Cerebral Cortex, 2024, 34, bhae210

[Link]
Advance access publication date 20 May 2024
Original Article

Gray matter volume of functionally relevant primary

motor cortex is causally related to learning a hand
motor task
1
Derin Cobia , Marc W. Haut2,3,4 , Kate P. Revill5 , Stephanie L. Rellick2 , Randolph J. Nudo6 , Miles Wischnewski7 ,
Cathrin M. Buetefisch7,8,9, *
1 Department of Psychology and Neuroscience Center, 1036 KMBL, Brigham Young University, Provo, UT 84602, USA
2 Department of Behavioral Medicine and Psychiatry, Rockefeller Neuroscience Institute, 33 Medical Center Dr., West Virginia University, Morgantown,
WV 26506, USA
3 Department of Neurology, Rockefeller Neuroscience Institute, 33 Medical Center Dr., West Virginia University, Morgantown, WV 26506, USA
4 Department of Radiology, Rockefeller Neuroscience Institute, 33 Medical Center Dr., West Virginia University, Morgantown, WV 26506, USA
5 Department of Psychology, Emory University, 36 Eagle Row, Atlanta, GA 30322, USA
6 Department of Rehabilitation Medicine, University of Kansas Medical Center, 3901 Rainbow Boulevard, Kansas City, KS 66160, USA
7 Department of Neurology, Emory University, 1441 Clifton Road NE, Suite 236 C, Atlanta, GA 30322, USA
8 Department of Rehabilitation Medicine, Emory University, 1441 Clifton Road NE, Suite 236 C, Atlanta, GA 30322, USA
9 Department of Radiology, Emory University, 1441 Clifton Road NE, Suite 236 C, Atlanta, GA 30322, USA

*Corresponding author: Departments of Neurology, Rehabilitation Medicine, and Radiology, Emory University, School of Medicine, 1441 Clifton Road NE, Suite 236
C, Atlanta, GA 30322. Email: [Link]@[Link]

Variability in brain structure is associated with the capacity for behavioral change. However, a causal link between specific brain areas
and behavioral change (such as motor learning) has not been demonstrated. We hypothesized that greater gray matter volume of a
primary motor cortex (M1) area active during a hand motor learning task is positively correlated with subsequent learning of the task,
and that the disruption of this area blocks learning of the task. Healthy participants underwent structural MRI before learning a skilled
hand motor task. Next, participants performed this learning task during fMRI to determine M1 areas functionally active during this
task. This functional ROI was anatomically constrained with M1 boundaries to create a group-level “Active-M1” ROI used to measure
gray matter volume in each participant. Greater gray matter volume in the left hemisphere Active-M1 ROI was related to greater motor
learning in the corresponding right hand. When M1 hand area was disrupted with repetitive transcranial stimulation (rTMS), learning
of the motor task was blocked, confirming its causal link to motor learning. Our combined imaging and rTMS approach revealed greater
cortical volume in a task-relevant M1 area is causally related to learning of a hand motor task in healthy humans.

Key words: BA4; hand area; motor skill; motor learning; neuroimaging.

Introduction (Kanai and Rees 2011;Sampaio-Baptista et al. 2014 ; Lehmann et al.

Advances in neuroimaging analysis have led to the association 2020), whether these brain areas were of functional relevance for
of exceptional behavior, particularly the acquisition of complex learning the particular skill has not been demonstrated. For exam-
cognitive and motor skills, with specific brain structures (Elbert ple, greater gray matter volume of the medial occipito–parietal
et al. 1995; Gaser and Schlaug 2003; Hyde et al. 2009; Hamzei areas at baseline correlated with steeper learning slopes for jug-
et al. 2012; Gerber et al. 2014; Sampaio-Baptista et al. 2014). For gling (Sampaio-Baptista et al. 2014). Similarly, greater gray matter
example, memorization of detailed spatial maps over an extended volume in right orbitofrontal cortex was related to a participant’s
period (years) was associated with increased hippocampal volume ability to improve performance on a whole-body balancing task
(Maguire et al. 2000; Woollett and Maguire 2011) while learning on a seesaw-like-platform (Lehmann et al. 2020). However, these
complex visual motor skills, such as juggling, were associated tasks involve the whole body and the brain areas identified as
with increases in gray matter of the temporal/occipital junction being related to skill learning were not examined for functional
(Draganski et al. 2004, 2006). These findings have advanced the relevancy (Sampaio-Baptista et al. 2014; Lehmann et al. 2020).
fundamental question of whether skill learning can alter brain Therefore, whether individual variability in functionally-relevant
structure in the mature brain. However, the question of whether brain structures is causally linked to observed differences in
initial, or preexisting, brain structure determines the capacity behavioral change remains to be determined.
for skill learning is less studied. While a relationship between In the present study, we sought to determine the relationship
inter-individual differences in brain structure and inter-individual between gray matter volume in a task-relevant brain area in the
differences in learning has been reported in multiple studies mature healthy human brain and inter-individual differences in

Received: January 31, 2024. Revised: April 24, 2024. Accepted: April 28, 2024
© The Author(s) 2024. Published by Oxford University Press.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License ([Link]
licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For
commercial re-use, please contact [Link]@[Link]
2 | Cerebral Cortex, 2024, Vol. 34, No. 5

