Journal of Electromyography and Kinesiology 20 (2010) 1036–1043

Contents lists available at ScienceDirect

Journal of Electromyography and Kinesiology

journal homepage: www.elsevier.com/locate/jelekin

Alternative methods of normalising EMG during cycling

Yumna Albertus-Kajee *, Ross Tucker, Wayne Derman, Michael Lambert
UCT/MRC Research Unit for Exercise Science and Sports Medicine, Department of Human Biology, Faculty of Health Sciences,
The University of Cape Town and the Sports Science Institute of South Africa, Cape Town, South Africa

a r t i c l e i n f o a b s t r a c t

Article history: We evaluated possible methods of normalisation for EMG measured during cycling. The MVC method,
Received 3 March 2010 Sprint method and 70% Peak Power Output Method were investigated and their repeatability, reliability
Received in revised form 16 July 2010 and sensitivity to change in workload were compared.
Accepted 16 July 2010
Thirteen cyclists performed the same experimental protocol on three separate occasions. Each day, sub-
jects firstly performed MVCs, followed by a 10 s maximal sprint on a cycle ergometer. Subjects then per-
formed a Peak Power Output (PPO) test until exhaustion. After which they cycled at 70% of PPO for 5 min
Keywords:
Electromyography
at 90 rpm. Results indicated that normalising EMG data to 70% PPO is more repeatable, the intra-class
Normalisation correlation (ICC) of 70% PPO (0.87) was significantly higher than for MVC (0.66) (p = 0.03) and 10 s sprint
Cycling (0.65) (p = 0.04). The 70% PPO method also demonstrated the least intra-subject variability for five out of
Repeatability the six muscles. The Sprint and 70% PPO method highlighted greater sensitivity to changes in muscle
Reliability activity than the MVC method. The MVC method showed the highest intra-subject variability for most
muscles except VM.
The data suggests that normalising EMG to dynamic methods is the most appropriate for examining
muscle activity during cycling over different days and for once-off measurements.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction 1999; Kollmitzer et al., 1999). Over the past few months there have
been a high number of publications investigating alternative nor-
Measuring biological variation in the EMG signal is important in malisation methods to the MVC method including, single leg
studies where surface EMG is used to gain understanding of phys- stance (Norcross et al., 2010), isokinetic squat jumps and 20 m
iological regulation, it is therefore necessary to minimise the vari- sprints (Ball and Scurr, 2010), maximal and sub-maximal contrac-
ation caused by intrinsic and extrinsic factors that affect the raw tions (Murley et al., 2010; Chapman et al., 2010). These publica-
signal (Burden and Bartlett, 1999). The extrinsic or experimenter- tions have all attempted to find alternative methods to
related factors can be controlled through the correct application normalising EMG during dynamic activity. However, normalisation
and placement of electrodes supported by a thorough understand- of EMG data during dynamic activities such as cycling is relatively
ing of electrophysiology and anatomy. Whereas the intrinsic fac- poorly understood, and there are presently few alternatives to the
tors including muscle fiber type, muscle fiber diameter and use of the isometric maximal voluntary contraction (MVC) method
length and the amount of tissue between muscle vary between (Marsh and Martin, 1995; MacIntosh et al., 2000; Ricard et al.,
subjects and cannot be controlled. A means in reducing the varia- 2006).
tion in the EMG signal is in part achieved through the appropriate The use of the MVC as a method of normalisation has theoreti-
method of normalisation. Normalisation involves the comparison cal constraints, since, during cycling there are variations in actions
of the EMG signal to a reference value obtained during standard- of the muscles involved. For example, changes in joint angle, joint
ised reproducible conditions, thus allowing comparison of activity angle velocity, and muscle lengthening and shortening all occur
between different muscles, across time, and between individuals. during cycling (Farina et al., 2004a), raising questions about the
Methods of normalisation for EMG activity measured during efficacy of the MVC method as a normalisation tool for EMG data
both static and dynamic exercise have been reviewed in detail measured during cycling (Farina et al., 2004a; Hunter et al.,
(Bolgla and Uhl, 2007; Knutson et al., 1994; Burden and Bartlett, 2002). This is particularly the case when the EMG activity is being
used in an attempt to quantify how much skeletal muscle may be
active. Researchers have in the past made direct comparisons be-
* Corresponding author. Address: UCT/MRC Research Unit for Exercise Science tween the EMG signal measured during cycling and the EMG activ-
and Sports Medicine, Sports Science Institute of South Africa, P.O. Box 115, ity measured during an isometric MVC, expressing the resultant
Newlands 7725, South Africa. Tel.: +27 21 6504572; fax: +27 21 6867530.
value as a percentage of maximum (for example, 45% of the muscle
E-mail address: [email protected] (Y. Albertus-Kajee).

