R OLE OF SLEEP

A second factor may have critical effects on the consolidation of memory: sleep. Before reviewing the principal studies that show sleep-dependent memory consolidation, I will briefly describe what sleep is.

2.5.1. What is sleep?

Sleep is defined as a natural periodic state of rest for the mind and body, in which the eyes are usually closed and consciousness is completely or partially lost. However, it does not refer to a single state. Instead it covers several brain states that differ in their physiology (e.g., presence or absence of eye-movements), chemistry (e.g., variation in Acetylcholin secretion), and phenomenological experiences (e.g., presence or absence of dreams). Humans and other mammals, sleep is composed of characteristic 90-min cycles of brain-wave activity that includes two major ones, the Non-Rapid Eye Movement sleep (NREM, also called Slow Wave Sleep, SWS), mostly present in the first half period of sleep, and Rapid Eye Movement sleep (REM), occurring mostly in the second half of the night. REM sleep is characterized by muscle atonia, fast cortical activity similar to the waking state (as measured by EEG), and ocular saccades. REM sleep was also originally called Paradoxical sleep (Jouvet, 1965).

NREM sleep has been divided in four stages (1-4), and is characterized by slow cortical activity as well as specific oscillations, called spindles (Rechtschaffen & Kales, 1968).

Despite the increasing knowledge on the mechanisms generating and regulating sleep, still little is known about its function. Among several hypotheses (detoxification, restoration or energy conservation), it has been proposed that sleep periods are favorable for brain plasticity and, for learning and memory. According to this hypothesis, sleep could participate in the consolidation of memory traces. The information acquired during wakefulness would be actively restructured, strengthened, and integrated in a network of pre-existing memories during sleep

2.5.2. Influence of sleep on memory processes

Accumulated evidence led to suggest that mechanisms of brain plasticity underlying memory formation and consolidation might be modulated by specific brain-states (e.g., waking or sleep-stages) both in animals and in humans (for reviews, see Buzsaki, 1998; Ribeiro &

Nicolelis, 2004; Hobson & Pace-Schott, 2002; Maquet, 2001; Walker, 2005).

At the neurophysiological level, changes in some sleep parameters during post-training sleep episodes have been shown to relate to subsequent performance levels (e.g., Hennevin et al., 1995; Gais et al., 2002). Experience-dependent replay of hippocampal and neocortical ensemble activity have been observed in rodents and birds (Pavlides & Winson, 1989; Wilson

& McNaughton, 1994; Wilson & McNaughton, 1994; Louie & Wilson, 2001; Battaglia et al., 2004; Dave & Margoliash, 2000). Reactivation of brain areas in humans (Maquet et al., 2000;

Peigneux et al., 2004; Rasch et al., 2007) during both REM and NREM sleep have also been found. These results suggested that an information acquired during awakening can be further processed in sleep and might underlie subsequent enhancement of behavioral performance (Stickgold et al., 2001).

In addition, fMRI studies in humans demonstrated that post-training sleep could have a critical effect upon lasting changes in brain responses. For example, Maquet and colleagues (Maquet, Schwartz et al., 2003) investigated sleep-dependent brain activity changes using a visuo-motor pursuit task. Subjects were trained to follow a visual target moving along a partially predictable trajectory with a joystick. After sleep, they were better at following the target moving along that trajectory as compared to a new trajectory. This was not the case for subjects who had been sleep-deprived. Performance improvement was correlated with an enhancement of activity in the superior temporal sulcus. Moreover, some changes in connectivity were identified between the supplementary eye field and the frontal eye field, implicated in the visual tracking of the target. Other studies found sleep-related increase of activity in the hippocampus (Gais et al., 2007; Walker et al., 2005), and in the early visual cortex (Schwartz et al., 2002; Yotsumoto et al., 2009).

At the behavioral level, post-training sleep deprivation can drastically reduce learning-related performance improvement in many tasks, including perceptual (Karni et al., 1994; Stickgold, James et al., 2000; Gais et al., 2000; Gaab et al., 2004) or motor skills learning (Walker et al., 2002; S. Fischer et al., 2002), as well as both explicit and implicit memory in more cognitive tasks (Plihal & Born, 1997; Gais & Born, 2004b; Wagner et al., 2004).

