Sleep-dependent consolidation of procedural motor memories in children and adults: the pre-sleep level of performance matters

PAPER

Sleep-dependent consolidation of procedural motor memories inchildren and adults: the pre-sleep level of performance matters

Ines Wilhelm,1,2 Maila Metzkow-Meszaros,3 Susanne Knapp3

and Jan Born1,2

1. Department of Neuroendocrinology, University of L�beck, Germany2. Department of Medical Psychology and Behavioral Neurobiology, University of Tbingen, Germany3. Department of Psychology, Martin Luther University Halle-Wittenberg, Germany

Abstract

In striking contrast to adults, in children sleep following training a motor task did not induce the expected (offline) gain in motorskill performance in previous studies. Children normally perform at distinctly lower levels than adults. Moreover, evidence inadults suggests that sleep dependent offline gains in skill essentially depend on the pre-sleep level of performance. Against thisbackground, we asked whether improving children’s performance on a motor sequence learning task by extended training tolevels approaching those of adults would enable sleep-associated gains in motor skill in this age group also. Children (4–6 years)and adults (18–35 years) performed on the motor sequence learning task (button-box task) before and after �2-hour retentionintervals including either sleep (midday nap) or wakefulness. Whereas one group of children and adults, respectively, received thestandard amount of 10 blocks of training before retention intervals of sleep or wakefulness, a further group of children receivedan extended training on 30 blocks (distributed across 3 days). A further group of adults received a restricted training on onlytwo blocks before the retention intervals. Children after standard training reached lowest performance levels, whereas in adultsperformance after standard training was highest. Children with extended training and adults after reduced training reachedintermediate performance levels. Only at these intermediate performance levels did sleep induce significant gains in motorsequence skill, whereas performance did not benefit from sleep in the low-performing children or in the high-performing adults.Spindle counts in the post-training nap were correlated with performance gains at retrieval only in the adults benefitting fromsleep. We conclude that, across age groups, sleep induces the most robust gain in motor skill at an intermediate pre-sleepperformance level. In low-performing children sleep-dependent improvements in skill may be revealed only after enhancing thepre-sleep performance level by extended training.

Introduction

Numerous studies have convincingly demonstrated inadults that sleep after practicing a new motor skill sup-ports the consolidation of these skill memories (Walker,Brakefield, Hobson & Stickgold, 2003a; Fischer, Hall-schmid, Elsner & Born, 2002; Wilhelm, Diekelmann &Born, 2008; Doyon, Korman, Morin, Dostie, Hadj,Benali, Karni, Ungerleider & Carrier, 2009). On a behav-ioral level, this consolidation manifests itself in greatergains in speed of motor performance across retentionintervals containing sleep compared with wakefulness,despite the absence of any training during the retentioninterval (Walker, Brakefield, Seidman, Morgan, Hobson& Stickgold, 2003b; Fischer et al., 2002; Debas, Carrier,Orban, Barakat, Lungu, Vandewalle, Hadj Tahar, Bellec,Karni, Ungerleider, Benali & Doyon, 2010). During earlydevelopment, children acquire a great variety of basicprocedural skills, and this period coincides with intense

and long periods of sleep (Ohayon, Carskadon, Guille-minault & Vitiello, 2004; Campbell & Feinberg, 2009).This makes it tempting to assume that the obviously highcapabilities of learning in children are functionallyrelated to their intense and superior quality of sleep. All themore astonishing is recent evidence that children, unlikeadults, do not show sleep-dependent gains in proceduralmotor skills (Wilhelm et al., 2008; Fischer, Wilhelm &Born, 2007; Prehn-Kristensen, Gçder, Chirobeja, Bress-mann, Ferstl & Baving, 2009), whereas processes ofmemory consolidation appeared to be even accelerated intime during wakefulness (Dorfberger, Adi-Japha &Karni, 2007). These findings agree with studies in youngbirds (zebra finches) learning a song, which likewisefailed to exhibit any improvement in singing the tutoredsong after overnight sleep (Deregnaucourt, Mitra, Feher,Pytte & Tchernichovski, 2005). The mechanisms thatcould explain the lack of benefit for motor skill memoriesfrom sleep in children are at present entirely obscure.

Address for correspondence: Ines Wilhelm, Department of Neuroendocrinology, University of L�beck, Ratzeburger Allee 160, Haus 50.1, 23538L�beck, Germany; e-mail: [email protected]

� 2012 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

Developmental Science (2012), pp 1–10 DOI: 10.1111/j.1467-7687.2012.01146.x

Studies in adults have indicated that the pre-sleepperformance level is one factor determining whethernewly encoded procedural memories benefit from sleepor not (Diekelmann, Wilhelm & Born, 2009). In anocculomotor sequence task, sleep induced significantperformance gains after a 24-hour retention interval onlyin fast learning subjects but not in slow learners (Albouy,Sterpenich, Balteau, Vandewalle, Desseilles, Dang-Vu,Darsaud, Ruby, Luppi, Degueldre, Peigneux, Luxen &Maquet, 2008). On the other hand, in a finger sequencetapping task, sleep-dependent gains in reaction timesoccurred only for the difficult and unusual sequencetransitions where subjects showed slow reaction timesduring learning (Kuriyama, Stickgold & Walker, 2004).In combination these results appear contradictory, inas-much as one of the studies showed sleep benefits only forfast responding subjects and the other only for slowresponses during learning. However, general familiaritywith the task procedures may be a modulating factor, assubjects are less familiar in learning an occulomotorsequence than tapping a finger sequence. Thus, it hasbeen suggested that sleep benefits do not occur consis-tently at the beginning of learning a completely new skilland also when a skill is already highly over-trained,whereas robust sleep-dependent gains occur with inter-mediate levels of task performance (Stickgold, 2009).

