Corticospinal excitability during motor preparation of upper extremity reaches reflects flexor muscle synergies: A novel principal component-based motor evoked potential analyses

Abstract

Background:

Previous research has shown that noninvasive brain stimulation can be used to study how the central nervous system (CNS) prepares the execution of a motor task. However, these previous studies have been limited to a single muscle or single degree of freedom movements (e.g., wrist flexion). It is currently unclear if the findings of these studies generalize to multi-joint movements involving multiple muscles, which may be influenced by kinematic redundancy and muscle synergies.

Objective:

The objective of this study was to characterize corticospinal excitability during motor preparation in the cortex prior to functional upper extremity reaches.

Methods:

20 participants without neurological impairments volunteered for this study. During the experiment, the participants reached for a cup in response to a visual “Go Cue”. Prior to movement onset, we used transcranial magnetic stimulation (TMS) to stimulate the motor cortex and measured the changes in motor evoked potentials (MEPs) in several upper extremity muscles. We varied each participant’s initial arm posture and used a novel synergy-based MEP analysis to examine the effect of muscle coordination on MEPs. Additionally, we varied the timing of the stimulation between the Go Cue and movement onset to examine the time course of motor preparation.

Results:

We found that synergies with strong proximal muscle (shoulder and elbow) components emerged as the stimulation was delivered closer to movement onset, regardless of arm posture, but MEPs in the distal (wrist and finger) muscles were not facilitated. We also found that synergies varied with arm posture in a manner that reflected the muscle coordination of the reach.

Conclusions:

We believe that these findings provide useful insight into the way the CNS plans motor skills.

Keywords

Motor control muscle synergy movement planning cortical excitability stroke

1 Introduction

An essential part of daily life for many humans is manipulating their physical environment in reaction to sensory stimuli. For instance, someone who feels thirsty will see a glass of water and reach forward to pick it up and drink it. At a neuromuscular level, this process starts with sensory (e.g., visual) stimuli producing afferent neural signals that the central nervous system (CNS) uses as a cue to initiate the movement. The process of initiating the movement can be roughly broken down into a “planning” and “preparation” phases. During planning, the premotor cortex is believed to combine sensory and task-specific information to generate possible motor plans. Following this, the motor cortex enters the preparation phase by increasing the excitability of neurons representing the selected motor plan and inhibiting neurons representing alternative plans. This increase in excitability of the desired motor plan causes the release of efferent action potentials through the corticospinal tract that activate the relevant muscles (Bestmann & Duque, 2016; Cisek & Pastor-Bernier, 2014). Studying this process, particularly motor preparation, provides valuable insight into the CNS. Additionally, we can compare the preparation process with survivors of neurological injuries (e.g., stroke) to study how the injuries influence motor preparation.

We can examine motor preparation by measuring changes in cortical activity during the interval between the cue to initiate movement and movement onset. Researchers will typically use transcranial electrical (Rossini et al., 1988) or magnetic (Bestmann & Duque, 2016; Chen et al., 1998; Ibanez et al., 2020) stimulation (TES or TMS, respectively) to stimulate a participant’s primary motor cortex just before a real (Copithorne et al., 2015; Davey et al., 1998; Leocani et al., 2000) or imaged (Bianco et al., 2012; Facchini et al., 2002; Kumru et al., 2008; Zschorlich & Kohling, 2013) movement. Responses to the stimulation are recorded in the peripheral muscles groups that span the moving joint. These responses, called motor evoked potentials (MEPs), show the activity of the corticospinal tract (Bestmann & Duque, 2016). Previous studies using these approaches have shown that cortical excitability leading up to movement onset is characterized by initial inhibition, followed by an increase in excitability around 100 ms prior to movement onset (Chen & Hallett, 1999; Ibanez et al., 2020; Rossini et al., 1988). This initial inhibition is thought to allow the motor cortex to select between competing movement plans and prevent premature movements, while the subsequent increase in excitability corresponds to the facilitation of the selected motor plan and the inhibition of alternative plans (Bestmann & Duque, 2016; Labruna et al., 2014). Previous studies have also found that MEPs in muscles related to the movement are facilitated, while MEPs in unrelated or antagonist muscles are inhibited (Ganguly et al., 2021; Lebon et al., 2019; Reuter et al., 2015; Sohn & Hallett, 2004a, 2004b). This facilitation/inhibition is thought to represent the plan formed by the motor cortex: facilitated muscles will be activated during the movement, while inhibited muscles will not. Previous studies have also found that these processes are influenced by age (Reuter et al., 2015) and neurological disorders (Sohn & Hallett, 2004a).

However, these previous studies have usually only examined a single muscle or single degree of freedom movements (e.g., wrist flexion/extension), and have not considered more complex, multi-joint movements involving coordinated actions between multiple muscles. This distinction is particularly important when considering the upper extremity because it is a kinematically redundant system i.e., different paths in the joint space can create the same path in the end-point space. This redundancy has several potential consequences on motor preparation in the cortex. First, the distinction between muscles that are relevant and irrelevant to a movement are less clear. MEPs in muscles that seem irrelevant to the movement may be facilitated because the muscles serve a secondary function in the movement (e.g., biceps brachii providing anti-gravity compensation during elbow extension). Second, different paths in the joint space require different muscle coordination, and therefore the representation of the motor plan in the cortex may vary depending on the movement. Finally, previous research has hypothesized that the CNS reduces the complexity of musculoskeletal control by synchronously activating groups of muscles in fixed patterns, called “muscle synergies” (Sherrington, 1910; Ting, 2007; Ting & McKay, 2007; Tresch & Jarc, 2009). If these synergies are indeed fixed, they may additionally influence corticospinal excitability during movement preparation. Therefore, measuring changes in corticospinal excitability prior to the onset of multi-joint, upper extremity movements could reveal useful information regarding how the central nervous system plans and executes movement.

Previous studies have not examined changes in corticospinal excitability during preparation for a multi-joint upper extremity movement because such an examination would require producing MEPs in multiple muscles simultaneously. This requirement conflicts with conventional usage of TMS in neuroscience research, which relies on using the somatotopic properties of the motor cortex to find the position that produces the largest response in a single muscle (i.e., the muscle’s “hot spot”). In effect, this conventional process studies how neurons from multiple locations in the motor cortex converge on the same muscle i.e., the “convergent properties” of the corticospinal pathway (DeJong et al., 2021). However, previous studies have also shown that closely associated muscles have overlapping representations in the motor cortex (Raffin et al., 2015; Wilson et al., 1993). This finding led some researchers to use TMS to study how neurons from a single location in the motor cortex diverge to multiple muscles i.e., the “divergent properties” of the corticospinal pathway (DeJong et al., 2021; Melgari et al., 2008). These studies have found that it is possible to use TMS to produce MEPs in multiple upper extremity muscles (both proximal and distal) simultaneously (DeJong et al., 2021; Melgari et al., 2008), and that the coupling between muscle MEPs is independent of stimulus intensity (DeJong et al., 2021). Therefore, it may be possible to use the divergent properties of the corticospinal pathway to study how corticospinal excitability modulates during preparation of a multi-joint upper extremity task.

Therefore, the objective of this study was to leverage the divergent properties of the corticospinal pathway to examine how motor plans are expressed in the cortex prior to the onset of upper extremity movements. Specifically, we used single-pulse TMS to measure MEPs prior to the onset of a functional reaching task in several key upper extremity muscles. During the experiment, we varied muscle coordination by altering each participant’s initial arm posture so that different muscles were required to initiate the movement. We then employed a novel, synergy-based analysis to examine which MEP patterns were being facilitated during movement preparation. We hypothesized that MEPs would be facilitated in all upper extremity muscles relative to rest due to the multiple joints and muscles necessary to initiate the reach. Further, we hypothesized that corticospinal excitability would modulate with initial arm posture, as the arm posture would necessitate different muscle coordination to complete the reach.

