See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/6850661

Prediction of One Repetition Maximum Strength

From Multiple Repetition Maximum Testing and
Anthropometry

ARTICLE in THE JOURNAL OF STRENGTH AND CONDITIONING RESEARCH · SEPTEMBER 2006

Impact Factor: 2.08 · DOI: 10.1519/R-15304.1 · Source: PubMed

CITATIONS READS

38 503

3 AUTHORS, INCLUDING:

Robert Robergs
Charles Sturt University
212 PUBLICATIONS 2,887 CITATIONS

SEE PROFILE

Available from: Robert Robergs

Retrieved on: 15 October 2015
Journal of Strength and Conditioning Research, 2006, 20(3), 584–592
䉷 2006 National Strength & Conditioning Association

PREDICTION OF ONE REPETITION MAXIMUM

STRENGTH FROM MULTIPLE REPETITION MAXIMUM
TESTING AND ANTHROPOMETRY
JEFF M. REYNOLDS, TORYANNO J. GORDON, AND ROBERT A. ROBERGS
Exercise Physiology Laboratories, Exercise Science Program, University of New Mexico, Albuquerque, New Mexico
87131.

ABSTRACT. Reynolds, J.M., T.J. Gordon, and R.A. Robergs. Pre- frame, or with loads inducing fatigue within a specific
diction of 1 repetition maximum strength from multiple repeti- range of repetitions. The range of repetitions used in such
tion maximum testing and anthropometry. J. Strength Cond. testing has been between 2 (6, 7) and 15 (21), with linear
Res. 20(3):584–592. 2006.—The purpose of this study was to
equations apparent for repetitions less than 10 and non-
quantify the decrease in the load lifted from 1 to 5, 10, and 20
repetitions to failure for the flat barbell bench press (chest press; linear equations available for repetition numbers up to
CP) and plate-loaded leg press (LP). Furthermore, we developed 15 (19, 21).
prediction equations for 1 repetition maximum (RM) strength Mayhew et al. (21) developed their nonlinear equation
from the multiple RM tests, including anthropometric data, gen- using data from multiple subjects who lifted a fixed
der, age, and resistance training volume. Seventy subjects (34 weight repeatedly, recording as many correct repetitions
men, 36 women), 18–69 years of age, completed 1, 5, 10, and of the bench press as possible in 1 minute, with a load of
20RM testing for each of the CPs and LPs. Regression analyses 55–95% (selected randomly by computer) of 1RM for each
of mean data revealed a nonlinear decrease in load with increas-
subject. The range of repetitions was based on the be-
ing repetition number (CP: linear Sy.x ⫽ 2.6 kg, nonlinear Sy.x ⫽
0.2 kg; LP: linear Sy.x ⫽ 11.0 kg, nonlinear Sy.x ⫽ 2.6 kg, respec- tween-subject variability and the randomly assigned per-
tively). Multiple regression analyses revealed that the 5RM data centage of 1RM. Conversely, Lombardi (19) proposed a
produced the greatest prediction accuracy, with R2 data for 5, nonlinear equation in his textbook but provided no data,
10, and 20RM conditions being LP: 0.974, 0.933, 0.915; CP: or evidence of data, from which this equation was devel-
0.993, 0.976, and 0.955, respectively. The regression prediction oped. Consequently, there is no experimental evidence for
equations for 1RM strength from 5RM data were LP: 1RM ⫽ a nonlinear relationship between the decrease in load lift-
1.0970 ⫻ (5RM weight [kg]) ⫹ 14.2546, Sy.x ⫽ 16.16 kg, R2 ⫽ ed and repetition number for any given individual.
0.974; CP: 1RM ⫽ 1.1307 ⫻ (5RM weight) ⫹ 0.6999, Sy.x ⫽ 2.98
In addition to repetition number, other factors may
kg, R2 ⫽ 0.993. Dynamic muscular strength (1RM) can be ac-
curately estimated from multiple repetition testing. Data reveal affect the maximum amount of weight an individual can
that no more than 10 repetitions should be used in linear equa- lift. Age, sex, ethnicity, limb lengths and circumferences,
tions to estimate 1RM for the LP and CP actions. body mass, muscle mass, training routine and status, the
rate of contractions, and the time distribution between
KEY WORDS. chest press, leg press, regression, fatigue, training
concentric, eccentric, and recovery phases of a contraction
cycle could all possibly influence the load able to be lifted
INTRODUCTION for a specific number of repetitions.
he use of 1 repetition maximum (RM) testing Given the limited research of 1RM strength prediction

T in resistance training has been applied to quan-

tify strength in order to prescribe training pro-
grams by health and fitness professionals, ath-
letic trainers, rehabilitation specialists, and strength
coaches. Research on 1RM testing began over 50 years
for a variety of weight lifting exercises, and the absence
of guidelines for using linear vs. nonlinear equations for
1RM strength prediction, additional research of 1RM pre-
diction is needed. Furthermore, because of the limited
number of independent variables used in past regression
ago (9), and the use of 1RM testing has become a reliable equations, it is possible that additional variables could
method of strength assessment in trained and untrained improve the accuracy of 1RM strength prediction.
subjects (1, 4–6, 8, 17–23). However, for some popula- The purpose of this study was to (a) examine the re-
tions, age and preexisting medical conditions may be con- lationship of decreases in the load lifted and increases in
traindications to the safe completion of 1RM testing. repetitions to failure, (b) determine if there are gender
Investigators have identified the difficulty in complet- differences in the decrease in loads lifted from 1, 5, 10,
ing 1RM testing on certain populations, and several 1RM and 20RM testing, (c) assess which of the loads lifted dur-
strength prediction equations have been developed. How- ing 5, 10, and 20RM testing most accurately predicts 1RM
ever, some of these equations are only for specific exer- strength, and (d) identify if the addition of anthropome-
cises, such as the leg extension (1) or bench press (1, 4– try, gender, age, and training history data increases the
9, 11, 15, 18–21, 23–26, 30), while others are only for cer- accuracy of 1RM strength prediction in a large, diverse
tain populations, such as college-aged men (8, 21, 22, 24, population.
30). These equations are all based on having subjects lift We chose to study the leg press (LP) and chest press
the greatest load possible for a predetermined number of (CP) actions, as these have been included in past research
repetitions (such as in RM testing), a given load for as and are common actions included in most resistance
many repetitions as possible in a predetermined time training programs.

