Proximity to Failure and Total Repetitions Performed in a Set Influences Accuracy of Intraset Repetitions in Reserve-Based Rating of Perceived Exertion

Abstract

Zourdos, MC, Goldsmith, JA, Helms, ER, Trepeck, C, Halle, JL, Mendez, KM, Cooke, DM, Haischer, MH, Sousa, CA, Klemp, A, and Byrnes, RK. Proximity to failure and total repetitions performed in a set influences accuracy of intraset repetitions in reserve-based rating of perceived exertion. J Strength Cond Res 35(2S): S158–S165, 2021—The aim of this study was to assess the accuracy of predicting repetitions in reserve (RIR) intraset using the RIR-based rating of perceived exertion (RPE) scale. Twenty-five men (age: 25.3 ± 3.3 years, body mass: 89.0 ± 14.7 kg, height: 174.69 ± 6.7 cm, and training age: 4.7 ± 3.2 years) reported to the laboratory. Subjects performed a 1 repetition maximum (1RM) squat followed by one set to failure at 70% of 1RM. During the 70% set, subjects verbally indicated when they believed they were at a 5RPE (5RIR), 7RPE (3RIR), or 9RPE (1RIR), and then continued to failure. The difference between actual repetitions performed and participant-predicted repetitions was calculated as the RIR difference (RIRDIFF). The average load used for the 70% set was 123.10 ± 24.25 kg and the average repetitions performed were 16 ± 4. The RIRDIFF was lower (RPEs were more accurate) closer to failure (RIRDIFF at 9RPE = 2.05 ± 1.73; RIRDIFF at 7RPE = 3.65 ± 2.46; and RIRDIFF at 5RPE = 5.15 ± 2.92 repetitions). There were significant relationships between total repetitions performed and RIRDIFF at 5RPE (r = 0.65, p = 0.001) and 7RPE (r = 0.56, p = 0.004), but not at 9RPE (r = 0.01, p = 0.97). Thus, being farther from failure and performing more repetitions in a set were associated with more inaccurate predictions. Furthermore, a multiple linear regression revealed that more repetitions performed per set was a significant predictor of RIR prediction inaccuracy at the called 5 (p = 0.003) and 7 (p = 0.011) RPEs, while training age (p > 0.05) was not predictive of rating accuracy. These data indicate RIR predictions are improved during low to moderate repetition sets and when there is close proximity to failure.

Introduction

The amount of repetitions that can be performed at a specific percentage of 1 repetition maximum (1RM) is highly individual (^12,17). Specifically, a recent investigation observed that a range of 6–26 repetitions were performed in trained lifters at 70% of 1RM in the back squat (⁴). Thus, prescribing a predetermined number of repetitions at a percentage of 1RM would likely produce highly variable degrees of muscular stress and fatigue between individuals. For example, prescribing 4 sets of 8 repetitions at 70% of 1RM on a specific exercise could lead to multiple sets to failure for some, whereas others could be far from failure during each set. Importantly, recent evidence has demonstrated failure training to elongate recovery time vs. nonfailure training (¹⁴), which could lead to diminished training frequency and volume. An additional limitation of percentage-based or RM zone load prescription is the inability to account for day-to-day readiness, which could lead to missed training volume and diminished long-term adaptations (¹⁰).

Therefore, to account for the limitations of percentage and RM-based load prescription, the repetitions in reserve (RIR)-based rating of perceived exertion (RPE) scale was developed to allow individuals to estimate the number of RIR at the completion of a set (¹⁹). Thus, the RIR-based RPE scale can be used in lieu of percentage of 1RM to prescribe training load. Specifically, a training program could prescribe 4 sets of 8 repetitions at an 8RPE (2RIR) rather than assigning a percentage (i.e., 70% of 1RM) to the prescribed repetitions. When using this approach, the lifter would select a load that they believe they can perform 10 repetitions, stopping the set at an 8RPE (2RIR). Furthermore, the RIR-based RPE scale can be used to implement autoregulation (¹⁵), thus when RPE scores are too high or low, training load can be adjusted for subsequent sets to meet the RPE goal (¹⁰), which can be beneficial considering day-to-day variation performance. Importantly, RIR-based RPE seems to have more utility at equating for effort across individuals than traditional Borg RPE (¹), as data have demonstrated submaximal ratings on the Borg scale during resistance training even when a set has been taken to failure (⁹).