Here, we addressed the question whether the lack ofbenefit from sleep for motor sequence memories that hasbeen consistently observed in children in previous studiesmight result from their overall lower familiarity with andperformance speed on these tasks. To this purpose weinvestigated the effects of different performance levelsinduced by varying the amount of pre-sleep training onsleep-dependent motor memory consolidation in chil-dren (4–6 years) and adults using a coarse motorsequence task (i.e. the button-box task). Following pre-vious suggestions (Stickgold, 2009), we hypothesizedthat, independent of age, sleep would preferentially affectmotor sequence memories once an intermediate perfor-mance level is reached by the subject. Due to the greatdifferences in performance levels between adults andchildren, to reach intermediate levels children wereextensively trained on the motor sequence task comparedto controls who received a standard amount of training.By contrast, to induce this intermediate level of perfor-mance in adults, training was restricted to a minimumcompared with the standard training of controls.

Methods

Participants

Thirty-five healthy children between 4 and 6 years(mean € SEM: 5.44 € 0.75 years) and 33 adults(24.80 € 3.96 years) participated in the study. The par-ticipants were recruited via advertisements placed at theuniversity and local kindergardens. Interviews with theparents and children as well as standardized question-

naires, i.e. German versions of the K-SADS-PL (Kauf-man, Birmaher, Brent, Rao, Flynn, Moreci, Williamson& Ryan, 1997) and the Pittsburg Sleep Questionnaire(Buysse, Reynolds, Monk, Berman & Kupfer, 1989)ensured that the children had no behavioral problems,cognitive impairments or sleep disorders. Children aswell as adults had no history of any neurological orpsychiatric disorder and did not take any medication atthe time of the experiment. All subjects were adapted topolysomnographic recordings during a nap preceding theexperiments proper. The study was approved by the localethics committee, and subjects gave written informedconsent before participating. For the children this wasundertaken by a parent. In addition, all children pro-vided verbal assent. To compensate for participating inthe study, children received vouchers for leisure activities(to visit the zoo, the cinema, or the swimming bath), andadults were paid for participation.

Procedure

Subjects in each age group were randomly assigned toone of two experimental groups differing in the amountof training performed before the retention periods (i.e.low and medium amounts of training in adults, andmedium and high amounts of training in children;Figure 1). Each subject participated in a sleep and awake condition with the order balanced across subjects,i.e. half of the subjects started with the wake conditionand were then tested on the sleep condition, for the otherhalf this order was reversed. For children and adults withmedium amounts of training, this training (Tr) was partof the learning phase immediately before the retentioninterval, and consisted of 10 blocks of performance onthe button-box task, with each block including five eight-element sequences. For children with high amounts oftraining, the training was extended by two times 10blocks (Tr +1, Tr +2) which were performed on two dayspreceding the day of the learning phase (in addition tothe 10 blocks performed during the learning phase). Thetwo days of additional training were separated from thelearning phase by one day without any intervention. Foradults with low amounts of training, the training duringthe learning phase was restricted to two blocks on thebutton-box task. By manipulating the amount of train-ing, distinct performance levels could be induced beforesubjects entered the retention intervals of sleep andwakefulness: Children with medium amounts of trainingperformed at the lowest level before the retention interval(‘low performers’) whereas adults with the same amountof training performed best (‘high performers’). Adultswith low amounts of training as well as children withhigh amounts of training both showed intermediatelevels of performance at the end of the learning phase(‘intermediate performers’, Figure 2).

Task performance at the end of 10 blocks of training isdecelerated due to fatigue developing in the course oftraining. Thus, performance on the last blocks of train-

2 Ines Wilhelm et al.

� 2012 Blackwell Publishing Ltd.

ing underestimates the actual performance level (Keisler,Ashe & Willingham, 2007). In order to achieve a moreaccurate baseline for evaluating sleep-dependent perfor-mance gains, our subjects performed on three furtherblocks of the button-box task 30 minutes after thetraining phase of the experimental session (Test). The30-minute delay period was used to fixate electrodes forpolysomnographic recordings. Immediately after the firsttest phase (Test), the retention interval started whichtook �120 min and in which subjects in the sleep con-ditions took a nap. Subjects in the wake conditionsstayed awake during the entire retention interval. Napstook a maximum of 90 min, and subjects were awakenedlatest 30 min before the second test (Retest) in order toavoid any effects of sleep inertia on retrieval perfor-mance. In the wake conditions, the experimenter read

books to the children or played games with them duringthe retention interval, while adult subjects engaged innon-arousing and non-exhausting activities such aswatching TV or playing card games. The Retest includedthree blocks of the button-box sequence. Before both,Test and Retest, subjects rated their actual mood, tired-ness and motivation on 5-point (adults) and 3-pointrating scales (children), respectively.

Memory task

To investigate memory consolidation we used the ‘but-ton-box task’ which is a novel coarse motor sequencelearning task specifically adapted to the motor abilities inchildren. The button box is a white 50 cm · 22 cm · 7 cmbox with eight colored buttons placed on its upper panelin two rows that are consecutively flashed up accordingto a repeating eight-element sequence. Subjects wereinstructed to press the button flashing up as fast aspossible. Pressing the respective button turned off theillumination, and the next button flashed up immediatelyafterwards. Each block during training and both testphases consisted of five sequences with a 20-sec breakbetween blocks during which the subject received feed-back on his ⁄ her individual performance level during thepreceding block. This was done by informing the chil-dren verbally whether their performance was better orworse than on the block before. For adults, mean reac-tion time was displayed on the screen. The individualmean reaction times per block were also used for

Figure 1 Experimental design: Children performed on thetask either with standard, i.e. medium amounts of training(Tr) or extended training and adults either with medium orrestricted amounts of training which resulted in three differentperformance levels before retention intervals (i.e. low-per-forming children, intermediate-performing children and adults,high-performing adults). Medium amounts of training consistedof 10 blocks (bl) on the button-box task which were performedduring the learning phase immediately before the 120-minretention interval during which subjects either took a nap(sleep condition) or remained awake (wake condition). Highamounts of training in the children included, in addition to the10 blocks performed during the learning phase, practice ontwo times 10 blocks (Tr +1, Tr +2) which were performed three(Day -3) and two days (Day -2) before the learning phase.There was no intervention on the day before the learning phase(Day -1). Restricted training in the adults consisted of perfor-mance on only two blocks during the learning phase. Thirtyminutes after the end of training in the learning phase, per-formance was tested the first time (Test) which was used asbaseline performance to assess performance improvementsacross the succeeding 120-min retention intervals of eithersleep or wakefulness. This first test consisted of three blocks,except for the adults with restricted training who performed ononly two blocks again. After the 120-min retention intervals allsubjects were tested again on the button-box task in a Retestconsisting of three blocks.

