UNIVERSITY OF AMSTERDAM
Procedural memory
consolidation of motor
coordination during a
nap
Netherlands Institute for Neuroscience Royal Netherlands Academy Of Arts And Sciences
Department of Sleep and Cognition Internship report – 41 EC Supervisor: Ysbrand van der Werf
UvA representative/co-assessor: Winni Hofman Annemarie Tuominen, 10256849
16.01.-16.08.2012, 20.10.2012
TABLE OF CONTENTS Abstract 1 Introduction 1 Methods 3 Subjects 3 Procedure/Experiment outline 4 Task 4 Sleep recordings 6 Analyses 6 Statistical analyses 7 Results 8 General 8
Napping and performance 10
Correlations with slow waves and spindles 11
Discussion 17
Learning to perform 17
Relationship between general learning processes
and sleep stages 18
Two sides of having a nap 18
The possible contribution of a nap to a better
performance 19
Conclusion and future direction 20
FIGURES AND TABLES
Figure 1. A K-complex and a sleep spindle example 2 Figure 2. One-button and two-button presses on the modified
SISL task 3
Figure 3. Schematic of the experiment procedure 4
Figure 4. An outline of the finger tapping task used in this
experiment 5
Figure 5. The order of the fixed and the random time interval
conditions in the task 6
Figure 6. Selection of 72 channels from 128- and 256-channeled
EEG caps 7
Figure 7. Settings for spindle detection 8
Figure 8. Settings for slow wave detection 8
Figure 9. Topographical view of the spindle and slow wave
distribution on the scalp during sleep 10 Figure 10. Mean behavioural performance across all trials
before/after sleep in fixed and random time interval
conditions 12
Figure 11. Mean behavioural performance across all trials before/after a rest in fixed and random time interval
conditions 13
Figure 12. A summary of mean behavioural performance between the sleep group and the wake group before
and after a sleep/rest 14
Figure 13. Reaction time of the sleep group and the wake group before and after a sleep/rest in fixed and random
time interval conditions 15
Figure 14. Steepness of the learning curves against spindle
density in random interval 16
Figure 15. Mean spindle power against steepness of the
learning curves 16
Table 1. Ranking of finger combinations 9
Procedural memory consolidation of motor coordination
during a nap
Abstract
Motor behavioural performance improvement has been linked to sleep in many studies. A positive correlation between motor task improvement (motor
learning) and the procedural memory consolidation after the sleep has been observed, but relatively little evidence exists for the effects of short-duration sleep on motor learning. In this study we investigated a short sleep (90 min) effect on motor learning by using a novel finger tapping task that requires anticipation, while measuring the EEG signals throughout the experiment. We discovered napping to have a positive effect on behavioural performance (t-test, p = 0.04).
Acknowledgments
The author would like to thank Ysbrand van der Werf, Ilse Verweij and Yoshiyuki Onuki for excellent supervision and guidance, the Sleep and Cognition group with special thanks to Jeroen Benjamins and Nico Romeijn, and fellow interns.
_____________________________________________________________________________________________ Introduction
It is common knowledge, due to many studies in the past, that sleep is an integral part of learning in consolidating previously acquired memories (Stickgold, 2005; Rasch et al., 2007; Gais and Born, 2004). Be it motor, social, cognitive or
behavioural type, having a good sleep promotes subsequent performance overall (Turner at al., 2007; Born et al., 2006; Walker et al., 2005). When it comes to motor coordination, studies have shown that best behavioural performance is achieved when a sufficient amount of sleep has occurred, allowing the synaptic networks to reconstruct and integrate the newly learned information of previous experiences, later facilitating memory consolidation post-sleep when performing the motor task again (Kuriyama et al., 2004; Philal and Born, 1997). Moreover, recent papers have further shown that there is a correlation between different sleep stages and the post-learning reprocessing of different categories of memory (Fogel et al., 2006). For example, stage 2 (S2) sleep is associated with simple motor tasks (Smith et al., 2004), while rapid eye movement (REM) sleep has been hypothesized to be involved with tasks that require a new cognitive strategy. With the possibility of categorizing different sleep stages to types of memory consolidation post-learning, it becomes interesting and challenging to investigate the relationship of various sleep stages with different tasks to identify possible specific relationships between sleep and memory function. This paper will focus on S2 sleep.
One of the hallmarks of S2 sleep is sleep spindles. Spindles by definition oscillate in the 11-17 Hz range and are characterized as waxing-and-waning rhythmic waves of amplitude lasting between 0.5-2 seconds (Gibbs and Gibbs, 1950). The functional role of spindles remains unclear although indications of their role in memory consolidation have been shown (Schabus et al., 2004). A study by Fogel et al. (2006) showed that the overall spindle count in S2 sleep increased as performance of a task improved, which can be considered as an indication of motor consolidation. Another study by Morin et al. (2008) confirms this in an experiment where NREM sleep characteristics were investigated. Changes in NREM sleep following motor learning were specific to consolidation and learning.