1050-6411/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jelekin.2010.07.011
 Y. Albertus-Kajee et al. / Journal of Electromyography and Kinesiology 20 (2010) 1036–1043 1037

was active at volitional exhaustion). Quite clearly, this comparison Watkins, 1990). Reliability also refers to the stability and
is questionable on the grounds that the two types of muscle con- consistency of measures with respect to time so that
tractions are very different in nature. This creates the need to eval- changes between the measurements can be attributed to
uate other possible means for normalising EMG activity during the intervention (Keskula et al., 1995). One of the main con-
cycling. tributors to reliability is the within-subject variation which
The MVC method has proved popular amongst researchers (Ri- affects the precision of estimates of change in the variable
card et al., 2006; Potvin, 1997; Marsh and Martin, 1993; Ericson of an experimental trial (Hopkins, 2000). The coefficient of
et al., 1986; Arsenault et al., 1986; Dubo et al., 1976) but has cer- variation (CV) is the statistical method often used to evalu-
tain limitations. For example, it is assumed that subjects provide ate intra-subject variability.
maximum effort during testing and that the maximal contraction (c) Sensitivity
achieved represents 100% of muscle activity (Burden and Bartlett, Sensitivity is defined as the ability to detect true biological
1999). Although researchers can encourage the subject to generate variations. In the context of this study, the ability of the
a MVC, the force generated ultimately depends on the level of method to assist in the measurement of actual change in
motivation of the subject. muscle activity rather than the measurement of the combi-
Furthermore, the MVC typically involves an isometric knee nation of the biological and experimental factors that affect
extension movement at an angle of 60° from full extension, a the signal, is a requirement of a good method of normalisa-
movement which does not occur during cycling and can be argued tion. It is reasonable to expect muscle force output and EMG
to be non-functional. Despite these limitations few studies have amplitude to have a linear relationship, as they both depend
addressed the appropriateness of the MVC method or proposed on the number of motor units recruited (Farina et al., 2004b).
alternative methods of normalisation during exercise trials. One Thus an ideal method of normalisation should to some
of the first studies that investigated alternative methods of nor- extent, be able to identify changes in muscle activity due
malisation was done by Hunter et al. (2002), investigating EMG to the change in workload or exercise intensity, if fatigue
methods of normalisation in cycling by using two angles (60° is not present.
and 180°) with four types of pedal contractions. The investigation
found that the isometric MVC produced a higher integrated EMG It is important that the EMG measured during the normalisation
(iEMG) value than the protocols using knee angles at 60°, 180° or procedure is consistent from 1 day to the next (repeatability), and
one dynamic maximal cycle pedal revolution. This finding suggests also that when the EMG activity measured during dynamic activity
that muscle activity during cycling does not provide the highest is normalised against the chosen procedure, the value obtained
EMG amplitude. This study however did not investigate the repeat- must reflect changes in EMG that occur as the exercise workload
ability of the various normalisation protocols but rather only inves- increases (sensitivity).
tigated which of the four protocols produced the greatest iEMG Thus the aim of the study was to examine and evaluate possible
activity, and therefore has the same potential limitation with re- methods for normalisation of EMG measured during cycling,
gards to the comparison between cycling exercise and static, iso- according to the requirements for a good normalisation technique
metric contractions. A more recent study by Rouffet and Hautier which are described above. Three methods for normalising EMG
(2008) investigated the use of maximal cycling sprints compared during cycling, which are based on previous research (Bolgla and
to isometric MVC’s for EMG normalisation in cycling. The study Uhl, 2007; Yang and Winter, 1984, 1983), were evaluated and their
found that the maximal sprint was as repeatable as the MVC meth- variability and sensitivity to change was compared to the method
od in measuring peak EMG amplitude. In addition they highlighted of normalisation using isometric MVC.
the fact that the sprint method was less time and energy consum-
ing. This study however did not measure between day repeatability  MVC method – maximal static normalisation and is the current
of the normalisation methods, only the within day repeatability. widely-used ‘‘standard” method
Therefore the most appropriate method for normalising muscle  Sprint method – maximal dynamic normalisation
activity during cycling has yet to be established, especially for  70% Peak Power Output Method – sub-maximal dynamic normal-
studies conducted over a number of days. To address this chal- isation, this method was based on studies that found sub-max-
lenge, the first requirement was to determine an appropriate imal MVC’s to be more repeatable and reliable than 100% MVC’s
method of normalisation that fulfils certain criteria namely. (Kollmitzer et al., 1999; Mathur et al., 2005; Yang and Winter,
1983). We therefore aimed to investigate the repeatability and
(a) Repeatability reliability of using a sub-maximal dynamic method.
An appropriate method of normalisation would need to be
highly repeatable. Repeatability should consider two differ- 2. Methods
ent and complementary aspects; (1) reliability, which
addresses between-day variations of measured variables; 2.1. Subject selection
and (2) constancy, which addresses within-day variations
of measured variables. Repeatability would ensure precision Thirteen well-trained cyclists (age 27 ± 8 years; height 1.8 ± 0.1;
during the measurement and lowest variation in repeated mass 72 ± 6 kg) were recruited from local cycling clubs to partici-
trials with altered electrode positioning (Burden et al., pate in this study. Subjects were included if they were between
2003; Rainoldi et al., 2001). Intra-class correlation (ICC) is the ages of 18–35 years old and if they were able to complete a lo-
the most commonly used statistical method to identify cal 109 km cycling race in less than 3 h 30 min. The study was ap-
repeatability of the EMG signal (Bolgla and Uhl, 2007; proved by a local Research and Ethics Committee and performed in
Laplaud et al., 2006; Mathur et al., 2005; Rainoldi et al., accordance with the principles of the Declaration of Helsinki
2001; Ng and Richardson, 1996). A high ICC (closest to 1) (2008). All subjects signed the informed consent form and were
is associated with a small within-subject variance relative asked to maintain their same physical activity pattern over the
to the between-subjects variance (Hopkins, 2000). course of the study and not to begin any new training or recrea-
(b) Reliability tional programs. Subjects performed the same experimental proto-
Reliability refers to the extent to which measurements are col on each of the three testing days. Each testing day was
consistent, dependable, and free from error (Portney and separated by 5–7 days of normal training.
1038 Y. Albertus-Kajee et al. / Journal of Electromyography and Kinesiology 20 (2010) 1036–1043

