Past, present, and the future: discussions surrounding a new model of sleep-dependent learning and memory processing

Commentary/ Walker: A refined model of sleep and the time course of memory formation which settles the debate about the exclusiveness of memory con- tical inversion of the visual field. In the second study, the persons who experienced incorporations of the inverted visual field in In describing the findings regarding procedural memory and their dreams performed better on tasks (reading and writing) sleep in humans, Walker states that the evidence is “incredibly ro- measuring adaptation. This relation makes sense since research bust.” Reviewing the literature, however, one must say that the (Schredl 2000) has shown that dream content is related to specific number of studies is quite small, and direct replication studies brain activation patterns and other physiological parameters mea- carried out in different laboratories are scarce. Often different sured during sleep. Moreover, this is in line with the continuity hy- tasks (e.g., a visual discrimination task [Stickgold et al. 2000b], pothesis of dreaming (cf. Schredl 2003), which states that waking- motor skills like finger tapping [Walker et al. 2003b], acquisition life experiences, for example, the evening learning sessions, are of probabilistic rules [Peigneux et al. 2003], and priming [Plihal probably incorporated into subsequent dreams. An experimental & Born 1999a]) as well as different manipulation techniques (e.g., approach to this topic could be the technique of lucid dreaming, early versus late sleep [Plihal & Born 1997], REM sleep depriva- since it is possible to carry out assigned tasks during the dream tion [Karni et al. 1994], and correlations between sleep parame- (e.g., LaBerge & Rheingold 1990). For a simple motor activity ters and improvement [Stickgold et al. 2000b]) have been used.
(hand clenching), Erlacher et al. (2003) were able to demonstrate In our laboratory, we are currently conducting a correlation study that the related area of the motor cortex was active during the lu- applying the mirror trace task used by Plihal and Born (1997). The cid dream (EEG measure). This approach makes sense in the light preliminary findings (N ϭ 12) are promising: a significant corre- of the extensive literature on the effect of mental training on per- lation (r ϭ .430, p Ͻ .05, one-tailed) between percentage of REM formance (e.g., Driskell et al. 1994). Single cases of successful sleep and improvement in speed from the evening session to the training of sport skills in lucid dreams have been reported morning session was found. This is not completely consistent with (LaBerge & Rheingold 1990; Tholey 1981). On the other hand, the finding of Stickgold et al. (2000b) for the visual discrimination one should consider that dreaming as reportable subjective expe- task; they reported a much higher correlation (r ϭ .74; N ϭ 14).
riences during sleep is only a small part of the total activity of the Next, my coworkers (Orla Hornung, Francesca Regen, Heidi sleeping brain (comparable to consciousness during the waking Danker-Hopfe, and Isabella Heuser) and I utilized a modified state), so it remains unclear how close the relationship between version of the mirror-tracing task in a study of memory in elderly, dream content and learning processes during sleep might be.
healthy persons and were also able to demonstrate a correlation To summarize, the model proposed by Walker is a promising between the percentage of REM sleep and performance (this is a starting point for future research investigating, in addition to the preliminary result; the study is still in progress). On the other time course, influential factors such as task type, experimental dif- hand, the insignificant finding regarding non-REM Stage 2 sleep ficulty, and performance level in the relationship between sleep and performance is not in line with the findings of Walker et al.
(2003b). In addition to these conflicting results, other inconsis-tencies between the different studies in the field can be pointedout. Karni et al. (1994), for example, found an effect of REM sleepdeprivation on the improvement in the visual discrimination task Memory consolidation during sleep:
but not for slow wave sleep deprivation, whereas Stickgold et al.
A form of brain restitution
(2000b) reported correlations for slow wave sleep and REM sleep.
To summarize, although the amount of evidence supporting a close relationship of procedural memory and sleep is growing,many inconsistencies have to be clarified by futures studies.
University of Houston, Houston, TX 77204. bhavin@alum.mit.edu
If sleep plays a crucial role in memory consolidation, one of the next steps will be to study patients with primary sleep disorders.