Indeed, Helms et al. (¹⁰) recently demonstrated that RIR-based RPE load prescription resulted in greater back squat (effect size [ES] = 0.50) and bench press (ES = 0.28) strength over 8 weeks than percentage-based load prescription in well-trained males. Despite these findings, it is important to note that the utility of intrasession RPE is predicated on the accuracy of the rating. It is clear that with close proximity to failure (^7,15,19) and greater training age (¹⁸), the accuracy of RIR is improved (⁸). However, the accuracy of intraset RIR-based RPE has not been assessed at multiple points on the scale during the same set in a compound movement. Thus, to further advance RIR-based RPE programming, additional data are required to determine exactly how far from failure RIR predictions become inaccurate, if prior experience using RIR improves rating accuracy, and if the number of repetitions performed in a set affects accuracy. Although RIR-based RPE training improves with training age, intraset RIR-based RPE accuracy has not been assessed in truly well-trained individuals with a similar training status to the long-term data from Helms et al. (squat 1RM = 141.3 ± 21.2 kg, relative strength = 1.82 times body mass [BM]). Furthermore, the number of repetitions in a set has not been investigated in relation to rating accuracy; however, it is possible that high-repetition sets would be associated with more inaccurate RIR predictions due to the potential presence of metabolic fatigue, which is not present to the same degree during low-repetition sets (²). Importantly, the RIR-based RPE scale has gained significant popularity in both the athletic and fitness communities; thus, elucidating these details will provide practitioners the ability to implement the scale when appropriate and to find alternative methods of autoregulation and gauging proximity to failure in situations when ratings are inaccurate.

Therefore, the primary aim of this study was to examine the accuracy of gauging intraset RIR-based RPE when verbally called by the lifter at a perceived “5,” “7,” and “9” RPE (5, 3, and 1 RIR) before continuing the set to volitional failure in well-trained males. A secondary aim was to examine if total repetitions performed, training age and RPE rating experience affected the accuracy of intraset RIR predictions. We hypothesized that intraset accuracy would improve closer to failure, in that the “9” would be more accurate than the “7,” which would be more accurate than the “5” RPE in gauging RIR. In addition, we hypothesized that both training age and RPE experience would be positively related to accuracy and that higher numbers of repetitions in a set would be associated with more inaccurate RIR predictions.

Methods

Experimental Approach to the Problem

Subjects reported to the laboratory on one single day for data collection. After preliminary paper work, subjects had anthropometrics assessed (height, total BM, and body fat percentage). Next, subjects performed a five-minute dynamic warm-up and then completed a 1RM back squat test following validated procedures (¹⁹). After the 1RM test, a 10-minute rest interval was administered. Subjects then performed one set to volitional failure at 70% of 1RM in the squat. During the 70% set, subjects verbally called when they believed they were at a 5RPE (5RIR), 7RPE (3RIR), and 9RPE (1RIR) using the RPE scale from Zourdos et al. (¹⁹) (Figure 1). Furthermore, subjects were blinded to the load being lifted and the percentage of 1RM being used for the set to volitional failure (not during the 1RM test). To blind subjects to the load, opaque trash bags were positioned over the weight discs on each side of the barbell and subjects waited outside of the laboratory after the 1RM test until 70% of 1RM was loaded onto the barbell and the trash bags were in position.

Figure 1.: Resistance training–specific rating of perceived exertion scale. Adapted from Zourdos et al. (¹⁹) with permission.

Subjects

Twenty-five well-trained men participated in this study and the specific subject characteristics can be seen in Table 1. For inclusion, all subjects must have performed the back squat an average of once per week for at least 2 consecutive years as determined through a Physical Activity Questionnaire. Subjects who had any contraindications to exercise (e.g., heart disease, serious musculoskeletal disorders, etc.), as determined through the Health History Questionnaire, were excluded. In addition, subjects were self-reported to be free of performance-enhancing drugs and were required to refrain from exercise for 48 hours before testing. Florida Atlantic University Review Board approved this investigation and all subjects provided written informed consent before participation within the study.

