Neural Drive is Greater for a High-Intensity Contraction Than for Moderate-Intensity Contractions Performed to Fatigue : The Journal of Strength & Conditioning Research

Secondary Logo

Journal Logo

Scientific Manuscript Excellence Honor: Dr. Gary A. Dudley Memorial Paper

Neural Drive is Greater for a High-Intensity Contraction Than for Moderate-Intensity Contractions Performed to Fatigue

Miller, Jonathan D.1; Lippman, Jeremy D.1; Trevino, Michael A.2; Herda, Trent J.1

Author Information
Journal of Strength and Conditioning Research 34(11):p 3013-3021, November 2020. | DOI: 10.1519/JSC.0000000000003694
  • Free


Dr. Gary A. Dudley, 1952-2006


The potential of resistance training performed at lower loads to stimulate muscle hypertrophy and strength gains similar to high-load resistance training paradigms is of high research interest (12,36). A primary question remains, are motor unit (MU) activation patterns similar between lower-load contractions performed to fatigue and high-intensity contractions? It is hypothesized that the activation of the entire MU pool, as reported during high-intensity contractions, would maximally stimulate muscle protein synthesis and, in theory, result in overall increases in muscle hypertrophy and strength (21).

A recent meta-analysis comparing low- and high-load resistance training protocols concluded that high-load leads to greater increases in strength, but increases in muscle cross-sectional area are similar between low- and high-load resistance training (36). Schoenfeld et al. (36) reported the average percent increase in 1 repetition maximum (1RM) strength was 35.3% for high-load (>60% 1RM) in comparison to 28.0% for low-load (<60% 1RM) training, and increases in muscle size were similar between low- (7.0%) and high-load (8.3%) training protocols. Muscular endurance receives considerably less attention, although Schoenfeld et al. (37) reported a 16.0% increase in muscular endurance following low-load training in comparison to no improvement for high-load training. Although there is conflicting evidence (27,28), some report muscle fiber type specificity of training, such that low-load training may preferentially hypertrophy muscle fibers that primarily express Type I characteristics while high-load training may preferentially hypertrophy muscle fibers that primarily express type II characteristics (32,40). Although muscles fibers are more accurately represented as possessing characteristics along a continuum rather than distinct groupings (7), it is generally believed that low-threshold MUs are primarily composed of Type I fibers and high-threshold MUs are primarily composed of Type II fibers (2,3).

Analysis of MU activity during low-intensity fatiguing and high-intensity contractions may clarify whether similar MUs are involved in such contractions. In voluntary efforts in humans, MUs are recruited in order of increasing size, according to the size principle (15), and action potential amplitudes are correlated with the diameter of muscle fibers within a MU (14). Thus, the size principle can be observed by regressing action potential amplitudes against recruitment thresholds (RTs) during linearly increasing isometric muscle actions (17,26). A recently developed model of fatigue analyzed recruitment patterns and activity of MUs during fatigue across different contraction intensities and indicated lower-intensity contractions are far more fatiguing to earlier recruited (lower-threshold) smaller MUs in comparison to high-intensity contractions, which resulted in more fatigue to the later recruited (higher-threshold) larger MUs (33). Therefore, it is plausible that low-intensity fatiguing, and high-intensity contractions will yield differing MU recruitment and firing patterns.

Recruitment thresholds of high-threshold MUs are reported to decrease as fatigue develops (4,30). This is observed during prolonged contractions as new MUs with larger action potential amplitudes are recruited, and during subsequent contractions, as MUs are recruited earlier at lower force levels. Based on these findings, one assumption of the model (33) is that all MUs, including the largest highest-threshold MUs, would be recruited before exhaustion during lower- or moderate-intensity contractions performed to fatigue. However, previous studies have reported greater electromyographic (EMG) amplitude of the vastus lateralis (VL) during higher-intensity (75–90% of 1RM) than lower-intensity resistance training (30–50% 1RM) of the leg extensors performed to fatigue, including during leg extensions (20), leg press (35), and back squat (22). This would indicate that recruitment during low-to-moderate-intensity fatiguing contractions does not equal recruitment during high-intensity contractions. There are limitations, however, to interpreting EMG amplitude as a measure of excitatory drive or MU recruitment (8). There is limited research comparing excitatory drive between lower- and higher-intensity contractions in terms of MU firing rates and recruitment (30). One recent study (30) observed greater MU firing rates and larger action potential amplitudes of the VL during high-intensity contractions (70% maximum voluntary contraction [MVC]) performed to fatigue in comparison to low-intensity contractions (30% MVC) of the leg extensors performed to fatigue. The results indicated that the repetitive 30% MVCs did not necessitate the excitatory drive or recruitment equal to that of the repetitive 70% MVCs before exhaustion was reached, which contrasts what would be predicted by the model (33) and by other researchers suggesting that low-to-moderate-intensity contractions performed to fatigue recruit the entire MU pool (27,29). However, no study has investigated differences in neural drive and MU recruitment between a moderate-intensity contraction performed to fatigue and a single, near maximal, high-intensity contraction.

