OUP user menu

Inhibitory control of acquired motor programmes in the human brain

Friedhelm Hummel, Frank Andres, Eckart Altenmüller, Johannes Dichgans, Christian Gerloff
DOI: http://dx.doi.org/10.1093/brain/awf030 404-420 First published online: 1 February 2002


An important basis of skilled human behaviour is the appropriate retrieval of acquired and memorized motor programmes (‘motor memory traces’). Appropriate retrieval is warranted if motor programmes are only activated if necessary and are, probably more often, inhibited if required by the context of a given situation. It is unknown how this type of inhibition is accomplished in the brain. We studied context‐dependent modulation of motor memory traces in 18 volunteers and six patients with focal dystonia. Cortical function was assessed with transcranial magnetic stimulation over the primary motor cortex (M1) and with task‐related analysis of oscillatory EEG activity. An activation (ACT) and inhibition (INH) condition were compared. In both, visual cues were presented at 1/s. In ACT, subjects had to respond to these cues with individual finger movements as learned in a preceding training session. In INH, subjects had to observe the cues without retrieval of motor responses. During INH, inhibitory control of the motor memory trace was confirmed by significant amplitude reduction of motor evoked potentials (MEPs) compared with baseline. This was accompanied by a significant increase of 11–13 Hz oscillatory activity over the sensorimotor areas during INH. During active retrieval of the motor memory traces, the reverse was true (increased MEP amplitudes, decreased oscillatory 11–13 Hz activity). In a small sample of dystonic patients (n = 6), the increase of 11–13 Hz oscillatory activity during INH was consistently absent. The present data demonstrate for the first time cortical correlates of appropriate, context‐dependent inhibition of motor memory traces. We propose that focal increases of oscillatory activity are instrumental for inhibitory control at the cortical level. This concept is supported by the preliminary observations in dystonic patients who are known to have deficits of inhibitory motor control and in whom these context‐dependent focal increases of oscillatory activity were absent.

  • Keywords: motor learning; reorganization; oscillation; movement disorders; dystonia
  • Abbreviations: ACT = activation condition; ANOVA = analysis of variance; α‐ERD = event‐related α‐desynchronization (event‐related α‐power decrease); α‐ERS = event‐related α‐synchronization (event‐related α‐power increase); EPSP = excitatory post‐synaptic potential; FDI = first dorsal interosseus muscle; FOI = frequency of interest; INH = inhibition condition; IPSP = inhibitory post‐synaptic potential; LFP = local field potential; LSM = left sensorimotor region; MEG = magnetoencephalography; MEP = motor evoked potential; ML = midline sensorimotor areas; MT = motor threshold; M1 = primary motor cortex; OP = optimal point; ROI = region of interest; RSM = right sensorimotor region; TMS = transcranial magnetic stimulation; α‐TRPD = task‐related α‐power decrease; α‐TRPI = task‐related α‐power increase; TRPow = task‐related power


Learning can modulate cortical motor processing in experimental animals (Nudoet al., 1996a) and humans (Pascual‐Leoneet al., 1994; Karniet al., 1995). A basis for adapting to environmental changes and optimizing future behaviour is the learning of associations between external cues and motor acts. Learning‐induced changes in motor behaviour can be represented in the brain as motor memory traces that facilitate task‐related activation of the primary motor cortex (M1) (Pascual‐Leoneet al., 1994; Karniet al., 1995; Classenet al., 1998). A simplified example is the acquisition of novel finger movement sequences cued by external stimuli (Hondaet al., 1998).

To be effective, the retrieval of motor memory traces must be appropriate. Appropriateness is warranted if motor memory traces are only activated if necessary and are, probably more often, inhibited if required by the context of a situation. Inhibitory (GABA‐ergic) neurones are numerous in the human cortex (Keller, 1993; Micheva and Beaulieu, 1997) and are major local circuit neurones (Moore, 1993). Yet, most research on learning and complex human behaviour has been focused on neuronal activation in terms of excitation. There are no standard neuroimaging procedures for measuring inhibition, and inhibition seems to be underscored. Clinical data suggest a role of deficient inhibition in the pathogenesis of diseases such as dystonia (Berardelliet al., 1998), Tourette’s syndrome (Petersonet al., 1998) and attention deficit hyperactivity disorder (Barkley, 1997).

Task‐related changes of regional brain activation are a consequence of altered firing rates of single neurones involved in the processing of the task. At the level of neuronal assemblies, this can be measured as local field potentials (LFPs), and LFPs frequently have oscillatory properties (Gray and Singer, 1989; Murthy and Fetz, 1992). It is assumed that synchronized groups of action potentials in afferent fibres generate wave‐like excitatory post‐synaptic potentials (EPSPs) in the dendritic areas and produce the corresponding field and surface potentials in the EEG (Speckmann and Elger, 1999). In awake cats, for example, direct evidence has been provided that fast oscillatory brain activity (15–75 Hz) is correlated with tonic increases of neuronal firing rates (Destexheet al., 1999). In order to generate signals that are large enough to be detected in the surface EEG, ∼30–40 macro‐columns, comprising ∼1 cm2 of cortical surface, need to fire in synchrony (Lopes da Silva and Pfurtscheller, 1999). These and previous results (Singer, 1993; Contreras and Steriade, 1995; Donoghueet al., 1998; Frieset al., 2001) emphasize a close relationship between EEG oscillatory activity, oscillatory changes in LFPs and variations of single neurone firing rates, and the generation of EPSPs and inhibitory post‐synaptic potentials (IPSPs). In the human EEG, power decreases in the alpha band (8–13 Hz) have been linked to cortical activation. This phenomenon has been termed event‐related α‐desynchronization (α‐ERD) (Pfurtscheller and Aranibar, 1979) or task‐related α‐power decrease (α‐TRPD) (Gerloffet al., 1998b). It can also be seen in other frequency ranges, but has been studied most extensively in the alpha band (Steriade and Llinas, 1988; Pfurtschelleret al., 1997). On the contrary, focal increases of oscillatory alpha activity have been observed in areas not engaged in the task tested. This phenomenon has been termed event‐related α‐synchronization (α‐ERS) or task‐related α‐power increase (α‐TRPI) and has been suggested to be a correlate of an idling (Pfurtscheller, 1992) or ‘nil‐working’ state (Mulholland, 1995). In simple motor tasks, α‐ERS has also been linked with the concept of surround inhibition (Pfurtscheller and Andrew, 1999; Suffczynskiet al., 1999; Pfurtschelleret al., 2000). However, direct evidence for the association of α‐ERS (or α‐TRPI) and effective inhibition is lacking. We have addressed this question in a paradigm involving skilled motor behaviour and the inhibitory control of acquired motor programmes.