2.2. Experimental design USA). Two electrodes (Blue Sensor, Medicotest, Denmark) were
placed on the belly of the following lower limb muscles during
Subjects were required to visit the laboratory on four separate all the tests; Vastus lateralis (VL), Vastus medialis (VM), Rectus
occasions. During the first visit, a complete anthropometrical femoris (RF), Biceps femoris (BF) and Medial and Lateral gastrocne-
assessment and familiarisation with all the equipment was per- mius (MG and LG, respectively). Prior to placing the electrodes on
formed with each subject. On trial days, subjects were firstly pre- the skin, the skin over the muscle was shaved and cleaned with
pared for the placement of the electrodes on six muscles (Vastus ethanol. The placement and location of the electrodes are accord-
medialis (VM); Vastus lateralis (VL), Rectus femoris (RF), Biceps ing to the recommendations by SENIAM (Surface EMG for non-
femoris (BF) and Medial and Lateral gastrocnemius (MG and LG, invasive Assessment of Muscles) (Hermens et al., 2000). Therefore,
respectively)) on the right leg. Following this each subject per- two electrodes were carefully taped to the belly of each muscle,
formed three maximum voluntary contractions (MVC) using the parallel to the muscle fibers with an inter-electrode distance of
Biodex Dynamometer 3 (Biodex Medical Systems, New York). They 20 mm. A telemetric signal was relayed to an antenna connected
were then allowed a 10 min warm-up on the electronically braked to an online computer and captured at 2000 Hz. The wire-leads
cycle ergometer (Lode, Groningen, Holland), after which they per- connected to the electrodes were well secured with tape to avoid
formed a 10 s maximal sprint, where each subject started the artefacts from lower limb movements during cycling.
sprint from a power output of 200 Watts (W). After a 30 min rest
period, subjects performed their Peak Power Output (PPO) test un- 2.6. EMG analysis
til exhaustion on the ergometer (Hawley and Noakes, 1992). After
another 30 min rest period, each subject then cycled at 70% of their The raw digital EMG signal was processed using Noraxon’s
PPO for 5 min at 90 revolutions per minute. Myoresearch software (Version 2.11). The raw EMG data was fil-
tered using a 50 Hz notch filter to remove any electrical interfer-
2.3. Maximal voluntary contractions ence from external sources (MyoResearch 2.02). The signal was
then filtered a second time using a 15–500 Hz band pass filter. This
Before the start of each trial, each subject’s peak isometric force allowed noise or movement interference below 15 Hz and other
of the lower limb was measured using the Biodex Dynamometer, non-physiological signals above 500 Hz to be removed. The data
with simultaneous EMG recordings captured for muscle activity were smoothed using route mean squared analysis (RMS), which
measurement. The subject’s hips, thighs and upper bodies were was calculated for a 50 ms window.
firmly strapped to the seat of the dynamometer. The subjects’
crossed their arms over their chests for each test. All isometric tests 2.7. EMG analysis of PPO trial
were conducted with the knee at an angle of 60° from full leg
extension for the VM, RF and VL as well as BF muscles. The lateral EMG was recorded for 10 s halfway through each workload and
condyle of the femur was carefully aligned with the axis of rotation at exhaustion during the PPO trials. From this recording 3 s of EMG
of the dynometer and the distal end to the calf was secured to the data were extracted and analysed (used for MVC, Sprint and 70%
lever arm of the dynamometer using a padded Velcro strap. When PPO method of normalisation). The EMG value (lv s) was calcu-
measuring the force output for the MG and LG, the right leg was lated for each contraction and then time-normalised to 1 s
elevated with the thigh resting on a holder and the foot firmly [(lv s) s 1] (referred to as mean amplitude).
strapped against a foot-plate, setting a knee angle of 30°. The ankle
was carefully aligned to the axis of rotation of the dynamometer. 2.7.1. MVC method
The ankle at a 15° angle from full vertical (reference plane at When normalising using the MVC method, 3 s of EMG data were
0°), from where the subjects had to push against the foot-plate extracted from the middle period of the 5 s MVC which produced
using the gastrocnemius muscles. A standardised warm up in- the highest force. The amplitude of the 3 s periods taken from each
cluded four isometric contractions of the knee and calf muscle workload measurement during the PPO was normalised to the 3 s
extensors at 50%, 60%, 70% followed by 80% for each subject’s sub- period obtained during the MVC.
jective maximum. The isometric test included three maximum vol-
untary contractions (MVC) of 5 s each separated by 60 s intervals. 2.7.2. Sprint method
Subjects were verbally motivated to encourage them to achieve When normalising to the 10 s sprint, three 3 s period within the
their maximum potential. The seating position and length of leg 10 s measurement (taken in the sprint) were isolated and aver-
attachments of subjects were standardised over all 3 days. aged. The averaged amplitudes of these sections were used for
normalisation.
2.4. Peak Power Output test
2.7.3. 70% PPO method
Subjects performed their PPO test by starting at workload of When normalising to 70% PPO, EMG was captured for 10 s at the
3.33 W kg 1 body mass. After 150 s, the workload increased by end of each minute (of cycling at 70% PPO for 5 min), and 3 s peri-
50 W and then increased by 25 W every 150 s until exhaustion. ods were then analysed from each 10 s measurement, and the
Exhaustion was defined as the inability to maintain pedalling fre- resulting amplitude was then averaged (Fig. 1).
quency above 60 revolutions per minute (rpm). PPO was defined
as the highest power output the subject completed for 150 s, plus 2.8. Statistical analysis
the fraction of time spent in the final work rate multiplied by 25 W.
The position on the bicycle was standardised for each subject over All results were analysed using a statistical software pro-
the 3 days of testing. Subjects were also required to remain seated gramme (Statistica 7, StatSoft, Tulsa, OK, USA). Results were ex-
throughout the PPO trials. pressed as means ± standard deviation (SD) The repeatability was
investigated using the Intra-class Correlation Coefficient (ICC).
2.5. EMG data collection The ICC scores ranged between R = 0.80 and 1.00 were defined as
representing ‘‘good” reproducibility, scores between R = 0.60 and
The EMG activity of the muscles was recorded using the tele- 0.79 a ‘‘fair” reproducibility and less than R = 0.60 a ‘‘poor” repro-
metric EMG system (Telemyo 900, Noraxon, USA, Inc., Arizona, ducibility (Sleivert and Wenger, 1994). The 95% confidence
 Y. Albertus-Kajee et al. / Journal of Electromyography and Kinesiology 20 (2010) 1036–1043 1039

