Neuroscience Letters 388 (2005) 27–32

Effects of sleep deprivation on event-related fields and alphaactivity during rhythmic force production

T.W. Boonstra∗, A. Daffertshofer, P.J. BeekFaculty of Human Movement Sciences, Institute for Fundamental and Clinical Human Movement Sciences,

Vrije Universiteit, Van der Boechorststraat 9, 1081BT Amsterdam, The Netherlands

Received 26 April 2005; received in revised form 8 June 2005; accepted 11 June 2005

Abstract

The influence of sleep deprivation (SD) on event-related fields and the distribution of power over the scalp of MEG imaged brain activitywas studied during acoustically paced rhythmic force production. At the behavioral level, SD resulted in a reduction of the lag (negativeasynchrony) between produced forces and acoustic stimuli at higher movement tempos. Principal component analysis of the accompanyingMEG activity showed that auditory- and motor-evoked fields were attenuated after SD and revealed an anterior shift of power towards morefrontal channels. These results were interpreted in terms of a change of central processing of afferent sensory input due to SD.©

K

T(twasto[ma

ltTlwc

wf

rcesn)imi-syn-wasin

rdi-ncy

asivity.own

) par-ancethealand

eadweredue

0d

2005 Elsevier Ireland Ltd. All rights reserved.

eywords:Sleep deprivation; MEG; Event-related fields; Spectral analysis; Principal component analysis

he effects of central fatigue on electroencephalographicEEG) activity have been studied extensively in the con-ext of the sleep–wake cycle. During the transition fromakefulness to sleep, EEG activity changes considerablycross frequencies, particularly in the alpha band[8]. Nearleep onset, a posterior–anterior shift of alpha power occurshat corresponds with a decrease of alpha power aboveccipital areas and an increase above more frontal areas

7,13]. To date, however, changes in alpha power lack aore functional interpretation, although they readily pointt a link with attention[30,33].

Apart from relations with fatigue or attention, oscil-atory EEG activity in the alpha band (mu-rhythm) ishought to play an important role in motor tasks[23,27].he mu-rhythm seems to reflect an information-processing

oop between motor cortices and sub-cortical structures,hich entails reverberating activity in thalamo–cortical andortical–cortical circuits[8,21].

To uncover the role of alpha activity in motor functioning,e examined the effects of sleep deprivation (SD) on the per-

ormance of a simple motor task and the accompanying MEG

activity. In this task, subjects had to produce isometric foeither in synchrony with auditory stimuli (synchronizatioor in-between successive stimuli (syncopation). In a slar experimental setup, transitions from syncopation tochronization occurred when the inter-response intervalshortened[12,19,20]. Based on previous results obtainedexperiments on the influence of attention on interlimb coonation[24], we expected SD to affect the (critical) frequeof these transitions. With regard to brain activity, SD wexpected to have an effect on the motor-related mu actSince the transition from wakefulness to sleep is also knto alter the N1–P2 component of the auditory ERP[6,17],SD was further expected to affect event-related activity.

Four male subjects (between 25 and 45 years of ageticipated in the experiment that was conducted in accordwith the Declaration of Helsinki and the guidelines ofMedical Ethical Committee of the Vrije Universiteit MedicCenter. All subjects were self-proclaimed right-handerssigned an informed consent prior to participation.

Neuromagnetic activity was recorded using a whole-hMEG (CTF Systems Inc., Vancouver, Canada). Subjectslying on a bed in a comfortable position to avoid artifacts

∗ Corresponding author. Tel.: +31 20 5988506; fax: +31 20 5988529.E-mail address:t.boonstra@fbw.vu.nl (T.W. Boonstra).

to head motion or involuntary contractions of head and shoul-der muscles. Subjects were instructed to produce isometric

304-3940/$ – see front matter © 2005 Elsevier Ireland Ltd. All rights reserved.

oi:10.1016/j.neulet.2005.06.045

28 T.W. Boonstra et al. / Neuroscience Letters 388 (2005) 27–32

forces by adducting their thumb against a MEG-compatibleforce sensor[5] that was fixed on the bed allowing for (almostmaximal) arm extension. Adduction forces had to be pro-duced either simultaneously with or in-between the acousticstimuli that were delivered binaurally using EARTone 3AInsert Earphones (Cabot Safety Corporation). Stimulus trainsconsisted of 100 tones (pitch: 400 Hz; 50 ms duration each)that were presented with decreasing inter-stimulus intervals(ISIs: ranging from 1000 to 357 ms over 10 plateaus), i.e.with increasing tempo from 1 to 2.8 Hz in steps of 0.2 Hz[12,20]. Each trial lasted about 1 min.