Figure 2 Motor performance at Training 1 and 2 and atLearning in Children: Motor performance in the children withhigh amounts of training at Training 1, 2 and at Learning on themain experimental day (open and filled triangles) and motorperformance in children with intermediate amounts of trainingat Learning on the main experimental day (open and filledcircles). For both experimental groups, performance in thesleep (filled symbols) and wake condition (open symbols) isindicated. Two additional training sessions resulted in signifi-cantly better motor performance at Learning in the group ofchildren with high amounts of training compared to interme-diate amounts of training (p < .001).

Sleep-dependent consolidation of motor memories 3


statistical analyses of motor sequence performance. Thetask did not include presentation of a random sequence.

Sleep recordings and analysis

Standard polysomnographic recordings were obtainedusing a portable amplifier (SOMNOscreen EEG 10–20,Somnomedics, Kist, Germany). Recordings were visuallyscored offline according to the criteria pf Rechtschaffenand Kales (1968). For each night, sleep onset, total sleeptime, and the time as well as the percentage of total sleeptime spent in the different sleep stages were determined.Sleep stages are wake, NonREM (REM – rapid eyemovement) sleep stages 1, 2, 3, and 4, slow wave sleep(SWS, i.e. the sum of stages 3 and 4 sleep) and REMsleep. Sleep onset latency (i.e. the first occurrence of aperiod of stage 1 sleep followed by stage 2 sleep) wasdetermined with reference to the time of lights off.Latencies of SWS and REM sleep were determined withreference to sleep onset.

EEG spindles were identified automatically in Non-REM sleep stage 2 and SWS separately at frontal, centraland parietal sites (F3, F4, C3, C4, P3, P4) using a cus-tom-made software tool (SpindleToolbox V.3) that wasbased on an algorithm adopted from previous studies(Mçlle, Marshall, Gais & Born, 2002). Briefly, first foreach individual and recording channel the average powerspectrum was calculated enabling the user to visuallydetect the individual peak of the spindle frequency band.Then, the root mean square (RMS) of the band-passfiltered signal in the range € 1.5 Hz around the detectedspindle peak of each 200-msec interval was calculatedand the events were counted for which the RMS signalexceeded a constant threshold of 5 lV for 0.5–3 sec. Themean number of spindles were calculated separately forfrontal (F3, F4), central (C3, C4) and parietal (P3, P4)recording sites.

Statistical analyses

Separate analyses of variance were run to evaluate (i) thelevel of motor sequence performance reached at the endof training, (ii) the improvement in motor performancefrom the end of training to the first test taking placeimmediately before the retention interval (Test), and (iii)the gain in motor performance across the retentioninterval, i.e. from Test to Retest. The performance levelreached at the end of training was indicated by theaverage reaction time across the last three blocks of thetraining period; in the case of the adults with lowamounts of training, reaction times were averaged acrossthe total two blocks of training they received. Statisticalanalyses of the performance levels relied on analyses ofvariance (ANOVA) including group factors for ‘age’(children vs. adults), a repeated measures factor ‘sleep ⁄ -wake’, and a group factor for the relative amount of‘training’ each of these groups received, whereby in thechildren this amount was either ‘intermediate’ or ‘high’

and in adults it was either ‘low’ or ‘intermediate’,respectively. Additional analyses were run to evaluate theimprovement in motor sequence performance across thetraining blocks with these ANOVAs including a separate‘block’ factor.

To analyze the improvement in motor performanceacross the two retention intervals (i.e. from the end oftraining to Test as well as from Test to Retest), percentdifferences were calculated, i.e. the difference between theindividual average reaction times during the last threeblocks of the training and at Test, with performance atthe end of training set to 100%, as well as the differencebetween the individual average reaction times at Test andat Retest, with performance at Test set to 100%. (For theadults with low amounts of training, average reactiontimes were calculated for only two blocks.) Here, we usedpercent values of improvement to account for baselinedifferences between children and adults (althoughANOVA on absolute values revealed essentially thesame results as for percentages). The respective ANOVAincluded factors for age, sleep ⁄ wake and training.Additional ANOVAs were run on the absolute reactiontimes before and after the respective retention intervals(including a ‘before ⁄ after’ factor). As these ANOVAsessentially confirmed results for the percent improve-ments, they are not reported here.

Sleep parameters and subjective ratings were analyzedby ANOVA including the factors age, sleep ⁄ wake andtraining and, for subjective ratings, also a factor ‘time’(Test vs. Retest). Analyses of sleep spindles incuded anadditional factor for ‘topography’. Post-hoc t-tests werecalculated if ANOVA revealed significant interactions.As post-hoc t-test only planned comparisons were madewhich were not corrected for multiple comparisons.Greenhouse-Geisser correction of degrees of freedomwas introduced where appropriate. A p- value < .05 wasconsidered significant.

Results

Training performance

Performance continuously improved across the blocks oftraining on the motor sequence task in both children andadult groups (p < .001 for main effect of ‘block’), andalso between the three training sessions the children withhigh amounts of training performed on three successivedays (p < .01, for main effect of ‘session’, Figure 2).Importantly, the four experimental groups differed dis-tinctly according to the motor performance they reachedat the end of training (Figure 3A). As expected, at the endof training children with high amounts of training (threetimes 10 blocks on three different days) showed fasterreaction times than children with intermediate amountsof training (10 blocks of training on the main experi-mental day, p < .001); and adults with intermediatetraining showed faster reaction times than adults with low



amounts of training (two blocks on the main experi-mental day, p < .001; p < .001, for main effect of ‘train-ing’). Moreover, adults were generally faster than children(p < .001, main effect of age). The extension of trainingfrom 10 to 30 blocks in children improved motorsequence performance to a greater extent than the exten-sion from two to 10 blocks in adults (p = .014, for training· age interaction). Accordingly, by manipulating theextent of training we were able to induce three differentlevels of motor performance reached during the learningphase: low-performing children, intermediate-performingchildren and adults, and high-performing adults.