There are limitations to assessing neural drive and MU recruitment according to global EMG characteristics, not the least of which is amplitude cancellation (8). Therefore, the current study will also use surface EMG decomposition techniques to analyze MU firing rates, a more robust measure of neural drive (8), and MU recruitment via RTs and action potential amplitudes (4,17,26). There are still limitations of analyzing MU recruitment via action potential amplitudes, because they may be altered in the presence of fatigue (6). To partially combat this, we implemented a spike trigger averaging procedure (17–19,38) to validate the firing times and to exclude any MUs from analyses where the action potential waveform was unstable during each repetition. Each measure of muscle activation has limitations, which further supports the multi-faceted approach to analyzing muscle activation in the current study.

Therefore, the purpose of the current project is to analyze MU activity during repetitive fatiguing moderate-intensity contractions (50% MVCs) in comparison to a single high-intensity nonfatigued contraction (90% MVC). Results from the present study will provide further knowledge regarding the potential differences in MU activity during such tasks. The purpose of this paper is not to infer hypertrophic or other adaptations which may result from training; however, the findings may offer possible explanations for training load-specific adaptations which have been previously reported.


Experimental Approach to the Problem

This investigation used a repeated-measures design in which all subjects performed isometric leg extensions for a high-intensity contraction and for a series of repeated moderate-intensity contractions, which was performed to volitional fatigue. The contractions were performed at relative intensities of 90% (high-intensity) and 50% (moderate-intensity) of MVC. Surface EMG decomposition techniques were used for a multi-faceted approach to compare muscle activation via MU activity of the VL during the contractions. This allowed for a more thorough comparison of MU recruitment by observing action potential amplitudes and neural drive via the firing rates of similar sized MUs between contractions than EMG amplitude alone.


Subjects were comprised of 9 healthy individuals between the ages of 18 and 29 (mean ± SD, 7 males, 2 females, age = 22.78 ± 4.15 years, height = 173.78 ± 14.19 cm, mass = 87.39 ± 21.19 kg) who volunteered for the study. Subjects ranged from recreationally active (∼1–3 h·wk−1) to resistance trained (4–8 h·wk−1). Each subject was informed of the potential risks and benefits of participating in the study and voluntarily signed an institutionally approved written informed consent form before any data collection. This study was approved by the Human Subjects Committee—Lawrence at the University of Kansas.


The subjects visited the laboratory 2 times separated by at least 24 hours. The first visit was a familiarization trial followed by an experimental trial. During the first visit, the subjects practiced the isometric MVCs and submaximal trapezoidal contractions several times. For all isometric testing, each subject was seated with restraining straps over the pelvis, trunk, and contralateral thigh, and the lateral condyle of the femur was aligned with the input axis of the Biodex System 3 isokinetic dynamometer (Biodex Medical Systems, Inc., Shirley, NY). All isometric leg extensor strength assessments were performed on the right leg with the knee flexed at a 90° joint angle.

During the experimental trial, subjects completed a warm-up consisting of 3–5 brief voluntary isometric contractions from 30 to 80% MVC. Subjects then performed 3 isometric MVCs with strong verbal encouragement. The peak force from the 3 MVCs was used to determine the target force amplitude for subsequent isometric trapezoidal muscle actions at 90% MVC (REP90) and for repetitive contractions at 50% MVC which were performed to failure. The first 50% MVC was considered REP1 and the last completed 50% MVC was considered REPL. At least 2 minutes of rest were given between each MVC, the 90% MVC, and the start of the repetitive 50% MVCs. The rest interval was reduced to 8–9 seconds between each repetitive 50% MVCs. The trapezoidal trajectories consisted of a ramp-up period, where force increased linearly at a rate of 10% MVC·s−1, a constant force segment at 90% or 50% MVC, which was 12 seconds in duration, and a ramp-down where force decreased linearly at 10% MVC·s−1. Therefore, the duration of the 90% MVC was 30 seconds and the 50% MVCs were 22 seconds. Each subject was instructed to maintain their force output as close as possible to the target force presented digitally in real time on a computer monitor. The subjects did have difficulties completing the 90% MVC. The force was relatively unsteady and, in most instances, the subjects completed the isometric trapezoidal templates before 30 seconds because their rate of decline was quicker than the template.

Electromyographic Signal Detection and Processing

During the contractions, surface EMG signals were recorded from the VL using a 5-pin surface array sensor (Delsys, Inc., Natick, MA). The diameter of each pin is 0.5 mm, and they are placed at the corners of a 5 × 5-mm square, with the fifth pin in the center of the square. Before sensor placement, the surface of the skin was prepared by shaving and sterilized with an alcohol swab. To remove the dead layers of skin, hypoallergenic tape (3M; St. Paul, MN) was applied to the site, then peeled back to remove contaminants (Delsys, Inc., dEMG User Guide). The surface EMG sensor was placed over the belly of the VL and fixed with adhesive tape, whereas the reference electrode was placed over the contralateral patella.