Table 1
Mean Intra-class correlations for Peak Power Output EMG normalised to MVC, Sprint,
70% PPO methods.

Muscle MVC Sprint 70% PPO

VM 0.70 (0.52–0.88) 0.75 (0.59–0.91) 0.88 (0.80–0.97)
RF 0.47 (0.11–0.85) 0.56 (0.35–0.77) 0.82 (0.71–0.93)
VL 0.82 (0.76–0.87) 0.70 (0.61–0.79) 0.91 (0.85–0.97)
BF 0.87(0.84–0.91) 0.80 (0.72–0.88) 0.87 (0.83–0.92)
MG 0.86 (0.82–0.92) 0.90 (0.86–0.94) 0.88 (0.83–0.94)
LG 0.83 (0.78–0.88) 0.82 (0.77–0.88) 0.92 (0.89–0.96)

Mean ICC values of methods of normalisation for each muscle; Vastus medialis
(VM), Rectus femoris (RF), Vastus lateralis (VL), Biceps femoris (BF), Medial gas-
trocnemius (MG) and Lateral gastrocnemius (LG), over 3 separate days of cycling.
Values presented as mean (of six work loads) with 95% confidence intervals. Bolded
values indicate the highest ICC value for each muscle.

normalisation with the highest percentages greater than

CV = 20%, has the most intra-subject variability. The 70% PPO
method showed the least intra-subject variability for five out of
six muscles (VL, RF, BF and LG) (Table 2). The Sprint method the
range of <12%. The MVC method showed the highest percentage in-
Fig. 1. Schematic guideline for normalising EMG during cycling that is suitable for
tra-subject variability for all muscles except VM, where most of the
the study design, research question and skeletal muscles being investigated. The
muscles are displayed from highest to lowest ranking order. The muscles CV values were greater than 20% (ranging between 46% and 77%).
corresponding to the Sprint method is highlighted in grey.

3.3. Sensitivity
intervals were calculated for the ICC using software downloaded
from http://www.newstats.org (Hopkin, 2007). Analysis of Vari- For VM over six loads (Table 3), the Sprint method showed mus-
ance (ANOVA) with repeated measures was used to detect signifi- cle activity to be significantly different over more loads than the
cant differences between the methods of normalisation. MVC and 70% PPO method. The Sprint method found muscle activ-
Sometimes there was a negative ICC meaning the with-in subject ity in loads 1–4 to be significantly lower than the final load,
variance exceeded the between subject variance, thus ICC R = 0 whereas MVC and 70% PPO methods were only able to detect loads
which indicates no repeatability (Larsson et al., 1999). A Tukey post 1–3 to be significantly lower than the final load.
hoc test was used to detected differences in muscle activity be- The Sprint and 70% PPO methods showed loads 1 and 2 to be
tween the six stages of the PPO test. The method of normalisation significantly lower than the final load for VL, whereas the MVC
variability over 3 days for each subject (intra-subject variability) method found no significance between the workloads. For RF, the
was assessed using coefficient of variation (CV) analysis. Accept- MVC and 70% PPO methods showed similar findings where loads
able variability has been defined as CV values less than 10% (Win- 1 and 2 were significantly lower than 4, 5 and final.
ter, 1991; Menz et al., 2004), however Taylor and Bronks (1995) For BF (Table 3), the Sprint method showed the greatest sensi-
have shown CV values between 9% and 12% for the quadriceps tivity in detecting change in muscle activity over the workloads,
muscles. Thus, as a cautionary measure, the definition for an
‘‘acceptable” CV value is regarded as less than 12% and ‘‘unaccept-
Table 2
able” as greater than 20%. p < 0.05 has been considered as signifi- Percentage of intra-subject coefficient of variation in each range.
cant in all statistical test.
Muscles Norm Method 612% 12> 620% >20%
VM MVC 8 46 46
3. Results
Sprint 58 33 8
70%PPO 77 8 15
3.1. Repeatability
VL MVC 8 15 77
Sprint 46 31 23
The 70% PPO method was the only method of normalisation to 70%PPO 54 23 23
yield ‘‘good” ICC values for all muscles, with average ICC values for RF MVC 8 17 75
all muscles being greater than R = 0.82 (Table 1). The MVC method Sprint 0 38 61
was repeatable for MG, LG, VL and BF as seen by average ICC values 70%PPO 15 61 23
greater than 0.80. This method however, showed ‘‘poor” repeat- BF MVC 0 25 75
ability for RF (R = 0.47 (0.11–0.85)). The Sprint method showed Sprint 8 54 39
similar repeatability patterns as the MVC method, where MG and 70%PPO 17 75 8