Abstract: Does sleep restore brain function or does it consolidate mem-
ory? I argue that memory consolidation during sleep is an offshoot of resti-
Although Fulda and Schulz (2001) published a extensive meta- tution. Continual learning causes local synapse-specific neural fatigue, analysis on the cognitive impairment in patients with sleep disor- which then masks expression of that learning, especially on time-limited ders, detailed studies using paradigms including evening training tests of procedural skills. Sleep serves to restore the fatigued synapses, re- sessions and morning retest sessions have not yet been carried out vealing the consolidation-based enhancement observed as a “latent” in these patient groups. Keeping in mind the reduced daytime vig- ilance in these patients, it will be interesting to search for correla-tions between sleep architecture (total sleep time, percentage of Evidence for the involvement of sleep in memory consolidation REM sleep) and performance improvements in procedural as well comes in many forms, such as the effects of learning on postlearn- ing sleep and the re-expression of behavior-specific patterns dur- Assuming that REM sleep plays a crucial role in consolidation ing postlearning sleep. However, a cause-and-effect relationship of procedural memory (e.g., Plihal & Born 1997), studying the ef- or even a robust correlation between the effects of learning on fects of REM sleep augmentation on learning will be of interest.
sleep or the replay of patterns during sleep, on the one hand, and Schredl et al. (2001) have published the first human study in the magnitude of consolidation, on the other, has yet to be effec- which donepezil, an acetylcholinesterase inhibitor, was adminis- tively demonstrated. Improved learning following a period with tered to enhance REM sleep. A significant correlation (r ϭ .669, sleep, compared to one without, remains the most consistent evi- p Ͻ .05, one-tailed) between percentage of REM sleep and the dence to date; I propose an explanation for this.
improvement of a task (relearning a word list) that comprises de- I begin by noting that there exists emerging evidence for sleep clarative and implicit features was found for the donepezil nights.
as a localized brain process. While Rechtschaffen (1998) suggests Although this pilot study leaves many questions unanswered, this that it is “difficult to arrive at a widely acceptable theory of sleep research area is of interest because it was found that patients with function because that function is not reflected at the organ or sys- Alzheimer’s disease have reduced REM sleep (Bliwise 1993), and tem level,” he and others (e.g., Moruzzi 1966) propose that sleep cholinergic agents, which often enhance REM sleep – one of the is a localized process that provides basic cellular resources. In- measurable effects of these agents on the cholinergic system – deed, no brain lesion has ever successfully eliminated sleep totally (see Schredl et al. 2000), are widely used in the treatment of Alz- for long periods (Rechtschaffen 1998). In certain marine animals, sleep is sometimes localized to one brain hemisphere at a time The last topic to be addressed here is the possible relationship (Oleksenko et al. 1992). Continual tactile stimulation of the right between dream content and learning. Some preliminary evidence hand prior to sleep results in increased spectral power in the delta has been reported by De Koninck et al. (1988) for intense lan- band during early non-REM sleep in the contralateral so- guage learning, and De Koninck et al. (1996) for adaptation to ver- matosensory cortex (Kattler et al. 1994).
BEHAVIORAL AND BRAIN SCIENCES (2005) 28:1 Commentary/ Walker: A refined model of sleep and the time evidence exists but has eluded researchers, or perhaps proceduraland declarative memories differ in the same respects that makethe former more susceptible to sleep. Procedural learning is usu-ally dependent on the context and modality in which the materialwas presented initially (Squire 1986), are “realized as cumulativechanges stored within the particular neural systems engaged dur-ing learning” (Squire 1986), and typically require training for sev- Figure 1 (Sheth). A hypothetical evolution of local brain eral minutes to several hours on the procedure. In contrast, de- processes as a function of behavioral state, and the effects on clarative learning is flexible, accessible to all modalities and can be memory performance. Various processes (A, B, and C) combine “one-shot.” The weaker synaptic specificity and quicker learning to affect memory test performance (ordinate). With continual of declarative as opposed to procedural learning implies less lo- practice while the observer is awake, learning occurs (A), which calized declarative storage, which means, by our hypothesis, less improves performance. However, neuronal fatigue (B), which oc- synapse specific fatigue, and smaller benefits of sleep.
curs hand in hand with the learning, impairs performance. Dur- In sum, two issues are critical in the present account: (1) Synap- ing sleep, neural circuits slowly recover (C), which gives rise to “la- tic specificity: With greater synaptic specificity, there is greater tent learning.” A, B or C are transparent to the experimenter.
impact of sleep on local synaptic recovery; and (2) the need for Observed performance is some (nonlinear) combination of them.
speed: On perceptual as well as motor learning tasks, perceptionand/or motor action must be conducted within a finite period oftime for optimal performance (e.g., Stickgold et al. 2000b; Walkeret al. 2002). With time constraints, inefficiency of synaptic trans- With these ideas as a basis, I propose that two separate local mission takes on even greater significance, and, because speed-ac- brain processes are involved in the learning of a procedure or skill curacy tradeoffs are commonplace, the effects of sleep depriva- (see Fig. 1). Over repeated trials, the awake observer practices tion are observable on measures of speed as well as accuracy.
