Motor performance fluctuates trial by trial even in a well-trained motor skill. Here we show neural substrates underlying such behavioral fluctuation in humans. We first scanned brain activity with functional magnetic resonance imaging while healthy participants repeatedly performed a 10 s skillful sequential finger-tapping task. Before starting the experiment, the participants had completed intensive training. We evaluated task performance per trial (number of correct sequences in 10 s) and depicted brain regions where the activity changes in association with the fluctuation of the task performance across trials. We found that the activity in a broader range of frontoparietocerebellar network, including the bilateral dorsolateral prefrontal cortex (DLPFC), anterior cingulate and anterior insular cortices, and left cerebellar hemisphere, was negatively correlated with the task performance. We further showed in another transcranial direct current stimulation (tDCS) experiment that task performance deteriorated, when we applied anodal tDCS to the right DLPFC. These results indicate that fluctuation of brain activity in the nonmotor frontoparietocerebellar network may underlie trial-by-trial performance variability even in a well-trained motor skill, and its neuromodulation with tDCS may affect the task performance.
Human motor performance does not always end up with the same behavioral consequences, but it rather fluctuates trial by trial even in a fully acquired and well-trained motor skill [
Previous neurophysiological studies in nonhuman primates have shown that fluctuation in reaching movements is caused by the variability of preparatory neuronal firing in the premotor cortex (PM) and in the primary motor cortex (M1) [
In the present study, in order to elucidate distribution of neuronal cause for a motor-performance fluctuation, we conducted a functional magnetic resonance imaging (fMRI) experiment (fMRI experiment). In a next experiment, we applied transcranial direct current stimulation (tDCS) to the dorsolateral prefrontal cortex (DLPFC), which was identified in the fMRI experiment as one of the brain regions where the activity changes in association with a fluctuation of task performance, and tested if the tDCS affects the task performance (tDCS experiment).
In the fMRI experiment, we measured brain activity with fMRI while participants repeatedly performed a well-trained skillful sequential finger-tapping task. We identified brain regions where the activity is negatively correlated with the task performance across trials. Since we found a negative correlation between the task performance and activity in a broader range of the frontoparietocerebellar network, including the DLPFC, in the following tDCS experiment, we tried to modulate brain activity in the DLPFC and evaluated the performance change in order to examine causal relationship between the DLPFC activity and the task performance.
Fifteen healthy male volunteers participated in the fMRI experiment (
The participants performed a sequential finger-tapping task with their dominant right fingers [
(a) A sequential finger-tapping task was performed utilizing the dominant right hand fingers. (b) Participants were required to repeatedly press a five-element sequence (sequence “32413”) as quickly and accurately as possible during a 10 s trial period. The fixation color turned yellow from white and the numeric sequence (“32413”) appeared for 3 s, indicating the ready period. A red fixation color indicated the 10 s execution period. (c) Experimental procedure of the tDCS experiment. During the 1st and 2nd runs, the participants did not receive any stimulation. The third run was begun 5 min after the start of tDCS. The participants performed the 3rd and 4th runs when receiving tDCS (anodal, cathodal, or sham) to the right dorsolateral prefrontal cortex (DLPFC).
To examine the fluctuation of the well-trained motor performance, we required the participants to intensively train the tapping task prior to the fMRI experiment (two participants who only participated in the tDCS experiment also completed this training). Training consisted of two parts: training in the laboratory and then in their own home. On the first day of training, the participants practiced a total of 50 trials of the 10 s sequential finger tapping in the laboratory. In the following days, they were asked to practice the tapping task as many times as possible in their daily life in order to feel confidence to perform the task until the date when they participated in the fMRI experiment. The interval between the laboratory training and the fMRI experiment was
We used a 3.0 T MRI scanner (Trio Tim, Siemens, Germany) with a head-coil to obtain T1-weighted anatomical images and functional
In the scanner, the participants rested comfortably in a supine position. Their right arms were orientated parallel to their torso, and their forearms were pronated and supported by a cushion, allowing them to relax their arms. During the scan, the participants were allowed to move only their right fingers to press the buttons placed just beneath their hands (HHSC-1x4-D, Current Designs Inc., Philadelphia, USA).
