Original Investigations

International Journal of Sports Physiology and Performance, 2008, 3, 16-30

© 2008 Human Kinetics, Inc.

Quantifying Training Load: A Comparison

of Subjective and Objective Methods
Jill Borresen and Michael I. Lambert

Purpose: To establish the relationship between a subjective (session rating of per-

ceived exertion [RPE]) and 2 objective (training impulse [TRIMP]) and summated-
heart-rate-zone (SHRZ) methods of quantifying training load and explain char-
acteristics of the variance not accounted for in these relationships. Methods:
Thirty-three participants trained ad libitum for 2 wk, and their heart rate (HR) and
RPE were recorded to calculate training load. Subjects were divided into groups
based on whether the regression equations over- (OVER), under- (UNDER), or
accurately predicted (ACCURATE) the relationship between objective and sub-
jective methods. Results: A correlation of r = .76 (95% CI: .56 to .88) occurred
between TRIMP and session-RPE training load. OVER spent a greater percentage
of training time in zone 4 of SHRZ (ie, 80% to 90% HRmax) than UNDER (46%
± 8% vs 25% ± 10% [mean ± SD], P = .008). UNDER spent a greater percentage
of training time in zone 1 of SHRZ (ie, 50% to 60% HRmax) than OVER (15%
± 8% vs 3% ± 3%, P = .005) and ACCURATE (5% ± 3%, P = .020) and more
time in zone 2 of SHRZ (ie, 60% to 70%HRmax) than OVER (17% ± 6% vs 7% ±
6%, P = .039). A correlation of r = .84 (.70 to .92) occurred between SHRZ and
session-RPE training load. OVER spent proportionally more time in Zone 4 than
UNDER (45% ± 8% vs 25% ± 10%, P = .018). UNDER had a lower training HR
than ACCURATE (132 ± 10 vs 148 ± 12 beats/min, P = .048) and spent more
time in zone 1 than OVER (15% ± 8% vs 4% ± 3%, P = .013) and ACCURATE
(5% ± 3%, P = .015). Conclusions: The session-RPE method provides reason-
ably accurate assessments of training load compared with HR-based methods,
but they deviate in accuracy when proportionally more time is spent training at
low or high intensity.

Keywords: TRIMP, session RPE, summated heart-rate zones

Physical activity is interpreted by the body as a physiological stress. If exposure

to the stress of exercise is repeated regularly, adaptations to various physiological
systems will be induced that might be associated with improved performance.1 Fur-
thermore, athletes respond differently to the same training load.2–4 Thus, the ability
to measure and monitor training loads would provide valuable information for the
prescription of individualized exercise programs. An athlete’s heart-rate response to

The authors are with the Research Unit for Exercise Science and Sports Medicine, University of Cape
Town, Newlands, 7725 Cape Town, South Africa.

16
 Quantifying Training Load   17

exercise, combined with exercise duration, might represent an objective measure of

physical effort that would allow the quantification of a training session into a unit
“dose” of physical effort. Banister and coworkers named this concept the training
impulse (TRIMP), the equation for which is explained in detail by Banister.5 The
ability to quantify training to a single figure or factor is appealing in terms of its
practical application. Use of the TRIMP equation requires steady-state heart-rate
measurements, however, which limits the quantification of exercise consisting of
alternating bouts of high-intensity exercise and recovery.
Other methods of quantifying training load have evolved and are currently
used in research. For example, the summated-heart-rate-zones method, proposed by
Edwards,6 might facilitate the quantification of interval training because it divides the
training session into duration spent in each of 5 heart-rate zones (50% to 60%, 60% to
70%, 70% to 80%, 80% to 90%, and 90% to 100% of maximal heart rate). Duration
in each zone is multiplied by a different factor, which weights the higher intensity
zones more than the lower intensities. Thereafter the adjusted scores are summated.6
Recently a modified version of the summated-heart-rate-zones equation was used by
Earnest et al7 and Lucia et al8 and referred to as Lucia’s TRIMP by Impellizzeri et
al.9 In this method the duration spent in each of 3 heart-rate zones (zone 1, below the
ventilatory threshold; zone 2, between the ventilatory threshold and the respiratory-
compensation point; and zone 3, above the respiratory-compensation point) is
multiplied by a coefficient (k) relative to each zone (k = 1 for zone 1, k = 2 for zone
2, and k = 3 for zone 3), and the adjusted scores are then summated. The original
source of this equation, however, was not referenced in these articles.
Foster et al10 simplified the quantification of training load by substituting a ses-
sion rating-of-perceived-exertion (RPE) measure for heart-rate data. RPE is based
on the understanding that athletes can inherently monitor the physiological stress
their body is experiencing during exercise.11 According to this procedure an exercise
session load is calculated by multiplying session RPE by the duration of aerobic
exercise.10 The session RPE was developed with the goal of further simplifying, yet
remaining equivalent to, the objective measures of assessing training load (ie, TRIMP
and summated-heart-rate-zones methods). To our knowledge the session RPE is the
only subjective measure of training load that has been developed and popularized.
Foster et al10,12 evaluated the relationship between the subjective session-RPE method
and the objective summated-heart-rate-zones method and found that the individual
correlations between the 2 techniques ranged from r = .75 to .90. They concluded
that the session-RPE method was a valid and reliable measure of exercise intensity
in aerobic exercise when compared with heart-rate-based methods.10,12
Impellizzeri et al9 studied the relationship between the session-RPE method
and 3 objective methods—Banister’s TRIMP, Edwards’ summated-heart-rate-zones
method, and Lucia’s TRIMP—in soccer players during training and match play.
Individual correlations between the session-RPE method and Banister’s TRIMP
method ranged between r = .50 and .77; individual correlations between the session-
RPE method and Edwards’ summated-heart-rate-zones method ranged from r
= .54 to .78, and individual correlations between the session-RPE method and
Lucia’s TRIMP method ranged between r = .61 and .85. To our knowledge, how-
ever, neither the TRIMP, summated-heart-rate-zones score, nor Lucia’s TRIMP
equation has been validated, posing the question of the legitimacy of validating
the session-RPE method against these heart-rate-based methods. Collectively the
18   Borresen and Lambert

