Participants

Eleven participants were included in this study (age: 22.6 ± 3.1 years; height: 176.4 ± 7.9 cm; weight: 70.5 ± 14.9 kg). The participants were healthy men without history of musculoskeletal injuries within the last 6 months. Although all participants were active physical education students, they did not have any resistance training background. All participants provided informed written consent and the recruitment took place between September 2022 and July 2023. This study was approved by the Faculty’s Ethics Committee (CEIFMH N.º: 24/2021) and all procedures adhered to the Declaration of Helsinki.

Data collection

Data processing

Intermuscular coherence (IMC).

Coherence analysis was performed using Matlab 2018a (Mathworks, Natick, MA, United States). Before data analysis, all data files were visually inspected to detect possible artifacts. In the cases where signals demonstrated some disruption or noise the repetitions were excluded from the analysis. IMC analysis was performed during 2500ms of each repetition, starting from 500ms after the onset of force production. A total of 7.5 seconds were analyzed. Before IMC analysis, raw EMG signals were full wave rectified [46,52,53], and data was divided into non-overlapping windows of 500 samples in length. The signals were subjected to fast Fourier transform, which gave a frequency resolution of 2 Hz. IMC was calculated as the cross-spectrum normalized by the auto-spectrum:

(1)

where is the magnitude squared coherence between two rectified EMG signals, and , for a given frequency . is the cross-spectrum and and are the auto-spectrum of and , respectively. Coherence above was considered significant [46,54]:

(2)

where L is the total number of sections and the significance level α was set to 0.05. The L-value was 15.

IMC was computed for the follow pairs of muscles: VL-VM, VL-RF, VM-RF, VL-GM, VL-GL, VL-BF, RF-BF and GL-TA. Total coherence between 10–100 Hz was averaged across participants at each timing of evaluation, as well as coherence sectioned into 15–35, 40–60 and 60–100 Hz bands. Within 15–35 Hz band representing beta band, two additional 15–24, 25–35 Hz bands were separated. Average coherence within each band was used to compare the evaluation timings (BSL, W0, W2, W4 and POS).

Statistics

Statistical analysis was made using jamovi software version 2.3.28.0 (Sydney, Australia). We verified normality through the Shapiro-Wilk test. A repeated measures ANOVA with four levels (W0, W2, W4 and POS time-points) was used to assess differences during the time course of the training program. When normality was not assumed, a Friedman test was used, and when sphericity was not assumed, a Greenhouse-Geisser correction was made. Tukey adjustments were used in Post hoc tests of ANOVA, and Durbin-Conover adjustments were used in Post hocs of Friedman test. The previously described analysis was made to all variables (PF, PRFD, absolute and normalized RFD in time intervals, RER of each knee extensor muscle as well as average coherence within each band). As IMC values within each band were not normally distributed we applied a log 10 transformation to allow parametric tests to be performed.

We assessed the relative reliability of the measurements between the BSL and W0 time points by calculating two-way mixed-effects intraclass correlation coefficients [ICC(3,1)] [55], with 95% confidence interval. ICC values were considered: poor, < 0.5; moderate, 0.5–0.75; high, 0.75–0.9; excellent, > 0.9. Additionally, we calculated standard error of measurement [SEM] as an absolute reliability measurement [56].

All statistical analyses were performed using a significance level of p < 0.05. Data is presented as mean ± SD. Cohen’ d was calculated as measure of effect size. Effect size was considered small, medium or large for Cohen’ d of 0.2–0.5, 0.5–0.8 and > 0.8, respectively. A post-hoc power analysis was conducted (G*Power 3.1.9.4, Düsseldorf, Germany) using the observed effect sizes of the primary outcomes showing differences from W0 to POS, using two-tailed parameters at α = 0.05. The achieved power was 0.805 for PF, 0.756 for PRFD, and 0.771 for RER of the RF. For the primary IMC outcomes presenting significant changes with training, the achieved power was 0.820 for TA-GL total IMC, and 0.727 for VL-RF at 15–24 Hz band. These values indicate that the study had adequate sensitivity to detect large effects, while small or moderate effects, especially concerning several IMC variables, may have been underpowered.

Reliability

Most mechanical variables regarding force production (PF, PRFD, absolute RFD at time intervals and %MVC at 50ms) revealed high to excellent reliability between BSL and W0 (0.85 < ICC < 0.98), with exception of %MVC at 100ms that was moderate (ICC = 0.73) and poor at 150 and 200ms (ICC = 0.37 and 0.29, respectively). The SEM of the variables with high to excellent reliability was consequently low, with the SEM of %MVC at 100, 150 and 200ms, that presented poor to moderate reliability varying from 4.69 to 7.34%.