We studied context‐dependent modulation of motor memory traces in an externally cued sequential motor task. Cortical function was assessed with transcranial magnetic stimulation (TMS) over M1 (Cohenet al., 1998; Rothwell, 1991; Gerloffet al., 1998a) and task‐related analysis of oscillatory activity in multichannel surface EEG. The two main conditions were retrieval (activation, ACT) and non‐retrieval (inhibition, INH) of the acquired and memorized motor programmes. Following a training session, subjects performed visually paced sequences of 16 finger movements each (ACT) or observed the visual cues, but stayed relaxed throughout (INH). The behavioural concept of this paradigm might be best illustrated with a car waiting in front of a red traffic light. As the light turns green, the normal driver’s behaviour is to accelerate (retrieval, ACT). Non‐retrieval (INH) would correspond to the same situation with the signal of an ambulance approaching so that acceleration would be fatal, and the normal motor programme needs to be inhibited. Under physiological conditions, acquired motor programmes (motor memory traces) are readily retrieved and are thus the substrate of effective motor learning. We hypothesized that, rather than being ‘activated or disregarded’, these programmes are ‘activated or actively inhibited’ and that modulation of cortical oscillatory activity is instrumental for this aspect of context‐dependent motor control during complex behaviour. Under pathophysiological conditions with deficient inhibitory circuits (as in dystonia), we speculated that this type of inhibitory control should be disturbed.

Material and methods

This study consisted of two main experiments (labelled ‘original experiment’ and ‘TMS experiment’) and three control experiments (labelled ‘control experiments 1–3’). Original, TMS and control experiments 2 and 3 were each composed of a training session (Day 1) and a recording session on the subsequent day (Day 2).


We studied 18 healthy subjects (nine females, nine males), aged 21–33 years [mean ± SD (standard deviation), 25.9 ± 2.8 years] and six patients (two females, four males) suffering from idiopathic focal dystonia of the right hand, aged 29–68 years (43.0 ± 13.5 years). Further information on the patients is given in Table 1. All subjects and patients were right‐handed (Oldfield, 1971), naive to the experimental purpose of the study and did not play the piano regularly. All subjects gave their written informed consent for the study, which was approved by the University of Tübingen Medical School.

View this table:
Table 1

Dystonic patients (idiopathic dystonia)

PatientAge (years)SexSyndromeYear of manifestationTherapy
Dyst_0168MFocal dystonia of the right arm/hand (writer’s cramp)1975Botulinum toxin (13 months after last injection)
Dyst_0245FFocal dystonia of the right arm/hand (writer’s cramp)1991No
Dyst_0338MFocal dystonia of the right arm/hand (music dystonia, guitar player)1994Botulinum toxin (11 weeks after last injection)
Dyst_0436MFocal dystonia of the right arm/hand (music dystonia, guitar player)2000Botulinum toxin (6 weeks after last injection)
Dyst_0542MFocal dystonia of the right arm/hand (writer’s cramp); focal dystonia of the left hand (music dystonia, violin player)2000No botulinum toxin to the right hand
Dyst_0629FFocal dystonia of the right arm/hand (writer’s cramp) 1999Botulinum toxin (9 weeks after last injection)

Experimental setup

Sequences of finger movements

The motor sequences consisted of 16 consecutive finger movements (key presses on an electronic keyboard). The fingers were numbered as follows: 2 = index finger, 3 = middle finger, 4 = ring finger and 5 = little finger (Fig. 1A).

Fig. 1 (A) Right, electric keyboard. Left, the index finger was labelled ‘2’, the middle finger ‘3’, the ring finger ‘4’ and the little finger ‘5’. (B) Electrode montage according to the 10/20 system of electrode placement. Regions of interest (ROI) are marked in grey: LSM = FC3, C3, CP3; ML = FCz, Cz, CPz; RSM = FC4, C4, CP4.

In all sequences, rate (1 Hz), total number of key presses (n = 16) and external pacing by visual cues were equal. To ensure a comprehensive motor learning process, subjects had to perform three different types of motor sequences: (i) the first consisted of 16 repetitive key presses of the same finger (e.g. 2‐2‐2‐2‐2‐2‐2‐2‐2‐2‐2‐2‐2‐2‐2‐2); (ii) the second consisted of key presses in consecutive order up or down using four fingers (e.g. 4‐5‐2‐3, 4‐5‐2‐3, 4‐5‐2‐3, 4‐5‐2‐3); and (iii) the third consisted of key presses of the four fingers in a complex non‐consecutive order (e.g. 5‐5‐4‐2‐3‐4‐3‐5‐3‐5‐4‐4‐2‐4‐4‐2). The sequences of the third type were generated randomly; sequences with >3 repetitive key presses of the same finger were excluded. Over the whole experimental session, sequence group and performing fingers were matched. In a preceding training session, subjects had to learn one sequence of each type. These sequences were labelled ‘learned’. A common problem of performing pre‐learned motor acts repeatedly in an experimental setting is loss of attention. To minimize this problem, we randomly introduced unknown, visually instructed sequences during which the visual cues contained the information of which key to press next (termed ‘new’).

Visual cues

Two groups of visual stimuli were presented: (i) symbols (without specific information content) and (ii) numbers from 2 to 5 (indicating fingers). Symbols and numbers consisted of the same number of pixels. The symbols served as pacemakers for the learned sequences. The numbers were used to instruct subjects which finger they had to press while performing the new, visually instructed sequences (see ‘training session’ and ‘experimental setup’). The visual stimuli were presented for 350 ms at a rate of 1 Hz (onset to onset).

Training session (Day 1)

Subjects were asked to learn the three motor sequences described above (i–iii) to the beat of a metronome (1 Hz). The goal of the learning session was to perform the motor sequences 10 times in a row without errors. At this level of performance, the sequences were considered ‘overlearned’ (Gerloffet al., 1998b). During the training session, subjects were seated comfortably in an armchair with their right hand placed on the keyboard. All recordings were carried out on the day following the training session (Day 2).

Original experiment

During the recording session, subjects were seated comfortably in an armchair in front of a video‐screen in a light‐dimmed room with the right arm relaxed and resting on a pillow. The visual stimuli presented on the screen were subtending a horizontal visual angle of 4°. The right hand was positioned palm down so that the fingers could be moved freely and the keys could be pressed without wrist movements (Fig. A).

During this experimental session, subjects had to play the learned and the novel sequences (the novel condition was included solely for control of attention). The blocks of new sequences were presented randomly with the previously learned sequences. All blocks with movement were labelled ‘activation’ conditions. The actual target condition of the present study was the ‘inhibition’ condition, where visual cues were presented but no movement was required (Fig. 2). Key presses were paced externally by the visual cues at a rate of 1 Hz. Subjects were instructed to perform as accurately as possible and to continue without correction movements irrespective of errors they may have made. EEG was recorded continuously during the randomly presented blocks. As a rest condition, a 5 min pre‐ and post‐experimental EEG was recorded at unconstrained rest.