Table 1. - Subject characteristics.^*

	Age (y)	Height (cm)	Body mass (kg)	Body fat (%)	TA (y)	REX (mo)	1RM:BM	70% load (kg)	TR 70%	Mean ACV at 1RM (m·s−1)
Mean ± SD	25	174.69	88.95	12.94	5	10	1.98	123.10	16	0.27
	3	6.70	14.72	4.50	3	7	2.36	24.25	4	0.07

^*TA = training age; REX = rating of perceived exertion experience; BM = body mass; 1RM:BM = the ratio of 1RM (1 repetition maximum) strength to body mass to represent subjects' relative strength; 70% load = load used for 70% of 1 repetition maximum; TR 70% = total repetitions performed during 70% set; Min = minimum; Max = maximum; ACV = average concentric velocity.

Procedures

Training Age and Repetitions in Reserve–Based Rating of Perceived Exertion Experience

After providing informed written consent, all subjects completed a Physical Activity Questionnaire. On this questionnaire, subjects were asked for how many years they had been consistently engaged in resistance training and for how many months (if any) they had been using the RIR-based RPE scale in training. Those responses were used to determine training age and RPE experience, respectively.

Anthropometrics

Height (cm) was measured to the nearest 0.01 cm using a wall-mounted stadiometer (SECA, Hamburg, Germany.). Total BM (kg) was assessed by a calibrated digital scale (Mettler-Toledo, Columbus, OH, USA) to the nearest 0.01 kg. Body fat percentage (BF%) was estimated using the average sum of 2 measurements of skinfold thickness acquired from 3 sites (abdomen, front thigh, and chest); if any site measurement differed by more than 2 mm, a third measurement was taken. The Jackson and Pollock (¹³) equation was used to compute body fat percentage.

One Repetition Maximum Test

Squat 1RM testing was administered in accordance with previously validated procedures (¹⁹). First, subjects performed 5 repetitions with 20% of their estimated 1RM (obtained from the Physical Activity Questionnaire), followed by 3 repetitions at 50% of estimated 1RM, 2 repetitions at 70%, one repetition at 80%, and one repetition at 90% of 1RM. Following the 90% load, increases in subsequent 1RM attempts were administered at the investigator's discretion. Average concentric velocity (m·s⁻¹), which was measured with the Open Barbell System (Squats & Science Labs LLC, Seattle, WA, USA), and RPE were collected on each 1RM attempt so that investigators could use this information to aid in selecting the next attempt. Five to 7 minutes of rest were allowed between each attempt. A 1RM was considered valid if one of 3 conditions were met: (a) Participant reported a “10” on the RIR scale and the investigator determined a subsequent attempt with increased body mass could not be successfully completed, (b) Participant reported a “9.5” on the RIR/RPE scale and failed the subsequent attempt with a load increase of 2.5 kg or less, and (c) Participant reported a “9” or lower on the RIR/RPE scale and failed the subsequent attempt with a load increase of 5 kg or less. Finally, Eleiko barbells and lifting discs (Chicago, IL, USA), calibrated to the nearest 0.25 kg, were used to ensure accuracy of the load lifted.

Repetitions to Failure and Intraset Rating of Perceived Exertion

After the 10-minute post-1RM testing rest period, subjects performed repetitions until volitional failure at 70% of 1RM while being blinded to the load with opaque trash bags positioned over the weight discs. Volitional failure was determined as the subject either failing on a repetition or both the subject and investigator determining that another repetition could not be completed. To predict RIR during the set to failure, subjects verbally indicated when they believed they had reached a 5RPE (5RIR), 7RPE (3RIR), and 9RPE (1RIR) using the RPE scale from Zourdos et al. (¹⁹) (Figure 1). Both predicted repetitions to failure at each called RPE and actual repetitions performed were recorded. Next, the difference between the actual and predicted repetitions performed (actual repetitions − predicted repetitions) was recorded as the RIR difference (RIRDIFF) for all 3 intraset RPEs. Three different points on the scale were used for intraset RPE calls to achieve RIR predictions at different proximities to failure. We chose a 5RPE as starting point, as recent evidence has shown training at a 5–6 RPE over 8 weeks produced similar hypertrophy to training at a 7–8 RPE (¹⁰), thus performing repetitions at ≥5RPE ensures an effective stimulus to evoke adaptation.