For the 50 and 90% MVCs, action potentials were extracted into firing events of single MUs from the 4 separate EMG signals, sampled at 20 kHz, via the precision decomposition (PD) III algorithm (version 1.1.0) as described by De Luca et al. (5). Initially, the accuracy of the decomposed firing instances were tested with the reconstruct-and-test procedure (31). Only MUs decomposed with >90% accuracies were included in the analyses. In addition, a secondary spike trigger average (STA) procedure was included to validate the firing times and action potential waveforms generated via the PDIII algorithm. The derived firing times from the PDIII algorithm were used to STA the 4 raw EMG signals (17–19,24,25). A MU was included in further analyses if there were high correlations (r > 0.70) across the 4 channels between the PDIII algorithm (version 1.1.0) and STA derived action potential waveforms and the coefficient of variation of the STA derived peak-to-peak amplitudes across time was <0.30 (17). It is possible to observe seemingly valid MU action potential waveforms from trigger events that do not correspond with MU discharges (9). To examine this possibility, we added Gaussian noise to the discharge times identified from the PDIII algorithm (18,38). The Gaussian noise added to the firing times was set at 1% of the standard deviation of the interspike interval for each MU (18,38). Correlations were performed between the MU action potential waveforms created from the STA procedure with the small amount of noise (<2 ms shift in firing times) added to the firing times and the action potential waveforms derived from the PDIII algorithm. In addition, the peak-to-peak amplitudes of the STA action potential waveforms were compared with the STA action potential waveforms with the addition of Gaussian noise. A small amount of noise added to the firing times should reduce the correlation between action potential waveforms derived from the PDIII algorithm and the STA procedures. In addition, the peak-to-peak amplitudes of the STA action potentials should be diminished with the addition of small shifts in the firing times if no true action potential waveform is consistently present (Figure 1B, C).

Figure 1.:
Motor unit (MU) action potential templates as derived from the precision decomposition system III (PDIII), the spike trigger averaging procedure (STA) and the STA with firing times shifted according to Gaussian noise (STA + S) from channels 1–4 (A–D respectively). It is clear that the motor unit action potential waveforms were not present with the shift in firing times. Column scatter plots with mean (SD) bars for the average correlation of STA and STA + G MUAP waveforms with the PDIII derived waveforms (E) and for mean MU action potential amplitudes (MUAPAMPS) derived via STA and the STA + S (F). *Indicates that STA + S was significant less than STA.

For each MU, RT (RT [expressed relative to MVC]), MU action potential amplitude (MUAPAMP), and the mean firing rate (MFR) during the steady force plateau were recorded. A 2000-ms hanning window was applied to the MU firing instances to create the MFR curves. MUAPAMPS were calculated for each MU according to previous methods (17), because the average peak-to-peak amplitude values from each of the 4 unique action potential waveform templates using a custom-written software program (LabVIEW 2015; National Instruments, Austin, TX).

If the range of RTs of the sample of MUs recorded in any contraction was less than 12% MVC, the contraction was excluded from the RT-based relationship analyses, because this may lead to spurious relationship coefficients that do not fall in the physiological ranges for MU data. In addition, if less than 10 MUs were observed for 50% MVCs or 6 MUs for a 90% MVC, the data were excluded from all relationship-based analyses. Data from our laboratory have indicated when such procedures are followed, the interday reliability of the relationship coefficients of interest have excellent reliability (intraclass correlation coefficients = 0.801–0.901, p = 0.285–0.884). EMG from REP1 and REPL and REP90 were expressed as root mean squared amplitude for analysis.

Statistical Analyses

Linear regressions were performed on the MFR vs. RT and MUAPAMP vs. RT relationships for each subject with the y-intercepts and slopes used for statistical analysis. MFRs for MUs recruited at 40% MVC were predicted for each subject using their y-intercept and slope of the MFR vs. RT relationship for each contraction. Mean firing rate vs MUAPAMP relationships were fitted with an exponential model in accordance with previous methods (17,25) using the following equation: MFR = AeB(MUAPAMP). Where A is the theoretical MFR of a MUAPAMP of 0 mV, e is the natural constant, and B is the decay coefficient of MFR with increments in MUAPAMP, and the A and B terms were and used for statistical analysis.

A total of 10 one-way mixed-factorial ANOVAs (REP1 vs. REPL vs. REP90) were used to examine possible differences in the A and B terms for the MFR vs MUAPAMP relationships, the slopes and y-intercepts for the MFR vs. RT and MUAPAMP vs. RT relationships, the predicted MFRs for MUs recruited at 40% MVC, EMG amplitude, and mean and maximum MUAPAMP (recorded MU with the greatest APAMP) for each contraction. The difference in correlation coefficients between the PDIII algorithm and STA action potential waveforms with and without Gaussian noise shifted firing times were collapsed across contraction, converted to z-scores using Fisher's Z-Transformation, and the z-scores were analyzed using a dependent samples t-test. In addition, a dependent samples t-test was used to analyze differences in mean MUAPAMP between waveforms from STA and STA with Gaussian noise shifted firing times to further demonstrate deterioration of the action potentials with small errors introduced to the firing times identified by the PDIII algorithm. When appropriate, paired-samples t-tests were used as follow-ups for significant main effects and Cohen's d effect sizes were calculated for each pairwise comparison. The alpha level was set at p ≤ 0.05, and all statistical analyses were performed using SPSS, version 25 (IBM Corp., Armonk, NY).