2 Methods and materials

2.1 Participants

20 right-hand dominant adults (12 males, 8 females, age 22.1 yr±4.5 yr) with no history of major orthopedic or neurological conditions, recent injuries in their upper extremities, uncontrolled diabetes or hypertension, and/or contraindications to transcranial magnetic stimulation (i.e., recent head injury, metallic implants in the brain/skull, cardiac pacemaker, cochlear implants, recent history of seizures, history of recurrent fainting episodes, pregnant or actively trying to conceive, etc.) participated in this study. Participants provided informed, written consent prior to participation, and all protocols received approval from the University of Michigan Institutional Review Board.

2.2 Experimental apparatus

Participants were seated at a table with a cup placed in front of them (Fig. 1(A)). A computer monitor was also placed in front of the participant that displayed a green indicator light. Participants were instructed to wait for the green indicator to illuminate (i.e., the “Go Cue”), and then extend their dominant arm towards the target cup and extend their fingers, as if they were reaching out to grab it. Participants were instructed to react as quickly as possible to the Go Cue, but to move their arm towards the cup at a comfortable, self-selected pace. Participants were also instructed to keep their trunk stationary during the reach. The Go Cue was controlled by a custom program written in LabView (National Instruments Corp., Austin, TX USA) that could trigger a transcranial magnetic stimulator following an experimenter-defined time delay after the Go Cue (Fig. 1(B)). When a participant’s reaction time was known (i.e., the time between the Go Cue and movement onset), the delay could be used to deliver a stimulation at a desired time prior to movement onset.

Fig. 1

(A) The experimental apparatus. Each participant sat at a height-adjustable table facing a computer monitor. A cup was placed on the table in between the participant and the monitor, just outside of the participant’s reach. A green light was displayed on the monitor that would periodically light up. Participants were instructed to wait for the green light to illuminate, which was a “Go Cue” for the participants to extend their dominant arm (the left arm in the figure) forward as if they were going to grab the cup. During the experiment, a figure-of-eight transcranial magnetic stimulation (TMS) coil would periodically produce a single, monophasic magnetic pulse to stimulate the participant’s contralateral primary motor cortex. The responses to these stimulations, called MEPs, were measured using EMG sensors placed over key upper extremity muscles on the dominant arm. Marker clusters placed on the coil and on lenseless glasses worn by the participant were measured by a motion capture system to ensure consist coil placement. (B) TMS stimulation procedure: The Go Cue was controlled by a custom LabView program which triggered a TMS pulse following an experimenter-defined time delay. The time delay of the TMS pulse was configured so that it occurred at a desired time interval prior to movement onset.

Participants reached for the cup starting from one of two possible arm postures: “Elbow Down” and “Elbow Up” (Fig. 2). In the “Elbow Down” position, the participants’ shoulder and forearm were in the neutral positions with the elbow flexed to 90 degrees. Their hand rested on their thigh with their fingers lightly flexed into a fist. In the “Elbow Up” position, the participants’ forearm and elbow rested on the table in front of them with their fingers lightly flexed into a fist in their sagittal plane. In this position, the shoulder was abducted, the elbow was flexed to 90 degrees, and the forearm was pronated. In the Elbow Up position, the height of the table was adjusted such that the participant’s elbow was just below the height of their shoulder.

Fig. 2

The two initial arm postures used in the experiment. (A) The Elbow Down posture. In this posture, the participant sat with their shoulder in the neutral position and elbow resting against their side. The elbow was flexed to 90°, the forearm was in the neutral position, and the hand rested on the participant’s ipsilateral thigh with the fingers lightly flexed into a fist. (B) The Elbow Up posture. In this posture, the participant sat with their shoulder abducted and their elbow resting on the table. The elbow was flexed to > 90° with the forearm pronated, the fingers flexed into light fist, and the hand in the sagittal plane.

We used these two postures because each required different muscles to initiate the reach. Specifically, the Elbow Down posture initially required shoulder flexion and elbow extension, with wrist and finger extension later in the reach to open the fingers. It is important to note, however, that gravity assisted elbow extension in this posture, therefore the biceps brachii both assisted in shoulder flexion and maintained a constant elbow position before elbow extension occurred. Following this, we expected that the muscle coordination pattern immediately following movement onset would be dominated by the anterior deltoids, biceps brachii, and triceps brachii. The Elbow Up posture, on the other hand, initially required horizontal shoulder adduction and vertical shoulder abduction to rotate and lift the humerus, respectively, as well as elbow extension to extend the arm, and wrist extension to counteract gravity and keep the hand parallel with the forearm. Later in the reach, the participant would extend their wrist and fingers to grasp the cup. Thus, we expected that the muscle coordination just after movement onset in the Elbow Up posture would be dominated by the anterior and middle deltoids, triceps brachii, and wrist extensors. When comparing these two postures, we expected that the anterior deltoid and biceps brachii to play larger roles in the Elbow Down posture than the Elbow Up posture, and the middle deltoid, triceps brachii, and wrist extensors to play larger roles in the Elbow Up posture. Therefore, if corticospinal excitability prior to movement onset reflects the subsequent muscle coordination, we expected that these muscle coordination patterns would emerge in the evoked responses prior to movement onset.

2.3 Experimental protocol

The experiment was split into two sections, one for each arm posture. The order of the postures was pseudorandomized for each participant. In the first section, participants performed a rest block, a movement onset block, and five reaching blocks in the first arm posture. The resting block consisted of ten TMS stimulations to the participant’s contralateral motor cortex while the participant kept their dominant arm relaxed. The movement onset block consisted of ten reaches from which we computed the mean movement onset time during reaching. During the movement onset block, participants reached for the cup as soon as they saw the Go Cue. During this block, we measured the time between the Go Cue and the participant’s movement onset for each reach to quantify movement onset, and no stimulations were delivered. Following the movement onset block, we performed the five reaching blocks. During each reaching block, the participant reached for the cup in response to the Go Cue ten times, and each time a TMS stimulation occurred following the appearance of the Go Cue and prior to movement onset. In the first reaching block, stimulations occurred 200 ms prior to the onset of each reach. In the following reaching blocks, we reduced the time between stimulations and movement onset in 50 ms increments until the last reaching block, where the stimulations occurred 0 ms prior to (or at) movement onset. The timing of the stimulation, i.e., the stimulation time delay, used in these reaching blocks was determined based on the movement onset time computed from the movement onset block. In the second section, participants repeated the rest and the five reaching blocks in the second arm posture. Following data collection, we recorded each participant’s maximum voluntary isometric contraction in shoulder flexion and abduction, elbow flexion and extension, and wrist flexion and extension while the participants’ arm were in the Elbow Down posture. Specifically, participants attempted to perform each of these joint motions with their maximum effort, and an experimenter provided manual resistance to these joint motions so that the contractions were isometric and remained in the same arm posture (Krishnan et al., 2013; Washabaugh et al., 2019; Washabaugh et al., 2016).

2.4 Electromyography and TMS

During the experiment, we recorded each participant’s electromyography (EMG) to measure their responses to the TMS (i.e., motor evoked potentials [MEPs]), as well as their muscle activity. To measure EMG, we placed single-use, Ag/AgCl electrodes (22 mm electrode spacing, NOROTRODE 20 Bipolar SEMG Electrodes, Myotronics, Kent, WA) over the muscle bellies of the anterior deltoid (AD), middle deltoid (MD), biceps brachii (BB), triceps brachii (TB), extensor digitorum (WE), and flexor digitorum superficialis (WF) of each participant’s dominant arm. Prior to electrode placement, the skin at each electrode site was cleaned with alcohol pads to minimize skin impedance, and electrode cream (Signacreme, Parker Laboratories, Inc., Fairfield, NJ) was applied to electrodes to increase conductivity between the skin and the electrode. Electrodes were fixed to the skin with the double-sided adhesive layer (applied to the electrode backing by the manufacturer) and further secured with tape. Electrodes were connected to preamplifiers (x20 differential gain, MA-422, Motion Lab Systems, Baton Rouge, LA) with snap connectors. Signals from each preamplifier were filtered with a 1000 Hz low-pass analog filter (x350 gain, MA300, Motion Lab Systems, Baton Rouge, LA) to eliminate signal aliasing, and then sampled at 2000 Hz with an 18-bit National Instruments Data Acquisition system (NI USB-6218, National Instruments Corp., Austin, TX USA). Sampled signals were recorded using a custom program written in LabView 2014 (National Instruments Corp., Austin, TX USA). Prior to data collection, the quality of each EMG signal was visually inspected to ensure proper electrode placement.

