Understanding perceived exertion in power-aimed resistance training: the relationship between perceived exertion and muscle fatigue

Introduction

Strong evidence suggests that resistance training is effective in maintaining and enhancing muscle mass, functional capacity, and athletic performance (Garber et al., 2011). However, when performing resistance exercises, individuals often experience subjective sensations such as fatigue and exertion (Ekkekakis, Parfitt & Petruzzello, 2011). A previous study reported that perceived exertion is associated with endurance performance outcomes, such as time to exhaustion (Di Fronso et al., 2020). Therefore, understanding these subjective exertion is crucial for developing appropriate resistance training guidelines, achieving optimal athletic performance, and designing long-term periodized training programs.

Perceived exertion during physical exercise can be quantified using the rating of perceived exertion (RPE) scale. RPE is a subjective, perception-based method commonly used to assess exercise intensity (Lagally et al., 2002a; Kilpatrick et al., 2009; Lea et al., 2022). Originally developed for evaluating the intensity of rehabilitation and other medical treatments, the RPE scale has since been widely adopted in physiological and sports-science research due to its simplicity and its validity in reflecting physiological responses to physical activity (Robertson et al., 2004; Fontes et al., 2010). Over the past 30 years, with the growing popularity of resistance training worldwide, researchers and practitioners have increasingly incorporated RPE into resistance exercise protocols as supporting evidence for exercise prescription (Lagally & Robertson, 2006; Gearhart et al., 2009; Zhao, Nishioka & Okada, 2022). During aerobic exercise, many factors affect RPE, including environmental temperature, body temperature, and hypoxia (Tucker, 2009). In resistance exercise, several key factors have been examined as contributors to increased RPE, including load, volume, and rest-interval configuration (Lagally et al., 2002a; Hiscock et al., 2018; Zhao, Yamaguchi & Okada, 2020). As a result, RPE has become a valuable physiological marker for resistance training, offering both practicality and insight.

Increased fatigue levels have been identified as one of the primary contributors to elevated RPE scores (Fontes et al., 2010; Blain et al., 2016; Weakley et al., 2017; Cruz-Montecinos et al., 2019). However, most previous studies have examined the relationship between RPE and either peripheral (e.g., cortisol, blood lactate) or central (e.g., command from motor cortex) responses in isolation, typically under controlled laboratory conditions (De Morree, Klein & Marcora, 2012; Vargas-Molina et al., 2020). As a result, it remains unclear whether RPE during actual resistance exercise scenarios is predominantly influenced by central or peripheral factors.

With the growing adoption of velocity-based training, velocity loss has been recognized as an indirect marker of fatigue in resistance exercise (Sánchez-Medina & González-Badillo, 2011). Studies have demonstrated that velocity loss is significantly correlated with fatigue-related metabolic markers, such as blood lactate and ammonia concentrations, during power-aimed resistance exercises (Izquierdo et al., 2006; Sánchez-Medina & González-Badillo, 2011). However, despite its growing application, the validity of velocity loss as a fatigue marker lacks sufficient neuromuscular evidence.

Surface electromyography (sEMG) is a non-invasive tool widely used to assess neuromuscular function in physiological, sports science, and clinical settings (González-Izal et al., 2012; Campanini et al., 2020). When sEMG is used to grasp fatigue induced by peripheral mechanisms, changes in power spectral density serve as a marker of fatigue (Brody et al., 1991). These changes, caused by intramuscular metabolic disturbances, result in a shift of the power spectrum toward lower frequencies (Dimitrov et al., 2006; González-Izal et al., 2012; Campanini et al., 2020). In contrast, when sEMG is used to measure central drive, signal amplitude is typically employed as an indicator of motoneuronal activity (De Morree, Klein & Marcora, 2012; Blain et al., 2016; Campanini et al., 2020). Overall, sEMG provides a valuable means to investigate both peripheral and central nervous system responses during muscle contractions.