The group completed an average of 10 ± 5 repetitions. After the removal of 39 MUs that failed to meet the validation criteria according to the STA procedures, a total of 576 MUs were analyzed in the current study. As expected, the minor Gaussian noise shift in firing times resulted in a significant decrease in the mean peak-to-peak amplitudes (98%, p < 0.001) of the STA action potentials and the correlations coefficients (11%, p < 0.001) with the PDIII algorithm derived waveforms (Figure 1). The changes in peak-to-peak amplitudes and correlations with the shifted firing times are similar to what is previously reported by Thompson et al. (2018) and Hu et al. (2013a) who used 2 different EMG decomposition methods. These findings further validate the firing times and action potential shapes of the MUs recorded in the current study.

Two contractions failed to meet the criteria for inclusion in the RT-based relationship analyses because their RT ranges were <12%. Each subject's relationships used for analyses were significant (r2 = 0.25–0.92) and demonstrated the expected properties of MU activity regardless of contraction. Together, the 3 calculated relationships depict that the later recruited higher-threshold MUs were larger and possessed lower firing rates at steady force for each subject and contraction. The recorded MU activity in the present study conforms to the size principle and the onion-skin scheme of firing rates (19). An illustration of the decomposed firing instances of MUs during REP90 and REPL contractions and the action potential shapes of a few representative MUs are presented in Figure 2. Motor unit and other EMG data for each repetition are presented in Table 1.

Figure 2.:
The normalized force and individual motor unit (MU) firing instances during the 90% maximum voluntary contraction (REP90) (A) and the final 50% maximum voluntary contraction (REPL) of the fatiguing protocol (B) for the same subject. Action potential templates (all 4 recorded channels are shown) from the first (C) and final (D) MUs recruited that were observed during REP90 and the final MU (E) recruited that was observed during REPL are presented along with their recruitment thresholds (RTs), mean firing rates (MFRs), and MU action potential waveform amplitudes (MUAPAMP). Of note, the defining characteristics of the MU action potentials waveforms are lost due to the scaling.
Table 1 - Mean ± SDs of the slopes and y-intercepts from the mean firing rate (MFR [pulses per second]) vs. recruitment threshold (RT [% MVC]) relationships and the motor unit action potential amplitude (MUAPAMP) vs. RT relationships, and the A and B terms from the MFR vs. relationships, as well as EMG amplitude, mean and maximum MUAPAMPS for the first (REP1) and last (REPL) repetition of the fatiguing protocol and the 90% MVC (REP90).*
MFR vs. RT
 Slopes (pps/%MVC) −0.486 ± 0.256 −0.532 ± 0.209 −0.374 ± 0.061
 Y-intercepts (pps) 30.39 ± 9.35 32.45 ± 11.25 33.33 ± 3.31
 Predicted MFR (pps) 10.92 ± 2.32 11.14 ± 3.48 18.38 ± 2.60
 Slopes (%MVC·mV−1) 0.00587 ± 0.00413 0.00596 ± 0.00280 0.00913 ± 0.00585
 Y-intercepts (mV) −0.0685 ± 0.103 −0.0318 ± 0.0862 −0.258 ± 0.210
A term (pps) 26.21 ± 5.40 28.07 ± 5.56 24.08 ± 4.49
B term (pps·mV−1) −4.96 ± 1.95 −4.77 ± 1.82 −2.63 ± 1.00
 Amplitude (mV) 42.99 ± 20.19 69.57 ± 27.18 100.64 ± 61.16
 Mean amplitude (mV) 0.109 ± 0.0468 0.178 ± 0.0668 0.263 ± 0.128
 Maximum amplitude (mV) 0.219 ± 0.0810 0.320 ± 0.127 0.520 ± 0.234
*MVC = maximum voluntary contraction; RT = recruitment threshold.
Significantly greater than REP1.
Significantly greater than REPL.

Mean Firing Rate vs. Recruitment Threshold Relationships

There was no significant main effect for contraction for the slopes (p = 0.277) or for the y-intercepts (p = 0.766) of the MFR vs. RT relationships. Therefore, the slopes and y-intercepts of the MFR vs. RT relationships were not significantly different between the high-intensity contraction and the first and last moderate-intensity contraction (Figure 3A). For the predicted MFRs of MUs recruited at 40% MVC, there was a significant main effect for contraction (p < 0.001). Dependent samples t-tests indicated predicted MFRs of MUs recruited at 40% MVC were similar between REP1 and REPL (p = 0.897, d = 0.07, 95% confidence interval (CI) [-0.98 to 1.12]), but were greater for REP90 than REP1 (p < 0.001, d = 3.03, 95% CI [1.03 to 4.97]) and REPL (p = 0.005, d = 2.45, 95% CI [0.71 to 4.13]) (Figure 4D).

