Concurrent contractions of antagonistic muscles about a joint (cocontraction) are involved in various daily activities. People tend to cocontract muscles when performing an unfamiliar motor task, standing on an unstable surface (e.g., surfing, train), holding a steering wheel on rough terrains, and reaching for or holding an object with accuracy and steadiness. In individuals with various movement disorders such as cerebral palsy and Parkinson's disease, the cocontraction of antagonist muscles is often different from that of healthy individuals in terms of timing and magnitude (11,21). In sports and other physical activities, the ability to control accurate and steady cocontraction would influence motor performance that requires joint stabilization such as gymnastics, tai chi, yoga, sumo wrestling, archery, shooting, biathlon, and car racing. In producing comparable magnitudes of torque in opposite directions for stabilizing a joint with cocontraction, the activation level can be different (i.e., unbalanced) between antagonistic muscles (22) because the capability for force generation can be nonuniform across muscles because of variable muscle architecture and moment arm (14). For identifying potential people with compromised ability for steady cocontraction and developing effective training or rehabilitation strategies for improvement, it is important to understand fundamental neural characteristics that are related to the ability of individuals for controlling unbalanced cocontraction of antagonistic muscles as steady as possible.
When controlling a certain level of motor output as steady as possible using agonistic muscle(s), steadiness in motor output is primarily influenced by low-frequency oscillations (<5 Hz) of discharges of motor units (15,20). Low-frequency neural oscillations <5 Hz are associated with the generation of muscle force (2,8,9). In this frequency range, correlated modulation of motor unit discharges is observed within and across muscles, and they are believed to be produced by a “common drive” (3). In steady contractions, the correlated oscillations are in-phase across motor units and muscles, including antagonistic muscles (5). The amount of low-frequency common input to motor neurons is estimated to be 60%–80% of the total synaptic input (16); hence, the independent synaptic input would constitute only the remaining 20%–40%. In controlling antagonistic muscles independently to perform an unbalanced cocontraction, a greater amount of low-frequency common input can be unfavorable because it would leave a smaller amount of independent input. Although various factors may influence the performance of steady cocontraction (e.g., familiarity to the task, excitability of corticospinal neurons, and Ia presynaptic inhibition) (12,19), it is presumable that individuals who have greater magnitudes of low-frequency common input have fundamental difficulties in steady unbalanced cocontraction. Hence, we hypothesized that the capability of individuals for maintaining steady unbalanced cocontraction is related to the amount of low-frequency correlated neural oscillations between muscles. Accordingly, the first purpose of the study was to determine whether there is an underlying association between the low-frequency correlated neural oscillations between muscles and the performance of steady unbalanced cocontraction when their variability across healthy young adults is analyzed.
The performance of steady cocontraction is known to be improved by a bout of motor practice. In antagonistic muscles about the ankle joint, a bout of steady cocontraction practice resulted in a reduction in the variability of motor output in healthy young adults (12,19). It is unknown whether an adaptation in low-frequency correlated neural oscillations is involved in practice-induced improvement in steady cocontraction performance. Assuming the previously mentioned relationship exists between the low-frequency correlated neural oscillations and steady cocontraction performance across individuals as a cause and effect relationship, an acute reduction in the low-frequency common neural oscillations within individuals may lead to an improvement in steady cocontraction performance. During a force-varying task involving antagonistic muscles, correlated neural oscillations in motor units between antagonistic muscles were observed out of phase as well as in phase (5). It is possible that a practice that enforces the out-of-phase drive to antagonistic muscles may induce adaptation that counteracts the in-phase common drive and improve steady cocontraction performance. We thus hypothesized that out-of-phase cocontraction practice acutely causes an adaptation to reduce the net in-phase correlated neural oscillations and improve steady cocontraction performance. Therefore, the second purpose of the study was to determine whether a bout of out-of-phase cocontraction practice reduces in-phase low-frequency correlated neural oscillations and improves the performance of steady unbalanced cocontraction of antagonistic muscles in healthy young adults.
For these purposes, surface EMG of arm muscles was used to provide visual feedback and the assessment of steady performance and correlated neural oscillations during steady unbalanced cocontraction about the elbow joint. The low-frequency band of rectified EMG was used for assessing low-frequency oscillations of pools of motor units according to previous studies (15,22,23).