data from these studies suggest that the relationship between the different methods
of quantifying training load varies. The reasons for the variations have not been
clearly explained. Therefore, the aim of our study was to assess the relationships
between the session-RPE method and the TRIMP and summated-heart-rate-zones
methods in the quantification of ad libitum training and to identify characteristics
that can explain the variance not accounted for in the relationship between the
objective and subjective methods of quantifying training load.

Methods
Study Design and Subjects
Thirty-three habitually physically active men (n = 15) and women (n = 18) were
recruited for this 2-week trial. Before the start of the trial the body composition,
maximal oxygen consumption, and maximal heart rate of the participants were
measured. Thereafter the participants trained ad libitum for 2 weeks during which
time their training was quantified simultaneously using the session-RPE,10 TRIMP,5
and summated-heart-rate-zones6 methods.
This prospective observational cohort study was approved by the University
of Cape Town Ethics and Research Committee and carried out in accordance with
the principles outlined in the Declarations of Helsinki.13 The participants gave their
informed consent after the testing protocol had been explained to them.

Methodology
Anthropometric and Physiologic Measurements.  Body-fat percent was assessed
with a near-infrared measurement on the right biceps using the Futrex-6100A/ZL
(Kett Electric Laboratory, Futrex Inc, Gaithersburg, MD). A reliability study in
our laboratory showed that the relationship between body composition using
dual-photon X-ray absorptiometry and near-infrared was r = .84, and the limits of
agreement were –2.8% to 5.0% (n = 59; unpublished data). Body mass was recorded
to the nearest 100 g, and the height of each athlete was recorded to the nearest
millimeter (Seca model 708, Germany). Resting heart rate, recorded immediately
after the participants awoke in the morning with a Vantage XL heart-rate monitor
(Polar Electro, Kempele, Finland), was taken at least 3 times during the trial.
An incremental exercise test to exhaustion was conducted on a motor-driven
treadmill to determine subjects’ peak oxygen consumption (VO2max), peak treadmill
running speed, and maximum heart rate (beats/min). A face mask was secured over
each athlete’s nose and mouth to collect expired O2 and CO2, and an Oxycon Alpha
gas analyzer (Jaeger/Mijnhardt, Groningen, Netherlands) was used to measure VO2
during the test. A 5-minute warm up at 8 to 10 km/h and 0% gradient preceded the
start of the test. The subject began the test running at 10 to 12 km/h, after which
the speed was increased by 0.5 km/h every 30 seconds until the subject was unable
to maintain the pace. VO2max (mL · kg–1 · min–1), peak treadmill running speed
(km/h), and maximum heart rate (beats/min) were defined as the highest respective
measurements recorded during the last full 30-second interval completed.
Quantification of Training Load.  All the training data collected and analyzed in
this study were cardiovascular in nature. The training of 19 of the 33 participants
 Quantifying Training Load   19

consisted exclusively of running, whereas 4 other runners also engaged in some

gym training. Gym training consisted mainly of interval-type training that combined
cycling, stepping, spinning, and treadmill/interval running. Subjects also reported
using elliptical machines, the supercircuit, cross-trainers, or rowing machines. Five
participants engaged in gym training exclusively. Four subjects did approximately
equal amounts of running and cycling, and 1 subject only cycled. During the 2
weeks of the trial each participant wore a heart-rate monitor for each exercise
training session to record heart rate and exercise duration. Session training durations
(minutes) were summated for each subject for the 2 weeks to calculate total training
duration (over the 2 weeks) in minutes. Subjects also completed a daily training
diary in which they recorded session RPE (Table 1) for each training bout. Three
equations were used to calculate training load for each of the training sessions
performed during the trial:
TRIMP (Equation 1)

Men: duration (min) × (HRex – HRrest)/(HRmax – HRrest) × 0.64e1.92x

Women: duration (min) × (HRex – HRrest)/(HRmax – HRrest) × 0.86e1.67x
where e = 2.712, x = (HRex – HRrest)/(HRmax – HRrest), HRrest = average heart
rate during rest, and HRex = average heart rate during exercise (Banister et al5).
Summated-heart-rate-zones method (Equation 2)
(duration in zone 1 × 1) + (duration in zone 2 × 2) + (duration in zone 3 × 3)
+ (duration in zone 4 × 4) + (duration in zone 5 × 5)
where zone 1 = 50% to 60% of maximum heart rate, zone 2 = 60% to 70% HRmax,
zone 3 = 70% to 80% HRmax, zone 4 = 80% to 90% HRmax, and zone 5 = 90%
to 100% HRmax (Edwards6).