task learning. For this purpose, we focused on the primary motor was designed to establish causality between processes in M1 and
cortex (M1) for its known involvement in the early stage of motor learning of the skilled hand motor task. We conducted two exper-
consolidation with new skill acquisition. We hypothesized that iments in a randomized order (see below for details). We used low
baseline gray matter volume within an M1 area that is also active frequency subthreshold rTMS or sham stimulation of the hand
in a specific hand motor learning task is related to the capacity for area of left M1 to transiently disrupt M1 activity immediately after
gains in this task, and that disruption of this M1 area blocks the participants practiced the hand motor learning task.
ability to learn the task. To test this hypothesis, we performed two
studies in a group of healthy, middle-aged to older, right-handed Study 1
participants. In the first study, we obtained structural magnetic Participants
resonance imaging (MRI) of participants’ brains prior to learning Thirty healthy, right-handed adult participants (mean age ± SD:
how to manipulate a joystick with their fingers to reach targets 61.9 ± 9.0 years, range 50–78, 16 females) were tested and partici-
of different sizes on a computer screen. Subsequently, task-based pated in Study 1.
fMRI was used to determine, at a group level, the cortical area
functionally active when executing this task. This region of inter- MRI acquisition parameters
est (ROI) was further structurally constrained to fall within atlas- Structural and functional MR images of all participants in Study
defined boundaries of M1 (Fischl et al. 2008). We measured each 1 were collected on a Siemens Prisma 3 T MRI system with
participant’s gray matter volume within this group-level “Active- a 64-channel head coil. Structural images were acquired using
M1” ROI and related it to functional gains during learning of the HCP Lifespan Project MPRAGE protocol (Harms et al. 2018):
the skilled hand motor task. The second study was designed to TR = 2400 ms, TE = 2.24 ms, f lip angle = 8 degrees, FOV = 256 × 240
establish causality between processes in M1 and learning of the × 208 mm, 0.8 mm isotropic voxel dimensions. Eight task-based
skilled hand motor task. We used low frequency subthreshold functional MRI runs were collected with the following protocol:
repetitive transcranial magnetic stimulation (rTMS) of the hand TR = 2000 ms, TE = 28.0 ms, f lip angle = 90 degrees, FOV = 192 × 192
area of left M1 to transiently disrupt M1 activity immediately after × 102 mm, voxel size = 3 mm3 , 163 volumes per run, total duration
participants practiced the hand motor learning task (Muellbacher 44:00 min. Structural scans were collected and considered for
et al. 2002). Given evidence for the critical involvement of M1 in inclusion into the study according to the inclusion criteria prior
the early stage of motor consolidation (Muellbacher et al. 2002), to participants’ involvement in other parts of the study.
we expected that learning of the skilled hand motor task would
be disrupted by rTMS, but not sham stimulation, to the hand Experimental design—hand motor learning task
area of M1, thereby providing evidence for a causal link between All participants completed a single exposure hand motor learning
processes in the task-relevant M1 gray matter and learning of the task that lasted 20–30 min per trained hand (Buetefisch et al. 2014;
skilled hand motor task. Barany et al. 2020). Participants completed 1 to 3 runs depending
on their performance in the first run and their gains in skillfulness
with subsequent practice. This learning paradigm was selected
Materials and methods to avoid overlearning in participants with high baseline skillful-
Participants in both studies were recruited through Emory Univer- ness. Training was conducted for both the right and left hand
sity and fulfilled the following criteria for inclusion: age between separately, with the order of initial left versus right hand training
50 and 80 years; no neurological or psychiatric disorders; normal randomized across participants. The training task is a computer-
neurological examination; no current usage of central nervous ized behavioral paradigm (Presentation® software, [Link].
system-active drugs; no contraindication to MRI; and no struc- com) developed to measure a participant’s ability to manipulate
tural MRI brain abnormalities as evaluated by a board-certified a joystick (Mag Design and Engineering, [Link])
neurologist. All participants also demonstrated normal cogni- between the thumb, middle, and index fingers to precisely guide
tion on the Repeatable Battery for the Assessment of Neuropsy- a cursor to an array of target positions. Manipulating the joystick
chological Status (RBANS; Total Scale index scores >2 standard to reach the target required wrist extension/f lexion movements
deviations below the mean were considered abnormal) (Ran- of ∼5◦ in addition to finger movements. Four different target sizes
dolph et al. 1998). Handedness was determined by the Edin- were presented, which allows for parametric variation in the level
burgh Handedness Inventory (Oldfield 1971). All procedures were of required precision. According to the speed-accuracy tradeoff
approved by the Institutional Review Board at Emory University described in Fitts’s law, decreasing the size of a target increases
and were in accordance with the ethical standards of the insti- the level of difficulty and results in longer movement times (Fitts
tutional research committee on human experimentation (institu- 1954). By forcing participants to complete the task in a short,
tional and national) and with the Helsinki Declaration of 1975, as predefined period of time, accuracy was expected to decrease for
revised in 2008. All participants gave written informed consent smaller target sizes. Training-related improvement in accuracy
and were blinded to the stated hypothesis of the experiments. at a similar or increased speed meets the criteria for motor skill
The studies were conducted as secondary analysis of healthy learning (Censor et al. 2012).
participants from two registered clinical trials ([Link] Participants sat comfortably in a dental chair ∼165 cm in front
identifier: NCT01726218 and NCT02544503); data are available of a computer screen (29 × 51 cm). The forearm and proximal arm
upon request. were supported to eliminate their involvement in the execution
Overview of experiments: First, all participants underwent a of the task. At the beginning of each trial, participants were
structural magnetic resonance imaging (MRI). Second, partici- presented with a blank gray screen followed by the display of a
pants learned how to manipulate a joystick with their fingers to central cue square indicating the target size at the center of the
reach targets of different sizes on a computer screen. Third, task- screen (Fig. 1A). In the upper half field of the computer screen,
based fMRI was used to determine, at a group level, the cortical either a small (5.3 × 5.3 mm), medium (9.3 × 9.3 mm), large
area functionally active when executing the hand motor learning (13.2 × 13.2 mm), or extra-large (17.2 × 17.2 mm) target (Fig. 1B)
task. These data were collected at least 1 day after the motor was presented at four possible locations (300◦ , 330◦ , 30◦ , and 60◦ ).
learning task. This concluded data collection for Study 1. Study 2 Participants were instructed to move the cursor square as quickly
 Cobia et al. | 3

Fig. 1. Task design. A) each trial began with a central square cuing the upcoming target size. After the cue, the target appeared in one of four possible
locations (300◦ , 330◦ , 30◦ , 60◦ relative to vertical meridian) after which the participants were instructed to use the joystick to move the cursor to the
target as quickly and accurately as possible. If the cursor began at the starting point and was inside the target 2 s after target presentation, participants
received “hit” feedback; otherwise, they received “miss” feedback. B) Four different target sizes were used to vary demand on precision. C) during fMRI
acquisition, participants used an MRI-compatible joystick secured on their body to complete the task with pads placed to support the arm and minimize
elbow and shoulder movement. EMG electrodes were placed on participants’ right- and left- extensor carpi ulnaris (ECU) muscles to monitor unimanual
performance.

as possible into the target square, the appearance of which served overlearning the task. All other participants continued to practice
as a “Go-signal.” After the target square appeared, participants until they reached either a minimum performance level of 50%
had 2 s to move the cursor into the target and maintain it in that hits for the extra-large target or until they completed 2 additional
position until the end of the 2-s trial. runs. For participants with small target accuracies <50% but
Correct movements were defined as originating from the start extra-large target accuracies >50% after the first run, practice
position at the center of the screen and resulting in an overlap continued until their performance improved to >50% hits on the
between the center of the cursor and the target at the end of the largest target size for which they were less than 50% accurate
2-s movement period. The movement time was defined as the after the first run or until two additional runs were completed
time elapsed between the Go signal and the time the center of the (Wischnewski et al. 2016).
cursor reached the target. Accordingly, a movement was recorded
as inaccurate when the participant did not reach the target within Experimental design—hand motor learning task-based
the 2-s period, failed to hold the cursor position within the target, fMRI
or left the start position early. Participant feedback was provided Subsequent to the motor learning task described above, a varia-
in the form of presentation of the word “hit” or “miss” on the tion of the motor learning task was administered to participants
screen for 500 ms for each correct or incorrect trial, respectively during active acquisition of fMRI protocols in the scanner, with
(Fig. 1A). The total duration of each trial was 4 s. Trials were target sizes modified to accommodate the different participant
grouped into blocks of 28 trials with three blocks per run, and a positioning and viewing distance. Stimuli were displayed using
12-s rest period between blocks. In each block, each target size Presentation® software ([Link]), and in separate runs
was presented seven times at a random selection of the four participants manipulated the joystick with either the right or left
possible locations (4 sizes × 7 trials/size = 28 trials). At the end of hand as described for the motor learning task above (Barany et al.
the run, the percentage of “hit” trials for each target size across all 2020). Targets were projected onto a screen that was viewed via
three blocks was calculated. The performance level determined a mirror mounted onto the head coil. With the participant in a
whether participants continued to practice the pointing task for supine position on the scanner bed, the base of the joystick base
up to 2 additional runs. For participants achieving at least 50% was strapped to the participant’s torso with Velcro straps and
hits on the small target size, no further practice was done to avoid positioned so participants could rest their wrist on the base of
4 | Cerebral Cortex, 2024, Vol. 34, No. 5