Sixty healthy young adults (22.5 ± 3.0 yr old, 30 men and 30 women) without any history of neurological disorder participated in the study. Subjects were divided into three groups that received different intervention: cocontraction, contraction, and control. All subjects were right-handed, as confirmed with the Edinburgh Handedness Inventory (18). They all gave their written informed consent in accordance with the Institutional Review Board at Georgia Institute of Technology.
Subjects were seated on a chair with a backrest. The upper body of the subjects was upright and attached to the backrest using seat belts. The right shoulder was flexed to 20° from the anatomical position, i.e., the upper arm was placed forward from the trunk by 20°. The elbow was rested on a table, and the forearm was constrained to a padded attachment at the wrist in the neutral position, using a Velcro strap. With their fingers fully extended, the elbow joint was kept at 80° from the anatomical position during the tasks. Surface EMG signal was recorded from two elbow flexor muscles, biceps brachii (BB) and brachioradialis (BR), and one elbow extensor muscle, triceps brachii (TB). EMG of the BB–TB pair was used as the antagonistic muscle pair for the cocontraction tasks. Because EMG from this muscle pair could be influenced by potential cross talk because of their anatomical proximity, EMG recording of BR allowed for an additional antagonistic muscle pair (BR–TB) for analysis, which is less susceptible to cross talk. Before bipolar electrodes were attached to these muscles, the skin surface was prepared by shaving the hair, gently exfoliating the skin, and cleaning with alcohol. The electrodes were equipped with a differential amplifier (×300) and band-pass filter (15–2000 Hz) with the interelectrode distance of 18 mm (Z03 EMG; Motion Lab Systems, Baton Rouge, LA). EMG signals were acquired at 1000 samples per second using an analog-to-digital converter (NI USB-6216; National Instruments, Austin, TX) and Matlab (Mathworks, Natick, MA).
All participants performed a target-matching test using smoothed EMG amplitude (AEMG) of BB and TB before and after a bout of intervention. For normalizing AEMG, maximal voluntary contraction (MVC) was performed at the beginning of the experiment. Each subject then performed the steady cocontraction test before and after an intervention to assess cocontraction performance and neural oscillations. In addition, steady contractions were performed before and after the intervention to assess potential neuromuscular fatigue.
MVC was performed to obtain maximal AEMG of BB, BR, and TB in all subjects. The tasks were conducted by isometrically flexing and extending the elbow joint separately for the elbow flexors (BB and BR) and the elbow extensor (TB), respectively. MVC was performed as the maximal contraction of each muscle group independently, and not concurrently. The task consisted of a gradual increase in activation from zero to maximum for 3 s with the maximum held for 2–3 s. Subjects were verbally encouraged to achieve maximal activation while they observed AEMG of each muscle displayed on a monitor (10). At least three trials were performed, and the peak AEMG across the trials was obtained. When the difference in the peak AEMG of two highest MVC was >5%, additional trials were performed until the difference <5% was achieved. The maximal peak AEMG across trials was determined as maximal AEMG in each muscle. The same experimental setup and measurements were used across tasks.
Steady cocontraction test
All subjects were tested on their ability to control steady cocontraction of BB and TB by matching their AEMG to two pairs of target templates before and after an intervention. In all templates, the baseline was a resting level for both muscles, followed by different levels of AEMG between the muscles (Fig. 1). In one pair, the target template for BB started with 4% MVC for 3.5 s followed by 12% MVC (HIGH target) for 24.5 s, whereas the template for TB started with 12% MVC for 3.5 s followed by 4% MVC (LOW target) for 24.5 s (Fig. 1A). This target pair was termed as TB-LOW/BB-HIGH target. In another pair of templates, the roles of BB and TB were swapped, and termed as BB-LOW/TB-HIGH target (Fig. 1B).
The test was composed of six trials for each pair of templates. In each trial, one pair of target templates appeared concurrently on the screen. Approximately 10 s before the templates showed up, a visual alarm was displayed. Then approximately 3–5 s before the templates showed up, an auditory alarm signaled to be ready. Subjects were instructed to “reach and match both targets as fast, accurate, and steady as possible.” Before starting the test, they were presented with two trials for familiarization of the task. The order of the pair of templates was pseudorandomized, and there was a 32-s rest in between trials. After the completion of the test, subjects were shown the traces of individual trials and grand average traces of all the trials to provide knowledge of results.
