Learning by doing versus learning by thinking: An fMRI study of motor and mental training

Neuropsychologia 44 (2006) 711–717

Learning by doing versus learning by thinking: An fMRI studyof motor and mental training

Lars Nyberg a,∗, Johan Eriksson a, Anne Larsson b, Petter Marklund a

a Department of Psychology, Umea University, S-901 87 Umea, Swedenb Department of Radiation Sciences, Umea University, Umea Sweden

Received 30 June 2004; received in revised form 16 August 2005; accepted 16 August 2005Available online 7 October 2005

Abstract

Previous studies have documented that motor training improves performance on motor skill tasks and related this to altered functional brainactivity in cerebellum, striatum, and frontal motor cortical areas. Mental training can also improve the performance on motor tasks, but the neuralbasis of such facilitation is unclear. The purpose of the present study was to identify neural correlates of training-related changes on a finger-tappingtask. Subjects were scanned twice, 1 week apart, with fMRI while they performed two finger-tapping sequences with the left hand. In-betweens(sccvi©

K

1

ahfTMrshi(ft

0d

cans, they practiced daily on one of the sequences. Half of the participants received motor training and the other half received mental trainingmotor imagery). Both training procedures led to significant increases in tapping performance. This was seen for both the trained and the untrainedequence (non-specific effect), although the gain was larger for the trained sequence (sequence-specific effect). The non-specific training effectorresponded to a reduction in the number of activated areas from an extensive set of brain regions prior to training to mainly motor cortex anderebellum after training. The sequence-specific training effect involved the supplementary motor area and the cerebellum for motor training andisual association cortex for mental training. We conclude that gains following motor and mental training are based on distinct neuroplastic changesn the brain.

2005 Elsevier Ltd. All rights reserved.

eywords: fMRI; Motor skill learning; Mental training; Cerebellum; SMA; Imagery

. Introduction

Learning by doing may be the most effective way of learningnew motor skill but it is not the only way. Controlled studiesave demonstrated that mental practice leads to improved per-ormance on tests of motor skill (e.g., Feltz & Landers, 1983).he neural bases for such improvements are not well understood.otor learning has been shown to be associated with training-

elated changes in several brain regions, notably cerebellum,triatum, and frontal motor cortical areas (e.g., Doyon, Pen-une, & Ungerleider, 2003). A recent study of mental-trainingnduced strength gains included EMG and EEG recordingsRanganathan, Siemionow, Liu, Sahgal, & Yue, 2004). It wasound that strength gains following mental training were relatedo elevated cortical EEG potentials, but the EMG indicated that

∗ Corresponding author. Tel.: +46 907 866 429; fax: +46 907 866 695.E-mail address: [email protected] (L. Nyberg).

the resulting signal did not go down to the muscle level. Based onthese observations, Ranganathan et al. proposed that the trainingaffected higher-order motor cortical regions, such as supple-mentary motor and prefrontal regions. In turn, these areas caninfluence primary motor areas, and there is some evidence thatmental training actually can affect primary motor cortex (e.g.,Pascual-Leone et al., 1995).

The purpose of the present study was to identify neural cor-relates of training-related changes on a finger-tapping task. Ourprocedure was patterned after a previous study of motor train-ing (Karni et al., 1995). During the first of two fMRI session theparticipants performed finger tapping according to two differentnovel sequences. A second identical fMRI session followed after1 week. In the time in-between sessions, the participants receiveddaily training on one of the sequences. Half went through amotor-training program, whereas the other half received men-tal training. The mental training involved visualization (motorimagery). For both groups we evaluated gains in tapping perfor-mance after compared to before training for both the trained and

028-3932/$ – see front matter © 2005 Elsevier Ltd. All rights reserved.

712 L. Nyberg et al. / Neuropsychologia 44 (2006) 711–717

untrained sequence (non-specific training effect). In addition,and of main concern, we evaluated sequence-specific trainingeffects by contrasting the trained and untrained sequences aftercompletion of the training program.