We used a custom program written in LabView to determine EMG onset from the movement time block. This program rectified the electromyography recorded during the block, segmented the data into 300 sample sections starting from the appearance of the Go Cue, and then ensemble averaged these segments to get an average activation profile for each muscle. Following this, the program provided an initial guess of EMG onset based on the instance that biceps brachii activation exceeded five times its resting standard deviation (SD). The experimenter would then visually examine the EMG onset of all muscles and compare their activations with the initial guess. If the experimenter felt that this initial guess provided a suitable approximation of EMG onset, then this guess was used. However, if the initial guess provided an inaccurate approximation (e.g., another muscle activated more quickly than the biceps brachii or EMG onset was slow such that 5x the resting SD appeared notably later than visible EMG onset), then a manually adjustable cursor was used to set the EMG onset time based on the muscle with the shortest EMG onset latency.

We performed TMS with a 70 mm figure-of-eight coil connected to a monophasic magnetic stimulator (Magstim 200, Magstim, U.K.). The coil was positioned over the participant’s primary motor cortex contralateral to their dominant upper extremity. The coil was oriented to produce posterior-to-anterior current flow in cortex; the handle of the coil oriented 45° from the participant’s midline (pointed posterior). To find each participant’s optimal coil position for stimulation (i.e., the “hot spot”), we initially placed the coil at the estimated hot spot (over the contralateral hemisphere, 3 cm lateral and anterior to the vertex), and made small adjustments. During hot spot finding, we looked for a location that produced visible MEPs in all muscles. We note that there were instances where the only location that could be found was that TMS produced MEPs in all muscles except one deltoid. In this case, we used the location that produced MEPs in middle deltoid and all other muscles, as middle deltoid was generally more excitable during both Elbow Up and Elbow Down conditions. To ensure that we retained consistent coil position and orientation during data collection, we fixed a cluster of three retroreflective markers (9 mm diameter) to the coil and had participants wear custom, lenseless glasses with five retroreflective markers attached to them. The positions of both the coil and glasses were tracked using an OptiTrack V120: TRIO camera and Motive Motion Capture Software (Version 1.8.0, 120 Hz). The positions and orientations of the coil and the participant’s head, as well as the hot spot, were displayed to the experimenter during data collection using NeuRRoNav, a low-cost, open-source software for navigated TMS (Rodseth et al., 2017). All stimulations were performed at 100% resting motor threshold (RMT), which was determined as the single lowest stimulator intensity that produced a distinct MEP waveform in each muscle more than 50% of the time. 100% RMT was chosen to minimize the intensity of the stimulus, which is recommended when studying divergent properties of the motor cortex to avoid excessive cortical spread of the stimulus (Melgari et al., 2008). Threshold finding was conducted using the MTAT 2.0 program (MTAT 2.0)) (Awiszus & Borckardt, 2011).

2.5 Data analysis

To measure the MVIC of each muscle, we used a custom program written in MATLAB (R2019b, MathWorks, Natick, MA) to process the raw EMG signals recorded while the participant produced their maximum contractions. The raw EMG signals were processed using a linear envelope technique. This processed involved digitally filtering the EMG signals first with a 20-500 Hz band-pass filter (zero-lag, 4^th order Butterworth) and then a 58-62 Hz band-rejection filter (zero-lag, 4^th order Butterworth) to eliminate surrounding electrical noise. We then deducted the mean from the filtered signals to remove the DC gain, rectified the signals, and smoothed them with a digital, 6 Hz low-pass filter (zero-lag, 4^th order Butterworth). The maximum smoothed value observed in each muscle was used as that muscle’s MVIC in subsequent analysis.

To measure MEPs, each EMG channel was segmented by the 100 ms (200 samples) following each stimulation. Using a custom MATLAB program, we manually selected the time window in each data segment during which the MEP occurred, and then computed the peak-to-peak amplitude (i.e., voltage maximum minus minimum) in this time window. This peak-to-peak value was then normalized to the corresponding muscle’s MVIC value to quantify the MEP. Note that the MEP data was not processed using the same linear envelope technique, as it would remove the negative component and create signal distortion, which would make peak-to-peak analysis impossible.

To examine how MEP patterns across muscles varied between arm postures during movement preparation, we performed a muscle synergy-based analysis (Augenstein et al., 2020; Ranganathan & Krishnan, 2012; Roh et al., 2012) to examine what MEP couplings emerged as stimulations were delivered closer to movement onset. Specifically, for each participant, the MEPs from all blocks in the same arm posture (i.e., the resting block and five reaching blocks) were arranged into a matrix $D \in ℝ^{60 \times 6}$ . In D, the rows denoted MEP instances organized by block (i.e., rows 1-10 were MEPs from the resting block, rows 11-20 were from the reaching block when stimulations occurred 200 ms prior to movement onset, rows 21-30 were from the reaching block when stimulations occurred 150 ms prior to movement onset, etc.) and the columns denoted each of the six muscles. We performed principal component analysis on the matrix D to decompose it into two separate matrices: a synergy matrix $W \in ℝ^{6 \times N}$ and an activation matrix $H \in ℝ^{60 \times N}$ , such that N is the number of extracted synergies (1 ≤ N <the number of muscles) and $D \approx \bar{D} + H W^{T}$ (the $\bar{D}$ denotes the mean of D). Here, each column of W can be thought of as an MEP coordination patterns, or “MEP synergy”, that emerged from the participant’s MEP data in a single arm posture. The value of each element in the columns of W reflects the relative contribution of each muscle to the synergies in that arm posture (i.e., W (i, j) is the contribution of the ith muscle to the jth synergy). In H, each column j can be thought of as the activation of the jth synergy that varies with the MEPs from the same arm posture (i.e., H (i, j) is the activation of the jth synergy in the ith MEP). It is important to note that because we analyzed all blocks within an arm posture, a W and H matrix were computed for each arm posture (i.e., there were W and H matrices for the Elbow Up posture, and separate W and H matrices for the Elbow Down posture). For each arm posture, we then produced a simplified version of H, called $\tilde{H} \in ℝ^{6 \times N}$ , whose elements were elements of H averaged across rows that came from the same block (e.g., the first row represented of $\tilde{H}$ the resting block, the second row represented the reaching block when stimulations occurred 200 ms prior to movement onset, etc.). Therefore, each row of $\tilde{H}$ denotes the average synergy activation in each block. Following this analysis, each participant had a W and $\tilde{H}$ matrix for each arm posture. We aligned each synergy across participants based on minimum deviation, defined as inverse cosine of the dot product (Ranganathan & Krishnan, 2012). We then averaged each W and $\tilde{H}$ matrix across all participants to get average W and $\tilde{H}$ matrices for each arm posture, and then aligned these average synergies across arm postures. N was selected such that the extracted synergies explained, on average, 90% of the variance in D across all participants (Augenstein et al., 2020; Roh et al., 2012).

It is important to consider how the behavior of each muscle would appear in this synergy-based analysis and what conclusions we can draw about motor preparation. This is because PCA is a measure of covariance, and therefore some results may seem counterintuitive when comparing them with average MEPs. Specifically, muscles whose MEPs covary (e.g., individual MEPs rise and fall simultaneously) will be placed into the same synergy and will receive larger weights in W than muscles that do not covary. Therefore, muscles with low resting MEPs that vary across reaching blocks will have larger weights than muscles with larger resting MEPs that do not vary throughout the experiment. Furthermore, muscles with large resting MEPs that vary across reaching blocks will receive equal weights to muscles with smaller resting MEPs that vary the same amount. This reveals two critical features of this analysis: it allows us to (1) separate out potentially conflating contributions of resting excitability on our analysis, and (2) use weights to compare excitability changes between muscles. Furthermore, if a subset of muscles varied across blocks, these muscles would be grouped into a single synergy whose activation (column of $\tilde{H}$ ) increased as stimulations were delivered closer to movement onset. In this synergy, a positive weight would mean that the corresponding muscle was becoming more facilitated during movement preparation. Likewise, a negative value would mean that the muscle was becoming more inhibited during movement preparation. This MEP synergy would therefore indicate the motor plan being facilitated by the corticospinal tract as part of movement preparation. Conversely, if activation of a synergy did not increase as stimulations were delivered closer to movement onset, this indicated that the observed synergy was not important to the motor plan formed in the cortex.