Regarding muscle activation, there were variables presenting high (RER of RF, RER of RF between 0 and 50ms and between 50 and 100ms), moderate (RER of VL and VM, RER of VL and VM between 0 and 50ms) and poor (RER of VL and VM between 50 and 100ms) relative reliability. The SEM values for RER of VL, VM and RF were similar ranging from 0.07 to 0.08%/ms, and the RER at time intervals for the three muscles ranging from 0.12 to 0.18%/ms.

For the IMC analysis, ICC values revealed that most of the 48 variables (8 pairs of muscles x 6 bands of interest) present moderate to excellent reliability (moderate, 42% of the variables; high, 14%; excellent, 5%), although 39% of the variables present poor reliability.

Force production

A time effect was shown for PF during the leg press, with the final moment of evaluation (POS) being significantly different from the first (W0; p = 0.011 and d = 0.944). Although there was a trend for a progressive increment of PF along the weeks, there were not significant differences in post hoc comparisons between W0 and W2 or W4 even with medium effect sizes (p = 0.064; d = 0.516 and p = 0.073; d = 0.583, respectively). Concerning PRFD, only differences between W0 and POS were found (p = 0.015 and d = 0.888). Absolute and relative RFD only presented differences in post hoc comparisons between W2 and POS for absolute RFD between 50 and 100ms that was greater after training (p = 0.048; d = 0.302), and in the overall ANOVA for the relative RFD represented by %MVC at 200ms (p = 0.013; d = 0.917) that tended to reduce with training. Results are detailed in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Force and muscle activation related variables.

https://doi.org/10.1371/journal.pone.0346417.t001

Knee extensor activation

Muscle activation measured by RER from the onset of electrical activity to the point at PRFD showed to be unchanged for VM and VL. For RF, the ANOVA revealed a significant decreasing effect of training overtime (p = 0.045; d = 0.777) with post hoc comparisons presenting differences between W0 and W4 (p = 0.030; d = 0.904). Regarding RER at time intervals for VM and VL there was not any significant training effect. For RF, although there were not significant overall differences, there was post hoc differences between W0 and W2 for RER between 0 and 50ms (p = 0.044; d = 0.904) and between W0 and W4 for RER between 0 and 50ms (p = 0.020; d = 1.005) as well as between 50 and 100ms (p = 0.028; d = 0.863), with all the significant differences representing a decreased activation. Results are detailed in Table 1.

Intermuscular coherence

Fig 1 shows all the muscle pairs analyzed at each time point. A visual inspection of the coherence data series revealed that IMC at 25–45 Hz, encompassing the higher beta band as well as the initial frequencies of gamma band, presented significant peaks between VL and VM. IMC between VM and RF has also presented a significant peak at this frequency only at W0. Significant peaks at occasional time points have been shown between VL and BF at 15–35 Hz bands, and between GL and TA at 25–45 Hz only at the two first evaluations (W0 and W2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Intermuscular coherence of the analyzed pairs of muscles at W0 and POS between 0 and 100 Hz.

The horizontal dashed line represents significant coherence, and frequency-specific effects are reported quantitatively in the Results section. BF – biceps femoris; GL – gastrocnemius lateralis; GM – gluteus maximum; POS – Ending of training; RF – rectus femoris; TA – tibialis anterior; VM – vastus medialis; VL – vastus lateralis; W0 – Training beginning moment.

https://doi.org/10.1371/journal.pone.0346417.g001

The overall result of the ANOVA for each pair of muscles at each time point did not show any significant effect of training in IMC with exception of GL and TA IMC between 10–100 Hz that tended to be reduced with training (p = 0.025; d = 0.963). However specific differences between time points have been revealed by post hoc comparisons. Concerning agonist-agonist pairs, increased IMC between VL and RF at 15–24 Hz (p = 0.017; d = 0.857) was shown between W0 and POS. Regarding synergist pairs acting at different joints, VL and GL presented decreased IMC at 25–35 Hz after the six weeks of training (p = 0.047; d = 0.784). Agonist-antagonist pairs have also revealed occasional differences, with decreased IMC between VL and BF after the first two weeks of training at 15–24 Hz (p = 0.017; d = 0.567) band, although these changes had not been maintained during the subsequent weeks. Additionally, decreased IMC at 10–100 Hz frequency range has been shown between RF and BF between W4 and POS (p = 0.008; d = 0.065) and between GL and TA between W2 and W4 (p = 0.048; d = 0.725).