the joystick comfortably and manipulate it without moving their were used to extract each individual’s cortical volume data for
distal or proximal arm. Foam pads further supported the arm to statistical analysis. In each hemisphere a surface-based ROI was
avoid additional muscle activity (Fig. 1C). Movement epochs (16 s computed that only contained overlapping vertices from both the
× 12) alternated with resting epochs (9 s × 11). Trials were blocked FS M1 label and the respective fMRI left- and right-hand group
by target size, with four trials per movement epoch. Each target masks; for example, overlapping vertices in the left hemisphere
size block was presented three times per run. The order of target from M1 and the fMRI group mask (derived from right-hand per-
size blocks was randomized within each run. The participants formance) were isolated and combined into a single ROI labeled
completed four runs with one hand (right or left) and then four “Left Hemi Active-M1”. A similar ROI was constructed in the right
runs with the other hand, in a counterbalanced order across hemisphere from overlapping vertices in the FS M1 label and fMRI
participants. mask derived from left-hand performance labeled “Right Hemi
Active-M1” (see Fig. 2 for all mapped ROIs). The resulting group
MRI data processing Active-M1 ROIs were mapped to all participants and cortical gray
Structural T1-weighted images were processed using the FreeSurfer matter volume, calculated as the product of cortical thickness
(FS) toolkit version 6.0 ([Link] (Dale and surface area, from each label was extracted (in mm3 ) using
et al. 1999; Fischl et al. 1999). Each MR image was intensity embedded FS algorithms. Taking this approach, we were able to
corrected, skull stripped, and then automatically segmented delineate gray matter volume in M1 areas from both hemispheres
into gray and white matter. The segmentations and surfaces that were active when executing the motor learning task. As
were inspected and manually edited for accuracy according to indicated in Fig. 2 Active-M1 ROIs were primarily located in the
established guidelines (Segonne et al. 2007). A surface-based anterior wall of the central sulcus and posterior aspect of the
map of the primary motor cortex (M1) was isolated in its precentral gyrus, areas referred to as hand area where corticomo-
entirety by combing both the anterior (BA4a) and posterior tor (CM) neurons are found in non-human primates (Strick et al.
(BA4p) labels generated as part of the default FS pipeline in both 2021).
hemispheres; these labels have demonstrated robust accuracy
in their definition of the motor cortex according to cortical Statistical analysis—motor learning task
folding patterns (Fischl et al. 2008). Global intracranial volume Baseline accuracy was defined as the percentage of hits in the
was also calculated using the FS estimated Total Intracranial first block of the first run. Final accuracy was defined as the
Volume measure (Buckner et al. 2004) and examined as a potential percentage of hits in the last block of the final run (which was
covariate. either the last block of run 1, 2, or 3 depending on the participant’s
accuracy in the first run and their gain in skillfulness with subse-
Experimental design—identifying M1 regions related to the quent practice). A standardized learning value (SLV) (Wischnewski
hand motor learning task et al. 2016) was calculated to assess the training-related gain in
Functional ROIs ref lecting activation by the hand motor learning precision, defined as:
task were derived from eight runs of the functional imaging data Standardized Learning Value = Learning Value ∗ (max possible
collected as described above. For each hand, data were collapsed gain)−1 .
across target size and blocks of task performance were compared Learning Value = (final accuracy − baseline accuracy/number of train-
with rest blocks. All processing of the fMRI data was accomplished ing blocks) ∗ 100.
using Analysis of Functional NeuroImages (AFNI) (Cox 1996); fMRI Given that a typical learning curve asymptotes near 100%
preprocessing steps included slice time correction, head motion accuracy and that improvement upon high accuracy is more
correction, 12-parameter affine alignment between the structural difficult than improvement from low accuracy, the learning value
and functional images, nonlinear warping between the structural was multiplied by the inverse of the maximum possible gain
image and the MNI152 2009 template in MNI space, smoothing (Ammons 1947). The maximum possible gain was calculated
with a FWHM 6.0 mm smoothing kernel, and conversion to per- as the difference between the maximum possible hits and the
cent signal change. The transformations for head motion correc- number of actual hits in the first block of the first run. Repeated-
tion, co-registration, and normalization were concatenated and measures analysis of variance (RM-ANOVA) models were used to
applied in a single step to the functional data before smoothing examine the within-subjects effects of target size (S, M, L, XL) and
to reduce the number of interpolation steps. Following prepro- hand (right, left) on accuracy (% hits) and movement time. Paired
cessing, GLM analysis was performed using AFNI’s 3dDeconvolve t-tests were used to compare the baseline skillfulness and SLV
tool. In addition to the left- or right-hand task regressors, six head between hands. Two-tailed correlation coefficients were gener-
movement vectors were included as regressors of no interest. ated between baseline skillfulness and the change in skillfulness
Volumes with more than 0.9 mm total head movement were across the experiment; all data are expressed in mean ± SEM.
censored. Beta values for the left- and right-hand conditions for
each participant were submitted to one-sample t-tests at the Statistical analysis—cortical volume of M1 regions related
group level to determine areas where left- and right-hand activity to motor learning task
were significantly greater than rest. Group left- and right-hand To address the proposed hypothesis that greater gray matter
masks were created by thresholding these statistical maps at an volume within the Active-M1 ROIs relates to greater gain in skill
uncorrected P = 10−10 , a threshold more stringent than required acquisition on a hand motor learning task, a series of Pear-
by multiple comparisons correction, but that resulted in a single son correlation coefficients were calculated. First, to assess the
cluster of activation for each hand regressor centered in the potential confound of age, Pearson correlation coefficients were
contralateral M1. Left- and right-hand group activation masks generated between age and all variables of interest, including
were then registered to FS atlas space (MNI 305) and mapped to baseline accuracy and SLV motor performance for both hands, as
the cortical surface using embedded FS algorithms. well as volume of the Active-M1 ROIs for both hemispheres. To
Using the FS M1 label to anatomically constrain the group-level assess the potential confound of global head/brain size, Pearson
functional activation ROIs to M1, two ROIs (one per hemisphere) correlation coefficients were generated between the FS estimated
 Cobia et al. | 5

Fig. 2. Group labels for active-M1 ROI in each hemisphere: Visualization of the anatomically defined M1 (outlined in white) atlas-based regions of interest
on both pial and inf lated surfaces derived from the FreeSurfer toolkit. Regions within anatomically defined M1 that were active at the group level during
the motor learning task, as identified by fMRI (noted in yellow), are defined as “active-M1” ROI and are localized in anterior wall of the central sulcus
and posterior aspects of the precentral gyrus included in the anatomically defined hand area. (n = 30).

Total Intracranial Volume measure and the gray matter volume eight participants who did not contribute data to Study 1 were
of the Active-M1 ROIs for both hemispheres. To examine the recruited using the same recruitment strategies and inclusion
relationship between Active-M1 gray matter volume and baseline criteria (see overview of the studies); their neuroimaging data
accuracy on the hand motor task, a priori hypothesized Pear- were only used for inclusion criteria evaluation as it had been
son correlation coefficients were calculated between accuracy in collected under a different scan protocol on a different MRI
each hand relative to the contralateral M1 ROIs (e.g. right hand machine (NCT01726218). All participants had the same exposure
baseline accuracy with Left Hemi Active-M1 ROI). Similarly, to to the motor learning task.
examine the relationship between Active-M1 gray matter volume
and motor skill acquisition gains (SLV), a priori hypothesized Pear- Experimental design—disrupting the retention of
son correlation coefficients were calculated between SLV motor training-related improvement in accuracy via rTMS
performance in each hand relative to the contralateral M1 ROI.
For Study 2, 1 Hz rTMS or sham stimulation was applied to left
Finally, to determine whether any relationships observed above
M1 (LM1) hand area as defined by the hot spot of the targeted
were unique to M1 and not to other primary sensory cortical
wrist extensor muscle, a muscle that is involved in the training
regions, a “control region” analysis was performed by constructing
task and represented within the hand area (see below) (Siebner
Pearson correlation coefficients between baseline accuracy/SLV
et al. 2022) to determine the effects of subthreshold stimula-
motor performance and cortical volume of the primary auditory
tion on pointing task accuracy and motor learning as previously
cortex (A1) in the contralateral hemisphere. The A1 ROI was
described (Muellbacher et al. 2002; Hadipour-Niktarash et al. 2007;
defined as the anterior transverse temporal gyrus label (Heschl’s
Buetefisch et al. 2011). Brief ly, the subthreshold intensity refers to
gyrus) from a standard FS parcellation scheme (Destrieux et al.
the percentage of the maximum output of the stimulator that is
2010). All statistical analyses were accomplished using SPSS v27
below the level of a response in the electromyographic recording
(IBM Corp 2020) and R (R Core Team 2017).
of a targeted muscle and does not produce a muscle twitch. rTMS
applied at similar intensities and frequencies or as a timed single
Study 2 pulse exerts specific effects in M1 (Siebner and Rothwell 2003;
Participants Lazzaro et al. 2008) that differ from effects seen when applied
Twenty healthy right-handed participants (mean age ± SD: more anterior or posterior to M1 (Johansen-Berg and Matthews
60.1 ± 7.2 years, range 50–74, 13 females), contributed TMS data 2002; Muellbacher et al. 2002; Shadmehr and Krakauer 2008). In
to Study 2; 12 of the participants also participated in Study 1 two separate sessions, participants performed two runs of the
(data were collected under the NCT02544503). The remaining motor learning task described above with each hand. In between
6 | Cerebral Cortex, 2024, Vol. 34, No. 5