Every subject participated in two experimental sessionsat consecutive days, roughly starting at the same time ofday. During those days, subjects had to abstain from drink-ing alcohol and coffee. On the second day, the subjects weresleep deprived as they had been kept awake for 24 h underthe supervision of one of the experimenters. The record-ings of the first day served as control condition. On bothdays, subjects performed 2× 12 = 24 trials (total duration,about 24 min per day). Synchronization and syncopation tri-als were randomized within subjects. The produced force wasmonitored on-line and visually fed back by projecting it onthe ceiling (feedback signals were low-pass filtered to elimi-nate oscillatory components; fourth order Butterworth filter;cut-off, 0.33 Hz). Subjects were instructed to maintain theforce level within a certain range indicated by two referencel ion,w ent).T ed tofi them

rderg elmets ffec-t s off tane-o italc fored

ther minedv (e.g.[ ativep ntsw wasd

EGs fre-q ver-l tionv owers trialsa fre-q weree r thiss s were

concatenated for each MEG channel in both fatigued andcontrol conditions, resulting in a set of 2× 149 signals (forboth the fatigued and control condition) of 10× 60 frequencybins (0–60 Hz with resolution of 1 Hz). Principal componentswere calculated[12,20]and studied by analyzing the result-ing eigenvalue distributions, the spatial modes (per condition)and the corresponding frequency spectra. Note that by con-catenating frequency spectra we focused on persistent spa-tial activity patterns irrespective of movement tempo. Notefurther that, in this particular application, the PCA servedas an unbiased multivariate clustering measure classifyingconditions similar to a (time-resolved) multivariate analysisof variance, rendering additional statistical tests superfluous[11].

Subsequently, both auditory- and motor-related fieldswere calculated using tone onsets and maxima of the forcecycles, respectively, as event-defining indices. Before aver-aging over events, MEG data were band-pass filtered (fourth-order Butterworth: 0.5–60 Hz). For the auditory-evoked fields(AEFs), epochs were individually averaged for each move-ment tempo (1.0–2.8 Hz) and condition over an interval from−100 to 300 ms (with 0 ms referring to the moment of toneonset). Similarly, motor-evoked fields (MEFs) were calcu-lated by averaging over a−150 to 250 ms interval (with0 ms referring to the moment of maximal force productionfor each cycle). Artifact-contaminated epochs were excludedf bye f thea thes thee aver-a r thee ion,P encys jectsa ition,r -t ndi-t rtedf on thei

andt for-m priort( aledt tem-p ass2 n,a wasfn nd:i ragea lc

ines (2.5% and 7.5% of the maximum voluntary contracthich was determined at the beginning of the experimo minimize eye movements, subjects were further askxate their gaze at a small cross that was displayed iniddle of the force-feedback region.The MEG comprised of 151 SQuID sensors (third-o

radiometers) distributed homogeneously across the hurface. Two channels were not operational so that, eively, 149 MEG signals were recorded. The voltageorce sensor and acoustic stimuli were sampled simulusly with the MEG using two additional analogue–dighannels. All signals were low-pass filtered at 415 Hz beigitization at a sampling rate of 1250 Hz.

To quantify the effects of SD on motor performance,elative phase between tone and force signals was deteria the Hilbert transform of the band-pass filtered signals29]). Per movement tempo, we computed the mean relhase[22] and, for the syncopation trials, transition poiere established as center of a sigmoid function thatetermined by a least squares fit.