Maximal and rapid strength

Six weeks of squat training led to increased PF. Previous research showed increased isometric MVC PF following six weeks of dynamic strength training [24,25,63,64]. Although the sample size is small and caution should be taken when interpreting the results, it is interesting that our results only demonstrated a training effect after the sixth week of training. Previous studies reported increased isometric knee extension PF following only 3 weeks of training but only when training is performed with high loads (80% 1-RM) [24]. The same study showed that PF continued to increase until the completion of the 6 weeks for the group who trained with high loads, and that significant increases in PF with low loads were only achieved after 6 weeks. Others also found increases in isometric knee extension torque/force after 10 days [65] or 3 weeks of dynamic knee extension training [66], in isometric wrist flexion after 3 weeks of biceps curl training [67] or 2 weeks of dynamic wrist flexion training [18]. However, 2 weeks did not seem enough to increase isometric strength following dynamic elbow flexion training, with significant increases being observed after 4 weeks only with high loads [68]. Although significant results were only observed at the final evaluation, a progressive trend to an increase may be observed. This progressive trend to an increase may also be observed after 2 [64] and 3 weeks of strength training [25] although significant results were only achieved after the sixth week. The study of Knight and Kamen [64] has shown that individuals without training background present an increase in isometric force due to familiarization even without training. The familiarization session and the previously BSL session performed in our study may have helped to minimize the magnitude of this learning effect, although BSL and W0 measurements have shown good reliability. Additionally, task specificity may also have influenced the temporal pattern of the observed adaptations. The training program involved dynamic squats, while the testing task was an isometric leg-press. It is well-known that strength gains and the neural adaptations are specific to the training modality (as velocity and load) and contraction type [19,69,70]. Dynamic resistance training has been shown to transfer only partially to isometric testing, with reduced improvements compared to task-matched training conditions [71]. Therefore, the limited changes observed at intermediate time points may reflect incomplete transfer of training-induced adaptations to the isometric task rather than an absence of neuromuscular adaptations. This consideration is especially relevant when considering intermuscular coordination which is highly task dependent.

We have also found greater PRFD after six weeks of training. Previous research showed increased RFD after 6 weeks of dynamic strength training [25,72]. In our study we did not find differences in 0–50ms interval, but a trend for increased 50–100ms absolute RFD was observed. Previously, the 0–50ms interval in RFD analysis have shown to be greater in strength trained individuals when compared to untrained ones, although the 50–100ms interval may be increased for the untrained individuals possibly because of delayed activation of the agonist muscles [3–5]. These intervals are related to greater neural input to the muscles, namely greater firing frequency of motor units and muscle fiber conduction velocity [3,14]. The fact that we only verified significant results at 50–100ms time window may suggest that 6 weeks led to increased neural output to the muscles, although this activation may be delayed when comparing with long-term trained individuals. Concerning later intervals, our results have also shown decreased relative RFD at 200ms. The later intervals are related with cross-sectional area or muscle thickness, as well as PF [28,31]. Although we do not present muscular adaptations to the 6-week training program, it is plausible that the load dynamics adopted in this study led to gains in muscle mass. The decreased relative RFD at 200ms may be surprising, although it results from an unchanged absolute RFD at 150–200ms time interval, and a great magnitude increase in PF during the training period.

Furthermore, dynamic strength training programs have shown to increase RFD at specific time windows. Increased maximal rate of torque development of knee extensor and flexor muscles was observed after 10 weeks of barbell deadlift training, but also at time windows of 50 and 200ms [73]. Similar results were also achieved after 14 weeks of heavy lower limb training [30]. Also, heavy leg press training (90% 1-RM) during 8 weeks led to increased RFD at early and later time windows when performed with or without eccentric overload [27]. It is important to note that specific results may be found in current research. For example, 8 weeks of bench press training have also increased early RFD at 50ms during a bench press, but only when the velocity loss during the training program was 0%, while greater velocity losses resulted in increased later RFD [26]. Additionally, velocity losses above 20% have shown to decrease the early 50ms RFD in isometric squat following 8 weeks of squat training [74]. One study comparing 1 vs 4 vs 8 exhausting squat sets also found decreased RFD at all time intervals [75]. The previous cited studies observed differences in previously trained men, and as it was stated above, different training background will lead to different neuromuscular adaptations. Also, differences in study design may lead to different results.