Figure 3.:
Average predicted mean firing rate (MFR [pulses per second]) vs. recruitment threshold {RT (% maximum voluntary contraction [MVC])} relationships (A) for the 90% MVC (REP90) and the first (REP1) and the last (REPL) 50% MVC performed during the fatiguing protocol. The dotted vertical line shows the predicted MFR of motor units (MU) recruited at 40% MVC for each of the contractions. Average predicted MU action potential amplitude (MUAPAMP) vs. RT relationships (B) and MFR vs. MUAPAMP relationships (C) for REP90, REP1 and REPL.
Figure 4.:
Spaghetti plots illustrating the change in mean motor unit action potential amplitude (MUAPAMP) (A), maximum MUAPAMP (B), EMG amplitude (C), and predicted mean firing rate (pulses per second) at 40% maximum voluntary contraction (MVC) (D) from the first (REP1) and the last (REPL) 50% MVC performed during the fatiguing protocol and for the 90% MVC. *Significantly greater than REP1. †Significantly greater than REPL.

MUAPAMP vs. Recruitment Threshold Relationships

For the slopes, there was no significant main effect for contraction (p = 0.137). For the y-intercepts there was a significant main effect for contraction (p = 0.032); however, dependent samples t-tests indicated there were no significant differences between REP1 and REPL (p = 0.532, d = 0.39, 95% CI [-0.69 to 1.44), REP1 and REP90 (p = 0.057, d = 1.15, 95% CI [-0.11 to 2.34]), or REPL and REP90 (p = 0.059, d = 1.41, 95% CI [0.07 to 2.68]) for the y-intercepts (Figure 3B).

Mean Firing Rate vs. MUAPAMP Relationships

For the A terms, there was no significant main effect for contraction (p = 0.201), indicating MFRs of MUs with the smallest APAMPS were similar between contractions. However, there was a significant main effect for contraction for the B terms (p = 0.001). Paired samples t-tests indicated no significant difference between REP1 and REPL (p = 0.731, d = 0.10, 95% CI [−0.83 to 1.02]), but B terms were less negative for REP90 than REP1 (p = 0.005, d = 1.50, 95% CI [0.30–2.64]) and REPL (p = 0.001, d = 1.46, 95% CI [0.27–2.59]). The decrement in MFRs for MUs with larger APAMPS was less pronounced for REP90, which indicated larger MUs maintained greater firing rates during the high-intensity contraction than the first or last moderate-intensity contractions (Figure 3C).

EMG Amplitude

There was a significant main effect for contraction (p = 0.002). Dependent samples t-tests indicated EMG amplitude was greater during REPL than REP1 (p < 0.001, d = 1.11, 95% CI [0.02–2.15]) and during REP90 than REP1 (p = 0.007, d = 1.27, 95% CI [0.14–2.35]), but was not significantly greater for REP90 than REPL (p = 0.076, d = 0.66, 95% CI [-0.34 to 1.62]). EMG amplitude was greater for both the high-intensity contraction and the last moderate-intensity contraction in comparison to the first. A moderate effect size suggested EMG amplitude was also greater for the high-intensity contraction in comparison to the last moderate-intensity contraction (Figure 4C).

Mean and Maximum MUAPAMPS

There was a significant main effect for contraction for the mean MUAPAMPS (p < 0.001). Dependent samples t-tests indicated mean MUAPAMPS were greater for REPL (p < 0.001, d = 1.20, 95% CI [0.08–2.26]) and REP90 (p = 0.004, d = 1.60, 95% CI [0.37–2.77]) than REP1 and were greater for REP90 than REPL (p = 0.038, d = 0.83, 95% CI [−0.20 to 1.82]) (Figure 4A). In addition, there was a significant main effect for contraction for maximum MUAPAMPS (p < 0.001). Dependent samples t-tests indicated maximum MUAPAMPS were greater for REPL (p = 0.005, d = 0.94, 95% CI [−0.11 to 1.95]) and REP90 (p = 0.002, d = 1.72, 95% CI [0.45–2.93]) than REP1, and were greater for REP90 than REPL (p = 0.008, d = 1.06, 95% CI [−0.02 to 2.09]) (Figure 4B). Observed MUAPAMPS were largest during the high-intensity contraction despite that MUAPAMPS were greater for the last moderate-intensity contraction in comparison to the first when considering the average observed MUAPAMP and the largest MUAPAMP observed for each contraction.


Previous research has reported that over the course of fatiguing protocols consisting of low-to-moderate-intensity contractions, EMG amplitude increases (20,35,39), additional MUs are recruited (4), and firing rates increase (4,30). It is suggested that as fatigue is approached, these measures of excitation and recruitment increase until all MUs are recruited and excitation is maximal (27,33). Some of the findings of the current study agree with this theory. There was no significant difference between the 90% MVC and the final 50% MVC for EMG amplitude or the slopes and y-intercepts of the mean firing rate and action potential amplitude vs. RT relationships. The initial conclusion from these analyses would be that there were no differences in MU activity between the 90% MVC and the final 50% MVC of the fatiguing protocol.