After the completion of the initial steady cocontraction test, subjects received different interventions depending on the group. The duration of the interventions was approximately 52 min. In the control group, subjects did not perform practice but rested. In the cocontraction group, subjects performed practice of out-of-phase cocontraction of BB and TB concurrently. In the contraction group, subjects performed practice of repeatedly adjusting contraction levels of BB or TB independently. In both cocontraction and contraction groups, there were 32 practice trials, with a 32-s rest in between. Subjects were presented with two trials for familiarization of task before starting the practice. Visual and auditory alarms for start of trials were similar to those of the steady cocontraction test. After every eight trials, a resting interval of 5 min was provided. During each break, traces of individual trials and their average were shown to subjects to provide knowledge of results.
In the cocontraction group, both flexor and extensor muscles contracted concurrently. The following protocol was designed so subjects voluntarily produce out-of-phase low-frequency correlated oscillations between BB and TB repeatedly. A pair of target sequences for one 28-s trial consisted of repetitions of out-of-phase cocontraction: concurrently increasing AEMG to 12% MVC in one muscle and decreasing AEMG to 4% MVC in another muscle, which was alternated every 3.5 s (sequences A and B combined in Fig. 2). Starting at the time stamp of 3.5 s, each trial had seven epochs for out-of-phase activation: one muscle increases its activation level (“I” in Fig. 2), whereas another muscle decreases its activation level (“D” in Fig. 2) concurrently. Subjects were instructed to “reach and match both targets as fast, accurate, and steady as possible.” This protocol and target were created based on our pilot study for subjects to be able to repeat out-of-phase activations of BB and TB to alternating targets without confusion. The total number of the epochs for out-of-phase activation was 224 (7 epochs × 32 trials) across practice. The order of assigning the target sequence A or B (Fig. 2) to BB or TB was pseudorandomized.
In the contraction group, either the flexor or the extensor muscles contracted at a time. The following protocol was designed so subjects voluntarily used only one muscle (BB or TB) for a similar template as in the cocontraction group. The target sequence for one trial consisted of repetitions of alternating AEMG of one muscle between 4% MVC and 12% MVC, 3.5 s for each, for a total of 28 s. The sequence started with 4% MVC in one half of the trials (sequence A, Fig. 2) and 12% MVC and in another half of the trials (sequence B, Fig. 2) in each muscle. Subjects were shown one of the sequences at a time and instructed to “reach and match the target as fast, accurate, and steady as possible.” The order of muscle and sequence was pseudorandomized.
To obtain EMG signals for determining the median frequency as an indirect measure of neuromuscular fatigue, steady contractions were performed immediately before the steady cocontraction test, immediately before the intervention, and after the steady cocontraction retest postintervention. In each trial, subjects were asked to contract either the elbow flexor or the extensor muscles independently to match the AEMG of BB or TB, respectively, to the target of 12% MVC for 28 s. This trial was repeated three times for each muscle.
Amplitude of EMG signal was computed through a series of steps. Large spikes (>3.8 SD) in the raw EMG signal likely due to movement artifacts were first removed, if present, and replaced by neighboring samples. The mean value of raw EMG signal across each trial was computed and subtracted from the signal of the corresponding trial. The resulting signal was then full-wave rectified and smoothed using a moving average window of 125 samples (125 ms). Resting background noise at the beginning of the experiment was subtracted. The signal was then normalized by the maximal value during MVC for each muscle (AEMG). Thus, processed AEMG data of BB and TB were used for providing visual feedback to the subjects. To assess the variability in maintaining steady cocontraction, the variance of AEMG was calculated across the last 21 s (from time 7 s to time 28 s) before going back to the baseline. To assess the accuracy in matching the cocontraction level about the steady target (i.e., slow deviations from the target), the mean squared error between AEMG and target was calculated for the same period after applying a 1-s moving average. These performance variables were determined on BB and TB only but not on BR because there was no instruction or visual feedback on BR for the motor task.
For assessing the oscillatory characteristics of EMG signals, the following processing was performed. The signal was first normalized by the maximal value during MVC for each muscle. In each trial, EMG signal from the onset of warning (time −4 s) to the end (time 32 s) was used. An eighth-order Butterworth high-pass filter of 15 Hz cutoff was applied, using a zero-phase forward and reverse digital IIR filter. After removing the mean value, the signal was full-wave rectified. Resting background noise was subtracted. The signal for the last 21 s of the constant target (from time 7 s to time 28 s) was extracted, and the mean value was subtracted. The signals for all six trials were concatenated together to form a 126-s long segment. To assess the power content of oscillations, estimates of the event-related spectral perturbation (ERSP) (6) or shifts in the power spectrum of each muscle in time (t) and frequency (f) were derived using a Hanning window with size of 2048 samples (2.048 s) with 512 samples (0.512 s) overlap for n trials using the Fourier (F) transform (equation 1).