2. Materials and methods

2.1. Subjects

Sixteen young, neurologically healthy subjects participated in the study.They were between 24 and 37 years old. All participants were right handed by selfreport and had normal or corrected to normal vision. They were randomly dividedinto one of two groups; motor (four women and four men, mean age = 29.9) ormental (three women and five men, mean age = 29.9). The study was approvedby the ethical committee at the University Hospital of Northern Sweden and allparticipants gave informed consent prior to participation.

2.2. Procedure

The first part of the study consisted of the initial fMRI-session. Prior toscanning, the subjects were instructed that the left index to little fingers werenumbered from 1 to 4 and that they were to perform finger tapping sequencesas fast and accurate as possible according to visually presented sequences. Theywere also told that when series of “x x x x x” were presented they should rest andlie still. In the scanner two different sequences (A = 4 1 3 2 4; B = 4 2 3 1 4) werepresented in a blocked fashion intermixed with blocks of x’s (block length = 30 s;block order = A-X–B-X, repeated three times per subject).

After the first scanning session, the participants received four daily sessionsohIiwhfsetfit

tIt

2

Slgr6iwp

smhw

tww

A program was written in Matlab (Mathworks Inc., MA, USA) to count thenumber of correct sequences performed by each participant, where a correctsequence was defined as a complete motor replicate of the visually presentedsequence (e.g., 4 1 3 2 4).

2.4. Statistical analyses

The fMRI images were transferred to a PC and converted to Analyze for-mat using the program MRIcro (Rorden & Brett, 2000). The data was thenpre-processed and analysed using SPM2 (Wellcome Department of CognitiveNeurology, London, UK) implemented in Matlab 6.5.1 (Mathworks Inc., MA,USA). The pre-processing steps were: slice timing correction, realignment withrespect to the first image volume in the series, unwarping to reduce residualmovement related variance, normalisation to an EPI template in the MontrealNeurological Institute (MNI) space, and smoothing with an isotropic 8.0 mmGaussian filter kernel. Single-subject statistical contrasts were then set up usingthe general linear model. The blocks of finger tapping and rest were modelled asfixed response (box-car) waveforms convolved with the hemodynamic responsefunction (HRF). Statistical parametric maps (SPMs) were generated using t-statistics to identify regions activated according to the model for individualsubjects, and random effects analyses were then used to reveal results for thewhole group.

Finger tapping performance was analyzed as a series of planned comparisons(one-tailed within-subjects t-tests). To test for non-specific training effects, tap-ping performance after training was contrasted with tapping performance beforetraining for both the trained and untrained sequences in each group. To testfor sequence-specific training effects, tapping performance after training wascontrasted for the trained versus untrained sequence. Two outliers with veryfew performed sequences were identified in the motor group for the untrainedsequence after training, possibly signalling a registration failure. These werertd

3

3

srpfpnfa

Fa

f individual training on one of the sequences. In each group (motor and mental),alf of the subjects practiced on sequence A and the other half on sequence B.n the motor group, training consisted of tapping the left-hand fingers on a tablen front of the subject according to the specified sequence (the sequence wasritten on a paper that was placed on the table). To prevent visual feedback theand was concealed with a cardboard box. The subjects tapped for 90 s, restedor 60 s, and then tapped again. They tapped a total of four times (6 min) peression (a total of 16 times or 24 min across sessions). The mental group followedxactly the same training program with the exception that they were instructedo visualize that they performed the finger-tapping sequence. To prevent actualnger movements they had their left and right hand fingers crossed, visible on

he table.One week after the first scanning session the subjects were again placed in

he scanner and they performed an identical run as during the first fMRI session.mportantly, at the time of the second session, they had practiced on one of thewo sequences that were presented and performed.