Motor sequence performance 30 minutes after training(Test)

Reaction times at the end of extended training can beslowed due to fatigue developing during training. In orderto obtain baseline measures for evaluating offlineconsolidation that are not contaminated by fatigue-related processes, subjects performed on three blocks (twoblocks in adults with low amounts of training) of thebutton-box task 30 minutes after the end of training.Compared with performance at the end of training,motor sequence performance at this first test was signif-icantly enhanced (p < .001). Percent improvements inmotor performance across the 30-min interval (withperformance at the end of training set to 100%) weresignificantly greater when performance levels at the end oftraining were relatively low, i.e. in children with 10 blocksof training and in adults with two blocks of training(p < .001, for main effect of training, see Figure 3B).

Gains in motor sequence performance across sleep andwake retention intervals

Reaction times on the button-box task were again tested onthree blocks after the �120-min retention intervals filled

with a nap or a comparable period of wakefulness (Retest).Generally, motor performance further improved across theretention interval, with reference to performance at the firsttest before the retention interval (p < .001, Figure 4).Independent of age, subjects who had slept during theretention interval showed greater performance gains thanthe subjects who had stayed awake (p = .017, for sleep ⁄ -wake main effect, see Figure 5). However, the effect ofsleep distinctly depended on the pre-sleep performancelevel (p = .048, for sleep ⁄ wake · age · training). In low-performing children (i.e. after 10 blocks of training),reaction time gains did not significantly differ between thesleep and wake retention interval (sleep: 12.54 € 2.41%,wake: 9.83 € 2.55%; p > .49) whereas in intermediate-performing children (i.e. after three times 10 blocks oftraining) the performance gain was distinctly greater afterretention sleep than wakefulness (7.83 € 2.58 versus)0.60 € 2.36%; p = .006). Complementary to this patternin children, in adults performance gains across theretention interval did not significantly differ for the high-performing group (i.e. after 10 blocks of training; sleep:6.76 € 3.19%, wake: 8.00 € 2.44%; p > .76), whereas thegain in motor performance was significantly greater afterretention sleep than wakefulness in the intermediate-per-forming adults (i.e. after two blocks of training;13.38 € 2.22 vs. 7.10 € 1.64%, p = .019). Thus, acrossage, sleep induced most robust gains in motor sequenceperformance when pre-sleep performance was at anintermediate level, whereas no sleep-associated gains inreaction time were obtained when pre-sleep performancewas either very high or very low.

Possible modulating factors

Sleep architecture

Table 1 summarizes polysomnographical results. Com-pared with adults, children in the sleep conditions spent

(A) (B)

Figure 3 Motor performance at training and 30 minutes later (Test): (A) Motor performance in low- and intermediate-performingchildren (filled and open circles, respectively) and intermediate- and high-performing adults (filled and open triangles, respectively)during training and at Test. (B) Gains in motor performance at Test, i.e. 30 minutes after training (with performance at the end oftraining set to 100%) for low-, intermediate- and high-performing children and adults, respectively. Since there were no differencesbetween the respective sleep and wake conditions (p > .23), data are pooled across these conditions. Means (± SEM) are indicated.



less time awake and in NonREM sleep stage 1 and stage2 (p < .001), but more time in slow wave sleep (SWS;p < .001). Children displayed a shorter SWS latency(p < .001) than adults. There were no significant differ-ences between the age groups for time in REM sleep orany of the other sleep parameters (for all comparisonsp > .10).

Spindles

Analyses of EEG spindles revealed that the number ofspindles was generally higher in adults than children(p = .023, for main effect of age), and also generallyhigher over the anterior than the posterior cortex (p < .001,for main effect of topography), with this topographicaldistribution being more evident in children than adults(p = .005, for age · topography). Spindle counts weregenerally lower in children than adults (average across

recording sites in low-performing children 193.66 € 19.74,medium-performing children 182.31 € 13.10, medium-performing adults 262.95 € 25.79, high-performing adults214.73 € 29.19, p = .02, for effect of age). Spindle countswere associated with the absolute gain in reaction timeperformance over the retention interval, but only in theintermediate-performing adults who showed robustbenefits from sleep after training, with these correlationsbeing greatest at parietal recording sites (frontal:r = 0.50, p < .08; central: r = 0.52, p < .07; parietal:r = 0.56, p < .05). There were no similar correlationsin the high-performing adults or in the children (allp > .12)

Feelings of tiredness and motivation

Ratings indicated that subjects generally felt more tiredbefore the retention interval (Test) than after (i.e. Retest;children: p = .04; adults: p = .005, for main effect oftime) with this decrease in tiredness being more pro-nounced before retention sleep than retention wakeful-ness (children: p < .001; adults: p < .001, for sleep ⁄ wake· time). The amount of training did not affect tiredness(all p > .12). Adults in the wake conditions in compari-son to the sleep conditions indicated being more moti-vated before and after the retention interval (p = .01, forsleep ⁄ wake main effect). However, rated motivation didnot correlate with the gain in motor performance (allr < 0.17, p > .32) excluding a confounding impact ofmotivation. In children, rated motivation tended toincrease after retention sleep and to decrease afterretention wakefulness (p = .056, for sleep ⁄ wake · time).Again, the amount of training did not affect ratedmotivation, either in children or in adults (all p > .18).