For phase and amplitude coherence between muscles, the event-related phase coherence (ERPCOH) and the event-related linear coherence (ERLCOH) (6) between each pair (a and b) of muscles (BB–TB, BR–TB, and BB–BR) were estimated using the window size of 2048 samples (2.048 s) with 512 samples (0.512 s) overlap (equations 2 and 3). Coherence between the agonist BB–BR pair was included in the analysis because the low-frequency correlated oscillations (“common drive”) are common to contracting muscles whether they are agonists or antagonists (5):
Relatively high coherence amplitude was observed less than 3 Hz, so we decided to focus on the 0- to 3-Hz range. For spectral power (EMG power), amplitude coherence (EMG coherence), and phase coherence (EMG phase), the mean value in the 0- to 3-Hz range was computed across the 21 s. During the steady cocontraction test, EMG phase with positive or negative sign was used to understand the deviation in reference to 0° (i.e., complete in phase) with a direction.
In the cocontraction group, the practice protocol was designed so subjects voluntarily produce out-of-phase low-frequency correlated EMG oscillations between muscles. To examine the characteristics of EMG oscillations during the intervention in this group, EMG coherence and EMG phase between muscle pairs were calculated in a similar manner as above except for the extraction of the samples. The last 21 s of each trial (after discarding first 7 s) are divided into 3.5-s windows corresponding to decreasing (D) or increasing (I) activation level referenced to one of the muscles (see Fig. 2). In each muscle, a total of three D and three I segments were extracted for each trial. For each segment, amplitude and phase coherence were computed. To obtain a variable that reflects the deviation of EMG phase in reference to 180° phase (i.e., complete anti-phase) with a direction, the following calculation was made in EMG phase between antagonistic muscle pairs (i.e., BB–TB pair and BR–TB pair) during the practice. If EMG phase was positive, the deviation of EMG phase from 180° was determined as positive deviation from 180°. If EMG phase was negative, the negative sign was attached to its absolute deviation from 180°. These deviations of EMG phase were compared with 90° for the former and with −90° for the latter in the statistical analysis. In the agonistic BB–BR pair, the EMG coherence value was used as it is to represent the deviation of EMG phase in reference to 0° phase (i.e., complete in phase).
To obtain a variable that may help infer the potential neuromuscular fatigue, the raw EMG signals during steady contractions were analyzed. Using the last 21 s of each trial, median power frequency was computed using power spectral density estimate via periodogram method.
Dependent variables for accuracy and variability of AEMG in the steady cocontraction test were mean squared error and variance of AEMG. Dependent variables for oscillatory characteristics of EMG in the steady cocontraction test included EMG power for each muscle and EMG coherence and EMG phase for each pair of muscles. These variables were tested with a three-way ANOVA with factors being time (before and after intervention period), muscle (BB and TB, with BR added only for power), and group (control, contraction, and cocontraction) with repeated measures. EMG coherence and EMG phase were tested with a three-way ANOVA with factors being time, muscle pair (BB–TB, BR–TB, and BB–BR), and group with repeated measures. EMG phase during the steady cocontraction was compared with 0° using Student’s t-test. The deviation of EMG phase during the intervention in the cocontraction group was compared with 90°, −90°, or 0° using Student’s t-test, depending on the muscle pair (see Data Analysis section). EMG coherence during the intervention in the cocontraction group was tested with a two-way ANOVA with factors being muscle pair and segment (D and I) with repeated measures. For assessing neuromuscular fatigue, median power frequency of raw EMG during steady contractions was tested with a three-way ANOVA with factors being time, muscle, and group with repeated measures. When appropriate, post hoc comparisons were performed using Tukey’s test. Linear regression analysis was performed between each of the oscillation-related variables and the performance-related variables across all subjects for each muscle and muscle pair before and after the intervention period. Pearson product–moment correlation coefficient (r) was obtained for these correlations. An alpha level of 0.05 was chosen for determining statistical significance. P < 0.05 and P < 0.01 are noted when significant.