Another significant feature of sleep to study and analyse during any sleep experiments are slow waves (SW). SW sleep is categorized as stages 3-4, where the waves are typically seen in the 0.5-4 Hz range. In S2 sleep distinct large amplitude waveforms called K-complexes are also considered a type of slow wave. A K-complex consists of a brief negative high-voltage peak, usually greater than 100 µV, followed by a slower positive complex around 350 and 550 ms, and at 900 ms a final negative peak emerges (Fig. 1). In this paper, K-complexes will be considered as slow waves, as seen in S2 sleep data.
Figure 1. A K-complex and a sleep spindle example. A representation of a typical K-complex and a sleep spindle occurring in stage 2 sleep over the course of time; time (ms) shown on the x-axis and the voltage (µV) on the y-axis.
In a paper of Walker et al. (2002), subjects’ performance in a finger tapping sequence task was positively correlated with amount of S2 sleep in the last quarter of the night. A variant of the finger tapping task is the Serial Interception Sequence Learning (SISL) task where subjects are required to precisely time motor responses to intercept moving cues (Sanchez et al., 2010). The current task implemented in this study was slightly modified from the original SISL, mainly to increase the difficulty of completing it successfully; whereas in the original task the subjects used only one finger key presses and the cues appeared at a constant time interval, in this version, besides having one finger key presses, subjects also had two finger key presses, and the timing of the cues were altered
between a fixed and a random time interval (Fig. 2). Because the SISL task demands better capture of real-world skills in which responses are not made as soon as possible but are instead timed to relevant cues in the environment (Gobel et al., 2011), we decided to use it in this experiment. One of the many research questions one can address using the SISL task is whether sleep helps in consolidating only specific timed sequences of motor actions or in generalising the visuo-motor integration. To distinguish between these two scenarios, a fixed time interval condition was implemented to investigate consolidation of only specific timed sequences of motor actions, whereas a random time interval condition was used for generalising the visuo-motor integration. Instead of having a full night’s sleep, this experiment focused on a daytime nap to
investigate short sleep’s influence on memory consolidation and learning. This study set out to investigate if performance is sensitive to naps, and whether there is a performance correlation with spindles and/or slow waves. We expected a correlation between the performance and a 90 minute sleep, specifically more so for the fixed time interval condition. In more detail we expected a correlation between performance and spindle/slow wave density.
Figure 2. One-button and two-button presses on the modified SISL task. The current study implemented two-button presses to the SISL task pictured on the right that requires subjects to simultaneously press two assigned keys on a keyboard when the cue markers match the target markers.
Methods
Subjects
26 subjects (16 female) were recruited under the following criteria: aged
between 18 and 35 years old, right handed, slept at least between 12:00 am and 06:00 am on regular weekdays without having troubles sleeping in an unknown environment, do not play rhythmic games (e.g. Guitar Hero, Patapon, Dance
Dance Revolution) or instruments such as the piano, organ, drums or guitar. All subjects were screened for a normal sleeping pattern using Athens Insomnia Scale (mean 1.6, 1.8), Pittsburgh Sleep Quality Index (mean 2.5, 1.7), Sleep Diagnosis List (mean 1.9, 0.9) and Epworth Sleepiness Scale (mean 3.5, 2.8). Keeping a sleep diary and continuously wearing an Actiwatch (a device that uses an accelorometer to detect and log wrist movement, Philips) was required for a minimum of 3 days prior to the experiment day. Participants were excluded from the experiment on the basis of irregular sleeping patterns and apparent
substance abuse that would affect the cognitive performance and possibly the quality of sleep. Subjects who spent >50% of the nap time sleeping and had >20% of S2 sleep were categorized in to the sleep group. After eliminating the outliers based on criteria explained above, the remaining subjects were further divided in to a sleep group (11 subjects, 6 female) and wake group (7 subjects, 5 female). The experiment was approved by the ethical committee of the
Department of Psychology, the University of Amsterdam, the Netherlands.
Procedure
Experiment outline
A general outline of the experiment was explained to the subjects beforehand, which was to perform task 1, have a nap on a normal 200 x 90 cm bed or have a rest of 1.5 hours, followed by task 2 and a questionnaire (Fig. 3). Before starting task 1, an EEG cap (EGI) was placed on the subjects’ heads. For the sleep group, two more electrodes were attached underneath the chin to record muscle activity during the nap. All lights were turned off and the surroundings were kept as silent as possible for the sleep group. During the nap, EEG waves were recorded for scoring the sleep stages. After 1.5 hours the subjects were woken up, had a 10-20 min break to recover from the nap, and were asked to do task 2. For the wake group, subjects kept the EEG cap on their heads during their rest of 1.5 hours between tasks, and also had a 10-20 min break before starting task 2. For both groups, a questionnaire about the difficulty of finger coordination to perform the task was given after completing task 2.
Figure 3. Schematic of the experiment procedure. An overview of the experiment protocol where different stages of the experiment with allotted durations are represented, also showing the EEG recording time lasting throughout the experiment.