Knee extensor activation

Muscle activation was measured by RER, and it was observed that the monoarticular portions of the quadriceps remained similar from W0 to POS, while RF decreased its activation overtime. This may be surprising, since RF is agonist of knee extension, and it was expected a greater activation of the knee extensors. Previous studies have found increased activation during the initial epochs of the contraction of plantar flexors following 4 weeks of isometric plantar flexion training [20], knee extensors of strength trained athletes during the initial 25ms of contraction [5], elbow flexors during the initial 30ms only after 2 weeks of dynamic elbow flexion training [76], VL during a jump squat after 10 weeks of power training only for a stronger group [22], or knee extensors during isometric knee extension after 14 weeks of heavy lower limb training [30]. The fact that the participants of the present study did not have any strength training experience may be related to the absence of adaptations concerning VL and VM activation, as suggested by James, Gregory Haff [22]. Also, the remaining enumerated studies evaluated single-joint while in our study we have evaluated a multi-joint exercise. This fact may have resulted in different neuromechanical patterns of force production since there is evidence of different neural input to agonist muscles between single- and multi-joint exercises [77] as well as between closed (as the leg press) and open kinetic chain (as knee extension) [78,79]. Indeed, the decreased activation of RF observed in the present study, although unexpected, aligns with recent evidence concerning the role of biarticular muscles, may be related with better efficiency of RF activation as a biarticular muscle, agonist of knee extension and antagonist of hip extension. First, studies using high-density surface EMG have shown that RF is regionally and task-dependently activated [80,81], also depending on hip flexion angle [82]. Second, the activity of RF during closed chain exercises has been demonstrated to be lower when compared to open chain exercises [77,83]. Third, during multi-joint exercises combining knee extension and hip extension, as the leg-press, RF activation is decreased when compared to single-joint knee extension, which may contribute to increased hip extension torque [77]. Consequently, it has been shown that single-joint knee extension training is preferential to enhance RF activation rather than multi-joint exercises [84]. Therefore, a reduction in RF RER after squat training may reflect a more efficient distribution of the neural drive during the task, reducing RF early activation when knee extension is combined with hip extension, and enhancing coordination with the monoarticular portions of the quadriceps. Nevertheless, recent work from our research group presented a trend for reduced RER of RF in highly trained individuals when compared to untrained ones, with different fatigue-related responses between groups (with the trained individuals increasing and the untrained individuals decreasing RF activation after fatigue) [85]. These findings further support that RF activation patterns are modified with training experience, and therefore the observed reduction in RF RER may reflect a task-specific adaptation to optimize the intermuscular coordination patterns. Future studies using high-density surface EMG should explore the regional adaptations in RF activation to confirm if the observed activation results from a redistribution of the neural drive or an overall neural input to the muscle.

Intermuscular coherence

One technique to study intermuscular coordination has been through muscle coupling between pairs of muscles with distinct functional roles, and it has been suggested that neural control of motor units of different muscles may not be independent, merging the concept of common input to motoneurons and synergistic coordination of movement [86,87]. Within this conceptual framework, IMC analysis has been used to distinguish signal input frequencies common between a pair of muscles, and concerning the correlation between the respective power spectral densities it allows to infer about the neural origin of movement control [43]. It is important to note that the frequency bands used in EMG-based IMC analysis are not physiologically equivalent to EEG-derived alpha-, beta-, or gamma-bands. Surface EMG reflects the summation of motor unit action potentials, whereas EEG reflects synchronized synaptic activity at the cortical level. Consequently, IMC frequency bands should be interpreted as reflecting frequency-specific common neural input to motor neuron pools from spinal and supraspinal origins, rather than direct cortical oscillatory activity [40,41,43]. The adoption of EEG-like terminology in IMC studies is largely used, reflecting established conventions in the literature, allowing comparison across studies. Nevertheless, interpretations should remain cautious, and future work combining EMG coherence with cortical measures (e.g., EEG–EMG coherence) may further clarify the neural origins of frequency-specific intermuscular coupling.