Statistical Analyses

The absolute RIRDIFF (actual repetitions − predicted repetitions) for the set to volitional failure was calculated at all intraset RPEs. For example, if a participant completed 15 total repetitions and called a 5RPE after 7 repetitions (predicting he could do a total of 12 repetitions), then the RIR difference would equal 3. Therefore, a smaller RIRDFF indicates a higher degree of accuracy and vice versa. To determine the accuracy of intraset RIR-based RPEs across several called RPEs, the RIRDIFF was statistically compared between the called 5, 7, and 9 RPEs using a repeated-measures analysis of variance. Paired t-tests were used for multiple comparison purposes. Furthermore, 5 subjects failed to call at least one of the intraset RPEs, thus 20 subjects were included in the repeated-measures analysis. In addition, ES were calculated between each RIRDIFF with the formula: ES = (Mean₁ − Mean₂)/SD_pooled. The magnitude of each ES was interpreted in accordance with Cohen (³). Pearson product-moment correlations were used to assess relationships between the independent variables (total repetitions performed, training age, RPE experience, and chronological age) with the absolute RIRDIFF for each called RPE. Furthermore, to assess the influence of these variables in predicting RIRDIFF, a multiple linear regression was performed to predict dependent variables when at least 2 bivariate correlations were significant or approaching significance (i.e., p < 0.10). All statistical analyses were performed using Statistica (StatSoft, Tulsa, OK, USA) for Windows and SPSS version 25.0 (IBM Corp., Armonk, NY, USA). Significance was set at p ≤ 0.05.

Results

Repetitions in Reserve Difference

The specific values for RIRDIFF at each intraset RPE are displayed in Table 2. There was a significant condition effect (p < 0.001) indicated that the RIRDIFF was lower (i.e., RPEs were more accurate) closer to failure. Specifically, the RIRDIFF at the called 7 was significantly lower (p < 0.001, ES = 0.56) than the 5RPE. The RIRDIFF at the called 9RPE was significantly lower than both the called 7 (p < 0.001, ES = 0.75) and 5 RPEs (p < 0.001, ES = 1.29).

Table 2. - Average RIRDIFF at each intraset called RPE.^*†

	RIRDIFF at called 5RPE	RIRDIFF at called 7RPE	RIRDIFF at called 9RPE
Mean ± SD	5.15 ± 2.92	3.65 ± 2.46‡	2.05 ± 1.73‡§
Closest	0 (1 time)	0 (1 time)	0 (4 times)
Farthest	11 (1 time)	7 (5 times)	6 (1 time)

^*RIRDIFF = rating of perceived exertion difference (actual repetitions − predicted repetitions); RPE = rating of perceived exertion; closest = the closest RIRDIFF and the number of times it occurred; farthest = the farthest RIRDIFF and the number of times it occurred.

^†Data are mean ± SD. N = 20.

^‡RIRDIFF significantly lower than called 5RPE (p < 0.001).

^§RIRDIFF significantly lower than called 7RPE (p < 0.001).

Bivariate Correlations

The specific r and p values for all relationships can be seen in Table 3. The total repetitions performed on the 70% set were 16 ± 4 with a minimum of 9 repetitions and a maximum of 26 repetitions. Total repetitions performed in the 70% of 1RM set were positively and significantly related to RIRDIFF at the called 5 (r = 0.65, p = 0.001) and 7 RPEs (r = 0.56, p = 0.004), but not at the called 9RPE, indicating that more repetitions per set were associated with less accurate RIR prediction. An inverse association between training age and RIRDIFF at the intraset 5 (r = −0.35, p = 0.094) and 7 RPEs (r = −0.34, p = 0.096) approached significance, but did not at the called 9RPE (p = 0.32), indicating that more years of training experience was potentially associated with more precise prediction of RIR when further from failure. Chronological age was inversely and significantly associated with RIRDIFF at the called 9 intraset RPE (r = −0.50, p = 0.021) and approached significance at the called 7RPE (r = −0.36, p = 0.077), but not at the 5RPE (p = 0.20), signifying that older lifters predicted RIR more accurately than younger lifters when closer to failure. Interestingly, there was no significant relationship at any intraset RPE between experience with the RPE scale and RIRDIFF.

Table 3. - Relationships between independent variables and RIRDIFF.^*

	TR RPE5	TR RPE7	TR RPE9	TA RPE5	TA RPE7	TA RPE9	REX RPE5	REX RPE7	REX RPE9	CA RPE5	CA RPE7	CA RPE9
R value	0.65†	0.56†	0.01	−0.35‡	−0.34‡	−0.23	−0.08	0.11	0.16	−0.27	−0.36‡	−0.50†
p value	0.001†	0.004†	0.97	0.094‡	0.096‡	0.32	0.51	0.85	0.49	0.20	0.077‡	0.021†

^*RIRDIFF = rating of perceived exertion difference (actual repetitions − predicted repetitions); TR = total repetitions performed; RPE = rating of perceived exertion; TA = training age; REX = RPE experience; CA = chronological age.

^†Significant relationship.

^‡Relationship approached significance.

Multiple Linear Regression

Based on the bivariate correlations, 2 multiple linear regressions were conducted. Specifically, total repetitions performed and training age were entered as predictors with the RIRDIFF at the called 5RPE as the dependent variable for one regression. For the second regression, total repetitions performed, training age, and chronological age were entered with the RIRDIFF at the called 7RPE as the dependent variable. The R² for the RIRDIFF at the 5RPE model was 0.436, whereas the R² for the RIRDIFF at 7RPE model was 0.377. In each regression, the total repetitions per set was shown to be a significant predictor of RIRDIFF (5RPE: p = 0.003; 7RPE: p = 0.011), whereas training age and chronological age were not significant predictors (p > 0.05). Further details of the regression analyses can be seen in Table 4.

Table 4. - Multiple linear regression results.^*

Predictor	Unstandardized beta ± SE	Standardized beta	T value	p
Results with RIRDIFF at called 5RPE as dependent variable
Total repetitions performed	0.569 ± 0.167	0.591	3.411	0.003†
Training age	−0.140 ± 0.152	−0.159	−0.920	0.368
Results with RIRDIFF at called 7RPE as dependent variable
Total repetitions performed	0.350 ± 0.127	0.513	2.771	0.011†
Training age	0.023 ± 0.189	0.030	0.122	0.904
Chronological age	−0.207 ± 0.178	−0.274	−1.159	0.259

^*RIRDIFF = rating of perceived exertion difference (actual repetitions − predicted repetitions); RPE = rating of perceived exertion.

^†Significant predictor of RIRDIFF (repetitions in reserve difference).

Discussion

The primary aim of this study was to assess the accuracy of intraset RIR-based RPE. Our hypotheses were that RIR prediction would be more accurate closer to failure (supported), more repetitions per set would be related to more inaccurate RIR predictions (supported), and greater training age and RPE experience would be related to increased accuracy of RIR prediction (not supported). Therefore, our main findings were as follows: (a) RIR prediction was more accurate closer to failure in that the 9RPE calls were more accurate than the 7RPE calls, which were more accurate than the 5RPE calls, (b) more repetitions per set were predictive of more inaccurate RIR predictions at the 5 and 7RPE calls, and (c) both training age and RPE experience were not significantly related to RPE accuracy. Overall, these findings suggest that the RIR-based RPE scale is best used when closer to failure during moderate- to low-repetition sets.