LG had ‘‘high” average ICC values, (R = 0.90 (0.86–0.94) and MG MVC 15 38 46

R = 0.82 (0.77–0.88), respectively), and VM, VL and BF had only Sprint 61 23 15
70%PPO 54 23 23
‘‘fair” repeatability (average ICC between R = 0.70 and 0.75) but
slightly less ICC values than that derived from the MVC method. LG MVCSprint 15 15 69
70%PPO 31 31 38
47 15 38
3.2. Intra-subject reliability
Percentage of CV values in each range for each method of normalisation and
muscle; Vastus medialis (VM), Vastus lateralis (VL), Rectus femoris (RF), Biceps
The method of normalisation with the highest percentage of CV femoris (BF), Medial and Lateral gastrocnemius (MG and LG), respectively. The
values that is less than CV = 12% is seen to show the least intra- methods with highest percentages in each CV range for each muscle, is highlighted
subject variability over three trials. Furthermore, the method of in bold.
1040 Y. Albertus-Kajee et al. / Journal of Electromyography and Kinesiology 20 (2010) 1036–1043

Table 3 0.82 and 0.91, respectively) when normalised to the sub-maximal

Muscle activity which is significantly different between the six power output loads dynamic 70% PPO method (Table 3). Laplaud et al. (2006), whose
when normalised to maximal voluntary contraction (MVC), Sprint and 70% PPO
methods.
method of normalisation involved expressing root mean square
(RMS) values as a percentage of the maximal RMS value recorded
Muscle MVC Sprint 70% PPO during the incremental cycling trial, found RF to be the least
VM F load >1–3 F load >1–4 F load >1–3 repeatable. It is important to note that the present study showed
5th Load >1–2 5th Load >1–2 5th Load >1–2 RF to have low ICC values when normalised to the ‘maximal’ meth-
VL F load >1–2 F load >1–2 ods of normalisation of the MVC and Sprint (Table 1). However, the
5th Load >1–2 5th Load >1–2 sub-maximal 70% PPO method did show RF to have good repeat-
RF F load >1–3 F load >1–3 F load >1–3 ability where ICC was R = 0.82 (0.71–0.93). The fact that RF is a
5th >1–3 5th Load >1–3 5th Load >1–3 biarticulate muscle which is involved in both leg extension and
4th >1–2 4th Load >1 4th Load >1–2
flexion cannot be the sole reason for the variability found in the
BF F load >1–2 F load >1–2 F load >1–2 measurements. If this was the reason, then the BF which is in-
5th Load >1–2 5th Load >1–2 5th Load >1–2
4th Load >1
volved in leg flexion and extension would show similar trends.
Yet in the present study, BF had good repeatability for all three
MG NS NS NS
LG NS NS NS
methods of normalisation (Table 1). Another possible reason for
RF variability could be that RF comprises a higher percentage of
Summary of the post hoc repeated measures analysis indicating the significant fatigable fast twitch muscle fibers (Mathur et al., 2005; Komi and
differences between the power output loads of the PPO test. The bold results
Tesch, 1979), which would explain the high variability found using
indicate the most sensitive method for each muscle. NS indicates no significant
differences between the loads. F, final load. the ‘maximal’ and fatiguing methods of normalisation used in the
present study and in the study conducted by Laplaud et al.
(2006). The use of the sub-maximal 70% PPO method may reduce
where load 1 was significantly lower than loads 4, 5 and final. 70% this variability by activating fewer fast twitch fibers and not fatigu-
PPO and MVC method were only able to detect significant differ- ing these fibers to the same degree as maximal methods of
ences of load 1 compared to the 5th and final load. None of the normalisation.
methods were able to detect significant differences in muscle Moreover, the fact that the 70% PPO Method is dynamic in nat-
activity between the workloads for MG and LG, even though the ure may result in a more relevant representation of muscle activity
MVC and Sprint methods showed a significant stage effect during the trial. This method is representative of regular neural
throughout the trial for MG (p = 0.029 and p = 0.032, respectively). muscle control strategy (Felici, 2004) and takes into account the