the first and second runs, stimulation was applied to the extensor Table 1. Demographic and clinical characteristics of Studies 1
and 2 samples combined.
carpi ulnaris (ECU) hot spot of the left M1 using an air-cooled
figure-of-8 coil (70-mm wing diameter) or sham air-cooled figure- (n = 38)
of-8 coil (70-mm wing diameter) connected to a Magstim Rapid2
Mean (SD)
(Magstim Company, UK) (Buetefisch et al. 2011). The rTMS was
applied at a 1 Hz frequency for a total of 900 pulses for 15 min to Age, mean (SD) 61.9 (8.4)
the LM1 at subthreshold intensity of 90% resting motor threshold. Gender M/F (% male) 16/22
rTMS at subthreshold intensity reduces the spread of the rTMS- RBANSa
induced electric field and avoids muscle twitches during rTMS Immediate Memory 105.2 (13.9)
Visuospatial/Constructional 104.0 (16.1)
that could modulate central processing via sensory afferents and
Language 100.1 (7.9)
compromise blinding of participants with respect to the sham
Attention 106.4 (15.1)
stimulation. In the sham condition, the coil is discharged at
Delayed Memory 104.3 (12.8)
a constant minimal current regardless of the intensity set for Total 105.3 (12.3)
the stimulator. The sham stimulation induces an electric field
a Mean = 100, standard deviation = 15. Note because 12 participants’ data
amplitude roughly 5% that of real TMS with a similar spatial
were included in both Study 1 and Study 2, the number of total participants
extent in cortex (Opitz et al. 2015). In our study, the acoustic is smaller (n = 38) than the sum for each study (30 and 20 participants).
by-product associated with discharging the coil was matched in
the sham and rTMS condition which helps with the blinding of Table 2. Motor learning task and motor cortex characteristics of
the participants. The order of the sham and rTMS sessions was Study 1 sample.
counterbalanced across participants; however, all participants
(n = 30)
from Study 1 completed both the hand motor learning task and
structural MRI session prior to both Study 2 sessions. As part of Mean (SD)
a separate research question, participants also completed a third Baseline Motor Accuracya
session with stimulation at another strength; that data are not Right Hand 0.573 (0.164)
reported here. Participants were blinded to the type of stimulation Left Hand 0.545 (0.183)
and told they would receive two different stimulation protocols. Standardized Learning Value
The subthreshold intensity rTMS, the identical appearance of the Right Hand 0.384 (0.301)
sham and rTMS coil, and the matched auditory noise ensured Left Hand 0.253 (0.213)
the blinding of participants. Participants remained seated in a Active M1 ROI (mm3 )
dental chair throughout the procedure. Surface EMG was recorded Right Hemisphere 1401.77 (174.84)
Left Hemisphere 1131.10 (191.26)
from the right ECU muscle throughout the stimulation process
A1 ROI (mm3 )
and monitored at a sensitivity of 0.01 mV/div to ensure muscle
Right Hemisphere 777.37 (180.49)
relaxation. This sensitivity was also used to confirm subthreshold Left Hemisphere 883.37 (205.22)
stimulation. At the beginning of each session, the hot spot was Estimated Total Intracranial 1530124.21 (167988.10)
defined as the location of the largest MEP (motor evoked potential) Volume (mm3 )
amplitude in response to a single TMS pulse at a low intensity. The
a Proportion of hits in the first block of run1.
location was verified by absence of MEPs in locations within 1 mm.
This was accomplished by marking the preliminary hotspot on
the reconstructed MRI of the participants brain using a frameless
neuronavigation system (BrainSight, Rogue Research, Montreal,
Results
Canada) and then systematically probe the locations around this Study 1
preliminary hotspot. If other locations produced larger responses, Participants met all inclusion criteria and completed all aspects
another preliminary hotspot was defined and tested in a similar of the study with usable imaging data for analysis. No outliers
manner. Once the hotspot was verified, it was marked as the (> ±3 SD) or missing data were observed among any variables.
definite hotspot for this person to ensure accuracy and precision See Table 1 for demographic information on participants from
in coil position for targeting the M1 ECU hot spot during all both studies, and Table 2 for motor learning and cortical volume
measurements. The threshold in this hotspot was determined characteristics for Study 1.
using PEST (Parameter Estimation by Sequential Testing) (Awiszus Pearson correlation coefficients between age and all brain/be-
2003). havior variables revealed a significant relationship between age
and baseline accuracy for the right hand (r = −0.670, P < 0.001)
but not the left (r = −0.348, P = 0.060). Age did not correlate with
Statistical analysis—rTMS related effects on retention of SLV in either the right (r = −0.263, P = 0.160) or left (r = −0.090,
training-related improvement in accuracy P = 0.636) hand, Active M1-ROI volume in the right (r = −0.181,
Motor learning task accuracy was calculated for each participant P = 0.339) or left (r = −0.245, P = 0.191) hemisphere, A1 volume
for each combination of hand (contralateral/right or ipsilater- in the right (r = −0.122, P = 0.521) or left (r = −0.033, P = 0.862)
al/left), intervention (rTMS 90% or sham), and time (preinter- hemisphere, or estimated Total Intracranial Volume (r = −0.148,
vention or postintervention) according to the criteria described P = 0.436).
above. A mixed model ANOVA was calculated to determine
the main effects of hand, intervention and time on accuracy, Baseline motor skillfulness and motor learning
with subsequent ANOVAs or paired t-tests calculated to explore As expected, the hand motor learning task had demands on
interaction effects among these variables. Greenhouse–Geisser precision that varied with target size and fulfilled the criteria of
corrections were applied to adjust for lack of sphericity where Fitts’ law (Fitts 1954; Censor et al. 2012). The repeated-measures
appropriate. 2 (hand) × 4 (target size) ANOVA showed significant effects
 Cobia et al. | 7

Fig. 3. Performance on the motor learning task (n = 30). Task performance for each block (collapsed across target size) by a) mean accuracy (proportion
hits), and B) movement time (ms), plotted as a function of the number of training blocks. Error bars indicate standard error. As participants completed
different numbers of runs (each containing three blocks) depending on their performance, the final block of each participant served as the reference
point. The x-axis is structured in reference to the final training block, with preceding training blocks indicated as negative numbers in reference to the
final target block. Note the number of participants at each data point is different due to the differing numbers of runs needed to reach criterion (left
hand: Final to −2, n = 30; −3 to −5, n = 25; −6 to −8, n = 6; right hand: Final to −2, n = 30; −3 to −5, n = 19; −6 to −8, n = 11).

