Exercise type, training load, velocity loss threshold, and sets affect the relationship between lifting velocity and perceived repetitions in reserve in strength-trained individuals

Introduction

Velocity-based strength training (VBST) has become a popular objective approach for managing exercise intensity/load and neuromuscular fatigue (Bastos, Machado & Teixeira, 2024; Jovanovic & Flanagan, 2014; Maroto-Izquierdo et al., 2025; Suchomel et al., 2021). Velocity-based strength training differs from traditional strength training by focusing on lifting velocity rather than loads expressed as a percentage of one repetition maximum (1RM) or a specific number of repetitions. Typically, VBST is based on the mean velocity of the bar (ascending phase), measured by devices such as a linear encoder or an accelerometer (Jovanovic & Flanagan, 2014; Weakley et al., 2021). A prerequisite with VBST—and using mean velocity as a variable—is maximal effort in the ascending phase of each repetition (Jovanovic & Flanagan, 2014).

It is well established that desired strength training outcomes rely on load intensity (e.g., % of 1RM) and volume (Atha, 1981). However, the effort in each lift (intention to move) and proximity to set failure can also be used to tune the adaptations to strength and power training (Grgic et al., 2022; Kawamori & Newton, 2006). While hypertrophy appears to be most effectively achieved with (near) failure sets, non-failure sets are recommended as preferable for developing power and maximal strength in well-trained individuals and athletes (Carroll et al., 2018). Managing and controlling neuromuscular fatigue during exercise may improve training quality and promote specific adaptations (Izquierdo et al., 2006; Larsen, Kristiansen & van den Tillaar, 2021; Suchomel et al., 2021).

As the mean velocity gradually declines with successive repetitions, a percentage loss of velocity or pre-set velocity threshold (in ms⁻¹) can be used to decide when to terminate the set. In practice, low velocity loss thresholds (i.e., <25%) appear preferred for developing muscle power and maximal strength, while hypertrophy seems to require higher velocity loss thresholds (>25%), i.e., close to contraction failure or RM (Baena-Marín et al., 2022; Hickmott et al., 2022; Weakley et al., 2021).

The mean velocity (ms⁻¹) has been demonstrated to correlate with both the percentage of 1RM and repetitions in reserve (RIR) (Pelland et al., 2022). Repetitions in reserve refer to the difference between the number of repetitions performed and the maximum possible repetitions in a set. However, it is evident that the relationship between mean velocity and RIR depends on the type of exercise, e.g., the bench press and the squat, and the high inter-individual variability seems to necessitate individual adjustments for practical application (Garcia-Ramos et al., 2018; Jukic et al., 2024; Moran-Navarro et al., 2019).

An alternative to objective measures of mean velocity is the subjective evaluation of effort and fatigue through the rate of perceived exertion (RPE) and perceived repetitions in reserve (pRIR) (Helms et al., 2016; Helms et al., 2020; Zourdos et al., 2016). The main argument for using RPE and pRIR is autoregulation, which is an individual, intuitive, and dynamic way to account for biological day-to-day variations and readiness to train (Suchomel et al., 2021). Perceived RIR is undeniably simplistic, practical, and low-cost, with no need for technological devices.

Both the objective mean velocity (Garcia-Ramos et al., 2018; Moran-Navarro et al., 2019) and subjective pRIR (Hickmott, Butcher & Chilibeck, 2024; Varela-Olalla et al., 2019) have been reported to predict the actual RIR in a set with acceptable accuracy. The pRIR seems, however, more influenced by the proximity to failure and training status than the objective, mean velocity measurements (Halperin et al., 2022; Moran-Navarro et al., 2019). That said, the applicability of the mean velocity method is also debatable, as some find good accuracy and reliability (Jukic et al., 2024), while others concluded that the mean velocity does not appropriately align with actual RIR values across loads and sets (Mansfield et al., 2023).

Previous studies examining the relationship between mean velocity, RIR, and pRIR have done so in controlled laboratory settings, using only a few test sessions (Garcia-Ramos et al., 2018; Jukic et al., 2024; Mansfield et al., 2020; Varela-Olalla et al., 2019). In such experimental settings, when multiple conditions tested in series, e.g., different loads and repetition ranges, the order and proximity of the conditions may affect the perceived efforts and ratings. First, the number of lifts and sets becomes restricted due to the development of neuromuscular fatigue, meaning that the total number of data points from each individual becomes limited. Second, the participants are, in fact, aware of the purpose of the study and may develop biased responses when assessing similarities or differences between the various conditions, compared to real-world scenarios where the focus is on the training itself (i.e., the Hawthorne effect). This raises questions about the ecological validity of the available literature on the topic. An alternative approach is to collect mean velocity and pRIR data from consecutive real-life training sessions that vary in load intensity and proximity to failure.

This study explored how exercise type (bench press and squat), load intensity (% of 1RM), velocity loss thresholds, number of sets, and sex may influence the relationship between objectively measured bar mean velocity and pRIR across multiple strength training sessions in well-trained individuals.

We recognize both VBST and pRIR as potentially valuable tools for managing exercise effort and training specificity (Helms et al., 2016; Weakley et al., 2021), and our results may offer clearer insights into how these tools can be applied by individuals involved in strength training.

Methods

Results

A total of 2,972 mean velocity and pRIR combinations were analyzed. Table 2 presents a summary of the number of lifts differentiated by exercise, velocity loss group (LVL and HVL), sex (male and female), and type of session (low, moderate, or high loads).

The individual regression lines in Fig. 1 depict the relationship between mean velocity and pRIR (measured per set) for each participant. For the squat exercise, the correlation coefficients (r-values) between mean velocity and pRIR ranged from 0.3 to 0.9, with an average correlation of 0.6 ± 0.2 (mean ± standard deviation) across participants. For the bench press, the individual correlations ranged from 0.1 to 0.9, with an average correlation of 0.6 ± 0.2.

In Figs. 2–8, pRIR is plotted as a function of mean velocity. The data is differentiated on exercise, velocity group (LVL vs. HVL), type of session (low, moderate, and high load), and sex. Table 3 gives an overview of the number of data points for the different plots.

The LMM analysis (Table 4) revealed that the mean velocity was strongly associated with pRIR (β = 6.99, p < 0.001), indicating that faster movements correspond with higher pRIR values. Group (LVL vs. HVL) was the second strongest predictor (β = 0.94, p < 0.001), suggesting that individuals in the HVL group tend to report higher pRIR. Exercise type demonstrated a negative effect (β = −0.67, p < 0.001), meaning that the bench press was associated with lower pRIR values than the squat. The interaction between group and exercise type was significant and negative (β = −0.33, p < 0.001), which suggests that the difference in pRIR between squat and bench press was smaller in the HVL group than in the LVL group. Additionally, the number of sets per session was inversely related to pRIR (β = −0.11, p < 0.001). Weeks of training (β = −0.01, p = 0.27) and sex (β = 0.35, p = 0.26) did not significantly affect pRIR. The random effects revealed considerable between-subject variability, with a standard deviation of 0.96 for the intercept and 2.25 for the slope of mean velocity and a strong negative correlation (r = −0.73) between these random components. The within-subject residual variability was 0.69.

Discussion

This study explored the relationship between mean velocity of the bar and pRIR across a 6-week training period, analyzing nearly 3,000 data points. The key findings were: (1) The mean velocity was strongly associated with pRIR but with individual differences from trivial to very large correlations. (2) The type of exercise, i.e., bench press and squat, the pre-set velocity loss thresholds (LVL or HVL), the session load (% of 1RM), and the number of sets influenced the relationships between bar velocity and pRIR. (3) The pRIR was not appreciably influenced by sex and the number of training weeks.

Perceived or predicted RIR (pRIR) is a subjective variable with inherent biases (Hackett et al., 2017; Hackett et al., 2012; Helms et al., 2017a). Hackett et al. (2012) observed that the participants tended to underestimate actual RIRs in the first sets (sets 1 and 2) but more accurately predicted RIR in the later sets (sets 3 and 4). Hackett et al. (2017) reported an accuracy of ∼1 repetition when the actual RIR was 0–5, while the inaccuracy increased to >2 repetitions with an actual RIR of 7–10. These observations were supported by Zourdos et al. (2021). In our study, pRIRs were <5, suggesting that participants likely demonstrated acceptable RIR estimation accuracy. In support of this, the LMM analysis showed a within-subject residual variability of 0.69, indicating that the participants’ reported pRIR values, on average, deviated from the model’s predictions by less than one repetition.

All participants reported lower pRIR for a given mean velocity in the squat than in the bench press. This aligns with observations that 1RM is achieved with a significantly lower mean velocity in the bench press (∼0.1–0.2 ms⁻¹) than in the squat (∼0.2–0.3 ms⁻¹) (Gonzalez-Badillo & Sanchez-Medina, 2010; Helms et al., 2017b; Iglesias-Soler et al., 2019; Zourdos et al., 2016). The greater range of motion and bar displacement, combined with a shorter sticking region and stronger acceleration in the post-sticking phase of the squat compared to the bench press, may explain the difference (Kompf & Arandjelović, 2017; van den Tillaar, Andersen & Saeterbakken, 2014).

In the present study, about half of the participants trained with low bar velocity loss thresholds (LVL) for terminating each set (i.e., 20% for squat and 30% for bench press), while the other half trained with high bar velocity loss thresholds (HVL: 40% in the squat and 60% in the bench press)—i.e., close to contraction failure in each set. Visual inspection of Figs. 3 and 4 shows that the HVL group reported higher pRIR values for a given mean velocity than the LVL group, which was most pronounced for the bench press exercise. This finding was confirmed by the LMM, which detected a significant group effect of about one pRIR and an interaction effect between exercise type and velocity group. By design, the LVL group terminated the sets at higher mean velocities than the HVL group, meaning the LVL group was likely to underestimate the RIR. This suggestion aligns with observations from previous studies (Hackett et al., 2012; Zourdos et al., 2021), which have shown systematically more accurate RIR prediction closer to failure. Nonetheless, it remains possible that the HVL group overestimated their RIR.

Our participants conducted three different sessions per week, consisting of low-, medium-, and high-load sessions (Table 2). We observed a clear load effect on pRIR, with low and moderate loads resulting in lower pRIR values at a given mean velocity compared to heavy loads. This was evident for both exercises, especially for the squat (Figs. 7 and 8). Interestingly, the sensation and experience of effort and exertion seem to depend on several factors, including neural drive from the motor cortex, which increases with higher muscle force requirements, and neuromuscular fatigue (Pageaux, 2016). Heavy loads challenge the lifter’s maximal capacity, where even minimal fatigue can lead to failure. In contrast, lighter loads allow for continued repetitions despite severe neuromuscular fatigue (Behm et al., 2002). Thus, the sensation of muscle fatigue and discomfort (Pollak et al., 2014) might lead to an underestimation of RIR with lower loads. Alternatively, the participants were overly optimistic in their RIR predictions with heavy loads due to low muscular fatigue and discomfort, “it just feels heavy”. In correspondence with this, Mansfield et al. (2023) observed an effect of load (60% vs. 80% of 1RM) in the prediction (estimates) of RIR in the bench press, where the mean velocity at 0-3 RIR was higher with 60% of 1RM load than with 80% of 1RM. Physiologically, this suggests that neuromuscular fatigue during low loads—likely due to an increased number of repetitions—causes a steeper decline in mean velocity during the final repetitions of a set compared to heavier loads with fewer total repetitions performed (hence, the drop in velocity may be non-linear).

Mansfield et al. (2023) observed that RIR decreased with each successive set. We found the same; for a given mean velocity, the pRIR was lower in, for example, the third set compared to the first. This can be due to neuromuscular fatigue per se and/or the fact that the participants became less optimistic about their RIRs for a given mean velocity due to the perception of fatigue. This effect was, nevertheless, rather small, accounting for about −0.1 RIR per set (according to the LMM analysis).

We observed that the correlations between mean velocity and pRIR varied considerably at the individual level, but no systematic improvements (changes) in the relationship between the mean velocity and the pRIR occurred during the 6-week training period, inferring that the participants did not improve (or worsen) their RIR prediction accuracy. This finding is in agreement with recent studies (Halperin et al., 2022; Jukic et al., 2024; Remmert, Laurson & Zourdos, 2023) that revealed no apparent effects of training status (experience) on the accuracy of RIR prediction. However, it would be intriguing to see if individuals with poor correlations between bar velocity and pRIR could improve by deliberate practice, as no feedback was given during the training period in the current study.