Correlation between oscillations and performance
To address the first purpose, correlation coefficients between EMG coherence and performance variables across subjects are summarized in Figure 3 for various muscle pairs at each target. In the antagonistic pairs (i.e., BB–TB and BR–TB pairs, top 2 panels in Fig. 3), EMG coherence had no significant positive correlation with either performance variable before the intervention period. After the intervention period, correlation coefficients between EMG coherence and performance variables across subjects changed to more positive values in all cases in both of the antagonistic muscle pairs. As a result, significant positive correlations emerged in many of these correlations (10 of 16 cases, 63% frequency). After the intervention period, significant correlations were more prevalent with the BB-LOW/TB-HIGH target, in which performance variables had significant positive correlations with EMG coherence of the BB–TB pair (four of four cases) and BR–TB pair (three of four cases). For the other BB-HIGH/TB-LOW target, performance variables had significant positive correlations with EMG coherence of the BR–TB pair in three of four cases, but not of the BB–TB pair, after the intervention period.
In the agonistic BB–BR pair, EMG coherence already had a significant positive correlation before the intervention period with mean squared error of AEMG in all cases (four of four cases). After the intervention period, correlation coefficients between EMG coherence and mean squared error remained significant except for one case (but just below the level of significance) in this agonistic muscle pair. For the variance of AEMG, the correlation with BB–BR coherence was significant only in one case before the intervention period (one of four cases). It then became significant in all cases after the intervention period (four of four cases). Overall, significant correlations between BB–BR coherence and performance variables were found in five of eight cases before the intervention period and in all eight cases (100% frequency) after the intervention period.
Collectively, significant positive correlations between EMG coherence and performance variables were found only in 6 of 24 cases (25% frequency) before the intervention period when all muscle pairs and targets were collapsed. After the intervention period, significant positive correlations were found in 17 of 24 cases (71% frequency): 7 of 12 cases (58% frequency) for the TB-LOW/BB-HIGH target and 10 of 12 cases (83% frequency) for the BB-LOW/TB-HIGH target.
EMG power was not significantly correlated with performance variables in all but one case. Variances of AEMG and EMG power in TB were significantly correlated for the BB-LOW/TB-HIGH target after the intervention period (r = 0.26, P < 0.05).
The second aim was to examine if out-of-phase cocontraction practice acutely reduces the correlated neural oscillations and improves steady cocontraction performance. To determine whether correlated oscillations were changed differently depending on the type of intervention, the statistical significance of the time–group interaction was looked for in EMG coherence, EMG phase, and EMG power <3 Hz.
For coherence of EMG, there was a main effect of time (P < 0.01), showing reduced EMG coherence by 6% after the intervention period (Table 1). For other main effects of group (P < 0.01) and muscle pairs (P < 0.01), EMG coherence was greater in contraction (0.445 ± 0.068) compared with cocontraction (0.431 ± 0.059, P < 0.01) by 3% and control (0.427 ± 0.052, P < 0.01) by 4%, greater in BB–BR (0.452 ± 0.062) compared with BB–TB (0.431 ± 0.066, P < 0.01) by 5% and BR–TB (0.419 ± 0.047, P < 0.01) by 8%. There was no significant interaction that contained time–group interaction.
For phase of EMG, there was a main effect of muscle pair (P < 0.01). In reference to TB, the positive EMG phases of the BB–TB pair (2.04° ± 8.40°) and the BR–TB pair (2.25° ± 8.66°) were different from 0° (P < 0.01 for both pairs), indicating the flexor BB and BR led the extensor TB by approximately 2° on average. In reference to BB, EMG phase of the agonist BB–BR pair (−0.26° ± 8.75°) was different from both antagonistic pairs (P < 0.01 for both) but not significantly different from 0° (P > 0.05). There was no significant interaction that contained time–group interaction.
For low-frequency EMG power, EMG power was smaller by 14% after the intervention period with the main effect of time (P < 0.01) (Table 1). With other main effects of muscle (P < 0.01), EMG power in BB (20.04% ± 8.85% MVC2) was greater compared with TB (17.31% ± 6.53% MVC2, P < 0.01) by 16% and BR (14.92% ± 10.19% MVC2, P < 0.01) by 34% with significant difference between TB and BR (P < 0.01). There was no significant interaction that contained time–group interaction.