of both target size (F3,87 = 245, P < 0.001, η2 G = 0.55) and hand Relationships between M1 cortical volume and motor
(F1,29 = 10.5, P < 0.01, η2 G = 0.03) on accuracy, but no significant learning
interaction (F3,87 = 1.68, P = 0.18, η2 G = 0.006). Similarly, both Pearson correlation coefficients between estimated Total Intracra-
target size (F3,87 = 197, P < 0.001, η2 G = 0.65) and hand (F1,29 = 16.4, nial Volume and volume of the Active-M1 ROIs yielded no
P < 0.001, η2 G = 0.052) affected movement time, but target size significant results, indicating that variability in global head size
and hand did not interact with each other (F3,87 = 1.22, P > 0.2, did not relate to variability in gray matter volume of Active-
η2 G = 0.006). Consistent with Fitts’ law, accuracy increased M1 ROIs. As stated above, our primary goal was to determine
(S = 43.4 ± 2.3%, M = 65.7 ± 2.1%, L = 79.9 ± 1.8%, XL = 85.4 ± 1.5%) whether the gray matter volume in contralateral Active-M1 ROIs,
and movement time decreased (S = 879 ± 13 ms, M = 775 ± 12 ms, measured prior to motor learning, contributed to subsequent
L = 654 ± 10 ms, XL = 554 ± 13 ms) as a function of target size. learning of the hand motor task. Pearson coefficients revealed
Across all target sizes and trials, participants demonstrated that right hand SLV was significantly correlated with gray matter
greater accuracy and faster task performance with the dominant volume of the Left Hemi Active-M1 in the positive direction
right hand (accuracy = 71.4 ± 2.0%, movement time = 694 ± 15 ms) (r = 0.410, P = 0.025) (see Fig. 4) but right hand baseline accuracy
relative to the left hand (accuracy = 65.8 ± 2.1%, movement was not (r = 0.264, P = 0.158). For the left hand, neither SLV
time = 737 ± 13 ms). However, at baseline (the first block of the (r = −0.26, P = 0.892) nor baseline accuracy (r = −0.067, P = 0.724)
first run), accuracy between the right and left hands did not were significantly correlated with gray matter volume in the right
significantly differ (right hand: 57.3 ± 3.0%, left hand: 54.5 ± 3.3%, hemisphere Active-M1 ROI.
t29 = 0.85, P > 0.2). To determine whether the observed relationships between
As indicated in Fig. 3A–B, participants’ accuracy improved and M1 volumes and motor hand performance were specific to
movement time decreased with practice. Paired t-test results M1, primary auditory cortex (A1) was selected as a control
show that accuracy increased significantly between the first and cortical region given it is not directly involved in motor learning.
last blocks (left hand t29 = 7.36, P < 0.001; right hand t29 = 9.74, Pearson correlation coefficients between A1 gray matter volume
P < 0.001) and movement time significantly decreased (left hand in each hemisphere with baseline accuracy and SLV in the corre-
t29 = 5.06, P < 0.001; right hand t29 = 4.03, P < 0.001), indicating sponding contralateral hand were all nonsignificant: Left Hemi
that motor learning occurred. Participants with lower baseline A1 with right hand baseline accuracy (r = 0.063, P = 0.740) and
accuracy had a higher maximum possible gain in precision SLV (r = −0.080, P = 0.674); Right Hemi A1 with left hand baseline
across trials and did show larger gains in precision across the accuracy (r = 0.060, P = 0.755) and SLV (r = −0.032, P = 0.866).
experiment. In both hands, baseline accuracy was significantly
correlated with the amount of change in accuracy between Study 2
the first and last blocks (left hand r = 0.699, P < 0.001; right Twenty right-handed participants completed Left M1 rTMS pro-
hand r = 0.628, P < 0.001), indicating that participants who began tocols for Study 2. Data from the sham session were unavailable
with lower baseline accuracy experienced greater gains in for two participants—one participant reported migraine headache
precision across training. Overall, these results support the use after the first rTMS session and was excluded from participat-
of SLV as a measure of motor skill learning that accounts for ing in additional sessions, and one participant had a schedul-
differences in baseline accuracy and maximum possible gain in ing conf lict. Twelve Study 2 participants also participated in
precision. Study 1. The imaging data of the remaining eight participants
8 | Cerebral Cortex, 2024, Vol. 34, No. 5

the capacity for behavioral changes. Our specific approach to

identifying a region of M1 that was active during the execution of
the motor skill learning task using fMRI (“Active-M1”) is unique.
We show that greater gray matter volume within a task-relevant
M1 structure is related to better learning of a hand motor task
in the healthy mature human brain. As expected, this identified
M1 area (Active-M1) was in the anterior wall of the central sulcus
and posterior aspects of the precentral gyrus and included the
anatomically defined hand area in the omega-shaped hand knob
(Yousry et al. 1997). In studies of non-human primates, this M1
area also contains the CM neurons with direct monosynaptic out-
put from M1 to the spinal alpha motor neurons that skilled hand
movements depend on (Rathelot and Strick 2006). We, therefore,
demonstrate that the task-based fMRI labeling of the atlas based
M1 area (“Active-M1”) and the neurophysiologically TMS defined
hotspot (Siebner et al. 2022) co-localize in M1 hand area.
Performance on the skilled motor learning task was not related
Fig. 4. Relationship between skilled motor learning performance
(SLV = standardized learning value) of the right hand and gray matter to the cortical volume of A1, a brain area not involved in motor
volume of the active-M1 region of interest (ROI) in the left hemisphere. execution or motor learning, which supports the specificity of its
Right hand SLV is plotted on the y-axis against left hemisphere gray relationship to M1.
matter volume of active-M1 ROI along the x-axis. This correlation was Furthermore, the finding that learning of the left hand did not
statistically significant in the positive direction (n = 30).
significantly relate to the gray matter volume of Active-M1 of the
right hemisphere confirms the specificity of its relationship to the
were collected on a different scanner under a different imaging dominant hemisphere. These findings are in line with evidence to
protocol not optimized for cortical volumetric analysis. Their support that left M1 is involved with left and right hand learning a
data were therefore not considered for Study 1 but they par- serial reaction task (Grafton et al. 1995, 2002) which is consistent
ticipated in the same motor learning task as the remaining 12 with our findings of disrupted motor learning for both hands
participants. following rTMS of the left M1 (see below). Anatomical asymmetry
of the hand area in M1 has been demonstrated and related to
Functional relevance of the left M1 hand area to motor handedness using MRI of the brain (Amunts et al. 1996, 1997,
learning 2000) and cytoarchitectonic studies of histological sections of
Subthreshold stimulation of left M1 at 90% resting motor postmortem brains (Amunts et al. 1996). Finally, disruption of
threshold (RMT, rTMS90) disrupted motor learning compared to learning was only seen when M1 hand area was stimulated with
sham stimulation (Fig. 5). A 2 (hand: right/left) × 2 (intervention: rTMS but not with sham, which provides evidence for a causal link
rTMS90/sham) × 2 (time: preintervention/postintervention) between M1’s critical participation in the learning task.
ANOVA showed a significant main effect of hand (F1,17 = 23.4, Specifically, we assessed whether inter-individual differences
P < 0.001, η2 G = 0.087) and a significant time by intervention inter- in the gray matter volume of Active-M1 was related to inter-
action (F1,17 = 15.2, P < 0.001, η2 G = 0.016). No other main effects or individual differences in motor learning. We observed a positive
interactions reached significance. To better understand the time relationship between the amount of motor learning with the
by intervention interaction, we performed a two-way ANOVA for dominant right hand and the gray matter volume of left hemi-
each intervention separately. When sham stimulation was applied sphere Active-M1, a relationship not observed between motor
during the intervention period, there was a significant effect of learning in the non-dominant left hand and gray matter volume
time (F1,17 = 9.66, P < 0.01, η2 G = 0.041), indicating accuracy differed in right hemisphere Active-M1. To our knowledge, this is the first
between the preintervention (73.7 ± 2.2%) and postintervention demonstration of a link between M1 gray matter volume in a region
(78.3 ± 1.7%) assessments. There was also a significant effect of specifically delineated as task-relevant using fMRI and subsequent
hand (F1,17 = 14.6, P < 0.001, η2 G = 0.102) such that participants motor learning in the healthy mature human brain.
were more accurate with the right hand than the left hand; the Some early work described increased size in dorsal aspects of
time × hand interaction was not significant. For the rTMS90 the intrasulcal length of the precentral gyrus, thought to repre-
condition, only a significant main effect of hand was seen sent the hand area, in skilled keyboard players versus control
(F1,19 = 15.9, P < 0.001, η2 G = 0.067). Similar to the sham condition, participants, but the actual involvement of this M1 area in the
participants were more accurate with the right hand than the left task was not demonstrated (Amunts et al. 1997). In addition, using
hand, but there was no effect of time. The performance before voxel-based morphometry in a time-series design Wenger and
(76.8 ± 1.6%) and after (76.4 ± 1.7%) the application of rTMS at colleagues (Wenger et al. 2016) found expansion of the right M1 in
90% RMT did not significantly differ (F1,19 < 1, η2 G = 0.0004). response to 7 weeks of a left-hand writing task which normalized
despite continued practice and proficiency; however, the finding
was not localized to a particular M1 area and the functional
Discussion relevance is more indirect as the involvement in execution of the
Here, we report that a combined multimodal approach can be task was not demonstrated. (Huang et al. 2013; Lissek et al. 2013;
used to delineate task-relevant M1 structure and that greater gray Hamano et al. 2021).
matter volume in this task-relevant M1 structure is directly and Importantly, we provide evidence for a causal link between
causally related to subsequent gains in motor skill learning of the task relevant left hemisphere M1 area (left Active-M1 ROI,
the corresponding hand. This demonstrates that a functionally which is located in the M1 hand area) and learning with the
relevant structure in the healthy mature human brain determines corresponding right hand by disrupting learning when rTMS is
 Cobia et al. | 9