Discussion

Our data show that the level of performance reachedafter different amounts of training on a motor sequencetask crucially affects whether this motor skill benefits

(A) (B)

Figure 4 Motor performance at Test and Retest in children and adults: Motor performance in each of the blocks 30 minutes afterlearning (Test) and after retention periods (Retest) which were either filled with sleep (filled symbols) or wakefulness (open symbols)in the intermediate- and low-performing children (A, circles and triangles, respectively) and the intermediate- and high-performingadults (B, triangles and circles, respectively). Means (± SEM) are indicated.

Figure 5 Motor memory consolidation: Gains in motor per-formance after retention sleep (black bars) and wakefulness(white bars) in low-, intermediate- and high-performingsubjects indicated by the percent difference in performancebetween Test and Retest (with performance at Test set to 100%).** p < .01, * p < .05, for pairwise comparisons between sleepand wake conditions. Means (± SEM) are indicated.



from a subsequent nap, both in children and adults.Whereas our children after a standard amount oftraining did not show sleep-dependent gains in motorsequence performance, they did so after extensive train-ing, indicating that children’s low performance level canbe one factor that prevented the development of sleep-associated benefits for skill memory in previous studies(Wilhelm et al., 2008; Fischer et al., 2007; Prehn-Kris-tensen et al., 2009). Whereas our adults after a standardamount of training did not exhibit nap-relatedimprovements in motor skill, they did so when pre-sleeptraining was reduced to only two blocks of practice.Because with this low amount of training the adults’pre-sleep performance came close to the children’s per-formance after extensive training, our findings overallsuggest that, independent of age, the beneficial effect ofsleep on motor skill is most robust with an intermediatelevel of pre-sleep performance.

Although considerable evidence has been accumulatedthat sleep supports the consolidation of skill memory, therespective findings have often been questioned based onthe presence of confounding factors, most importantlycircadian factors and fatigue developing during learning(Cai & Rickard, 2009; Rickard, Cai, Rieth, Jones & Ard,2008; Song, Howard & Howard, 2007). Specifically, itwas argued that in studies comparing nocturnal sleepwith daytime wakefulness, the behavioral expression of

learning at the end of training in the evening is muchlower than the actual learning performance due to (i) acircadian low in learning capability in the evening and (ii)fatigue gradually increasing during the course of training(Keisler et al., 2007). Here, we excluded these potentialconfounding factors by (i) investigating effects of sleepand waking on motor performance at the same circadiantime, i.e. before and after a midday-nap or a parallelperiod of wakefulness, and (ii) assessing pre-sleep per-formance level in a separate test that took place30 minutes after training had ended to allow recoveryfrom fatigue. Our results confirmed in the presence ofthese controls that compared with waking the increase inmotor sequence performance after a 2-hour retentioninterval was significantly greater when sleep occurredduring this interval (Walker et al., 2003a).

Several studies in adults have indicated an impact ofthe pre-sleep level of performance on sleep-dependentmemory consolidation (Albouy et al., 2008; Hauptmann,Reinhart, Brandt & Karni, 2005; Kuriyama et al., 2004)which led us to hypothesize that the lack of benefit fromsleep for motor skill memories that was consistentlyfound in children resulted from the children’s slower andless automated pre-sleep task performance (Fischer et al.,2007; Wilhelm et al., 2008; Prehn-Kristensen et al.,2009). Confirming this hypothesis we found that inchildren who received a threefold increased amount oftraining on the motor sequence sleep indeed improvedthe motor skill. Reaction times in these children– although remaining still significantly longer – approx-imated those in adults who received the minimumtraining of only two blocks on the sequence. As these twoexperimental groups, i.e. children after extended trainingand adults after restricted amounts of training, benefit-ted most from the nap, whereas the low-performingchildren and the high-performing adults after the stan-dard amount of training both showed no consistentbenefit from the nap, our findings agree well with the

Table 2 Subject data

Adults Children

Amount of training

Low Medium Medium High

N 18 15 18 17Female 8 9 8 5

Number of subjects as well as the number of females in each experimental con-dition.

Table 1 Sleep data

Children lowperformer

Children intermediateperformer

Adults intermediateperformer

Adults highperformer p-values

Total sleep time (min) 66.6 € 4.0 66.1 € 2.9 70.71 € 6.6 66.04 € 7.6 < 0.001Sleep onset (min) 20.7 € 2.3 25.0 € 2.6 26.06 € 4.1 28.5 € 5.3 nsSWS latency (min) 8.00 € 0.7 8.5 € 0.9 29.1 € 6.2 43.0 € 6.4 < 0.001REM latency (min) 63.9 € 4.7 59.4 € 3.5 68.9 € 7.2 71.5 € 5.2 nsSleep stages – time in min

Wake 3.1 € 0.7 3.1 € 0.9 7.5 € 1.1 12.1 € 3.2 < 0.001Stage 1 3.7 € 0.4 4.3 € 0.7 7.9 € 0.9 7.5 € 1.4 < 0.001Stage 2 13.8 € 1.9 14.5 € 2.0 35.6 € 3.9 33.7 € 4.8 < 0.001SWS 43.0 € 2.9 40.5 € 3.6 14.7 € 2.8 8.3 € 2.8 < 0.001REM 2.9 € 1.0 3.6 € 0.8 4.5 € 1.5 4.1 € 2.1 ns

Sleep stages – % of TSTWake 4.6 € 1.1 4.9 € 1.6 11.8 € 1.9 19.8 € 4.4 < 0.001Stage 1 5.6 € 0.6 6.6 € 1.1 12.8 € 2.1 11.7 € 1.7 < 0.001Stage 2 20.1 € 2.1 21.9 € 2.6 49.5 € 2.4 52.0 € 4.7 < 0.001SWS 65.6 € 3.2 60.9 € 3.9 20.2 € 3.4 11.4 € 4.4 < 0.001REM 4.0 € 1.5 5.6 € 1.2 5.0 € 1.7 4.5 € 2.4 ns

Note: Mean (€ s.e.m.) total sleep time (TST), latency of SWS and REM sleep (in minutes, with reference to sleep onset) and time spent awake, in stage 1 sleep, stage 2 sleep,slow wave sleep (SWS) and rapid eye movement (REM) sleep in minutes and percentage of total sleep time. Right column indicates p-values for the main effect in theANOVA. Differences in sleep parameters within the children or adults performing at the different levels were not significant.



view that sleep preferentially benefits procedural motormemories at an intermediate performance level (Stick-gold, 2009).