Visual stimuli were projected onto a screen in the scanner. The participants viewed the stimuli via a mirror in front of their eyes. Throughout the experiment, a fixation cross was present in the center of the screen, and the participants were instructed to maintain their gaze on this point and to avoid unnecessary eye movements (Figure
At the beginning of each trial, the numeric sequence “32413” was presented just above the fixation cross and the fixation cross turned yellow for 3 s (Figure
In the fMRI experiment, all the participants completed 15 experimental runs, with 2 min interrun intervals. Each run was composed of 10 trials with a 9 s ITI, and thus a total of 150 trials were completed. Each run included a 12 s rest period before the first trial and another 12 s rest period after the last trial. In total, each run lasted for 232 s and 116 functional volumes were collected per run.
To evaluate the task performance, we calculated the number of correct sequences completed within the 10 s execution period in each trial [
It is well known that a motor performance tends to be worse at the beginning of an experimental session even when a motor task is well trained (warm-up decrement) [
In the preprocessing of the functional volumes, we initially performed slice timing correction and head motion correction (realignment). After these corrections, both the functional and anatomical images were normalized to the Montreal Neurological Institute (MNI) template brain using the standard SPM8 defaults. The functional images were smoothed with an isotropic 8 mm full-width-at-half-maximum Gaussian kernel. Finally, high-pass filters (128 s) were applied to the fMRI time series in each run to remove low frequency noise and global changes in the signals.
The statistical analysis was performed on two levels. A first-level analysis was done in each participant as follows. A linear regression model (general linear model) was fitted to the data obtained from each participant. The model included the following three regressors: (1) a performance-related regressor (PERFORMANCE), which was a parametric modulation regressor for the number of correct sequences in each trial; (2) an error-related regressor (ERROR), where each timing of incorrect button press was modeled as the event-related regressor [
To accommodate interparticipant variability, each participant’s contrast image was entered into a random-effect group analysis [
In the fMRI experiment, we identified the clusters of active voxels in the bilateral DLPFC, bilateral frontoparietal regions, and the left cerebellum in the PERFORMANCE image (see Section
To stimulate the right DLPFC, the target electrode was placed on F4 according to the EEG international 10-20 system [
A 15 min tDCS with 2 mA was applied from an electrical stimulator (DC-stimulator-Pulse M, neuroConn, Germany) via two saline-soaked surface sponge electrodes (5 × 7 cm). We used three types of stimulation: anodal (anodal electrode on F4), cathodal (cathodal electrode on F4), and sham. Sham stimulation was comprised of a short-period (30 s) current stimulation with the same polarity as the anodal stimulation. A fade-in and fade-out period was set at 30 s at the beginning and at the end of stimulation. In the sham session, we did not inform the participants of this (single blind).
Each participant completed three sessions (anodal, cathodal, or sham session) on separate days. The order of the stimulation was randomized. Each session was conducted at least 7 days apart in order to minimize the risk of contamination via the carryover effects from the previous tDCS application. In the tDCS experiment, the participants sat comfortably in a chair with the right forearm on an armrest, allowing them to relax their arms. Beneath their hands, the same button device used in the fMRI experiment was placed and the same visual instructions as in the fMRI experiment were presented on a computer monitor in front of the participant (Figure
In the first two runs, the participants performed the task without receiving any stimulation (1st and 2nd runs; PRE). These were done to measure the baseline level of their performance for a session. Then, they performed the task by receiving either type (anodal, cathodal, or sham) of the tDCS to the right DLPFC in the next two runs (3rd and 4th runs; DURING). We started the tDCS immediately after the 2nd run was completed and the stimulation lasted until the end of the 4th run in the anodal and cathodal sessions. We set a 5 min interrun interval between the 2nd and the 3rd runs, because it is shown that neuronal modulation effect by tDCS may emerge around 5 min after the initiation of the stimulation [
When we looked at the change in the number of correct sequences (= task performance) in each participant, we found that the task performance fluctuated trial by trial even though no gradual performance improvement was observed throughout the experimental runs (Figure
(a) Trial-by-trial fluctuation in the number of correct sequences in a representative participant. (b) Mean number of correct sequences across all 15 participants. The error bar indicates ±1 SD.
Despite this stable performance, we found warm-up decrement effect in the performance of the 1st and 2nd runs. Namely, the mean number of correct sequences in the 1st and 2nd runs (
These lines of evidence suggested that even though the participants well trained the task and the performance was stable during the experiment, the performance could still fluctuate trial by trial.
When we analyzed the PERFORMANCE image, we found that activity in a broader range of brain regions was negatively correlated with task performance across trials (Figure
Brain regions where activity was negatively correlated with the number of correct sequences (PERFROMANCE image).