Further, we estimated the frequency spectra of the Mignals using Welch’s periodogram method providing auency resolution of 1 Hz (window size: 1250 samples; o

ap: 625 samples). For all four conditions (synchronizaersus syncopation; fatigued versus control), the log-ppectral densities were averaged on a sensor level overnd subjects for each movement tempo. Differences inuency spectra between fatigued and control conditionsxamined using principal component analysis (PCA). Foake, the frequency spectra at the ten movement tempo

rom averaging. Magnetic field distortions (e.g. elicitedye movements and blinks) were marked as artifacts implitude of a MEG channel was larger than eight timestandard deviation for the entire trial. In total, 6.5% ofpochs were removed, leaving at least 350 epochs forging for each movement tempo. To examine whethevent-related fields were different in the fatigued conditCA was used equivalently to the analysis of the frequpectra. Both AEFs and MEFs were averaged over subnd concatenated for every MEG channel in each condesulting in a matrix of 2× 149 channels by 10× 0.4 s duraion for both the synchronization and the syncopation coion. Notice that given the consistency of the here repoeatures of the ERFs, we used grand averages to focusnter-subject commonalities.

The relative Hilbert phase between produced forcesones disclosed that, in the synchronization condition, theer anticipated the latter, that is, peak forces occurred

o tone onsets (negative asynchrony). A tempo (10)× fatigue2) analysis of variance of the relative Hilbert phase revehat this anticipation was reduced at higher movementos (F(9, 27) = 3.8,p< 0.005) and that this reduction wtronger in the fatigued than in the control condition (F(9,7) = 3.0,p< 0.05; Fig. 1A). In the syncopation conditioclear transition from syncopation to synchronization

ound around a movement tempo of 2 Hz (Fig. 1B). No sig-ificant effect of SD on the transition frequency was fou

n the fatigued condition the transition occurred on avefter 35.6± 10.0 s compared to 35.8± 12.0 s in the controondition (the end of the 1.8 Hz plateau).

T.W. Boonstra et al. / Neuroscience Letters 388 (2005) 27–32 29

Fig. 1. (A) Mean relative Hilbert phase (φ) between force and tone for all10 plateaus in the synchronization condition (0 and 2� represent in-phaseforce production). Error bars display the circular standard deviation of meanphases over trials and subjects. (B) Idem, but now for the syncopation con-dition.

PCA of the frequency spectra yielded a drastic reduction ofthe data as the first principal mode of the power spectral densi-ties already covered more than 99% of the data’s spread, withthe parts of the eigenvector corresponding to the fatigued andthe control conditions being rather similar. The power spectracorresponding to the first mode turned out to be almost iden-tical for every movement tempo. In other words, there wasa dominant spectral distribution reflecting the consistency ofthe frequency content during the experiment. Interestingly,the spatial distribution of the power differed after SD (Fig. 2)in that the power was decreased at occipital and temporalchannels (maximal at channel MLT32:−13% for synchro-nization and−11% for syncopation) and increased at frontal

and central channels (maximal at channel LC31: +5.8% forsynchronization and RC31: +7.8% for syncopation). The sec-ond principal mode (about 0.1% of the data’s spread) reflectedthe commonly reported concentration of alpha power aroundoccipital channels. This spatial pattern, however, was lesspronounced after SD because of an anterior shift of alphaactivity. The increase was maximal at channel RC41 in thesynchronization condition and at channel RC42 in the syn-copation condition. The frequency band extracted by modetwo was roughly bounded between 9 and 13 Hz peaking at11 Hz for all movement tempos.

By and large, three principal modes sufficed to describeAEFs and MEFs (eigenvalues for AEFs: 0.53 + 0.18 + 0.08 =0.79 for synchronization and 0.59 + 0.08 + 0.08 = 0.75 forsyncopation; similarly for MEFs: 0.6 + 0.21 + 0.05 = 0.86 and0.59 + 0.14 + 0.05 = 0.78). Importantly, AEF modes appearedto be permutated MEF modes (modes one and two wereswapped (Fig. 3, first two rows)). The first mode in the AEFsdisplayed two bipolar activity patterns (bilateral auditory cor-tices), whereas the first mode in the MEFs reflected a bipolarfield above the left contra-lateral motor area. The observedaverages always contained fields that originated from bothauditory- and motor-related processes, although the auditory-related fields were clearly stronger in the auditory-relatedaverage due to better alignment and visa versa. The left/rightsymmetric mode three was similar in both averages buta harpp

osti notc fereds thep ndi-t ectorc ooka e of

F s. (A) T lotted on2 ndition; e difb longing t frequep

ig. 2. The first two modes of the PCA of the power spectral densitieD scalp; (B) the part of the eigenvectors belonging to the control coetween conditions; (D) PCA projections; the frequency spectra belateaus.

ppeared more event-locked in the case of AEFs (cf. seak aroundt= 100 ms in panel F ofFig. 3).