Consequently, the use of RPE to quantify exertion during resistance training has increased; however, it remains unclear whether exertion is influenced predominantly by central or peripheral factors. Accordingly, elucidating the contributors to RPE using neuromuscular evidence may help practitioners in training, sports science, and clinical settings better understand the relationship between perceived exertion and actual performance. The present study aims to: (1) examine the relationship between RPE and neuromuscular fatigue, as assessed by sEMG, to determine whether RPE is primarily influenced by central or peripheral responses; and (2) evaluate the validity of velocity loss as a fatigue indicator by analyzing its relationship with neuromuscular fatigue during power bench press (BPress) exercises. We hypothesized that (1) peak root mean square (RMS) would increase and (2) median frequency (MDF) would decrease as RPE increased. However, (3) velocity loss may reflect neuromuscular fatigue only under certain conditions.

Materials & Methods

Results

Significant differences in overall RPE were observed across the experimental conditions (χ² = 31.524, p < 0.001). Post hoc comparisons revealed significant differences between L and M (p = 0.024), M and H (p = 0.011), and L and H (p < 0.001) conditions (Fig. 1A). For average velocity loss, significant differences were found between L and M (p = 0.001, Cohen’s d = 1.616), as well as between L and H (p < 0.001, Cohen’s d = 2.090) (Fig. 1B). Significant differences in average MDF were observed between M and H (p = 0.027, Cohen’s d = 0.831) and between L and H (p = 0.003, Cohen’s d = 1.288) (Fig. 1D). In contrast, no significant differences were found in average RMS across the conditions (Fig. 1C).

Intra-set data were analyzed using a two-way ANOVA, with the first, mid-point, and last repetitions extracted for comparison. Significant interaction effects between condition and repetition were observed for RPE (p < 0.001, F = 39.060, partial η² = 0.723), velocity loss (p < 0.001, F = 13.266, partial η² = 0.469), and MDF (p < 0.001, F = 19.934, partial η² = 0.571). A significant main effect of repetition was also found for RPE (p < 0.001, F = 73.833, partial η² = 0.831), velocity loss (p < 0.001, F = 62.850, partial η² = 0.807), and MDF (p < 0.001, F = 74.986, partial η² = 0.833), indicating consistent changes throughout the lifting tasks. Within-condition comparisons showed significant increases in RPE across repetitions for all three conditions (L, M, H; all p < 0.001). For velocity loss, significant increases were observed in the M (p < 0.001, F = 32.041, partial η² = 0.681) and H (p < 0.001, F = 46.172, partial η² = 0.755) conditions, but not in the L condition (p = 0.545, F = 0.619, partial η² = 0.040). For MDF, significant decreases were observed in the L (p < 0.001, F = 25.893, partial η² = 0.633), M (p <0.001, F = 28.440, partial η² = 0.655), and H (p < 0.001, F = 77.326, partial η² = 0.838) conditions. Between-condition comparisons revealed significant differences in RPE at the mid-point (p < 0.001, F = 34.748, partial η² = 0.698) and last repetition (p < 0.001, F = 73.708, partial η² = 0.831). Velocity loss also differed significantly between conditions at the mid-point (p < 0.001, F = 9.805, partial η² = 0.395) and last repetition (p < 0.001, F = 37.089, partial η² = 0.712).

For MDF, a significant between-condition difference was found at the last repetition (p < 0.001, F = 28.436, partial η² = 0.655). For peak RMS, only a significant main effect of repetition was found (p < 0.001, F = 22.654, partial η² = 0.602). No significant main effect of condition or interaction effect was observed for intra-set RMS.

The results of the Spearman correlation analysis are presented in Table 1. No significant correlations were found between MDF and velocity loss across the three experimental conditions.

Discussion

The purpose of this study was twofold: (1) to examine the relationship between RPE and neuromuscular responses, as indexed by RMS and MDF of sEMG, and (2) to evaluate the validity of velocity loss as an indicator of fatigue during power BPress exercises. The key findings of the study are as follows: (1) MDF, velocity loss, and RPE showed similar trends across conditions, suggesting that the experimental tasks induced fatigue, and that RPE is primarily influenced by peripheral fatigue-related responses. (2) No significant differences were observed between conditions in either intra-set peak RMS or average RMS, indicating that central motor command remained relatively stable despite increasing fatigue levels during power BPress exercises. (3) Although velocity loss increased alongside MDF, no significant correlations were found between the two, suggesting that while velocity loss may reflect aspects of muscle fatigue, it only has limited precision in the context of power BPress.