In the assessment of accuracy, there was a main effect of time (P < 0.01), showing reduced mean squared error of AEMG by 33% after the intervention period (Table 1). For other significant main effects of muscle (P < 0.01), mean squared error of AEMG was 40% greater in TB (7.60% ± 10.05% MVC) compared with BB (5.41% ± 6.91% MVC). There was no significant interaction on mean squared error of AEMG that contained time–group interaction.
In the assessment of variability, there was a main effect of time (P < 0.01), showing a decreased variance of AEMG by 30% (3.50% ± 4.25% vs 2.46% ± 2.96% MVC2) after the intervention period. For other main effects of group (P < 0.05) and muscle (P < 0.01), the variance of AEMG was lower in cocontraction (2.33% ± 2.18% MVC2) compared with contraction (3.34% ± 4.47% MVC2, P < 0.05) by 30% and control (3.26% ± 3.97% MVC2, P < 0.05) by 29%, 32% greater in TB (3.39% ± 3.98% MVC2) compared with BB (2.57% ± 3.34% MVC2). There was no significant interaction on variability of AEMG that contained time–group interaction.
Correlated oscillations during cocontraction practice
In the cocontraction group, subjects performed out-of-phase activation of antagonist muscles (Fig. 2). EMG coherence and EMG phase <3 Hz during the intervention were determined in each target sequence assignment, depending on BB targeting the D or I segments in Figure 2. There was an interaction of muscle pair and segment on EMG coherence (P < 0.01). EMG coherence between the BB–TB pair was 0.426 ± 0.048 for D segments and 0.458 ± 0.057 for I segments. For the BR–TB pair, EMG coherence was 0.440 ± 0.055 for D segments and 0.467 ± 0.057 for I segments. For the agonistic BB–BR pair, EMG coherence was 0.561 ± 0.067 for D segments and 0.545 ± 0.063 for I segments.
To examine if out-of-phase EMG oscillations were produced between antagonistic muscles, the deviation of EMG phase from 180° (i.e., complete anti-phase) was determined for the antagonistic pairs. In reference to TB, the deviation of EMG phase in BB–TB pair was 20.8° ± 42.0° for D segments for BB (i.e., I segments for TB) (P < 0.01 compared with 90°) and −38.0° ± 42.4° for I segments for BB (P < 0.01 compared with −90°). Similarly, the deviation of EMG phase in the BR–TB pair was 22.7° ± 45.1° for D segments for BR (P < 0.01 compared with 90°) and −34.5° ± 46.7° for I segments for BR (P < 0.01 compared with −90°). The positive and negative values indicate that BB and BR followed and preceded TB, respectively. These deviations from 180° demonstrate that the average EMG phase between the antagonistic muscles was 159.2°, −142.0°, 157.3°, and −147.5°. In addition, the EMG phase between agonistic muscles was determined to examine if EMG oscillated in phase. In reference to BB, EMG phases in the agonistic BB–BR pair were different from 0° but only slight deviations for both segments (P < 0.01): 0.047° ± 0.23° for D segments for BB and 0.067° ± 0.22° for I segments for BB.
Potential neuromuscular fatigue
The median power frequency of raw EMG during the steady contraction was compared as an indirect measure of neuromuscular fatigue. There was only a significant main effect of muscle (85.6 ± 14.2 Hz in TB vs 78.9 ± 12.2 Hz in BB, P < 0.01), but not time or group. When collapsed across groups and muscles, median power frequency of raw EMG was 83.1 ± 13.5 Hz before the steady cocontraction test, 82.3 ± 14.3 Hz immediately before the intervention, and 81.3 ± 13.3 Hz after the steady cocontraction retest postintervention (P > 0.05). There were no significant interactions.
The study aimed 1) to determine an association between low-frequency correlated EMG oscillations and performance across individuals and 2) to determine whether a bout of out-of-phase cocontraction practice reduces the in-phase correlated oscillations and improves performance in steady unbalanced cocontraction. The major findings are as follows: 1) there were positive correlations between low-frequency EMG coherence and performance variables (i.e., mean squared error of AEMG and variance of AEMG) across subjects, which became more prevalent after the intervention period, and 2) there were marginal reductions in low-frequency EMG coherence and large improvements in performance variables after the intervention period, but the type of intervention did not influence the reductions in these variables.