change in muscle length during cycling. Kellis and Baltzopoulos
(1996) found dynamic method of normalisation to be more
4. Discussion appropriate because this method considers the effects of muscle
activation, muscle length and angular velocity on EMG. Yang and
The aim of the study was to examine three different methods of Winter (1984) investigated various methods of normalisation dur-
normalising EMG activity measured during incremental cycling to ing gait. They normalised gait EMG to isometric methods; 50% MVC
exhaustion, with specific emphasis on identifying the most repeat- (since they previously showed 100% MVC to be less reliable) and
able and suitable one for dynamic cycling exercises. These methods mean EMG per unit isometric force (lv Nm 1). They also used dy-
included normalisating to an isometric MVC (MVC method) and namic methods, peak and mean within-subject ensemble averages.
two methods with normalisation to cycling (Sprint method and Their study found the dynamic methods reduced inter-subject
70% PPO method). variability.
The result of the present study and previous research pertaining
4.1. Repeatability of the methods of normalisation to the repeatability of the methods indicates that the use of static
isometric methods is not as appropriate for normalisation in dy-
Repeatability is an indication of reproducibility, thus the meth- namic cycling exercise. The dynamic methods of normalisation,
od should produce similar results over different testing trials predominantly the 70% PPO method, were found to be ‘highly’
(within day and between days). The primary finding of this study repeatable for most muscles.
was that while there are variations between the specific muscles
and with specific loads, the highest repeatability of EMG activity 4.2. Intra-subject reliability of the methods of normalisation
of the lower limb muscles was found when using the 70% PPO
Method (Table 1). Since the 70% PPO method utilises a sub-maxi- The reliability tested in this study refers to the reproducibility
mal exercise profile, this finding is consistent with earlier research of EMG activity of the same individual performing repeated trials
that showed sub-maximal exercise to be more repeatable than (Hopkins, 2000). This study found the best intra-subject reliability
maximal efforts in isometric contractions. Mathur et al. (2005) of EMG amplitude when normalising the cycling trials to 70% PPO
studied the repeatability of RF, VL and VM during sustained iso- method. For most muscles (VM, VL, RF, BF and LG) (Table 2),
metric contractions at 80% and 20% MVC maintained to exhaustion. summed percentage scores ranged from 14% to 21%. The Sprint
EMG amplitude was reliable (moderate to high) during 20% MVC method produced similar intra-subject CV values to that of the
where ICC ranged from 0.58 to 0.99. In contrast, the 80% MVC trial 70% PPO method, and the MVC methods had the highest variability
had a greater degree of variability and produced low ICC values. for all muscles (as seen by the highest scores produced). These
Likewise, Kollmitzer et al. (1999) showed EMG to be more repeat- findings suggest that EMG amplitude during cycling is most reli-
able for 50% MVC knee extensions than 100% MVC, where RF was able when normalised to dynamic methods, preferably the sub-
more reliable than VL and VM. maximal methods.
With respect to specific muscles, Mathur et al. (2005) found no This finding is similar to that of Bolgla and Uhl (2007). They
differences in the repeatability of VM, VL and RF EMG activity dur- investigated the repeatability and reliability of three methods of
ing the 20% MVC, which they attributed to the lower force require- normalisation; MVC, Mean dynamic (m-Dyn) and Peak Dynamic
ment whilst performing the task. The present study showed that (pk-Dyn) methods for analysing hip abductor rehabilitation exer-
VM, RF and VL had high ICC (averaged over six loads, R = 0.88, cises on healthy subjects. Subjects performed three open kinetic
 Y. Albertus-Kajee et al. / Journal of Electromyography and Kinesiology 20 (2010) 1036–1043 1041

and three closed kinetic chain hip abductor exercises. Intra-subject and LG. This could possibly be answered by studies investigating
CV’s ranged between 11% and 22% for all methods; however the m- the effects of cycling cadences on muscle activity. A study con-
Dyn method provided the lowest intra-subject CVs for five out of ducted by Marsh and Martin (1995) showed that an increase in ca-
the six exercises performed. Interestingly the pk-Dyn method dence resulted in a significant increase in average EMG amplitude
had the highest CV values. Based on these results, Bolgla and Uhl of the gastrocnemius muscles despite the cyclists’ cycling experi-
(2007) agreed with other studies (Burden et al., 2003; Winter ences. This study was in agreement with study conducted by Goto
and Yack, 1987; Yang and Winter, 1984), that the m-Dyn method et al. (1976), which found gastrocnemius EMG amplitude increased
provides greater measurements of reliability. Recently Rouffet linearly with an increase in cadence. More interestingly, Gregor
and Hautier (2008) and Ferndández-Peña et al. (2009) showed dy- et al. (1985) showed gastrocnemius activation to be associated
namic methods of normalisation to have lower intra-subject CVs. more strongly with cadence changes than changes in power out-
Rouffet and Hautier (2008) explains that by normalising to an all put. They found similar gastrocnemius EMG amplitude when cy-
out torque velocity test (maximal sprint), each muscle is being cal- cling at 90 rpm for power outputs of both 90 W and 270 W,
ibrated in reference to the maximal activity obtained when its acti- despite significant power output differences; for instance when
vation is regulated by identical afferent feedback and neural EMG amplitude was lower when cycling at a cadence of 60 rpm
processing. at 180 W. Even though the cadence was not recorded during the
It is of interest to note is that in the present study the maximal trials in the present study, it can thus be assumed that all the cy-
methods of normalisation (MVC and Sprint methods) show the clists maintained their cadence until exhaustion, since no method
highest variability, which could possibly be explained by an in- of normalisation showed any significant changes in gastrocnemius
verse relationship existing between force production and force var- activity throughout the PPO trials.
iability (Mathur et al., 2005), where higher forces show greater There are certain limitations in this study, not all the major low-
variability and hence increased intra-subject variability (Ng and er extremity muscles were measured due to limited EMG recording
Richardson, 1996). By performing a maximal task/contraction, we channels. We therefore focused on measuring the main quadriceps
assume the person is performing and providing a maximal effort, and gastrocnemius muscles. Furthermore, in the 70% PPO method,
which adds to the variability of the method (Yang and Winter, the 70% is an arbitrary intensity selection for a sub-maximal load;
1983) as well as the motivational factors (Komi and Buskirk, therefore other sub-maximal loads (65%, 75% etc.) could possibly
1970). In addition, EMG measurements from MVC’s have previ- provide the same or different results.
ously been shown to be less reliable than sub-maximal contrac- In summary, the most appropriate method of normalisation is
tions (Mathur et al., 2005; Yang and Winter, 1984; Kollmitzer one that reflects reproducibility over different testing times, has
et al., 1999), and because some individuals are able to activate low variation and the ability to detect changes in EMG amplitude
their muscles closer to the maximal level of activation than others with changes in exercise intensity. The following offers evidence-
(Gandevia, 2001). Felici (2006) explained that MVC was a poor rep- based guidance (based on this study) for the selection of the most
resentation of muscle function during normal activity since it rep- appropriate method of normalisation for different research de-
resents a simplified version of movement. This could be a possible signs. Fig. 1 presents a schematic guideline for normalising EMG
reason for the high intra-subjects CV’s found for the MVC methods. during cycling studies. If a researcher is investigating muscle
The finding of the present study is in agreement with previous activity on a once-off cycling trial, and intends to measure maxi-
research that has found dynamic methods of normalisation to have mal muscle activity or changes in muscle activity, the Sprint
less variability than the maximal static isometric method. method of normalisation is recommended. This method has dem-
onstrated ‘fair’ to ‘good’ repeatability for most of the muscles
4.3. Sensitivity of the methods of normalisation measured, where the MG and LG have shown to have the highest
repeatability. The MG and VL are the recommended muscles to
Sensitivity of the method of normalisation refers to the meth- use for low intra-subjection variation. The VM and BF are the
ods’ ability to track changes in workload or exercise intensity. most sensitive muscles in tracking changes in workload, followed
For this analysis, the EMG measurement was able detect differ- by the remaining four muscles (VL, RF, MG and LG). Important to
ences in EMG activity with increasing workload. This however does note is that RF is not recommended as a muscle to measure when
not take into account the presence of fatigue and the affect this has investigating maximal muscle activity, since RF has ‘poor’ repeat-
on the EMG amplitude. Taylor and Bronks (1995) found during ability and intra-subject reliability. However, RF activity does
incremental cycling, the relationship between EMG amplitude track changes in workload to the same extent as does VL, MG
and exercise intensity to be linear. There been additional studies and LG.
that have found EMG amplitude increases with increasing power It is important to point out that the specific, practical applica-
output/workload (MacIntosh et al., 2000; Ericson, 1986; Ericson tion of measured EMG data is also dependent on the research ques-
et al., 1985; Bigland-Ritchie and Woods, 1974). tion and the nature of the study being performed. For example, for
Since muscle activity has been shown to be repeatable during a once-off trial in which the absolute EMG activity or percentage of
incremental cycling (Laplaud et al., 2006; Gamet et al., 1996), it total active muscle is to be estimated using surface EMG measure-
could be assumed that the changes found in muscle activity during ments, repeatability is less important than intra-individual CV.
each cycling load was highlighted more by the type of method of Rather, in this situation, it is important to use a method of normal-
normalisation used and not entirely due to the variation in muscle isation that represents what may be considered a true maximum of
activity caused by fatigue or from the intrinsic and extrinsic influ- skeletal muscle activation. In contrast, if the study is designed to
ences on the EMG signal. In this study the aim was to measure the measure physiological/biological changes as a result of some inter-
ability of each method to detect changes in muscle activity over vention or fatigue process, then the reliability, repeatability and
five stages of the PPO test and also during the final stage of the sensitivity become more important. Therefore, in considering the
PPO trial. Sprint method (Table 3) was the most sensitive in detect- optimal normalisation technique, the three previously identified
ing changes in muscle activity over all the six stages for most mus- factors (repeatability, reliability and sensitivity) must be inter-
cles except MG and LG. However, this method was not highly preted in the context of the overall research question.
repeatable for RF and VL (Table 1). Accordingly, the present study has shown that normalising
It is important to note that no method was able to detect EMG during cycling to the 70% PPO method meets the three
changes in EMG amplitude over the six power out stages for MG requirements and is therefore recommended for studies that are
1042 Y. Albertus-Kajee et al. / Journal of Electromyography and Kinesiology 20 (2010) 1036–1043

