves to maintain requisite levels of central nervous system (CNS) activity throughout sleep. REM is the mechanism used by the brain
to promote recovery from sleep. We believe that the cumulative evidence indicates that REM sleep serves no role in the processing or
consolidation of memory.
Keywords: antidepressant drugs, brain stem lesions; dreams; functional imaging; memory consolidation; REM deprivation; REM sleep;
theta rhythm
1. Introduction
sleep and memory as those that claimed such a relationship(Horne 1988; Horne & McGrath 1984; McGrath & Cohen Although its origin is difficult to establish precisely, the view that memories are processed and consolidated in sleep, or There has been a renewed interest in the role of sleep specifically in REM sleep, dates back at least to the report and memory stemming in part from two complementary ar- of Jenkins and Dallenbach (1924) claiming that human ticles that appeared in Science in 1994: one by Wilson and recall improves following an intervening period of sleep.
McNaughton (1994) on rats and the other by Karni et al.
There was intense interest in the possible role of sleep inmemory in the late 1960s to the 1980s as evidenced by thewealth of scientific papers on animals (and to lesser extent Robert P. Vertes is Professor of Neuroscience in the on humans) devoted to this issue. The position that memo- Center for Complex Systems and Brain Sciences at ries are consolidated in REM has been championed by, Florida Atlantic University. His main research interests among others, Pearlman (Pearlman 1971; 1978; 1979; include sleep, the anatomy and physiology of the brain- Pearlman & Becker 1973), Fishbein (Fishbein 1970; 1971; stem, and subcortical systems controlling the theta Fishbein & Gutwein 1977; Gutwein & Fishbein 1980a; rhythm of the hippocampus. He has written several 1980b); Hennevin and colleagues (Bloch et al. 1979; Hars articles and reviews on these topics. He is co- et al. 1985; Hennevin et al. 1995b; Leconte et al. 1974), and editor of Brainstem Mechanisms of Behavior. He cur-rently holds a research career award from the National Smith (1985; 1995; 1996; Smith & Butler 1982; Smith & Kelly 1988; Smith & Lapp 1991; Smith & Rose 1996; 1997).
There was a marked decline in the number of studies de- Kathleen E. Eastman is Assistant Professor of Psy- voted to this area beginning about the mid-1980s. As dis- chology at Northern Arizona University. She has written cussed below, the principal reason for this fall-off was that publications in the areas of visual motion perception, on balance the early work failed to convincingly demon- pattern recognition, perceptual stability, and the neural strate a relationship between sleep and memory. There were as many studies that failed to describe a link between 2000 Cambridge University Press Vertes & Eastman: Absence of memory consolidation in REM sleep (1994) on humans. In a follow-up to a study by Pavlides and review of this area, but rather is meant to serve as a general Winson (1989), Wilson and McNaughton (1994) reported background and critical assessment of some important is- that ensembles of hippocampal “place” cells tend to repeat patterns of activity of waking in subsequent episodes of slow wave sleep (SWS). Karni et al. (1994) showed that im-provement on a visual task in humans depended on REM 2.1. Effects of heightened experiences of waking
sleep. The two studies supported the view that memories on subsequent REM sleep
are consolidated in sleep. It is interesting to note that, pro- The rationale behind this set of studies is as follows: If pelled by these reports, this area reached national public at- REM serves to consolidate learning/memory, then expo- tention in the United States when Jonathan Winson and sure to enhanced learning situations or enriched environ- Matt Wilson appeared on the Charlie Rose television pro- ments in waking should result in increases in REM to gram explaining and promoting their shared belief that process and consolidate these experiences. We will only sleep is vital for memory consolidation.
briefly discuss this work for we do not believe that it rep- This area has recently received a further boost from Al- resents a particularly powerful test of the REM consolida- lan Hobson and colleagues who have recently come out in tion hypothesis owing, among other things, to confounding favor of the hypothesis that memories are consolidated in effects of natural variations in REM sleep and the difficulty REM sleep (Stickgold et al. 2000b). This recent position of establishing, at least for animals, that enriched experi- seems very much at odds with their earlier proposal, termed ences represent a significant departure from normal rou- the activation-synthesis hypothesis (Hobson 1988b; Hob- tines. Additionally, there is a certain degree of circularity in son & McCarley 1977), claiming that dreams (the cognitive this position, in that enhanced learning experiences in wak- component of REM sleep) represent the best cognitive fit ing presumably trigger increases in REM to consolidate (synthesis) to the undifferentiated and random action (ac- them, yet they only become “learning experiences” after tivation) of the brain stem on the forebrain, and as such being processed and consolidated in REM sleep.
would have little value to the organism and presumably The findings of several reports in animals and humans us- ing this paradigm have been mixed. In general, the majority As indicated by our title, we do not subscribe to the view of studies in animals have reported that heightened learning that memories are consolidated in REM sleep. This target experiences or enriched conditions in waking produce article evolved from an earlier piece by Vertes (1995) which increases in the amount of REM sleep (Horne 1988; Horne appeared as part of a series in Sleep Research Society Bul- & McGrath 1984; McGrath & Cohen 1978; Smith 1985); on letin on the topic of sleep and memory. In the same series, the whole, human studies have not shown this to be the case Hennevin et al. (1995a), supporters of a role for sleep in (Allen et al. 1972; Bowe-Anders et al. 1974; Horne 1976; memory consolidation, acknowledged why others may be Horne & Walmsley 1976; Zimmerman et al. 1978).
Horne and McGrath (1984) have raised objections to the The hypothesis of memory processing in sleep has always had animal work, pointing out, for instance, that in many of these to face criticism both from people working in the field of sleep, reports: (1) increases in REM appeared to be an “artifact” who predominantly consider that sleeping serves more basic bi- of an overall increase in total sleep time (TST); that is, the ological functions, and from people in the field of learning and proportion of REM to TST was not increased (Gutwein & memory, who do not easily accept the idea that information Fishbein 1980a; 1980b; Kiyono et al. 1981; Krech et al. 1962; processing can take place in a non-conscious state.
Mirmiran et al. 1982; Tagney 1973;); and (2) control animals As researchers involved in both sleep (Vertes 1984; 1990) were generally confined to impoverished environments, and memory work (Vertes 1986a; Vertes & Kocsis 1997), we raising the possibility that differences between control and remain skeptical on both counts, largely for the reasons put experimental animals involved decreases in REM (re- forth by Hennevin et al. (1995a); that is, sleep involves basic flecting decreases in TST) for controls rather than increases biological functions and memory requires consciousness.
for the experimental animals (Gutwein & Fishbein 1980a;1980b; Krech et al. 1962; Tagney 1973).