Table 1  Scale for Session Rating of Perceived Exertiona

0 Rest
1 Really easy
2 Easy
3 Moderate
4 Sort of hard
5 Hard
6
7 Really hard
8
9 Really, really hard
10 Just like my hardest race
a
From Foster et al.10
20   Borresen and Lambert

Session-RPE method (Equation 3)

duration × session RPE (Foster et al10)
In this method the athlete’s perception of the overall difficulty of the training bout
was recorded 30 minutes after the completion of the exercise10 according to the
scores in Table 1. The session-RPE scale (Table 1) is based on the Borg category-
ratio (CR-10) RPE scale, which translates the athlete’s perception of effort into a
numerical score between 0 and 10. This test is designed to encourage the athlete
to respond to a simple question—How was your workout?—with the goal of get-
ting an uncomplicated response that reflects the athlete’s global impression of the
workout.10
Session training loads were summated for each subject for the 2 weeks to
calculate total training load for each method. The equations and the standard error
of the estimate (SEE) defining the relationship between the total training loads for
the 3 methods were calculated via linear regression using GraphPad Prism version
3 for Windows (GraphPad Software, San Diego, CA, www.graphpad.com). The
data points of the subjects that occurred outside the SEE of the regression equa-
tion were identified and grouped according to whether they were above (OVER) or
below (UNDER) the SEE of the regression line. The ACCURATE group consisted
of subjects within the SEE around the regression equation. The anthropometric,
physiological, and training characteristics of the 3 groups were then compared.

Statistical Analyses
A Pearson product–moment correlation was performed to assess the relation-
ship between the objective (HR-based) and subjective (RPE-based) methods of
quantifying training load (for each session for each subject and with pooled data
for each subject). The 95% confidence intervals around the correlation coefficient
were calculated using a spreadsheet for this purpose downloaded from Hopkins.14 A
Levene test was used to check for the homogeneity of variance between the groups.
This test showed unequal variance, so nonparametric tests were used thereafter.
A Kruskal–Wallis test was used to compare the 3 groups (OVER, UNDER, and
ACCURATE) that were identified based on their position relative to the SEE of
the regression equation. Statistical analysis was performed using STATISTICA 7.0
data-analysis software system (StatSoft, Inc, Tulsa, OK), and statistical significance
was accepted at P < .05.

Results
The 33 participants of this study had a mean (± SD) age of 30 ± 5 years, mass of
67 ± 13 kg, height of 171 ± 10 cm, body fat of 20% ± 7%, VO2max of 56 ± 8 mL
· min–1 · kg–1, and resting heart rate of 55 ± 7 beats/min. They had an average of
11 ± 4 training sessions in the 2 weeks. The variation in training load was large
for each subject. For example, the average coefficient of variation for total train-
ing load for each subject determined over the 2 weeks was 75% (95% CI: 63% to
87%) for the session-RPE method, 78% (68% to 89%) for TRIMP, and 74% (63%
to 85%) for the summated-heart-rate-zones method. Relationships between the 3
 Quantifying Training Load   21

different methods were then evaluated to establish whether the 2 HR-based and the
RPE-based methods provided similar assessments of training load.

TRIMP Versus Summated-Heart-Rate-Zones Method

A correlation of r = .98 (95% CI: .96 to .99) was found between total training load
calculated using the TRIMP and the summated-heart-rate-zones methods (pooled
data).

TRIMP Versus Session-RPE Method

A correlation of r = .76 (95% CI: .56 to .88) occurred between total training loads
calculated using TRIMP and the session-RPE method (Figure 1; pooled data). The
95% CI for the correlation coefficients determined for the session training loads of
each subject was r = .84 to .93 (ie, within-subject data).
Baseline anthropometric, physiological, and training characteristics are shown
in Table 2 for the OVER group (n = 6), UNDER group (n = 4), and ACCURATE
group (n = 23), that is, the 3 groups identified based on their position relative to
the SEE of the regression equation for the correlation between the session-RPE
method and TRIMP.
The position of the data points for the OVER group lying above the regression
line in Figures 1 and 2 indicates that the TRIMP method overestimated the

Figure 1 — Relationship between total training load calculated using TRIMP and the session-
RPE method. Groups were based on whether the regression equations overpredicted (OVER),
underpredicted (UNDER), or accurately predicted (ACCURATE) the relationship between
the methods.
22   Borresen and Lambert

Table 2  Subject Characteristics for Athletes in OVER (n = 6),

UNDER (n = 4), and ACCURATE (n = 23) in the Correlation Between
the Session-RPE Method and the TRIMP Method (mean ± SD)a