Task
The finger tapping task is based on the serial interception sequence learning task (Gobel et al, 2011) where subjects are required to intercept vertically upward moving cues from the bottom of the screen towards the target cues at the top of the screen by pressing corresponding target cues when the moving cues
intercept at a constant speed (Fig. 4A). The target cues consisted of a row of four circles of 120 pixel diameter, equally apart from one another (32 pixels), placed on the top of a 17 inch computer screen at a distance of 40 cm from the subject. The moving cues lined up vertically beneath the targets, and were of 10% smaller size than the targets for an easier visualization of overlap between the target and the moving cue. From left to right, the targets were assigned to four horizontal keys of a computer keyboard and subjects needed to press with their small (A), ring (B), middle (C) and index (D) fingers respectively using their left, non-dominant, hand when the moving cues overlapped with the target markers (Fig. 4B). Keys needed to be pressed either individually (one key press) or
simultaneously (two key presses at the same time). The moving cues appeared in a fixed order (C-D-AB-CD-BC-A-BD-AD-B-AC) with either a fixed interval (0.43-0.45-0.37-0.50-0.47-0.37-0.47-0.50-0.37 seconds, mean 0.44 seconds) or a random interval (mean 0.44 seconds) of key presses. The random interval is based on the shuffled intervals used in the fixed interval condition. Task 1 consisted of 21 trials of blocks of rest, fixed order and fixed interval, rest, fixed order and random interval, and rest. Task 2 consisted of 21 trials of blocks of rest, fixed order and random interval, rest, fixed order and fixed interval, and rest. Each block lasted for 30 seconds, and the fixed sequence of moving cues repeated itself 5 times every block. To reject the possibility of the order effect on each conditions, the order of these two conditions were alternated across
subjects (Fig. 5). Tasks were carried out in a separate room from the experiment control room so that the experimenters did not interfere with the subjects’ performance.
Figure 4. An outline of the finger tapping task used in this experiment. (A) The cue markers move upward from the bottom of the screen towards the top of the screen where the target markers are. The red arrow points to a situation where a cue marker overlaps with the target marker, indicating the moment for the subject to press a correct key on the keyboard. (B) Assigned fingers for the finger tapping experiment. Subjects used their left non-dominant hand in the
experiment where A= small finger, B = ring finger, C = middle finger and D = index finger.
Figure 5. The order of the fixed and the random time interval conditions in the task. Fixed and random order of markers alternate with rest conditions, each block lasting for 30 seconds. Task 1 and 2 were alternated across subjects to prevent the order effect on each conditions.
Sleep recordings
All subjects had their head measured to determine the middle reference point of a 256-channeled or 128-channeled EEG cap. The EEG cap was then placed on the subject’s head, and each channel cup was filled with Spectra 360 electrically conductive gel and tested for a strong signal transmission. EEG signals were amplified with an EGI amplifier and digitized at a rate of 500 Hz. A hardware highpass filter of 0.01Hz and a bandpass filter of 50 Hz for visualization purposes on the monitor display were applied.
Analyses
Behavioural performance of the task was measured as the mean performance of the last 3 trials before sleep/rest and the first 3 trials after sleep/rest as
successful key presses over a total amount of key presses. Steepness of the learning curve was measured as individual steepness of linearly fitted learning curves after sleep. Sleep stage scoring was done according to the standards of the American Association of Sleep Medicine (Iber et al., 2007), and to verify the reliability of sleep scoring, an independent scorer also scored the sleep data. The interrater agreement over the sleep stage scores was a correlation of 0.89 that was considered sufficient for the purposes.
Since both 256 and 128-channel caps were used in the experiment, a selection of 72 channels was picked to represent the common channels from both caps (Fig. 6). This was done using the guide of Luu and Ferree (2000) in which the common electrode positions for 128 and 256 were acquired. S2 sleep spindles were filtered within the range of 11-17 Hz and detected using the Hilbert transform (Hilbert, 1953) with 2 criteria: spindle detection threshold was +3 standard deviations (SD) of spindle power and the duration above threshold longer than 0.5-1 seconds (Fig. 7). Spindle density was calculated as the number of spindles over the duration of S2 sleep in percentages times the total naptime per subject. Spindle power was calculated through a Matlab algorithm using 50% overlapping Hanning windows of two seconds. Slow waves were filtered within the range of 0.5-4 Hz and detected applying the method used in Massimini et al. (2004) with 3 criteria: slow wave detection threshold was less than -40 V, peak-to-peak
amplitude 75 V and the slow wave zero-crossing between 0.2-1 seconds (Fig. 8). Topographical representations of both spindle density and slow wave density were calculated as an average from S2 duration while using the entire scalp as one region of interest in Matlab with the FieldTrip toolbox.
Statistical analyses
Statistical analyses were done using t-tests and Pearson’s correlation coefficients in Matlab (v7.13, MathWorks) in the following manner: comparisons of
performance between the fixed and the random time interval conditions before and after a nap/rest in both subject groups, and the reaction time between the fixed and the random time interval conditions before and after a nap/rest in both subject groups were done using a paired-sample t-test. Correlations of
performance with sleep parameters spindles, slow waves and duration of S2 sleep were done using Pearson’s correlation coefficient. Significance was considered when p ≤ 0.05. We used directed t-tests for the separate groups instead of ANOVAs because the data analysis was incomplete and therefore
performed on a limited number of participants.