McGrath and Cohen (1978) reviewed 15 studies in hu- 2. Background
mans examining the effects of enhanced waking experienceson REM sleep (nondeprivation studies) and reported a lack Memory consolidation refers to neural processing that oc- of effect in 10 of the 15 reports. They concluded: “nondepri- curs after information is initially registered, which con- vation studies employing humans seemingly provide little tributes to its permanent storage in memory (Nadel & Mos- support for a relationship between REM sleep and learning.” covitch 1997). As mentioned, several reports appeared inthe 1970s exploring the possible role of sleep in memoryconsolidation. These studies were of two basic types: (1) ex- 2.2. REM deprivation (REMD) studies in animals
aminations of potential increases in REM sleep following REM deprivation (REMD) studies in animals and humans heightened experiences in waking; and (2) examinations of are of two types: prior REMD and post (or subsequent) the effects of REM sleep deprivation on previously learned REMD, reflecting whether the REM deprivation period tasks. A number of reviews (Dujardin et al. 1990; Fishbein precedes (prior) or follows (post) the learning situation.
& Gutwein 1977; Horne 1988; Horne & McGrath 1984;McGrath & Cohen 1978; Pearlman 1979; Smith 1985), in- 2.2.1. Post-REMD studies. Specifically, the post-REMD
cluding recent ones (Hennevin et al. 1995b; Rechtschaffen procedure involves training animals to criterion on a task(s), 1998; Smith 1995; 1996), have been devoted to the topic of depriving them of REM sleep for varying periods of time, sleep and memory. The following is not intended as a re- and then retesting them on the task(s). If REM is critical BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Vertes & Eastman: Absence of memory consolidation in REM sleep for learning/memory, REMD should severely disrupt these fects of the deprivation. Attempts to separate learning from functions; and if REM is not critical, REMD should have performance deficits primarily by looking at short term ver- sus long term effects of REMD have largely shown that im- The most widely used technique for depriving animals of pairments are short term or, in effect, performance deficits REM sleep is the water tank (or pedestal) technique. In (Fishbein 1970; 1971; van Hulzen & Coenen 1982).
brief, animals are placed on top of a small pedestal (usually For example, Fishbein (1971) trained groups of mice on a small inverted flower pot) that is surrounded by water. As a passive avoidance task, deprived them of REM sleep for animals enter REM sleep, they lose postural tone (atonia), 1, 3, 5, or 7 days using the pedestal technique, and then partially or fully slip from the pedestal into the water, and retested them on the task 30 min, 3 h, and 24 h following awaken. The procedure is thought to fairly selectively de- removal from the pedestal. The results showed that: (1) prive animals of REM sleep. Controls are placed on larger mice deprived of REM for 1 day showed no impairments at diameter pedestals or allowed normal sleep in their home any of the three retest intervals (i.e., 30 min, 3 h, or 24 h) and (2) mice deprived for 3, 5, or 7 days showed marked It is widely acknowledged that the pedestal technique in- deficits when retested at 30 min and 3 h but no impairments troduces several spurious and uncontrolled variables that when retested at 24 h. In essence, mice deprived of REM are generally recognized to confound results obtained with for 3, 5, or 7 days were very impaired on short term but not this method; these include isolation, wetness, heat loss, on long term retest (i.e., 24 h), indicating that deficits were high levels of stress, muscle fatigue, and a significant loss of most likely performance and not learning/memory deficits.
slow wave sleep as well as REM (Coenen & van Luijtelaar1985; Ellman et al. 1978; Fishbein & Gutwein 1977; Grahn- 2.2.2. Prior REMD studies. A number of reports (Bueno et
stedt & Ursin 1985; Horne & McGrath 1984; Kovalzon & al. 1994; Danguir & Nicolaidis 1976; Fishbein 1970; Hart- Tsibulsky 1984; Youngblood et al. 1997). The pedestal tech- mann & Stern 1972; Linden et al. 1975; Sagales & Domino nique is a severe method for REMD; alternatives are 1973; Stern 1971; van Hulzen & Coenen 1982; Venkat- presently used such as the multiple platform and pendulum krishna-Bhatt et al. 1978) have shown that depriving ani- techniques (van Hulzen & Coenen 1980; 1982; van Luijte- mals of REM sleep prior to training (prior REMD) impairs laar & Coenen 1986) as well as the recently developed disk- acquisition/learning on a variety of tasks. These studies, over-water method of Rechtschaffen and Bergmann (Recht- however, do not seem to test the REM consolidation hy- schaffen 1998; Rechtschaffen & Bergmann 1995).
pothesis since the deprivation period precedes training/ac- It appears that the problems inherent in the pedestal quisition and there is no potential carryover of information technique have significantly clouded findings obtained with pre to post REMD as in the post-REMD design.
it. In fact, Bill Fishbein, an advocate of the REM consoli- Aside from their intended purpose, we suggest that the dation hypothesis, recently acknowledged (Fishbein 1995) prior REMD studies support the position that the deficits that he abandoned REMD work in mice because he was not seen in post-REMD reports were performance and not able to respond adequately to criticisms leveled at the tech- memory deficits. With both paradigms (prior and post REMD) animals are impaired to similar degrees on the have anticipated all the flack that I received, in the years to same types of tasks. In the post-REMD paradigm, however, come, about the “stress factor” produced by the mouse-on-the the claim is made that deficits involve the inability of ani- pedestal technique. I spent a great deal of time trying to prove mals to use information learned prior to deprivation, as a di- that there was no stress factor. Despite my efforts to design ex- rect result of the loss of REM; that is, animals perform periments in a way that training and retention testing were not poorly following REM deprivation because without REM confounded by the pedestal procedure, it became clear that no they are unable to process, store, and utilize information ac- matter what control experiment I did, I was never going to con- quired before deprivation to meet the demands of the task vince everyone. Eventually this controversy led me to com- – a memory deficit. Although impairments are similar with pletely abandon the REM deprivation procedure and look in- the prior REMD paradigm, the claim could not be made stead at the effects of learning on REM enhancement.
that this involves a memory dysfunction. We suggest that With the caveat, then, that many of the REMD studies in both cases the impairments are mainly performance supporting a role for REM in memory consolidation may deficits due in large part to the debilitating effects of dep- lack validity based on the use of the pedestal technique, a rivation procedures. The following report (van Hulzen & review of the REMD work in animals shows studies to be Coenen 1982) is consistent with this view.
about equally divided among those showing that REMD van Hulzen and Coenen (1982) deprived two groups of disrupted learning/memory (Fishbein 1971; Leconte et al.
rats of REM sleep for three days – one group with the 1974; Pearlman & Becker 1973; 1974; Smith & Butler 1982; pedestal (or water tank) technique and the other with the Smith & Kelly 1988) and those showing that this was not the less stressful pendulum technique. Immediately following case (Albert et al. 1970; Dodge & Beatty 1980; Joy & Prinz deprivation, both groups were trained on a two-way shuttle 1969; Miller et al. 1971; Shiromani et al. 1979; Sloan 1972; avoidance task (acquisition) and then retested six days later.