.3. MRI methods

Scanning was conducted on a Philips Intera 1.5 T system (Philips Medicalystems, Netherlands), equipped for echo-planar imaging (EPI). Blood-oxygen

evel-dependent contrast images were acquired using a T2*-weighted single-shotradient echo EPI sequence with the following parameters: echo time: 50 ms,epetition time: 3000 ms, flip angle: 90◦, field of view: 22 × 22 cm, matrix size:4 × 64 and slice thickness: 4.4 mm. Thirty-three transaxial slices positioned tonclude the whole brain volume were acquired every 3.0 s. Five “dummy scans”ere run before the image acquisition started to avoid signals resulting fromrogressive saturation.

The finger tapping sequences were presented visually on a semi-transparentcreen at the end of the scanner bore, which the subject could view via a mirrorounted on the head coil. Cushions inside the head coil were used to reduce

ead movement, and headphones (Silent ScanTM SS-3100, Avotec, FL, USA)ere used to dampen the scanner noise.

The finger tapping task was executed with a four-button response pad (Lumi-ouch reply system, Lightwave Medical Industries, Canada). The response padas connected to a computer running the program E-prime (Psychology Soft-are Tools, PA, USA) which registered the responses.

emoved from the analyses (if they had been included this would have servedo magnify the training effect in the motor group, i.e., an effect in the expectedirection).

. Results

.1. Finger tapping

On average, the participants performed 26.7 correctequences before training (range = 24–28). The proportion cor-ectly performed sequences was 88% of the total number oferformed sequences. The mean increase in finger tapping per-ormance after training is shown in Fig. 1. At this session, theroportion correctly performed sequences was 90% of the totalumber of performed sequences. Thus, the vast majority of per-ormed sequences before as well as after training were correctnd only a minority represented incorrect responses. In both

ig. 1. Mean increases in finger-tapping performance as a function of sequencend group. S.E. = standard error of the mean increase.

L. Nyberg et al. / Neuropsychologia 44 (2006) 711–717 713

Fig. 2. Group-averaged non-specific training effect for all conditions and sequences ((A) motor training/trained sequence; (B) motor training/untrained sequence;(C) mental training/trained sequence; (D) mental training/untrained sequence). Before training (left column), bilateral regions of motor and parietal cortex and leftcerebellum were activated relative to baseline. After training (right column), activity was restricted to contralateral (right) motor cortex and ipsilateral cerebellum.Activated cerebral regions are visualized on a dorsal cortical render image (anterior pointing upwards) from SPM2. Cerebellar regions are visualized on a cerebellaranatomical image from Caret (http://brainmap.wustl.edu/caret; Van Essen et al., 2001). (A), (C) and (D) show the dorsal cerebellar surface; (B) shows the ventralsurface. All images were thresholded at p < 0.01 (FDR corrected, k > 10), except for (C) and (D) after training for which p < 0.05 (FDR corrected). The peak coordinatesfor activated regions are listed in Table 1.

groups, the performance increased significantly after trainingfor the trained as well as the untrained sequence (all t’s >5.6;all p’s <0.01). That the training effect generalized to untrainedsequences provided evidence for a non-specific trainingeffect.

The increase for the untrained sequences was about 15sequences (Fig. 1), whereas the increase for the trained sequencewas significantly higher in both groups (motor: t(5) = 2.43,p < 0.05; mental: t(7) = 1.85, p = 0.05). This provided evidencefor a sequence-specific training effect.

714 L. Nyberg et al. / Neuropsychologia 44 (2006) 711–717

Table 1Peak coordinates (x, y, z in MNI space) for regions that were consistently activated before or after training for both the trained and untrained sequences (see Fig. 2)

Region Sequence Motor condition Mental condition

Before After Before After

Right motor cortex Trained 40, −16, 54 38, −14, 54 34, −18, 62 32, −28, 56Untrained 42, −16, 60 34, −18, 64 54, −18, 50 38, −32, 64

Left motor cortex Trained −40, 4, 40 – −34, 0, 66 –Untrained −28, −2, 70 – −46, −6, 58 –

Left parietal cortex Trained −28, −58, 56 – −24, −62, 52 –Untrained −30, −58, 58 – −26, −66, 54 –