Group
 Subject characteristic OVER UNDER ACCURATE
Age (years) 31 ± 6 27 ± 5 30 ± 5
Mass (kg) 58 ± 10 73 ± 14 68 ± 13
Height (cm) 170 ± 9 175 ± 12 170 ± 10
Body fat % 19 ± 6 17 ± 3 20 ± 7
VO2max (mL · min · kg )
–1 –1
57 ± 10 61 ± 7 55 ± 8
HRmax (beats/min) 181 ± 10 193 ± 9 192 ± 9
Peak treadmill running speed (km/h) 17 ± 3 18 ± 1 17 ± 2
Total training duration over 2 weeks (min) 50 ± 4 52 ± 10 57 ± 7
Average training HR (beats/min) 894 ± 325 b
698 ± 614 461 ± 300
Resting HR (beats/min) 144 ± 11 132 ± 10 148 ± 13
Percent time in zone 1 (%) 3±3 c
15 ± 8 d
5±3
Percent time in zone 2 (%) 7±6 c
17 ± 6 13 ± 6
Percent time in zone 3 (%) 22 ± 6 30 ± 9 27 ± 11
Percent time in zone 4 (%) 46 ± 8c 25 ± 10 36 ± 9
Percent time in zone 5 (%) 17 ± 10 3±1 d
17 ± 14
Training load (RPE method, AU) e
3383 ± 1671 4717 ± 4164 2298 ± 1549
Training impulse (TRIMP, AU) 1896 ± 480b,c 907 ± 879 853 ± 461
Training load (summated-HR-zone
method, AU) 3084 ± 881b 1819 ± 1701 1520 ± 840
a
Groups were based on whether the regression equations over-, under-, or accurately predicted the
relationship between the methods (pooled data).
b
OVER significantly different from ACCURATE.
c
OVER significantly different from UNDER.
d
UNDER significantly different from ACCURATE.
e
This comparison should be interpreted with caution because the interpretation of RPE scales might
vary between individuals.

calculation of training load when compared with the session-RPE method for
this group. Conversely, the data points of the UNDER group lying below the
regression line in Figures 1 and 2 indicate that the TRIMP method underestimated
the calculation of training load compared with the session-RPE method for the
UNDER group.
OVER had a higher total training duration over 2 weeks (894 ± 325 vs 461
± 300 minutes, P = .031) and training load (using the summated-heart-rate-zones
 Quantifying Training Load   23

Figure 2 — Relationship between total training load calculated using the summated-heart-
rate-zones method and the session-RPE method. Groups were based on whether the regres-
sion equations overpredicted (OVER), underpredicted (UNDER), or accurately predicted
(ACCURATE) the relationship between the methods.

method, 3084 ± 881 vs 1520 ± 840 AU, P = .015) than ACCURATE (Table 2).
OVER spent 21% more of their total training time in heart-rate zone 4 than UNDER
(46% ± 8% vs 25% ± 10%, P = .008). OVER had a higher total TRIMP than both
ACCURATE (1896 ± 480 vs 853 ± 461 AU, P = .004) and UNDER (1896 ± 480
vs 907 ± 879 AU, P = .045). UNDER spent 12% more of their total training time
in heart-rate zone 1 than OVER (15% ± 8% vs 3% ± 3%, P = .005) and 10% more
of their training time in heart-rate zone 1 than ACCURATE (15% ± 8% vs 5% ±
3%, P = .020). UNDER also spent 10% more of their total time training in heart-
rate zone 2 than OVER (17% ± 6% vs 7% ± 6%, P = .039). ACCURATE spent
14% more of their total training time in zone 5 than UNDER (17% ± 14% vs 3%
± 1%, P = .031). UNDER had a nonsignificantly lower average training heart rate
than ACCURATE (P = .052). All 3 groups spent a similar proportion of their total
training time in heart-rate zones 2 and 3.

Summated-Heart-Rate-Zones Versus Session-RPE Method

A correlation of r = .84 (95% CI: .70 to .92) was found when comparing the summated-
heart-rate-zones method with the session-RPE method in the calculation of total
training load (Figure 2; pooled data). The 95% CI for the correlation coefficients
determined for the session training loads of each subject were r = .86 to .94 (ie,
within-subject data).
24   Borresen and Lambert

Baseline anthropometric, physiological, and training characteristics are shown

in Table 3 for OVER (n = 5), UNDER (n = 4), and ACCURATE (n = 24), the 3
groups identified based on their position relative to the SEE of the regression
equation for the correlation between the session-RPE method and the summated-
heart-rate-zones method.

Table 3  Subject Characteristics for OVER (n = 5), UNDER (n = 4),

and ACCURATE (n = 24) for the Correlation Between the Session-
RPE Method and the Summated-Heart-Rate-Zones Method
(mean ± SD)a

Group

 Subject characteristic OVER UNDER ACCURATE

Age (years) 32 ± 6 27 ± 5 30 ± 5
Mass (kg) 55 ± 9 73 ± 14 68 ± 13
Height (cm) 167 ± 8 175 ± 12 171 ± 10
Body fat % 21 ± 3 17 ± 3 20 ± 8
VO2max (mL · min · kg )
–1 –1
54 ± 6 61 ± 7 56 ± 9
HRmax (beats/min) 179 ± 11 193 ± 9 191 ± 9
Peak treadmill running speed (km/h) 16 ± 3 18 ± 1 17 ± 2
Total training duration over 2 weeks (min) 944 ± 336b 698 ± 614 469 ± 296
Average training HR (beats/min) 143 ± 11 132 ± 10c 148 ± 12
Resting HR (beats/min) 50 ± 5 52 ± 10 56 ± 7
Percent time in zone 1 (%) 4 ± 3d 15 ± 8c 5±3
Percent time in zone 2 (%) 8±6 17 ± 6 12 ± 6
Percent time in zone 3 (%) 22 ± 7 30 ± 9 26 ± 11
Percent time in zone 4 (%) 45 ± 8d 25 ± 10 37 ± 9
Percent time in zone 5 (%) 17 ± 11 3 ± 1c 17 ± 13
Training load (RPE method, AU)e 3504 ± 1839 4717 ± 4164 2314 ± 1508
Training impulse (TRIMP, AU) 1974 ± 492b,d 907 ± 879 880 ± 470
Training load
(summated-HR-zone method, AU) 3226 ± 905b 1819 ± 1701 1555 ± 839
a
Groups were based on whether the regression equations over-, under-, or accurately predicted the
relationship between the methods (pooled data).
b
OVER significantly different from ACCURATE.
c
UNDER significantly different from ACCURATE.
d
OVER significantly different from UNDER.
e
This comparison should be interpreted with caution because the interpretation of RPE scales might
vary between individuals.
 Quantifying Training Load   25

OVER had a higher training load (summated-heart-rate-zones method, 3226 ±

905 vs 1555 ± 839 AU, P = .020), TRIMP (1974 ± 492 vs 880 ± 470 AU, P = .007),
and total training duration (944 ± 336 vs 469 ± 296 min, P = .044) than ACCU-
RATE and higher TRIMP than UNDER (1974 ± 492 vs 907 ± 879 AU, P = .043).
OVER spent 20% more of their total training time in zone 4 than UNDER (45% ±
8% vs 25% ± 10%, P = .018). UNDER had a lower average training heart rate than
ACCURATE (132 ± 10 vs 148 ± 12 beats/min, P = .048). UNDER spent 11% more
of their total training time in zone 1 than OVER (15% ± 8% vs 4% ± 3%, P = .013)
and 10% more of their total training time in zone 1 than ACCURATE (15% ± 8%
vs 5% ± 3%, P = .015). ACCURATE spent 14% more of their total training time
in zone 5 than UNDER (17% ± 13% vs 3% ± 1%, P = .028). All 3 groups spent a
similar amount of their total training time in heart-rate zones 2 and 3.

Discussion
The first finding of this study was that training load calculated for ad libitum train-
ing using the TRIMP equation correlated best with the summated-heart-rate-zones
method. This is not surprising, however, because both methods use the direct
physiological measure of heart rate as a fundamental part of the calculation. The
summated-heart-rate-zones method also correlated well with the session-RPE
method, explaining approximately 71% of the variance. The relationship between
TRIMP and the session-RPE method was the weakest of the 3 correlations, only
accounting for about 58% of the variance.
Foster et al12,15 evaluated the relationship between the subjective session-RPE
method and the objective summated-heart-rate-zones method of quantifying training
load during different forms of exercise. They found that although the summated-
heart-rate-zones method gave lower scores than the session-RPE method (because
of different units), the pattern of differences between the 2 methods was very consis-
tent; however, no correlation coefficients were provided by Foster et al.15 Individual
correlations between the 2 methods ranged between r = .75 and .90 in Foster et al,12
but no statistical methods were explained. They concluded that session RPE was a
valid and reliable measure of exercise intensity in aerobic exercise when compared
with heart-rate-based measues.12,15 Impellizzeri et al,9 however, found only moder-
ate correlations (r = .50 to .85 for individuals and r = .71 for mean team training
load) between training loads calculated using the session-RPE method and the
summated-heart-rate-zones method for members of a club soccer team. The content
of the 50 training sessions performed by 7 subjects in the study of Foster et al12 was
not explained, but the 12 subjects in the study of Foster et al15 performed 3 steady-
state and 5 interval sessions. Most training in the Impellizzeri et al9 study appears
to have consisted of “small-sided” games, with sprint and plyometric exercises
performed once a week. The increased use of the anaerobic-energy system in the
more intermittent-type exercise involved in soccer might contribute to an increase
in RPE, which might explain some of the difference in the correlations reported
in each study.9 Impellizzeri et al9 suggested that the RPE-based method cannot yet
replace heart-rate-based methods as a valid measure of exercise intensity because
only about 50% of the variation they measured in heart rate could be explained
by the session-RPE method. We could explain about 71% of the variance between
26   Borresen and Lambert

training load measured using the session-RPE and summated-heart-rate-zones

methods in subjects who engaged in 2 weeks of ad libitum training.
Because the absolute values or scores of training load for each of the methods
cannot be compared directly, because of differences in units as explained by Foster
et al,12 our conclusions are based on the position of the data points relative to the
regression line in Figures 1 and 2. The data points for the OVER group lie above
the regression line and thus indicate that the TRIMP method overestimated the
calculation of training load when compared with the session-RPE method for this
group. Conversely, the data points of the UNDER group lying below the regression
line indicate that the TRIMP method underestimated the calculation of training
load compared with the session-RPE method for the UNDER group.
There are many limitations to each of the 3 methods that might partially explain
the unaccounted-for variation between the objective and subjective methods of
quantifying training load. The complex interaction of many factors contributing
to one’s personal perception of physical effort might be something to consider
when assessing the variation between session-RPE and heart-rate-based methods.
These interactions might include hormone and substrate concentrations, personal-
ity traits, ventilation rate, neurotransmitter levels, environmental conditions, and
psychological states.16 Limitations such as these therefore need to be considered
when quantifying training using RPE-based methods. A further limitation was
expressed by Robinson et al,11 who observed no correlation between mean effort
ratings and mean relative training speed between subjects. They suggest that current
RPE scales might not be useful in comparing or prescribing training intensities for
different runners, but RPE scales might still be useful within individuals. Although
the limitations for using the session RPE do need to be acknowledged, the practical
value of this measure in quantifying training load is emphasized in situations when
measuring and monitoring training load are necessary but heart-rate monitors are
not available or when an easier means of reporting and calculating training load
is required. In these cases the session-RPE method might still give reasonably
accurate assessments of training load. RPE might also be a more valid measure
of training intensity than the HR-based methods when both aerobic and anaerobic
metabolic systems are activated.9
Foster et al12 suggest that the accuracy of the TRIMP equation might be limited
by the inability of heart-rate data to quantify high-intensity or non-steady-state
exercise such as resistance training, high-intensity interval training, or plyometric
exercise. This is because heart rate usually increases disproportionately during
resistance exercise and the cardiac responses required for the calculation of
TRIMP are not elicited; thus, objective measurements of heart rate cannot be used
to quantify resistance exercise’s intensity.17 The use of average heart rate in the
TRIMP equation might thus not adequately represent very high-intensity exercise.12
Results from our study might help in evaluating these suggestions. The correlation
between the TRIMP method and the session-RPE method of quantifying training
load revealed significant differences between the 3 groups, which were identified
based on their position relative to the SEE of the regression equation. Differences
in total TRIMP and training volume were found between the groups, along with
a nonsignificant difference in average heart rate. Because the TRIMP equation
includes average heart rate during exercise, maximum heart rate, resting heart rate,
and exercise duration, significant differences in the first 3 of these factors (because
 Quantifying Training Load   27

exercise duration is a common factor in both equations) might suggest a quantifiable

reason for the poor relationship between TRIMP and the session-RPE method in
groups OVER and UNDER in the current study. In addition, endurance training
has an effect on resting,18–20 submaximal,19,21,22 and possibly maximal23 heart rate.
Overtraining has also been found to decrease heart rate at the same submaximal
intensity.24 These changes might influence the usefulness of HR-based methods
for quantifying training loads, especially if these are going to be used to monitor
an athlete’s training over time.
The TRIMP equation also includes a weighting factor (Y) that is used to
emphasize high-intensity exercise in an attempt to avoid giving disproportion-
ate importance to higher volumes of low-intensity exercise compared with low
volumes of intense activity. Results from our study suggest that, compared with
the training load calculated using the session-RPE method, the TRIMP equation
might in fact be giving disproportionate importance to high-intensity exercise for
participants who spent a greater percentage of their total training time exercising
in higher intensity heart-rate zones (group OVER) and underestimating (compared
with the session-RPE method) the effect of low-intensity exercise on training
load for those who spent a greater percentage of their total training time training
in low-intensity heart-rate zones (group UNDER). The Y weighting factor in the
TRIMP equation is based on the lactate profiles of trained men and women as
exercise intensity increases.5 It has been shown, however, that the exponential
relationship between lactate and work load can change with training.25 As such,
the use of a standard weighting factor (Y) in the TRIMP equation, which is based
on a fixed lactate–workload relationship, might be inappropriate for quantifying
training load in subjects who differ in training status. It is possible that a different
weighting factor would be required for different training states. Thus, we suspect
that the use of the generic TRIMP equation for all the subjects in our study (who
varied in training status) might have contributed to the low correlations observed
between the session-RPE method and the TRIMP method.
In addition, the lactate response to exercise can be affected by many external
factors such as ambient temperature, dehydration, mode of exercise, exercise dura-
tion, intensity and the rate of change in exercise intensity, prior exercise, diet, and
muscle glycogen content.25,26 Improvements in training status and overtraining have
both been associated with decreases in maximal and submaximal blood lactate
concentration,27,28 which might lead to erroneous interpretations of lactate measure-
ments.26 For these reasons the fact that the TRIMP weighting factor is based on a
fixed lactate–workload relationship might introduce error in the quantification of
training load when an athlete’s training status changes over time or when comparing
training loads of subjects who differ with respect to training status.
The correlation between the summated-heart-rate-zones method and the ses-
sion-RPE method revealed differences among the 3 groups similar to those that
occurred between the TRIMP and session-RPE methods. The summated-heart-rate-
zones method focuses on the duration spent in 5 heart-rate zones and weights each
zone such that zone 1 is weighted the least and zone 5 is weighted the most. This
weighting system might limit the accuracy of this equation. Because a weighting
factor is applied to each zone comprising a range of heart rates, the lowest heart rate
and the highest heart rate in each zone will be weighted the same despite a differ-
ence in the physiological load. Under certain circumstances a change in heart rate
28   Borresen and Lambert