Rats deprived with the pedestal technique showed severe As discussed above, it is generally acknowledged that de- impairments in acquisition but not on retest; those de- priving animals of REM sleep with the pedestal technique prived by the pendulum method showed no deficits on ac- or other means is debilitating. This has led to the view that the impairments seen following REMD are not true learn- The results show that prior REMD by a stressful tech- ing/memory deficits but merely performance deficits; that nique (pedestal), as opposed to a more moderate procedure is, animals are simply unable to perform the required (pendulum), affects immediate performance, while neither task(s), in large part owing to the physically debilitating ef- procedure impairs performance/learning when rats are fully BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Vertes & Eastman: Absence of memory consolidation in REM sleep recovered from REMD – that is, six days after deprivation.
and/or per day. For instance, in separate reports, the times The findings suggest that stress (or other factors) associated (post training) of the “REM window(s)” were: 9–12 and with REMD and not necessarily the loss of a particular stage 17–20 h (Smith & Butler 1982), 48–72 h (Smith & Kelly of sleep is largely responsible for the disruptive effects of 1988), 53–56 h (Smith & MacNeill 1993), 5–8 h (Smith & REMD. This was indicated by the authors when they stated: Rose 1996), and 1–4 h (Smith & Rose 1997). In fact, the shuttle box avoidance performance [was found] to be severely last two studies (Smith & Rose 1996; 1997) involved virtu- disrupted following 72 hrs of PS [paradoxical sleep] deprivation ally identical conditions (place learning with rats on the by means of the water tank technique. Similar effects could not Morris water maze) yet the window shifted from 5–8 h in be replicated in using the pendulum technique. Therefore, the the earlier report to 1–4 h in the later one. Apparently, the possibility that these phenomena are not due to PS deprivation only difference was a change from distributed (Smith & per se must seriously be considered. (van Hulzen & Coenen Rose 1996) to massed trials (Smith & Rose 1997).
It appears that REM windows (at least as defined for an- imals) are not present in humans. Smith and Lapp (1991) 2.2.3. Summary and conclusions. A review of REMD
examined patterns of REM sleep (potential windows) in studies in animals shows that they are about equally divided college students following an intense learning experience in showing that REMD does or does not disrupt learning/ (post exams) compared to baseline periods (summer vaca- memory. As developed above, it has been argued that re- tion), and reported that aside from an increase in the total ports claiming that REMD disrupts learning/memory are number of (rapid) eye movements in test versus control confounded by the use of very stressful deprivation proce- conditions (most prominent in the fifth REM period), there dures. It appears that stress (and associated factors) rather were no changes in sleep/REM sleep under the two condi- than the loss of sleep/REM sleep is responsible for the tions. They stated: “No other REM-related measure (min- learning/memory deficits seen in these studies. While these utes of REM sleep, % REM sleep or latency from stage 2 reports are open to other interpretations, there appears to onset to any of the five REM periods) was found to be sig- be no alternative explanation for studies that fail to show nificant. Further, there were no changes in any of the other sleep parameters measured” (Smith & Lapp 1991).
Following a comprehensive review of the REMD litera- Finally, although there is some suggestion from recent work in humans that information is differentially processed The memory consolidation theories for REM sleep function are in distinct phases of SWS and/or REM sleep (Plihal & Born having increasing difficulty in handling REM sleep deprivation 1997; Stickgold et al. 2000b), to our knowledge “REM win- findings, as it is clear from both animal and human studies that dows” has not been independently demonstrated outside of even the longest periods of deprivation do not incapacitate the laboratory of Smith and colleagues (see Smith 1996). It memory, and at best only produce modest decrements.
seems that this potentially important phenomenon would And further, “In sum, and in relation to the memory con- be considerably strengthened if confirmed in other labora- solidation hypothesis for REM sleep, I find the field of REM sleep deprivation and learning in animals unconvinc-ing.” 2.4. REMD studies in humans: Early reports
Compared to their numbers on animals, relatively few re- 2.3. REM windows
ports on humans have examined the effects of REMD on Carlyle Smith, a foremost advocate of the REM consolida- learning/memory. In contrast to the case with animals in tion hypothesis and a major contributor to this area, has put which reports were about equally divided among those forth and provided supporting evidence for the existence of showing, or not, that REMD affects learning, the majority “REM windows”; that is, specific segments of REM sleep of studies in humans have described minimal or no effects that are enhanced following learning and corresponding of REM deprivation on learning/memory (Castaldo et al.
segments which when disrupted (REMD) impair learning/ 1974; Chernik 1972; Ekstrand et al. 1971; Lewin & Glaub- memory. According to the proposal, memories are selec- man 1975; Muzio et al. 1972). If anything, complex tasks tively consolidated during the period of the REM windows (Empson & Clarke 1970; Tilley & Empson 1978), as op- (for review, see Smith 1985; 1995; 1996).
posed to simple tasks (Castaldo et al. 1974; Chernik 1972), The REMD studies of Smith and coworkers focusing on REM windows appear subject to some of the same prob- Following a review of early REMD studies in humans, lems as other REMD studies, foremost of which is the in- ability to adequately control for the stress factor associated It is clear that, given before or after learning, REM sleep de- with the use of the pedestal technique for REMD. How- privation does not lead to any greater learning impairment on ever, in defense of Smith and colleagues it should be noted simple tasks, but difficult tasks are more affected. Whilst these that their work is less vulnerable to this criticism because latter findings can reach statistical significance, the effects are their REMD periods are generally short, about 4–12 h.
still relatively small, and not convincing enough to support any On the other hand, there are difficulties with “REM win- theory that REM sleep has a crucial role to play in the consol- dows” not encountered by other REMD studies. Of signif- icant concern is the shifting nature of the REM window. Asreadily acknowledged by Smith (1985; 1996), the precise lo-cation of the window in REM varies widely, dependent on 2.5. REM sleep and memory consolidation
such factors as species and even strain of animals, the na- in humans: Recent reports
ture of the training tasks, and the number and distribution Karni and Sagi (1993) initially showed that improved per- (concentrated or dispersed) of training trials per session formance on a perceptual learning task required the pas- BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Vertes & Eastman: Absence of memory consolidation in REM sleep sage of time; that is, subjects showed no improvement til these discrepancies are resolved, it is difficult to evaluate immediately following training but marked improvement the reliability of the findings using this perceptual learning 8–10 h following training. As discussed below, they have ex- tended their original findings to sleep: performance wasshown to improve not only with an intervening period ofwaking but also of sleep (Karni et al. 1994).