The presence of significant positive correlations between EMG coherence and both mean squared error and variance of AEMG in the majority of cases after the intervention period indicates that accuracy and steadiness of steady cocontraction tend to be degraded in individuals who have greater low-frequency correlated oscillations between muscles. The magnitude of low-frequency neural oscillations are suggested to be one of the major contributors to steady performance in both simulation (7,20) and experimental studies (13,15,23) on contractions primarily within agonistic muscles. In the current study, the absence of significant positive correlations between EMG power in individual muscles and performance variables in most cases indicates that the amount of independent neural oscillations in each muscle is not associated with steady performance when the steady task is antagonistic cocontraction. This is a new set of intriguing findings suggesting that people who tend to perform steady cocontraction less skillfully are not those who have greater low-frequency neural oscillations in each muscle but greater correlated oscillations between muscles. Hence, the results support the first hypothesis that capability for steady cocontraction performance is related to the amount of low-frequency correlated neural oscillations.
The currently used steady test was unique not only because it was a cocontraction task but also because it used unbalanced activation levels between antagonistic muscles. Maintaining a higher activation level in one muscle and a lower activation level in an antagonistic muscle requires the motor command to concurrently excite both muscles while partially inhibiting one of the antagonistic muscles as steady as possible. Concurrent excitation of multiple muscles involves low-frequency common oscillations in motor unit discharges, called “common drive” of central origin (4). As an indirect measure of common oscillations of pools of motor units, low-frequency correlated oscillations in rectified EMG have negligible time lags (<50 ms) between elbow antagonistic muscles (22). Very slight deviations of EMG phase from zero during steady cocontractions confirm that correlated oscillations in the current study were also in phase practically. In the presence of this in-phase common drive, the unbalanced cocontraction requires partial inhibition for adjusting to a lower target level with one of the muscles. Because spinal Ia reciprocal inhibition is strongly depressed during cocontraction (12,17), this inhibition is likely to involve central mechanisms, including reciprocally organized anti-phase drive between antagonistic muscles (5). It is unknown whether the decreased corticospinal excitability with improved cocontraction performance (19) is related to this implied inhibition. Observed amounts of in-phase low-frequency correlated oscillations are the net results of excitatory and inhibitory input to the motor neurons. Collectively, it is possible that individuals who have lower net in-phase common drive perform better in steady cocontraction by producing less in-phase common drive or by using central inhibition (or both). Although these possibilities are yet to be determined, lower in-phase correlated oscillations indicate less coupled activity (i.e., more decoupled activity) between muscles, which would help independent control of unbalanced activity between antagonistic muscles.
The correlations between EMG coherence and performance variables emerged after the intervention period likely because of the elimination of other factors that can influence performance. Before the initial steady cocontraction test, subjects were familiarized in two trials with the requirements of the task, including the production of unbalanced cocontraction with the target muscles and visuomotor coordination between their muscle activities and target lines on the monitor. Less prevalent positive correlations between performance variables and EMG coherence before the intervention period indicate the involvement of such other factors. Acute large improvements in performance (~30%) and marginal reductions in EMG coherence (6%) after the intervention period demonstrate that performance and EMG coherence do not change with corresponding magnitudes. This noncorrespondence also implies that factors not associated with EMG coherence may influence cocontraction performance. Before the intervention period, performance was low probably due to the involvement of continued familiarization and explorations of the novel task for understanding and satisfying the difficult requirements. A large improvement in performance ensued after the intervention period when the explorations were assumingly less. Nonsystematic variability of such familiarization and explorations as well as the consequent performance across subjects before the intervention can conceal the underlying correlations between performance variables and EMG coherence.
The degree of significant positive correlations between EMG coherence and performance variables was not uniform or high across muscle pairs (r ≤ 0.66). It should be noted that, despite smoothing, AEMG used for providing visual feedback and assessing performance still has a large amount of high-frequency components (originating from the shape of motor unit spikes) that are not directly related to the control of muscle activation level. This study purposefully used such signals so visual feedback responds to the activation changes intuitively without delay, which is not possible with substantial low-pass filtering. Hence, the limited degree of significant correlations is inevitable in this research design. There were also several cases without significant correlations, including the BB–TB pair for HIGH target. The currently available data set does not allow us to identify the specific causes for these cases. Although we focused on low-frequency correlated oscillations in the current study, they cannot be the sole neural mechanism that determines the performance of steady cocontraction. For example, steady cocontraction performance is suggested to be associated with excitability of corticospinal neurons and Ia presynaptic inhibition (12,19). It is possible that the contributions of listed and undefined other factors are variable depending on the task. Nonetheless, it is important to note that significant positive correlations between performance variables and EMG coherence were present in the muscle pairs (BR–TB and BB–BR pairs) that were not specifically used in visual feedback or task requirements. This interesting finding ensures that the significant correlations were observed not because of using the same original signal sources (i.e., EMG) between the assessments of performance and the correlated neural oscillations but because of fundamental activation characteristics across muscle pairs.