Fig. 2 Schematic of the experimental paradigm with activation and inhibition conditions. Top: sequence in the inhibition condition. In this condition, subjects observed the visual cues without retrieval of motor responses. Each square wave indicates a visual cue (onset, presentation, offset). Depending on the condition, the visual cue was a symbol or a number. Middle: sequence in the activation condition. In this condition, subjects responded to visual cues with individual finger movements, as learned in a preceding training session. Each square wave indicates a visual cue (symbol or number). The grey rectangles indicate the motor actions (key presses) of the subject. Bottom left: exact timing of the first two sweeps in an activation condition. Bottom right: exact timing of two sweeps in an inhibition condition. Numbers indicate time in milliseconds. Note that in the inhibition condition, the subjects knew that there would be no motor act throughout each entire sequence.

The three learned sequences were announced on the screen by ‘learned simple’, ‘learned scale’ and ‘learned complex’. The new sequences consisted of three types of sequences with parameters similar to the learned sequences and were announced by ‘new’. Thus, the movement (‘activation’) condition consisted of six variations: three types of learned sequences and three types of new sequences. The ‘inhibition’ sequences also consisted of six different variations, presentation of the same symbols as in the ‘learned’ conditions (three types of sequences), and presentation of the numbers (‘2’, ‘3’, ‘4’ and ‘5’) as in the ‘new’ condition (three types of sequences). The parameters (rate, symbols and numbers) of the ‘inhibition’ sequences were matched with the sequences used for the ‘activation’ conditions. In addition, the number of key presses per finger and sequence during activation was matched to avoid any bias due to the use of different fingers. In the ‘learned’ sequences, a target symbol (square) was presented with a probability of 20%. To keep visual attention in the learned conditions on a similar level across conditions, subjects were asked after each sequence whether it contained the target symbol or not.

The experimental session consisted of eight blocks, each block having 20 sequences; thus, subjects had to perform 160 sequences. The different sequences were presented pseudo‐randomized across the eight experimental blocks.

TMS experiment (cortical excitability)

The setup was as in the original experiment (preceding training session, experimental session consisting of activation and inhibition conditions). In the inhibition conditions, one to four magnetic stimuli per (16‐key press) sequence were delivered to the motor cortex over the optimal point (OP) for stimulation of a small hand muscle (first dorsal interosseus, FDI) of the right hand in four subjects that had participated in the original experiment. Three of the subjects were stimulated additionally during activation conditions. The TMS pulses were triggered automatically with a delay of 120 ms after the onset of the visual cue. A delay of 120 ms was used, based on theoretical considerations, to allow for transmission of the visual information to the occipital cortex and for some further higher order visual processing (Sokol, 1976). It has to be noted, however, that the effects seen in the EEG (original experiment) were long lasting (several 100 ms) so that exact timing in the range of milliseconds did not appear to be critical. The interval between two subsequent magnetic pulses was kept ≥4 s. Before and after each experiment, 20–30 MEPs were recorded as baseline.

Control experiments

A subgroup of the volunteers who were part of the original experiment participated in control experiment 3. A new (naive) group of subjects participated in control experiments 1 and 2. Experimental settings are explained in the paragraphs referring to each control experiment. Data acquisition and data analysis for control experiments 1, 2 and 3 were the same as in the original experiment.

Control experiment 1. We tested the relevance of the preceding training session for the changes of oscillatory activity in the inhibition condition. Here, the inhibition sequences were presented to five training‐naive subjects (no preceding training session). All parameters were kept equal to the inhibition conditions in the original experiment. Subjects did not have to perform any movements, and they had not been informed that the study was related to (finger) movements at all. Furthermore, the subjects’ right hands were not placed on the keyboard to avoid any associations of the stimuli with finger movements.

Control experiment 2 (‘context’ specificity). We assessed if the changes of oscillatory activity in the inhibition condition were only present if the inhibition condition was alternated rapidly with the activation condition in the same session. Five subjects (who had also participated in control experiment 1) performed the preceding training session as in the original experiment. One subject was unable to achieve the overlearned level in the training session and was excluded from this control experiment. The recording session included only the ‘inhibition’ condition with equal parameters as in the original experiment, but not alternated with the activation condition. As in control experiment 1, a pre‐/post‐experimental rest EEG was also part of this experiment.

Control experiment 3 (specificity of visual cue). We tested the possibility that the changes of oscillatory activity in the inhibition condition were an effect of unspecific visual stimulation. The same setup as described for the original experiment was used in this control experiment. Additionally, after the first rest EEG, a 5 min video clip (black and white) was presented and EEG was recorded throughout. After the video presentation, subjects had to answer three questions about its content (to keep attention at a constant level).

Data acquisition

Continuous EEG was recorded from 28 (Ag/AgCl) surface electrodes (Fig. B), mounted in a cap (Electro‐Cap International, Inc., Eaton, OH, USA). Impedance was kept below 5 kΩ. Data were sampled at 250 Hz, upper cutoff was 50 Hz, and the time constant was set to DC (Synamp amplifiers and software by NeuroScan Inc., Herndon, Va., USA). Linked earlobe electrodes served as reference. Two bipolar EMG channels were recorded from surface electrodes positioned over the right and left forearm flexors, with each pair of electrodes located ∼15 cm apart (distal tendon reference). The high‐pass filter for EMG was set to 30 Hz.

Key presses and visual trigger stimuli (symbols or numbers) were documented automatically with markers in the continuous EEG file and were used for stimulus‐locked averaging.

For digitization of head shape and electrode positions, a magnetic field digitizer (3Space Fastrak®, Polhemus, Colchester, Vt., USA) was used, as described elsewhere in detail (Andreset al., 1999).

In the TMS experiment, EMG was recorded from surface electrodes over the right FDI muscle for a time window of –150 ms to +250 ms relative to trigger onset. Sampling rate was 2500 or 5000 Hz. The recording bandpass filter was 30–1000 Hz.

During all conditions, subjects looked at a stationary fixation point in the centre of the screen to minimize eye movement artefacts. They were instructed to avoid eye blinks, swallowing or any movement other than the required finger movements during the performance of a sequence.


Single‐pulse TMS was carried out with the subjects at complete rest (absence of visually detectable background EMG). We used a ‘MAGSTIM rapid’ stimulator (Magstim Company Ltd, Whitland, Wales, UK) with a peak magnetic field of 2.5 T, a biphasic pulse of 250 µs pulse duration and a magnetic field rise time of 60 µs. A figure of eight‐shaped magnetic coil (inner diameter of each loop 5 cm) was used with the handle pointing backwards and laterally at an ∼45° angle to the sagittal plane. The optimal scalp position (OP) for activation of the FDI muscle was identified using a stimulus intensity sufficient to evoke small index finger movements and clear MEPs [∼120% motor threshold (MT)]. To facilitate constant positioning of the coil during the experimental session, a cap was used for marking OP, coil position and orientation on the subject’s head. MT was defined as the minimal intensity of stimulation capable of inducing five MEPs of >50 µV out of 10 stimulation pulses in the relaxed FDI muscle. Stimulation intensity was adjusted so as to obtain stable baseline MEPs of 1–2 mV. On average, this required an intensity of 119.0 ± 2.7% (mean ± SD) motor threshold. With the intensities used, subjects did not report relevant discomfort during the magnetic stimulation.