2.6. Theta rhythm and REM sleep
The task involved identifying the orientation of three di- In a variation of the REM consolidation hypothesis, Jona- agonal lines (arranged either horizontally or vertically) em- than Winson has proposed and provided supporting docu- bedded in a background of horizontal lines. The stimulus mentation for the position that certain types of memory, (target and background elements) was presented briefly (10 specifically memories that are critical for the survival of the msec) in one quadrant of the visual field followed by a blank species, are selectively processed and consolidated in REM screen and then a patterned mask (100 msec). The interval sleep (Pavlides & Winson 1989; Winson 1985; 1990; 1993).
between the onset of the stimulus and onset of the mask The theta rhythm of the hippocampus figures prominently (stimulus-to-mask onset asynchrony, SOA) was varied, and in this proposal (Greenstein et al. 1988; Pavlides et al. 1988; the measure of performance was an 80% correct identifi- cation (threshold SOA) of the stimulus (horizontal or verti- Winson (1972) reviewed the behavioral correlates of the cal lines) at a set interval. The index of improved perfor- theta rhythm of waking in several species and showed that mance was a decrease in threshold SOA (Karni & Sagi 1993; theta was selectively present during certain behaviors char- acterized as species-specific behaviors that are critical for In the sleep study, Karni et al. (1994) trained subjects on survival; for example, exploration in rats, defensive behav- the task and then tested them after a normal night of sleep, iors in rabbits, and predation in cats. In addition, theta is sleep without SWS, or sleep without REM. They described present throughout REM sleep (Vanderwolf 1969).
significantly improved performance following a normal A number of recent reports (including those of Winson night of sleep as well as sleep that included REM but not and colleagues) have shown that theta is directly involved SWS (SWS deprivation condition), but no gains in perfor- in mnemonic functions of the hippocampus (for review, mance in the absence of REM sleep (REM deprivation Vertes & Kocsis 1997). For example, it has been demon- condition). Karni et al. (1994) concluded that learning of strated that: (1) long term potentiation (LTP) is optimally this perceptual skill was a slow latent process requiring con- elicited in the hippocampus with stimulation at theta fre- solidation over time. The period of consolidation could be quency (i.e., 5–7 Hz or pulses separated by 170–200 msec) in waking or sleep, but if in sleep, it required REM sleep (Diamond et al. 1988; Greenstein et al. 1988; Larson & Lynch 1986; 1988; Larson et al. 1986; Leung et al. 1992; Using the identical visual display, Stickgold et al. (2000b) Rose & Dunwiddie 1986; Staubli & Lynch 1987); (2) stim- recently reported, like Karni et al. (1994), that subjects ex- ulation delivered in the presence but not in the absence of hibited marked improvement on the task following sleep.
theta potentiates population responses in the hippocampus Specifically, they reported: (1) no improvement on the task (Bramham & Srebro 1989; Huerta & Lisman 1993; Pavlides over the course of waking; (2) no improvement unless sub- et al. 1988); and (3) discrete medial septal (MS) lesions that jects obtained at least 6 h of sleep; (3) improved perfor- abolish theta produce severe learning/memory deficits, as mance proportional to the total amount of sleep after 6 h of do MS lesions with unexplored effects on the hippocampal sleep; and (4) improved performance proportional to the EEG (Berger-Sweeney et al. 1994; Dutar et al. 1995; Ha- amount of SWS in the first quartile of the night (SWS1) and gan et al. 1988; Hepler et al. 1985; Kesner et al. 1986; Leut- to the amount of REM in the last quartile (REM4). They geb & Mizumori 1999; M’Harzi & Jarrard 1992; Mizumori proposed that learning was a two-step process requiring et al. 1990; Poucet et al. 1991; Shen et al. 1996; Stackman & Walsh 1995; Walsh et al. 1996; Winson 1978).
Although there are parallels between the two sets of find- In brief, then, Winson’s position is that theta serves to en- ings (Karni et al. 1994; Karni & Sagi 1993; Stickgold et al.
code survival-enhancing information during waking and to 2000b), there are several pronounced differences. A major consolidate this information during REM sleep. In this difference involves the performance of subjects during scheme, theta is essential for the acquisition of skills for sur- waking. As discussed above, Karni and Sagi (1993) origi- nally showed and subsequently confirmed (Karni et al.
The primary focus of the research of the senior author is 1994) that performance significantly improved over time the theta rhythm of the hippocampus. In fact, the senior au- during waking. By contrast, Stickgold et al. (2000b) reported thor was introduced to this area by Jonathan Winson and no improvement during post training waking behavior, even remains enormously grateful for the opportunity to learn after 12 h, commenting: “12 hours of wake behavior was in- from him. As is evident, however, we do not share Winson’s adequate to produce reliable improvement while as little as view that theta is instrumental in consolidating memories 9 hours of sleep reliably produced improved performance.” Additional differences were as follows: (1) Stickgold et al.
We believe that the case is strong for the involvement of (2000b) demonstrated a direct relationship between im- theta in mnemonic functions of waking but not of REM proved performance and total amounts of SWS, particularly sleep (Vertes 1986a; Vertes & Kocsis 1997). This seeming SWS1, whereas Karni et al. (1994) showed that depriving discrepancy was recently addressed by Fishbein (1996) subjects of SWS did not alter performance; and (2) Stick- stating, “Robert Vertes has published a variety of studies gold et al. (2000b) reported that a minimum amount of that would lead one to assume he would be a leading cham- sleep (6 h) was required for improved performance, and af- pion of the theory of memory consolidation in REM sleep.
ter 6 h gains were proportional to the total amount of sleep; Despite his important contributions he does not believe the neither was the case in the report by Karni et al. (1994). Un- BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Vertes & Eastman: Absence of memory consolidation in REM sleep Our position is that theta of REM is a by-product of the we believe that theta serves a mnemonic function in wak- intense activation of the pontine region of the brainstem in REM sleep; theta merely reflects this activation and as suchmay not have any functional significance in REM or at leastnot the same functional significance as in waking. In a se- 3. REM sleep and antidepressant drugs
ries of studies (Kocsis & Vertes 1994; 1997; Vertes 1979;1981; 1988; 1992; Vertes & Martin 1988), we have shown It is well recognized that virtually all major antidepressant that the theta rhythm is generated by a system of connec- drugs suppress REM sleep (for review, Vogel et al. 1990) tions from the pontine reticular formation (PRF) to the sep- and it has, in fact, been proposed that the clinical efficacy tum-hippocampus. In brief, cells of nucleus pontis oralis of of these drugs largely derives from their suppressant effects PRF fire tonically with theta and transfer this tonic barrage on REM sleep (Vogel 1975; 1983). The major classes of an- to the supramammillary nucleus of the hypothalamus tidepressant drugs are the monoamine oxidase inhibitors where it is converted into a rhythmical pattern of discharge (MAOIs), the tricyclic antidepressants (TCAs), and the re- and then relayed to the GABAergic/cholinergic pacemak- cently developed and widely used selective serotonin reup- ing cells of the medial septum to drive theta (Vertes & Koc- take inhibitors (SSRIs). A review of the actions of several members of these classes of antidepressants shows that they As previously described (Datta 1995; Jones 1991; Steri- ade & McCarley 1990a; Vertes 1984; 1990), pontine andlower mesencephalic regions of the brainstem contain dis-crete populations of cells that control individual events of 3.1. Monoamine oxidase inhibitors (MAOIs)
REM sleep; when activated together these cell groups trig- Of the antidepressants, the MAOIs have the strongest sup- ger each of the major indices of REM sleep (cortical EEG pressive action on REM sleep. A number of early reports desynchronization, hippocampal theta, muscle atonia, PGO using normal and patient populations showed that MAOIs spikes, rapid eye movements, myoclonic twitches, and car- virtually completely (or completely) suppressed REM sleep diorespiratory fluctuations), and hence the REM state. Part for weeks to several months. In an initial study, Wyatt et al.
of this orchestration of activity of the pontine RF in REM (1969) reported that the MAOIs, isocarboxazid, pargyline involves excitation of nucleus pontis oralis and conse- hydrochloride, and mebanazine, reduced REM from about quently theta. As argued above, theta of REM may simply 20–25% of TST to 9.7, 8.6, and 0.4% of TST, respectively, reflect a highly activated brainstem in REM, and thus bear and that in one subject REM was virtually eliminated for little functional relationship to its role in waking.