Left cerebellum Trained −12, −52, −26 −6, −76, −42 −20, −64, −20 −4, −70, −18Untrained −14, −54, −18 −8, −68, −44 −16, −64, −20 −2, −68, −14

3.2. fMRI data—non-specific training effect

Before training, in comparison with the baseline condition,finger tapping was associated with increased brain activity inan extensive set of regions (Fig. 2, Table 1). These includedbilateral sensory-motor cortex, parietal cortex, and cerebellum.After training, by contrast, the activation pattern was much morerestricted (Fig. 2, Table 1), and included right motor cortexand left cerebellum. Thus, after training, the neural correlatesof finger tapping were restricted to regions related to left handmovement. This change could be related to a non-specific train-ing effect as it was seen for both the trained and untrainedsequences.

3.3. fMRI data—sequence-specific training effects

The trained and untrained sequences were directly comparedafter training (fMRI session 2). Previous studies have relatedtraining-induced changes to increased magnitude of activityin certain voxels and also to increased size of activated clus-ters (e.g., Deiber et al., 1998). We therefore examined whethertraining had led to changes in the intensity or peak height ofactivations and to changes in the spatial extent of activations.First, the SPMs were thresholded with regard to peak height(p < 0.001 uncorrected). In the motor group, differential acti-vFtotrUii

ouiaapb

were then entered into a random-effects (group) t-contrast toexamine regional activity differences after compared to beforetraining. Significant t-values indicate that a differential responseof a given brain region during performance of the trained relativeto the untrained sequence was greater after than before training(i.e., a sequence-specific training effect).

In the motor group, the post-pre comparison revealedincreased post-training activity in approximately the same SMA(x, y, z = 8, 2, 78; t(7) = 5.53) and cerebellum (x, y, z = 48,−72, −20; t(7) = 5.59) areas as were identified in the preced-ing analysis of post-training data. In the mental group, thepost-pre comparison identified a region in right occipital cor-tex (x, y, z = 12, −68, 6; t(7) = 11.19). This region was inclose proximity to the region identified in the analysis of post-training data (BA 18 in both analyses). However, in the post-pre comparison, SMA activation was not seen in the mentalgroup.

Together, the above analyses converged in showing asequence-specific training effect in the SMA and right cerebel-lum following motor training, and in secondary visual cortexfollowing mental training. In Fig. 4, percentage signal changein these regions for the trained versus untrained sequence isplotted for pre- and post-training data. As can be seen, thereare no significant activation increases pre-training but a robustpost-training increase in all three regions.

3

cbAtswwrwapet

ation was observed in the supplementary motor area (SMA;ig. 3). In the mental group, increased activity was observed in

he SMA and secondary visual cortex (Fig. 3). The spatial extentf these activations exceeded 35 voxels. Next, the SPMs werehresholded with regard to the extent of activations (p < 0.05 cor-ected for multiple comparisons with a peak threshold of 0.01).sing this statistical criterion, differential activity was revealed

n the cerebellum in the motor group (Fig. 3) whereas no signif-cant activation changes were observed in the mental group.

The preceding analyses were based on post-training contrastsf trained and untrained sequences. These analyses are thereforeninformative with regard to potential differences prior to train-ng. To formally test whether the observed differences in brainctivity were selectively expressed post-training, we conducteddditional analyses (post-pre comparisons). For each subject, weerformed t-contrasts of the trained versus untrained sequencesefore as well as after training. The resulting contrast images

.4. Control analyses—rate effects

It is known that changes in motor performance (rate effects)an affect the results in studies of learning-related changes inrain activity (e.g., Riecker, Wildgruber, Mathiak, Grodd, &ckermann, 2003). Since more sequences were performed after

raining for the trained than the untrained sequence, it is pos-ible that some or all of the corresponding activation changesere driven by rate effects. In an attempt to address this issue,e identified brain regions that were sensitive to finger tapping

ate and found that these regions were distinct from those inhich training-related effects were seen. Specifically, we took

dvantage of substantial between-individual variability in tap-ing rate in session 1 (i.e., prior to when any training-relatedffect could have occurred). We divided our participants intowo groups of eight individuals each that differed with regard