The spatial distribution of the evoked fields was almdentical for synchronization and syncopation and didhange due to SD. In contrast, the modes’ strengths difignificantly between the fatigued and control condition:art of the eigenvector corresponding to the fatigued co

ions was, in general, smaller than the part of the eigenvorresponding to the control condition. A more detailed lt Fig. 3 revealed that the eigenvector of the first mod

he part of the eigenvectors belonging to the fatigued condition and pa(C) the difference between both parts of the eigenvector showing thferenceto the first two modes sliced in parts corresponding to the differenncy

30 T.W. Boonstra et al. / Neuroscience Letters 388 (2005) 27–32

Fig. 3. The first three PCA modes of the ERFs. Left panels: The parts of the eigenvectors belonging to the fatigued (A) and control (B) condition and thecorresponding time-series (projections) of the MEFs (C). Right panels (D–F): Idem, but now for the PCA for the AEFs.

the AEFs had a maximal strength at a right temporal chan-nel (RT32) and the MEFs were maximal at a left temporalchannel (LT13) in all four conditions. At channel RT32,the strength of the eigenvector for the AEFs was 19% and11% smaller in the fatigued compared to the control con-dition for synchronization and syncopation, respectively. Inthe synchronization condition, for instance, this decreasecorresponded to an amplitude reduction of the N1 from161 to 139 fT at an ISI of 1 s. Similarly, for the MEFs thestrength at channel LT13 was 9% (synchronization) and 19%(syncopation) weaker when sleep deprived. In the secondmodes, an even more pronounced decrease of amplitudewas found in the fatigued condition. To investigate the dif-

ference in amplitude reductions in the fatigued conditionbetween synchronization and syncopation, the ERFs of allfour conditions were analyzed concurrently after concate-nating ERFs of all four conditions into a ‘4× 149 chan-nels by 10× 0.4 s duration’ array. The resulting eigenvalueswere rather similar (AEFs: 0.57 + 0.11 + 0.07 = 0.75; MEFs:0.59 + 0.16 + 0.06 = 0.81). Again, the first eigenvector wasmaximal at channel RT32 for the AEFs and at channel LT13for the MEFs across conditions. Interestingly, at these chan-nels the first principal mode was almost identical to theoriginal ERFs, that is, the reduction to a single PCA modedid preserve the primary structure of the ERFs (Fig. 4). Theexplained variance at this channel (covariance of PCA recon-

F anels: t whE yncop dia %). Ri nc . Uppe

ig. 4. Results of the PCA for the ERFs over all four conditions. Left pRFs were maximal (MEF: LT13; AEF: RT32) (1, synchronization; 2, ss percentage change relative to the mean over all four conditions (100ondition at channel LT13–RT32. Thin line displays the original ERFs

The strength of the eigenvector in all four conditions at the channels aich theation; red, fatigued; blue, control). Differences between conditions aresplayedght panels: the reconstruction of the first mode in the synchronization,on-fatiguedr panels: MEFs; lower panels: AEFs.

T.W. Boonstra et al. / Neuroscience Letters 388 (2005) 27–32 31

struction and original data divided by the variance of theoriginal data) was always above 86%, both for AEFs andMEFs. The reduction of the amplitude of ERFs after SD isshown inFig. 4 (left panels): both for synchronization andsyncopation, amplitudes are reduced in the fatigued con-dition. Furthermore, there is a reduction of amplitude forsyncopation compared to synchronization, for the AEFs onlyin the fatigued condition and for the MEFs only in the controlcondition.