The presence of similar electrophysiological events in In a subsequent report in anxious-depressed patients, waking and sleep does not indicate that they serve the same Wyatt et al. (1971b) described the remarkable findings that (or even similar) physiological and/or behavioral func- the MAOI, phenelzine (Nardil), given at therapeutic doses, tion(s). For example, the cortical EEG desynchronization completely abolished REM sleep in six patients for periods of waking and REM by no means signifies identical pro- of 14 to 40 days. There was a gradual decline in amounts of cesses in the two states; that is, the EEG desynchronization REM sleep for the first two weeks on the drug and a total of waking is associated with diverse sensory, motor, emo- loss of REM after 3–4 weeks. In a complementary study tional, and cognitive processes that are notably absent in with narcoleptic patients, Wyatt et al. (1971a) reported that phenelzine completely abolished REM in five of seven pa- As indicated, we favor the position that theta is critically tients for the following lengths of time: 14, 19, 93, 102, and involved in memory processing functions of waking (Vertes 226 days. They stated that: “The complete drug-induced 1986a; Vertes & Kocsis 1997). Specifically, we propose that suppression of REM sleep in these patients is longer and theta serves to gate and/or encode information reaching the more profound than any previously described”; and further hippocampus simultaneously with it from various external that “no adverse psychological effects were noted during sources (e.g., the entorhinal cortex). In the awake state, the the period of total rapid-eye-movement suppression.” “information arriving with theta” is governed by the behav- Several other studies have similarly shown that MAOIs ioral situation (context); that is, the sum of internal and essentially abolish REM sleep. Akindele et al. (1970) re- external events relatively time locked to theta. If theta ported that phenelzine completely eliminated REM sleep were involved in memory processing functions in REM, it in four subjects (one normal and three depressed) for 2 to should, in a similar manner, gate information to the hip- 8 weeks, and addressing possible behavioral consequences pocampus in that state. Unlike waking, however, in which stated that, “Far from this leading to disastrous effects on the information reaching the hippocampus is dictated by mental functions, as some might have proposed, clinical im- behavioral circumstances, there appears to be no mecha- provement began.” Kupfer and Bowers (1972) showed that nism in REM for the selection and orderly transfer of in- phenelzine abolished REM in seven of nine patients, and formation to the hippocampus from other sources. If the drastically suppressed it in remaining patients from pre- transfer of information in REM is not orderly, or is essen- drug values of 23.1 and 24.8% of TST to 1.4 and 0.5% of tially chaotic, it would seem that there would be no func- TST, respectively. Finally, Dunleavy and Oswald (1973) re- tional value in consolidating or “remembering” this infor- ported that phenelzine eliminated REM in 22 depressed mation. In effect, dream-like material might be presented to the hippocampus in REM, but there would be no pur- If REM sleep were involved in memory consolidation, it pose in storing or consolidating it during REM. This may would seem that the total loss of REM with MAOIs for pe- be the reason that dreams (or other cognitive material of riods of several months to a year (Dunleavy & Oswald 1973; Kupfer & Bowers 1972; Wyatt et al. 1969; 1971a; 1971b) In sum, the theta rhythm is present in waking and REM; would affect memory. As indicated above, the loss of REM BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Vertes & Eastman: Absence of memory consolidation in REM sleep did not appear to be associated with any noticeable decline showed a 42% reduction in REM with paroxetine com- in cognitive functions in these largely patient populations.
pared to a 30% reduction with amitriptyline. Similar find- These studies, however, made no systematic attempt to as- ings have been described in other comparisons of these sess the effects of MAOIs on cognition.
classes of antidepressants (Kupfer et al. 1991; Nicholson & Other reports, however, have examined the actions of MAOIs, primarily phenelzine, on cognition/memory and Although the TCAs and SSRIs do not completely elimi- described an essential lack of impairment (Georgotas et al.
nate REM sleep, they significantly suppress it by as much 1983; 1989; Raskin et al. 1983; Rothman et al. 1962). For as 75–85% in the short term (days) and 40–50% in the long example, Raskin et al. (1983) observed no adverse effects of term (weeks/months). As discussed for the MAOIs, if mem- phenelzine on a battery of 13 psychomotor and cognitive ories are consolidated in REM sleep, it would seem that the tasks in a heterogeneous population of 29 depressed pa- sustained reductions in REM with TCAs/SSRIs would al- tients. Similarly, Georgotas et al. (1983; 1989) reported that elderly depressed patients given phenelzine for 2 to 7 weeks There is a substantial literature describing the effects of showed no alteration in several measures of cognitive func- TCAs and SSRIs on cognitive functions in normal and de- tion, and concluded that the lack of adverse effects with pressed subjects, including several reviews devoted to the phenelzine suggests that it is preferable to TCAs (see be- topic (Amado-Boccara et al. 1995; Deptula & Pomara 1990; low) in the treatment of depression in the geriatric popula- Knegtering et al. 1994; Thompson 1991; Thompson & Trimble 1982). Because these classes of antidepressants arein such widespread use, it is obviously important to know ifthey disrupt motor/cognitive functioning.
3.2. Tricyclic antidepressants (TCAs) and selective
serotonin reuptake inhibitors (SSRIs)
3.2.1. The effects of TCAs on cognition/memory. Al-
As discussed below, virtually all of the commonly used TCAs though there is conflicting evidence, mainly related to the and SSRIs significantly suppress REM sleep, but unlike the diverse procedures used to evaluate the effects of antide- MAOIs, do not eliminate it. Also, the TCAs and SSRIs pressants on cognition (Amado-Boccara et al. 1995; Dep- appear to exert immediate suppressive effects on REM tula & Pomara 1990; Thompson & Trimble 1982), the gen- (within the first few days of treatment); by contrast, the eral consensus is that some TCAs, primarily amitriptyline, MAOIs produce maximal effects on REM about 2–3 weeks impair memory, but most have minor or no effects on mem- ory (for review, Amado-Boccara et al. 1995; Deptula & Po- An early report by Dunleavy et al. (1972) in normal sub- mara 1990; Thompson 1991; Thompson & Trimble 1982).