Data analysis

Spectral power analysis

In the original experiment and control experiments 1, 2 and 3, continuous EEG was segmented into artefact‐free epochs of 1024 ms duration (±512 ms visual cue onset), separately for each activation and inhibition sequence, for video presentation and for the rest period. Each single sweep was inspected visually, and trials with artefacts were rejected. The number of artefact‐free epochs for further analysis of inhibition, video and rest condition were: (i) symbol presentation between 138 and 291 epochs (mean ± SD, 198 ± 45); (ii) number presentation between 127 and 291 epochs (196 ± 47); (iii) video presentation between 53 and 248 epochs (170 ± 49); and (iv) rest condition between 146 and 520 epochs (327 ± 107). For spectral power analysis, a discrete Fourier transformation was computed for each 1024 ms epoch and all electrodes. The power spectrum from 1 to 50 Hz was calculated for each single epoch and then averaged across epochs. In order to account for intersubject variability, task‐related power (TRPow) was expressed as the percentage of spectral power during activation (Powactivation) compared with the spectral power during the rest condition (Powrest). This normalization was computed according to:

%TRPow = [(Powrest – Powactivation )/Powrest] × 100.

Therefore, TRPDs are expressed as positive values, and TRPIs as negative values. Further analysis of the power spectra was focused on the alpha (7–13 Hz) band (because on visual inspection the most prominent task‐related spectral power changes were observed in this frequency range). The alpha frequency range was divided into an upper (11–13 Hz) and lower (7–10.9 Hz) band for analysis (Pfurtscheller, 1989; Pfurtschelleret al., 2000). Group grand averages of percentage power changes were then calculated.

Time course analysis of spectral power

Epochs were bandpass filtered (slope 48 dB/octave), and magnitudes were squared and averaged in the time domain for the frequencies of interest (FOI). FOIs were here defined on the basis of visual inspection of the spectral power maps as 11, 12 and 13 Hz (main effect in inhibition condition). The inhibition condition was analysed for these FOIs. Averaged data were baseline‐corrected. The baseline was computed from 400 to 300 ms before stimulus onset. Group grand averages of spectral power time courses were then calculated for non‐overlapping time windows of 40 ms.

EMG analysis (TMS, cortical excitability)

Each single sweep was inspected visually, and trials with artefacts (pre‐stimulus EMG activity) were rejected. The numbers of artefact‐free epochs for further analysis of the MEPs of the baseline 1 (pre‐experiment), inhibition, activation and baseline 2 (post‐experiment) condition were: (i) baseline 1, 24–30 epochs (mean ± SD, 26.3 ± 2.6); (ii) inhibition, 27–38 epochs (34.0 ± 5.2); (iii) activation, 29–33 epochs (31.0 ± 2.8); and (iv) baseline 2, 20–27 (24.3 ± 3.0). The peak‐to‐peak amplitudes were analysed with custom‐made software. MEPs during inhibition and activation were calculated as the percentage of the peak‐to‐peak amplitude of the pooled baseline conditions (baseline 1, baseline 2).

Statistical analysis

On the basis of prior anatomical and physiological knowledge (Homanet al., 1987; Gerloffet al., 1998b), a set of regions of interest (ROI) was defined, the left sensorimotor region (LSM) represented by electrodes FC3, C3 and CP3; midline sensorimotor areas (ML) represented by electrodes FCz, Cz and CPz and the right sensorimotor region (RSM) represented by electrodes FC4, C4 and CP4 (Fig. B).

A factorial analysis of variance (ANOVA) design was employed for statistical analysis of log‐transformed data. The variances of spectral power estimates can be stabilized by logarithmic (log) transformation (Hallidayet al., 1995; Gerloffet al., 1998b). As factors, we defined experiment (original, control experiment 1, control experiment 2), condition (activation, inhibition, rest, video), region (RSM, ML, LSM), stimuli (symbols, numbers) and sweep (first 20, last 20). To account for interindividual variability, the factor subject was included. Contrast analysis was employed to test specific hypotheses derived from visual inspection of the maps. As the main effects occurred in the 11–13 Hz range, statistical analysis was focused on this frequency band. For analysis of the TMS data, the non‐parametric Kruskal–Wallis matched pairs test was used. Differences were considered significant, if P < 0.05. Results from multiple comparisons of the same data pool were Bonferroni corrected. For analysis of power data for INH between normal subjects and dystonic patients, a non‐parametric Mann–Whitney U test was used. For analysis of the behavioural data (performance), a t‐test for independent samples was used. To stabilize the variance of the performance data, which were expressed as a percentage (% accuracy), the logodds transformation was applied. Differences were considered significant if P < 0.05.


Behavioural data

The practice time to reach an ‘overlearned’ level was 23.0 ± 9.1 min (mean ± SD), which was largely due to practising the complex sequence. The practice time was similar for subjects participating in the original experiment (23.1 ± 10.7 min) and in control experiment 2 (22.7 ± 2.9 min). The practice time in patients was longer (33.8 ± 12.2 min; P < 0.05). In the TMS experiment, the necessary practice times were shorter (14.6 ± 3.9 min, control group). This was expected as an effect of learning transfer, because these subjects had already participated in the original experiment or in control experiment 2. The accuracy of performance during the ACT conditions was 97.4 ± 1.4% (mean ± SD) for normal subjects and 91.8 ± 3.1% for dystonic patients (P < 0.05).

Original experiment

Normal subjects

The main finding of the study was a topographically distinct α‐TRPI over sensorimotor areas in the inhibition condition, i.e. when visual cues were presented on the screen, but no movement was required (LSM, α‐TRPI –21.7%; ANOVA, inhibition versus rest P < 0.05; Fig. 3A). This α‐TRPI occurred in the upper alpha band (11–13 Hz, maximum at 12 and 13 Hz), was prominent over frontocentral, central and parietal areas bilaterally and was slightly less pronounced over the midline (ANOVA, contrast analysis, LSM versus RSM, LSM versus ML, n.s.; Fig. 4). In the lower alpha band, no conspicuous task‐related power increases were seen.