To validate that observed MEPs were not influenced by muscle activity prior to stimulations, we examined the EMG signals during the rest and five reaching blocks. To process this EMG data, we used the same filtering and smoothing process that was applied to the MVIC (i.e., linear envelope) and then normalized this data to the MVIC. Using the smoothed and normalized signals, we computed the average activation within each muscle during the 100 ms prior to the TMS stimulation. We then averaged these values across participants to get the average muscle activity prior to the stimulation for each muscle in each block. To validate that recorded muscle activity after movement onset reflected the hypothesized muscle coordination in the two postures, we examined the EMG signals from the first reaching block. To process this EMG data, we used the same filtering, smoothing, and normalization process described above, and then computed the average activation within each muscle during the first 100 ms following movement onset. We averaged across participants to get the average muscle activity for each muscle in each arm posture.

2.6 Statistical analysis

We performed our statistical analysis using IBM SPSS (Statistical Product and Service Solution, Version 27). Descriptive statistics was computed for MEP amplitudes, movement onset times, and muscle activation before stimulation and following movement onset. For each MEP synergy (i.e., column of W), we performed a two-way repeated measures analysis of variance (ANOVA) with muscle (AD, MD, BB, TB, WE, and WF) and arm posture (Elbow Up and Elbow Down) as within-subjects factors. For each simplified synergy activation profile (i.e., column of $\tilde{H}$ ), we performed a two-way repeated measures ANOVA with block (resting and 200 ms, 150 ms, 100 ms, 50 ms, and 0 ms prior to movement onset reaching blocks) and arm posture (Elbow Up and Elbow Down) as within-subjects factors. A significant main or interaction effect was investigated by post-hoc paired t-test with Sidak correction for multiple comparisons. We used α=0.05 as the significance level for all analyses. If sphericity was found to be violated, we used Greenhouse-Geisser corrected significance levels.

3 Results

3.1 Descriptive statistics

The non-normalized resting MEPs in both arm postures are shown in Fig. 3. The average movement onset time was 270±6 ms (mean±SEM) and 263±11 ms for the Elbow Up and Elbow Down postures, respectively, and the average EMG onset for each muscle is shown in Supplemental Table 1. The MEPs during each reaching block from a representative participant are shown in Fig. 4. In this participant, MEPs in the proximal muscles increased as the stimulus was delivered closer to movement onset, with the largest MEPs occurring between 100 and 0 ms prior to movement onset. At stimulation timings 50 and 0 ms prior to movement onset, MEPs in the proximal muscles increased above the resting MEP. In this participant, the stimulation timing did not modulate the excitability of the distal muscles (WE and WF), and the BB showed the largest increases in MEP as the stimulation timing approached movement onset. The MD showed larger MEPs in the Elbow Up posture, while the BB showed larger MEPs in the Elbow Down posture. MEPs in the AD, TB, WE, and WF appeared similar between postures. The average MEP for each muscle and arm posture across all subjects are shown in Table 1 and Fig. 5. The average muscle activation prior to the stimulations is shown in Table 2, and the average muscle activation following movement onset is shown in Fig. 6. Qualitatively, we saw higher activation of the BB when initiating the reach from the Elbow Down position, and higher activation of the MD, TB, and WE when initiating the reach from the Elbow Up position. These findings roughly aligned with our hypothesized muscle coordination as described in the Methods and Materials.

Fig. 3

Average raw (non-normalized) resting MEPs in both arm postures. Error bars denote standard error of the mean. Note that the proximal muscles and distal muscles have different scaling on the vertical axis.

Fig. 4

Ensemble average MEP from a typical participant in the measured upper extremity muscles and the in the two arm postures. Here, AD = Anterior Deltoid, MD = Middle Deltoid, BB = Biceps Brachii, TB = Triceps Brachii, WE = Wrist and Finger Extensors, and WF = Wrist and Finger Flexors.

Table 1

Mean±St. Err. of the Mean MEP Amplitude (V/V) during the Resting and Reaching Trials

Elbow Down
Timing	AD	MD	BB	TB	WE	WF
Rest	0.070±0.014	0.102±0.025	0.247±0.073	0.125±0.026	1.414±0.375	1.025±0.200
200 ms	0.095±0.022	0.210±0.053	0.636±0.174	0.199±0.049	1.672±0.401	1.529±0.329
150 ms	0.225±0.058	0.247±0.055	0.812±0.229	0.239±0.052	1.709±0.403	1.515±0.321
100 ms	0.348±0.065	0.280±0.066	1.020±0.231	0.295±0.060	1.719±0.410	1.550±0.310
50 ms	1.609±0.350	0.645±0.107	2.819±0.562	0.574±0.123	1.885±0.368	1.881±0.300
0 ms	2.305±0.530	0.979±0.198	3.730±0.790	0.750±0.181	2.158±0.416	2.101±0.280
Elbow Up
Timing	AD	MD	BB	TB	WE	WF
Rest	0.134±0.040	0.210±0.061	0.345±0.124	0.201±0.045	1.396±0.388	1.036±0.187
200 ms	0.347±0.109	0.307±0.078	0.687±0.168	0.247±0.052	1.784±0.465	1.700±0.359
150 ms	0.371±0.097	0.334±0.061	0.787±0.217	0.240±0.047	1.700±0.429	1.651±0.335
100 ms	0.537±0.129	0.459±0.095	0.787±0.235	0.317±0.061	1.884±0.490	1.484±0.282
50 ms	1.582±0.324	1.037±0.165	1.783±0.410	0.676±0.109	2.163±0.475	1.360±0.231
0 ms	2.481±0.490	1.666±0.340	1.786±0.297	1.044±0.176	2.534±0.512	1.244±0.198

Fig. 5

Average MEP in each arm posture and block. Here, the error bars denote the standard error of the mean.

Fig. 6

Participant and average muscle activation following movement onset between arm postures. Here, the error bars denote standard error of the mean.

Table 2

Mean±St. Err. of the Mean EMG Magnitude (% MVIC) during the Resting and Reaching Trials

Elbow Down
Timing	AD	MD	BB	TB	WE	WF
Rest	0.27±0.04	0.93±0.15	0.21±0.03	1.29±0.20	0.92±0.22	2.06±0.27
200 ms	0.28±0.04	0.95±0.16	0.25±0.04	1.26±0.21	0.99±0.21	2.08±0.29
150 ms	0.28±0.05	0.93±0.15	0.23±0.04	1.28±0.21	0.94±0.21	2.01±0.27
100 ms	0.31±0.05	0.92±0.15	0.28±0.04	1.31±0.21	1.1±0.23	2.07±0.29
50 ms	0.83±0.17	1.03±0.16	0.84±0.16	1.32±0.21	1.35±0.25	2.20±0.30
0 ms	2.92±0.62	1.39±0.25	2.53±0.69	1.49±0.23	2.48±0.56	3.02±0.47
Elbow Up
Trial	AD	MD	BB	TB	WE	WF
Rest	0.5±0.1	1.16±0.19	0.38±0.07	1.45±0.21	1.32±0.27	2.69±0.40
200 ms	1.19±0.2	1.43±0.19	0.56±0.1	1.51±0.21	1.22±0.26	2.66±0.40
150 ms	1.09±0.21	1.37±0.18	0.54±0.09	1.48±0.20	1.13±0.24	2.57±0.32
100 ms	1.16±0.22	1.48±0.19	0.56±0.10	1.52±0.20	1.24±0.27	2.67±0.36
50 ms	1.69±0.31	1.64±0.20	0.71±0.21	1.62±0.20	1.57±0.33	2.59±0.36
0 ms	3.19±0.58	2.22±0.34	1.41±0.42	1.92±0.22	3.28±0.86	2.67±0.38

The average MEP synergies (columns of matrix W) are shown in Fig. 7(A). On average, two synergies were sufficient to explain at least 90% of variance in the MEP data (93.7±4.4% in Elbow Down, 91.1±6.4% in Elbow Up). In the Elbow Down posture, the first and second synergies explained 80.7% ±3.3% and 13.0% ±2.6%, respectively. In the Elbow Up posture, the first and second synergies explained 68.0% ±5.1% and 23.1% ±5.0%, respectively. The first synergy in both postures generally had large weights (i.e., principal component coefficients) in most of the muscles, while the second synergy generally only had large weights in the distal muscles (Fig. 7(A)). The average synergy activation profiles (columns of matrix $\tilde{H}$ for each arm posture) are shown Fig. 7(B).

