The velocity of resistance exercise does not accurately assess repetitions-in-reserve

ABSTRACT

This study assessed the reliability of mean concentric bar velocity from 3- to 0-repetitions in reserve (RIR) across four sets in different exercises (bench press and prone row) and with different loads (60 and 80% 1-repetition maximum; 1RM). Whether velocity values from set one could be used to predict RIR in subsequent sets was also examined. Twenty recreationally active males performed baseline 1RM testing before two randomised sessions of four sets to failure with 60 or 80% 1RM. A linear position transducer measured mean concentric velocity of repetitions, and the velocity associated with each RIR value up to 0-RIR. For both exercises, velocity decreased between each repetition from 3- to 0-RIR (p ≤ 0.010). Mean concentric velocity of RIR values was not reliable across sets in the bench press (mean intraclass correlation coefficient [ICC] = 0.40, mean coefficient of variation [CV] = 21.3%), despite no significant between-set differences (p = 0.530). Better reliability was noted in the prone row (mean ICC = 0.80, mean CV = 6.1%), but velocity declined by 0.019–0.027 m·s⁻¹ (p = 0.032) between sets. Mean concentric velocity was 0.050–0.058 m·s⁻¹ faster in both exercises with 60% than 80% 1RM with (p < 0.001). At the individual level, the velocity of specific RIR values from set one accurately predicted RIR from 5- to 0-RIR for 30.9% of repetitions in subsequent sets. These findings suggest that velocity of specific RIR values vary across exercises, loads and sets. As velocity-based RIR estimates were not accurate for 69.1% of repetitions, alternative methods to should be considered for autoregulating of resistance exercise in recreationally active individuals.

Highlights

• Bar velocity of bench press and prone row repetitions decreases on average from 3- to 0-repetitions in reserve (RIR) and is faster for lighter versus heavier loads

• The velocity of 3- to 0-RIR varied across four sets for the prone row but was more reliable for the prone row than the bench press

• At the individual participant level, there was not a consistent decrease in velocity between consecutive repetitions, and target velocities for specific RIR values were not correctly predicted in most cases.

• Using velocity stops may not be an appropriate method to predict specific RIR in recreationally active individuals.

KEYWORDS:

• Fatigue
• musculoskeletal
• strength
• training

Introduction

Resistance training is typically prescribed to target specific neuromuscular adaptations by manipulating acute exercise variables, including exercise selection, load lifted and the number of sets and repetitions (Bird et al., 2005). While this prescription technique can increase muscular power, strength and hypertrophy, it does not account for an individual’s readiness to exercise on a given day, which is known to fluctuate due to fatigue or supercompensation from previous training (Jovanovic & Flanagan, 2014; Hughes et al., 2019b). Without knowing if the planned exercise intensity or volume is suited for an individual on a training day, the parameters prescribed may not result in the desired stimulus, leading to suboptimal adaptations (Jovanovic & Flanagan, 2014). For example, participants may unintentionally train to failure which can exaggerate post-exercise neuromuscular fatigue and muscle damage (Gonzalez-Hernandez et al., 2021) and increase the time needed for recovery between training sessions (Morán-Navarro et al., 2017).

An emerging alternative to traditional training prescription is the concept of autoregulation, which can be defined as self-management of training variables (i.e. load or volume) based on the individual’s current performance capabilities to meet session goals (Kraemer & Ratamess, 2004). One of the simplest approaches to autoregulation is the subjective repetitions in reserve (RIR) method, whereby individuals estimate how many more repetitions they can perform before failure. Using this method, they can then regulate their training based on perception of proximity to failure within a set (Hackett et al., 2012). For example, an individual may finish their set when they feel they can only perform two more repetitions (i.e. 2-RIR). While some research has reported that individuals can accurately estimate RIR when they are close to failure in a set (e.g. 2-RIR) (Hackett et al., 2018; Zourdos et al., 2021), there is contrasting evidence that RIR is often under-predicted (Fisher et al., 2017), meaning that they could have completed more repetitions than they perceive. The validity of perceived RIR also decreases when attempting to quantify higher RIR values (Hackett et al., 2017), which limits the ability to use this autoregulation strategy when attempting to manage fatigue by training sub-maximally (i.e. 2- or 3-RIR).