jects analyzed the effects on sleep of six TCAs and showed Virtually all TCAs have some sedative and anticholinergic that four of them (imipramine, desipramine, chlorimip- actions (Hardman et al. 1996), and if cognitive dysfunctions ramine, and doxepin) markedly depressed REM, beginning are present with TCAs they reportedly involve these prop- with the first night of administration. Chlorimipramine had erties (Curran et al. 1988; Deptula & Pomara 1990; Spring the strongest suppressive effect on REM sleep, producing a complete loss of REM for the first three nights and an ap- A number of studies have shown that amitriptyline dis- proximate 50% reduction in REM for the remaining four rupts memory – whether given acutely or long term, to the depressed or nondepressed, and across all age groups (Bran- Several subsequent examinations of the actions on sleep connier et al. 1982; Curran et al. 1988; Lamping et al. 1984; of these and other TCAs (amitriptyline, amoxapine, nor- Linnoila et al. 1983; Spring et al. 1992; Warot et al. 1996).
triptyline, imipramine, maprotiline, clomipramine) have For instance, Spring et al. (1992) compared the effects of a demonstrated that, as a class, TCAs produce an immediate four-week treatment with amitriptyline and clovoxamine 40–70% reduction in REM and sustained 30–50% de- (an SSRI) on psychomotor and memory tests in depressed creases in REM sleep (Brebbia et al. 1975; Hartmann & Cra- outpatients, and reported that amitriptyline, despite allevi- vens 1973; Kupfer et al. 1979; 1982; 1991; 1994; Mendle- ating depression, significantly impaired performance on the wicz et al. 1991; Nicholson & Pascoe 1986; Passouant et al.
memory tasks. Clovoxamine, on the other hand, had no ad- 1975; Roth et al. 1982; Shipley et al. 1984; Staner et al.
verse effects of psychomotor/cognitive performance (see 1995; Ware et al. 1989). Of the TCAs, clomipramine ap- pears to be the strongest REM-suppressant (Passouant et Spring et al. (1992) attributed the disruptive effects of al. 1975; Sharpley & Cowen 1995; Thase 1998).
amitriptyline on cognition to its anticholinergic actions, The SSRIs, like the TCAs, produce an initial marked re- noting that, in general, anticholinergics (e.g., scopolamine) duction in REM sleep that slightly abates with time. Ex- disrupt memory (Caine et al. 1981; Drachman & Leavitt aminations of the effects on sleep of several SSRIs (in- 1974). They stated: “The decline in memory performance dalpine, fluvoxamine, fluoxetine [Prozac], paroxetine, and associated with amitriptyline apparently reflects the rela- zimelidine) show that on average they produce an initial re- tively high anticholinergic action of the drug, rather than duction in REM of 40–85% and long term decreases of 30– a deficiency in its antidepressant action.” And further, 50% (Kupfer et al. 1991; Nicholson & Pascoe 1986; 1988; “Among the tricyclics, amitriptyline has the most pro- Nicholson et al. 1989; Oswald & Adam 1986; Saletu et al.
nounced anticholinergic effects, and would, therefore be 1991; Sharpley et al. 1996; Shipley et al. 1984; Staner et al.
expected to have the most adverse effect on memory.” 1995; Vasar et al. 1994; Vogel et al. 1990).
Consistent with this interpretation, Curran et al. (1988) In general, the SSRIs exert stronger suppressive effects compared the effects on memory of four antidepressants on REM than do the TCAs. Staner et al. (1995) compared (amitriptyline, trazodone, viloxazine, and protriptyline) that the actions on sleep of long term treatment with paroxetine varied with respect to their sedative and anticholinergic (SSRI) and amitriptyline (TCA) in depressed patients, and properties, and showed that the sedating compounds BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Vertes & Eastman: Absence of memory consolidation in REM sleep (amitriptyline and trazodone) but not the nonsedating ones produced significant impairments. Lamping et al. (1984) (viloxazine and protriptyline) impaired performance on a reported that even though amitriptyline and clovoxamine battery of memory tests, and that disruptive effects were gave rise to comparable relief from depression, the two an- considerably greater with amitriptyline (an anticholinergic) tidepressants differentially affected memory; that is, “an than with trazodone (no anticholinergic properties) (Ger- impairment of memory after chronic amitriptyline admin- istration, as contrasted with an improvement in memory af- In contrast to amitriptyline, most other TCAs have mini- ter chronic administration of clovoxamine.” Spring et al.
mal or no adverse effects on memory/cognition. In a well- (1992) described virtually the identical findings using the designed study, Peselow et al. (1991) examined the effects on learning/memory of a four-week treatment with the TCA Finally, an early review of this area (Thompson 1991) con- imipramine (Tofranil) with 50 depressed outpatients, and cluded that: “Newer compounds devoid of antimuscarinic reported that imipramine improved memory in these pa- effects, particularly the serotonin reuptake inhibitors, if not tients. Although the improvement in memory was attributed sedative, have not been associated with memory impair- to the clinical efficacy of the compound (not to a memory- ment. Furthermore, a few more recent studies suggest that enhancing function for imipramine), Peselow et al. (1991) these drugs may exert a beneficial influence on memory clearly demonstrated, as have several others (Amin et al.
processes in memory-impaired individuals”; while a recent 1980; Friedman et al. 1966; Glass et al. 1981; Henry et al.
review (Amado-Boccara et al. 1995) similarly concluded 1973; Raskin et al. 1983; Rothman et al. 1962) that imip- that: “antidepressants which inhibit serotonin reuptake ramine did not impair memory – even though imipramine seem to have no deleterious cognitive effects.” is a powerful REM suppressant (Kupfer et al. 1994; Wareet al. 1989). For instance, Kupfer et al. (1994) showed that 3.2.3. Summary of the effects of antidepressants on cog-
imipramine produced sustained 35–40% reductions in nition/memory. In summary, (1) MAOIs virtually abolish
REM sleep for three years in depressed patients.
REM sleep but have no adverse effects on cognition/mem- Finally, several other TCAs (doxepin, desipramine, nor- ory. (2) TCAs suppress REM by 30–70%. While amitripty- triptyline, amoxapine, protriptyline, maprotiline, and chlor- line, a strong anticholinergic and sedative compound, dis- imipramine) that also suppress REM sleep reportedly pro- rupts memory, most other TCAs produce minimal, or duce little or no detrimental effects on memory (Allain et generally no, disruptive effects of cognitive/memory. (3) al. 1992; Curran et al. 1988; Georgotas et al. 1989; Liljequist SSRIs suppress REM sleep by 40–85% but do not alter et al. 1974; Linnoila et al. 1983; McNair et al. 1984; Pishkin memory or other cognitive functions.