Fig. 5. Effect of the rTMS intervention on left- and right-hand motor performance (Study 2, n = 20). Performance on the motor learning task indexed by
accuracy (proportion hits) is plotted for preintervention (pre) and postintervention (post) timepoints for the sham and rTMS90% conditions. Error bars
indicate standard error. Participants’ accuracy improved between preintervention and postintervention testing for sham stimulation but not rTMS90%.

applied to the left M1 hand area. When rTMS is applied to M1 hand between neighboring representations without the involvement of
area, at the frequencies and intensities employed in the current nonadjacent M1 regions (Nudo et al. 1996; Sanes and Donoghue
study, it has been demonstrated to disrupt learning of a hand 2000; Rathelot and Strick 2006). More specifically, there is evidence
motor task likely by transiently disrupting local processes that are that larger baseline intracortical microstimulation (ICMS) maps
related to learning (Muellbacher et al. 2002). The effect of rTMS correlated with better accuracy at baseline and after training
applied at similar intensities and frequencies or as a timed single of a skilled hand motor task. For example, the monkey with
pulse exerts inhibitory or excitatory effects that are specific to M1 the smallest baseline distal ICMS map had the poorest baseline
(Siebner and Rothwell 2003; Lazzaro et al. 2008) and differ from performance on the pellet-retrieval task, while the monkey with
effects seen when applied more anterior to premotor cortex (PMC) the largest baseline distal ICMS map had the best baseline per-
or posterior to primary somatosensory cortex (S1) (Johansen-Berg formance on the skilled hand motor task. This also was true of
and Matthews 2002; Muellbacher et al. 2002; Hadipour-Niktarash posttraining performance (Nudo et al. 1996). While anecdotal, this
et al. 2007). finding is in line with the results of our fMRI experiment where
The effect of left M1 rTMS on early motor consolidation of task related activity was in the M1 hand area and also consistent
ipsilateral left-hand performance has not been tested before. with the notion that M1 is crucially involved in executing and
However, our finding of a bilateral effect with left M1 rTMS is learning skillful hand and finger movements.
consistent with its reported bilateral effect on M1 excitability Rodent studies offer compelling evidence regarding the mech-
(Fitzgerald et al. 2006) mediated through homotopic connections anisms underlying motor learning and that M1 contains the neu-
between M1 in both hemispheres (Ferbert et al. 1992; Lazzaro et al. ronal substrate to support processes related to learning related
1999). The results could also be explained within the framework of functional and structural changes in M1 (Sanes and Donoghue
hemispheric specialization where each hemisphere contributes to 2000). These studies demonstrate that the acquisition of motor
the control of hand movements and motor learning/adaptation. skills in adult rodents involves activity-dependent plasticity that
Specifically, in this model, the left hemisphere provides predic- is ref lected in both neuronal changes and reorganization of move-
tive control mechanisms specifying aspects such as movement ment representations within M1 (Rioult-Pedotti et al. 1998; Kleim
direction for movements with both the contralateral and the et al. 2002a; Monfils et al. 2005; Jones and Jefferson 2011) includ-
ipsilateral arms (see Mutha et al. 2012). Accordingly, learning a ing the formation and strengthening of synapses, as well as
visuomotor task such as the pointing task in this study is expected changes in dendritic branching, axonal morphology, and angio-
to demonstrate improved performance in accuracy for both hands genesis (Kleim et al. 1998; Kleim et al. 2002b; Kleim et al. 2002a).
through processes in left M1. The similarity of the metrics of While our approach does not allow anatomical specification on
motor learning between the left and right hand in our sample the micro- and ultrastructural level, one possible explanation
would suggest that the behavioral acquisition and trajectory of for greater cortical volume may result from more abundance
motor learning for this particular task occurs irrespective of the of vascular, neuronal, or glial structures thereby providing criti-
hand as the control mechanisms specifying these aspects resides cal anatomical structures to support superior performance with
in left M1. It is, therefore, conceivable that left M1 disruption with skilled motor learning (Asan et al. 2021). An alternative explana-
rTMS affects motor consolidation for improved precision of both tion is that active M1 network expanded in to areas that were
hands. formerly not active. The finding from the second study where
While movement representation in M1 is distributed and over- rTMS to M1, but not sham stimulation, disrupted the retention
lapping, it is primarily contained within the border of major body of practice-related improvements in accuracy provides the causal
parts such as face, arm, trunk, and leg, and short-term and long- link between the identified M1 structures for motor learning.
term M1 reorganization is marked by dynamically shifting borders Our results confirm earlier findings where the disruptive effect
10 | Cerebral Cortex, 2024, Vol. 34, No. 5

of rTMS on retention of the behavioral improvement was only the National Institute of Child Health and Human Development
present when applied to M1 but not seen with rTMS to other brain (R21HD067906-01A1).
areas, and only when applied immediately after the motor prac-
tice but not late after practice (Muellbacher et al. 2002). Together, Conf lict of interest statement: Derin Cobia currently serves as a con-
the results support the notion of rTMS-related interference with sultant for Sage Therapeutics on a project unrelated to the present
processes in M1 actively involved in early consolidation of gains study. Marc Haut has received speaker fees from Medtronic and
in skillfulness (i.e. accuracy). These processes involve LTP-like receives compensation from Taylor & Francis for serving as Editor
mechanisms as demonstrated in rodents (Rioult-Pedotti et al. for Aging, Neuropsychology, and Cognition, and compensation
1998; Rioult-Pedotti et al. 2000; Sanes and Donoghue 2000), non- from Insightec for study work. Randolph Nudo receives compen-
human primates (Plautz et al. 2000), and humans (Bütefisch et al. sation from American Society of Neurorehabilitation for serving
2000). as Editor-in-Chief for Neurorehabilitation and Neural Repair, and
Previous studies examining relationships between perfor- compensation from Elsevier for serving as Deputy Editor for Brain
mance and brain structure note that an increase in gray matter Stimulation. Kate Revill, Stephanie Rellick, Miles Wischneswski,
volume was related to the acquisition of specific skills, such and Cathrin Buetefisch all declare no competing financial inter-
as professional keyboard playing, juggling, and visuospatial ests.
navigation among taxi drivers in London (Maguire et al. 2000;
Draganski et al. 2004, 2006; Woollett and Maguire 2011). Given that
exceptional skillfulness is a result of extensive training, the strong
References
relationships between brain structure and behavior observed Ammons RB. Acquisition of motor skill: I. Quantitative analysis and
in these studies may be explained by the impact of significant theoretical formulation. Psychol Rev. 1947:54(5):263–281. https://
practice. However, it is noteworthy that the participants in [Link]/10.1037/h0062455.
our study did not demonstrate extraordinary performance in Amunts K, Schlaug G, Schleicher A, Steinmetz H, Dabringhaus
a specific skill, nor engage in extensive practice. This suggests A, Roland PE, Zilles K. Asymmetry in the human motor cor-
even normal range motor skill performance and learning can be tex and handedness. NeuroImage. 1996:4(3):216–222. [Link]
determined by interindividual differences in volume in M1 for org/10.1006/nimg.1996.0073.
“everyday” individuals (Bütefisch et al. 2000; Gaser and Schlaug Amunts K, Schlaug G, Jäncke L, Steinmetz H, Schleicher A, Dabring-
2003). This is consistent with previous reported association of haus A, Zilles K. Motor cortex and hand motor skills: struc-
gray matter and learning (Kanai and Rees 2011). tural compliance in the human brain. Hum Brain Mapp. 1997:5(3):
In conclusion, our study demonstrates that primary motor 206–215. [Link]
cortex structure, as measured by gray matter volume, determines C;206::AID-HBM5&#x003E;[Link];2-7.
gains in precision with hand motor practice, suggesting that Amunts K, Jäncke L, Mohlberg H, Steinmetz H, Zilles K. Interhemi-
structural brain resources determine the capacity for behavioral spheric asymmetry of the human motor cortex related to hand-
changes in the healthy mature human brain. Future research edness and gender. Neuropsychologia. 2000:38(3):304–312. https://
is necessary to define the underlying cellular, anatomical, and [Link]/10.1016/S0028-3932(99)00075-5.
ultrastructural characteristics that determine the relationship Asan L, Falfán-Melgoza C, Beretta CA, Sack M, Zheng L, Weber–
between cortical volume in M1 and gain in skillfulness. Further- Fahr W, Kuner T, Knabbe J. Cellular correlates of gray matter
more, given the older age of our sample, determination of whether volume changes in magnetic resonance morphometry identified
our findings are consistent across the age spectrum would require by two-photon microscopy. Sci Rep. 2021:11(1):4234.[Link]
examination in younger samples. org/10.1038/s41598-021-83491-8.
Awiszus F. TMS and threshold hunting. Suppl Clin Neurophysiol.
[Link]–23. [Link]
Acknowledgments
Barany DA, Revill KP, Caliban A, Vernon I, Shukla A, Sathian K,
We thank our participants for their involvement in this study. Buetefisch CM. Primary motor cortical activity during unimanual
movements with increasing demand on precision. J Neurophysiol.
Author contributions 2020:124(3):728–739. [Link]
Derin Cobia (Conceptualization, Data curation, Formal Analysis, Buckner RL, Head D, Parker J, Fotenos AF, Marcus D, Morris JC,
Investigation, Methodology, and Writing—original draft, Writing— Snyder AZ. A unified approach for morphometric and func-
review & editing), Marc Haut (Conceptualization, Data curation, tional data analysis in young, old, and demented adults using
Formal Analysis, Investigation, Methodology, and Writing— automated atlas-based head size normalization: reliability and
original draft, Writing—review & editing), Kate Revill (Con- validation against manual measurement of total intracranial
ceptualization, Data curation, Formal Analysis, Investigation, volume. NeuroImage. 2004:23(2):724–738. [Link]
Methodology, and Writing—original draft, Writing—review & [Link].2004.06.018.
editing), Stephanie Rellick (Software, Data curation), Randolph Buetefisch CM, Hines B, Shuster L, Pergami P, Mathes A. Motor
Nudo (Writing—review & editing), Miles Wischneswski (Method- demand-dependent improvement in accuracy following low-
ology), and Cathrin Buetefisch (Conceptualization, Investigation, frequency transcranial magnetic stimulation of left motor cor-
Methodology, Project Administration, Resources, Supervision, tex. J Neurophysiol. 2011:106(4):1614–1621. [Link]
Funding Acquisition, and Writing—original draft, Writing—review jn.00048.2011.
& editing). Buetefisch CM, Revill KP, Shuster L, Hines B, Parsons M. Motor
demand-dependent activation of ipsilateral motor cortex. J Neu-
rophysiol. 2014:112(4):999–1009. [Link]
Funding 2014.
This project was funded by the National Institute of Neuro- Bütefisch CM, Davis BC, Wise SP, Sawaki L, Kopylev L, Classen J,
logical Disorders and Stroke (R56NS070879, R01NS090677) and Cohen LG. Mechanisms of use-dependent plasticity in the human
 Cobia et al. | 11