Subjective exertion during physical exercise is influenced by a range of physiological responses, spanning from central nervous system activity to peripheral muscle fatigue (Hollander et al., 2003; De Morree, Klein & Marcora, 2012). Previous studies have reported a direct association between RPE and central motor command, demonstrated through various neuromuscular measures during resistance exercise (Pincivero et al., 1999; De Morree, Klein & Marcora, 2012). For example, concurrent increases in RMS, RPE, and electroencephalogram amplitude were observed during elbow flexion tasks (De Morree, Klein & Marcora, 2012). Regarding peripheral mechanisms, RPE has also been shown to correlate significantly with metabolic and endocrine markers such as blood lactate, testosterone, and cortisol concentrations (Weakley et al., 2017; Hiscock et al., 2018). In the present study, we observed significant increases in RPE accompanied by significant decreases in MDF across all experimental comparisons. In contrast, no significant changes in RMS corresponding to RPE were found. Notably, the trends in average MDF and overall RPE were closely aligned (Figs. 1A and 1D), and RPE increased significantly as MDF decreased within conditions (Figs. 2A and 2D). Although the MDF difference between conditions at the mid-point did not reach statistical significance, the effect approached the threshold (p = 0.055), suggesting a possible trend. A decrease in MDF, as measured by sEMG, is typically attributed to a reduction in muscle fiber conduction velocity, which is strongly influenced by local physiological changes such as pH reduction and accumulation of metabolic byproducts. These factors alter the metabolic milieu at the muscular level (Brody et al., 1991; Amann et al., 2011; Gorostiaga et al., 2012; Keller et al., 2019). Additionally, these peripheral changes increase afferent feedback from group III/IV muscle fibers, contributing to heightened sensations of exertion (Blain et al., 2016; Broxterman et al., 2017). Neuromuscular fatigue is typically characterized by a decrease in MDF and an increase in RMS (Enoka & Duchateau, 2008; Cruz-Montecinos et al., 2019). However, in the current study, RMS did not increase significantly across repetitions in any condition. These findings suggest that fatigue was predominantly localized to peripheral musculature, and that peripheral mechanisms play a key role in RPE determination. In contrast, central motor command—as indicated by RMS—appears to remain relatively stable and may have only a limited influence on subjective exertion during power-aimed BPress exercises.

Previous studies have investigated the mechanisms of fatigue in the context of pedaling exercises, identifying contributions from both central and peripheral factors (Amann, 2011; Amann et al., 2011; Blain et al., 2016). However, these findings may not be directly applicable to resistance training due to the unique characteristics of this modality, such as abrupt intensity shifts driven by rest intervals, the lifting-to-rest phase ratio, and distinct physiological demands (Paulo et al., 2012; Weakley et al., 2017; Zhao, Yamaguchi & Okada, 2020). As a result, the specific contributions of central and peripheral fatigue in resistance exercise remain less clearly defined. In the present study, neuromuscular fatigue was assessed indirectly using sEMG, based on the relationship between muscle fatigue and bioelectrical responses. Although peripheral biomarkers (e.g., blood lactate) were not measured, a significant increase in velocity loss was observed alongside a decrease in MDF (Figs. 1B and 1D; Figs. 2B and 2D). These corresponding changes strongly suggest the presence of substantial fatigue induced by the lifting tasks (Sánchez-Medina & González-Badillo, 2011). However, no significant increases in RMS were observed, indicating that central motor command remained stable across conditions. This pattern supports the conclusion that, during power-aimed resistance exercises, neuromuscular fatigue is primarily driven by peripheral mechanisms. These findings have important implications for training prescription. For instance, if the goal is to minimize fatigue while maintaining high power output, coaches and personal trainers should avoid scheduling power-aimed training sessions immediately after endurance-focused workouts or game days, as these activities typically involve high workloads and may leave behind residual metabolic byproducts that impair subsequent performance (Thorlund et al., 2008). Conversely, the stable RMS levels observed across conditions offer a novel insight for time-efficient training strategies. Since power-aimed training does not appear to demand high central output, it may be effectively combined with strength-oriented sessions involving heavy loads and low repetitions (Rodriguez-Lopez et al., 2022). This strategy could help maximize training outcomes without imposing excessive fatigue.