The notion that RIR is gauged more accurately closer to failure has been previously reported (^7,15,19). However, this study is the first, to the best of our knowledge, to directly compare 3 different predictions (5RIR, 3RIR, and 1RIR) within the same set. The results indicated a fair amount of inaccuracy when gauging ≥3RIR in a high-repetition, moderate intensity barbell back squat set while blinded to the load. However, the level of accuracy when gauging 7RPE varied between individuals. Indeed, one participant had an RIRDIFF of 0 at 7RPE, indicating perfect accuracy. Therefore, it may be that RPE is useful for some, but not all, to assess proximity to failure. Furthermore, the present data demonstrated fairly accurate RIR prediction at the called 9RPE (RIRDIFF = 1.95), which is similar to Hackett et al. (⁷) who reported an RIR of ±1 when subjects were 0–3 repetitions from failure in the chest press and leg press. One explanation for the slightly less accurate predictions in this study is that subjects were blinded to the load, thus lifters could not have had a predetermined repetition target. This factor is both a limitation and benefit of the present investigation. It serves as a limitation because lifters are rarely blinded to a load under normal training conditions. Conversely, load-blinding may be viewed as a benefit, as subjects may have truly reached a repetition maximum rather than simply meeting a self-determined repetition target if using a known load.

Interestingly, the total repetitions performed per set was a significant predictor of RIRDIFF at the called 5 and 7 RPEs, suggesting that it is more difficult to gauge RIR during high-repetition sets. Existing data suggest it is more difficult to predict RIR when ≥3 repetitions from failure; however, the present data indicate that the amount of repetitions in a set also affects this rating. For example, in a 15–20 repetition set, our data suggest it is difficult to gauge RPE when ≥3 repetitions from failure; however, in a set ≤12 repetitions, a fairly accurate RIR rating can be given when ≥3 repetitions from failure and RIR can be predicted with precision when <3 repetitions from failure. A possible explanation for this phenomenon is that during high-repetition sets, there is a greater degree of metabolic fatigue coupled with neuromuscular fatigue, which may convolute an individual's ability to gauge RIR. Indeed, Buitrago et al. (²) demonstrated greater increases in blood lactate with repetitions to failure at 70 vs. 85% of 1RM. Therefore, it is plausible that significant metabolic fatigue was present in the current study in subjects who performed high repetitions, which in turn hindered their ability to accurately gauge intraset RIR.

Moreover, because high-repetition sets lead to more inaccurate RIR predictions, it is worth noting that the utility of the RIR-based RPE scale may be optimal with higher intensities (≥80% of 1RM), in which typically lower repetitions are performed. If using RPE to monitor proximity to failure or autoregulate during higher repetition sets, perhaps it is best to assign an RPE range (i.e., 5–7 or 6–8) rather than an exact number. In fact, this approach by Helms et al. (¹⁰) led to greater strength adaptations when using RPE-based loading than percentage-based loading over 8 weeks.

Data have shown RIR predictions before a set to improve with training age (¹⁸); however, the current regression analysis did not show training age or RPE experience to be a significant predictor of intraset RIR prediction. Zourdos et al. (¹⁹) and Ormsbee et al. (¹⁵) suggested that experienced lifters recorded more accurate RPEs at 100% of 1RM compared with novice lifters. This claim was made due to experienced lifters having higher RPE at 100% of 1RM than novice lifters. However, another explanation is that novice lifters had poor rate of force development; thus, after recording a submaximal RPE (∼8–9), they simply could not complete their next attempt. In that interpretation, the RPE scores were not necessarily inaccurate, but rather, novice individuals are incapable of performing a true 10RPE lift due to “neuromuscular inefficiency.” In addition, Steele et al. (¹⁸) demonstrated accuracy of RIR prediction improved with training experience; however, in this study, the subjects made a prediction of how many repetitions they could complete before a set rather than during a set. Therefore, in Steele et al.'s study, it is possible that lifters with more training experience were more familiar with how many repetitions they can typically perform at a given load, which is not necessarily indicative of intraset RIR prediction. In addition, this study only employed trained lifters, whereas previous studies either inferring greater RIR prediction accuracy (^7,15,19) or showing greater accuracy (¹⁸), comparing trained lifters with beginners. Therefore, it is possible that training age plays a role in RIR prediction accuracy; however, there may be a point of diminishing returns. Therefore, because all lifters in this study had ≥2 years of training experience, the present comparison of training age was different than that of previous investigations.

Notably, experience with the RIR-based RPE scale did not affect rating accuracy. One participant who reported “0” months of experience with RIR-based RPE was the only individual to have an RIRDIFF of zero (i.e., perfect predictions) at all intraset called RPEs. However, this participant also had the highest training age (12 years) and perhaps more importantly, only performed 10 repetitions during his 70% to failure set. Although this is only one data point, it is consistent with the totality of data in this study and highlights that repetitions per set seems to have a greater contribution to accuracy than prior experience with the RPE scale. Interestingly, the only significant bivariate correlation at the called 9RPE was chronological age to be significantly and inversely related to RIRDIFF at the called 9RPE (r = −0.50, p = 0.021), suggesting that when close to failure, a more mature individual may have a more realistic interpretation of their limitations. This specific finding seems tenuous and further corroboration is needed; however, it seems reasonable that an older lifter might be more aware of his limitations.

A limitation of this study is that it only used male lifters and examining only the squat exercise, thus we cannot extrapolate our findings to female lifters or other compound movements such as the bench press and deadlift. It is also possible that performing 1RM testing before the 70% to failure set could have negatively impacted the amount of repetitions to failure and the subsequent RPE ratings; however, the average repetitions performed of 16 ± 4 is quite high and suggests that performance was not harmed. An additional limitation, as stated previously, is that subjects were blinded to the load, which is rarely the case in practice. Importantly, this study is the first to examine the accuracy of intraset RPE scores in the squat during a multiple repetition set. Furthermore, this study is the first to examine the relationship of total repetitions per set, training age, and RPE experience with the accuracy of ratings in well-trained lifters. Therefore, the novelty of these findings advances RIR-based RPE programming and provides greater application for practitioners to individualize load prescription and adjustment.

In summary, well-trained males gauged intraset RPE more accurately closer to failure at 70% of 1RM in the squat. However, the total repetitions per set significantly impacted the accuracy of intraset RIR prediction, in that when greater repetitions were performed, less accurate RIR predictions were made at all called RPEs (5 and 7 RPEs). Ultimately, our results suggest that well-trained individuals can predict RIR accurately during low-repetition sets and that accuracy improves closer to failure.

Practical Applications

A critical observation of the present data is that a wide range (9–26) and high average (16 ± 4) number of repetitions were performed during the 70% of 1RM set to failure. The highly variable ability to perform repetitions to failure at a moderate intensity in the squat provides further support for use of the RIR-based RPE scale to equate for effort across individuals (¹¹). For example, 4 sets of 10 repetitions at 6–8 RPE could be programmed for the back squat instead of prescribing 10 repetitions at 70% of 1RM. By using RPE, coaches and athletes could ensure that athletes are performing effective repetitions (i.e., ≥5RPE per set) but are not reaching failure consistently. Importantly, training to failure has shown diminished performance for 48 hours (¹⁶) and is associated with longer recovery even when directly compared with volume-equated nonfailure training (¹⁴). Importantly, failure training is not required to optimize strength and hypertrophy (⁵) and because failure training elongates recovery, it could decrease training frequency and subsequent training volume limiting long-term adaptations. Although measuring barbell velocity might be more accurate to predict RIR than the RPE scale (⁶), not all coaches and athletes have access to valid and reliable velocity tools, whereas the RIR-based RPE scale can be used by anyone. Therefore, because velocity measurements are not feasible for everyone, the RIR-based RPE scale provides a viable alternative for intraset RIR prediction, especially in well-trained individuals during low-repetition sets.

Furthermore, athletes can view their load-repetition and RPE relationships from previous training sessions to guide their current and future sessions. Thus, it is possible that in practice, RIR prediction may be even more accurate than is represented in this acute laboratory study. Finally, RIR-based RPE could also be used to gauge progress overtime in lieu of always performing RM testing to mitigate fatigue. For example, if an athlete performs a back squat with 200 kg at a 9RPE before a training block and after the training block performs a 200 kg back squat for one repetition at a 6RPE, it can be concluded the progress was made without the stress of RM testing.