584
 PREDICTION OF ONE REPETITION MAXIMUM STRENGTH 585

METHODS CP). A nonadjustable Cybex plate-loaded squat press (we

Experimental Approach to the Problem will refer to this as the plate-loaded leg press) was used
for the LP exercise. All equipment was identical in both
Our stated purpose and hypotheses required multiple testing locations.
types of research design, consisting of analysis of variance Two 1-hour testing sessions consisting of 4 maximum
(ANOVA) (influence of repetitions to failure [1, 5, 10, 20] resistance bouts were conducted on each subject. During
on the loads lifted and differences in loads lifted between the first testing session, each subject completed a 20RM
men and women) and multiple regression (prediction of and a 10RM for the LP and CP. Loads were initially es-
1RM strength from multiple RM testing and anthropom- timated based on researcher experience and feedback
etry, gender, age, and training history). from verbal questions pertaining to training history. Sub-
Subjects sequent loads were based on the following estimations ob-
Seventy subjects (34 men and 36 women; 18–69 years of tained from a collection of past research (1, 7, 26)—5RM
age) of varied resistance training experience were recruit- ⬃80% 1RM, 10RM ⬃70% 1RM, 20RM ⬃60% 1RM. Each
ed from the university campus and from the surrounding subject reached muscular failure for each RM, and the
community. Subjects were recruited using a convenience last completed weight was recorded for the RM. Partial
sampling technique. Prior to the start of the study, sub- repetitions (incomplete extension) did not count as an
jects completed a health history and resistance and aer- RM. If a subject had to redo a given repetition number
obic training questionnaire. In addition, seated resting for a given condition, as a result of ease in obtaining the
blood pressure was measured on the nondominant arm desired repetitions or failure to attain the repetition num-
using manual sphygmomanometry. ber, a 5-minute rest period was given and the condition
Subjects were excluded based upon known disease or was attempted again at an altered load. No subject had
signs or symptoms of health-related problems that would to perform a given repetition number test condition more
interfere with their ability to complete the protocol or than 3 times. Each subject performed the LP exercise,
compromise their health, as recommended and detailed rested for 5 minutes, and then performed the CP exercise.
by the American College of Sports Medicine (2). For ex- A total of 10 minutes of rest was given between each set
ample, if a subject had more than one positive risk factor before the subject was asked to repeat the same exercise
other than age (men ⬎45, women ⬎55 years; family his- regimen again for a different RM. The second session con-
tory, cigarette smoking, hypertension, hypercholesterol- sisted of a 5RM and a 1RM for the same 2 exercises, and
emia, impaired fasting glucose, obesity, and sedentary these loads were again based on their 20 and 10RM loads,
lifestyle), the subject was excluded from participation in based on estimates obtained from past research, as pre-
the study. viously explained.
The research protocol was approved by the University A Timex portable metronome (Timex Corp., Middle-
Institutional Review Board. All details of the study were burg, CT) was used to standardize the 20RM, 10RM, and
explained to each subject on an individual basis prior to 5RM for each subject. The metronome was set at 60
reading of the informed consent and subsequent signing b·min⫺1, and the subject was asked to perform each phase
to confirm participation in the study. All exercise testing (concentric and eccentric) of the repetition in cadence
sessions were completed at the university recreational with the metronome, resulting in a repetition rate of 30
weight room or a community-based health club, both lo- per minute (12, 13). Although such a constrained lifting
cated at an altitude approximating 1,572 m (PB ⬃635 mm cycle is atypical, it was required to ensure similarity be-
Hg) and having identical equipment to that used in this tween subjects and trials for all repetitions. For the CP,
research. subjects had to touch the top of their chest with the bar-
bell for each complete repetition. For the LP, a goniome-
Procedures ter was used to ensure that each subject attained a 90⬚
After successful screening for inclusion and exclusion cri- angle during the eccentric phase and attained full exten-
teria and signing of the consent forms, subjects arrived sion during the concentric phase. Consequently, failure
for their next appointment at the university weight room was defined as the inability to contract to full extension
or health club. Subjects were informed of the need to not for both the LP and CP exercises. One week separated
train the muscles (or antagonists) to be assigned testing the first and second sessions. Subjects that were unable
on a given day for at least 48 hours prior to the scheduled to adhere to the guidelines of the first and second test
session. Height was measured while the subject was in sessions were retested.
mid-inspiration and barefoot. Body weight was measured
Statistical Analyses
to the nearest 0.1 kg on a precalibrated digital scale (Seca
Corporation, Columbia, MD), and subjects were prepared All data were entered into spreadsheet software (Excel;
for anthropometric and skinfold measurements. Subjects Microsoft Corporation, Seattle, WA). For the training
had the girth of their chest, upper arm, and upper thigh questionnaire, responses were coded based on the number
measured. Finally, subjects had 3 skinfold sites measured of repetitions completed each week for each of the LP and
(chest, abdominal, and thigh for male; triceps, suprailiac, CP exercises. For example, a subject who trained for 5
and thigh for women) to predict body density (17). Body sessions per week, with 4 sets of 10 repetitions for the LP
density was converted into a population-specific equation and 3 sets of 8 repetitions for the CP, scored 200 and 120
(14) to estimate percentage body fat based on a 2-com- for LP and CP, respectively. We did not include the loads
ponent model, from which data for fat-free mass (FFM) lifted during training within training volume, as this as-
was calculated. pect of strength was provided by the RM data with the
A standard 22-kg (45-lb) barbell and a nonadjustable multiple regression analyses. As such, training volume
Cybex weight bench (Cybex International, Inc., Medway, then became a unique variable with minimal theoretical
MA) were used for the bench press exercise (chest press, co-linearity to strength.
586 REYNOLDS, GORDON, AND ROBERGS