However, further analyses suggest greater excitation and MU recruitment for the 90% MVC in comparison to the final 50% MVC of the fatiguing protocol. These included: (a) greater predicted firing rates of MUs recruited at 40% MVC (Figures 3A and 4D), (b) greater firing rates of MUs with similar action potential amplitudes (smaller B terms for MUAPAMP vs. RT relationships) (Figure 3C), and (c) greater mean and maximum action potential amplitudes of observed MUs (Figure 4A, B) during the 90% MVC. The greater MU action potential amplitudes tentatively suggest larger MUs were active, whereas the greater firing rates for MUs with a given RT or action potential amplitude indicate greater neural drive (8) during the 90% MVC in comparison to the last 50% MVC. These findings provide evidence that the operating point of MU control (4) was greater for a high-intensity contraction than for the final repetition of a moderate-intensity fatiguing protocol.

Increased EMG amplitude and MUAPAMPS suggested excitation increased and additional MUs were recruited for the last 50% MVC in comparison to the first 50% MVC. Although there was a slight trend for greater firing rates for the final 50% MVC (Figure 3A, C), no significant increase in firing rates was observed. These findings generally agree with previous research investigating changes in neural drive and MU recruitment with fatigue during repetitive contractions. It is typically observed that firing rates increase, RTs decrease, and larger MUs are needed to sustain the contractions as fatigue develops (4,30). Only one study (30) has compared fatigue related changes in firing rates and MU recruitment between lower- (30% MVC) and higher- (70% MVC) intensity repetitive contractions. Motor unit recruitment and firing rates were increased with fatigue at both intensities, but greater firing rates and recruitment of larger MUs during the high-intensity condition was reported. Results from the current study agree with those found by Muddle et al. (30) indicating MU recruitment and neural drive during the final repetition of a moderate-intensity fatiguing protocol did not match that of a higher-intensity condition, although the high-intensity condition in the current study consisted of just a single 90% MVC.

The motoneuron inhibitory effect of central fatigue may explain the decreased neural drive and MU recruitment during the moderate-intensity isometric contraction (1). Gandevia et al. (13) investigated voluntary activation via interpolated twitch over the time course of sustained isometric MVCs of the elbow flexors and reported high initial voluntary activation (>99%), which decreased over the course of the contraction (90.7%). Feedback from Group III/IV afferents associated with metaboreceptors and mechanoreceptors that are activated by muscular contraction do not inhibit initial muscle activation, but increase their inhibitory effect (central fatigue) as a contraction is maintained or repeated to constrain peripheral fatigue (1,13). Furthermore, central fatigue prevents voluntary activation from reaching maximal levels during voluntary submaximal isometric contractions (23) such as in the moderate-intensity contractions in the current study. During prolonged submaximal contractions, firing rates are increased and new MUs are recruited to maintain force as peripheral fatigue develops (4,30). However, as the contraction continues, Group III/IV afferent feedback inhibits further increases in neural drive or MU recruitment to constrain peripheral fatigue and avoid potentially deleterious effects in the working muscles (1). Therefore, it is plausible that central fatigue inhibits the recruitment of the largest MUs during low-to-moderate-intensity submaximal contractions performed to fatigue.

EMG amplitude has previously been observed to be greater during higher- than lower-intensity leg extensions (20) and leg presses (35) to fatigue. However, in these studies, EMG amplitude seemed to be similar for the last repetitions of the lower-intensity contractions in comparison to the first repetition of the higher-intensity contractions. In the current study, the difference in EMG amplitude from the final rep of the fatiguing 50% MVCs (69.57 ± 27.18%) to the 90% MVC (100.64 ± 61.16%) did not reach significance, but the Cohen's d indicated a moderate effect size (0.66) which tentatively suggested EMG amplitude was greater for the single high-intensity contraction than the final moderate-intensity contraction. The EMG amplitude data in the present study does provide support to researchers that caution against over interpretation of this measurements of muscle activation to monitor fatigue. The lack of significant differences between contractions according to EMG amplitude was contrasted by the findings of significant differences in action potential amplitudes and firing rates between contraction intensities. Therefore, examining MU activity via decomposition techniques provides a more complete interpretation of neural drive via firing rates and recruitment via action potential amplitudes of MUs than global EMG amplitude (8).

Limitations of the current study include that we cannot determine to what extent action potential amplitudes may have been affected by changes in the metabolic environment because of fatigue and, therefore, interpretations of action potential amplitudes from the current study should be made with caution. In addition, some of the differences in RTs between contractions may be because of the algorithm's ability to detect certain MUs under fatiguing or high-intensity contractions and the physiological changes in RTs that occur with fatigue.