motor cortex. Proc National Acad Sci. 2000:97(7):3661–3665. https:// J Neurosci. 2007:27(49):13413–13419. [Link]
[Link]/10.1073/pnas.97.7.3661. JNEUROSCI.2570-07.2007.
Censor N, Sagi D, Cohen LG. Common mechanisms of human per- Hamano YH, Sugawara SK, Fukunaga M, Sadato N. The integra-
ceptual and motor learning. Nat Rev Neurosci. 2012:13(9):658–664. tive role of the M1 in motor sequence learning. Neurosci Lett.
[Link] [Link]. [Link]
IBM Corp. IBM SPSS statistics for Macintosh V27. Armonk, NY: IBM Corp; Hamzei F, Glauche V, Schwarzwald R, May A. Dynamic gray mat-
2020. ter changes within cortex and striatum after short motor skill
Cox RW. AFNI: software for analysis and visualization of functional training are associated with their increased functional interac-
magnetic resonance Neuroimages. Comput Biomed Res. 1996:29(3): tion. NeuroImage. 2012:59(4):3364–3372. [Link]
162–173. [Link] neuroimage.2011.10.089.
Dale AM, Fischl B, Sereno MI. Cortical surface-based analysis I. Harms MP, Somerville LH, Ances BM, Andersson J, Barch DM,
Segmentation and surface reconstruction. NeuroImage. 1999:9(2): Bastiani M, Bookheimer SY, Brown TB, Buckner RL, Burgess
179–194. [Link] GC, et al. Extending the human connectome project across
Destrieux C, Fischl B, Dale A, Halgren E. Automatic parcellation of ages: imaging protocols for the lifespan development and aging
human cortical gyri and sulci using standard anatomical nomen- projects. NeuroImage. [Link]–984. [Link]
clature. NeuroImage. 2010:53(1):1–15. [Link] neuroimage.2018.09.060.
neuroimage.2010.06.010. Huang Y, Zhen Z, Song Y, Zhu Q, Wang S, Liu J. Motor training
Draganski B, Gaser C, Busch V, Schuierer G, Bogdahn U, May increases the stability of activation patterns in the primary
A. Changes in grey matter induced by training. Nature. 2004: motor cortex. PLoS ONE. 2013:8(1):e53555. [Link]
427(6972):311–312. [Link] [Link].0053555.
Draganski B, Gaser C, Kempermann G, Kuhn HG, Winkler J, Buchel Hyde KL, Lerch J, Norton A, Forgeard M, Winner E, Evans AC,
C, May A. Temporal and spatial dynamics of brain struc- Schlaug G. The effects of musical training on structural brain
ture changes during extensive learning. J Neurosci. 2006:26(23): development. Ann N Y Acad Sci. 2009:1169(1):182–186. [Link]
6314–6317. [Link] org/10.1111/j.1749-6632.2009.04852.x.
Elbert T, Pantev C, Wienbruch C, Rockstroh B, Taub E. Increased Johansen-Berg H, Matthews P. Attention to movement modulates
cortical representation of the fingers of the left hand in string activity in sensori-motor areas, including primary motor cor-
players. Science. 1995:270(5234):305–307. [Link] tex. Exp Brain Res. 2002:142(1):13–24. [Link]
science.270.5234.305. s00221-001-0905-8.
Ferbert A, Priori A, Rothwell JC, Day BL, Colebatch JG, Mars- Jones TA, Jefferson SC. Reflections of experience-expectant devel-
den CD. Interhemispheric inhibition of the human motor opment in repair of the adult damaged brain. Dev Psychobiol.
cortex. J Physiol. 1992:453(1):525–546. [Link] 2011:53(5):466–475. [Link]
jphysiol.1992.sp019243. Kanai R, Rees G. The structural basis of inter-individual differences
Fischl B, Sereno MI, Dale AM. Cortical surface-based analysis II: infla- in human behaviour and cognition. Nat Rev Neurosci. 2011:12(4):
tion, flattening, and a surface-based coordinate system. NeuroIm- 231–242. [Link]
age. 1999:9(2):195–207. [Link] Kleim JA, Barbay S, Nudo RJ. Functional reorganization of the
Fischl B, Rajendran N, Busa E, Augustinack J, Hinds O, Yeo BTT, rat motor cortex following motor skill learning. J Neurophysiol.
Mohlberg H, Amunts K, Zilles K. Cortical folding patterns and 1998:80(6):3321–3325. [Link]
predicting Cytoarchitecture. Cereb Cortex. 2008:18(8):1973–1980. Kleim JA, Barbay S, Cooper NR, Hogg TM, Reidel CN, Remple MS,
[Link] Nudo RJ. Motor learning-dependent synaptogenesis is localized
Fitts PM. The information capacity of the human motor system in to functionally reorganized motor cortex. Neurobiol Learn Mem.
controlling the amplitude of movement. J Exp Psychol. 1954:47(6): 2002a:77(1):63–77. [Link]
381–391. [Link] Kleim JA, Cooper NR, VandenBerg PM. Exercise induces angiogen-
Fitzgerald PB, Fountain S, Daskalakis ZJ. A comprehensive review esis but does not alter movement representations within rat
of the effects of rTMS on motor cortical excitability and motor cortex. Brain Res. 2002b:934(1):1–6. [Link]
inhibition. Clin Neurophysiol. 2006:117(12):2584–2596. [Link] S0006-8993(02)02239-4.
org/10.1016/[Link].2006.06.712. Lazzaro VD, Oliviero A, Profice P, et al. Direct demonstration of
Gaser C, Schlaug G. Brain structures differ between musicians interhemispheric inhibition of the human motor cortex produced
and non-musicians. J Neurosci. 2003:23(27):9240–9245. [Link] by transcranial magnetic stimulation. Exp Brain Res. 1999:124(4):
org/10.1523/JNEUROSCI.23-27-09240.2003. 520–524. [Link]
Gerber P, Schlaffke L, Heba S, Greenlee MW, Schultz T, Lazzaro VD, Pilato F, Dileone M, et al. The physiological basis
Schmidt-Wilcke T. Juggling revisited—a voxel–based of the effects of intermittent theta burst stimulation of the
morphometry study with expert jugglers. NeuroImage. [Link] human motor cortex. J Physiol. 2008:586(16):3871–3879. https://
320–325. [Link] [Link]/10.1113/jphysiol.2008.152736.
Grafton ST, Hazeltine E, Ivry R. Functional mapping of sequence Lehmann N, Villringer A, Taubert M. Colocalized white matter plas-
learning in Normal humans. J Cogn Neurosci. 1995:7(4):497–510. ticity and increased cerebral blood flow mediate the beneficial
[Link] effect of cardiovascular exercise on long-term motor learn-
Grafton ST, Hazeltine E, Ivry RB. Motor sequence learning with ing. J Neurosci. 2020:40(12):2416–2429. [Link]
the nondominant left hand. Exp Brain Res. 2002:146(3):369–378. JNEUROSCI.2310-19.2020.
[Link] Lissek S, Vallana Guido S, Güntürkün O, Dinse H, Tegenthoff M. Brain
Hadipour-Niktarash A, Lee CK, Desmond JE, Shadmehr R. activation in motor sequence learning is related to the level of
Impairment of retention but not Acquisition of a Visuomotor Skill native cortical excitability. PLoS ONE. 2013:8(4):e61863. https://
through time-dependent disruption of primary motor cortex. [Link]/10.1371/[Link].0061863.
12 | Cerebral Cortex, 2024, Vol. 34, No. 5