Figure 6. Selection of 72 channels from 128- and 256-channeled EEG caps. 72 channels from both caps were picked to represent the common channels in order to have a uniform representation for analyses using a guide for the electrode position and their 10-10 international equivalents. Red dots mark the anterior, green left, purple posterior and blue right side of the scalp.
Figure 7. Settings for spindle detection. The raw sleep spindles (top row) were filtered in the range of 11-17 Hz (second row) and detected using the Hilbert transform (third row) while the detection threshold was set to 3 standard errors of the spindle power (fourth row). Spindle duration above the threshold needed also to be longer than 0.5-1 seconds.
Figure 8. Settings for slow wave detection. The slow waves were filtered in the range of 0.5-4 Hz and detected using 3 criteria: A = slow wave detection was less than -40 µV, B = peak-to-peak amplitude 75 µV, and C = zero-crossing between 0.2-1 seconds.
Results
General
The questionnaire subjects filled out at the end of the experiment concerned the difficulty of different finger combinations. By asking them to rank on a scale of 1-10 the easiest (1) and the most difficult (1-10) combination of used fingers, a general feedback was acquired that can be used in the future for developing either an easier or a more difficult task. Most subjects ranked the use of the index finger as easiest, and the use of the combination of middle and small fingers most difficult (Table 1). On average a difficulty rank of 4-5 was given to middle, ring, small fingers and to a combination of index and middle fingers. An average of rank 5 was given to index and ring, index and small, middle and ring, and ring and small finger combinations. Sleep scoring was done individually per subject and a general overview of the duration of sleep stages during the nap is
illustrated in table 2. Sleep spindle density and slow wave density during the nap were measured and plotted topographically to show distribution over the scalp. Spindles were concentrated frontal-centrally whereas slow waves were
predominantly frontally located (Fig. 9). This distribution was in accordance with previous literature (De Gennaro et al., 2000).
Table 1. Ranking of finger combinations. Subjects ranked their experiences on the difficulty (top row, 1 = easy, 10 = hard) of the finger combinations (left column) in the task. Numbers in the table represent subjects (e.g. five subjects ranked the index finger with difficulty level 1). Majority found using the index finger easiest and the combination of middle and small finger most difficult (highlighted). Mean values represent the rank of difficulty.
Table 2. Mean duration of sleep stages. Amount of sleep in percentages spent in each sleep stage of the nap (90 minutes) is shown (mean ± SD).
Figure 9. Topographical view of the spindle and slow wave distribution on the scalp during sleep. The average density of the sleep spindles and slow waves during the 90 minutes of sleep is focused frontal-centrally and frontally respectively.
Napping and performance
Performance was plotted across all trials and subjects in fixed and random time interval conditions before and after a nap/rest in the sleep group and in the wake group. The sleep group’s performance improvement in the fixed time interval condition was moderately significant (t-test, p = 0.04) similar to the random time interval condition (t-test, p = 0.03. Fig. 10). Wake group’s performance
improvement in the fixed time interval condition was not significant (t-test, p = 0.83) neither was there significance in the random time interval condition (t-test, p = 0.23. Fig. 11, 12). Reaction time was also measured as another indicator of overall performance. Reaction time in the sleep group was significantly higher after the nap compared to before the nap (t-test, p = 0.04 fixed and random interval). Reaction time in the wake group was similarly higher after the rest compared to before the rest (t-test, p = 0.04 fixed interval, p = 0.03 random interval. Fig. 13).
Correlations with slow waves and spindles
To find out more about the experiment and learning, correlations using spindles and slow waves were calculated. There was no correlation with performance and duration of S2 sleep (Pearson’s, p = 0.07 fixed interval, p = 0.84 random interval). With respect to performance and number of spindles, there was no significant relationship between these two (Pearson’s, p = 0.84 fixed interval, p = 0.81
random interval). Spindle density against the mean steepness of learning curves after sleep showed no significance in fixed time interval after the nap (Pearson’s, p = 0.55), but in the random time interval a significant relationship was observed (Pearson’s, p = 0.02) (Fig. 14). It would seem that better learners had higher spindle densities. Mean spindle power against mean steepness of learning curves after sleep showed no significance in fixed time interval after the nap (Pearson’s, p = 0.88), but in the random time interval a significant relationship was observed (Pearson’s, p = 0.04) (Fig. 15). For slow waves, no significance was found in correlations between mean slow wave power and mean steepness of learning curves after sleep (t-test, p = 0.94 fixed interval, p = 0.74 random interval), or between performance and the number of slow waves (Pearson’s, p = 0.72 fixed interval, p = 0.77 random interval).