Fig. 3 Original experiment, normal subjects. (A) Left panel: topographic spectral power maps (average of 11 subjects) in the frequency range of 11–13 Hz. Maps are projected onto a 3D digitized head. Left, inhibition condition; right, activation condition. Task‐related increases of oscillatory activity (α‐TRPIs) are given in blue, decreases of oscillatory activity (α‐TRPDs) are given in red. Note the similarity of the topographic distribution of α‐TRPI (inhibition condition) and α‐TRPD (activation condition). Right panel: difference map between activation and inhibition in the mentioned frequency range. (B) This figure summarizes the results by providing TMS and EEG data of a representative subject for all relevant conditions. Top panel: motor evoked potentials (MEPs) induced by TMS over the primary motor cortex (over the optimal point for stimulation of the FDI muscle) during the baseline (left), inhibition (middle) and activation (right) conditions. The results of the TMS experiment (reduced MEPs in the inhibition condition) confirmed that the excitability of the corticospinal system (stimulated over M1) was reduced in the inhibition condition at the time when α‐TRPI was observed in the original experiment. Middle panel: subject’s hand placed on the electric keyboard. The EMG activity of the first digital interosseus muscle (FDI) is shown during the performance of the baseline (left), inhibition (middle) and activation conditions (right). Lower panel left: 3D digitized individual head shape with realistic (co‐registrated) placement of the surface electrodes; Lower panel middle: inhibition condition; lower panel right: activation condition. Topographic spectral power maps in the frequency range of 11–13 Hz. Maps are projected to the individual 3D digitized head shape of the subject. α‐TRPIs are given in blue, and α‐TRPDs are given in red. In the inhibition condition (middle), i.e. when visual cues were presented on the screen, but no movement was required, there was a topographically distinct α‐TRPI bilaterally over the sensorimotor areas. During the activation conditions, there was α‐TRPD bilaterally over the sensorimotor areas. Note the similarity of α‐TRPI and α‐TRPD.

Fig. 4 Original experiment, normal subjects. Summary of bin‐by‐bin analysis of oscillatory activity (in percentage of increase/decrease) across the entire alpha and beta frequency spectrum for the sensorimotor region (LSM = left sensorimotor, top; ML = mesial, middle; RSM = right sensorimotor, bottom). The increase of oscillatory activity during INH is most prominent in the 11–13 Hz frequency range. Error bars = 1 SEM. Activation = white, inhibition = black.

For the activation conditions, the main result was, as expected with this type of motor behaviour, an α‐TRPD bilaterally over the sensorimotor and parietal areas (ANOVA, ‘learned’ and ‘new’ pooled, activation versus rest P < 0.05; Fig. A). This α‐TRPD was evident in the frequency range 9–13 Hz. In the upper alpha band (11–13 Hz), bilateral frontal, central and parietal α‐TRPD was present, with the maximum over left centroparietal areas (ANOVA, contrast analysis, LSM versus RSM, LSM versus ML, P < 0.05; Fig. ). On direct comparison, the activation patterns of inhibition and activation conditions in the frequency range 11–13 Hz were different at group level (ANOVA, inhibition versus activation, P < 0.05).

The symbols presented on the screen contained only information about the pace (one movement per second), while the numbers contained information about pace and type of movement (determining the finger which should be pressed). Nevertheless, when subjects only observed the visual cues and did not move, there was no significant difference of α‐TRPI for presented symbols and numbers (ANOVA, P = 0.73). This suggests that the pacing information but not the information as to the type of movement to be executed was crucial for the α‐TRPI to evolve.

Time course analysis during the inhibition condition showed a relatively constant event‐related power increase over sensorimotor areas throughout the entire time window of 1 s. This is indicative of a relatively constant level of inhibitory control in this paradigm. A moderate desynchronization over occipital areas was found from 80 to 400 ms relative to stimulus presentation, corresponding to the visual processing of the cues. Because of its short duration and low amplitude, this occipital activation (α‐ERD) was not obvious in the bin‐by‐bin analysis of α‐TRPD, which always averages across 1024 ms.

TMS experiment (cortical excitability)

The TMS experiment was designed to test if the cortical oscillatory phenomenon ‘α‐TRPI’ seen in the present paradigm in healthy subjects indeed reflected net inhibition of motor cortical output. As TMS activates the pyramidal neurones trans‐synaptically, the size of a response to TMS (MEP) is influenced by the level of cortical excitability (Ziemannet al., 1996; Gerloffet al., 1998a).

The amplitudes of MEPs (mean ± SD) evoked by single TMS pulses over the M1 in inhibition conditions (0.8 ± 0.7 mV) were significantly lower than MEPs induced during unconstrained rest (1.4 ± 1.1 mV) or during the activation condition (3.4 ± 2.1 mV) (Kruskal–Wallis, P < 0.05; Fig. 5).

Fig. 5 TMS of the motor cortex during inhibition (middle), unconstrained rest (‘baseline’) (left) and activation conditions (right). Bars indicate means of peak‐to‐peak amplitudes in percentage of average peak‐to‐peak amplitude of the baseline conditions. Error bars = 1 SD. The amplitudes of MEPs in inhibition conditions were significantly lower than MEPs induced during baseline or activation conditions; *P < 0.05. These results confirm that the excitability of the corticospinal system (stimulated over M1) was reduced in the inhibition condition compared with baseline or activation conditions.

The results of the TMS experiment confirmed that the excitability of the corticospinal system (stimulated over M1) was reduced in the inhibition condition at the time at which α‐TRPI is observed in the main experiment. Figure B summarizes these results by providing TMS and EEG data of a representative subject for all relevant conditions.

Control experiment 1 (effect of training)

Does the α‐TRPI effect depend critically on the preceding motor training? In this control experiment, visual cues were presented to ‘naive’ subjects. The activation patterns during presentation of the visual cues (symbols or numbers 2–5) were significantly different from the original experiment. In essence, without training, there was no α‐TRPI in the upper alpha band during the inhibition conditions (LSM, α‐TRPI –3.4%; ANOVA, control experiment 1 versus rest, P = 0.67; Fig. 6A). The direct comparison of the inhibition conditions of control experiment 1 and the original experiment confirmed the difference in activation patterns (ANOVA, control experiment 1 versus original, P < 0.05). These data show that the α‐TRPI effect depended critically on the preceding training session.

Fig. 6 Topographic spectral power maps in the frequency range of 11–13 Hz. Maps are projected onto a 3D digitized head. α‐TRPIs are given in blue, and α‐TRPDs are given in red. (A) Control experiment 1 (performance of the inhibition conditions without preceding training). Note that no significant α‐TRPI was found, indicating that the preceding training session was critical to induce α‐TRPI. (B) Control experiment 3 (presentation of a video clip): the main effect of video stimulation occurred in the frequency range of 9–11 Hz. Topographic spectral power maps in the frequency range of 9–11 Hz. The maps revealed widespread α‐TRPD predominantly over occipital areas. Note that there was no α‐TRPI, indicating that the results of the original experiment were not an unspecific result of visual stimulation. (C) Original experiment, dystonic patients (average of six patients). Left, inhibition condition; right, activation condition. Note that there is no increase of oscillatory activity during INH. (D) Original experiment, dystonic patients (average of six patients). Summary of bin‐by‐bin analysis of oscillatory activity (in percentage of increase/decrease) across the entire alpha and beta frequency spectrum for the left sensorimotor region. ACT is given in white, INH in black. Error bars = 1 SEM.