Brain areas | Coordinates of peaks |
|
||
---|---|---|---|---|
|
|
| ||
Right hemisphere | ||||
MFG | 36 | 44 | 26 | 4.52 |
IFG | 36 | 32 | 26 | 4.00 |
MCC | 10 | 24 | 38 | 4.65 |
45 | 56 | 20 | 30 | 4.28 |
6 | 10 | 12 | 56 | 4.91 |
Insula | 38 | 10 | −10 | 4.76 |
PF | 62 | −32 | 28 | 4.90 |
PFt | 50 | −36 | 48 | 4.84 |
hiP2 | 40 | −34 | 38 | 4.38 |
7PC | 42 | −48 | 58 | 4.46 |
7A | 34 | −56 | 54 | 5.48 |
|
||||
Left hemisphere | ||||
MFG | −34 | 40 | 16 | 4.26 |
MCC | −10 | 18 | 36 | 4.91 |
44 | −46 | 10 | 26 | 4.36 |
6 | −4 | 8 | 52 | 4.49 |
Insula | −42 | 16 | −2 | 4.67 |
STG | −50 | 4 | −2 | 4.23 |
PG | −62 | −18 | 40 | 4.28 |
2 | −52 | −24 | 38 | 4.90 |
PF | −62 | −36 | 20 | 4.89 |
PFt | −60 | −16 | 30 | 4.83 |
hiP2 | −40 | −44 | 48 | 4.41 |
hiP3 | −30 | −50 | 44 | 6.04 |
Lobule VI | −32 | −50 | −36 | 4.67 |
Lobule VII | −44 | −52 | −34 | 4.42 |
Peaks in brain activation that were more than 4 mm apart from each other were reported (voxel size = 2 × 2 × 2 mm). For anatomical identification of peaks, we only considered cytoarchitectonic areas with more than 30% probability available in the anatomy toolbox. Cytoarchitectonic area with the highest probability was reported for each peak. When cytoarchitectonic areas with more than 30% probability were not available for a peak, we simply provided its anatomical location. When several peaks were identified at the same cytoarchitectonic area or anatomical location, we only provided the peak coordinates with the highest
(a) Brain regions where activity was negatively correlated with the number of correct sequences (PERFROMANCE image). (b) Brain regions where activity was related to the occurrence of movement errors (ERROR image). (c) Brain regions active during the 10 s execution period (TASK image). Activities are superimposed on the MNI standard brain.
It should be noted that the task performance was negatively correlated with the number of movement errors (mean of
Brain regions where activity was related to the occurrence of movement errors (ERROR image).
Brain areas | Coordinates of peaks |
|
||
---|---|---|---|---|
|
|
| ||
Right hemisphere | ||||
ACC | 6 | 32 | 28 | 3.53 |
MCC | 10 | 20 | 42 | 4.16 |
IFG | 52 | 14 | −2 | 3.51 |
44 | 56 | 12 | 8 | 3.69 |
6 | 8 | 18 | 58 | 3.96 |
Insula | 40 | 18 | −4 | 3.70 |
Thalamus | 12 | −14 | 12 | 3.52 |
|
||||
Left hemisphere | ||||
ACC | −4 | 26 | 28 | 3.76 |
MCC | −8 | 20 | 38 | 4.01 |
44 | −38 | 20 | 12 | 3.78 |
SFG | −14 | 12 | 48 | 3.67 |
Insula | −32 | 20 | 4 | 4.37 |
SMG | −2 | 18 | 42 | 4.19 |
Thalamus | −4 | −18 | 12 | 4.44 |
For anatomical identification of peaks, we used the same criterion adopted in Table
On the other hand, it seems that the greater activity in the bilateral DLPFC, superior and inferior parietal lobules, premotor cortices, and left cerebellum may not reflect such error-related activity (Figures
Brain regions active during execution period (TASK image).