of only 1 beat/min will change the weighting factor of the zone, thereby increasing
or decreasing the calculated load disproportionately. It could be speculated from
these results that, similar to the TRIMP method, the summated-heart-rate-zones
method might disproportionately overestimate the impact that high-intensity exer-
cise has on training load and underestimate the effect of low-intensity exercise
on training load compared with training load calculated using the session-RPE
method. Another potential source of error in this equation is that the time spent
below 50% of HRmax is not included in the calculation. This might only affect
the accuracy of the calculation marginally (if at all) in high-intensity workouts,
but it is nonetheless worth noting, especially when quantifying training load for
submaximal or interval-training bouts.
The limitations of this study were that the subject numbers varied in each of the
3 groups (OVER, UNDER, and ACCURATE). Furthermore, the numbers were low
in the OVER and UNDER groups and the variances between groups were unequal,
thus requiring the use of less robust nonparametric statistics. Although differences
were found between groups using these statistics, there is a risk of making a type
1 error. We believe, however, that this risk was low because the magnitude of the
differences was in most cases large. Although the large variation in training load of
the subjects might be viewed as a weakness, a heterogeneous sample might in fact
strengthen the research design and reduce the risk of forming the wrong conclu-
sions, which might happen with a homogeneous or bias sample.
In conclusion, we found that training load calculated using the TRIMP equation
correlated best with training load calculated with the summated-heart-rate-zones
method, which is understandable because both are based on the same physiological
measure. Nonetheless, the session-RPE method also correlated well with the heart-
rate-based methods, and we therefore conclude that the subjective session-RPE
method of calculating training load remains useful. Further results suggest that in
athletes who spend a greater percentage of their training time doing high-intensity
exercise the TRIMP and the summated-heart-rate-zones equations might overesti-
mate training load compared with the session-RPE method. Conversely, in athletes
who spend proportionally more of their training time doing low-intensity exercise,
these heart-rate-based methods might underestimate training load when compared
with the training load calculated using the session-RPE method. An alternative
interpretation is that for athletes spending more time doing low-intensity exercise,
the session-RPE method might overestimate training load, whereas for other ath-
letes participating in proportionally more high-intensity exercise the session-RPE
method underestimates training load compared with the objective methods. This
will have to be confirmed in future studies.
Further investigation is also required to establish the exact cause of the poor
correlation between objective and subjective quantification methods found in
some athletes. Because the session-RPE method is a more global indication of the
difficulty of an exercise bout than the heart-rate-based methods, it may be that the
percentage variance not accounted for by the heart-rate-based methods represents the
numerous extraneous factors (other than heart rate) that contribute toward a person’s
personal perception of the difficulty of the session. Conversely, to our knowledge
none of the heart-rate-based equations have been validated, so it cannot be ruled
out that there might be inherent flaws in these equations that could affect their
relationship with the session-RPE method. It follows that if there are inherent flaws
 Quantifying Training Load   29

in the equations, further research could identify the modifications to the equations
required to more accurately calculate training load in each individual.

Practical Applications
The precise quantification of training load for each individual could contribute
to a more accurate assessment of how the athlete is responding to the prescribed
training and assist in subsequent changes to the training program. Heart rate has
previously been considered the most appropriate, practical method of quantify-
ing internal training load for aerobic activity. Nonetheless, some of the technical
problems associated with weighting heart rate in calculating training load might
negate the increased precision of the objective method. Another potential problem
of using a heart-rate-based method to quantify training load is a logistical one. For
example, if the heart-rate monitor does not record data for that session there will be
no information about the training load. In contrast, the session-RPE method does
not depend on equipment or technology, and therefore the risk of losing data about
a session is low. The practical usefulness of the session-RPE method in quantifying
training load is thus emphasized when monitoring training needs to be quick and
easy. This recommendation, however, needs to be considered in the context that
there are many factors that can affect an athlete’s personal perception of physical
effort and that these factors need to be considered when using RPE. Furthermore,
the validity, reliability, and sensitivity of this method still need to be established.

Acknowledgments
The research undertaken in this study was funded in part by the University of Cape Town,
Discovery Health, National Research Foundation, Ernst & Ethel Eriksen Foundation,
and Deutscher Akademischer Austausch Dienst.