Subjective assessments of effort and exertion (such as RPE and RIR) are reported to be practically similar between sexes across different modes of exercise and exertion (Losnegard et al., 2021; Morishita et al., 2018; Naclerio & Larumbe-Zabala, 2017a; Naclerio & Larumbe-Zabala, 2017b). This is in line with our observations; the effect of sex was not significant in the LMM analysis. From visual inspection of Figs. 5 and 6, we can, however, see a trend where the females recorded higher pRIR values at a given bar velocity. Accordingly, Odgers et al. (2021) observed lower bar velocities in females than males at 7-9 RPE (corresponding to 1-3 RIR) in the front squat exercise but not in the hexagonal bar deadlift exercise. Moreover, Naclerio & Larumbe-Zabala (2017b) observed no sex difference in squats in strength-trained athletes. Generally, females demonstrate similar or better neuromuscular endurance than males, which might, at least partly, be related to absolute lower strength and an on average higher distribution of type I fiber in females than males (Hunter, 2009). Note, however, that we had a limited number of participants (both females and males), causing a risk of missing a potential sex effect.

A key limitation of this study was measuring only pRIR, not actual RIR, precluding quantification of true estimation precision. To mitigate this, our analysis was restricted to a pRIR of 0–4, as estimations within this range are generally more accurate and considered practically acceptable (Bastos, Machado & Teixeira, 2024; Halperin et al., 2022; Jukic et al., 2024).

Another potential limitation is the use of mean velocity. It could be that mean propulsion velocity (the mean velocity of the acceleration phase) and/or the lowest/minimum velocity had resulted in different results and should be further investigated; however, with the loads used in the present study (∼77–89% of 1RM), the differences between mean propulsion and mean velocities of the full movement should be very small or negligible (Sanchez-Medina et al., 2017; Sanchez-Medina, Perez & Gonzalez-Badillo, 2010).

Despite its limitations and descriptive nature, the present study has high ecological value by virtue of an extensive data set from actual training sessions conducted by well-strength-trained individuals. Moreover, all sessions were supervised. The inclusion of both sexes and different pre-set velocity loss targets (low and high), as well as sessions with different loads (low, medium, and heavy) in both bench press and squat (an upper and a lower body exercise), provided us with several interesting observations that allude to important nuances to be aware of when applying VBST and pRIR in training.

Practical implications

Different strength exercises, e.g., bench press and squat, exhibit distinct velocity-RIR profiles at the individual level. Mean velocity and pRIR are complementary but not interchangeable methods, as also noted by Mansfield et al. (2023). Consequently, training sessions managed by VBST or pRIR may differ, even when the intended session goal is the same. However, we do not assert that one method is superior to the other, as both have advantages and limitations. For power development and maximal strength training, VBST offers a key advantage over pRIR, as higher RIR values can compromise estimation accuracy, and bar velocity is directly relevant for assessing performance and training quality via biofeedback (Weakley et al., 2021). In contrast, pRIR may be equally effective or even preferable for hypertrophy-focused training, particularly when working within 0–2 RIR (Grgic et al., 2022; Halperin et al., 2022; Suchomel et al., 2021). Combining both methods may provide optimal benefits for high-level athletes, whereas RIR alone is likely sufficient for beginners, recreational lifters, and non-athlete populations, such as patients (Bastos, Machado & Teixeira, 2024; Maroto-Izquierdo et al., 2025). Further research across diverse populations is needed to better evaluate the advantages of VBST compared to or in combination with pRIR, given that pRIR is cost-free and easy to implement.

Conclusion

In this study, we investigated the relationship between mean velocity of the bar and pRIR during multiple exercise sessions. We conclude that type of exercise (bench press and squat), load (in %RM), the number of sets, and the proximity to failure (velocity loss threshold), but not sex and the number of training weeks, influenced the relationships between mean bar velocity (of the ascending phase) and pRIR. Both mean velocity and pRIR can be used to manage strength training, but these metrics may not be used interchangeably. We recommend using VBST in conjunction with pRIR for optimal control during strength training.

Supplemental Information