The spreadsheet data were imported into a statistical TABLE 1. Descriptive characteristics of the subjects (mean ⫾
software program (Statistica; StatSoft, Tulsa, OK) as well SD).*
as a curve-fitting program (Prism; Graphpad Software, Total Men Women
San Diego, CA) for subsequent analyses. Variable (N ⫽ 70) (N ⫽ 34) (N ⫽ 36)
Mixed-design ANOVA (2 [gender] ⫻ 2 [action] ⫻ 4
Age (y) 30.6 ⫾ 11.2 30 ⫾ 11 31 ⫾ 11
[RM condition]) was used to determine if there were gen- Height (cm) 173.3 ⫾ 9.4 180.7 ⫾ 5.9 166.3 ⫾ 6.1
der differences in the change in strength across RM con- Weight (kg) 73.8 ⫾ 16.7 85.0 ⫾ 15.2 63.4 ⫾ 9.9
ditions and for a significant interaction between CP and FFM (kg) 60.4 ⫾ 14.2 72.82 ⫾ 8.82 48.87 ⫾ 6.18
LP. When interactions were significant, simple main ef- %BF 18.0 ⫾ 8.0 13.2 ⫾ 7.3 22.6 ⫾ 5.6
fects analyses were completed, followed by Tukey’s tests Arm girth (cm) 32.2 ⫾ 5.3 35.7 ⫾ 4.9 28.9 ⫾ 3.2
to assess specific mean differences. Chest girth (cm) 98.5 ⫾ 11.5 106.2 ⫾ 10.5 91.3 ⫾ 6.7
The linear or nonlinear profile of strength and RM Thigh girth (cm) 55.0 ⫾ 6.1 57.1 ⫾ 6.5 53.1 ⫾ 4.9
condition for each action was assessed using linear and 1RM LP/FFM 4.3 ⫾ 1.1 4.45 ⫾ 1.20 4.18 ⫾ 1.93
nonlinear (mono- and 2-function exponential decay) curve 1RM CP/FFM 1.1 ⫾ 0.6 1.30 ⫾ 0.37 0.84 ⫾ 0.19
fitting. The strength of the correlations (linear and non- 1RM LP (kg) 264.1 ⫾ 101.4 719 ⫾ 224 450 ⫾ 112
1RM CP (kg) 67.0 ⫾ 34.5 209 ⫾ 64 90 ⫾ 22
linear) was quantified by the correlation coefficient (r),
explained variance (r2), and standard error of estimate * FFM ⫽ fat-free mass; %BF ⫽ % body fat; 1RM ⫽ 1 repetition
(Sy.x). maximum; LP ⫽ leg press; CP ⫽ chest press.
Multiple regression analyses were used to explain the
variance in the predicted 1RM, using the independent
variables of gender, age, height, weight, lean body mass,
body fat percentage, arm girth, chest girth, thigh girth,
and the 5RM, 10RM, and 20RM for the specific action.
Stepwise regression was performed because of the lack of
prior research that has evaluated the additional indepen-
dent variables used in this study. The same procedures
were used for 1RM prediction equations based on 10 and
20RM data. Normality of the residuals was assessed
through raw residual plots for each of the independent
variables using the action-specific 1RM data as the de- FIGURE 1. Three-way interaction (gender ⫻ repetitions to
pendent variable. failure [RM] ⫻ action) for strength. (a) Women decreased less
Linear regression was used to determine correlations in chest press (CP) strength than men across RM conditions.
and resulting residuals from measured and predicted (b) Women also decreased less in leg press (LP) strength than
1RM strength for the cross-validation group using the men. However, compared to CP strength, there were larger
equations from this study. The same procedures were decrements across LP RM conditions for both men and women.
All means are significantly different from each other.
used to assess the accuracy of predicted 1RM strength
using previously published 1RM prediction equations.
The subject number (70) was determined to be appro-
data, the sample was heterogeneous. The average partic-
priate using a priori power estimates based on the rec-
ommendation of at least 10 subjects per independent var- ipant was a resistance trained individual, participating
iable (IV) when conducting biomedical or physiological re- in their own weight-training program 1–3 days per week.
search involving human subjects. As we anticipated ap- The subjects comprised 34 men and 36 women, and train-
proximately 5 IVs for each equation, this required us to ing status consisted of 16 untrained, 37 circuit weight-
have at least 50 subjects. We continued subject recruit- trained, and 17 volume-trained (split body, ⬎4 days per
ment through 70 subjects to gain further improvements week). Each of the 5RM LP and CP data residuals (com-
in statistical power in multiple regression research. The pared to 1RM data) were normally distributed based on
use of an additional cross-validation group further in- plots of raw residuals superimposed to the normal curve.
creased the meaningfulness of our findings. Such a large Interactions Between Gender, RM, and Action
sample size provided excellent statistical power for AN-
OVA-based statistical procedures, providing a power of The mixed-design 3-way ANOVA revealed significant 2-
0.9 for mean differences of 59 and 18 kg for the LP and way (gender ⫻ RM, gender ⫻ action, action ⫻ RM) and
CP, respectively (1-tailed t-test at p ⱕ 0.05). In reality, 3-way (gender ⫻ RM ⫻ action) interactions (all p ⬍
statistical power was far greater for all ANOVA analyses 0.0001) (Figure 1a,b). Obviously, LP strength was greater
(ability to detect a smaller mean difference as significant) than CP across all RM values, and for both genders. Men
as a result of the repeated-measures nature of the re- were stronger than women, and this was more often the
search design. case for CP than for LP. In addition, both men and women
The cross validation of the prediction equations was exhibited a larger decrement in strength with increasing
accomplished using 20 additional subjects for the 5RM RM for the LP than for the CP. The decrement in strength
condition for LP and CP. Statistical significance was ac- across RM values was less for women than men for both
cepted at p ⱕ 0.05. All mean data (text, tables, and fig- actions.
ures) are presented as mean ⫾ standard deviation (SD). To assess whether these gender and RM differences
resulted from the 1RM strength differences between gen-
RESULTS ders and actions, we also completed 2 analysis of covari-
Subjects ance (ANCOVA) analyses (for LP and CP) using the ac-
The physical characteristics of the participants (N ⫽ 70) tion-specific 1RM data as the covariate. For both LP and
are presented in Table 1. As evidenced by the range of CP, the ANCOVA did not alter the significance for any
 PREDICTION OF ONE REPETITION MAXIMUM STRENGTH 587

FIGURE 3. The mean relative strength data of the subjects of

the combined regression and cross-validation groups (n ⫽ 90)
for leg press (LP) and chest press (CP) data. The equations
provided can be used, when the 1 repetition maximum (1RM)
is known, to estimate the %1RM load based on any number of
repetitions.

TABLE 2. Correlation matrix for leg press (LP) and chest

press (CP) strength and pertinent variables (N ⫽ 70).*
Variable 20RM 10RM 5RM 1RM
LP
Training 0.22 0.21 0.21 0.23
20RM 1.00 0.98 0.97 0.96
10RM 1.00 0.99 0.97
5RM 1.00 0.99
1RM 1.00
FIGURE 2. The mean strength data of all subjects (n ⫽ 70) Gender 0.59 0.62 0.63 0.61
presented with linear (solid line) and nonlinear (dotted line) Age (y) ⫺0.40 ⫺0.33 ⫺0.33 ⫺0.34
regression lines. (a) Leg press (LP) data. (b) Chest press (CP) Height (cm) 0.47 0.49 0.48 0.47
data. Weight (cm) 0.59 0.63 0.61 0.58
LBM (kg) 0.71 0.75 0.74 0.71
Thigh girth (cm) 0.54 0.59 0.60 0.59
main effect or interaction from the ANOVA analyses, in- CP
dicating that the absolute 1RM strength differences be-
Training 0.13 0.13 0.12 0.13
tween genders did not contribute to gender differences in 20RM 1.00 0.99 0.99 0.98
the magnitude of load decrement across RM values. 10RM 1.00 0.99 0.99
Linearity vs. Nonlinearity of the RM Data 5RM 1.00 1.00
1RM 1.00
The gender-combined data sets for LP and CP mean data Gender 0.78 0.79 0.79 0.79
were best fit by a nonlinear model (Figure 2), with S y.x Age (y) ⫺0.38 ⫺0.38 ⫺0.35 ⫺0.35
criterion for assessing goodness of fit being LP: linear S y.x Height (cm) 0.55 0.57 0.58 0.59
⫽ 11.2, nonlinear Sy.x ⫽ 2.4 kg (p ⫽ 0.014); CP: linear Sy.x Weight (kg) 0.51 0.53 0.54 0.56
⫽ 2.6, nonlinear Sy.x ⫽ 0.2 kg (p ⫽ 0.018). LBM (kg) 0.73 0.74 0.76 0.77
When the data were expressed as relative decrements Arm girth (cm) 0.61 0.62 0.65 0.66
Chest girth (cm) 0.56 0.57 0.59 0.60
from the 1RM (Figure 3), the nonlinear change in load is
also seen, and the mean values for each %RM condition * RM ⫽ Repetition maximum; LBM ⫽ lean body mass.
were LP of 85.91, 70.10, and 51.58% for the 5, 10, and
20RM, respectively. For CP, the values were 87.45, 75.65,
and 61.61%, respectively. See Figure 3 for the specific maximum strength for LP and CP are presented in Table
equations for estimating loads to be used for a given num- 2. High correlations existed between all strength mea-
ber of repetitions based on the %1RM. sures, yet there was a consistent decrease in correlation
to 1RM as the RM increased for both CP and LP. Training
Univariate Regression Analyses volume had low correlations to 1RM for both LP and CP.
Chest Press and Leg Press. The univariate correlations Anthropometric Variables. The univariate correlations
that existed between the maximum repetition ranges and that existed between anthropometric variables and max-
588 REYNOLDS, GORDON, AND ROBERGS

FIGURE 4. (a) Predicted vs. measured 1 repetition maximum FIGURE 5. (a) Predicted vs. measured 1 repetition maximum
(1RM) leg press (LP) strength. (b) The residuals resulting from (1RM) chest press (CP) strength. (b) The residuals resulting
the prediction of 1RM LP strength from 5RM LP strength from the prediction of 1RM CP strength from 5RM CP
data. The solid line represents the mean residuals (0.01 kg), strength data. The solid line represents the mean residuals
and the dotted lines represent ⫾2 standard deviations (SD) (0.00 kg), and the dotted lines represent ⫾2 standard
(⫾32.08 kg). deviations (SD) (⫾5.92 kg).

ed 1RM LP strength (using 5RM), and Figure 4b presents

imum strength for LP and CP are also presented in Table the distribution of residuals resulting from the prediction.
2. 1RM strength decreased with increasing age and was Chest Press. Forward stepwise multiple regression
moderately correlated with each of the remaining vari- analysis resulted in one significant variable (5RM) enter-
ables. As expected, FFM had the highest correlation to ing into a prediction equation. We also performed the
1RM for LP, whereas gender and FFM revealed similar multiple regression analyses using each of the 10 and
high correlations for the CP. Arm girth was more highly 20RM variables. These results produced the following
correlated to all RM strength scores than thigh girth was equations and results: 5RM: 1.1307 (5RM) ⫹ 0.6998, R2
to LP RM strength scores. ⫽ 0.993, Sy.x ⫽ 2.98 kg; 10RM: 1.2321 (10RM) ⫹ 0.1752
(FFM) ⫺ 5.7443, R2 ⫽ 0.976, Sy.x ⫽ 5.38 kg; 20RM: 1.5471
Multiple Regression Analyses (20RM) ⫹ 3.834, R2 ⫽ 0.955, Sy.x ⫽ 7.36. As for the LP,
Leg Press. Forward stepwise multiple regression analysis CP 1RM prediction error increased with increasing rep-
resulted in only one significant variable (5RM) entering etition numbers, but to a lesser extent. Figure 5a presents
into a prediction equation. We also performed the multi- the relationship between measured and predicted 1RM
ple regression analyses using each of the 10 and 20RM CP strength (using 5RM), and Figure 5b presents the dis-
variables. These results produced the following equations tribution of residuals resulting from the prediction.
and results: 5RM: 1.09703 (5RM) ⫹ 14.2546, R2 ⫽ 0.974, Cross-Validation Group. Prediction equations have an
Sy.x ⫽ 16.16 kg; 10RM: 1.2091 (10RM) ⫹ 38.0908, R2 ⫽ inflated accuracy when based solely on the data from
0.933, Sy.x ⫽ 26.13 kg; 20RM: 1.3870 (20RM) ⫹ 69.2494, which the equations were derived. To reveal a more re-
R2 ⫽ 0.915, Sy.x ⫽ 29.41. Compared to the 5RM prediction, alistic error of prediction, cross-validation groups are rec-
the error (based on Sy.x) associated with using each of the ommended. We recruited an additional 20 subjects to cre-
10 and 20RM equations is considerably large. Figure 4a ate a cross-validation group, and we took care to provide
presents the relationship between measured and predict- a range of pertinent strength and demographic charac-
 PREDICTION OF ONE REPETITION MAXIMUM STRENGTH 589

TABLE 3. Descriptive characteristics of the subjects in the

cross-validation group (n ⫽ 20).*
Variable Mean ⫾ SD Range
Age (y) 12.8 ⫾ 12.8 20–53
Height (cm) 175.3 ⫾ 10.5 163.5–195.1
Weight (kg) 17.4 ⫾ 17.4 53.5–107.1
LBM (kg) 14.9 ⫾ 14.9 44.1–83.0
%BF 8.5 ⫾ 8.5 5.6–37.4
Arm girth (cm) 5.1 ⫾ 5.1 26.6–45.7
Chest girth (cm) 11.6 ⫾ 11.6 82.7–116.3
Thigh girth (cm) 8.7 ⫾ 8.7 36.0–69.3
1RM LP/LBM 302.7 ⫾ 1.5 2.0–7.5
1RM CP/LBM 4.1 ⫾ 0.4 0.6–1.8
1RM LP (kg) 255.4 ⫾ 118.7 133.9–469.6
1RM CP (kg) 63.2 ⫾ 36.7 29.5–142.9
* LBM ⫽ lean body mass; %BF ⫽ % body fat; 1RM ⫽ 1 repe-
tition maximum; LP ⫽ leg press; CP ⫽ chest press.

TABLE 4. Summary of the regression analyses for 1 repeti-

tion maximum (1RM) leg press (LP) and 1RM chest press (CP)
using the cross-validation data set.*
R R2 Sy.x Residuals
LP
5RM equation 0.994 0.988 13.51 ⫺0.47 ⫾ 13.26
10RM equation 0.981 0.962 23.79 ⫺2.36 ⫾ 24.65
20RM equation 0.962 0.926 33.08 ⫺4.33 ⫾ 33.80
CP
5RM equation 0.999 0.998 1.80 ⫺0.52 ⫾ 1.75
10RM equation 0.995 0.991 3.64 ⫺1.34 ⫾ 3.80
20RM equation 0.992 0.984 4.82 ⫺2.63 ⫾ 5.13
* Residuals (kg) ⫽ measured ⫺ predicted.

teristics that represented our initial subject population.

The descriptive characteristics of the cross-validation
group are presented in Table 3. FIGURE 6. Data from the cross-validation group (n ⫽ 20). (a)
Use of the LP and CP prediction equations revealed Predicted vs. measured 1 repetition maximum (1RM) leg press
slightly decreased prediction error compared to the orig- (LP) strength. (b) The residuals of the predicted (from the
5RM equation) and measured 1RM LP strength. The solid line
inal (n ⫽ 70) data set, with the results for both LP and
represents the mean residuals (⫺0.47 kg), and the dotted lines
CP presented in Table 4. Figure 6a presents the relation- represent ⫾2 standard deviations (SD) (⫾26.52 kg).
ship between measured and predicted 1RM LP strength
using 5RM data, and Figure 6b presents the distribution
of residuals resulting from the prediction. Similarly, Fig-
ure 7a presents the relationship between measured and Epley (11), had the smallest mean residuals, with only a
predicted 1RM CP strength using 5RM data, and Figure slight trend for overestimation. The equations of Brzycki
7b presents the distribution of residuals resulting from (7), Lander (18), and O’Connor (26) all underestimated LP
the prediction. strength. All equations were less accurate when the
10RM rather than 5RM data was used. The nonlinear
Comparison to Other Prediction Equations equations of Lombardi (19) and Mayhew (21) were less
In order to compare the prediction accuracy of our equa- accurate than all linear equations.
tions to past research, we compared our prediction equa-
Test–Retest Reliability
tions to 6 different linear prediction equations and 2 non-
linear prediction equations for our cross-validation data Twenty additional subjects were used for the test–retest
set. The equations and resulting goodness-of-fit criteria reliability. These subjects were also of varied training sta-
are presented in Table 5. tus and of identical gender distribution, but they were
Our equation was evaluated at the 5 and 10RM values younger, ranging from 19 to 42 years of age. Data collec-
for both LP and CP as a result of the nonspecific nature tion occurred in a manner identical to that of prior de-
of the RM condition for many of the other prediction equa- scriptions, but testing was only done on the 5RM condi-
tions. For previously published equations, the authors’ tion. The intraclass correlation coefficient was computed
recommendations were strictly followed, and RM values as the correlation between repeated test scores (Statisti-
were used that applied to these alternate equations. ca), and such test–retest correlations for the LP and CP
All of the linear prediction equations using the 5RM were 0.999 and 0.999, respectively. The proportion of sub-
data functioned with similar accuracy for the LP and CP. jects that scored the identical load for LP and CP were
However, our equations, and those of Abadie (1) and 65 and 60%, respectively, with ranges of residuals being
590 REYNOLDS, GORDON, AND ROBERGS

plication to all weight-training actions, but there was no

citation to published research.
These findings have application to both the prediction
of 1RM and the designation of weight to lift for a given
number of repetitions. We illustrated the latter concept
in Figure 3 and provided prediction equations based on
this relative (%RM) data. For example, if the 1RM and
number of training repetitions are known, the load to be
lifted can be calculated using the action-specific predic-
tion equations. For example, for a LP 1RM of 300 kg, the
load to be lifted for sets of 10 repetitions is calculated to
be 212 kg. For a CP 1RM of 150 kg, the load to be lifted
for sets of 10 repetitions is calculated to be 113 kg. For
comparison, the table of Landers et al. (18) yields a LP
load of 225 kg and a CP load of 112 kg. The equation of
Mayhew et al. (21) for CP load estimation yields a load of
115 kg. Interestingly, our CP equation produces similar
results to the studies of Landers et al. and Mayhew et al.,
but our LP equation produces meaningfully different
data. The only other research study on this topic was that
of Abadie et al. (1). However, these authors solely pre-
sented 1RM prediction equations and did not reassess
their data to evaluate %RM loads.
Research of RM strength testing has mainly focused
on the need to predict the 1RM (1, 4, 6–9, 18, 20–23, 25,
29, 30). Our results revealed acceptable prediction accu-
racy for 1RM for each of the LP and CP. For example,
our regression predictions for the LP and CP resulted in
similar prediction accuracy for the original 70 subjects, as
well as for the 20 subjects used in the cross-validation
group. This consistency reaffirms the accuracy of our
equations. We have been the only investigators to use a
cross-validation group, which we also used to cross-vali-
date all prior prediction equations (Table 5). The data
from Table 5 reveals that our equation and that of Abadie
FIGURE 7. Data from the cross-validation group (n ⫽ 20). (a) et al. (1) are the most accurate for LP. Conversely, our
Predicted vs. measured 1 repetition maximum (1RM) chest equation and those of Bryzcki (7) and O’Connor et al. (26)
press (CP) strength. (b) The residuals of the predicted (from are the most accurate for the CP.
the 5RM equation) and measured 1RM CP strength. The solid The inability of anthropometric variables, as well as
line represents the mean residuals (⫺0.52 kg), and the dotted gender and training volume, to improve the prediction
lines represent ⫾2 standard deviations (SD) (⫾3.54 kg). accuracy of LP and CP 1RM was unexpected. These find-
ings indicate that each of these variables is either unre-
lated to 1RM strength (e.g., training volume) or is so in-
5 to 10 and 2 to 5 kg, respectively. The mean errors for terrelated to strength (gender and all anthropometric
the LP and CP were both ⫾0.5%. variables) that the 5RM data sufficiently accounts for its
contribution to the explanation of between-subject vari-
DISCUSSION ance in 1RM strength. This is fortunate from the per-
spective of the strength coach or personal trainer, who
We reported a nonlinear relationship between strength can be assured that accurate 1RM strength prediction is
and increasing repetitions to failure for both LP and CP. as simple as applying a known 5RM strength value to a
Women were less strong than men, and they were more simple equation or to charts derived from these equa-
so for CP than for LP exercises. For both LP and CP, the tions. If loads to be used are required to be lower in test-
most accurate prediction of strength occurred from a 5RM ing, then prediction can still occur using our 10 and 20RM
test, with the accuracy of prediction worsening with in- test equations.
creasing repetitions to failure. Addition of anthropomet- Rest periods between sets determine how quickly an
ric, gender, and training volume variables to the predic- individual will recover from the previous bout of exercise.
tion equations did not significantly improve the accuracy Although we did not experimentally assess differences in
of prediction. the rest interval, the variable duration of the rest interval
Our data clearly reveal the trend for a nonlinear de- used by investigators warrants comment on this topic. In
crease in strength with increasing repetitions to failure. this study, 5-minute rest periods were used for the LP
Surprisingly, we have been the first to document such a and CP exercises. Furthermore, a warm up period of 5
trend. However, Mayhew et al. (21) reported a nonlinear minutes was used to transition subjects from the CP ac-
equation to estimate 1RM for the bench press from the tion to the LP action. In prior research, there have been
number of repetitions performed in 1 minute. Lombardi varying practices for these rest periods. Some investiga-
(19) published a book with a nonlinear equation with ap- tors employed a 2-minute rest period (1, 5), others utilized
 PREDICTION OF ONE REPETITION MAXIMUM STRENGTH 591