L. Nyberg et al. / Neuropsychologia 44 (2006) 711–717 715

Fig. 3. Group-averaged sequence-specific training effect (trained > untrained). Following motor training, increased activity was observed in SMA (A: x, y, z = 4, 0,78) and cerebellum (D: x, y, z = 42, −74, −22; 0, −78, −8). Following mental training, increased activity was observed in SMA (B: x, y, z = 0, 4, 56) and visualcortex (C: x, y, z = 4, −82, 14). Differential activity is visualized on a single-subject anatomical MRI image from SPM2.

to tapping rate: mean number of taps/session in the high-rategroup = 369 (S.D. = 87), mean number of taps/session in the low-rate group = 231 (S.D. = 11). A between-group contrast (highversus low) of brain activity during finger tapping was conducted

Fig. 4. Plots of percent signal change in the activated regions when the trainedand untrained conditions were contrasted before and after the training period(changes in SMA and cerebellum after motor training; changes in visual cortexafter mental training). The plot is based on the average regional activity, withregion defined as the entire cluster of activated voxels after training (see Fig. 3for illustration of activated regions and x, y, z coordinates).

to reveal regions for which activity was modulated by tappingrate.

In the main analysis of peak activity changes, training-relatedeffects were observed in the SMA and in visual cortex. In the rateanalysis, no SMA activity differences were observed at the samethreshold but a differential effect was seen in visual cortex (x, y,z = 14, −86, 2). The location of this effect was in primary visualcortex (BA 17, lingual gyrus), a site which previously has beenshown to be sensitive to rate effects (e.g., Kwong et al., 1992).By contrast, the learning-related effect in visual cortex in themain analyses was located to secondary visual cortex (x, y, z = 4,−82, 14; 12, −68, 6, BA 18). Thus, whereas the effect in primaryvisual cortex likely can be related to a visual rate effect (e.g., howmany times the numbers were read), the alteration in secondaryvisual cortex after mental training is unlikely to have been causedby rate effects. Further support for this interpretation comes fromthe fact that no differential visual activity was observed aftermotor training, although the tapping rate was higher followingsuch training than after mental training.

In the main analysis of training-related effects on the size ofactivated clusters, a differential effect was seen in the right lat-eral and medial cerebellum. In the rate analysis, using the samestatistical threshold, a difference was observed in left cerebel-lum (x, y, z = −18, −50, −38). Tapping was done with the left

716 L. Nyberg et al. / Neuropsychologia 44 (2006) 711–717

hand, i.e., ipsilateral to the cerebellar rate effect. This resultsis consistent with previous findings of ipsilateral rate effects inthe cerebellum during motor performance (see Riecker et al.,2003), and it indicates that the main effect in right (contralat-eral) cerebellum after motor training was not contaminated byrate effects.

4. Discussion

4.1. Modulation of finger-tapping performance by training

The behavioural results provided evidence that mental train-ing can improve the performance on a motor task (finger tap-ping). In part, the training effect reflected a general facilitationof tapping performance that influenced performance on bothtrained and untrained sequences. A cognitive component thatcould contribute to such a general effect is learning to associatethe fingers with their assigned numbers. That should facili-tate performance on trained as well as untrained sequences.Importantly, however, in both groups the training-related gainin tapping performance was greater for the sequence that hadbeen trained. This sequence-specific effect should more directlyreflect motor-skill learning. Numerically, the gain was largerin the motor group, but our observation of a robust sequence-specific effect in the mental group suggests that mental train-ing can promote skill acquisition as well. This conclusion isidc2