Control experiment 2 (‘context’ specificity)

In this control experiment, inhibition and activation conditions were not alternated within the session. Only inhibition conditions were tested. The patterns of electrocortical activity for the inhibition condition were similar to the original experiment, with significant α‐TRPI over sensorimotor areas (LSM, α‐TRPI –16.4%, ANOVA, control experiment 2 versus rest, P < 0.05). Statistical analysis of α‐TRPI amplitudes in the upper alpha band between original and control experiment 2 showed no significant difference (ANOVA, control experiment 2 versus original, P = 0.56). As expected, significant differences were detected when inhibition conditions of this control experiment and inhibition conditions of learning‐naive subjects were compared (ANOVA, control experiment 1 versus control experiment 2, P < 0.05). These data suggest that the α‐TRPI effect did not depend critically on rapid alternation of motor activity and stimulus observation without movement in the same session, i.e. the α‐TRPI effect also occurs if the inhibition condition is not embedded in blocks of movement.

Control experiment 3 (specificity of visual cue)

Testing of a different type of visual stimulation (video clip in black and white), embedded in the original experimental setting, showed a different pattern of electrocortical activity compared with the inhibition condition (ANOVA inhibition versus video, P < 0.05; Fig. B). The main effect of video stimulation occurred in a frequency range of 9–11 Hz and revealed widespread α‐TRPD predominantly over parieto‐occipital areas, consistent with other studies on the processing of visual information of higher complexity (Pfurtschelleret al., 1994). No α‐TRPI‐like pattern was evident in this condition, indicating that the α‐TRPI effect was not an unspecific response to visual stimulation after motor learning.

Comparison of first 20 with last 20 sweeps of the inhibition condition in the original experiment (intra‐experimental learning)

For a subgroup (n = 4), data of the original experiment were analysed so as to compare the first 20 sweeps and the last 20 sweeps of the inhibition conditions. The first and last 20 sweeps of artefact‐rejected epochs were averaged separately and analysed as described above. The spectral power amplitudes in the upper alpha band were larger in the last 20 sweeps than in the first ones (ANOVA first 20 versus last 20, P < 0.05). This intra‐session effect was predominant in the 12 Hz frequency band. These data indicate that, besides the effect of the preceding training, there is an accentuation of the α‐TRPI effect, if motor practice is continued during the session, perhaps in the sense of a ‘use‐dependent’ enhancement.

Combination of EEG and TMS

A methodological limit of non‐invasive EEG measurements is the relatively coarse topographic resolution. To acknowledge this, we confined our conclusions to the level of ‘regions’ (LSM, ML, RSM). Furthermore, in a skilled task as presented here, different steps of cortical processing are necessary and associated with extended bilateral centroparietal activation (Sadatoet al., 1996a; Manganottiet al., 1998). However, by combining EEG and TMS in the same subjects, we may say that the common final path from motor cortex to a target hand muscle was significantly inhibited during the inhibition conditions and, thus, during the presence of α‐TRPI in these very subjects. Further experiments are necessary to determine the role of the adjacent parietal regions and of the ipsilateral sensorimotor regions in mediating inhibitory control of the corticospinal system.

Dystonic patients

All patients completed the task without subjective difficulties. None of the patients experienced relevant dystonia of the affected right hand during the ACT condition. According to EMG monitoring, no muscle contraction of the affected or unaffected hand was present in the INH condition.

The main finding in dystonic patients was the absence of α‐TRPI in the INH condition (Fig. C). On the contrary, during INH, the oscillatory EEG activity showed a pattern similar to that during ACT across the whole alpha and beta band (Fig. D). On direct comparison (healthy subjects versus dystonic patients), this resulted in a significant difference for the INH condition, with less α‐TRPI in dystonia (Mann–Whitney U test, P < 0.05). The amount of oscillatory activity over the LSM is presented for patients and normal subjects in Fig. 7. There was some overlap between healthy subjects and patients. However, in none of the patients were notable focal power increases (α‐TRPI) detected. The mean (± SD) α‐TRPI over LSM was –21.9 ± 40.2% for healthy subjects, and the corresponding value was 15.0 ± 22.2% (= α‐TRPD) for dystonic patients. In the present study, we included a limited number of patients for ‘proof of principle’. Further studies on dystonic patients are necessary to characterize more precisely the deficit reflected here in cortical inhibition.

Fig. 7 Oscillatory activity during INH for dystonic patients and healthy subjects in the frequency range of 11–13 Hz over left sensorimotor areas (pooled for C3, CP3). Right column, patients with focal dystonia; left column, normal subjects. Negative power values correspond to α‐TRPI (inhibition). Error bars indicate 1 SEM. *P < 0.05, Mann–Whitney U test.

Additional analyses of TRPow during movement in the ACT condition in the dystonic patients revealed α‐TRPDs bilaterally over the sensorimotor and parietal areas, similar to normal subjects. This α‐TRPD was most prominent in the frequency range of 9–13 Hz and had a higher amplitude than in normal subjects. No significant differences were found in other frequency ranges (low alpha, low and high beta).

TMS experiment in dystonic patients (cortical excitability)

Cortical excitability was studied in two dystonic patients. In patient 1, the amplitudes of MEPs (mean ± SD) evoked by single TMS pulses over the M1 were 0.3 ± 0.1 mV during INH, 0.6 ± 0.1 mV during ACT and 0.2 ± 0.05 mV during unconstrained rest. In Patient 6, the amplitudes of MEPs were 3.0 ± 0.6 mV during INH, 4.0± 0.6 mV during ACT and 3.1 ± 0.4 mV during unconstrained rest. As opposed to normal subjects, there was no amplitude reduction of the MEPs during INH in these dystonic patients.


The main finding of the present study was appropriate, context‐dependent inhibitory control of previously acquired motor programmes. This behavioural and electrophysiological (decreased MEP amplitudes) phenomenon was paralleled by enhanced oscillatory alpha activity (α‐TRPI) over cortical sensorimotor areas. We favour the interpretation that increases of local oscillatory activity are instrumental for inhibitory control of neuronal activity. This interpretation is strengthened by the absence of α‐TRPI in a preliminary study on six patients with focal dystonia of the hand, as a model for diseases with deficient inhibitory circuitry (Riddinget al., 1995; Tinazziet al., 2000; Abbruzzeseet al., 2001).