specifics of the procedure (A). Learning is a multifaceted process, By varying each factor, this hypothesis can be experimentally one facet being the progressive restriction in the brain circuits that verified. One possibility is to vary the degree of synapse-specific influence performance (Edelman & Tononi 2000; James 1890). I adaptation in two sets of synapses that exhibit learning during propose that while this process does not depend on sleep, a sec- training. Visual discrimination skills that transfer to different con- ond independent process exists that does. With increasing neural ditions (Ahissar & Hochstein 1996; 1997) are suitable for this.
specialization during learning, the circuits or synapses repeatedly Synapse-specific sleep dependent recovery will accordingly differ engaged in the procedure adapt or fatigue (B). Synapse-specific between the two sets. Learning following sleep loss will be im- fatigue during procedural learning is unavoidable. Repeated stim- paired following sleep loss in both brain circuits, but less so in the ulus processing produces decreased responses in brain circuits as- brain circuit that learnt the procedure indirectly via transfer.
sociated with that processing – a “repetition suppression” effect(Brown & Xiang 1998; Desimone 1996; Wiggs & Martin 1998).
The inefficiency in local signal transmission that arises from the I thank Professor J. Siegel and Dylan Nieman for carefully proofreading synaptic fatigue or adaptation masks expression of the learning.
With prolonged training on a task that involves both speed andskill, the net product of these two contravening processes, mea-sured behaviorally, is asymptotic learning (Karni & Sagi 1991).
Over still longer training periods, a decline in performance is ob- The incredible, shrinking sleep-learning
served (Mednick et al. 2002). Several studies (e.g., McCollough connection
1965) have shown that the effects of adaptation can be long last-ing, particularly if a select few synapses, specific to certain stimuli or conditions and not others, are adapted.
Neurobiology Research 151A3, V.A. Greater Los Angeles Health System, The recovery of functions related to sensory transmission, such Sepulveda, CA 91343; and Department of Psychiatry, UCLA David Geffen as the restoration of neurotransmitters or the re-formation of re- School of Medicine, 16111 Plummer Street, North Hills, CA 91343. ceptors, likely involves protein synthesis, which has its own char- JSiegel@ucla.edu
http://www.npi.ucla.edu/sleepresearch
acteristic time course, one that is longer than the time course ofresource depletion in the synapse. Thus, sleep-dependent Abstract: Initial claims that REM sleep is important in the consolidation
synapse-specific recovery (Fig. 1) is independent of the training in of all memories have been revised and reduced to the claim that sleep hasa role only in the consolidation of procedural learning. Now, Walker hy- the wake state. It is, however, dependent on sleep; sleep cannot pothesizes that sleep has no role in the “stabilization phase of consolida- be replaced with awake resting, which fails to control internally tion” but only in the “enhanced learning” phase of procedural learning.
generated activity in key brain areas, or by reversible inactivations Evidence for this vague, truncated hypothesis remains as inconsistent as of brain areas engaged in the consolidation, which only delays the recovery and may even shrink the critical time window duringwhich the learnt information can be actively enhanced (target ar- The idea that REM sleep is important for memory consolidation is attractive, since it would explain the vivid imagery of dreams as Sleep is clearly not monolithic. SWS, and non-REM sleep in a repetition of the events of the prior day to enable the laying down general, are believed to have a restorative role in brain function of permanent memory traces. Unfortunately, dream reports do (Horne 1988). It is notable that there is correlational evidence for not support this idea. Most dreams concern emotions and activi- the role of non-REM sleep in memory consolidation in humans ties that did not occur during prior days. Furthermore, most (Stage 2 sleep for motor skill learning, Walker et al. 2002; early dreams are not subsequently recalled unless they are immediately SWS for visual discrimination skill consolidation, Gais et al. 2000; rehearsed in waking following the dream (Rechtschaffen & Siegel Stickgold et al. 2000b). This dovetails nicely with the idea of synapse-specific recovery. Replay of behavior-specific patterns Those working on the role of sleep in human learning have during late REM sleep, if short-lived, may reactivate and reinforce modified their hypotheses to include non-REM sleep as well as the task-related synapses (target article, sect. 2.4.2) with minimal REM sleep. Many studies of the relationship between human sleep and learning have focused on sleep’s role in learning of word In contrast to procedural memory, evidence that sleep improves recognition and associations between words and events – tasks declarative memory is inconsistent (sect. 2.2). Perhaps conclusive mimicking most of what goes on in school; this is what learning BEHAVIORAL AND BRAIN SCIENCES (2005) 28:1

Source: http://www.ee.uh.edu/sites/ece.ee.uh.edu/files/44/files/Procedural%20Learning%20during%20Sleep%20is%20Synaptic%20Restitution--BBS%202005.pdf