Figure 10. Mean behavioural performance across all trials before/after sleep in fixed and random time interval conditions. (A) Mean behavioural performance was measured in the fixed time interval condition across all trials before and after sleep. A significant increase in performance (t-test, p = 0.04) is observed after the sleep. (B) Mean behavioural performance was measured in the random time interval condition across all trials before and after sleep, and a significant increase in performance (t-test, p = 0.03) was also observed after the sleep. Errorbars represent standard deviations.
Figure 11. Mean behavioural performance across all trials before/after a rest in fixed and random time interval conditions. (A) Mean behavioural performance was measured in the fixed time interval condition across all trials before and after a rest. No significance was observed after the rest. (B) Mean behavioural performance was measured in the random time interval condition across all trials before and after a rest. No significance was observed after the rest. Errorbars represent standard deviations.
Figure 12. A summary of mean behavioural performance between the sleep group and the wake group before and after a sleep/rest. (A) Bar graphs in the sleep group show a significant increase in task performance after the sleep in both time interval conditions (t-test, p = 0.04 fixed interval, p = 0.03 random interval). (B) Bar graphs in the wake group do not show significant increase in task performance after a rest in either time interval conditions. Errorbars represent standard deviations.
Figure 13. Reaction time of the sleep group and the wake group before and after a sleep/rest in fixed and random time interval conditions. (A) Reaction time in the sleep group was measured across all trials before and after a sleep and found to be significantly higher after the sleep in both time interval conditions (t-test, p = 0.04). (B) Reaction time in the wake group was measured across all trials before and after a rest and also found to be significantly higher after the rest in both time interval conditions (t-test, p = 0.04 fixed interval, p = 0.03 random interval). Standard deviation errorbars are omitted for clarity.
Figure 14. Steepness of the learning curves after sleep against spindle density in random interval condition. The steepness of the mean individual learning curves were correlated with the density of sleep spindles in the random time interval condition, and found to be significantly correlated (r = 0.68, p = 0.02).
Figure 15. Mean spindle power against steepness of the learning curves after sleep. Mean sleep spindle power across the scalp was correlated against the steepness of the mean individual learning curves in the random time interval condition, and a significant correlation was found (r = 0.61, p = 0.04).
Discussion
We have here demonstrated napping to have a positive effect on behavioural performance when tested on a novel finger tapping task that requires
anticipation. Based on our results we can confirm the hypotheses of the correlation between the performance and a 90 minute sleep, and between performance and spindle density in the random time interval condition. The hypothesis of a correlation between performance and slow wave density is rejected. According to this, sleep appears to be beneficial for both specific timing and for visuo-motor integration.
Learning to perform
In the finger tapping task, the motor response is initiated as quickly as possible after first identifying the velocity of the moving cues and the distance from the target markers. After this, planning of motor action takes place, which is
integrated with timing to complete the action, resulting in key presses when cue markers overlap the target markers. By making response timing integral to successful task performance, it was hoped that timing information would become fully integrated with the representation of the repeating action sequence, which happened in the fixed time interval condition and in the random time interval condition in the sleep group, but did not happen in the wake group under any conditions. Hence adaptation was observed in the performance with the sleep group in fixed time interval as the subjects learned to anticipate cue markers, seen as a performance increase, but simultaneously some adaptation could be observed from the random time interval condition (Fig. 12). As we did not expect to see learning in the random time interval condition assuming one cannot learn the unexpected, it is curious to see the small behavioural performance increase in this condition. The reason behind this could be that the upwards scrolling cues prepare the subjects to press correct keys much like in the fixed condition, and perhaps the randomness of the time interval was not random enough. On the other hand, this small significance in the random time interval condition would speak for the visuo-motor integration, which in fact is a skill, thus speaking for sleep helping to establish said skill memory (Maquet et al., 2003). Graybiel (1998) did propose a theory where shorter interval sequences would be learnt as “performance units”, “chunks” for motor planning purposes, so that learning only two-three responses at a time would help in the integration of timing and motor execution. Perhaps subjects learnt to mentally group the cues in smaller units thus making it easier to time the responses.
Alternatively, differences between the fixed and the random time interval
conditions in respect to behavioural performance in the sleep group may suggest an interference from already learnt motor sequences to the execution of novel motor sequences: perhaps the motor system is preparing an incorrect action at a specific time during the task, resulting in decreased processing efficiency, since knowing the timing but not the motor sequence order does not facilitate
preparation of the appropriate motor movements (O’Reilly et al., 2008; Sin and Ivry, 2002). This could explain why behavioural performance was not
The reaction times that were considered as another way to view performance do not differ between the sleep and the wake group or between the interval
conditions, indicating that reaction time is not good enough of a measure in this study to draw conclusion in respect to the behavioural performance increase between these two groups. Having a rest or a nap however had an influence, albeit a somewhat unexpected one: the mean reaction time increased over time as the number of trials increased. Interestingly, performance after the nap/rest did not decrease or increase while the reaction time increased. This could be explained by general experiment fatigue or by residual fatigue from the nap/rest, and even perhaps by taking more time to focus on the task at hand.