TABLE 5. Linear and nonlinear 1 repetition maximum (1RM) prediction equations, with data from regression analyses using the
cross-validation data set.*
Author Prediction equation R2 Sy.x
Linear
Reynolds (N ⫽ 70) Varied subject population
1RM LP ⫽ (1.09703 ⫻ [5RM, kg]) ⫹ 14.2546
1RM CP ⫽ (1.1307 ⫻ [5RM, kg]) ⫹ 0.6998
5RM LP 0.988 13.23
10RM LP 0.962 21.64
5RM CP 0.998 1.78
10RM CP 0.991 3.49
Abadie (1) (N ⫽ 30) Women, college aged
1RM ⫽ 7.24 ⫹ (1.05 ⫻ weight lifted)
5RM LP 0.988 13.74
10RM LP 0.962 20.41
5RM CP 0.998 1.67
10RM CP 0.990 3.00
Bryzcki (7) (N ⫽ ?) Population unknown
1RM ⫽ (Weight lifted/(1.0278 ⫺ [0.0278 ⫻ No. of reps])
5RM LP 0.988 13.56
5RM CP 0.998 1.79
Epley (11) (N ⫽ ?) Population unknown
%1RM ⫽ ([0.033 ⫻ reps] ⫻ rep wt.) ⫹ rep wt.
5RM LP 0.988 14.05
5RM CP 0.998 1.85
Lander (18) (N ⫽ ?) Population unknown
%1RM ⫽ 101.3 ⫺ (2.67123 ⫻ reps)
5RM LP 0.988 13.71
10RM LP 0.962 24.00
5RM CP 0.998 1.81
10RM CP 0.990 3.84
O’Connor (26) (N ⫽ ?) Population unknown
%1RM ⫽ (0.025 ⫻ [rep wt. ⫻ reps]) ⫹ rep wt.
5RM LP 0.988 13.56
10RM LP 0.962 22.37
5RM CP 0.998 1.79
10RM CP 0.990 3.58
Nonlinear
Lombardi (19) (N ⫽ ?) Population unknown
1RM ⫽ (No. of reps0.1) ⫻ (weight lifted)
5RM LP 0.988 14.16
10RM LP 0.962 22.54
5RM CP 0.998 1.87
10RM CP 0.990 3.60
Mayhew (21) (N ⫽ 434) 185 college men; 251 college women
%1RM ⫽ 52.2 ⫹ 41.9e⫺0.055⫻reps
5RM LP 0.988 14.35
10RM LP 0.962 23.44
5RM CP 0.998 1.89
10RM CP 0.990 3.75
* LP ⫽ leg press; CP ⫽ chest press. Guidelines for authors’ equations: Bryzcki and Epley ⫽ ⬍10 reps; Abadie ⫽ 5–10 reps; Lombardi
⫽ ⬍11 reps; Mayhew et al. ⫽ ⬍15 reps; Lander and O’Connor ⫽ not specified.

a 5-minute rest period (9), and still others have used a here to the 5-minute rest interval between sets in RM
variable 1–3- or 3–5-minute rest period (21, 23, 24). testing.
Rest periods of less than 5 minutes are likely to be too Based on the results in this study, the following con-
short. Muscle creatine phosphate recovery reveals a dual clusions were drawn: (a) 1RM strength, compared to a
exponential curve having a fast and slow component (3, multiple repetition maximum load, is a nonlinear rela-
27). The fast component of creatine phosphate recovery is tionship, in which the magnitude of the decline in load
complete within less than 2 minutes and represents 80– becomes smaller with increasing repetitions used in RM
90% of complete creatine phosphate recovery (3, 27). Ad- testing; (b) Women are less strong than men, and this is
ditionally, the slow component may require up to 45–90 more often the case for the CP than for the LP exercises;
more seconds after the fast component (3, 27). Both com- (c) Of the 5, 10, and 20RM, the best repetition maximum
ponents of creatine phosphate recovery are slowed with range to use for prediction of 1RM strength in the LP and
increasing acidosis (3). We would encourage other inves- CP exercises is the 5RM; (d) In order to increase accuracy,
tigators, as well as strength coaches and trainers, to ad- prediction equations must be exercise specific; (e) The
592 REYNOLDS, GORDON, AND ROBERGS