4. Brain stem lesions and REM sleep in humans
3.2.2. The effects of SSRIs on cognition /memory. As is
well recognized, SSRIs are very widely used and cur-
Although sizeable lesions at rostral, mesencephalic levels of rently the most prescribed treatment for depression. As a the brainstem often result in persistent coma or death group the SSRIs do not appear to alter cognitive func- (Cairns 1952), those located more caudally within the pons tions. For instance, there is no indication that any of the are less severe and have been shown to give rise to a condi- following SSRIs have any detrimental effects on psy- tion termed the “locked-in” syndrome. As originally de- chomotor/cognitive functions in normal or patient popu- scribed by Plum and Posner (1966), patients with this syn- lations: fluvoxamine, zimeldine, clovoxamine (an SSRI drome are fully conscious, alert, and responsive, but are and partial noradrenergic reuptake inhibitor), sertraline, quadriplegic and mute. Most of the patients retain the abil- paroxetine, or fluoxetine (Curran & Lader 1986; Fair- ity to make eye movements and very limited facial/head weather et al. 1993; 1996; Geretsegger et al. 1994; Hind- movements and some can communicate by small facial ges- march & Bhatti 1988; Hindmarch et al. 1990; Lamping et tures. For instance, Feldman (1971) described a case of a al. 1984; Linnoila et al. 1983; Saletu & Grunberger 1988; woman with this syndrome who learned to communicate by Saletu et al. 1980; Spring et al. 1992).
Morse code using eye blinks and jaw movements.
Kerr et al. (1992) recently examined the actions of parox- A number of reports have examined sleep-wake profiles etine, alone or in combination with alcohol, on several psy- of these patients, and probably not surprisingly, have shown chomotor/cognitive tests in elderly nondepressed subjects that most of them (or at least those with bilateral pontine with the goal of determining whether SSRIs, unlike com- lesions) completely lack REM sleep (Chase et al. 1968; pounds with anticholinergic and/or sedative effects, may al- Cummings & Greenberg 1977; Markand & Dyken 1976).
ter cognitive functions. They speculated that SSRIs “are For instance, Markand and Dyken (1976) reported that unlikely to have detrimental cognitive and psychomotor ef- REM sleep was entirely absent in five of seven patients with fects because of their unique pharmacological profile,” and the “locked-in” syndrome; SWS was present in essentially noted further that “patients often report that treatment normal amounts. From case reports, the mental capacities with SSRIs leaves them feeling more able to think clearly.” of these patients, including memory for events and people, It was shown that paroxetine not only had no adverse effects on psychomotor and cognitive functions, but that it slightly Although rare, there have been a few reports of patients ameliorated performance deficits produced by alcohol (Kerr with bilateral pontine lesions who are conscious, ambula- tory, and verbally communicative (Lavie et al. 1984; Osorio Comparisons of the actions of amitriptyline and SSRIs on & Daroff 1980; Valldeoriola et al. 1993). It appears that the psychomotor/cognitive performance in healthy or de- lesions in these patients are less extensive than those with pressed subjects (Curran & Lader 1986; Fairweather et al.
the locked-in syndrome. Nonetheless, like patients with the 1993; Lamping et al. 1984; Linnoila et al. 1983; Spring et al.
locked-in syndrome, they lack REM sleep (Osorio & Daroff 1992) have demonstrated that amitriptyline but not SSRIs 1980; Valldeoriola et al. 1993). Osorio and Daroff (1980) BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Vertes & Eastman: Absence of memory consolidation in REM sleep described two such patients. Both of them showed similar 1994) that dreams are logical and meaningful, Hobson and sleep deficiencies, the most prominent of which was a com- colleagues (Hobson 1988b; Hobson et al. 1998b) have ar- plete loss of REM sleep. It was further pointed out that gued that dreams can be defined by such characteristics as aside from minor neurological deficits, the patients led nor- hallucinosis, bizarreness, delusion, and confabulation and mal lives. The authors stated: “Our two patients are the first have likened dreams to the “delirium of organic brain dis- awake and ambulatory humans in whom total absence of ease”(Hobson 1997b). Hobson et al. (1998b) have proposed REM sleep has been demonstrated. These REM deprived a purely physiological explanation for the amnesia of REM, patients behaved entirely appropriately and were by no pointing to the likely correspondence between memory loss means psychotic.” The “psychotic” reference alludes to the and underlying physiological changes in REM, stating: early notion, subsequently dispelled (Vogel 1975), that long “The loss of memory in REM sleep makes dreaming con- term REM deprivation produces psychosis.
sciousness much more difficult to recall than waking con- Lavie et al. (1984) described the interesting case of a man sciousness. This phenomenological deficit logically implies who at the age of twenty suffered damage to the pontine re- a physiological deficit: some functional process, present gion of the brainstem from shrapnel fragments from a gun- and responsible for memory in waking is absent, or at least shot wound. Following the injury, the man was comatose for 10 days, remained in critical condition for another two Independent of theories of dreams, recent functional weeks and then recovered. An examination of his patterns imaging studies in humans during sleep have revealed pat- of sleep at the age of 33 revealed that he essentially lacked terns of activity in REM that appear to reflect dream pro- REM sleep; that is, REM was absent on most nights and cesses, including its amnesic quality. Although differences averaged 2.25% of TST on the other nights. Similar to the exist among reports (Braun et al. 1997; 1998; Maquet et al.
study by Osorio and Daroff (1980), Lavie et al. (1984) re- 1996; Nofzinger et al. 1997), a fairly consistent pattern of ported that despite the virtually total loss of REM sleep, the brain activity in REM sleep in humans has emerged from man led a normal life. For instance, following the injury the these studies. Some important findings are as follows: (1) man completed college, then law school, and at the time of the pontine reticular formation is highly active in REM the study was a practicing attorney.
sleep; (2) primary sensory areas (e.g., striate cortex for the Although no systematic attempt was made to examine visual system) are inactive in REM; by contrast, extrastriate the cognitive capacities of these patients, the virtual total (visual) regions (as well as other sensory association sites) loss of REM sleep did not seem to result in any apparent are very active in REM; (3) limbic and paralimbic regions, including the lateral hypothalamus, the amygdala and ante-rior cingulate, and parahippocampal cortices, are intenselyactivated in REM; and (4) widespread regions of the frontal 5. Functional imaging studies of brain
cortex including the lateral orbital and dorsolateral pre- activity in REM sleep
frontal cortices show marked reductions in activity in REMsleep (Braun et al. 1997; 1998; Maquet et al. 1996; Nof- Recent functional imaging studies of human brain activity in REM sleep reveal patterns of activity that are consistent This general pattern of activity in REM has been viewed with dream processes but not with memory consolidation.
as a “closed system” (Braun et al. 1998); essentially, an in- The mental/cognitive content of REM sleep is dreams.
ternal network disconnected from inputs and outputs. For Although dreams are not restricted to REM, they are un- instance, the suppression of activity in the primary visual questionably a prominent feature of REM sleep. Dreams cortex (input) is consistent with the well-characterized sen- are the sole window to cognitive processes of REM sleep.