4

reetblrwbrolidb1

iSmStt

learning, and SMA is proposed as one site where sequencesare represented. Following motor training, a sequence-specificeffect was also observed in the cerebellum. This effect shouldreflect expression rather than acquisition of a motor skill.A previous study specifically associated cerebellum with theexpression of a learned motor skill (Seidler, Purushotham,Kim, Ugurbil, Willingham, & Ashe, 2002). On basis of find-ings that the cerebellum is more associated with motor exe-cution than motor imagery (Deiber et al., 1998; Nyberg etal., 2001), it might be speculated that the cerebellum was notengaged during mental training in-between the two fMRI scan-ning sessions and therefore not recruited during post-trainingexpression. Here, it must be stressed, though, that it is possi-ble that cerebellum can contribute to training effects follow-ing motor imagery, but the effect may be more subtle thanafter motor training and perhaps moderated by individual dif-ferences in kinds of motor imagery. Future studies involvingmore training, more subjects, and more constrained imageryinstructions will be needed to more conclusively address thisissue.

Mental training resulted in a sequence-specific effect in sec-ondary visual cortex (BA 18). This part of the brain has pre-viously been found to be active during various imagery-basedtasks (Cabeza & Nyberg, 2000). The mental training procedureinvolved imagination of performed finger sequences. It is possi-ble that this activity resulted in the formation of a visual memorytpm(

4

ocaciIwapobntscoC

A

(s

n agreement with several prior findings (e.g., Feltz and Lan-ers, 1983; Pascual-Leone et al., 1995), and we have repli-ated the behavioural pattern in an independent study (Olsson,004).

.2. Brain correlates of motor and mental training

The non-specific training effect was reflected in the fMRIesults as a restriction of the set of brain regions that wasngaged during tapping. Before training, the task recruited sev-ral regions in addition to those that most strongly are relatedo left-hand movements (i.e., right motor cortex and left cere-ellum). These additional pre-training regions included ispi-ateral sensori-motor cortex and parietal cortex. After training,ight motor cortex and left/medial cerebellum were the regionsith the strongest activity increases. Thus, with training, therain responses associated with the finger-tapping task came toesemble those associated with simple finger movements. Thebserved transformation of activation patterns as a function ofearning is in keeping with stage models of sequence learn-ng that propose that association areas are dominant structuresuring the early stages of learning, whereas motor structuresecome dominant structures at later stages (Hikosaka et al.,999).

For motor training the sequence-specific effect involvedncreased activity in the supplementary motor area (SMA). TheMA region has in several previous studies been associated withotor skill acquisition (e.g., Hikosaka et al., 1999). A role forMA might be related to the formation of memory represen-

ations for sequences. In the model of Hikosaka et al. (1999)he SMA is associated with later stages of motor sequence

hat influenced task-performance after training. In support of thisossibility, cognitive training effects related to a visual imagerynemonic have been found to implicate occipito-parietal cortex

Nyberg et al., 2003).

.3. Potential broader implications

The results of this study indicate that the neural correlatesf training gains following motor or mental training differ, witherebellum more strongly associated with the former and visualssociation cortex with the latter. Thus, although performancean be enhanced by motor as well as mental training, it seems asf this enhancement relies on at least partly different brain areas.f true, a potential implication of this finding is that an athleteho is using mental training as a booster or as a way of keeping

ctive during a period of injury might effectively introduce a newrocedure with its distinct neural correlates. The consequencesf this might be positive, for example because performance cane based on both motor and non-motor programs. However, theet result can conceivably also be negative, for example dueo interference effects. At the very least, athletes and coacheshould be aware that the brain bases of motor and mental trainingould be more different than might have been expected on basisf similarities between motor imagery and motor execution (seerammond, 1997).

cknowledgements

This research was supported by a grant to L.N. from CIFCentrum for Idrottsforskning). We thank Micael Andersson forupport with statistical analyses.

L. Nyberg et al. / Neuropsychologia 44 (2006) 711–717 717

References

Cabeza, R., & Nyberg, L. (2000). Imaging Cognition II: An empirical reviewof 275 PET and fMRI studies. Journal of Cognitive Neuroscience, 12,1–47.