Relationship between general learning processes and sleep stages
Learning is an acquisition of a skill which requires practice and is achieved when an observed behaviour has changed due to experience or practice. Zimbardo and Gerrig (1999) have proposed a model for information processing including the abovementioned experience/practice, which has 4 components: processing speed, breadth of declarative knowledge, breadth of procedural skill and processing capacity. The latter is important for procedural memory because through processing one stores procedural memory, as they are accessed and used without the need for conscious control or attention. Literature shows that
formation of all types of memory is greatly enhanced during sleep (Walker et al., 2002; Robertson et al., 2004; Huber et al., 2004). But it has also been suggested that not all stages of sleep are sufficient to improve procedural memory – Vertes and Eastman (2000) made a point in REM sleep not playing a major role in processing or consolidating said memory, also supported by Siegel (2001) who concluded that existing literature does not indicate a major role for REM sleep in said memory consolidation. Tucker and Fishbein (2009) instead showed that amount of S2 sleep is important for optimal motor memory processing. Since S2 sleep is associated with simple motor tasks, and since the finger tapping task is largely about motor learning, having subjects sleep briefly (nap) should elicit enough S2 sleep with spindles to see if a correlation between motor learning and S2 sleep exists when sleep time is only 90 minutes.
Two sides of having a nap
A multitude of research has investigated the effects of napping and has
consistently demonstrated that naps can counteract the effects of sleepiness by enhancing subjective and objective alertness, improving cognition, vigilance and psychomotor ability (Lovato and Lack, 2010).Napping has positive effects such as refreshed mood and improved performance levels, but also negative effects like sleep inertia (where one feels groggy, disoriented and even more sleepy than before the start of the nap) or impaired alertness usually experienced upon waking (Stampi, 1992). These effects are affected by the time of the day, duration of the nap and prior wakefulness (Naitoh, 1981) and can have an impact on the acquired results of this experiment.
In a related fashion, circadian rhythms dictate when most of us sleep. Because of this, and to measure learning before and after sleep, researchers often
manipulate sleep itself (Karni et al., 1994), the time of learning relative to sleep (Huber et al., 2004) or a combination of both (Fischer et al., 2002). This usually
means depriving subjects of sleep, which makes discriminating between the unfavourable effects of sleep deprivation on learning and the beneficial effect of sleep on learning difficult. Alternatively, allowing them to sleep normally and measure learning at particular times of the day makes controlling the possible circadian effects difficult. Cajochen et al. (2004) have shown that sequence learning (using a serial reaction time task) is modulated by circadian phase, emphasising the need for control of circadian effects. Perhaps in the current study timing of the nap was off in the sense that the circadian rhythm had a negative influence in observing a clear learning pattern with the subjects across all tests (reaction time, performance before and after sleep, spindle and slow wave correlations). If spindles reach their peak density late in the night, as suggested by De Gennaro et al. (2000), then the circadian rhythm could interfere with obtaining proper S2 sleep. Then again, the subjects seemed to continue to learn despite of their performance not improving, visible in the almost-plateaued performance curve after the nap both in the fixed and random time interval conditions. This has also been shown in the paper of Keisler et al. (2007) in which learning (using the alternating serial response time task) occurred during evenings, but was not expressed in performance.
Curiously enough, a notion of a postlunch dip could be implemented to these results as well. It is known that the effects of it can be intensified with a heavy high-carbohydrate lunch (Craig et al., 1981), but it can occur even when no lunch is consumed (Monk et al., 1996). Many studies of performance have reported a short-lived decrease of performance during the midafternoon hours (Carrier et al., 2000) so perhaps the lack of increase in performance in the current study occurred in the same time window that is customary for the postlunch dip to occur.
The possible contribution of a nap to a better performance
To further investigate any relationships, we correlated the number of spindles (spindle density) and spindle power with mean steepness of the learning curve and noticed a correlation in the random time interval condition after the nap. This indicates that the learning increases together with the spindle
density/power after having a nap, and that the nap helps performance in the subsequent random time interval condition task. Because the random interval is representative of visuo-motor integration, it would seem that the nap facilitated information processing, such that subjects could perform better during the task immediately following the nap. Sleep appears to provide spindle activity to process recently acquired memory traces, but at the same time spindles are not related to memory performance in general (Schabus et al., 2004), which could explain why there was no correlation in the fixed time interval condition.
Perhaps the number of spindles is determined by one’s general ability on the task and not related to consolidation of the task. We did not see any correlations in SW sleep and therefore cannot conclude whether or not SWs played any part in memory consolidation or behavioural performance. We are reluctant to say that the lack of SW correlation indicates lack of learning because this study used a very limited number of subjects which might obscure a possible effect, and because we only used S2 in our SW analyses. In general, slow wave activity after learning correlates with improved performance of a motor task after sleep, as
shown by Huber et al. (2004).