In the brain, established associations are memorized in such a way that the efficacy of the respective cue–reaction pairing is increased in similar future situations, which results in improved motor skills (Karni, 1996; Manganottiet al., 1998). In daily life, learned associations (memory traces) not only have to be retrieved, but also have to be inhibited under certain circumstances to warrant successful behaviour. Besides the example of the green traffic light with an approaching ambulance and the need to inhibit the memory trace ‘green light–acceleration’, also, in more general terms, appropriate behaviour is reliant on a person’s ability to inhibit inappropriate thoughts, impulses and (motor) actions. Certainly, context‐dependent focal inhibition is only one of several instruments for the implementation of skilled human behaviour, but it deserves particular consideration because it is an aspect of brain function that tends to be underscored in neuroimaging studies.

The role of inhibition in motor control and learning

In the training session of our experiment, associations between external cues and motor output had been established so that every stimulus prompted a finger movement. The motor learning in this training session induced an ‘external cue–motor output’ memory trace, that allowed for improved performance in the recording session when movements were required (activation conditions). It is known that the acquisition of motor memory can in fact alter the primary sensorimotor cortex. There are various examples of learning‐induced changes in primary cortical regions: extension of primary sensorimotor representations in Braille‐readers (Sadatoet al., 1996b; Cohenet al., 1997) and string players (Elbertet al., 1995), changes of representation patterns during recovery after brain injury (Nudoet al., 1996b; Hondaet al., 1997) or the modulation of representations in the M1 after training in healthy subjects (Classenet al., 1998).

In the inhibition condition of the present study, subjects stayed relaxed during presentation of the known stimuli. We assumed that the training‐induced motor memory traces needed to be suppressed in this situation. This assumption was strongly supported by the TMS results. The MEPs in small hand muscles were significantly reduced when subjects watched the cues but stayed relaxed.

The present data emphasize the important role of inhibition in the implementation of motor behaviour in the human brain. Jacobs and Donoghue (1991) blocked cortical inhibition pharmacologically in adult rats, and found that reduction of inhibition can temporarily change the proper relationship patterns between different limb representations in M1 areas. In humans, Liepertet al. (1998) demonstrated with TMS that intracortical inhibition may be increased overly for an actively relaxed muscle, if this relaxation was required explicitly. Extending these previous data, the present results highlight the existence of functionally relevant inhibitory control in the context of skilled active motor behaviour.

Clinical and electrophysiological data also emphasize the importance of inhibitory motor control mechanisms. Deficient inhibition at various levels of the neuroaxis is discussed as one pathogenetic factor in dystonia (Prioriet al., 1995; Berardelliet al., 1998; Lorenzanoet al., 2000; Tinazziet al., 2000). Using a TMS double‐pulse technique, Riddinget al. (1995) provided evidence for decreased inhibition at the level of primary motor cortex in dystonic patients. Corresponding results were obtained by Abbruzzeseet al. (2001) in an experiment on cortical somatosensory and motor processing. Toroet al. (2000) found reduced pre‐movement β‐ERD (20–30 Hz, ‘activation’) in dystonic patients during the performance of self‐paced simple index finger abductions. They also explained their results with deficient inhibition at rest resulting in a less dynamic evolution of movement‐related activation. The absence of α‐TRPI over sensorimotor areas and the associated absence of MEP amplitude reduction during INH in our patients is in line with these previous results. Moreover, our data extend the previous findings because deficient inhibition has been documented here for the first time in the setting of an actively executed, skilled motor task with changing demands depending on the context of each experimental block. The performance of complex sequential movements is frequently impaired in patients suffering from writer’s cramp or music dystonia, as studied here. In the present paradigm, this is reflected in the higher error rates in patients (accuracy, 91.8 ± 3.1%, mean ± SD) compared with normals (97.4 ± 1.4%), although all patients could perform the task without developing dystonic contractions in the right hand. Patients also needed more extended training than normal subjects to achieve the required level of motor performance. Perhaps deficient inhibition of memory traces is a factor that contributes to these deficits. This difference in behavioural output needs to be kept in mind when comparing the cortical activation patterns, but the following points should be noted. First, performance was still >90% for both groups. Secondly, the inclusion criterion in the present study was that each subject (patient or normal) had to achieve the ‘overlearned’ level on Day 1. This was taken as evidence that a similar level of motor memory had been achieved (and needed to be inhibited) in all participants. Thirdly, the main result occurred during the inhibition condition (INH). Slight differences in motor execution during the activation condition (ACT) are thus not likely to be responsible for differences in α‐TRPI between patients and controls during INH. Furthermore, we would like to stress that the dystonia model was used here for ‘proof of principle’. A detailed pathophysiological analysis of the situation in dystonia, including the necessary detailed assessment of spinal and cortical excitability, would have been beyond the scope of the present study.

How do our results relate to Go/No‐go paradigms?

Decreased motor cortical excitability has also been described for the No‐go condition in Go/No‐go tasks (Sasaki and Gemba, 1986; Hoshiyamaet al., 1996, 1997; Naito and Matsumura, 1996; Jacksonet al., 1999). There are fundamental differences between the paradigm used in the present study and such a task. (i) In a Go/No‐go task, the S1 stimulus prompts the subjects to prepare the motor act as quickly as possible, and the No‐go S2 signal cancels this dynamic, volitional preparation process (Naito and Matsumura, 1996; Graftonet al., 1998). Our subjects did not have to produce any fast voluntary inhibition of volitional motor actions. Whenever an inhibition sequence had been initiated, they were informed in advance that for the upcoming 16 visual cues no motor output was required. Voluntary rapid initiation and inhibition of responses as in Go/No‐go paradigms can therefore not play a major role in the present experiments. (ii) Go/No‐go tasks are reaction time paradigms in which speed of performance is critical (Shibataet al., 1997). In the present study, precision of the sequential finger movements was the critical factor and the movement rate was slow (1/s).

There is converging evidence that, depending on the context, cortical motor networks integrate facilitatory and inhibitory activity. The importance of ‘context’ for neuronal activity in premotor and motor areas has been emphasized recently by Hepp‐Reymondet al. (1999) who demonstrated context‐dependent modifications of the discharge rates in a grip force task in monkeys. How does the human brain achieve context‐dependent focal inhibition of memory traces?

Neurophysiological correlates of inhibition

Inhibitory activity in the brain is difficult to assess with non‐invasive methods. For example, in functional imaging techniques such as PET and fMRI, decreases of the rCBF (regional cerebral blood flow) (Shadmehr and Holcomb, 1999) and the BOLD (blood oxygen level‐dependent) signal (Garavanet al., 1999) might indicate inhibition. However, activation of large assemblies of inhibitory neurones might equally as well result in rCBF and BOLD signal increases or leave these signals unchanged (Waldvogelet al., 2000). As for EEG, only few data are available and are ambiguous (Pfurtscheller, 1992; Mulholland, 1995; Basaret al., 1997). Analysis and detection of cortical inhibition in the sensorimotor system has been the domain of stimulation methods such as TMS (Ziemannet al., 1996; Chenet al., 1998; Gerloffet al., 1998a).