Concerning agonist coupling, in this study, we identified at W0 significant IMC peaks between 25 and 45 Hz between the single-joint portions of quadriceps and between VM and RF. This frequency range encompasses frequencies of beta (15–35 Hz) and gamma (> 40 Hz) bands usually associated with corticospinal pathway [40,88–91]. Beta band has also been linked to oscillations within the sensorimotor cortex and the integration of information from other cortical and subcortical structures as well as Ia afferents or Renshaw cells [92,93], while gamma band has been associated to attentional focus and efferent drive during strong isometric voluntary contractions or dynamic effort [94–97]. The peaks observed in IMC between VL and VM have remained, although IMC at lower beta band (15–24 Hz) decreased after 6 weeks of training. On the other hand, IMC between VL and RF increased within this frequency range. The fact that rate of activation has decreased but coupling between VL and RF increased at this band may have been an adapting process to better allow the kinetic output transmission from proximal to distal joints. Namely, since RF is antagonist at hip joint, its rate of activation may be decreased to facilitate an increase in torque development at that joint. Concurrently, augmenting common neural input may be more effective to allow a better synchronization with the monoarticular powerful muscles of quadriceps. It was also observed greater IMC at beta band between VM and RF in strength trained individuals [85] when compared to untrained ones, which means that the greater coupling between RF and the monoarticular knee extensors may be a neural adaptation to optimize maximal force production. On the contrary, one study has found lower beta band IMC between VM and RF in strength trained individuals [44]. However, in this study the method of calculation of IMC was based on wavelet analysis, and the evaluated task was an isometric knee extension which requires different patterns of muscle activation as discussed previously. Furthermore, lower IMC at gamma band was observed between RF and VM or VL than between VM and VL during squats which may suggest a specific mechanism of control of biarticular and monoarticular muscles during dynamic tasks [98]. Nevertheless, no significant changes were observed between VM and VL, which is in line with previous research that did not find differences in IMC between those muscles at any band following 14 weeks of strength training, although greater IMC at gamma band has been found during submaximal contractions at 70% MVC [46]. Furthermore, it is interesting that corticomuscular coherence, measured between electroencephalographic and EMG of lower limb muscles as the BF, TA and soleus, was lower in ballet dancers and weightlifters than in untrained individuals, which reveals a non-synchronized discharge of corticospinal cells in chronically trained athletes, which suggests that long-term training leads to fine postural adjustments dependent on specific temporal sequences [45].

Regarding synergist muscles acting at different joints, we have also found differences concerning the neural input between VL and GL as well as VL and GM, although VL and GM IMC was an occasional event that resulted from increased IMC of three participants. In its turn, VL and GL showed a trend to a decreased IMC at higher beta band (25–35 Hz). As we mentioned before, this band may be associated to corticospinal pathway, and the decreased IMC observed after training may arise from greater efficiency in muscle recruitment to produce joint moments. The fact that we observed altered patterns concerning the integration of RF during the task may also be applied to GL, considering its role as a biarticular muscle, antagonist of knee extension and agonist of plantarflexion. The reduced common corticospinal input to VL and GL after 6 weeks of training may be related to an adaptive role of GL as an ankle stabilizer during lower limb extension tasks [99,100], providing a more efficient force transmission. Additionally, the changes in the shared neural input to both muscles may be the result of a flexible adaptation in the recruitment of muscle synergies [101,102], supporting the beta oscillations from cortex to muscles [103].

Regarding the agonist-antagonist analyzed pairs, GL and TA showed significant peaks between 25 and 45 Hz (encompassing frequencies from beta and gamma bands). General IMC (10–100 Hz) was reduced after the 6 weeks of training. These results may suggest that both muscles share corticospinal projections that tend to be decreased by training, thus reducing co-activation. However, different neural strategies may be adopted concerning the task and the training mode. For example, badminton players presented higher GL and TA synchronization than untrained individuals, although reducing EMG amplitude of the GL during an isokinetic dorsiflexion, which may be interpreted as a more effective control strategy while maintaining ankle stability [104]. Additionally, it has been shown that distal muscles present stronger corticospinal connections than proximal ones [45,105,106] with weaker Renshaw cell inhibitory feedback [107,108] and greater number of muscle spindles in some muscles as TA or soleus compared to RF or BF [45,109], which may suggest that modulation of those mechanisms contributes to adaptations in IMC after training. However, these interpretations are inferential, since the present data refers exclusively to EMG signals and do not provide direct evidence of corticospinal modulation.