For practitioners such as coaches and personal trainers implementing velocity-based training, accurately assessing fatigue is essential—particularly when the training objective is to maximize power output (Sánchez-Medina & González-Badillo, 2011; Mayo, Iglesias-Soler & Kingsley, 2019). To make fatigue more observable in real time, velocity loss has been proposed as a reliable indirect marker of neuromuscular fatigue during power-aimed training, and various commercial tools are currently available to facilitate this approach (Sánchez-Medina & González-Badillo, 2011; Martínez-Cava et al., 2020; Külkamp et al., 2024). Most previous studies have validated velocity loss using devices commonly employed in training environments, such as linear position transducers. In contrast, the present study used a high-speed camera and a reflective marker to measure barbell velocity, offering higher temporal precision than many commercial encoders (e.g., 100 Hz). Although linear position transducers have become popular in recent years, there is still a lack of validation against gold standards such as motion capture (Pérez-Castilla et al., 2019). Thus, when using velocity loss as a muscle-fatigue indicator, precision may be negatively affected if a linear encoder is used for this purpose. Under these conditions, our findings suggest that velocity loss may not be as reliable as previously reported, and caution is warranted when relying solely on this metric during power-aimed training. Although velocity loss showed a trend similar to MDF, no significant correlations were observed between the two in any condition. Notably, in the L condition, velocity loss did not change significantly during the lifting task (p = 0.545; Fig. 2B), whereas MDF exhibited a significant decline (p < 0.001). This discrepancy highlights a key limitation: while velocity loss may increase with fatigue, it does not always provide sufficient precision to accurately reflect neuromuscular fatigue. Moreover, if velocity-based training relies on measurement devices with limited precision, this may lead to underestimation of fatigue levels, potentially increasing injury risk and compromising performance. Therefore, it is recommended that velocity loss be used in conjunction with other fatigue markers—such as jump height and RPE—to enhance the accuracy of fatigue monitoring (Ilias, 1998; Zhao, Seo & Okada, 2024).

This study has several limitations that should be acknowledged. First, only upper-body power BPress was examined. Prior research has demonstrated that upper- and lower-body exercises may elicit different fatigue responses due to differences in physiological characteristics, such as lactate kinetics (Van Hall et al., 2003). Additionally, fatigue responses in the lower body have been reported to be more complex and variable (Zhao, Nishioka & Okada, 2022). Second, to capture neuromuscular responses of actual power-aimed training, we used sEMG in present study. While RMS is interpreted as an indicator of central motor command, it is an indirect measure and may not fully capture the complexity of central fatigue. Thus, interpretations regarding central output should be made with caution. Third, although the sample size was determined using statistical power analysis, it remains relatively small, and the predictive model may not be fully suited to the present design (e.g., within- or between-factors ANOVA). For example, the pairwise comparison of MDF at the mid-point approached but did not reach statistical significance (p = 0.055), suggesting that a larger sample may be needed to confirm certain effects. Future studies should aim to develop more effective methods for collecting intra-set RPE data, directly compare fatigue responses between upper- and lower-body power-aimed training, and recruit larger samples to enhance the statistical power and generalizability of the findings.

Conclusions

The present study demonstrated that MDF, RPE, and velocity loss exhibited similar trends across experimental conditions, while no significant changes were observed in RMS. These findings suggest that fatigue during power-aimed BPress is primarily induced by peripheral mechanisms, and that subjective exertion is largely influenced by peripheral rather than central responses. In contrast, power BPress at moderate intensity does not appear to require substantial central motor command. From a practical standpoint, coaches and practitioners may consider integrating power-aimed training with strength-focused sessions, particularly when time constraints prevent separate implementation of both modalities, as power-aimed training may not impose excessive central fatigue. No significant correlations were found between MDF and velocity loss, indicating that velocity loss alone may be insufficient for comprehensive fatigue monitoring, particularly when measured with lower-precision devices. Therefore, when applying velocity loss as a fatigue indicator in training settings, practitioners may consider combining it with additional markers, such as jump height or RPE, to improve the accuracy of fatigue assessment.

Supplemental Information