To counter these limitations in subjective RIR estimates, objective approaches based on measurement of concentric repetition velocity have been developed (Pelland et al., 2022). One method gaining popularity is defining individualised velocity stops for specific RIR values (Morán-Navarro et al., 2019; Pelland et al., 2022). Bar velocity corresponding to a certain RIR is reliable in a single set to failure, for upper and lower body Smith-machine exercises with loads between 65-85% 1-repetition maximum (1RM) (Morán-Navarro et al., 2019). This means that if an athlete wished to stop the set with two repetitions before failure (2-RIR), they would perform repetitions until reaching the predetermined velocity associated with 2-RIR (Pelland et al., 2022). As it is unlikely that the same exact velocity would be achieved for a given RIR value across multiple sets, Pelland et al. (2022) have recently proposed implementing velocity stops as a range to accommodate this variability. This range can be determined by assessing the velocity of each repetition during a single testing set to failure for a specific exercise; the velocity range for 0-RIR (i.e. last completed repetition) would be between the velocity of 0-RIR and 1-RIR, whereas the 1-RIR range would be the velocities between 1-RIR and 2-RIR and so on (Pelland et al., 2022). To our knowledge, the accuracy of this approach has not been assessed in published research.

It is also unknown whether velocities associated with certain RIR values remain consistent across multiple sets. It is possible that fatigue across multiple sets will affect the velocity of specific RIR values, considering that research has observed a decline in the average set velocity across three sets in Smith-machine bench press and squat exercises (Sánchez-Medina & González-Badillo, 2011). While this research provides some evidence on the effect of fatigue on repetition velocity, the specific velocities for each repetition, and whether they changed across sets, were not reported (Sánchez-Medina & González-Badillo, 2011). Furthermore, many of these previous studies have used Smith-machine exercise, whereas athletes likely to have access to velocity-monitoring technology typically train using free-weight exercises. It is possible that the greater stability requirements of free-weight exercise could increase the variability of velocity measurements compared to machine-based exercise. Further work is required to identify whether velocity associated with RIR remains consistent with free-weights across a number of sets, and if not, to what extent velocity changes for each RIR value.

Another important factor which requires attention in velocity-based RIR research is the influence of the load being lifted. It has previously been identified that there are no differences between 60 and 80% 1RM loads for subjective estimates of RIR (Mansfield et al., in press). However, velocity in the bench press and back squat decline more rapidly across a set with 50–70% 1RM than for loads of ≥80% 1RM (Rodríguez-Rosell et al., 2020). Given that velocity loss occurs at different rates between relative loads, it is plausible that the velocity associated with certain RIR values may be specific to load being lifted. This would mean that the velocity of the targeted RIR value would need to be determined for each load being used in a session, which would limit the viability of velocity-based methods in practice. Considering these current gaps in understanding of velocity-based RIR assessments, the purpose of this study was to investigate the reliability of the velocity associated with repetitions from 3- to 0-RIR in two commonly performed free-weight exercises (bench press and prone row), across four sets, with both moderate (60% 1RM) and high (80% 1RM) loads. A secondary aim was to examine the accuracy of applying velocity stops as a range between the target RIR and the velocity of the previous repetition (Pelland et al., 2022). It was hypothesised that the velocity of repetitions would be less reliable at 3- compared to 0-RIR across multiple sets and between different loads. By extension of this, we also hypothesised that velocity stops could accurately be used to as RIR.

Methods

Participants

Twenty recreationally resistance trained males (age: 25.9 ± 4.5 years, height: 181 ± 7 cm, body mass: 86.5 ± 13.7 kg) volunteered for this study. They reported at least 2 years training experience (≥2 resistance training sessions per week; mean of 6.0 ± 4.5 years), meeting the resistance training guidelines for being classified as “recreationally active” (McKay et al., 2022). Participants completed a pre-exercise screening questionnaire to confirm they were free from any musculoskeletal injuries that would impact on the results of the study. Prior to commencing, all participants were provided with information detailing the requirements and aims of the research and provided signed informed consent. This study and its methods were approved by the Institutional Human Research Ethics Committee (2018/051).

Experimental design

Participants reported to the laboratory on three separate occasions with 3–7 days in-between visits. The initial session served to determine bench press and prone row 1RM. This was followed by two experimental sessions, the order of which was randomised and counterbalanced. During these sessions, participants performed four sets to failure for bench press followed by prone row, using weights at 60% 1RM in one session and 80% 1RM in the other. The mean concentric velocity of each repetition was recorded in both exercises using a previously validated linear position transducer sampling at up to 50 Hz (GymAware Powertool; Kinetic Performance Technology, Canberra, Australia) (Mitter et al., 2021). This approach allowed us to identify the concentric velocity associated with each repetition prior to failure (e.g. 6- to 0-RIR) across multiple sets, using 60 and 80% 1RM loads in free-weight exercises. For the primary analysis (reliability and analysis of variance), the range of 3- to 0-RIR was used, as 42.5% of participants could not complete more than four repetitions in the final sets for the 80% 1RM trials. For secondary descriptive analyses (i.e. counts and proportions), repetitions up to 5-RIR were examined. This study was completed as part of a larger project also investigating subjective estimates of RIR (Mansfield et al., 2020); however, the primary outcome measures (velocity-based RIR) of the current study do not overlap with previous analyses.

Procedures

One-repetition maximum testing and familiarisation

The first visit to the laboratory was used to assess 1RM for the bench press and prone row, as per recommended guidelines (Sheppard & Triplett, 2015). Participants completed a five-minute warm-up on a rowing ergometer at a self-selected intensity, followed by five minutes of self-selected upper body dynamic stretching, which was recorded and replicated in all future sessions. Participants then completed five warm-up sets for bench press which involved: three repetitions at 20% 1RM (as estimated by participants for this session), 40 and 60% 1RM and one repetition at 80 and 90% 1RM, with 90 s rest between each set. The load was increased by ∼5% (minimum load increments of 1 kg) and participants performed a single repetition before resting three minutes. This process was repeated until participants could not complete a successful repetition or adhere to correct technique. Participants rested five minutes after their final bench press 1RM attempt before repeating the process for the prone row. The 1RM was defined as the heaviest completed repetition and was determined within 3–5 sets for both exercises.

For both exercises, participants were instructed to complete the eccentric portion of each repetition under control and the concentric portion as quickly as possible (Hughes et al., 2019b), without any pause between phases of each repetition. Participant grip width was standardised for both exercises so that the forearms remained as close to vertical as possible throughout the entire range of motion, and this was also recorded during the 1RM testing session (with reference to the bar knurling) to be repeated in subsequent trials. The bench press was performed on a flat bench with a standard Olympic bar (20 kg; Australian Barbell Company, Victoria, AUS). Each bench press repetition commenced with elbows extended, before the bar was lowered to contact the chest and immediately returned to full elbow extension, with no bouncing of the bar on the chest, or leg or torso drive, allowed. The prone row was performed using a specialised bench (Valkyrie prone row bench; Aussie Strength, New South Wales, AUS) and bar (15.6 kg fixed grip bench pull barbell, AlphaFit, Queensland, AUS), starting from full elbow extension before pulling the bar towards their body and returning it to the starting position (Sánchez-Medina et al., 2014). A customised elastic string-line was set at the height of the bar when participants performed full scapular retraction, and the bar was required to contact the string-line at the top of each repetition to signify that appropriate range had been achieved (Scott et al., 2014). All participants used lifting straps (VELO Weight Lifting Wrist Wraps, VELO) for the prone row, to ensure that grip strength did not limit exercise performance. A 10 kg bar was used for light warm-up sets when required (Pro Series barbell, 360 Strength, Victoria, AUS). Exercise performance was visually observed by a researcher, and repetitions which did not meet the guidelines indicated above were deemed incomplete.

Experimental exercise protocols

During the second and third sessions, participants completed four sets of bench press and prone row to failure using either 60% 1RM or 80% 1RM. After completing the same general warm-up as per session one, participants performed a specific bench press warm-up according to the prescribed load for the session (60% 1RM protocol: 10 repetitions with 20% 1RM, 10 with 40% 1RM and 5 with 50% 1RM; 80% 1RM protocol: 10 with 20% 1RM, 5 with 50% 1RM and 3 with 70% 1RM). Following the warm-up, participants rested for three minutes, before completing four sets of bench press to failure, with three minutes of rest between sets. Upon completion of the bench press sets, participants rested for five minutes before completing the same warm-up and protocol for the prone row. All sessions were performed with the same exercise order, as we observed during pilot testing that when the prone row was performed first, participants often could not perform the required number of repetitions needed to analyse up to 3-RIR in the 80% 1RM bench press protocol. These two exercises were selected as they are commonly performed by athletic populations and allowed us to assess both a horizontal push and pull movement in the upper body. Participants were instructed at the start of each session to perform the concentric phase of each repetition as fast as possible, and strong verbal encouragement was provided throughout each set to motivate them to adhere to these instructions. No augmented feedback (i.e. information on the concentric velocity of repetitions) was provided during experimental sessions.

Velocity measurement and analysis

During experimental trials, a linear position transducer recorded time and displacement data of the bar for each repetition. The mean velocity was calculated for the concentric phase of each repetition, which was automatically identified by the linear position transducer. The velocity of 3- to 0-RIR was then determined for each set to examine whether these values changed across multiple sets during each trial, and if they were affected by the load lifted, via the primary analysis. The mean difference in velocity between 4- to 0-RIR was also calculated for each participant across all sets. These data are represented via scatterplots to illustrate the variance in between-repetition velocity changes, and the number of instances where velocity did not decline between successive repetitions as would be theoretically expected.

The velocity range for each repetition from 5- to 0-RIR was also determined from the first set of each exercise at both loads, following the method described by Pelland et al. (2022). This analysis was conducted for up to 5-RIR, rather than just 3-RIR, to also show the proportion of participants who did not complete enough repetitions for higher RIR values. For example, if the velocities from the first set for 1- and 2-RIR were 0.24 and 0.28 m·s⁻¹, respectively, then the velocity range for 1-RIR would be 0.24–0.27 m·s⁻¹; a repetition in subsequent sets completed within this velocity range (e.g. 0.26 m·s⁻¹) would then be assessed as the participant having 1-RIR. To provide a practical example of how this approach to moderating exercise volume would be achieved, the proportion of participants who achieved their target velocity for each RIR value over sets 2–4 was calculated, as well as the proportion who undershot (i.e. achieved target velocity before corresponding RIR value and would have concluded a set prematurely) or overshot (i.e. achieved target velocity after corresponding RIR value and performed more repetitions than desired).

Statistical analysis

The mean ± SD and 95% confidence intervals (CIs) were calculated for all variables. A three-way ANOVA with repeated measures was used to assess the differences in velocity between the two loads, four sets and four RIR values, separately for both exercises. When a significant main effect or interaction was observed, post hoc Bonferroni adjusted pairwise comparisons were used to determine where these differences occurred. Distributions of error residuals from the repeated measures ANOVA did not indicate any marked breaches to the assumption of normality. Mauchly’s test of sphericity, however, indicated moderate evidence that the assumption of sphericity was violated for the main effect of RIR (Mauchly’s W_df = 5 = 0.380, p = 0.014) and load × set interaction (Mauchly’s W_df = 5 = 0.435, p = 0.032) for the bench press analysis and so Greenhouse-Geisser corrections were applied to the degrees of freedom for these effects. To examine the reliability of the velocity associated with 3- to 0-RIR between sets, difference in the mean and limits of agreement (1.96 × SD of difference scores) were determined. The intraclass correlation coefficients (ICC_3,1) and coefficients of variation (CV) with 95% confidence intervals were also calculated for the velocity of 3- to 0-RIR between set 1 and sets 2–4 (Hopkins, 2002). Significance was set with a type-I error rate of α ≤ 0.05. Analyses were performed using SPSS software (v.23, IBM, New York, USA).

Results

Participant strength levels were as follows: 1RM bench press of 98.4 ± 16.4 kg [1.1 ± 0.1 × body mass]; 1RM prone row or 72.0 ± 11.7 kg [0.8 ± 0.1 × body mass].

The average velocities associated with 3- to 0-RIR for both loads (60 and 80% 1RM) and exercises are presented in Figure 1. For mean concentric velocity in the bench press, there were significant main effects for both load (F_1,16 = 19.66, p < 0.001, η_p² = 0.551) and RIR value (F_1.9,30.9 = 194.84, p < 0.001, η_p² = 0.924), but not for set (F_3,48 = 0.75, p = 0.530, η_p² = 0.045). A significant load ×RIR value interaction was also observed (F_3,48 = 4.27, p = 0.009, η_p² = 0.210), though there was no other two-way (load × set: F_2.2,34.5 = 1.09, p = 0.350, η_p² = 0.064; set × RIR value: F_9,144 = 0.72, p = 0.689, η_p² = 0.043) or three-way (load × set × RIR value: F_9,144 = 0.48, p = 0.889, η_p² = 0.029) interaction. Post hoc analysis for the load × RIR value interaction indicated that velocity was faster when lifting 60% 1RM compared to the 80% 1RM trial for the bench press at RIR values from 3- to 0-RIR (mean difference in velocity between loads = 0.050 m·s⁻¹ [CI: 0.026, 0.074], p ≤ 0.008). Post hoc analysis also identified that for both loads, the velocity associated with 3-RIR was the fastest, followed by 2-, 1- and 0-RIR, with significant differences between each repetition (p ≤ 0.001).

Figure 1. Average values of mean concentric velocity for each repetition from 3- to 0-repetitons in reserve (RIR) for bench press and prone row at 60% and 80% 1-repetition maximum. Data are mean ± SD. * Different from set 1, ^# Different from set 2, ‡ Different between each RIR value.

For mean velocity in the prone row, there were significant main effects for load (F_1,15 = 61.61, p < 0.001, η_p² = 0.804), set (F_3,45 = 7.34, p < 0.001, η_p² = 0.329) and RIR value (F_3,45 = 56.14, p < 0.001, η_p² = 0.789). There were no significant two-way (load × set: F_3,45 = 0.97, p = 0.416, η_p² = 0.061; set × RIR value: F_9,135 = 0.963, p = 0.474, η_p² = 0.060; load × RIR value: F_3,45 = 0.59, p = 0.626, η_p² = 0.038) or three-way (load × set × RIR value: F_9,135 = 1.07, p = 0.388, η_p² = 0.067) interaction effects for the prone row. Similarly to the bench press, post hoc analysis indicated that velocity was faster when lifting 60% 1RM compared to the 80% 1RM trial in the prone row (mean difference in velocity between loads = 0.058 m·s⁻¹ [CI: 0.042, 0.074], p < 0.001), and that the velocity of 3-RIR was the fastest, followed by 2-, 1- and 0-RIR, with significant differences between each repetition (p ≤ 0.010). Additionally, post hoc tests indicated significant differences in mean velocity between sets 1 and 3 (0.019 m·s⁻¹ [CI: 0.003, 0.035], p = 0.013), 1 and 4 (0.027 m·s⁻¹ [CI: 0.009, 0.045], p = 0.003) and 2 and 4 (0.019 m·s⁻¹ [CI: 0.001, 0.037], p = 0.032).

The ICC and CV values are shown in Table 1. Overall, the velocity measures corresponding with each RIR value were less reliable in the bench press (mean ICC = 0.40 [range: −0.04–0.73], mean CV = 21.3% [range: 13.5–41.7%]) than in the prone row (mean ICC = 0.80 [range: 0.52–0.94], mean CV = 6.1% [range: 3.7–10.2%]). There were no consistent differences in reliability between loads, sets, or RIR values.

Table 1. Mean (± SD), limits of agreement, intraclass correlation coefficient (ICC; 95% CI) and coefficient of variation (CV; 95% CI) for between set comparisons of velocity at different repetition-in-reserve (RIR) values for 60% and 80% 1RM loads in the bench press and prone row.

Set Comparison	RIR Value	Velocity Difference	Limits of Agreement	ICC (95% CI)	CV (95% CI)
Bench Press 60% 1RM Protocol
1–2	0	0.00 ± 0.07	−0.14 to 0.14	0.51 (0.10–0.77)	28.2% (20.8–43.7%)
	1	0.00 ± 0.07	−0.14 to 0.14	0.43 (0.00–0.73)	21.4% (15.9–32.7%)
	2	0.02 ± 0.08	−0.14 to 0.18	0.40 (−0.04–0.71)	17.1% (12.7–25.9%)
	3	0.03 ± 0.08	−0.13 to 0.19	0.49 (0.07–0.76)	15.4% (11.5–23.3%)
1–3	0	0.00 ± 0.09	−0.18 to 0.18	0.31 (−0.14–0.66)	32.0% (23.5–50.0%)
	1	0.00 ± 0.07	−0.14 to 0.14	0.32 (−0.13–0.66)	19.9% (14.8–30.3%)
	2	0.03 ± 0.07	−0.11 to 0.17	0.62 (0.25–0.83)	17.3% (12.9–26.2%)
	3	0.02 ± 0.12	−0.22 to 0.26	−0.04 (−0.47–0.40)	24.2% (17.9–37.3%)
1–4	0	−0.01 ± 0.11	−0.23 to 0.21	0.05 (−0.39–0.47)	41.7% (30.3–66.4%)
	1	−0.02 ± 0.09	−0.20 to 0.16	0.51 (0.10–0.77)	21.9% (16.3–33.6%)
	2	−0.01 ± 0.10	−0.21 to 0.19	0.46 (0.04–0.75)	22.1% (16.4–33.9%)
	3	0.00 ± 0.13	−0.25 to 0.25	0.25 (−0.21–0.62)	25.3% (18.7–39.0%)
Bench Press 80% 1RM Protocol
1–2	0	0.00 ± 0.06	−0.12 to 0.12	0.22 (−0.23–0.60)	23.9% (17.7–36.8%)
	1	0.00 ± 0.06	−0.12 to 0.12	0.42 (−0.01–0.72)	20.1% (15.0–30.7%)
	2	0.00 ± 0.06	−0.12 to 0.12	0.48 (0.06–0.76)	16.3% (12.2–24.7%)
	3	−0.02 ± 0.07	−0.16 to 0.12	0.36 (−0.08–0.69)	17.2% (12.8–26.1%)
1–3	0	0.01 ± 0.05	−0.09 to 0.11	0.31 (−0.14–0.66)	27.8% (20.5–43.1%)
	1	0.01 ± 0.06	−0.11 to 0.13	0.47 (0.05–0.75)	20.5% (15.3–31.4%)
	2	0.00 ± 0.07	−0.14 to 0.14	0.17 (−0.28–0.56)	20.4% (15.2–31.2%)
	3	−0.01 ± 0.06	−0.13 to 0.11	0.73 (0.40–0.89)	13.5% (10.0–20.9%)
1–4	0	0.02 ± 0.03	−0.04 to 0.08	0.73 (0.44–0.89)	14.6% (10.9–22.0%)
	1	0.01 ± 0.06	−0.11 to 0.13	0.47 (0.04–0.75)	19.3% (14.3–29.3%)
	2	0.01 ± 0.05	−0.09 to 0.11	0.59 (0.19–0.82)	15.9% (11.8–24.3%)
	3	−0.01 ± 0.05	−0.11 to 0.09	0.42 (−0.06–0.74)	15.4% (11.3–23.9%)
Prone Row 60% 1RM Protocol
1–2	0	−0.02 ± 0.09	−0.20 ± 0.16	0.52 (0.11–0.78)	10.2% (7.6–15.2)
	1	−0.02 ± 0.07	−0.16 ± 0.12	0.74 (0.44–0.89)	6.6% (5.0–9.8%)
	2	0.00 ± 0.07	−0.14 ± 0.14	0.84 (0.63–0.93)	6.8% (5.1–10.1%)
	3	−0.03 ± 0.08	−0.19 ± 0.13	0.82 (0.60–0.93)	6.8% (5.2–10.2%)
1–3	0	−0.04 ± 0.08	−0.20 ± 0.12	0.69 (0.37–0.86)	8.2% (6.2–12.1%)
	1	−0.03 ± 0.04	−0.11 ± 0.05	0.89 (0.75–0.96)	4.3% (3.2–6.3%)
	2	−0.03 ± 0.06	−0.15 ± 0.09	0.86 (0.67–0.94)	5.6% (4.2–8.4%)
	3	−0.02 ± 0.07	−0.16 ± 0.12	0.79 (0.53–0.91)	6.9% (5.1–10.3%)
1–4	0	−0.06 ± 0.09	−0.24 ± 0.12	0.49 (0.08–0.76)	10.0% (7.5–14.9%)
	1	−0.03 ± 0.07	−0.17 ± 0.11	0.79 (0.53–0.91)	6.7% (5.0–9.9%)
	2	−0.04 ± 0.08	−0.20 ± 0.12	0.72 (0.42–0.88)	7.8% (5.9–11.6%)
	3	−0.06 ± 0.06	−0.18 ± 0.06	0.80 (0.55–0.91)	5.9% (4.5–8.8%)
Prone Row 80% 1RM Protocol
1–2	0	0.01 ± 0.05	−0.09 ± 0.11	0.82 (0.61–0.93)	5.6% (4.2–8.2%)
	1	0.00 ± 0.05	−0.10 ± 0.10	0.79 (0.54–0.91)	6.1% (4.6–9.0%)
	2	−0.01 ± 0.05	−0.11 ± 0.09	0.81 (0.56–0.92)	5.0% (3.7–7.6%)
	3	−0.01 ± 0.05	−0.11 ± 0.09	0.80 (0.54–0.92)	5.0% (3.7–7.6%)
1–3	0	0.01 ± 0.06	−0.11 ± 0.13	0.80 (0.57–0.92)	6.8% (5.1–10.1%)
	1	−0.01 ± 0.04	−0.09 ± 0.07	0.94 (0.84–0.97)	3.7% (2.8–5.4%)
	2	−0.01 ± 0.05	−0.11 ± 0.09	0.81 (0.56–0.92)	5.4% (4.0–8.0%)
	3	−0.02 ± 0.05	−0.12 ± 0.08	0.85 (0.63–0.94)	5.2% (3.9–8.0%)
1– 4	0	−0.01 ± 0.05	−0.11 ± 0.09	0.89 (0.75–0.96)	5.0% (3.8–7.4%)
	1	−0.02 ± 0.05	−0.12 ± 0.08	0.88 (0.72–0.95)	4.7% (3.6–7.0%)
	2	−0.01 ± 0.04	−0.09 ± 0.07	0.92 (0.80–0.97)	3.8% (2.9–5.7%)
	3	−0.03 ± 0.05	−0.13 ± 0.07	0.88 (0.69–0.95)	4.1% (3.0–6.3%)

Figure 2 shows the individual differences in velocity between consecutive repetitions up to 0-RIR. Each successive repetition should be theoretically slower than the previous, and so the difference should be negative. Positive data points therefore indicate instances when a repetition was performed faster than the previous one. For the bench press, there were 56 (out of 319; 17.6%) instances where a repetition was faster than the previous one in the 60% session, and 32 (out of 298; 10.7%) instances for the 80% session. For the prone row, there were 104 (out of 317; 32.8%) instances in the 60% session, and 65 (out of 293; 22.2%) instances for the 80% session. A summary of these data is also presented as an appendix, grouped by exercise, load and RIR. Across both exercises and loads, there were fewer instances where a repetition was faster than the previous one at lower RIR values (e.g. 48 out of 320 [15.0%] for 0-RIR – 1-RIR) than higher RIR values (e.g. 70 out of 284 [24.6%] for 3-RIR–4-RIR).

Figure 2. Individual differences in mean concentric velocity between consecutive repetitions up to 0-repetitions in reserve (RIR). Dotted line indicates a difference of 0.0 m·s⁻¹, with data points above this therefore representing repetitions which were faster (rather than slower) than the previous one. Shaded area represents 95% confidence interval.

Figure 3 shows the proportion of participants whose bar velocity was within the velocity stop range for each RIR value over sets 2–4 as per Pelland et al. (2022), and so would have correctly ceased the set at the correct RIR value if autoregulating training by this method. The proportion who undershot or overshot their target velocity is also represented. These figures represent greater proportions of correct values for the 0-RIR data, though it should be highlighted that for this repetition it is not possible to overshoot as it was the final repetition performed in each set. The analysis showed that the RIR value was undershot (32.9% of cases) or overshot (36.2%) more often than it was correctly achieved (30.9%).

Figure 3. The proportion of participants who achieved their target mean concentric velocity for each repetition in reserve (RIR) value over sets 2–4, based on the mean concentric velocities achieved in the corresponding repetitions in set 1. Undershoot indicates the target velocity was achieved before the corresponding RIR value (i.e. would have concluded a set prematurely), and overshot indicates the target velocity was achieved after corresponding RIR value (i.e. would have performed more repetitions than desired). These velocity stops were determined as per Pelland et al. (2022).

Discussion

This study examined the efficacy of using repetition velocity to estimate RIR values across multiple sets and with different loads in the bench press and prone row exercises. The main findings suggest that; (1) bar velocity is slowed between subsequent repetitions from 3- to 0-RIR, (2) the bar velocity of 3- to 0-RIR was faster for lighter loads and varied across the four sets for the prone row, but not bench press, (3) the velocity associated with each RIR value was more reliable across multiple sets for the prone row than the bench press, (4) change in velocity between consecutive repetitions varied considerably at the individual level and (5) estimated velocity stops for each RIR value were not consistent with the actual velocities measured during trials for either exercise or load. While mean concentric velocity does decline with each repetition across multiple sets at the group level, the findings contrast our hypothesis and suggest that using velocity stops based on a single set is not an appropriate method to estimate RIR in trained individuals.

The findings from this study indicate that within the 3- to 0-RIR range, mean concentric velocity was different between consecutive repetitions in both the free-weight bench press and prone row. These findings align with previous research that has observed a linear decrease in repetition velocity across a set of bench press to failure performed in a Smith-machine (Izquierdo et al., 2006; Rodríguez-Rosell et al., 2020; Sánchez-Medina & González-Badillo, 2011), and indeed form the theoretical basis for measuring repetition velocity to distinguish between successive RIR values. However, an interesting result was that for both exercises, average velocity across 3- to 0-RIR was faster for the 60% compared to the 80% 1RM load. This contrasts some previous work which reported no differences in velocity of specific RIR values between 65-85% 1RM for the Smith-machine bench press, squat, prone row and shoulder press (Morán-Navarro et al., 2019). A possible explanation is that participants in the current study performed free-weight exercise which allows for mediolateral and anteroposterior movement that is not possible with a Smith-machine. Research has shown that fatigue during a set of free-weight bench press results in altered lifting kinematics (Duffey & Challis, 2007), and heavier loads alter the recruitment patterns of agonist and synergist muscles compared to lighter loads (Król & Gołaś, 2017). As linear position transducers can only measure linear displacement and velocity, it is not surprising that more variation was observed in mean concentric velocity values in this study compared to previous work. The results of this research therefore indicate that while mean concentric velocity can differentiate between repetitions in the bench press and prone row, the targeted RIR velocity value is likely to vary between different loads, which limits the practicality of this autoregulation method for these free-weight exercise.

Differences were not observed in velocity between sets for the bench press. This corroborates recent data from Hackett (2022), who reported that the minimal mean concentric velocity during sets to failure (which would typically occur in the final repetition) did not differ between five sets of bench press. Nevertheless, the results from this study suggest that the mean concentric velocity associated with RIR values was not reliable across multiple sets for the bench press (Table 1). In addition to this, differences in concentric velocity between sets for the prone row exercise were observed. These results collectively indicate that the velocity associated with specific RIR values may differ, or exhibit unacceptable reliability, across multiple sets for the exercises examined, limiting the use of these data for autoregulating training. Variation in the reliability of velocity between the bench press and prone row may be due to variances in how the sticking point impacts on force production between these exercises. Sánchez-Medina et al. (2014) have previously reported different kinetics between these two exercises which they partly explained by changes in muscle moment arms, and therefore mechanical advantages, at specific joint angles throughout a repetition. Indeed, performance in the concentric phase of the bench press (Elliott et al., 1989), but not the prone row (Sánchez-Medina et al., 2014), seem to be hampered by this sticking point. While further research to this point is needed, it is possible that exercise-specific kinetics may influence the reliability of velocity-based RIR estimates over multiple sets.