Thus, in order to directly profit from sleep, motorskills in children need to be more intensely trained.Whether it is the mere strengthening itself of the skillrepresentation or the greater relevance attached to theskill after repeated training that enables the sleep-dependent skill improvement is presently not clear(Wilhelm, Diekelmann, Molzow, Ayoub, Molle & Born,2011). Likewise, we can only speculate about the neuro-physiological mechanisms underlying the consolidationof motor sequence memories at an improved but not atlow pre-sleep performance level in children. In adults,patterns of brain activity distinctly change in the courseof motor sequence learning. Activity in the prefrontalcortex, cerebellum and parietal cortex graduallydecreases whereas activation in the striatum, primarymotor cortex, supplementary motor areas and hippo-campus increases (Willingham, 1998; Karni, Meyer,Rey-Hipolito, Jezzard, Adams, Turner & Ungerleider,1998; Doyon & Benali, 2005; Albouy et al., 2008). It hasbeen proposed that activation in hippocampal areasduring learning determines to what extent a memoryundergoes sleep-dependent consolidation, and this mayalso hold for sleep-dependent benefits in motor sequencelearning (Diekelmann et al., 2009; Rauchs, Feyers, Landeau,Bastin, Luxen, Maquet & Collette, 2011; Spencer, Sunm& Ivry, 2006). In fact, activation in this region predictedmemory gains after sleep in a motor sequence learningtask in adults (Albouy et al., 2008). Children’s motorperformance is distinctly slower and less automatedcompared to adults (Meulemans, van der Linden &Perruchet, 1998; Wilhelm et al., 2008), and this coincidedwith comparably less hippocampal activation duringsequence learning in a (probabilistic) serial reaction timetask (Thomas, Hunt, Vizueta, Sommer, Durston, Yang &Worden, 2004). Relatively reduced hippocampal activa-tion during training might have also been a factorexplaining why in the present study children with a lowamount of training did not benefit from sleep. In thechildren, hippocampal recruitment only after extendedtraining on the motor sequence task might achieve levelswhere sleep produces a distinct improvement in theseskills.

We found a significant correlation between the numberof sleep spindles during NonREM sleep stage 2 and SWSand the sleep-associated gain in motor sequence perfor-mance in the adults whose motor skill improved by sleep,i.e. the group with restricted training. This finding con-firms a number of previous studies indicating a robustlink between sleep spindles and the sleep-dependentconsolidation of procedural motor memories (Rasch,Pommer, Diekelmann & Born, 2009; Morin, Doyon,Dostie, Barakat, Hadj, Korman, Benali, Karni,Ungerleider & Carrier, 2008; Nishida & Walker, 2007;Tamaki, Matsuoka, Nittono & Hori, 2008; Fogel &Smith, 2011). Spindles synchronize gamma band activity

between different neocortical networks and, in doing so,could be a mechanism enhancing the interlinkagebetween neocortical parts of the motor sequence repre-sentation (Ayoub, Mçlle, Preissl & Born, 2011). Sur-prisingly, this correlation between spindle counts andsleep-related improvement in motor sequence perfor-mance was not obtained in the children who profitedfrom sleep, i.e. the children with extended training.Although the difference in the respective correlationsbetween adults and children did not reach significance,this finding could point towards differences in memoryprocessing during sleep between the age groups, requir-ing further examination.

Conceptually, our findings of a sleep-dependentimprovement in motor sequence skill specifically withintermediate levels of pre-sleep performance may beexplained on the basis of an interaction between explicitand implicit memory systems. Procedural motor taskslike the button-box task of the present study compriseimplicit and explicit components, both determining thefinal response speed (Schendan, Searl, Melrose & Stern,2003; Fletcher, Zafiris, Frith, Honey, Corlett, Zilles &Fink, 2005; Shanks & Johnstone, 1999; Sun, Zhang,Slusarz & Mathews, 2007; Willingham, 1998). Sleep hasbeen found to preferentially facilitate the consolidationof explicit aspects of a task representation (Robertson,Pascual-Leone & Press, 2004; Spencer et al., 2006;Diekelmann & Born, 2010). Importantly, whether ex-plicit knowledge strengthened during sleep helps orhinders motor sequence performance strongly dependson the performance level. At an early stage of motorlearning, explicit knowledge may even decrease skillperformance because both hippocampus-dependent ex-plicit systems and cortico-striatal systems underlyingimplicit skill performance competitively interact (Will-ingham, 1998; Poldrack, Clark, Par�-Blagoev, Shohamy,Creso Moyano, Myers & Gluck, 2001; Albouy et al.,2008). At a more advanced stage of performance, im-plicit aspects of the representations may be strong en-ough not to be essentially disturbed by sleep-dependentgains in explicit knowledge, enabling sleep to directlyenhance implicit aspects of the task representation.Because children show a superior capacity for enhanc-ing explicit aspects in memory during sleep (Wilhelm,Rasch, Rose, B�chel & Born, submitted), in the low-performing children with only weak implicit skill rep-resentations, any sleep-dependent gains in implicit mo-tor performance might be nullified by competinginteractions with explicit aspects that were strengthenedduring sleep. It appears that children’s implicit skillmemories reach a level of strength and independenceonly after extended training, which is comparable towhat adults obtain with only a minimum of practice,and which allows for an unhampered emergence of again in motor skill after sleep. Such interaction betweenexplicit and implicit task aspects very likely occursduring memory retrieval but may also occur offlineduring consolidation.