Fig. 7

MEP synergy analysis. (A) The MEP synergy weights in each arm posture across muscles and (B) The MEP synergy facilitation in each arm posture across blocks. Here, the error bars denote standard error of the mean.

3.2 MEP synergy analyses: synergy weight matrix

In the first synergy, we detected a significant muscle-by-posture interaction effect (p < 0.001). Post-hoc analysis comparing postures within each muscle revealed that weights of the anterior deltoid (AD), middle deltoid (MD), and wrist extensors (WE) were significantly larger in the Elbow Up position (p = 0.015, p < 0.001, and p = 0.004, respectively), while the weights of the biceps brachii (BB) and wrist flexors (WF) were significantly larger in the Elbow Down position (p < 0.001 and p < 0.001, respectively). There were no significant differences between postures in the weights of the triceps brachii (TB, p = 0.11). Post-hoc analysis comparing muscles within posture indicated that, within the Elbow Down position, the weights corresponding to the AD and BB were significantly larger than all other muscles (all p < 0.02). Within the Elbow Up position, all muscle weights were significantly larger than the weights of the WF (all p < 0.001). Additionally, the AD weights were significantly larger than the MD and TB (p = 0.004 and p < 0.001, respectively). In synergy 2, we found a significant main effect of muscle (p < 0.001), but no significant main effect of posture (p = 0.264) or interaction effect (p = 0.186). Post-hoc analysis of the main effect of muscle revealed that the weights corresponding to the WE were significantly larger than all muscles except for the WF (all p except WF < 0.001, WF p = 0.177). The weight corresponding to the WF was significantly different than the MD (p = 0.044).

3.3 MEP synergy analysis: synergy activation matrix

In the first synergy activation profile, we detected a significant muscle-by-posture interaction effect (p < 0.036). Post-hoc analysis comparing block (resting, 200 ms, 150 ms, 100 ms, 50 ms, and 0 ms prior to movement onset) within each posture revealed that in the Elbow Up posture, synergy activation during resting stimulations was significantly lower than activation during stimulations 50 and 0 ms prior to movement onset (p = 0.001 and p < 0.001, respectively). Additionally, synergy activation during stimulations 200 ms prior to movement onset was significantly lower than stimulations 50 and 0 ms prior to movement onset (p = 0.001 and p < 0.001, respectively). Further, synergy activation during stimulations 150 ms prior to movement onset was significantly lower than stimulations 50 and 0 ms prior to movement onset (p = 0.001 and p < 0.001, respectively). Activation during stimulations 100 ms prior to movement onset was significantly lower than stimulations 50 and 0 ms prior to movement onset (p = 0.002 and p < 0.001, respectively). Synergy activation did not differ between stimulations delivered at 50 ms and 0 ms prior to movement onset (p = 0.239) or between stimulations delivered during rest or at 200, 150, or 100 ms prior to movement onset (all p > 0.05). In the Elbow Down posture, synergy activation during resting was significantly lower than activation following stimulations delivered 150, 100, 50, and 0 ms prior to movement onset (p = 0.022, p = 0.001, p < 0.001, and p < 0.001, respectively). Additionally, activation during stimulations 200 ms prior to movement onset was significantly lower than activation during stimulations 150, 100, 50, and 0 ms prior to movement onset (p = 0.036, p = 0.001, p < 0.001, and p < 0.001, respectively). Further, activation during stimulations 150 ms prior to movement onset was significantly lower than activation during stimulations 50 and 0 ms prior to movement onset (both p < 0.001). Activation during stimulations 100 ms prior to movement onset was significantly lower than activation during stimulations 50 and 0 ms prior to movement onset (both p < 0.001). Activation did not differ between stimulations delivered 50 and 0 ms prior to movement onset (p = 0.091). Post-hoc analysis comparing posture within blocks revealed synergy 1 activation was significantly higher in Elbow Up posture during resting and 200 ms prior to movement onset and significantly lower at 50 ms prior to movement onset (all p < 0.036). In synergy 2 activation profile, no main effects of timing (p = 0.101), posture (p = 0.629) or interaction effects (p = 0.842) were detected.

4 Discussion

The objective of this manuscript was to examine how motor preparation is expressed in the motor cortex prior to upper extremity reaching tasks, and how this expression is influenced by kinematic redundancy. To examine this, we asked participants to perform a reaching task in response to a visual cue. Participants started from one of two arm postures that required different muscle coordination to complete the reach. Using TMS, we measured corticospinal excitability at several different upper extremity muscle and at several different time points prior to movement onset. We then performed a novel, synergy-based MEP analysis to examine which MEP patterns emerged as TMS was delivered closer to movement onset. We hypothesized that MEPs would be facilitated in all upper extremity muscles relative to rest, and that the facilitation of each muscle would modulate with initial arm posture. We found that the excitability of a synergy reflecting the expected muscle coordination increased as the stimulation timing approached movement onset. Additionally, we found that the composition of this synergy was modulated by arm posture. These results generally support our hypotheses and suggest that changes in excitability prior to multi-joint, upper extremity movements reflects the muscle coordination observed after movement onset. These findings reveal valuable information regarding how the central nervous system prepares movements in the motor cortex.

A key finding of this study was that we found significant synergy differences between arm postures. It is also important to note that these differences matched the hypothesized differences in muscle coordination proposed in the Methods and Materials as well as the measured differences in muscle activation after movement onset between the two postures (Fig. 6). Previous studies have shown that MEPs are facilitated during movement preparation in muscles relevant to the movement (Ganguly et al., 2021; Lebon et al., 2019; Reuter et al., 2015; Sohn & Hallett, 2004a, 2004b). Our findings supplement these previous studies by showing that the magnitude of facilitation reflects the importance of the muscle in the subsequent movement (e.g., a movement requiring more middle deltoid activation will show great middle deltoid MEP facilitation during preparation). This is a finding unique to our study and arises from the inherent kinematic redundancy of the upper extremity. Because of the kinematic redundancy, it is possible to alter the contributions of muscles to complete the same functional task (e.g., reaching for a cup) in multiple ways. However, previous studies have only considered movements in one degree of freedom movements, which makes modulating the contribution of muscles much more difficult. Further testing with other postures and movements with other extremities is necessary to examine the full extent of how muscle coordination is reflected in the movement preparation phase.

Another interesting finding of this study was that the first synergy was generally characterized by weights (i.e., principal component coefficients/loadings) in most muscles, regardless of the initial posture of the upper extremity. This finding was somewhat surprising, as previous research has shown that MEPs during movement preparation are inhibited in irrelevant or antagonist muscles (Lebon et al., 2019; Reuter et al., 2015). It is likely that our findings differ from those of previous studies because the movements considered in this study include multiple degrees of freedom, while previous studies have only considered single-DOF movements. Additionally, it is more difficult to discern what muscles should be considered irrelevant or antagonistic. This is because many upper extremity muscles span multiple joints and/or stabilize motion via eccentric contractions. Therefore, these muscles may increase in excitability even though their contribution to the motion is not obvious. For example, the biceps brachii, which is generally considered to be an elbow flexor, assists in shoulder flexion and stabilizes the forearm during elbow extension when reaching from the Elbow Down position. This task requirement most likely caused the biceps brachii it had a large component in the first synergy (Fig. 7(A)). As a result, it is possible that the general increase in excitability observed in all proximal muscles results from all proximal muscles serving a purpose in the subsequent movement, with some more obvious than others.