Considering that autoregulation methods aim to inform personalised resistance training, it is important to consider not only group-level analyses, but also how these methods apply to individual participants. As shown in Figure 2, there were 257 out of 1227 repetitions (20.9%) captured between 4- to 0-RIR which were faster than the previous repetition. While researchers generally agree that concentric repetition velocity declines progressively across a set due to accumulating neuromuscular fatigue (Hughes et al., 2020; Weakley et al., 2021), and group-level analyses demonstrate significant reductions between each repetition from 3- to 0-RIR (Figure 1), this was not consistently observed for individual participants. The proportion of repetitions which increased in concentric velocity from the previous one also seemed to vary between loads, exercises and across RIR values (Figure 2 and Appendix). In addition to this, most participants did not achieve their targeted velocity across 5- to 0-RIR (Figure 3). This means that in most instances they would have either performed fewer or more repetitions that planned, with a higher proportion of undershoot or overshoot for the prone row than for the bench press. This observed variability in resistance exercise performance across multiple sets likely renders the velocity of individual repetitions too variable to implement absolute velocity stops. Based on these results, the current study does not support the use of velocity stops to autoregulate free-weight bench press and prone row resistance training. Nevertheless, it should be acknowledged that other approaches to prescribing or monitoring resistance training via measurements of velocity (e.g. velocity loss thresholds) have gained considerable attention from researchers and practitioners (Weakley et al., 2021), with demonstrated efficacy for improving neuromuscular performance (Held et al., 2022). Using velocity-based approaches in combination with subjective assessments of RIR (Balsalobre-Fernandez et al., 2021) may be useful for practitioners to implement autoregulated resistance training programmes. While both velocity stop and perceived RIR approaches can be incorporated into a training programme, practitioners should be aware that these techniques cannot be used interchangeably.

While this study provides novel findings on the use of velocity to determine RIR values for different loads across multiple sets, it is not without limitations. The reliability of velocity associated with different RIR was only examined in a single session; it remains unknown whether the same reliability would be identified across several consecutive sessions. Fatigue accumulated from previous sessions could affect the reliability of velocity associated with different RIR, as participants may experience intra-set fatigue at a faster rate due to an acute decline in sub-maximal strength (Hughes et al., 2019a). Furthermore, the findings of this research can only be applied to specific upper body exercises. Particularly considering the stark differences in the reliability of the velocity associated with different RIR values between the bench press and prone row, these findings may not generalise to other exercises. The findings from this study relate to recreationally active (McKay et al., 2022) participants, and may not translate to more highly trained populations, particularly considering that the relative bench press strength levels observed would be classified as “fair” by normative data (ACSM, 2022). The reliability of velocity measures obtained during resistance exercise may be lower when performed as “touch-and-go” compared to with a pause between eccentric and concentric phases (Pallares et al., 2014). While participants in real-world training environments are likely to perform “touch-and-go” resistance exercise as in this study, future research should examine whether the current findings are replicated if paused repetitions are used. Finally, it is difficult to confirm that participants performed each repetition with maximal effort. While strong verbal encouragement was provided to motivate participants, alternative approaches to motivating performance via providing augmented feedback (Nagata et al., 2020) may be useful for practitioners.

Conclusion

While group level analysis showed that velocity significantly declined in each set up to 0-RIR, analysis of individual participants showed that this was not the case in 20.9% of instances. Furthermore, individuals did not perform repetitions within the velocity stop ranges for each specific RIR value in most cases. The velocities associated with 3- to 0-RIR were faster for 60% 1RM than for 80% 1RM, and varied across multiple sets for the prone row and the bench press exercises. Taken together, these findings suggest that the velocity of specific RIR values varies considerably between exercises, loads and across multiple sets for individuals. As velocity-based RIR estimates do not appropriately align with actual RIR values, alternative methods to velocity stops should be considered for autoregulating of resistance training.

Acknowledgements

The authors would like to thank members of the MASS Laboratory group who assisted with data collection.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Author BRS is supported by a National Health and Medical Research Council Investigator Grant (APP1196462).

Appendix

Proportion (and number) of repetitions which were faster than the preceding repetition across the bench press and prone row exercises, and with 60% and 80% of 1-repetition maximum (1RM) loads.

Table

Exercise (load)	3-RIR–4-RIR	2-RIR–3-RIR	1-RIR–2-RIR	0-RIR–1-RIR
Bench press (60% 1RM)	19.0% (15/79)	20.0% (16/80)	17.5% (14/80)	13.8% (11/80)
Bench press (80% 1RM)	12.7% (8/63)	17.1% (13/76)	8.9% (7/79)	5.0% (4/80)
Prone row (60% 1RM)	41.8% (33/79)	30.4% (24/79)	32.9% (26/79)	26.3% (21/80)
Prone row (80% 1RM)	22.2% (14/63)	25.0% (18/72)	26.9% (21/78)	15.0% (12/80)
All	24.6% (70/284)	23.1% (71/307)	21.5% (68/316)	15.0% (48/320)