Adults with the standard amount of training (i.e. 10blocks) did not benefit from periods of sleep afterlearning, which appeared to contrast with findings fromseveral previous studies (e.g. Fischer et al., 2002; Debaset al., 2010). However, rather than fine motor fingertapping as used in those previous studies, we employedhere a coarse motor task that is highly adapted to chil-dren’s motor abilities and requires movements with thewhole hand. Adults rapidly reach a rather high level ofperformance on this task at which further improvementsin speed might be difficult to achieve. Thus, the lackingsleep effect in the adult group with 10 blocks of trainingon this simple task might reflect a ceiling effect thatemerges similarly in every motor memory task once highperformance levels are reached. Also, improvements werecalculated with reference to a separate retest occurring30 min after training which revealed a higher perfor-mance level than at the end of training. However, mostprevious studies in adults used the performance level atthe end of training as the reference for determining sleep-related benefits in motor skill. Finally, here we examinedeffects of a 90-min nap rather than a whole night of sleep.Sleep of longer duration might be more effective inunmasking small increases in reaction times (Diekel-mann et al., 2009).

Acknowledgements

This work was supported by the Deutsche Forschungs-gemeinschaft (SFB 654). The authors thank SusanneDiekelmann and Sabine Groch for fruitful discussion ofthe data.

References

Albouy, G., Sterpenich, V., Balteau, E., Vandewalle, G., Des-seilles, M., Dang-Vu, T., Darsaud, A., Ruby, P., Luppi, P.-H., Degueldre, C., Peigneux, P., Luxen, A., & Maquet, P.(2008). Both the hippocampus and striatum are involved inconsolidation of motor sequence memory. Neuron, 58, 261–272.

Ayoub, A., Mçlle, M., Preissl, H., & Born, J. (2011). Groupingof MEG gamma oscillations by EEG sleep spindles. Neuro-Image (Epub ahead of print).

Buysse, D.J., Reynolds, C.F., 3rd, Monk, T.H., Berman, S.R.,& Kupfer, D.J. (1989). The Pittsburgh sleep quality index: anew instrument for psychiatric practice and research. Psy-chiatry Research, 28, 193–213.

Cai, D.J., &Rickard, T.C. (2009).Reconsidering the role of sleepfor motor memory. Behavioral Neuroscience, 123, 1153–1157.

Campbell, I.G., & Feinberg, I. (2009). Longitudinal trajectoriesof non-rapid eye movement delta and theta EEG as indica-tors of adolescent brain maturation. Proceedings of theNational Acad.emy of Sciences, USA, 106, 5177–5180.

Debas, K., Carrier, J., Orban, P., Barakat, M., Lungu, O.,Vandewalle, G., Hadj Tahar, A., Bellec, P., Karni, A.,

Ungerleider, L.G., Benali, H., & Doyon, J. (2010). Brain

plasticity related to the consolidation of motor sequencelearning and motor adaptation. Proceedings of the NationalAcademy of Sciences, USA, 107, 17839–17844.

Deregnaucourt, S., Mitra, P.P., Feher, O., Pytte, C., & Tcher-nichovski, O. (2005). How sleep affects the developmentallearning of bird song. Nature, 433, 710–716.

Diekelmann, S., & Born, J. (2010). The memory function ofsleep. Nature Reviews Neuroscience, 11, 114–126.

Diekelmann, S., Wilhelm, I., & Born, J. (2009). The whats andwhens of sleep-dependent memory consolidation. SleepMedicine Reviews, 13, 309–321.

Dorfberger, S., Adi-Japha, E., & Karni, A. (2007). Reducedsusceptibility to interference in the consolidation of motormemory before adolescence. PLoS.ONE, 2, e240.

Doyon, J., & Benali, H. (2005). Reorganization and plasticity inthe adult brain during learning of motor skills. CurrentOpinion in Neurobiology, 15, 161–167.

Doyon, J., Korman, M., Morin, A., Dostie, V., Hadj, T.A.,Benali, H., Karni, A., Ungerleider, L.C., & Carrier, J. (2009).Contribution of night and day sleep vs. simple passage of timeto the consolidation of motor sequence and visuomotoradaptation learning. Experimental Brain Research, 195, 15–26.

Fischer, S., Hallschmid, M., Elsner, A.L., & Born, J. (2002).Sleep forms memory for finger skills. Proceedings of theNational Academy of Sciences, USA, 99, 11987–11991.

Fischer, S., Wilhelm, I., & Born, J. (2007). Developmentaldifferences in sleep’s role for implicit off-line learning: com-paring children with adults. Journal of Cognitive Neurosci-ence, 19, 214–227.

Fletcher, P.C., Zafiris, O., Frith, C.D., Honey, R.A., Corlett,P.R., Zilles, K., & Fink, G.R. (2005). On the benefits of nottrying: brain activity and connectivity reflecting the interac-tions of explicit and implicit sequence learning. CerebralCortex, 15, 1002–1015.

Fogel, S.M., & Smith, C.T. (2011). The function of the sleepspindle: a physiological index of intelligence and a mecha-nism for sleep-dependent memory consolidation. Neurosci-ence and Biobehavioral Reviews, 35, 1154–1165.

Hauptmann, B., Reinhart, E., Brandt, S.A., & Karni, A.

(2005). The predictive value of the leveling off of withinsession performance for procedural memory consolidation.Brain Research. Cognitive Brain Research, 24, 181–189.

Karni, A., Meyer, G., Rey-Hipolito, C., Jezzard, P., Adams,M.M., Turner, R., & Ungerleider, L.G. (1998). The acqui-sition of skilled motor performance: fast and slow experi-ence-driven changes in primary motor cortex. Proceedings ofthe National Academy of Sciences, USA, 95, 861–868.

Kaufman, J., Birmaher, B., Brent, D., Rao, U., Flynn, C.,Moreci, P., Williamson, D., & Ryan, N. (1997). Schedule foraffective disorders and schizophrenia for school-age children– Present and lifetime version (K-SADS-PL): initial reliabil-ity and validity data. Journal of the American Academy ofChild and Adolescent Psychiatry, 36, 980–988.

Keisler, A., Ashe, J., & Willingham, D.T. (2007). Time of dayaccounts for overnight improvement in sequence learning.Learning & Memory, 14, 669–672.