Brain areas | Coordinates of peaks |
|
||
---|---|---|---|---|
|
|
| ||
Right hemisphere | ||||
6 | 2 | −6 | 70 | 4.29 |
LG | 14 | −90 | −14 | 3.60 |
18 | 22 | −86 | −16 | 3.41 |
17 | 22 | −94 | −16 | 3.40 |
Vermis | 6 | −64 | −26 | 6.28 |
Lobule V | 2 | −56 | −2 | 5.03 |
Lobule VI | 22 | −56 | −26 | 6.77 |
Lobule VIIa | 40 | −72 | −24 | 4.13 |
|
||||
Left hemisphere | ||||
SFG | −22 | −8 | 56 | 4.53 |
6 | −40 | −18 | 64 | 4.98 |
4a | −34 | −24 | 56 | 5.37 |
1 | −54 | −20 | 48 | 5.41 |
2 | −38 | −30 | 46 | 5.66 |
PFop | −56 | −20 | 28 | 5.05 |
Lobule VI | −20 | −62 | −24 | 4.95 |
For anatomical identification of peaks, we used the same criterion adopted in Table
The fact that this network included the premotor cortex nicely supported the previous nonhuman primate’s finding [
When we evaluated the tDCS effect on the number of correct sequences, we found small performance improvement in the DURING (3rd and 4th runs) of the sham session (
Mean tDCS effect on the number of correct sequences across participants in each (anodal, cathodal, or sham) session. The error bar indicates ±1 SD.
On the contrary, when we evaluated the tDCS effect on the standard deviation (SD) of the number of correct sequences (= degree of performance variability), we could not find any significant change between the PRE and the DURING in any of the tDCS sessions (anodal: PRE
We should carefully discuss the performance improvement in the sham session (Figure
If the anodal tDCS did not give any impacts to the brain activity, we could have expected the performance improvement also in the anodal session. But the results in the anodal session showed the opposite and significantly different effect from the sham session. This seems to corroborate our view that the present anodal tDCS gave significant impacts to the brain activity and that this tDCS effect likely deteriorated the task performance without affecting the degree of performance variability.
In the present study, we showed the possible causal relationship between the anodal tDCS to the right DLPFC and the task performance. However, we could not elucidate exact neuronal mechanisms of how the tDCS modulated brain activity so as to affect the motor performance. One possibility is that anodal tDCS modulated local brain activity in the right DLPFC. Previous neurophysiological studies demonstrated that anodal tDCS may provide a potentiation effect on neuronal excitability in a stimulated brain region [
Another possibility is that anodal tDCS could affect the activity in the remote nonstimulated brain regions that are functionally and anatomically connected with the DLPFC. A previous study showed that anodal tDCS to the right DLPFC can modulate the activity not only in this region but also in the left DLPFC and in the bilateral posterior parts of the parietal lobules [
We should also carefully discuss the effect of cathodal tDCS (Figure
The effect of cathodal stimulation is still controversial [
The reason why we could not find stronger improvement effect in the cathodal session than in the sham session could also be explained by ceiling effect. In general, when a motor task is well trained, its further performance improvement is normally difficult to achieve. Indeed, it is shown that facilitative tDCS to the M1 only improves the performance of a motor skill in novices but not in experts [
Taken together, we showed in the tDCS experiment that neuromodulation with anodal tDCS to the DLPFC (the representative nonmotor region where activity showing neuronal fluctuation associated with the performance fluctuation in the fMRI experiment) may affect the task performance without affecting the degree of performance variability per se, though we could not elucidate exact neuronal mechanisms underlying the tDCS effect.
The present study aimed to investigate a distributed neuronal cause for trial-by-trial fluctuation in well-learned skillful motor performance. We showed that the fluctuation of brain activity in the nonmotor frontoparietocerebellar network is associated with the trial-by-trial performance fluctuation even in a well-learned skillful motor task and that neuromodulation with anodal tDCS to the representative nonmotor domain (DLPFC) may affect the task performance.
Nobuaki Mizuguchi’s present addresses are as follows: Japan Society for the Promotion of Science, 5-3-1 Kojimachi, Chiyoda-ku, Tokyo 102-0083, Japan; Faculty of Science and Technology, Keio University, 3-14-1 Hiyoshi, Kohoku-ku, Yokohama, Kanagawa 223-8522, Japan.
The authors declare that there are no competing interests regarding the publication of this paper.
Nobuaki Mizuguchi and Eiichi Naito designed the research; Nobuaki Mizuguchi, Shintaro Uehara, and Shinji Yamamoto performed the experiment; Nobuaki Mizuguchi, Shintaro Uehara, and Satoshi Hirose analyzed the data; Nobuaki Mizuguchi, Shintaro Uehara, Satoshi Hirose, Shinji Yamamoto, and Eiichi Naito wrote the paper.
This work was supported by JSPS KAKENHI (Scientific Research on Innovative Areas “Understanding brain plasticity on body representations to promote their adaptive functions” no. 26120003 to Eiichi Naito, Grant-in-Aid for Specially Promoted Research 24000012 to Eiichi Naito, and 24800092 and 26750242 to Nobuaki Mizuguchi).