In addition to the amplitude reduction of the ERFs, theprincipal modes also revealed changes of ERFs related tomovement tempo. While for different ISIs the MEFs did notalter, the AEFs changed considerably. The amplitude of theN1m decreased strongly with decreasing ISIs to disappearalmost entirely at an ISI of 500 ms where it converged ontoa later component peaking at 180 ms after tone onset. Thedecreasing strength of the part of the first AEFs’ eigenvec-tor corresponding to the fatigued condition might have beencaused solely by the decrease of the N1m amplitude in thefirst five plateaus. To test this, we finally examined whetherthe effect of SD on the AEF was indeed only present at longISIs. For this purpose, two separate PCAs were conductedby dividing the data in subsets containing plateau 1–5 andplateau 6–10, respectively. If the effect of SD was restrictedto the N1m component, one would expect the difference dueto SD to be only present in the PCA of plateau 1–5. Instead,h cipalm

onM cedr d toe tion)o tion,S syn-c igherm -i taska hes ce inp ation[ usedb tap.S ronym sen-s eadt tionc cults nsatet oneds mt

ndc mentt dly.T han-n

while the alpha band revealed an anterior shift of powertowards central channels. This anterior shift might reflect anincrease of motor-related mu-rhythm activity above centralchannels, because the mu-rhythm and occipital alpha activ-ity have similar frequencies[1]. Since desynchronization ofmu-rhythm has been associated with a functional alerting ofmotor areas[4], the observed anterior shift of alpha powercould imply a reduction of this functional alerting of motorareas. An alternative interpretation of the observed changesin MEG activity along the anterior–posterior axis is that theeffects of SD are local and pertain mostly to (pre-)frontalareas as these areas are most vulnerable to SD[18]. SD isknown to reduce the activity especially in the prefrontal cor-tex (e.g.[16]) and previously reported sleep-related effectson EEG imaged brain activity have also been interpreted asa result of deactivation of frontal areas[3,13,31].

Second, the amplitudes of both AEFs and MEFs wereattenuated after SD. The decrease of AEF amplitude confirmsprevious studies[3,17], although, in contrast to the more tra-ditional focus on the N1–P2 complex, we analyzed the AEFas a whole. N1 is typically related to timing aspects or to onsetinformation of the auditory stimulus[25], which agrees withthe subjects’ self-reported difficulties in evaluating their per-formance when sleep deprived. Interestingly, the drop of AEFamplitude was also found at short ISIs for which the typicalN1m was no longer present. Apparently, the effects of SDw t buta s. Ina en-e bothm veralr mpo-n hery[ ness d foret gestt singo

mu-l thel ivelyw nt ata andi onalps l, wec e N1a lyingn ogyo thattw on ofN

n thei ss’ of

owever, the effect was more pronounced in the first prinode of the subset plateau 6–10.In the present study, we examined the effects of SD

EG activity during performance of an acoustically pahythmic force production task. Subjects were instructeither produce adduction forces at the tones (synchronizar in-between the tones (syncopation). For synchronizaD resulted in a stronger reduction of the negative ahrony between the produced forces and the tones at hovement tempos. Recently, Doumas et al.[14] found a sim

lar reduction of the negative asynchrony in a tappingfter rTMS stimulation of the motor cortex. In line with tuggestion that negative asynchrony reflects a differenrocessing times of somatosensory and auditory inform

2], they hypothesized that the observed reduction was cay altered processing of somatosensory input from eachimilarly, the present reduction of the negative asynchight imply that SD affected the processing of somato

ory information. Contrary to our expectation, SD did not lo a change in the transition frequency in the syncopaondition. Perhaps, in the present experiment, the diffiyncopation task led to an increase of attention to compehe detrimental effects of SD, whereas in the aforementitudy of Monno et al.[24] attention was forced away frohe coordination task by means of a second dual task.

The effects of SD on MEG activity were twofold aonsistent with our expectations. First, across all moveempos, the distribution of MEG power changed markehe overall power decreased at occipital and temporal cels and increased at central and frontal channels (cf.[15]),

ere not restricted to a reduction of the N1m componenlso reduced later AEF component found at shorter ISIddition, we found a decrease in MEF amplitude. In gral, motor-related fields observed in motor tasks reflectotor outflow processes and sensory feedback, but se

esearch groups have shown that the largest MEF coent (MEF1) signifies sensory feedback from the perip

9,26,32]. Amplitude decreases after SD, similar to the oeen here in AEFs and MEFs, have also been reportevent-related potentials in the visual cortex[10] renderinghe fatigue-related drop in amplitude canonical and sughat SD might be related with a change of central procesf sensory input.