Crammond, D. J. (1997). Motor imagery: never in your wildest dreams.Trends in Neurosciences, 20, 54–57.

Deiber, M.-P., Ibanez, V., Honda, M., Sadato, N., Raman, R., & Hallett, M.(1998). Cerebral processes related to visuomotor imagery and generationof simple finger movements studied with positron emission tomography.NeuroImage, 7, 73–85.

Doyon, J., Penhune, V., & Ungerleider, L. G. (2003). Distinct contribution ofthe cortico-striatal and cortico-cerebellar systems to motor skill learning.Neuropsychologia, 41, 252–262.

Feltz, D. L., & Landers, D. M. (1983). The effects of mental practice onmotor skill learning and performance: A meta-analysis. Journal of SportsPsychology, 5, 25–27.

Hikosaka, O., Nakahara, H., Rand, M. K., Sakai, K., Lu, X., Nakamura, K.,et al. (1999). Parallel neural networks for learning sequential procedures.Trends in Neurosciences, 22, 464–471.

Karni, A., Meyer, G., Jezzard, P., Adams, M. M., Turner, R., & Ungerleider,L. G. (1995). Functional MRI evidence for adult motor cortex plasticityduring motor skill learning. Nature, 377, 155–158.

Kwong, K. K., Belliveau, J. W., Chesler, D. A., Goldberg, I. E., Weisskoff, R.M., Poncelet, B. P., et al. (1992). Dynamic magnetic resonance imagingof human brain activity during primary sensory stimulation. Proceedingsof the National Academy of Sciences, USA, 89, 5675–5679.

Nyberg, L., Petersson, K. M., Nilsson, L.-G., Sandblom, J., Aberg, C., & Ing-var, M. (2001). Reactivation of motor brain areas during explicit memoryfor actions. NeuroImage, 14, 521–528.

Nyberg, L., Sandblom, J., Jones, S., Stigsdotter Neely, A., Petersson, K.M., Ingvar, M., et al. (2003). Neural correlates of training-related mem-ory improvement in adulthood and aging. Proceedings of the NationalAcademy of Sciences, USA, 100, 13728–13733.

Olsson, C. -J. (2004). Imagery training improves performance on a fingertapping task. Unpublished Master’s Thesis, Department of Psychology,Umea University.

Pascual-Leone, A., Nguyet, D., Cohen, L. G., Brasil-Neto, J. P., Cammarota,A., & Hallett, M. (1995). Modulation of muscle responses evoked bytranscranial magnetic stimulation during the acquisition of new fine motorskills. Journal of Neurophysiology, 74, 1037–1045.

Ranganathan, V. K., Siemionow, V., Liu, J. Z., Sahgal, V., & Yue, G. H.(2004). From mental power to muscle power—gaining strength by usingthe mind. Neuropsychologia, 42, 944–956.

Rao, S. M., Bandettini, P. A., Binder, J. R., Bobholz, J. A., Hammeke, T. A.,Stein, E. A., et al. (1996). Relationship between finger movement rateand functional magnetic resonance signal change in human primary motorcortex. Journal of Cerebral Blood Flow and Metabolism, 16, 1250–1254.

Riecker, A., Wildgruber, D., Mathiak, K., Grodd, W., & Ackermann, H.(2003). Parametric analysis of rate-dependent hemodynamic responsefunctions of cortical and subcortical brain structures during auditorilycued finger tapping: a fMRI study. NeuroImage, 18, 731–739.

Rorden, C., & Brett, M. (2000). Stereotaxic display of brain lesions.Behavioural Neurology, 12, 191–200.

Seidler, R. D., Purushotham, A., Kim, S.-G., Ugurbil, K., Willingham, D.,& Ashe, J. (2002). Cerebellum activation associated with performancechange but not motor learning. Science, 296, 2043–2046.

Van Essen, D. C., et al. (2001). An integrated software system for surface-based analyses of cerebral cortex. Journal of American Medical Infor-matics Association, 41, 1359–1378.