Some studies showed that REM and non-REM sleep could have an impact on different memory systems, respectively on implicit/procedural memory and on explicit/declarative memory (Gais & Born, 2004b; Karni et al., 1994; Plihal & Born, 1997; Plihal & Born, 1999;

Yaroush et al., 1971; Drosopoulos et al., 2005; for a review on the role of sleep on explicit memory, see Marshall & Born, 2007). However, other findings could not confirm such a clear-cut dissociation between sleep-stages functions but rather suggested a sequential processing of memory traces across brain-states whereby memory formation would be prompted during early non-REM (or slow-wave) sleep and further consolidated during

subsequent REM sleep (Giuditta et al., 1995; Stickgold, Whidbee et al., 2000; Peigneux et al., 2001; Buzsaki, 1989).

Daytime sleep is also favorable to learning, as shown by nap studies (e.g., Doyon et al., 2009). Even a very short interval of sleep (6 minutes) can lead to better performances in a word memory task (Lahl et al., 2008). In this later experiment, but more generally in all sleep studies, the question of the absence of interference has to be considered. Does sleep actively consolidate memory or does it ‘only’ protect the new memory trace which would be weakened by interference? This question has not been resolved yet. However, in a recent study, Ellenbogen and colleagues (2009) tested for sleep-dependent memory in a word-association task, and introduced an interference task after the wake or sleep period, just before the memory test. They found that the sleep group was more resistant to interference than the wake group, suggesting an enhancement of memory robustness during sleep.

Recently, Fenn et al. (2009) showed for the first time that sleep reduces false recognition.

They tested memory for words, after sleep or a comparable period of wakefulness.

Participants were presented with lists of words at study, and with three categories of words at testing, the studied ones, new ones and ‘lures’ which were semantically related to the studied ones. Participants who slept between the study and the test phases made fewer ‘old’ responses for the lures. This suggests that sleep may have a strong influence on episodic memory accuracy.

Finally, consistently with the standard memory consolidation theory (see section 2.3.1), a transfer from the hippocampus to neocortex has been found in several fMRI studies (Gais et al., 2007; Sterpenich et al., 2007; Takashima et al., 2009).

Overall, these studies provide strong evidences of the importance of sleep in memory consolidation. Memory representations might thus undergo successive transformations to become more stable and more accurate.

2.5.3. Sleep and face learning

Despite the massive literature on face processing and learning, little is known about how sleep affects memory for faces.

A few behavioral studies investigated this issue. Clemens et al. (2005) tested the effect of sleep on verbal declarative memory and used faces as a nonverbal control task. Subjects learned face-name associations and were then tested after a night of sleep, on their memory

for names and memory for faces. They found an effect of sleep on verbal memory, with a correlation with the number of spindles during sleep, but also an effect of sleep on memory faces, correlating with the total time of NREM sleep. Wagner and colleagues investigated behaviorally both implicit and explicit memory for faces. They found increased repetition priming effect across REM sleep in a task requiring identification (Wagner et al., 2003), and enhanced memory accuracy in recognition memory, independently of the emotional valence of the face (Wagner et al., 2007). Hussain and colleagues also found an effect of sleep, but very weak, in a 1-to-10 face identification task, with varying levels of contrast and noise (Hussain et al., 2008). These studies assessed the effect of sleep on learning at the behavioral level.

Mograss and colleagues used ERPs to test for this effect (Mograss et al., 2006; Mograss et al., 2008). They found an increased old/new effect after nocturnal sleep and associated modifications of early and late frontal and posterior components. They interpreted these changes as possible manifestations of a facilitation of episodic memory. To date, there is no fMRI study investigating the effect of sleep on face learning.

This effect remains controversial as another recent study on individual face memory found very small effects of sleep together with a detrimental effect of wake. The authors argued that sleep was not actively consolidating face memory traces but slightly protecting them from being forgotten (Sheth et al., 2009). Thus, work is still needed to clarify these early results on the role of sleep on face learning and investigate the underlying neural correlates of this effect.

To conclude this section, even if the role of sleep in memory consolidation is can be controversial (Siegel, 2001), increasing evidence show that specific phenomenon occurs during this period. Sleep may be an optimal state for the consolidation of memory traces and their integration into existing networks, allowing the brain to constantly adapt to its environment.