Naps are informative when an effect of sleep is observed, but the modest result we show here is difficult to interpret: is it that sleep plays no role in learning or that a full night's sleep is necessary to gain benefit? Standard nap experiments consist of a 90 minute sleep time, which is also considered to be the length of a normal sleep cycle (Billiard, 2003), but most of the participants did not
experience a full cycle of sleep with all the sleep stages. These results are not significant enough to conclude napping to have a positive impact on motor learning, but would give a small indication in that direction. Perhaps the overall napping time of 90 minutes was too long; reports have said that 20-min power naps are most effective when it comes to revitalising and performing better in motor tasks (Dhand and Sohal, 2006; Milner et al., 2006). Entering deep, slow-wave sleep and failing to complete the normal sleep cycle can result in sleep inertia. On the other hand, the subjects in this study did have a recovery period of 10-20 minutes after waking up from the nap which would eliminate sleep inertia or reduce it greatly. It has been shown that in order to obtain optimal post-nap performance, a power nap should be limited to the beginning of a sleep cycle, as waking up from deeper sleep stages worsens performance (Bruck and Pisani, 1999; Frey and Wright, 2007). Other studies however show the opposite: Hayashi et al. (1999, Hayashi and Hori, 2008) say that a 20-min nap did not improve task performance (computer tasks). These papers show that earlier sleep might play a role, i.e. arousal level being already high due to a well slept previous night. Also, immediately following long naps (e.g. 2h) performance can actually decline for a period with eventual improvements that can last up to 24h (Lovato and Lack, 2010). Furthermore, despite of the exclusion criteria, subjects in this study may differ in degree of fatigue after the nap (sleep inertia) that might have an effect on the results. One must also keep in mind that performance combines learning and other factors such as motivation, fatigue and attention (Matthews et al., 2000).
Conclusion and future direction
Motor learning and sleep have been tightly linked together. A growing body of evidence is showing that in order to perform better in any given motor task, adequate amounts of sleep must be had (Turner at al., 2007; Born et al., 2006; Walker et al., 2005references). As previously mentioned, S2 sleep has been shown to be linked with simple motor tasks and spindle count with motor task performance improvement. Familiar sequences of motor actions that are used in everyday life frequently depend on accurate timing between movements. This is clear in expertly trained behaviours such as sports and music performance as well as in more basic processes like walking and speaking. The results reported here provide some preliminary evidence that 1) a nap may help consolidate learning new motor sequences and 2) that S2 sleep facilitates and prepares for subsequent motor memory performance and information processing where timing is important to the task.
For the future, we suggest repeating this experiment with more subjects equally distributed between the two experimental conditions to obtain a better idea of the effect of a nap, and also to minimize the effect of having many subjects spending much of the napping time lying awake. We also suggest the physical
experiment conditions to be more stable in the sense that the room for napping should be isolated from external noises. With every aspect taken care of, we see no reason why napping alone could not produce behavioural performance increase in motor tasks. The time for incorporating naps in to everyday life has come – maybe we can learn from siestas enjoyed in southern countries.
REFERENCES
Billiard, M. (2003). Sleep: physiology, investigations and medicine. Springer, p. 5 Born, J., Rasch, B., Gais, S. (2006). Sleep to remember. Neuroscientist, 12, 410-424 Bruck, D., Pisani, D. L. (1999). The effects of sleep inertia on decision-making performance. J. Sleep Res., 8, 95-103
Cajochen, C., Knoblauch, V., Wirz-Justice, A., Kräuchi, K., Graw, P., Wallach, D. (2004). Circadian modulation of sequence learning under high and low sleep pressure conditions. Behav. Brain Res., 151, 167–176
Carrier, J., Monk, T. M. (2000). Circadian rhythms of performance: new trends.
Chronobiol. Int., 17, 719-732
Craig, A., Baer, K., Diekmann, A. (1981). The effects of lunch on sensory-perceptual functioning in man. Int. J. Occup. Environ. Health, 49, 105–14 De Gennaro, L., Ferrara, M., Bertini, M. (2000). Topographical distribution of spindles: variations between and within NREM sleep cycles. Sleep Res. Online, 3, 155-160
Dhand, R., Sohal, H. (2006). Good sleep, bad sleep! The role of daytime naps in healthy adults. Curr. Opin. Pulm. Med., 12, 379-382
Fischer, S., Hallschmid, M., Elsner, A.L., Born, J. (2002). Sleep forms memory for finger skills. Proc. Natl. Acad. Sci., 99, 11987–11991
Fogel, S. M., Smith, C. (2006). Learning-dependent changes in sleep spindles and stage 2 sleep. J. Sleep Res., 15, 250-255
Frey, D. J., Wright Jr, K. P. (2007). Sleep inertia: Performance impairments upon awakening from sleep. The influence of sleep inertia on human performance: individual differences, aging, hypnotics, cognition and balance, 1.
Gais, S., Born, J. (2004). Declarative memory consolidation: mechanisms acting during human sleep. Learn. Mem., 11, 679-685
Gibbs, F. A., Gibbs, E. L. (1950). Atlas of Electroencephalography. Addison-Wesley Press, Cambridge
Graybiel, A. M. (1998). Basal ganglian and chunking of action repertoires.
Neurobiol. Learn. Mem., 70, 119-136
Hayashi, M., Hori, T. (1998). The effects of a 20-min nap before post-lunch dip.