Interestingly, we also found that the activation of the second synergy, which was characterized by larger weights in the distal muscles (WE and WF), was not facilitated by movement preparation. This second synergy emerged to account for residual variability in the WE and WF MEPs that could not be explained by the first, primary synergy (see Descriptive Statistics subsection for individual variance accounted for by each synergy). Because previous research has shown that corticospinal excitability increases as stimulations occur closer to movement onset (Chen et al., 1998; Ibanez et al., 2020; Rossini et al., 1988), we believe that the lack of activation in Synergy 2 suggests that a portion of variability in the distal muscle MEPs was unrelated to reaching. We believe that this occurred because the distal muscles play a relatively small role in the reach initiation. Specifically, both reaching movements in this study are characterized by initial shoulder and elbow movement to extend the arm, followed by a delayed opening of the fingers to grasp the cup. As such, the proximal muscles are more important to the facilitated motor plan at reach initiation, while the distal muscles are not recruited until the end of the movement. However, it is currently unclear what this lack of facilitation of the distal muscle synergy represents at the level of motor preparation in the cortex. For instance, it is possible that the lack of facilitation of MEP in the wrist muscles reflects a motor plan that, at the moment of measurement, is more focused on proximal muscles. Previous studies in serial motor tasks support this view, showing that when the cortex plans a sequence of tasks, all features of the sequence are developed in parallel prior to movement onset. The resulting movement corresponds to selective facilitation and inhibition of different features of the motor plan (Behmer & Crump, 2017; Behmer et al., 2023; Hurlstone et al., 2014). However, a competing theory is that preceding actions in a sequence act as cues for subsequent actions, triggering a new preparation and action phase (Behmer et al., 2023; Shiffrin & Cook, 1978). As such, the observed response in the distal upper extremity muscles could then represent surround inhibition (Sohn & Hallett, 2004b) or subject’s inattention to their distal muscles at this stage in the movement. Future research is needed to determine the underlying neural mechanisms causing this result.

We also found that the first synergy had a large biceps brachii (BB) weight in both postures, which in the Elbow Down posture was significantly greater than the triceps brachii (TB) weight. This observation is most likely due to BB both assisting shoulder flexion and stabilizing elbow extension against gravity. Interestingly, we also found that the weight of the BB was similar to that of the TB in the Elbow Up posture, a posture where there was no necessary BB contribution. We believe that this observation could be related to the emergence of the “flexor synergy”. Conventionally, the flexor synergy is defined as the abnormal coupling of shoulder abduction and elbow flexion following a stroke (Dewald et al., 1995; Twitchell, 1951). Previous studies have shown that this coupling interferes with arm movements requiring elbow extension with shoulder flexion (Zackowski et al., 2004) or shoulder abduction (Sukal et al., 2006). Here, we observed a similar phenomenon: increased excitability of the BB during shoulder abduction, despite the necessity for concurrent elbow extension. This may suggest that the flexor synergy is present in humans without neurological injury, although to a lesser extent. These findings shed light onto the possible origin of abnormal joint coupling in stroke survivors, which is currently unclear. Researchers have shown that abnormal joint couplings following stroke may have a cortical origin (Gerachshenko et al., 2008; Krishnan & Dhaher, 2012; Tan & Dhaher, 2017; Tan et al., 2016; Yao et al., 2009), while others have shown that couplings may come from increased reliance on either ipsilateral cortical projections (Schwerin et al., 2008) and/or the reticulospinal tract (Kuypers, 1964; McPherson et al., 2018; McPherson & Dewald, 2022). Those who argue for the latter claim that if the stroke causes damage to the corticospinal tract, the central nervous system will recruit these alternative motor control pathways to compensate for the lost capabilities. However, because these pathways are not evolved to control joints independently, they control them simultaneously, and thus, lead to the emergence of abnormal joint couplings (Krakauer & Carmichael, 2022). Neurologically intact adults, however, do not have damaged corticospinal tracts, and therefore have no need to recruit these alternative pathways. Therefore, the emergence of a pattern that reflects the flexor synergy in healthy adults supports the position that abnormal joint couplings after stroke have a cortical origin. Our approach proposed in this study could help to verify this premise in future studies. For example, if abnormal synergies are of cortical origin, then we would expect that stroke survivors would have exaggerated biceps synergy weights prior to movement onset.

The findings in this study, when considered in conjunction with evidence that abnormal synergies in stroke survivors have a cortical origin, have interesting implications for use in clinical populations. Specifically, TMS could be used as a tool to examine the contribution of the flexor synergy to impairment or even a metric to judge the effectiveness of a therapy intervention in treating the flexor synergy. However, eliciting MEPs is notoriously difficult in stroke survivors, who have higher motor thresholds, particularly in the proximal muscles, (Cicinelli et al., 1997; Schwerin et al., 2011) that relate to their level of motor function (Edwards et al., 2013). MEPs in stroke survivors can be produced more reliably using paired-pulse approaches (Schwerin et al., 2011), but many researchers still rely on the use of background contractions to increase MEP reliability (Krishnan & Dhaher, 2012). However, background contractions can confound examinations of abnormal muscle coordination because MEPs in the contracting muscle will be biased, and researchers must develop complex testing rigs to control for this bias (Krishnan & Dhaher, 2012; Schwerin et al., 2008). However, as this manuscript has shown, corticospinal excitability increases prior to the onset of multi-joint movements, and the excitability reflects the muscle coordination of the subsequent movement without being confounded by background contractions. Therefore, TMS prior to the onset of a multi-joint movement may be a suitable paradigm for facilitating MEPs in stroke survivors. However, it is unclear whether corticospinal excitability prior to movement onset is representative of corticospinal excitability during movement (Copithorne et al., 2015; Lockyer et al., 2021). Therefore, future testing with members of the stroke population is necessary to fully establish this paradigm.

One limitation of this study is that stimulations were delivered based on each participant’s computed average response time. Although this approach is similar to previous studies (Ibanez et al., 2020; Rossini et al., 1988), it has two potential drawbacks. First, slight intra-subject variability could lead to some stimulations occurring after EMG onset, which could potentially conflate detection of changes in corticospinal excitability because background EMG activity can cause larger MEPs. It is important to note, however, that these EMG contractions were only prevalent in the reaching block when stimulations occurred 0 ms prior to movement onset (Table 2). We also saw significant activation of the first synergy when stimulations occurred 50 ms prior to movement onset without a large increase in EMG activity prior to the stimulation in this block. Furthermore, this EMG activity prior to the stimulus was much lower than the activity observed after movement onset (Table 2 and Fig. 6), Indicating that a large portion of the stimulus happened during movement preparation. Therefore, we strongly believe that this synergy activation primarily reflected coordination during movement preparation. The second potential drawback of this approach is that movement onset was determined from one muscle and applied to all six muscles, therefore slight differences in EMG onset timings between muscles would mean that stimulations occurred at slightly different times relative to each muscle’s EMG onset. Unfortunately, there is no straightforward solution to this. For instance, stimulating according to each muscle’s EMG onset time is not feasible because it would require six times as many stimulations per participant. Likewise, retrospectively binning stimulations based on the latency between the MEP and EMG onset of each muscle is not possible because the MEP and subsequent inhibition (i.e., silent period) would often mask EMG onset. However, it is important to note that we selected movement onset time based on the fastest muscle so that the stimulations were occurring prior to EMG onset in all muscles. Therefore, we do not expect that an alternative method that corrected for slight differences in EMG onset between muscles would have yielded significantly different results.

In summary, we used transcranial magnetic stimulation to elicit MEPs during preparation of a multi-joint, upper extremity movement. We examined two different initial upper extremity postures that would produce differing muscle coordination during the movement and found that the differences in coordination were present in the MEPs even prior to the movement onset (i.e., during the motor preparation phase). We also found that MEPs were facilitated in all proximal muscles as stimulations were delivered closer to movement onset. Additionally, we found that MEPs expression reflected the flexor synergy, suggesting that the flexor synergy typically observed in stroke survivors may have a cortical origin. These results reveal interesting information regarding the mechanisms by which the motor cortex plans movements and may have utility for researchers examining muscle coordination in stroke survivors.