By combining EEG and TMS over the M1 in the same subjects, it was possible to demonstrate an association between M1 inhibition in a skilled sequential motor task and increased oscillatory activity in this and other regions. The topographic distribution of oscillatory activity during this skilled motor task extended over bilateral centroparietal regions, in line with studies on cortical activation in similar tasks (Sadatoet al., 1996a; Manganottiet al., 1998). By combining EEG and TMS in the same subjects, we may say that the common final path from motor cortex to a target hand muscle was significantly inhibited during the INH conditions and, thus, during the presence of α‐TRPI. This association is circumstantial and no proof of causality, but it has to be stressed that the MEP amplitude reduction in normals and the absence of this phenomenon in dystonia were obtained prospectively, and strictly based on the hypothesis that α‐TRPI reflects inhibition. Further experiments are necessary to determine the role of the adjacent parietal regions and the ipsilateral sensorimotor regions in mediating inhibitory control of the corticospinal system. Our data support the concept that modulation of oscillatory neuronal activity is instrumental for task‐related, context‐dependent cortical inhibition. TMS alone is not sufficient to distinguish cortical and subcortical influences on the response (MEP) size. However, since the local EEG oscillations are cortical in origin and since TMS and EEG changes occurred in parallel, it is unlikely that subcortical and spinal mechanisms could account for the results.

The present findings go beyond the concept of locally increased oscillatory activity as a correlate of idling (Pfurtscheller, 1992) or a ‘nil‐working’ state (Mulholland, 1995). The view of increased oscillatory activity as a correlate of inhibition is supported further by Worden and colleagues who showed α‐power increases in association with ignored visual stimuli (Wordenet al., 2000). Their paradigm did not allow for direct demonstration of inhibition with neurophysiological measures (like TMS in the present study).

On the basis of the present data, we may now propose that task‐related increases of oscillatory activity in sensorimotor areas represent a neurophysiological mechanism of context‐dependent inhibition of acquired motor programmes and thus an instrument for implementing context‐dependent inhibitory behaviour in the human brain.

Physiological significance of task‐related increases of oscillatory activity (α‐TRPI)

No α‐TRPI in training‐naive subjects

In control experiment 1, subjects were naive and had no indication that the experiment was related to motor functions. Activation conditions had been excluded to avoid the possibility of intra‐experimental (‘use‐dependent’) learning effects. This experiment did not reveal any α‐TRPI pattern comparable with the original experiment. Thus, motor learning in the preceding training session was critical to induce α‐TRPI in the inhibition condition.

Intra‐experimental learning

Trained subjects were tested without the possibility of intra‐experimental learning (exclusion of activation conditions in control experiment 2). Their spectral power maps showed a distribution of increased oscillatory alpha activity (α‐TRPI) similar to that in the original experiment. Thus, intra‐experimental learning was not critical for the induction of α‐TRPI. The α‐TRPI effect did not depend critically on rapid alternation of motor activity and stimulus observation without movement in the same session, i.e. the α‐TRPI effect also occurred if the inhibition condition was not embedded in blocks of movement. However, α‐TRPI was somewhat enhanced by intra‐session practice (predominantly in the 12 Hz frequency band) if ACT and INH were alternated randomly (‘use‐dependent’ enhancement).

Mental rehearsal and imagery

The instruction for the subjects before performing the INH condition was: ‘Watch the symbols on the screen and stay relaxed’. Subjects may or may not have thought about the motor sequences while they watched the visual cues during INH. However, if the sensorimotor cortex is involved in motor imagery at all (still under debate), it is activated and not inhibited during imagery. This is supported by various PET (Decety, 1996) and fMRI data (Erslandet al., 1996; Porroet al., 1996), as well as by increases in M1 excitability when tested with TMS (Abbruzzese et al., 1996; Fadigaet al., 1999). EEG analyses during motor imagery showed alpha and beta power decreases and not increases (Pfurtscheller and Neuper, 1997). Thus, it seems possible that our subjects thought about movements, but the electrophysiological results with α‐TRPI during the INH condition cannot be explained with imagery‐related activity patterns on the basis of the available literature.

Information content of the visual cues

The two groups of cues carried different information. Symbols carried pacing information only. Numbers also announced which finger had to be used. One might assume that, at least in activation conditions, the attentional level is lower for symbol than for number cueing. To minimize this theoretical difference in attention, we included a target symbol in the symbol cueing condition and instructed the subjects to report on the presence or absence of the target after each sequence. The comparison of symbol cueing and number cueing revealed no difference, indicating that the oscillatory activity over sensorimotor areas was independent of the type of cueing.

No α‐TRPI with neutral visual material

To exclude the possibility that the task‐related increase of oscillatory activity (α‐TRPI) was a non‐specific result of visual stimulation after training, neutral visual material was presented to trained subjects in control experiment 3 (video clip). The main effect of video stimulation occurred in the frequency range of 9–11 Hz and consisted of widespread α‐TRPD predominantly over parieto‐occipital areas, consistent with other studies on the processing of visual information of higher complexity (Pfurtschelleret al., 1994). α‐TRPI over sensorimotor areas could not be induced by neutral visual material, indicating that the motor memory traces were associated specifically with visual cues (symbols, numbers) that provided at least the pacing information for the trained movement sequences.

Temporal aspects of learning induced cortical activity changes

Our training sessions lasted for ∼20 min. Thus, all changes of cortical oscillatory activity and excitability were initiated in this short training period. This is in agreement with previous data showing that learning‐induced changes can occur within minutes (Saneset al., 1992; Classenet al., 1998) and can alter electrophysiological parameters rapidly (Brasil‐Netoet al., 1992). Long‐term changes after prolonged training have been demonstrated with fMRI (Karniet al., 1995), MEG (Elbertet al., 1995) and TMS (Pascual‐Leoneet al., 1993). The effects of training in the present experiments persisted for at least 1 day. A detailed analysis of the time course, however, was beyond the scope of the present study.

In conclusion, the present experiments demonstrate cortical correlates of appropriate, context‐dependent modulation of acquired motor programmes in neurologically healthy subjects. A likely candidate mechanism for this type of inhibitory modulation is the focal increase of oscillatory activity (α‐TRPI) in neuronal assemblies normally engaged in the volitional activation of the previously learned motor act. This type of inhibitory modulation (α‐TRPI) was absent in a small group of dystonic patients. We propose that local enhancement of alpha oscillations is an instrument in inhibitory motor control.


We wish to thank Rolf Kirsammer, Volker Munderich and Rüdiger Berndt for technical assistance, and Helge Topka, MD for referral of patients. This work was supported by the Deutsche Forschungsgemeinschaft (SFB 307/B12, SFB550/C5).


View Abstract