Maguire EA, Gadian DG, Johnsrude IS, Good CD, Ashburner J, Frack- Sampaio-Baptista C, Scholz J, Jenkinson M, Thomas AG, Filippini
owiak RSJ, Frith CD. Navigation-related structural change in the N, Smit G, Douaud G, Johansen-Berg H. Gray matter volume is
hippocampi of taxi drivers. Proc National Acad Sci. 2000:97(8): associated with rate of subsequent skill learning after a long
4398–4403. [Link] term training intervention. NeuroImage. [Link]–166. https://
Monfils M-H, Plautz EJ, Kleim JA. In search of the motor [Link]/10.1016/[Link].2014.03.056.
engram: motor map plasticity as a mechanism for encoding Sanes JN, Donoghue JP. Plasticity and primary motor cortex. Annu
motor experience. Neuroscientist. 2005:11(5):471–483. [Link] Rev Neurosci. 2000:23(1):393–415. [Link]
org/10.1177/1073858405278015. neuro.23.1.393.
Muellbacher W, Ziemann U, Wissel J, Dang N, Kofler M, Facchini S, Segonne F, Pacheco J, Fischl B. Geometrically accurate topology-
Boroojerdi B, Poewe W, Hallett M. Early consolidation in human correction of cortical surfaces using nonseparating loops.
primary motor cortex. Nature. 2002:415(6872):640–644. https:// Ieee T Med Imaging. 2007:26(4):518–529. [Link]
[Link]/10.1038/nature712. TMI.2006.887364.
Mutha PK, Haaland KY, Sainburg RL. The effects of brain lateral- Shadmehr R, Krakauer JW. A computational neuroanatomy for
ization on motor control and adaptation. J Mot Behav. 2012:44(6): motor control. Exp Brain Res. 2008:185(3):359–381. [Link]
455–469. [Link] org/10.1007/s00221-008-1280-5.
Nudo R, Milliken G, Jenkins W, Merzenich MM. Use-dependent alter- Siebner H, Rothwell J. Transcranial magnetic stimulation:
ations of movement representations in primary motor cortex of new insights into representational cortical plasticity. Exp
adult squirrel monkeys. J Neurosci. 1996:16(2):785–807. [Link] Brain Res. 2003:148(1):1–16. [Link]
org/10.1523/JNEUROSCI.16-02-00785.1996. 1234-2.
Oldfield RC. The assessment and analysis of handedness: the Edin- Siebner HR, Funke K, Aberra AS, Antal A, Bestmann S, Chen R,
burgh inventory. Neuropsychologia. 1971:9(1):97–113. [Link] Classen J, Davare M, di Lazzaro V, Fox PT, et al. Transcranial
org/10.1016/0028-3932(71)90067-4. magnetic stimulation of the brain: what is stimulated? – a
Opitz A, Legon W, Mueller J, Barbour A, Paulus W, Tyler WJ. Is sham consensus and critical position paper. Clin Neurophysiol: Off J
cTBS real cTBS? The effect on EEG dynamics. Front Hum Neurosci. Int Fed Clin Neurophysiol. [Link]–97. [Link]
[Link]. [Link] [Link].2022.04.022.
Plautz EJ, Milliken GW, Nudo RJ. Effects of repetitive motor training Strick PL, Dum RP, Rathelot J-A. The cortical motor areas and
on movement representations in adult squirrel monkeys: role the emergence of motor skills: a neuroanatomical perspective.
of use versus learning. Neurobiol Learn Mem. 2000:74(1):27–55. Annu Rev Neurosci. 2021:44(1):425–447. [Link]
[Link] annurev-neuro-070918-050216.
R Core Team. R: a language and environment for statistical computing. Wenger E, Kühn S, Verrel J, Mårtensson J, Bodammer NC, Lin-
Vienna, Austria: R Foundation for Statistical Computing; 2017. denberger U, Lövdén M. Repeated structural imaging reveals
Randolph C, Tierney MC, Mohr E, Chase TN. The repeatable battery nonlinear progression of experience-dependent volume changes
for the assessment of neuropsychological status (RBANS): pre- in human motor cortex. Cereb Cortex. [Link]hw141–bhw125.
liminary clinical validity. J Clin Exp Neuropsychology Neuropsychol- [Link]
ogy Dev Cognition Sect. 1998:20(3):310–319. [Link] Wischnewski M, Kowalski GM, Rink F, Belagaje SR, Haut MW, Hobbs
jcen.[Link]. G, Buetefisch CM. Demand on skillfulness modulates interhemi-
Rathelot J-A, Strick PL. Muscle representation in the macaque motor spheric inhibition of motor cortices. J Neurophysiol. 2016:115(6):
cortex: an anatomical perspective. Proc National Acad Sci. 2006: 2803–2813. [Link]
103(21):8257–8262. [Link] Woollett K, Maguire EA. Acquiring “the knowledge” of London’s
Rioult-Pedotti M-S, Friedman D, Hess G, Donoghue JP. Strengthening layout drives structural brain changes. Curr Biol. 2011:21(24):
of horizontal cortical connections following skill learning. Nat 2109–2114. : [Link]
Neurosci. 1998:1(3):230–234. [Link] Yousry TA, Schmid UD, Alkadhi H, Schmidt D, Peraud A, Buettner
Rioult-Pedotti MS, Friedman D, Donoghue JP. Learning-induced A, Winkler P. Localization of the motor hand area to a knob on
LTP in neocortex. Science. 2000:290(5491):533–536. [Link] the precentral gyrus. A new landmark. Brain. 1997:120(1):141–157.
org/10.1126/science.290.5491.533. [Link]