Psych. Clin. Neurosci., 52, 203-204
sleepiness, performance and EEG activity. Int. J. Psychophys., 32, 2, 173-180 Hilbert, D. (1953). Grundzüge einer allgemeinen Theorie der linearen Integralgleichungen. Chelsea Pub. Co.
Huber, R., Ghilardi, M. F., Massimini, M., Tononi, G. (2004). Local sleep and learning. Nature, 430, 78-81
Iber, C., Ancoli-Israel, S., Chesson, A. L., Quan, S. F. (2007). AASM Manual for the Scoring of Sleep and Associate Events. Rules, terminology and technical
specifications. Am. Ass. Sleep Med., Westchester, IL
Karni, A., Tanne, D., Rubenstein, B. S., Askenasy, J. J. M., Sagi, D. (1994).
Dependence on REM sleep of overnight improvement of aperceptual skill.
Science, 256, 679–682
Kuriyama, K., Stickgold, R., Walker, M. P. (2004). Sleep-dependent learning and motor-skill complexity. Learn. Mem., 11, 705-713
Lovato, N., Lack, L. (2010). The effects of napping on cognitive functioning. Human sleep and cognition, part 1: basic research. Pro. Brain Res., 155-166 Luu, P., Ferree, T. (2000). Determination of the geodesic sensor nets’ average electrode positions and their 10-10 international equivalents. Electrical Geodesics Inc.
Maquet, P., Schwartz, S., Passingham, R., Frith, C. (2003). Sleep-related
consolidation of a visuomotor skill: brain mechanisms as assessed by functional magnetic resonance imaging. J. Neurosci., 23, 1432-1440
Matthews, G., Davies, R. D., Westerman, S. J., Stammers, R. B. (2000). Human performance: Cognition, stress and individual differences. Taylor and Francis, Philadelphia, PA
Milner, C. E., Fogel, S. M., Cote, K. A. (2006). Habitual napping moderates motor performance improvements following a short daytime nap. Bio. Psych., 73, 141-156
Monk, T. H., Buysse, D. J., Reynolds, C. F. et al. (1996). Circadian determinants of the postlunch dip in performance. Chronobiol. Int., 13, 135–45
Morin, A., Doyon, J., Dostie, V. et al. (2008). Motor sequence learning increases sleep spindles and fast frequencies in post-training sleep. Sleep, 31, 1149-1156 Naitoh, P. (1981). Circadian cycles and restorative power of naps. Johnson, L. C., Tepas, D. I., Colquhoun, W. P., Colligan, M. J. (eds), Biological Rhythms, Sleep and Shift Work. Spectrum, New York, 553-580
temporal and ordinal structure of movement sequences in incidental learning. J.
Neurophysiol., 99, 2731-2735
Philal, W., Born, J. (1997). Effects of early and late nocturnal sleep on declarative and procedural memory. J. Cog. Neurosci., 9, 534-547
Rasch, B., Buchel, C., Gais, S., Born, J. (2007). Odor cues during slow-wave sleep prompt declarative memory consolidation. Science, 315, 1426-1429
Robertson, E. M., Pascual-Leone, A., Press, D. Z. (2004). Awareness modifies the skill-learning benefits of sleep. Curr. Biol., 14, 208-212
Schabus, M., Gruber, G., Parapatics, S. et al. (2004). Sleep spindles and their significance for declarative memory consolidation. Sleep, 27, 1479-1485 Shin, J. C., Ivry, R. B. (2002). Concurrent learning of temporal and spatial sequences. J. Exp. Psychol. Learn. Mem. Cogn., 28, 445-457
Siegel, J. (2001). The REM sleep-memory consolidation hypothesis. Science, 294, 1058-1063
Smith, C., Aubrey, J., Peters, K. (2004). Different roles for REM and stage 2 sleep in motor learning: a proposed model. Psychol. Belg., 44, 81-104
Stampi C. (1992). Why We Nap. Birkhauser, Boston
Stickgold, R. (2005). Sleep-dependent memory consolidation. Nature, 437, 1272-1278
Tucker, M. A., Fishbein, W. (2009). The impact of sleep duration and subject intelligence on declarative and motor memory performance: how much is enough? J. Sleep Res., 18, 304-312
Turner, T. H., Drummond, S. P. A., Salamat, J. S., Brown, G. G. (2007). Effects of 42 hr sleep deprivation on component processes of verbal working memory.
Neurophys., 21, 787-795
Vertes, R. P., Eastman, K. E. (2000). The case against memory consolidation in REM sleep. Behav. Brain Sciences, 23, 793-1121
Walker, M. P., Brakefield, T., Morgan, A., Hobson, J. A., Stickgold, R. (2002).
Practice with sleep makes perfect: Sleep dependent motor skill learning. Neuron, 35, 205–211
Walker, M. P., Stickgold, R., Alsop, D., Gaab, N., Schlaug, G. (2005).
Sleep-dependent motor memory plasticity in the human brain. Neurosci., 133, 911–917 Zimbardo, P. G., and Gerring, R. J. (1999). Psychology and life. (15th ed.).New York, Longman