We have also observed that pre-training IMC peaks at 15–35 Hz and < 15 Hz band between VL and BF decreased after two weeks of training, although this behavior has not been maintained. This may have been a compensatory neural strategy temporally associated with the early phase of training during which muscle soreness or damage is often reported in untrained individuals. Previous research focusing on adaptations following eccentric training reported that biceps brachii muscle soreness only dissipated after two weeks of training [110]. Although these findings should not be directly transposed to our study, given our concentric training program, and that in the absence of direct perceptual or biochemical measures this interpretation remain speculative, it would be expected that muscle soreness was present in the first weeks of training and that the respective prolonged fatigue led to altered agonist-antagonist coordination. Namely, decreased IMC at lower beta (15–24 Hz) and upper alpha between 10 and 14 Hz (since the EMG hardware do not allow to fully characterize alpha-band), may be a consequence of lower corticospinal connectivity, but also from afferent feedback associated to alpha band [111–113]. Although our data may not represent the full alpha band, since the used EMG electrodes present a lower band pass filter threshold of 10 Hz, alpha band is linked to physiological tremor, and although its origin is multifactorial, the modulation of afferent feedback appears to be one mechanism contributing to IMC in this band [111]. It has been observed that afferent feedback modulation during an acute fatiguing state affects the CNS control of muscle recruitment [114,115]. Thus, despite further research is needed to clearly support this, the initial weeks of training may have led to prolonged fatigue that elicited specific agonist-antagonist connectivity modulated by the integration of corticospinal and sensory information. On the other hand, the decreased general IMC between RF and BF after W4, suggests that initial neural adaptations after 2 weeks of training may be transitional adjustments until the system becomes more efficient with task-specific optimization of central commands achieved with long-term training [45]. Furthermore, long-term strength trained individuals have shown different mechanisms of motor control of agonist and antagonist muscles, with lower activation and greater corticomuscular beta band suppression of the antagonist muscles [47,116].

Taken together, the direction and magnitude of IMC changes with training were dependent on both muscle pair and frequency-band, suggesting that early resistance training adaptations do not follow an uniform pattern. While increased IMC may reflect enhanced common neural input or improved functional coupling between muscles, decreased IMC may indicate a refinement of motor strategies and reduced shared drive as task execution becomes more efficient. Importantly, these adaptations appear to be frequency- and muscle-specific, highlighting that intermuscular coordination emerges from multiple parallel neural mechanisms rather than a single unifying process.

The results of this study should be interpreted with caution. First, although valid, the general limitations of the applied data processing methods should be considered when confronting the results with previous research. For example, onset definition of force and EMG may have impact in the results concerning RFD or RER, respectively. Second, muscle selection is determinant for IMC analysis results, as well as the evaluated task. Third, further research should consider using EMG electrodes that do not provide an automatic band-pass filter to a better understanding of full power spectrum from 0 Hz. Fourth, the rectification of EMG signals prior to IMC calculation is still a matter of debate. Rectification is highly recommended for low-force contractions [117], although it was not the case of our study. We opted by rectifying EMG signals to a better comparison with relevant studies although both options may be found in the literature. Fifth, the methods of IMC calculation can lead to different conclusions, although most studies calculate magnitude squared coherence as it was done in this study. Sixth, the present study focused on intermuscular coherence, which quantifies linear synchronization between EMG signals at identical frequencies. While IMC provides valuable insights into shared neural drive and functional coupling, it does not capture non-linear or cross-frequency interactions between muscles. Recent methodological developments, such as Amplitude-Amplitude Cross-Frequency coupling approaches, have been proposed to assess transient and non-stationary coordination across distinct EMG frequency bands, showing fatigue-, sex- and age-related differences in patterns of intermuscular coordination [118–120]. These approaches may offer complementary information regarding intermuscular coordination, identifying distinct rhythms embedded in surface EMG patterns. Future studies could combine IMC with cross-frequency coupling analyses to provide a more comprehensive characterization of neuromuscular connectivity during strength training adaptations. Seventh, it is important to consider that some IMC variables presented only poor reliability between BSL and W0, which may represent interday variability of untrained individuals, and that the significant results should be interpreted with caution because of the small sample size and generally low effect sizes in IMC adaptations. Furthermore, the conducted post-hoc power analysis concerning the main mechanical and electromyographic variables indicated adequate sensitivity for detecting large effects, although small and moderate effects may not have been detectable. Finally, another limitation of the study is the absence of a non-training control condition. Although two baseline assessments were performed (BSL and W0) without training between them, and reliability analysis confirmed that the primary neuromuscular variables were reliable across sessions, we cannot fully exclude the possibility that part of the observed adaptations over the six-week training period reflect learning. Nevertheless, it is also relevant to note that, within the context of the study, learning-related improvements should not be viewed solely as confounding influences. Rather, they represent a fundamental aspect of early neural adaptations to training, in which intermuscular coordination resulting in more efficient task execution is intrinsic to the physiological processes underpinning initial strength gains.