Besides effects of SD, we also found effects of stius presentation. The N1m component of the AEF hadargest amplitude at long ISIs and decreased progressith decreasing ISIs to transform into a later componen ISI of 500 ms. This effect has been studied in depth

s thought to display the slow refractoriness of the neuropulation producing the N1 responses[25]. Although theeparation of MEFs and AEFs appears quite successfuannot doubtlessly determine whether the reduction of thmplitude is caused either by attenuation of the undereural activity or by a change in timing-related morpholf this component. Indeed, it has recently been shown

he N1 amplitude also depends on motor-related activity[28],hich may have also affected the here reported reducti1m amplitude.In sum, the present experimental findings converge o

nterpretation that SD causes a reduced ‘responsivene

32 T.W. Boonstra et al. / Neuroscience Letters 388 (2005) 27–32

sensory areas to peripheral input. However, to examine thespecifics of this general conclusion, future studies are neededthat involve larger groups of subjects. Such studies may helpto uncover the neuronal mechanism(s) that link deactivationof frontal (or other) brain areas with the attenuation of ERFcomponents.

Acknowledgement

This project was financed in part by the Dutch ScienceFoundation (NWO grant #051.02.050).

References

[1] C. Andrew, G. Pfurtscheller, On the existence of different alphaband rhythms in the hand area of man, Neurosci. Lett. 222 (1997)103–106.

[2] G. Aschersleben, W. Prinz, Synchronizing actions with events:the role of sensory information, Percept. Psychophys. 57 (1995)305–317.

[3] M. Atienza, J. Cantero, C. Escera, Auditory information processingduring human sleep as revealed by event-related brain potentials,Clin. Neurophysiol. 112 (2001) 2031–2045.

[4] C. Babiloni, F. Carducci, F. Cincotti, P.M. Rossini, C. Neuper, G.Pfurtscheller, F. Babiloni, Human movement-related potentials vs.desynchronization of EEG alpha rhythm: a high-resolution EEG

vanhods

inhi-t. J.

e andates,

s introen-ysiol.

backnger

[ rci,tials

[ udy-tol,

[ vent-–302.

[ rioreuro-

[ ects

[15] C. Drake, J. Moran, H. Scofield, N. Tepley, T. Roth, MEG imagedbrain activation during the transition from wake to sleep, in: E.Halgren, S. Ahlfors, M. Hamalainen, D. Cohen (Eds.), Proceedingsof the 14th International Conference on Biomagnetism, Biomag 2004Ltd., Boston, MA, 2004, p. 211.

[16] S.P. Drummond, G.G. Brown, J.L. Stricker, R.B. Buxton, E.C. Wong,J.C. Gillin, Sleep deprivation-induced reduction in cortical functionalresponse to serial subtraction, Neuroreport 10 (1999) 3745–3748.

[17] M. Ferrara, L. De Gennaro, F. Ferlazzo, G. Curcio, Topographicalchanges in N1–P2 amplitude upon awakening from recovery sleepafter slow-wave sleep deprivation, Clin. Neurophysiol. 113 (2002)1183–1190.

[18] J.A. Horne, Human sleep, sleep loss and behaviour. Implications forthe prefrontal cortex and psychiatric disorder, Br. J. Psychiatry 162(1993) 413–419.

[19] J.A. Kelso, Phase transitions and critical behavior in human bimanualcoordination, Am. J. Physiol. 246 (1984) R1000–R1004.

[20] J.A.S. Kelso, S.L. Bressler, S. Buchanan, G.C. DeGuzman, M. Ding,A. Fuchs, T. Holroyd, A phase transition in human brain and behav-ior, Phys. Lett. A 169 (1992) 134–144.

[21] F.H. Lopes da Silva, J.E. Vos, J. Mooibroek, A. Van Rotterdam,Relative contributions of intracortical and thalamo–cortical processesin the generation of alpha rhythms, revealed by partial coherenceanalysis, Electroencephalogr. Clin. Neurophysiol. 50 (1980) 449–456.

[22] K.V. Mardia, Statistics of Directional Data, Vol. XX, AcademicPress, 1972, 357 pp.

[23] T. Mima, N. Simpkins, T. Oluwatimilehin, M. Hallett, Force levelmodulates human cortical oscillatory activities, Neurosci. Lett. 275(1999) 77–80.

[24] A. Monno, A. Chardenon, J.J. Temprado, P.G. Zanone, M. Laurent,ordi-rosci.

[ sen-859.