In testing the second hypothesis, we had expected that the subjects in the cocontraction group would achieve greater acute reductions in EMG coherence and accompanying performance variables than other groups. This was not the case based on the absence of a significant interaction of time and group on EMG coherence and performance variables. The out-of-phase cocontraction practice was designed with the expectation of attenuating in-phase low-frequency correlated oscillations by voluntarily tracking the target that required the repetition of out-of-phase low-frequency correlated oscillations between antagonistic muscles. Voluntary production of out-of-phase low-frequency correlated oscillations during the practice is confirmed by EMG phase around 160° or −145° in the antagonistic pairs in the cocontraction group. Although EMG phase was somewhat deviated from the expected 180° probably because of the difficulty of quickly alternating unbalanced cocontractions, the intervention goal of producing out-of-phase activation of antagonistic muscles was accomplished. The current findings with this practice are thus against our second hypothesis of specific effect of out-of-phase cocontraction practice on acute reductions in correlated neural oscillations and accompanying improvements of steady cocontraction performance. It is possible that the currently used conscious production of out-of-phase oscillations may not be able to influence the unconscious production of in-phase oscillations. In the literature, acute effects of practice on steady cocontraction performance have only been tested in distinct protocols: steady cocontraction practice with leg muscles (12,19). Hence, the current results are new important findings that demonstrate limited acute adaptability of (unconsciously produced) low-frequency correlated neural oscillations due to a bout of consciously produced out-of-phase oscillations via cocontraction practice at least in arm muscles.
One might think the invariant EMG coherence may imply the inability of the correlated oscillations for acute alterations. Indeed, low-frequency correlated oscillations during steady cocontraction are not altered acutely between leg agonist muscles, for example, because of neuromuscular fatigue (1). However, the low-frequency correlated oscillations are acutely adaptable at least between the BB and the TB based on an acute increase in low-frequency correlated oscillations between the muscles due to neuromuscular fatigue (22). The invariant median frequency of raw EMG in the present study also eliminates the possible confounding effect of neuromuscular fatigue. Thus, the current results against the second hypothesis would rather imply the independence of in-phase common drive from the out-of-phase drive. It is possible that the balance of using descending pathways of in-phase common drive and out-of-phase drive depends on the task and is not easy to be altered with an acute bout of intervention. Nonetheless, it would be interesting to examine whether a long-term training (e.g., practice several days per week for several weeks) with the out-of-phase cocontraction practice can modulate the low-frequency correlated oscillations during steady cocontraction.
The current neural findings on an association between the low-frequency correlated neural oscillations and steady cocontraction performance presumably have functional significance in stabilizing an object. As a follow-up, a next focused study on the roles of low-frequency correlated neural oscillations for mechanical motor output (i.e., force and position) is being planned. Nonetheless, the current results by itself open a possibility toward classifying individuals based on their potential ability for stabilization by analyzing EMG during cocontraction. This demonstration of a proof of concept would lead to its application to clinical, athletic, and factory settings, in which individuals who may have compromised capability for stabilization can be identified. Because a bout of used practice did not produce its specific efficacy for modulating neural oscillations, exploration into other practice strategies is warranted for improving the capability for stabilization.
The amount of in-phase low-frequency correlated neural oscillations between muscles was associated with the capability of healthy young adults for maintaining steady activation level during cocontraction of antagonistic muscles. Practice of enforcing out-of-phase low-frequency drive to antagonistic muscles did not specifically reduce the in-phase low-frequency correlated oscillations or improve steady cocontraction performance acutely. These findings suggest that individuals with less correlated neural oscillations tend to perform steady cocontraction more skillfully, and the low-frequency correlated oscillations may not be acutely modulated by one bout of out-of-phase cocontraction practice.
The authors thank Cole Simpson for the initial work on the project. This work was supported, in part, by the National Science Foundation (IIS 1317718). The authors acknowledge no conflict of interest. The results of the present study do not constitute endorsement by the American College of Sports Medicine. The results of the present study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation.