Exp Brain Res (1999) 125:417–425
R E S E A R C H A RT I C L E
Margret Hund-Georgiadis · D. Yves von Cramon
Motor-learning-related changes in piano players
and non-musicians revealed by functional magnetic-resonance signals
Received: 11 August 1998 / Accepted: 23 November 1998
Abstract In this study, we investigated blood-flow-re-
lated magnetic-resonance (MR) signal changes and the
time course underlying short-term motor learning of the
dominant right hand in ten piano players (PPs) and 23
non-musicians (NMs), using a complex finger-tapping
task. The activation patterns were analyzed for selected
regions of interest (ROIs) within the two examined
groups and were related to the subjects’ performance. A
functional learning profile, based on the regional blood-
oxygenation-level-dependent (BOLD) signal changes,
was assessed in both groups. All subjects achieved sig-
nificant increases in tapping frequency during the train-
ing session of 35 min in the scanner. PPs, however, per-
formed significantly better than NMs and showed in-
creasing activation in the contralateral primary motor
cortex throughout motor learning in the scanner. At the
same time, involvement of secondary motor areas, such
as bilateral supplementary motor area, premotor, and cer-
ebellar areas, diminished relative to the NMs throughout
the training session. Extended activation of primary and
secondary motor areas in the initial training stage (7–14
min) and rapid attenuation were the main functional pat-
terns underlying short-term learning in the NM group;
attenuation was particularly marked in the primary motor
cortices as compared with the PPs. When tapping of the
rehearsed sequence was performed with the left hand,
transfer effects of motor learning were evident in both
groups. Involvement of all relevant motor components
was smaller than after initial training with the right hand.
Ipsilateral premotor and primary motor contributions,
however, showed slight increases of activation, indicat-
ing that dominant cortices influence complex sequence
learning of the non-dominant hand. In summary, the in-
volvement of primary and secondary motor cortices in
motor learning is dependent on experience. Interhemi-
spheric transfer effects are present.
M. Hund-Georgiadis (✉) · D.Y. von Cramon
Max-Planck-Institute of Cognitive Neuroscience,
Stephanstr. 1b, D-04103 Leipzig, Germany
e-mail: hund@cns.mpg.de
Tel.:+49-341-9940-161, Fax: +49-341-9940-221
© Springer-Verlag 1999
Key words Motor learning · Motor cortex · Magnetic
resonance imaging · Musicians · Hand motor skill
Introduction
Rehearsal of a novel motor task leads to improvement in
performance and accuracy. A number of recent imaging
studies, mainly performed with positron emission to-
mography (PET), have addressed the underlying brain
mechanisms of motor learning (Seitz 1990; Friston 1992;
Schlaug 1994). These have focused on assessment of re-
gional cerebral blood-flow changes of cortical and sub-
cortical motor areas during the course of practicing, i.e.,
changes in premotor and primary motor cortices as well
as supplementary motor area (SMA), cerebellar motor
areas, and basal ganglia.
Motor learning is the effort of acquiring a certain mo-
tor skill. Two processes account for successful learning
(Squire 1986): implicit (“how to do it”) and explicit
(“what is to be done”) knowledge. There is kinematic
and imaging evidence that both implicit and explicit
knowledge belong to anatomically different and, to a
certain extent, independently operating memory systems
(Pascual-Leone et al. 1994; Rauch et al. 1995; Hazeltine
et al. 1997). The transfer of knowledge from an implicit
to an explicit state has been demonstrated to be paral-
leled by changes in the patterns of cortical motor activa-
tion (Rauch et al. 1995). This functional plasticity of cor-
tical outputs is regarded as the underlying mechanism of
motor learning.
There is evidence from many imaging studies that the
plastic cortical changes during motor and non-motor
practicing occur rapidly. Raichle and colleagues (1994)
demonstrated that relevant brain areas – active in a non-
motor learning task – changed after less than 15 min of
practice. They concluded that the degree to which a task
is learned or automatized determines the cortical net-
work involved. Shadmehr and Holcomb (1997) showed
that the representation of a motor skill is reorganized
within 6 h after completion of practice, while perfor-
418
mance remains unchanged. Karni et al. (1995) studied
short-term learning in a brief time window of one contin-
uous learning session lasting between 30 min and 2 h.
He demonstrated that habituation, i.e., a decline of corti-
cal activation, is the main characteristic of short-term
learning.
Prior imaging research has suggested a strong rela-
tionship between neuronal activity and measurable indi-
ces of function: Schlaug et al. (1996) studied blood-
flow-related magnetic-resonance (MR) signal changes in
response to different rates of repetitive movements of the
index finger. The authors concluded that increased firing
of neuronal aggregates or recruitment of additional neu-
ronal units within the primary motor cortex are necessary
for increased output to target neurons. Jenkins et al.
(1997) found comparable results using PET techniques.
The complexity of the movement has been shown to de-
termine the functional activation patterns as well. Using
PET methods, Shibasaki et al. (1993) and Sadato et al.
(1997) found that complex finger movements recruit a
set of brain areas in addition to those areas underlying
the execution of simple movement sequences. These in-
clude greater cerebral blood-flow (CBF) increases in the
SMA during the performance of a complex task
(Shibasaki et al. 1993) and increasing involvement of the
right dorsal premotor cortex and the right precuneus with
increasing complexity (Sadato et al. 1997). Rao and co-
workers (1993) found comparable results employing
functional magnetic resonance (fMR-) techniques.
Previous PET-studies have focused on the nature and
the extent of brain reorganization in response to motor
learning based on different task types (Seitz et al. 1990;
Grafton et al. 1992; Friston et al. 1992; Jenkins et al.
1994; Schlaug et al. 1994). One study (Seitz et al. 1990)
demonstrated that motor learning of a complex tapping
task was accompanied by cerebral blood flow (rCBF) in-
creases in the cerebellum with increasing movement
rates. Significant activation of prefrontal, lateral premo-
tor, and cerebellar cortices were claimed to be the main
functional patterns of new motor learning, as opposed to
the execution of an overlearned sequence (Jenkins et al.
1994).
Certain human skills, such as musical ability, have
been associated with differences in brain structure and
function. There is a great deal of evidence from structur-
al anatomical and electrophysiological studies that the
representation area of primary cortices depends on prac-
tice and adapts in order to conform to the current needs
and experiences (Elbert et al. 1995; Schlaug et al. 1995;
Pantev et al. 1998). Hence, long-term motor training,
such as piano or string playing, should induce an exten-
sion of the representation area of primary cortices. Elbert
and colleagues (1995) demonstrated by means of magne-
toencephaolography (MEG) that the cortical sensory rep-
resentation area of the left-hand digits of professional
string players is more extended than that of untrained
controls. Pantev et al. (1998) also attributed the capacity
of use-dependent plasticity to the auditory cortex after
studying a group of non-musicians by means of MEG.
The work of Elbert and Pantev further argues for the ca-
pacity for plastic change in primary cortices. The authors
claim that use-dependent functional reorganization ex-
tends across sensory cortices to reflect the patterns of
sensory input that is processed by the individual during
the development of musical skill. Earlier findings in this
field are based mainly on structural and bioelectric imag-
ing studies of human (Elbert et al. 1995; Amunts et al.
1997) and non-human primates (Milliken et al. 1992;
Recanzone et al. 1992).
Recent research on motor learning suggests that
short-term and long-term motor learning depend strongly
on the experience of the individual subject, the task type
itself, and the time frame of learning. In the present
study, we investigated the process of motor practicing re-
lated to different manual skills by means of functional
magnetic resonance imaging (fMRI) to address the ques-
tion of a use-dependent organization in motor cortices. A
group of ten experienced piano players (PPs) and a con-
trol group of 23 age-matched non-musicians (NMs) were
examined during the performance of a sequential tapping
task.
The present experiment was designed to answer the
following questions:
1. Does pre-practice experience determine the MR-de-
fined motor learning process? The hypothesis is that
distinct primary and secondary motor cortices exhibit
characteristic activation patterns that can be related to
prepracticing experience. We were particularly inter-
ested in signal changes of M1, assuming – according
to previous work – that the primary motor cortex
could be an appropriate indicator of general motor ex-
perience, regardless of a singular motor sequence. Al-
ternatively, M1 activation changes could evolve pure-
ly as a result of long-term motor training in each sin-
gular tapping sequence (Karni et al. 1995). We hy-
pothesized that experienced PPs would show – due to
their experience – a larger M1-activation, regardless
of the training stage, than the controls.
2. How does performance interfere with MR signal
changes in the course of motor learning within the
two practicing groups?
3. Does motor training of the dominant hand transfer to
the non-dominant one?
Material and methods
Selection of control subjects
Thirty three healthy subjects (18 females and 15 males, mean age:
27.3, SD: 4.1, all right-handed) were included in the study. Ten
were experienced PP (daily training of more than 1 h for more
than 10 years). The control group consisted of 23 NMs, who could
not play any instrument or touch-type. Written informed consent
was obtained from all participants before they entered the study.
The Edinburgh handedness inventory was performed to deter-
mine the degree of hand dominance (group mean 81,5%±16%, no
significant difference was found in between groups, P=0.87, t-test
for unpaired samples).

Motor tapping task and instructions
All subjects were asked to practice a predefined motor sequence
with their dominant hand in the MR-scanner. The movement was
to be performed continuously and as rapidly and accurately as
possible without any external trigger, in alternation with a resting
stage. The subjects were instructed immediately prior to the fMRI
session. The task consisted of finger tapping opposed to the
thumb: 1–3, 1–5, 1–2, 1–4, 1–2, 1–2. During the 35-min practic-
ing session in the MR-scanner, the tapping frequency was con-
trolled via a video camera. The tapping rates across MR-scans
were acquired in time steps of 7 min by an independent observer.
Each subject executed a behavioral test of the tapping sequence,
both prior to MR-scanning and after the MR-session, to assess the
improvement due to motor training. The tapping frequency as well
as the number of errors were measured.
fMRI technique
fMRI data were collected using a 3 T-Bruker 30/100 Medspec
system (Bruker Medizintechnik, Ettlingen, Germany). A FLASH
sequence was chosen in all subjects, using TE=40 ms and TR=80
ms with a 128×64 matrix. This allowed acquisition of an image in
less than 6 s. 64 consecutive images were acquired per scan, alter-
nating between four images of rest and four images of motor train-
ing. Slice thickness was 5 mm with a 2-mm interslice gap. Three
axial slices were chosen at a distance of 3.5 cm above and parallel
to the AC-PC line. In order to obtain precise anatomical informa-
tion about the precentral region within subjects, considerable care
was taken to visualize the angle between the superior frontal sul-
cus and the precentral sulcus. Thus, the hemodynamics in the mid-
portion of the precentral gyrus could be recorded. This region
largely contains the primary motor cortex (Broca’s knee) and in-
cluded parts of the classically defined arm area of Penfield and
Boldrey (1937). It also covered relevant portions of the supple-
mentary motor area (SMA), going down to the cingulate sulcus
and the premotor cortices of both hemispheres. In addition, a
fourth slice was planned on the midsagittal slice parallel to the up-
per portion of the fourth ventricle, incompletely covering the cere-
bellar dentate nucleus and the basal ganglia. We will refer to MR-
signal intensity as increases in blood flow of M1, SMA, premotor,
and cerebellar regions for a movement versus the resting condi-
tion, taking into account that no full coverage of the anatomically
defined area was achieved. Rest and activation states were con-
trolled via ERTS software (Behringer 1995) with a steady red light
marking rest and a green light marking the training state.
Experimental protocol
fMRI-scanning consisted of a tapping-learning block of 35 min re-
corded in five consecutive scans, lasting seven minutes each, and
performed with the dominant right hand. After dominant-hand
practicing, a further scan acquired fMRI data during non-dominant
finger tapping. Subjects were asked to execute the learned se-
quence with their non-dominant hand for 7 min. The behavioral
test was executed immediately before and after the MR-session.
Data analysis
Functional MRI data were processed by using the in-house soft-
ware package BRIAN (Kruggel and Lohmann 1996). The data
were first motion corrected for possible motion-induced signal
variability in the transverse direction and then smoothed by a
Gaussian filter with a 3-voxel length to increase signal-to-noise ra-
tio. Statistical analysis was carried out by comparing voxelwise the
tapping period with the resting period. Bonferroni-corrected t-tests
(P<0.05) were used to find significant brain activation. For further
analysis, regions of interest (ROI) were selected anatomically and
characterized by size, volume, number of activated pixels, and in-
tensity of the blood-oxygenation-level-dependent (BOLD) signal
419
during the course of motor learning. Normalized activation sizes
were then calculated, using a connected region r of N voxel size of
the functional data set, which was determined as significantly acti-
vated according to the method of Friston et. al. (1992) on the basis
of a z-score threshold of 3.5. Using this region as a binary mask,
we reexamined the raw functional data set. A voxel i at timestep t
of this region was denoted as as v(i, s, t), whereby the index s cor-
responds to the stimulation state (0 for a baseline period and 1 for
an activated state). T was defined as the number of timesteps in this
experiment, T0 as the number of base-level timesteps, and T1 as the
number of timesteps in the activated level. The intensities of all
voxels in this region over all timesteps, separated by its activation
status were subsequently averaged according to:
T N
sbase = 1 ∑ ∑ v(i, s, t ) if s = 0,
T0 t i
Likewise,
T N
sact = 1 ∑ ∑ v(i, s, t ) if s = 1,
T1 t i
The activation norm n was then defined as:
n=sact–sbase
and the relative activation a as:
s −s
a = act base
sbase
The analysis was performed in each individual case and for the en-
tire subject group related to the training stage. Again, Bonferroni-
corrected t-tests (P<0.05) were used to test for differences in areal
activation. The functional results are presented as normalized-acti-
vation areas. The following ROIs were defined based on areal ac-
tivation found in the primary motor cortex (M1), the SMA, bilat-
eral premotor cortex, and primary somatosensory cortex as well as
in the cerebellar dentate nucleus.
Results
Behavioral task
All subjects improved their tapping skills significantly
within the 35-min training session (Fig. 1a): The mean
frequencies prior to and after training were 33/min±20 (7
errors) and 58/min±15 (5 errors), respectively (P<0.001,
paired t-test). PPs tapped significantly faster than NMs
(pre-training: 52±7 versus 26±12, post-training: 75±8
versus 53±14; P<0.001, t-test for two samples). Thus,
performance rates showed an improvement of 100% in
the NM group and of 41.5% in the PP group. The num-
ber of errors did not show significant differences pre-
and post-training between groups. During practicing in
the scanner, a linear increase of tapping rates was evi-
dent in PPs and NMs (Fig. 1b). Significant differences
within groups were found after 7, 21, and 28 min of
practicing (P<0.05, t-test for unpaired samples).
fMRI data analysis
The activation patterns during motor learning showed
dynamic changes in all subjects and exhibited significant
differences between the PP and the NM group.
Contralateral M1. The normalized activation area during
initial tapping did not show significant differences be-
420
Fig. 1 a Behavioral test of motor tapping. Mean group differences
and standard deviations in tapping frequency and number of errors
pre- and post-training are presented for non-musicians (NM) and
piano players (PP). b Behavioral test of motor tapping. Mean
group differences of performance rates during practicing in the
scanner are shown. n.s. Non-significant differences, * significant
differences
tween PP and NM group (Figs. 2a, 4a). During the
course of motor learning significant M1-attenuation only
occurred in the NM group with activation levels of
1.8±1.1 normalized area in the first seven minutes of
training and 1.2±1.0 normalized area after 35 min of
training (P<0.04, paired t-test). The PPs, however, exhib-
ited increasing M1 activation during motor learning
(1.9±1.0 normalized area after seven minutes seven min-
utes and 2.3±1.0 normalized area after 35 min of motor
training, P=0.149, paired t-test). In all but the first train-
ing stage (7 min) significant differences between groups
were assessed (P<0.05, t-test for unpaired samples).
Ipsilateral M1. Dynamic changes of activation as well as
significant group differences were also observed in the
ipsilateral M1 with considerably lower degrees of activa-
tion (NMs 0.24±0.39 and PPs 0.07±0.07 normalized ac-
tivation area, P<0.001, t-test for unpaired samples) as
compared to contralateral M1 (Fig. 2 a). During the time
course of learning, the NMs showed significant attenua-
tion effects of activation areas (0.24±0.39 after 7 min to
Fig. 2a–d Changes of functional magnetic-resonance (fMR) acti-
vation patterns during the time course of motor learning. a The
changes of normalized activation areas are given for the contralat-
eral primary motor cortex (cM1) and the ipsilateral primary motor
cortex (iM1) in both groups, non-musicians (NM) and piano play-
ers (PP). b Dynamic changes of the normalized activation areas in
the supplementary motor area (SMA) during motor learning.
c Time course of motor learning in terms of fMR-activation pat-
terns in ipsilateral premotor cortices (i_premot.) for both groups,
NM and PP. d Dynamic changes of normalized activation areas in
the contralateral premotor cortices (c_premot.)
 421

0.08±0.24 after 35 min, P<0.001, paired t-test). Sus-

tained small activations, however, were found in the PP
group throughout motor learning (0.07±0.07 after 7 min
to 0.08±0.09, P>0.05, paired t-test).

SMA. Involvement of the SMA was significantly less

marked in the PP as compared to the NM group. This
held true throughout the course of motor learning of the
right hand and during the tapping task with the left hand
(Figs. 2b, 4a). Declines of activation sizes during motor
learning from 7 min to 35 min of training time were ob-
served in both groups (PP 0.54±0.53 to 0.29±0.28 and
NM 0.97±0.72 to 0.48±0.49 normalized activation area,
P<0.5, t-test for paired samples). Statistically significant
group differences were evident in all training stages ex-
cept the last one (P<0.05, t-test for two samples).

Premotor cortex. During the time course of motor learn-

ing, the tapping task involved the premotor areas to a
different degree in the two examined groups. NM sub-
jects showed significantly higher degrees of activation in
ipsi- and contralateral premotor cortices during the first
7 min of motor learning (0.32±0.28 contralateral and
0.33±0.35 ipsilateral for NMs vs. 0.12±0.1 contralateral
and 0.06±0.1 ipsilateral for PPs; P=0.025 and P=0.014
for contra- and ipsilateral comparison, respectively, t-test
for unpaired samples). In the NM group, the activated
premotor areas showed an initial decline between 7 and
14 min, a subsequent increase between 14 and 21 min,
and finally resolved in a decrease between 21 and 35 min
(Fig. 2c, d). This was the case for both ipsi- and contra-
lateral premotor cortices. Less overall activation in the
premotor cortex was seen during motor learning in the
PP group. Whereas the same patterns of activation de-
scribed for NMs were also seen in the ipsilateral premo-
tor cortex of the PP group, they were absent in their con-
trolateral premotor cortex.
Somatosensory cortex. Similar patterns of activation
without significant group differences were observed in
both of the examined groups (NMs 0.67±0.99 normalized
area vs. PPs 0.54±55 normalized area, P>0.05, t-test for
unpaired samples). The time course of motor learning was
characterized by a decrease of activation in both groups,
showing an extended attenuation after 14 min of training
and sustained activation for the remaining training period.
Cerebellar activation. Ipsilateral cerebellar dentate nu-
cleus activation was involved in the beginning (7 min of
tapping) of motor learning in all subjects, with the ex-
ception of one PP. On the whole, significantly lower acti-
vations were detected in the PP group than in the NM
group (NMs 0.95±0.45 normalized area vs. PPs
0.57±0.16 normalized area; P<0.05, t-test for unpaired
samples; Figs. 3, 4b). In the last stage of motor learning,
a slight decline of activation was evident. This was more
pronounced in the NM than in the PP group (NMs
0.83±0.58 normalized area vs. PPs 0.48±0.26 normalized
area, P<0.05, t-test for unpaired samples). No significant
Fig. 3 a Mean activation changes in ipsi- and contralateral cere-
bellar dentate nucleus (i_cerebellar, c_cerebellar) are given in ini-
tial and advanced motor-training stages, after 7 and 35 min of tap-
ping time. b Comparison of normalized activation areas related to
the conditions dominant (dom.) and non-dominant (non-dom.) tap-
ping. The arrow marks significant differences within groups for
the following ROIs: PP: contralateral and ipsilateral primary mo-
tor cortex (cM1 and iM1), supplementary motor area (SMA), ipsi-
lateral premotor cortices (i_prem.), and somatosensory cortices
(S1); NM: SMA and somatosensory cortices. c_prem. Contralater-
al premotor cortices
group difference was evident between the final stage and
initial practicing within groups (P>0.05, t-test for paired
samples). Contralateral cerebellar activation was missing
in all PP subjects, while it was strictly associated with
learning of a complex finger tapping sequence in the
group of NMs. A statistically insignificant attenuation
occurred in this group with proceeding motor learning
(P>0.05, t-test for paired samples).
Non-dominant tapping. Tapping of the newly learned se-
quence with the non-dominant, i.e., left hand, showed
significant group differences in distinct regions of inter-
est. Thus, the contralateral M1 exhibited increased acti-
vation in the PP group when the non-dominant hand was
performing the tapping task as compared with dominant
hand tapping (group means: PPs vs. NMs: 2.3±0.5 vs.
1.5±0.6; P<0.05). However, significantly less involve-
ment was assessed in the contralateral premotor cortices
of the PPs compared with NMs (group means: PP vs.
NM: 0.08±0.1 vs. 0.37±0.3; P<0.05).
Comparable involvement of the ipsilateral motor cor-
tex was found in both training groups, i.e., the supposed
dominant M1 was activated equally, (NMs 0.26±0.33,
422

Fig. 4a, b Magnetic-resonance

(MR) activation patterns during
motor learning overlayed on
T1-weighted high-resolution
MR in the axial plane in a nor-
mal control subject (NM) and a
piano player (PP): functional
images were created by gener-
ating statistical z-maps with z-
values >3.5 on a single pixel
level, a=0.05. a The dynamic
changes of activation patterns
and the decreasing involvement
of the supplementary motor ar-
ea and premotor areas during
motor learning are presented in
one typical NM subject. In-
creasing primary motor cortex
activation and less contribu-
tions of secondary motor com-
ponents were the main charac-
teristics in the PP subject.
b Functional activation patterns
of non-dominant finger tapping
are presented in a typical NM
and PP subject. Dynamics of
cerebellar activation decreases
are shown in final as opposed
to initial training in both sub- b
jects

PPs 0.2±0.27; P>0.05, t-test for unpaired samples). No
further significant group differences were found related
to the other selected regions of interest, i.e., ipsilateral
premotor cortex, SMA, and somatosensory cortices.
However, NMs showed, as a tendency, more extended
activation areas than PPs.
Comparison of initial dominant and non-dominant
training within groups revealed, on the whole, less ex-
tended activation in the non-dominant tapping condition
(Fig. 3b). Three exceptions were evident: (1) in the PP
group, involvement of the contralateral primary motor
cortex was significantly more pronounced in non-domi-
nant tapping; (2) the ipsilateral M1 showed more extend-
ed activation during non-dominant tapping in the PP
group; and (3) in the NM group, non-dominant tapping
recruited, as a tendency, more extended contralateral pre-
motor activation than dominant tapping, but this differ-
ence failed to reach statistical significance.
A number of ROIs, however, showed significant dif-
ferences between the dominant and non-dominant tapping
conditions (Fig. 3b); PPs: contralateral and ipsilateral
M1, SMA, ipsilateral premotor cortices, and somatosen-
sory cortices; NMs: SMA and somatosensory cortices.
Discussion
This study was designed to examine local cortical chang-
es associated with short-term motor learning in 10 PPs
and 23 NMs related to different motor training experi-
ences. To the best of our knowledge, no other functional
study addressed short-term learning related to different
manual skills and experience. We demonstrated that PPs
– in contrast to NMs – already recruited larger M1 areas
at an early training stage and showed increasing M1-acti-
vation throughout the learning session with increasing
tapping rates: in the PP group, M1 enlargement was
5.3% after 7 min of training and 47.8% after 35 min of
training, compared with the cortical M1 representation
areas of the controls. Performance rates, however, in-
creased by 41.5% in the PP group and by 100% in the
NM group when pre-practicing and post-practicing be-
havioral tests were compared.
Rao et al. (1996) showed a direct positive proportion-
ality between movement rate and fMRI signal change in
the primary motor cortex. In accordance with this find-
ing, Seitz et al. (1990) described – in contrast to our ob-
servation – increased M1 peak activation even in normal
controls during the course of short-term motor learning
with increasing tapping rates. However, the length of
training periods differed in both studies (100 min vs. 35
min in our study). The question of whether the size of
M1-activation in our PP group purely depends on in-
creasing tapping rates remains to be addressed. Both
groups, PPs and NMs, practiced with increasing rates in
the scanner. NMs even showed shrinking M1 activation
with improving performance. We therefore conclude that
the influence of the tapping frequency does not provide a
sufficient explanation of our findings.
423
Enlargement of M1 activation is a known phenome-
non in long-term motor learning: Karni et al.(1995) re-
ported M1-expansion of about 25% in normal controls
due to practicing over several months. Based on normal
control studies, Karni et al.(1998) further proposed that
performance is acquired in several stages: “fast” learn-
ing, an initial, within-session improvement phase, is fol-
lowed by a period of consolidation of several hours dura-
tion. “Slow” learning, consists of delayed incremental
gains in performance emerging after continued practice.
This time course may reflect basic mechanisms of neuro-
nal plasticity in the adult brain, which subserve the ac-
quisition and retention of many different skills. We hy-
pothesize that these basic mechanisms of motor learning
are determined by pre-practice experience. The learning
profile of PPs in our study resembles – at least with re-
spect to M1-activation patterns – the plastic changes typ-
ical of “slow” learning described by Karni et al. (1998).
Hence, long-term motor experience may induce a slowly
evolving expansion of primary motor cortices. This has
already been shown for primary sensory (Elbert et al.
1995) and auditory cortices (Pantev et al. 1998).
Furthermore, our data demonstrate that experience
clearly influences the contribution of secondary motor ar-
eas in sequential motor learning. Our data reveal an over-
all minor SMA-involvement in the PP group and show at-
tenuation effects in both groups during motor learning,
regardless of the complexity of the task and increasing
tapping rates in both groups. Recently, the functional role
of the SMA has been much debated. It has been suggest-
ed that this area is involved in self-paced sequential mo-
tor tasks of non-visually controlled movements (Freund
1996). Shibasaki (1993) described greater rCBF increases
of the SMA during performance of a complex task than
during a simple movement task. The same finding was re-
ported by Rao et al. (1993) using fMRI-technique. Fur-
thermore, the authors detected differences that depended
on whether the movements were paced or unpaced, with
the latter resulting in larger SMA activation. Complexity
could be considered as an indicator of SMA-involvement
in our study as well. The performance data clearly
showed that the complexity of the tapping sequence has a
different effect in NMs and PPs. The underlying SMA-in-
volvement would then reflect the distinct effort of motor
practicing in both groups. Our findings are in contrast to
previous functional PET studies on new motor learning,
which report an increasing involvement of the SMA with
early motor learning (e.g., Grafton et al. 1992; Seitz et al.
1997). However, it must be noted that both PET studies
used different tasks (pursuit motor tasks/writing task).
The SMA has been hypothesized to be particularly en-
gaged when the execution of a sequential movement de-
pends on internally stored, memorized information (Tanji
et al. 1993). A differential role has been attributed to the
SMA in motor learning by Jenkins and colleagues (1994).
Automatic performance of an overlearned sequence re-
cruited more SMA activation than new learning of a mo-
tor task. The prefrontal cortex and the lateral premotor
cortex, however, were activated during new sequence
424
learning, but were not involved in performance of a pre-
learned sequence.
The exact time course of SMA-involvement in motor
learning is unknown and most likely escapes the tempo-
ral resolution of a fMRI-experiment. Supposing that the
switch towards an automatic performance occurs during
the first minutes of training in both groups, this would
reflect the highest SMA involvement after 7 min of
training in our data, followed by attenuation effects dur-
ing all other stages of motor learning. The behavioral da-
ta again suggest that the setting up of task-specific motor
processing occurs earlier in PPs than in NMs. Thus, the
first observation point at 7 min post-training in the PP
group may reflect an overlap of activations due to auto-
matic performance and due to attenuation with proceed-
ing practicing. The important influence of performance
on functional patterns has already been described by
Raichle et al. (1994). Shadmehr and coworkers (1997)
claimed that motor learning involves acquisition of an
internal model of the dynamics of the task, which gradu-
ally evolves along with the learning process and induces
a shift in the functional recruitment of the motor areas
involved after completing practice.
Again, the degree of cerebellar involvement in short-
term motor learning was influenced by the pre-practice
experience of the subjects. In particular, the involvement
of the contralateral cerebellum was a reliable marker of
the training stage. PPs did not recruit the contralateral
dentate nucleus during practicing, whereas it was regu-
larly involved in NM subjects. The ipsilateral cerebellum
showed significant increases of activation after 7 min of
practicing in the NM relative to the PP group. A slight,
statistically insignificant attenuation of cerebellar activa-
tion paralleled the process of motor training in all sub-
jects, regardless of the pre-training experience.
The role of the cerebellum in motor learning has been
a subject of much controversy. Right cerebellar activation
seems to be a marker of naive performance in contrast to
practiced performance in a non-motor learning task
(Raichle et al. 1994). A shift towards premotor, posterior
parietal and cerebellar structures has been associated with
proceeding motor learning (Shadmehr and Holcomb
1997). Our findings are consistent with observations of
Seitz et al. (1990), who demonstrated that, in the right an-
terior lobe of the cerebellum, the mean rCBF remained in
the same intensity from initial learning to skilled perfor-
mance, although the frequency and speed of finger move-
ments almost doubled. However, the training time and the
observation points in the PET scanner (50 min and 100
min post-training) differed from our task design. Our
findings contrast with observations of Friston and col-
legues (1992) on repeated performance of a simple motor
task in a time window similar to our task design. They
detected a decrease in activation with practice.
Two main findings were evident in non-dominant tap-
ping. Firstly, the PPs exhibited larger contra- and ipsilat-
eral activation in M1 areas than during dominant tap-
ping. Secondly, involvement of all other regions of inter-
est in both study groups was less pronounced in the non-
dominant tapping condition, with one exception: the con-
tralateral premotor cortex in the NM group. Our results
are consistent with a model of interhemisphric communi-
cation that is associated with intermanual transfer of
training and, thereby, emphasizes facilitatory mecha-
nisms (Parlow and Dewey 1991).
In contrast, Karni et al. (1995, 1998) claimed that cor-
tical plastic changes co-occurring with long-term learn-
ing do not generalize to the contralateral hand. Thut and
collegues (1997) reported that right-hand transfer learn-
ing after previous opposite-hand training even activated
neural mechanisms within the prefrontal cortex that
might have a transient inhibitory function. When the re-
hearsed sequence was performed with the left hand in
our study, contralateral cortical activation patterns re-
sembled those of short-term learning of the right hand in
advanced stages, i.e., less activation of primary and sec-
ondary motor areas was seen. Additionally, the ipsilater-
al primary M1 and premotor cortex of the dominant
hemisphere showed considerable involvement in the new
left-hand tapping in all subjects. This finding is in accor-
dance with Kawashima et al. (1993), who found that
movements of the non-dominant hand also elicited sig-
nificant ipsilateral increases in rCBF in the primary mo-
tor cortex and the premotor cortices.
The finding that PPs recruited larger ipsi- and contra-
lateral M1 during non-dominant tapping than during
dominant tapping is intriguing. The left motor cortex is
known to be substantially activated during ipsilateral
movements in left-handed subjects and even more so in
right-handed subjects (Kim et al. 1993). This supports
the concept of a hemispheric asymmetry in the function-
al activation of the human motor cortex during contralat-
eral and ipsilateral finger movements, especially in right-
handed subjects. As knowledge of the task has probably
been transferred into an explicit state during dominant
practicing, activation of non-dominant M1 underlies the
same mechanism as the dominant M1.
A limitation of this study was certainly that not all
motor areas were fully covered. Especially, the hemody-
namics of cerebellar motor areas and the basal ganglia
were incompletely recorded. A further drawback of the
study may be that behavioral data were acquired before
and after the fMRI session. Performance during the ses-
sion was controlled via video monitoring, but not contin-
uously recorded and evaluated on-line by EMG, due to
technical limitations in the scanner.
Conclusions
1. Learning of a complex finger tapping task with the
dominant hand exhibited involvement of different pri-
mary and secondary motor areas, depending on the
training stage.
2. Increasing activation of the contralateral M1 hand ar-
ea and small contributions of secondary motor areas,
such as premotor and somatosensory as well as cere-
bellar cortices, characterized the learning process in
PP subjects.

3. NM controls were characterized by recruitement of
ipsi- and bilateral primary and secondary motor re-
gions in initial tapping training and a decline in M1
activation across the training period.
4. Attenuation of all motor components accompanied
motor learning in the NM group. In the PP group,
however, no decrease of activation was observed in
ipsi- and contralateral M1 or in the contralateral pre-
motor cortex. Contralateral M1 even showed an in-
crease in activated pixels during the final stage of mo-
tor learning.
5. There is evidence of interhemispheric transfer effects
in motor sequence learning.
References
Amunts K, Schlaug G, Jäncke L, Steinmetz H, Schleicher A, Dab-
ringhaus A, Zilles K (1997) Motor cortex and hand motor
skills: structural compliance in the human brain. Hum Brain
Map 5:206–215
Behringer J (1995) Experimental run time system (ERTS). User’s
manual. BeriSoft Cooperation, Frankfurt am Main
Elbert T, Pantev C, Wienbruch C, Rockstroh B, Taub E (1995) In-
creased cortical representation of the fingers of the left hand in
string players. Science 270:305–307
Freund HJ (1996) Functional organization of the human supple-
mentary motor area and dorsolateral premotor cortex. In:
Lüders HO (ed) Advances in neurology, vol 70. Lippincott-
Raven, Philadelphia, pp 263–269
Friston KJ, Frith CD, Passingham RE, Liddle PF, Frackowiak RS
(1992) Motor practice and neurophysiological adaptation in
the cerebellum: a positron tomography study. Proc R Soc Lond
B Biol Sci 248:223–228
Grafton ST, Mazziotta JC, Presty S, Friston KJ, Frackowiak RS,
Phelps ME (1992) Functional anatomy of human procedural
learning determined with regional cerebral blood flow and
PET. J Neurosci 12:2542–2548
Hazeltine E, Grafton ST, Ivry R (1997) Attention and stimulus
characteristics determine the locus of motor-sequence encod-
ing. A PET study. Brain 120:123–140
Jenkins IH, Brooks DJ, Nixon PD, Frackowiak RSJ, Passingham
RE (1994) Motor sequence learning: a study with positron
emission tomography. J Neurosci 14:3775–3790
Jenkins IH, Passingham RE, Brooks DJ (1997) The effect of
movement frequency on cerebral activation: a positron emis-
sion tomography study. J Neurol Sci 151:195–205
Karni A, Meyer G, Jezzard P, Adams MM, Turner R, Ungerleider
LG (1995) Functional MRI evidence for adult motor cortex
plasticity during motor skill learning. Nature 377:155–158
Karni A, Meyer G, Rey-Hipolito C, Jezzard P, Adams MM, Turner
R, Ungerleider LG (1998) The acquisition of skilled motor
performance: fast and slow experience-driven changes in pri-
mary motor cortex. Proc Natl Acad Sci USA 95:861–868
Kawashima R, Yamada K, Kinomura S, Yamaguchi T, Matsui H,
Yoshioka S, Fukuda H (1993) Regional cerebral blood flow
changes of cortical motor areas and prefrontal areas in humans
related to ipsilateral and contralateral hand movement. Brain
Res 623:33–40
Kim SG, Ashe J, Hendrich K, Ellermann JM, Merkle H, Ugurbil
K, Georgopoulos AP (1993) Functional magnetic resonance
imaging of motor cortex: hemispheric asymmetry and handed-
ness. Science 261:615–617
Kruggel F, Lohmann G (1996) BRAIN (brain image analysis) – a
toolkit for the analysis of multimodal brain datasets. In:
Lemke HU, Inamura K, Vannier MW (eds) Proceedings of
computer aided radiology (International symposium on com-
puter and communication systems for image guided diagnosis
and therapy). Elsevier, Paris, pp 323–328
425
Milliken GW, Nudo RJ, Grenda R, Jenkins WM, Merzenich MM
(1992) Expansion of distal forelib representation in primary
motor cortex of adult squirel monkeys following motor train-
ing. Soc Neurosci Abstr 22:214.13
Pantev C, Oostenveld R, Engelien A, Ross B, Roberts LE, Hoke
M (1998) Increased auditory representation in musicians. Na-
ture 392:811–814
Parlow SE, Dewey D (1991) The temporal locus of training be-
tween hands: an interference study. Behav Brain Res 46:1–8
Pascual-Leone A, Grafman J, Hallett M (1994) Modulation of cor-
tical motor output maps during development of implicit and
explicit knowledge. Science 263:1287–1289
Penfield W, Boldrey E (1937) Somatic motor and sensory repre-
sentation in the cerebral cortex of man as studied by electrical
stimulation. Brain 60:389–443
Raichle ME, Fiez JA, Videen TO, MacLeod AM, Pardo JV, Fox
PT, Petersen SE (1994) Practice-related changes in human
brain functional anatomy during nonmotor learning. Cereb
Cortex 4:8–26
Rao SM, Binder JR, Bandettini PA, Hammeke TA, Yetkin FZ,
Jesmanowicz A, Lisk LM, Morris GL, Mueller WM, Estkow-
ski LD (1993) Functional magnetic resonance imaging of
complex human movements. Neurology 43:2311–2318
Rao SM, Bandettini PA, Binder JR, Bobholz JA, Hammeke TA,
Frost JA, Stein EA, Hyde JS (1996) Relationship between fin-
ger movement rate and functional magnetic resonance signal
change in human primary motor cortex. J Cereb Blood Flow
Metab 16:1250–1254
Rauch SL, Savage CR, Brown HD, Curran T, Alpert NM, Kendrick
A, Fischman AJ, Kosslyn SM (1995) A PET investigation of im-
plicit and explicit sequence learning. Hum Brain Map 3:271–286
Recanzone GH, Merzenich MM, Jenkins WM, Grajski KA, Dinse
HR (1992) Topographic reorganization in cortical area 3b of
owl monkeys trained in a frequency-discrimination task. J
Neurophysiol 67:1031–1056
Sadato N, Yonekura Y, Waki A, Yamada H, Ishii Y (1997) Role of
the supplementary motor area and the right premotor cortex in
the coordination of bimanual finger movements. J Neurosci
17:9667–9674
Schlaug G, Knorr U, Seitz RJ (1994) Inter-subject variability of
cerebral activations in acquiring a motor skill: a study with
positron emission tomography. Exp Brain Res 98:523–534
Schlaug G, Jäncke L, Huang Y, Steinmetz H (1995) In vivo evi-
dence of structural brain asymmetry in musicians. Science
267:699–701
Schlaug G, Sanes JN, Thangaraj V, Darby DG, Jancke L, Edelman
RR, Warach S (1996) Cerebral activation covaries with move-
ment rate. Neuroreport 7:879–883
Seitz RJ, Roland PE, Bohm C, Greitz T, Stone-Elander S (1990)
Motor learning in man: a PET study. Neuroreport 1:75–66
Seitz RJ, Canavan AG, Yaguez L, Herzog H, Tellmann L, Knorr
U, Huang Y, Homberg V (1997) Representations of grapho-
motor trajectories in the human parietal cortex: evidence for
controlled processing and automatic performance. Eur J Neu-
rosci 9:378–389
Shadmehr R, Holcomb HH (1997) Neural correlates of motor
memory consolidation. Science 277:821–825
Shibasaki H, Sadato N, Lyshkow H, Yonekura Y, Honda M, Nag-
amine T, Suwazono S, Magata Y, Ikeda A, Miyazaki M (1993)
Both primary motor cortex and supplementary motor area play
an important role in complex finger movement. Brain 116:
1387–1398
Squire LR (1986) Mechanisms of memory. Science 232:1612–161
Tanji J, Mushiake H, Masahiko I (1993) Premotor and supplemen-
tary motor cortex in sequential motor task. In: Ono T, Squire
LR, Raichle M, Perrett DI, Fukuda M (eds) Brain mechnan-
isms of perception and memory. Oxford University Press, Ox-
ford, pp 464–472
Thut G, Halsband U, Roelcke U, Nienhusmeier M, Missimer J,
Maguire RP, Regard M, Landis T, Leenders KL (1997) Inter-
manual transfer of training: blood flow correlates in the human
brain. Behav Brain Res 89:129–134