Kuriyama, K., Stickgold, R., & Walker, M.P. (2004). Sleep-dependent learning and motor-skill complexity. Learning &Memory, 11, 705–713.

Meulemans, T., van der Linden, M., & Perruchet, P. (1998).Implicit sequence learning in children. Journal of Experi-mental Child Psychology, 69, 199–221.



Mçlle, M., Marshall, L., Gais, S., & Born, J. (2002). Groupingof spindle activity during slow oscillations in human non-rapid eye movement sleep. Journal of Neuroscience, 22,10941–10947.

Morin, A., Doyon, J., Dostie, V., Barakat, M., Hadj, T.A.,Korman, M., Benali, H., Karni, A., Ungerleider, L.G., &Carrier, J. (2008). Motor sequence learning increases sleepspindles and fast frequencies in post-training sleep. Sleep, 31,1149–1156.

Nishida, M., & Walker, M.P. (2007). Daytime naps, motormemory consolidation and regionally specific sleep spindles.PLoS.ONE, 2, e341.

Ohayon, M.M., Carskadon, M.A., Guilleminault, C., & Viti-ello, M.V. (2004). Meta-analysis of quantitative sleepparameters from childhood to old age in healthy individuals:developing normative sleep values across the human lifespan.Sleep, 27, 1255–1273.

Poldrack, R.A., Clark, J., Par�-Blagoev, E.J., Shohamy, D.,Creso Moyano, J., Myers, C., & Gluck, M.A. (2001). Inter-active memory systems in the human brain. Nature, 414, 546–550.

Prehn-Kristensen, A., Gçder, R., Chirobeja, S., Bressmann, I.,Ferstl, R., & Baving, L. (2009). Sleep in children enhancespreferentially emotional declarative but not proceduralmemories. Journal of Experimental Child Psychology, 104,132–139.

Rasch, B., Pommer, J., Diekelmann, S., & Born, J. (2009).Pharmacological REM sleep suppression paradoxicallyimproves rather than impairs skill memory. Nature Neuro-science, 12, 396–397.

Rauchs, G., Feyers, D., Landeau, B., Bastin, C., Luxen, A.,Maquet, P., & Collette, F. (2011). Sleep contributes to thestrengthening of some memories over others, depending onhippocampal activity at learning. Journal of Neuroscience, 31,2563–2568.

Rechtschaffen, A., & Kales, A. (1968). A manual of standard-ized terminology, techniques, and scoring system for sleepstages of human subjects. Bethesda, MD: US Department ofHealth, Education, and Welfare (NIH).

Rickard, T.C., Cai, D.J., Rieth, C.A., Jones, J., & Ard, M.C.(2008). Sleep does not enhance motor sequence learning.Journal of Experimental Psychology: Learning, Memory, andCognition, 34, 834–842.

Robertson, E.M., Pascual-Leone, A., & Press, D.Z. (2004).Awareness modifies the skill-learning benefits of sleep. Cur-rent Biology, 14, 208–212.

Schendan, H.E., Searl, M.M., Melrose, R.J., & Stern, C.E.(2003). An FMRI study of the role of the medial temporallobe in implicit and explicit sequence learning. Neuron, 37,1013–1025.

Shanks, D.R., & Johnstone, T. (1999). Evaluating the rela-tionship between explicit and implicit knowledge in asequential reaction time task. Journal of Experimental Psy-chology: Learning, Memory, and Cognition, 25, 1435–1451.

Song, S., Howard, J.H., Jr., & Howard, D.V. (2007). Sleep doesnot benefit probabilistic motor sequence learning. Journal ofNeuroscience, 27, 12475–12483.

Spencer, R.M., Sunm, M., & Ivry, R.B. (2006). Sleep-depen-dent consolidation of contextual learning. Current Biology,16, 1001–1005.

Stickgold, R. (2009). How do I remember? Let me count theways. Sleep Medicine Reviews, 13, 305–308.

Sun, R., Zhang, X., Slusarz, P., & Mathews, R. (2007). Theinteraction of implicit learning, explicit hypothesis testinglearning and implicit-to-explicit knowledge extraction. Neu-ral Networks, 20, 34–47.

Tamaki, M., Matsuoka, T., Nittono, H., & Hori, T. (2008).Fast sleep spindle (13–15 hz) activity correlates with sleep-dependent improvement in visuomotor performance. Sleep,31, 204–211.

Thomas, K.M., Hunt, R.H., Vizueta, N., Sommer, T., Dur-ston, S., Yang, Y., & Worden, M.S. (2004). Evidence ofdevelopmental differences in implicit sequence learning: anfMRI study of children and adults. Journal of CognitiveNeuroscience, 16, 1339–1351.

Walker, M.P., Brakefield, T., Hobson, J.A., & Stickgold, R.

(2003a). Dissociable stages of human memory consolidationand reconsolidation. Nature, 425, 616–620.

Walker, M.P., Brakefield, T., Seidman, J., Morgan, A., Hob-son, J.A., & Stickgold, R. (2003b). Sleep and the time courseof motor skill learning. Learning & Memory, 10, 275–284.

Wilhelm, I., Diekelmann, S., & Born, J. (2008). Sleep in chil-dren improves memory performance on declarative but notprocedural tasks. Learning & Memory, 15, 373–377.

Wilhelm, I., Diekelmann, S., Molzow, I., Ayoub, A., Molle,M., & Born, J. (2011). Sleep selectively enhances memoryexpected to be of future relevance. Journal of Neuroscience,31, 1563–1569.

Wilhelm, I., Rasch, B., Rose, M., B�chel, C., & Born, J.

(submitted). The sleeping child outplays adults’ capacity forinsightful understanding.

Willingham, D.B. (1998). A neuropsychological theory ofmotor skill learning. Psychological Review, 105, 558–584.

Received: 27 May 2011Accepted: 18 January 2012



Documents

Sleep-dependent consolidation of procedural motor memories in children and adults: the pre-sleep level of performance matters