most accurate equations to predict 1RM strength for the 11. EPLEY, B. Poundage chart. In: Boyd Epley Workout. Lincoln,
LP are ours and those of Abadie et al. (1). The most ac- NE: Body Enterprises, 1985. p. 86.
curate equations to predict 1RM strength for the CP are 12. GILLETTE, C.A., R.C. BULLOUGH, AND C.L. MELBY. Postexercise
energy expenditure in response to acute aerobic or resistive
ours and those of Bryzcki (7) and O’Connor et al. (26);
exercise. Med. Sci. Sports Exerc. 30:512–522. 1994.
and (f) Our data can also be used to predict the load need- 13. HALTOM, R.W., R.R. KRAEMER, R.A. SLOAN, E.P. HERBERT, K.
ed for efforts to failure for specific repetition conditions FRANK, AND J.L. TRYNIECKI. Circuit weight training and its
for LP and CP. effects on excess post exercise oxygen consumption. Med. Sci.
Sports Exerc. 31:1613–1618. 1999.
PRACTICAL APPLICATIONS 14. HEYWARD, V., AND L. STOLARCZYK. Applied Body Composition
It is important that practitioners use prediction equations Assessment. Champaign, IL: Human Kinetics, 1996.
for specific actions. We used the free-weight barbell bench 15. HOEGER, W.K., D. HOPKINS, S. PARRETTE, AND D. HALE. Re-
press for the CP action and the Cybex plate-loaded leg lationship between repetitions and selected percentages of 1-
RM: A comparison between untrained and trained males and
press for the LP action. As the CP action used free
females. J. Appl. Sports Sci. Res. 4:47–54. 1990.
weights, the range of contraction and rate or contraction 16. HORTOBAGYI, T.F., V. KATCH, P. LACHANCE, AND A. BEHNKE.
variables are all that are required for correct application Relationships of body size, segmental dimensions, and ponder-
of the prediction equations. However, there are different al equivalents to muscular strength in high strength and low-
types of LP equipment in weight rooms, and we discour- strength subjects. Int. J. Sports Med. 11:349–356. 1990.
age any application of our equations to multiple RM test- 17. JACKSON, A.S., AND M.L. POLLACK. Practical assessment of
ing for prediction of 1RM on equipment other than the body composition. Phys. Sport Med. 13:76–90. 1985.
Cybex plate-loaded LP, or when contraction or rest du- 18. LANDERS, J. Maximum based on reps. Natl. Strength Cond. As-
rations that differ from 1-second concentric, 1-second ec- soc. J. 6:60–61. 1985.
19. LOMBARDI, V.P. Beginning Weight Training. Dubuque, IA: W.C.
centric pattern are used.
Brown, 1989.
When assessing 5RM prediction of 1RM, care should 20. MAYHEW, J.L., T. BALL, AND M. ARNOLD. Prediction of 1-RM
be given to the quality of repetitions during the 5RM. For bench press from submaximal bench press performance in col-
example, based on our equations, a ⫾5-kg error in 5RM lege males and females. J. Appl. Sport Sci. Res. 3:73. 1989.
load calculates to a 1RM prediction error of 19.7 kg for 21. MAYHEW, J.L., T. BALL, M. ARNOLD, AND J. BOWEN. Relative
the LP and 6.3 kg for the CP. Clearly the 5RM test needs muscular endurance performance as a predictor of bench press
to be conducted with strict attention to methodological strength in college men and women. J. Appl. Sport Sci. Res. 6:
detail. Once the 1RM is known (whether measured or pre- 200–206. 1992.
dicted), our equations can be used to estimate the load 22. MAYHEW, J.L., T.E. BALL, T.E. WARD, C.L. HART, AND M.D.
needed for sets to failure for a specific repetition number. ARNOLD. Relationships of structural dimensions to bench press
strength in college males. J. Sports Med. Phys. Fitness 31:135–
REFERENCES 141. 1991.
23. MAYHEW, J.L., J.L. PRINSTER, J.S. WARE, D.L. ZIMMER, J.R.
1. ABADIE, B.R., AND M. WENTWORTH. Prediction of 1-RM ARABAS, AND M.G. BEMBEN. Muscular endurance repetitions to
strength from a 5–10 repetition submaximal strength test in predict bench press strength in men of different training levels.
college-aged females. J. Exerc. Physiol. [serial online] 3:1–5. J. Sports Med. Phys. Fitness 35:108–113. 1995.
2000. 24. MAYHEW, J.L., F. PIPER, AND J. WARE. Anthropometric corre-
2. AMERICAN COLLEGE OF SPORTS MEDICINE. ACSM’s Guidelines lates with strength performance among resistance-trained ath-
for Exercise Testing and Prescription (6th ed.). Philadelphia: letes. J. Sports Med. Phys. Fitness 33:159–165. 1993.
Lippincott, Williams, & Wilkins, 2000. 25. MORALES, J., AND S. SOBONYA. Use of submaximal repetition
3. ANDERSON, D.L., P.M. MATHEWS, AND G.K. RADDA. Metabolic tests for predicting 1-RM strength in class athletes. J. Strength
recovery after exercise and the assessment of mitochondrial Cond. Res. 10:186–189. 1996.
function in vivo in human skeletal muscle by means of 31P 26. O’CONNOR, B., J. SIMMONS, AND P. O’SHEA. Weight Training
NMR. Magnet. Reson. Imaging 1:307–315. 1984. Today. St. Paul, MN: West Publishing, 1989.
4. BALL, T.E., AND K.S. ROSE. A field test for predicting maximum 27. ROBERGS, R., AND S. ROBERTS. Exercise Physiology: Exercise,
bench press lift of college women. J. Appl. Sports Sci. Res. 5: Performance, and Clinical Applications (1st ed.). St. Louis: Mos-
169–170. 1992. by-Year Book, Inc., 1997.
5. BRAITH, R.W., J.E. GRAVES, S.H. LEGGETT, AND M.L. POLLACK. 28. SALE, D.G., AND D. MACDOUGALL. Specificity in strength train-
Effect of training on the relationship between maximal and ing: A review for the coach and athlete. Can. J. Appl. Sport Sci.
submaximal strength. Med. Sci. Sports Exerc. 25:132–138. 6:87–92. 1981.
1993. 29. SCANLAN, J.M., K. BALLMANN, J.L. MAYHEW, AND C. LANTZ.
6. BRZYCKI, M. Strength testing: Predicting a one-rep max from Anthropometric dimensions to predict 1-RM bench press in un-
reps to fatigue. J. Health Phys. Ed. Rec. and Dance 64:88–90. trained females. J. Sports Med. Phys. Fitness 39:54–60. 1999.
1993. 30. WARE, J.S., C.T. CLEMENS, J.L. MAYHEW, AND T.J. JOHNSTON.
7. BRZYCKI, M. Assessing strength. Fitness Manage. June:34–37. Muscular endurance repetitions to predict bench press and
2000. squat strength in college football players. J. Strength Cond.
8. CHAPMAN, P.P., J. WHITEHEAD, AND R. BINKERT. The 225-lb Res. 9:99–103. 1995.
reps-to-fatigue test as a submaximal estimate of 1-RM bench
press performance in college football players. J. Strength Cond. Acknowledgments
Res. 12:258–261. 1998.
9. CUMMINGS, B., AND K. FINN. Estimation of a repetition maxi- The authors wish to thank Defined Fitness, Inc., for the use of
mum bench press strength for untrained women. J. Strength their facility in the data collection process.
Cond. Res. 12:262–265. 1998.
10. DELORME, T., AND A. WATKINS. Techniques of progressive re- Address correspondence to Robert A. Robergs,
sistance exercise. Arch. Phys. Rehab. 29:263–273. 1948. [email protected].