Footnotes

Acknowledgments

This work was partly supported by grants from the National Science Foundation (Grant #’s 1804053 and 1256260), the National Institutes of Health (Grant # R01-EB019834), and the NICHD National Center of Neuromodulation for Rehabilitation (Grant # P2C-HD086844)

Conflict of interests statement

None of the authors have any conflicts of interest.

The supplementary material is available in the electronic version of this article: .

References

Augenstein,

T. E.

, Washabaugh,

E. P.

, Remy,

C. D.

, Krishnan,

(2020) Motor Modules are Impacted by the Number of Reaching Directions Included in the Analysis, IEEE Trans Neural Syst Rehabil Eng 28(9), 2025–2034. https://doi.org/10.1109/TNSRE.2020.3008565.

Awiszus,

, Borckardt,

(2011). TMS motor threshold assessment Tool (MTAT 2.0). In ClinicalResearcher.org.

Behmer,

L. P.

, Crump,

M. J.

(2017) The dynamic range of response set activation during action sequencing, J Exp Psychol Hum Percept Perform 43(3), 537–554. https://doi.org/10.1037/xhp0000335.

Behmer,

L. P.

, Crump,

M. J. C.

, Jantzen,

K. J.

(2023) Motor-evoked potentials for early individual elements of an action sequence during planning reflect parallel activation processes, Motor Control. 1–20. https://doi.org/10.1123/mc.2022-0106.

Bestmann,

, Duque,

(2016) Transcranial magnetic stimulation: decomposing the processes underlying action preparation, Neuroscientist 22(4), 392–405. https://doi.org/10.1177/1073858415592594.

Bianco,

, Feurra,

, Fadiga,

, Rossi,

(2012) Bi-hemispheric effects on corticospinal excitability induced by repeated sessions of imagery versus observation of actions, Restor Neurol Neurosci 30(6), 481–489. https://doi.org/10.3233/RNN-2012-120241.

Chen,

, Hallett,

(1999) The time course of changes in motor cortex excitability associated with voluntary movement, Can J Neurol Sci 26(3), 163–169. https://doi.org/10.1017/s0317167100000196.

Chen,

, Yaseen,

, Cohen,

L. G.

, Hallett,

(1998) Time course of corticospinal excitability in reaction time and self-paced movements, Ann Neurol 44(3), 317–325. https://doi.org/10.1002/ana.410440306.

Cicinelli,

, Traversa,

, Rossini,

P. M.

(1997) Post-stroke reorganization of brain motor output to the hand: a 2-4 month follow-up with focal magnetic transcranial stimulation, Electroencephalography and Clinical Neurophysiology 105(6), 438–450. https://doi.org/10.1016/s0924-980x(97)00052-0.

10.

Cisek,

, Pastor-Bernier,

(2014) On the challenges and mechanisms of embodied decisions, Philosophical Transaction of the Royal Society B 369(1655). https://doi.org/10.1098/rstb.2013.0479.

11.

Copithorne,

D. B.

, Forman,

D. A.

, Power,

K. E.

(2015) Premovement changes in corticospinal excitability of the biceps brachii are not different between arm cycling and an intensity-matched tonic contraction, Motor Control 19(3), 223–241. https://doi.org/10.1123/mc.2014-0022.

12.

Davey,

N. J.

, Rawlinson,

S. R.

, Maskill,

D. W.

, Ellaway,

P. H.

(1998) Facilitation of a hand muscle response to stimulation of the motor cortex preceding a simple reaction task, Motor Control 2(3), 241–250. https://doi.org/10.1123/mcj.2.3.241.

13.

DeJong,

S. L.

, Bisson,

J. A.

, Darling,

W. G.

, Shields,

R. K.

(2021) Simultaneous recording of motor evoked potentials in hand, wrist and arm muscles to assess corticospinal divergence, Brain Topography 34(4), 415–429. https://doi.org/10.1007/s10548-021-00845-1.

14.

Dewald,

J. P.

, Pope,

P. S.

, Given,

J. D.

, Buchanan,

T. S.

, Rymer,

W. Z.

(1995) Abnormal muscle coactivation patterns during isometric torque generation at the elbow and shoulder in hemiparetic subjects, Brain 118(Pt 2), 495–510. https://doi.org/10.1093/brain/118.2.495.

15.

Edwards,

J. D.

, Meehan,

S. K.

, Linsdell,

M. A.

, Borich,

M. R.

, Anbarani,

, Jones,

P. W.

, Ferris,

, Boyd,

L. A.

(2013) Changes in thresholds for intracortical excitability in chronic stroke: more than just altered intracortical inhibition, Restorative Neurology and Neuroscience 31(6), 693–705. https://doi.org/10.3233/RNN-120300.

16.

Facchini,

, Muellbacher,

, Battaglia,

, Boroojerdi,

, Hallett,

(2002) Focal enhancement of motor cortex excitability during motor imagery: a transcranial magnetic stimulation study, Acta Neurol Scand 105(3), 146–151. https://doi.org/10.1034/j.1600-0404.2002.1o004.x.

17.

Ganguly,

, Kulshreshtha,

, Almotiri,

, Jog,

(2021) Muscle Tone Physiology and Abnormalities, Toxins (Basel) 13(4). https://doi.org/10.3390/toxins13040282.

18.

Gerachshenko,

, Rymer,

W. Z.

, Stinear,

J. W.

(2008) Abnormal corticomotor excitability assessed in biceps brachii preceding pronator contraction post-stroke, Clin Neurophysiol 119(3), 683–692. https://doi.org/10.1016/j.clinph.2007.11.004.

19.

Hurlstone,

M. J.

, Hitch,

G. J.

, Baddeley,

A. D.

(2014) Memory for serial order across domains: An overview of the literature and directions for future research, Psychol Bull 140(2), 339–373. https://doi.org/10.1037/a0034221.

20.

Ibanez,

, Hannah,

, Rocchi,

, Rothwell,

J. C.

(2020) Premovement suppression of corticospinal excitability may be a necessary part of movement preparation, Cereb Cortex 30(5), 2910–2923. https://doi.org/10.1093/cercor/bhz283.

21.

Krakauer,

J. W.

, Carmichael,

S. T.

(2022). Broken movement: the neurobiology of motor recovery after stroke. MIT Press.

22.

Krishnan,

, Dhaher,

(2012) Corticospinal responses of quadriceps are abnormally coupled with hip adductors in chronic stroke survivors, Exp Neurol 233(1), 400–407. https://doi.org/10.1016/j.expneurol.2011.11.007.

23.

Krishnan,

, Ranganathan,

, Dhaher,

Y. Y.

, Rymer,

W. Z.

(2013) A pilot study on the feasibility of robot-aided leg motor training to facilitate active participation, PLoS One 8(10), e77370. https://doi.org/10.1371/journal.pone.0077370.

24.

Kumru,

, Soto,

, Casanova,

, Valls-Sole,

(2008) Motor cortex excitability changes during imagery of simple reaction time, Exp Brain Res 189(3), 373–378. https://doi.org/10.1007/s00221-008-1433-6.

25.

Kuypers,

H. G.

(1964) The descending pathways to the spinal cord, their anatomy and function, Prog Brain Res 11, 178–202. https://doi.org/10.1016/s0079-6123(08)64048-0.

26.

Labruna,

, Lebon,

, Duque,

, Klein,

P. A.

, Cazares,

, Ivry,

R. B.

(2014) Generic inhibition of the selected movement and constrained inhibition of nonselected movements during response preparation, J Cogn Neurosci 26(2), 269–278. https://doi.org/10.1162/jocn_a_00492.

27.

Lebon,

, Ruffino,

, Greenhouse,

, Labruna,

, Ivry,

R. B.

, Papaxanthis,

(2019) The neural specificity of movement preparation during actual and imagined movements, Cereb Cortex 29(2), 689–700. https://doi.org/10.1093/cercor/bhx350.

28.

Leocani,

, Cohen,

L. G.

, Wassermann,

E. M.

, Ikoma,

, Hallett,

(2000) Human corticospinal excitability evaluated with transcranial magnetic stimulation during different reaction time paradigms, Brain 123(Pt 6), 1161–1173. https://doi.org/10.1093/brain/123.6.1161.

29.