sory blockade of REM, whereas the deactivation of the pre- Although the function(s) of dreams have been, and con- frontal cortex (output) parallels the failure of dreams to in- tinue to be, strongly debated (see Revonsuo, this issue), a fluence executive systems for behavior. With respect to the generally agreed-upon feature of dreams is that they are latter, Braun et al. (1997) stated: “REM sleep may consti- poorly remembered. Similar to its function, diverse expla- tute a state of generalized brain activity with the specific ex- nations have been put forth to account for the amnesic clusion of executive systems which normally participate in the highest order analysis and integration of neural infor- Foulkes and coworkers (Foulkes 1982a; 1985; 1999; Foulkes & Fleisher 1975; Foulkes et al. 1989), leading pro- In effect (and not unexpectedly), the brain in REM sleep ponents of the view that dreams are a meaningful extension mirrors the dreaming brain; that is, internally generated vi- of waking mental life, have suggested that the reason sual images are fed to (or recruited by) the limbic system.
dreams are so easily forgotten is that the brain in REM They are then incorporated into dreams but due to the sup- sleep is in a reflective mode (akin to reminiscing about, or pression of activity of the prefrontal cortex dream scenarios reflecting on, events during waking) rather than in an en- are not often recorded and generally do not influence wak- coding mode. An important difference, however, between ing behavior. In this regard, in an article on the neural ba- the reflections of dreams and waking is that during waking sis of consciousness, Jones (1998) commented that the re- one can rapidly switch from the reflective to the encoding cent demonstration in imaging studies (Braun et al. 1997; mode to integrate and possibly store information. This can- Maquet et al. 1996) that activity in the frontal cortex is de- not readily be done in REM sleep and as a result the re- pressed in REM suggests “an attenuation of processes im- flections/reminiscences of REM (dreams) are lost to mem- portant in episodic and working memory and perhaps ex- ory (Foulkes 1985; Foulkes & Fleisher 1975).
plaining why unless awakened from a dream, a sleeping At the opposite end of the spectrum to the position of Foulkes and others (Domhoff 1969; 1996; Domhoff & Finally, if dream material is so readily forgotten in REM Schneider 1998; Hall & Van de Castle 1966; Van de Castle sleep (reflecting the state of the brain in REM), it seems BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Vertes & Eastman: Absence of memory consolidation in REM sleep unlikely that other mental phenomena that are not incor- longer periods of REM throughout sleep serve to prime the porated into dreams would be processed and permanently brain for a return to consciousness as waking approaches.
With respect to the latter, the disorientation experienced on In summary, the pattern of brain activity in REM sleep sudden, unexpected awakenings from sleep (middle of the is consistent with dreams but inconsistent with the orderly night), compared to natural awakening, may reflect an in- evaluation, organization, and storage of information which adequate preparation of the brain for waking due to in- is the domain of attentive, waking consciousness.
In line with the foregoing, reductions in REM, seen par- ticularly with antidepressants, are generally accompanied 6. A proposed function for REM sleep
by a reorganization of sleep; that is, marked increases inlight SWS and corresponding decreases in deep SWS as It appears that the active state of the brain during REM has well as frequent awakening (Cohen et al. 1982; Kupfer et fueled claims that REM sleep is involved in complex, al. 1989; 1991; Nicholson & Pascoe 1988; Saletu et al. 1983; higher order functions, including memory (for review, 1991; Schenk et al. 1981; Shipley et al. 1984; Staner et al.
1995; Wyatt et al. 1971b). For the SSRIs, this has been re- It is tempting to speculate, as several theories do, that ferred to as the “alerting” effect on sleep of these antide- magical processes occur during REM sleep; that is, that pressants (Kupfer et al. 1989; 1991; Nicholson & Pascoe during the unconscious state of REM sleep some pro- 1988; Saletu et al. 1983; 1991; Schenk et al. 1981; Shipley grammed or purposeful reordering of mental events occurs so that a nightly replay of daytime events during REM en- In accord with others (Benington & Heller 1995; Berger hances the storage or consolidation of these events. In con- & Phillips 1995), we believe that the general purpose of trast to the view that the effects of REM extend beyond sleep is restitution/recuperation for the CNS, and within sleep to influence waking activities, we propose that REM this context, the primary function of REM sleep is to pre- can be entirely understood within the context of sleep with- pare the brain/CNS for recovery from sleep.
out invoking mental phenomena or quasi-conscious pro-cesses (for review, Vertes 1986b). REM is a state of sleep;as such, it would seem that attempts to describe its function 7. Conclusions
should look to sleep and not to waking.
As described in detail in our earlier theoretical paper (see We believe that the evidence reviewed in this report dis- Vertes 1986b), we propose that the primary function of REM putes the claim that REM sleep serves a role in the consol- sleep is to provide periodic endogenous stimulation to the idation of memory. Numerous studies have shown that de- brain which serves to maintain minimum requisite levels of priving animals of REM sleep has no effect on learning/ CNS activity throughout sleep. REM is the mechanism used memory. Although other reports have shown that REM de- by the brain to ensure and promote recovery from sleep. We privation (REMD) disrupts memory, many of them have argued that the brain is strongly depressed in SWS, particu- been questioned based on the use of the stressful pedestal larly in delta sleep, and incapable of tolerating long continu- technique for REMD leading to the view that reported ous periods of relative suppression. REM serves the critical deficits were performance and not learning/memory function of periodically activating the brain during sleep deficits. The majority of REM deprivation studies in hu- without awakening the subject or disturbing the continuity mans have failed to show that REMD disrupts memory.
of sleep. By analogy, the process of induction and recovery Perhaps the strongest evidence against the memory con- from general anesthesia is a delicate one requiring the spe- solidation hypothesis comes from the demonstration that cial skills of highly trained medical professionals. The brain antidepressant drugs or brain stem lesions profoundly sup- performs a very similar function daily and seemingly flaw- press, or eliminate, REM sleep, yet neither appears to alter lessly. REM is an integral part of this process.
memory/cognitive functions. Finally, recent imaging stud- Our theory is consistent with sleep state organization; the ies in humans during sleep have described patterns of ac- main elements of which are that: (1) the percentage of tivity that are consistent with dreams, including their am- REM sleep is very high in early infancy (about 50% of total nesic quality, but inconsistent with the orderly processing, sleep time) and declines sharply at 2–3 months of age; (2) evaluation, and storage of information that characterizes sleep continuously cycles from light to deep sleep and back waking consciousness. In conclusion, we believe that the to lighter stages of sleep as the cycle repeats itself; and (3) weight of evidence, as reviewed herein, fails to support a REM sleep is quite evenly distributed throughout sleep role for REM sleep in the processing or consolidation of (occurring about every 90 minutes) and the duration of REM periods become progressively longer throughout sleep.
Regarding this organization, we would suggest that the high percentage of REM sleep in neonates serves to offset We thank Alison M. Crane for her critical reading and construc- equally high amounts of SWS in newborns (see also, Ben- tive comments on the manuscript. This work was supported by ington & Heller 1994); that sleep cyclically alternates be- NIH Grant NS35883 and NIMH Research Career Scientist tween light and deep sleep to prevent the brain from dwelling too long in deep SWS; and that the progressively BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6

Source: http://www.maths.tcd.ie/~mnl/store/VertesEastman2000a.pdf