References
1. Coyle EF. Physical activity as a metabolic stressor. Am J Clin Nutr. 2000;72(2,
Suppl):512S–520S.
2. Hellard P, Avalos M, Millet G, Lacoste L, Barale F, Chatard JC. Modeling the residual
effects and threshold saturation of training: a case study of Olympic swimmers. J
Strength Cond Res. 2005;19(1):67–75.
3. Avalos M, Hellard P, Chatard JC. Modeling the training-performance relationship using
a mixed model in elite swimmers. Med Sci Sports Exerc. 2003;35(5):838–846.
4. Skinner JS, Wilmore KM, Krasnoff JB, et al. Adaptation to a standardized training
program and changes in fitness in a large, heterogeneous population: the HERITAGE
Family Study. Med Sci Sports Exerc. 2000;32(1):157–161.
5. Banister EW. Modeling elite athletic performance. In: MacDougall JD, Wenger HA,
Green HJ, eds. Physiological Testing of the High-Performance Athlete. 2nd ed. Cham-
paign, IL: Human Kinetics; 1991:403–425.
6. Edwards S. The Heart Rate Monitor Book. Sacramento, CA: Fleet Feet Press; 1993.
7. Impellizzeri FM, Rampinini E, Coutts AJ, Sassi A, Marcora SM. Use of RPE-based
training load in soccer. Med Sci Sports Exerc. 2004;36(6):1042–1047.
8. Lucia A, Hoyos J, Santalla A, Earnest C, Chicharro JL. Tour de France versus Vuelta
a España: which is harder? Med Sci Sports Exerc. 2003;35(5):872–878.
30   Borresen and Lambert

9. Earnest CP, Jurca R, Church TS, Chicharro JL, Hoyos J, Lucia A. Relation between
physical exertion and heart rate variability characteristics in professional cyclists during
the Tour of Spain. Br J Sports Med. 2004;38(5):568–575.
10. Foster C, Daines E, Hector L, Snyder AC, Welsh R. Athletic performance in relation
to training load. Wis Med J. 1996;95(6):370–374.
11. Robinson DM, Robinson SM, Hume PA, Hopkins WG. Training intensity of elite male
distance runners. Med Sci Sports Exerc. 1991;23(9):1078–1082.
12. Foster C, Florhaug JA, Franklin J, et al. A new approach to monitoring exercise train-
ing. J Strength Cond Res. 2001;15(1):109–115.
13. World Medical Association. Declaration of Helsinki: ethical principles for medical
research involving human subjects. 2004. Available at: http://www.fda.gov/oc/health/
helsinki89.html.
14. Hopkins WG. A spreadsheet for calculating confidence intervals of correlation coef-
ficients. SportScience, 2006; 10. Available at: www.sportsci.org/resource/stats/index.
html.
15. Foster C. Monitoring training in athletes with reference to overtraining syndrome. Med
Sci Sports Exerc. 1998;30(7):1164–1168.
16. Williams JG, Eston RG. Determination of the intensity dimension in vigorous exercise
programmes with particular reference to the use of the rating of perceived exertion.
Sports Med. 1989;8(3):177–189.
17. Sweet TW, Foster C, McGuigan MR, Brice G. Quantitation of resistance training using
the session rating of perceived exertion method. J Strength Cond Res. 2004;18(4):796–
802.
18. Uusitalo AL, Uusitalo AJ, Rusko HK. Exhaustive endurance training for 6–9 weeks did
not induce changes in intrinsic heart rate and cardiac autonomic modulation in female
athletes. Int J Sports Med. 1998;19(8):532–540.
19. Wilmore JH, Stanforth PR, Gagnon J, et al. Heart rate and blood pressure changes
with endurance training: the HERITAGE Family Study. Med Sci Sports Exerc.
2001;33(1):107–116.
20. Wilmore JH, Stanforth PR, Gagnon J, et al. Endurance exercise training has a
minimal effect on resting heart rate: the HERITAGE Study. Med Sci Sports Exerc.
1996;28(7):829–835.
21. Hedelin R, Kentta G, Wiklund U, Bjerle P, Henriksson-Larsen K. Short-term overtrain-
ing: effects on performance, circulatory responses, and heart rate variability. Med Sci
Sports Exerc. 2000;32(8):1480–1484.
22. Skinner JS, Gaskill SE, Rankinen T, et al. Heart rate versus %VO2max: age, sex,
race, initial fitness, and training response—HERITAGE. Med Sci Sports Exerc.
2003;35(11):1908–1913.
23. Zavorsky GS. Evidence and possible mechanisms of altered maximum heart rate with
endurance training and tapering. Sports Med. 2000;29(1):13–26.
24. Snyder AC, Kuipers H, Cheng B, Servais R, Fransen E. Overtraining following intensi-
fied training with normal muscle glycogen. Med Sci Sports Exerc. 1995;27(7):1063–
1070.
25. Jacobs I. Blood lactate. implications for training and sports performance. Sports Med.
1986;3(1):10–25.
26. Swart J, Jennings CL. Use of blood lactate concentration as a marker of training status.
S Afr J Sports Med. 2004;16:3–7.
27. Jeukendrup AE, Hesselink MK. Overtraining—what do lactate curves tell us? Br J
Sports Med. 1994;28(4):239–240.
28. Urhausen A, Gabriel HH, Weiler B, Kindermann W. Ergometric and psychological
findings during overtraining: a long-term follow-up study in endurance athletes. Int J
Sports Med. 1998;19(2):114–120.