The results of the current study also highlight the importance of analyzing this phenomenon from multiple perspectives. There may be many measures by which high-intensity contractions are indistinguishable from fatiguing moderate-intensity contractions, but conclusions should not be drawn from these alone. Many studies have reported similarities between fatiguing low- or moderate-intensity and high-intensity resistance training in hormone responses (28), fiber type specific muscle glycogen depletion (29), and one study reported similar peak EMG amplitude (34). In addition, studies have reported similar responses to such training programs for strength (11,16) and muscle hypertrophy, in Type I and Type II fiber cross-sectional area (28) or whole muscle cross-sectional area and volume (10,16,27). However, most EMG studies report greater muscle activation (20,22,30,35) for high-intensity contractions and the majority of resistance training studies report greater strength gains from high-load programs (27,28,36,37). In addition, although gains in hypertrophy from training may be similar, there is evidence for preferential Type II muscle fiber hypertrophy from higher-intensity training and preferential Type I muscle fiber hypertrophy for lower-intensity training (32,40). The current study is not well suited to infer hypertrophic adaptations, which may be gained through longitudinal training, but the findings may offer an explanation for training load-specific adaptations that have been previously reported. That is, it is plausible this effect is because of higher-threshold MUs being fatigued to a greater extent during high-intensity contractions, whereas lower-threshold MUs are fatigued to a greater extent during lower-intensity contractions (33). Variability among responses in neural drive and recruitment patterns among the contractions was observed (Figure 4), which would likely result in different adaptions following resistance training. Therefore, variability in neural drive during low- and high-load resistance training among individuals may help explain variability in adaptions following these modes of resistance training.

In summary, greater neural drive and MU recruitment were observed during a single high-intensity contraction than moderate-intensity contractions before or at volitional fatigue. The current findings agree with previous research investigating MU activity during fatigue and comparing muscle activation during high- and low-intensity fatiguing protocols. However, a multifaceted approach to examining muscle activation by analyzing the activity of individual MUs provided a more comprehensive investigation than previous studies analyzing EMG amplitude alone. It is speculated that Group III/IV afferent inhibition prevented neural drive and MU recruitment during the fatiguing moderate-intensity contractions to equal that of the high-intensity contraction.

Practical Applications

The current findings have implications for models of MU activity during fatigue and resistance training paradigms. We have provided further evidence that moderate-intensity contractions performed to volitional fatigue appear to not equal the levels of neural drive and recruitment that are observed in high-intensity contractions. The different contractions (high-intensity vs. fatigued moderate-intensity) in the current study would presumably stress and adapt a different subpopulation of MUs and comprising muscle fibers. Individuals, coaches, and practitioners should recognize that fatiguing bouts performed at low-to-moderate intensities may not include a significant amount of activity of the higher-threshold MUs where the largest skeletal muscle fibers are generally present.


The authors would like to thank the undergraduate students who assisted in the project, as well as each of the subjects for their selfless participation. The authors have no conflicts of interest to disclose.


1. Amann M. Central and peripheral fatigue: Interaction during cycling exercise in humans. Med Sci Sports Exerc 43: 2039–2045, 2011.
2. Bottinelli R, Canepari M, Pellegrino MA, Reggiani C. Force-velocity properties of human skeletal muscle fibres: Myosin heavy chain isoform and temperature dependence. J Physiol 495: 573–586, 1996.
3. Burke RE, Levine DN, Tsairis P, Zajac FE. Physiological types and histochemical profiles in motor units of the cat gastrocnemius. J Physiol 234: 723–748, 1973.
4. Contessa P, De Luca CJ, Kline JC. The compensatory interaction between motor unit firing behavior and muscle force during fatigue. J Neurophysiol 116: 1579–1585, 2016.
5. De Luca CJ, Adam A, Wotiz R, Gilmore LD, Nawab SH. Decomposition of surface EMG signals. J Neurophysiol 96: 1646–1657, 2006.
6. Dimitrova NA, Dimitrov GV. Interpretation of EMG changes with fatigue: Facts, pitfalls, and fallacies. J Electromyogr Kinesiol 13: 13–36, 2003.
7. Enoka RM, Duchateau J. Inappropriate interpretation of surface EMG signals and muscle fiber characteristics impedes understanding of the control of neuromuscular function. J Appl Physiol 119: 1516–1518, 2015.
8. Farina D, Holobar A, Merletti R, Enoka RM. Decoding the neural drive to muscles from the surface electromyogram. Clin Neurophysiol 121: 1616–1623, 2010.
9. Farina D, Merletti R, Enoka RM. The extraction of neural strategies from the surface EMG: An update. J Appl Physiol 117: 1215–1230, 2014.
10. Fink J, Kikuchi N, Nakazato K. Effects of rest intervals and training loads on metabolic stress and muscle hypertrophy. Clin Physiol Funct Imaging 38: 261–268, 2018.
11. Fisher JP, Steele J. Heavier and lighter load resistance training to momentary failure produce similar increases in strength with differing degrees of discomfort. Muscle Nerve 56: 797–803, 2017.
12. Fry AC. The role of resistance exercise intensity on muscle fibre adaptations. Sports Med 34: 663–679, 2004.
13. Gandevia SC, Allen GM, Butler JE, Taylor JL. Supraspinal factors in human muscle fatigue: Evidence for suboptimal output from the motor cortex. J Physiol 490: 529–536, 1996.
14. Hakansson CH. Conduction velocity and amplitude of the action potential as related to circumference in the isolated fibre of frog muscle. Acta Physiol Scand 37: 14–34, 1956.
15. Henneman E, Olson CB. Relations between structure and function in the design of skeletal muscles. J Neurophysiol 28: 581–598, 1965.
16. Hisaeda H, Miyagawa K, Kuno S, Fukunaga T, Muraoka I. Influence of two different modes of resistance training in female subjects. Ergonomics 39: 842–852, 1996.
17. Hu X, Rymer WZ, Suresh NL. Motor unit pool organization examined via spike-triggered averaging of the surface electromyogram. J Neurophysiol 110: 1205–1220, 2013.
18. Hu X, Rymer WZ, Suresh NL. Assessment of validity of a high-yield surface electromyogram decomposition. J Neuroengineering Rehabil 10: 99, 2013.
19. Hu X, Rymer WZ, Suresh NL. Reliability of spike triggered averaging of the surface electromyogram for motor unit action potential estimation. Muscle Nerve 48: 557–570, 2013.
20. Jenkins NDM, Housh TJ, Bergstrom HC, et al. Muscle activation during three sets to failure at 80 vs. 30 % 1RM resistance exercise. Eur J Appl Physiol 115: 2335–2347, 2015.
21. Kraemer WJ, Ratamess NA. Fundamentals of resistance training: Progression and exercise prescription. Med Sci Sports Exerc 36: 674–688, 2004.
22. Looney DP, Kraemer WJ, Joseph MF, et al. Electromyographical and perceptual responses to different resistance intensities in a squat protocol: Does performing sets to failure with light loads produce the same activity?. J Strength Cond Res 30: 792–799, 2016.
23. Löscher WN, Cresswell AG, Thorstensson A. Excitatory drive to the alpha-motoneuron pool during a fatiguing submaximal contraction in man. J Physiol 491: 271–280, 1996.
24. McManus L, Hu X, Rymer WZ, Suresh NL, Lowery MM. Muscle fatigue increases beta-band coherence between the firing times of simultaneously active motor units in the first dorsal interosseous muscle. J Neurophysiol 115: 2830–2839, 2016.
25. Miller JD, Sterczala AJ, Trevino MA, et al. Motor unit action potential amplitudes and firing rates during repetitive muscle actions of the first dorsal interosseous in children and adults. Eur J Appl Physiol 119: 1007–1018, 2019.
26. Milner-Brown HS, Stein RB, Yemm R. The orderly recruitment of human motor units during voluntary isometric contractions. J Physiol 230: 359–370, 1973.
27. Mitchell CJ, Churchward-Venne TA, West DWD, et al. Resistance exercise load does not determine training-mediated hypertrophic gains in young men. J Appl Physiol 113: 71–77, 2012.
28. Morton RW, Oikawa SY, Wavell CG, et al. Neither load nor systemic hormones determine resistance training-mediated hypertrophy or strength gains in resistance-trained young men. J Appl Physiol 121: 129–138, 2016.
29. Morton RW, Sonne MW, Farias Zuniga A, et al. Muscle fibre activation is unaffected by load and repetition duration when resistance exercise is performed to task failure. J Physiol 597: 4601–4613, 2019.
30. Muddle TWD, Colquhoun RJ, Magrini MA, et al. Effects of fatiguing, submaximal high- versus low-torque isometric exercise on motor unit recruitment and firing behavior. Physiol Rep 6: e13675, 2018.
31. Nawab SH, Chang S-S, De Luca CJ. High-yield decomposition of surface EMG signals. Clin Neurophysiol 121: 1602–1615, 2010.
32. Netreba A, Popov D, Bravyy Y, et al. Responses of knee extensor muscles to leg press training of various types in human. Ross Fiziol Zh Im I M Sechenova 99: 406–416, 2013.
33. Potvin JR, Fuglevand AJ. A motor unit-based model of muscle fatigue. PLOS Comput Biol 13: e1005581, 2017.
34. Schoenfeld BJ, Contreras B, Vigotsky AD, et al. Upper body muscle activation during low-versus high-load resistance exercise in the bench press. Isokinet Exerc Sci 24: 217–224, 2016.
35. Schoenfeld BJ, Contreras B, Willardson JM, Fontana F, Tiryaki-Sonmez G. Muscle activation during low- versus high-load resistance training in well-trained men. Eur J Appl Physiol 114: 2491–2497, 2014.
36. Schoenfeld BJ, Grgic J, Ogborn D, Krieger JW. Strength and hypertrophy adaptations between low- vs. high-load resistance training: A systematic review and meta-analysis. J Strength Cond Res 31: 3508–3523, 2017.
37. Schoenfeld BJ, Peterson MD, Ogborn D, Contreras B, Sonmez GT. Effects of low- vs. high-load resistance training on muscle strength and hypertrophy in well-trained men. J Strength Cond Res 29: 2954–2963, 2015.
38. Thompson CK, Negro F, Johnson MD, et al. Robust and accurate decoding of motoneuron behaviour and prediction of the resulting force output: Neural drive to muscle and resulting force output. J Physiol 596: 2643–2659, 2018.
39. Vila-Chã C, Falla D, Correia MV, Farina D. Adjustments in motor unit properties during fatiguing contractions after training. Med Sci Sports Exerc 44: 616–624, 2012.
40. Vinogradova OL, Popov DV, Netreba AI, et al. Optimization of training: Development of a new partial load mode of strength training. Fiziol Cheloveka 39: 71–85, 2013.

muscle activation; action potential amplitude; motor unit firing rate; vastus lateralis

© 2020 National Strength and Conditioning Association