Lockyer,

E. J.

, Compton,

C. T.

, Forman,

D. A.

, Pearcey,

G. E.

, Button,

D. C.

, Power,

K. E.

(2021) Moving forward: methodological considerations for assessing corticospinal excitability during rhythmic motor output in humans, J Neurophysiol 126(1), 181–194. https://doi.org/10.1152/jn.00027.2021.

30.

McPherson,

J. G.

, Chen,

, Ellis,

M. D.

, Yao,

, Heckman,

C. J.

, Dewald,

J. P. A.

(2018) Progressive recruitment of contralesional cortico-reticulospinal pathways drives motor impairment post stroke, J Physiol 596(7), 1211–1225. https://doi.org/10.1113/JP274968.

31.

McPherson,

L. M.

, Dewald,

J. P. A.

(2022) Abnormal synergies and associated reactions post-hemiparetic stroke reflect muscle activation patterns of brainstem motor pathways, Front Neurol 13, 934670. https://doi.org/10.3389/fneur.2022.934670.

32.

Melgari,

J. M.

, Pasqualetti,

, Pauri,

, Rossini,

P. M.

(2008) Muscles in “concert”: study of primary motor cortex upper limb functional topography, PLoS One 3(8), e3069. https://doi.org/10.1371/journal.pone.0003069.

33.

Raffin,

, Pellegrino,

, Di Lazzaro,

, Thielscher,

, Siebner,

H. R.

(2015) Bringing transcranial mapping into shape: Sulcus-aligned mapping captures motor somatotopy in human primary motor hand area, Neuroimage 120, 164–175. https://doi.org/10.1016/j.neuroimage.2015.07.024.

34.

Ranganathan,

, Krishnan,

(2012) Extracting synergies in gait: using EMG variability to evaluate control strategies, J Neurophysiol 108(5), 1537–1544. https://doi.org/10.1152/jn.01112.2011.

35.

Reuter,

E. M.

, Behrens,

, Zschorlich,

V. R.

(2015) Age-related differences in corticomotor facilitation indicate dedifferentiation in motor planning, Exp Gerontol 65, 79–84. https://doi.org/10.1016/j.exger.2015.03.008.

36.

Rodseth,

, Washabaugh,

E. P.

, Krishnan,

(2017) A novel low-cost approach for navigated transcranial magnetic stimulation, Restor Neurol Neurosci 35(6), 601–609. https://doi.org/10.3233/RNN-170751.

37.

Roh,

, Rymer,

W. Z.

, Beer,

R. F.

(2012) Robustness of muscle synergies underlying three-dimensional force generation at the hand in healthy humans, J Neurophysiol 107(8), 2123–2142. https://doi.org/10.1152/jn.00173.2011.

38.

Rossini,

P. M.

, Zarola,

, Stalberg,

, Caramia,

(1988) Pre-movement facilitation of motor-evoked potentials in man during transcranial stimulation of the central motor pathways, Brain Res 458(1), 20–30. https://doi.org/10.1016/0006-8993(88)90491-x.

39.

Schwerin,

, Dewald,

J. P.

, Haztl,

, Jovanovich,

, Nickeas,

, MacKinnon,

(2008) Ipsilateral versus contralateral cortical motor projections to a shoulder adductor in chronic hemiparetic stroke: implications for the expression of arm synergies, Exp Brain Res 185(3), 509–519. https://doi.org/10.1007/s00221-007-1169-8.

40.

Schwerin,

S. C.

, Yao,

, Dewald,

J. P.

(2011) Using paired pulse TMS to facilitate contralateral and ipsilateral MEPs in upper extremity muscles of chronic hemiparetic stroke patients, J Neurosci Methods 195(2), 151–160. https://doi.org/10.1016/j.jneumeth.2010.11.021.

41.

Sherrington,

C. S.

(1910) Flexion-reflex of the limb, crossed extension-reflex, and reflex stepping and standing, J Physiol 40(1-2), 28–121. https://doi.org/10.1113/jphysiol.1910.sp001362.

42.

Shiffrin,

R. M.

, Cook,

J. R.

(1978) Short-term forgetting of item and order information, Journal of Verbal Learning and Verbal Behavior 17(2), 189–218.

43.

Sohn,

Y. H.

, Hallett,

(2004a) Disturbed surround inhibition in focal hand dystonia, Ann Neurol 56(4), 595–599. https://doi.org/10.1002/ana.20270.

44.

Sohn,

Y. H.

, Hallett,

(2004b) Surround inhibition in human motor system, Exp Brain Res 158(4), 397–404. https://doi.org/10.1007/s00221-004-1909-y.

45.

Sukal,

T. M.

, Ellis,

M. D.

, Dewald,

J. P.

(2006) Source of work area reduction following hemiparetic stroke and preliminary intervention using the ACT3D system, Conf Proc IEEE Eng Med Biol Soc 177–180. https://doi.org/10.1109/IEMBS.2006.259311.

46.

Tan,

A. Q.

, Dhaher,

Y. Y.

(2017) Contralesional Hemisphere Regulation of Transcranial Magnetic Stimulation-Induced Kinetic Coupling in the Poststroke Lower Limb, Front Neurol 8, 373. https://doi.org/10.3389/fneur.2017.00373.

47.

Tan,

A. Q.

, Shemmell,

, Dhaher,

Y. Y.

(2016) Downregulating aberrant motor evoked potential synergies of the lower extremity post stroke during TMS of the contralesional hemisphere, Brain Stimul 9(3), 396–405. https://doi.org/10.1016/j.brs.2015.12.006.

48.

Ting,

L. H.

(2007) Dimensional reduction in sensorimotor systems: a framework for understanding muscle coordination of posture, Prog Brain Res 165, 299–321. https://doi.org/10.1016/S0079-6123(06)65019-X.

49.

Ting,

L. H.

, McKay,

J. L.

(2007) Neuromechanics of muscle synergies for posture and movement, Curr Opin Neurobiol 17(6), 622–628. https://doi.org/10.1016/j.conb.2008.01.002.

50.

Tresch,

M. C.

, Jarc,

(2009) The case for and against muscle synergies, Curr Opin Neurobiol 19(6), 601–607. https://doi.org/10.1016/j.conb.2009.09.002.

51.

Twitchell,

T. E.

(1951) The restoration of motor function following hemiplegia in man, Brain 74(4), 443–480. https://doi.org/10.1093/brain/74.4.443.

52.

Washabaugh,

, Guo,

, Chang,

C. K.

, Remy,

, Krishnan,

(2019) A portable passive rehabilitation robot for upper-extremity functional resistance training, IEEE Trans Biomed Eng 66(2), 496–508. https://doi.org/10.1109/TBME.2018.2849580.

53.

Washabaugh,

E. P.

, Claflin,

E. S.

, Gillespie,

R. B.

, Krishnan,

(2016) A novel application of eddy current braking for functional strength training during gait, Ann Biomed Eng 44(9), 2760–2773. https://doi.org/10.1007/s10439-016-1553-2.

54.

Wilson,

S. A.

, Thickbroom,

G. W.

, Mastaglia,

F. L.

(1993) Transcranial magnetic stimulation mapping of the motor cortex in normal subjects. The representation of two intrinsic hand muscles, Journal of the Neurological Sciences 118(2), 134–144. https://doi.org/10.1016/0022-510x(93)90102-5.

55.

Yao,

, Chen,

, Carmona,

, Dewald,

J. P.

(2009) Cortical overlap of joint representations contributes to the loss of independent joint control following stroke, Neuroimage 45(2), 490–499. https://doi.org/10.1016/j.neuroimage.2008.12.002.

56.

Zackowski,

K. M.

, Dromerick,

A. W.

, Sahrmann,

S. A.

, Thach,

W. T.

, Bastian,

A. J.

(2004) How do strength, sensation, spasticity and joint individuation relate to the reaching deficits of people with chronic hemiparesis? Brain 127(Pt 5), 1035–1046. https://doi.org/10.1093/brain/awh116.

57.

Zschorlich,

V. R.

, Kohling,

(2013) How thoughts give rise to action –conscious motor intention increases the excitability of target-specific motor circuits, PLoS One 8(12), e83845. https://doi.org/10.1371/journal.pone.0083845.