[ ticaltudy,

[ inElec-

[ ryer,otor-

[ n of

[ ilva,con-hys-

[ theNeu-

[ , A.kedClin.

[ ida,an

study, Neuroimage 10 (1999) 658–665.[5] T.W. Boonstra, H.E. Clairbois, A. Daffertshofer, J. Verbunt, B.W.

Dijk, P.J. Beek, MEG compatible force sensor, J. Neurosci. Met144 (2005) 193–196.

[6] K. Campbell, I. Colrain, Event-related potential measures of thebition of information processing: II. The sleep onset period, InPsychophysiol. 46 (2002) 197–214.

[7] J. Cantero, M. Atienza, C. Gomez, R. Salas, Spectral structurbrain mapping of human alpha activities in different arousal stNeuropsychobiology 39 (1999) 110–116.

[8] J. Cantero, M. Atienza, R. Salas, Human alpha oscillationwakefulness, drowsiness period, and REM sleep: different eleccephalographic phenomena within the alpha band, NeurophClin. 32 (2002) 54–71.

[9] D. Cheyne, H. Endo, T. Takeda, H. Weinberg, Sensory feedcontributes to early movement-evoked fields during voluntary fimovements in humans, Brain Res. 771 (1997) 196–202.

10] M. Corsi-Cabrera, C. Arce, I.Y. Del Rio-Portilla, E. Perez-GaM.A. Guevara, Amplitude reduction in visual event-related potenas a function of sleep deprivation, Sleep 22 (1999) 181–189.

11] A. Daffertshofer, C.J. Lamoth, O.G. Meijer, P.J. Beek, PCA in sting coordination and variability: a tutorial, Clin. Biomech. (BrisAvon) 19 (2004) 415–428.

12] A. Daffertshofer, C.E. Peper, P.J. Beek, Spectral analyses of erelated encephalographic signals, Phys. Lett. A 266 (2000) 290

13] L. De Gennaro, M. Ferrara, G. Curcio, R. Cristiani, Antero-posteEEG changes during the wakefulness–sleep transition, Clin. Nphysiol. 112 (2001) 1901–1911.

14] M. Doumas, P. Praamstra, A. Wing, Low frequency rTMS effon sensorimotor synchronization, Exp. Brain Res. (in press).

Effects of attention on phase transitions between bimanual conation patterns: a behavioral and cost analysis in humans, NeuLett. 283 (2000) 93–96.

25] R. Naatanen, I. Winkler, The concept of auditory stimulus repretation in cognitive neuroscience, Psychol. Bull. 125 (1999) 826–

26] M. Oishi, S. Kameyama, M. Fukuda, K. Tsuchiya, T. Kondo, Coractivation in area 3b related to finger movement: an MEG sNeuroreport 15 (2004) 57–62.

27] G. Pfurtscheller, A. Aranibar, Occipital rhythmic activity withalpha band during conditioned externally paced movement,troencephalogr. Clin. Neurophysiol. 45 (1978) 226–235.

28] P. Praamstra, M. Turgeon, C.W. Hesse, A.M. Wing, L. PerNeurophysiological correlates of error correction in sensorimsynchronization, Neuroimage 20 (2003) 1283–1297.

29] M. Rosenblum, A. Pikovsky, J. Kurths, Phase synchronizatiochaotic oscillators, Phys. Rev. Lett. 76 (1996) 1804–1807.

30] P. Suffczynski, S. Kalitzin, G. Pfurtscheller, F.H. Lopes da SComputational model of thalamo–cortical networks: dynamicaltrol of alpha rhythms in relation to focal attention, Int. J. Psychopiol. 43 (2001) 25–40.

31] E. Werth, P. Achermann, A.A. Borbely, Brain topography ofhuman sleep EEG: antero-posterior shifts of spectral power,roreport 8 (1996) 123–127.

32] H. Woldag, G. Waldmann, M. Schubert, U. Oertel, B. MaessFriederici, H. Hummelsheim, Cortical neuromagnetic fields evoby voluntary and passive hand movements in healthy adults, J.Neurophysiol. 20 (2003) 94–101.

33] N. Yamagishi, D.E. Callan, N. Goda, S.J. Anderson, Y. YoshM. Kawato, Attentional modulation of oscillatory activity in humvisual cortex, Neuroimage 20 (2003) 98–113.