investigating changes in muscle activity over different days, where Hawley JA, Noakes TD. Peak power output predicts maximal oxygen uptake and
performance time in trained cyclists. Eur J Appl Physiol Occup Physiol
reliability and repeatability of measurements becomes more
1992;65(1):79–83.
important. However, for investigating the maximal muscle activity Hermens HJ, Freriks B, Disselhorst-Klug C, Rau G. Development of recommendations
during a trial, the Sprint method of normalisation is recommended, for SEMG sensors and sensor placement procedures. J Electromyogr Kinesiol
since it provides a reliable method that is similar in nature to the 2000;10(5):361–74.
Hopkins WG. Measures of reliability in sports medicine and science. Sports Med
subsequent cycling trial. Further, the Sprint method was sensitive 2000;30(1):1–15.
to changes in workload. For a once-off trial where the absolute le- Hopkins WG. A new view of statistics, http://www.sportsci.org/resources/stats;
vel of EMG activity (and hence muscle activity) is to be indirectly 2007 [accessed 01.03.2010].
Hunter AM, St Clair GA, Lambert M, Noakes TD. Electromyographic (EMG)
measured, it is still recommended that researchers utilise the normalization method for cycle fatigue protocols. Med Sci Sports Exerc
Sprint method as a method of normalisation for cycling activity. 2002;34(5):857–61.
Kellis E, Baltzopoulos V. The effects of normalization method on antagonistic
activity patterns during eccentric and concentric isokinetic knee extension and
Acknowledgments flexion. J Electromyogr Kinesiol 1996;6(4):235–45.
Keskula DR, Dowling JS, Davis VL, Finley PW, Dell’Omo DL. Interrater reliability of
isokinetic measures of knee flexion and extension. J Athl Training
This study was funded by the National Research Foundation of 1995;30(2):167–70.
South Africa, University of Cape Town Postgraduate Funding, DAAD Knutson LM, Soderberg GL, Ballatyne BT, Clarke WR. A study of various
South Africa and the Waddel Scholarship. normalization procedures for within day electromyographic data. J
Electromyogr Kinesiol 1994;4(1):47–59.
Kollmitzer J, Ebenbichler GR, Kopf A. Reliability of surface electromyographic
measurements. Clin Neurophysiol 1999;110(4):725–34.
Appendix A. Supplementary data Komi PV, Buskirk ER. Reproducibility of electromyographic measurements with
inserted wire electrodes and surface electrodes. Electromyography
Supplementary data associated with this article can be found, in 1970;4:357–67.
Komi PV, Tesch P. EMG frequency spectrum, muscle structure, and fatigue during
the online version, at doi:10.1016/J.JELEKIN.2010.07.011. dynamic contractions in man. Eur J Appl Physiol Occup Physiol
1979;42(1):41–50.
Laplaud D, Hug F, Grelot L. Reproducibility of eight lower limb muscles activity level
References in the course of an incremental pedaling exercise. J Electromyogr Kinesiol
2006;16(2):158–66.
Arsenault AB, Winter DA, Marteniuk RG. Bilateralism of EMG profiles in human Larsson B, Mansson B, Karlberg C, Syvertsson P, Elert J, Gerdle B. Reproducibility of
locomotion. Am J Phys Med 1986;65(1):1–16. surface EMG variables and peak torque during three sets of ten dynamic
Ball N, Scurr J. An assessment of the reliability and standardisation of tests used to contractions. J Electromyogr Kinesiol 1999;9(5):351–7.
elicit reference muscular actions for electromyographical normalisation. J MacIntosh BR, Neptune RR, Horton JF. Cadence, power, and muscle activation in
Electromyogr Kinesiol 2010;20(1):81–8. cycle ergometry. Med Sci Sports Exerc 2000;32(7):1281–7.
Bigland-Ritchie B, Woods JJ. Integrated EMG and oxygen uptake during dynamic Marsh AP, Martin PE. The association between cycling experience and preferred and
contractions of human muscles. J Appl Physiol 1974;36(4):475–9. most economical cadences. Med Sci Sports Exerc 1993;25(11):1269–74.
Bolgla LA, Uhl TL. Reliability of electromyographic normalization methods for Marsh AP, Martin PE. The relationship between cadence and lower extremity EMG
evaluating the hip musculature. J Electromyogr Kinesiol 2007;17(1):102–11. in cyclists and noncyclists. Med Sci Sports Exerc 1995;27(2):217–25.
Burden A, Bartlett R. Normalisation of EMG amplitude: an evaluation and Mathur S, Eng JJ, MacIntyre DL. Reliability of surface EMG during sustained
comparison of old and new methods. Med Eng Phys 1999;21(4):247–57. contractions of the quadriceps. J Electromyogr Kinesiol 2005;15(1):102–10.
Burden A, Trew M, Baltzopoulos V. Normalisation of gait EMGs: a re-examination. J Menz HB, Latt MD, Tiedemann A, Mun SK, Lord SR. Reliability of the GAITRite
Electromyogr Kinesiol 2003;13(6):519–32. walkway system for the quantification of temporo-spatial parameters of gait in
Chapman AR, Vicenzino B, Blanch P, Knox JJ, Hodges PW. Intramuscular fine-wire young and older people. Gait Posture 2004;20(1):20–5.
electromyography during cycling: repeatability, normalisation and a Murley GS, Menz HB, Landorf KB, Bird AR. Reliability of lower limb
comparison to surface electromyography. J Electromyogr Kinesiol 2010;20(1): electromyography during overground walking: a comparison of maximal- and
108–17. sub-maximal normalisation techniques. J Biomech 2010;43(4):749–56.
Dubo HI, Peat M, Winter DA, Quanbury AO, Hobson DA, Steinke T, et al. Ng JK, Richardson CA. Reliability of electromyographic power spectral analysis of
Electromyographic temporal analysis of gait: normal human locomotion. Arch back muscle endurance in healthy subjects. Arch Phys Med Rehabil
Phys Med Rehabil 1976;57(9):415–20. 1996;77(3):259–64.
Ericson MO. On the biomechanics of cycling: a study of joint and muscle load Norcross MF, Blackburn JT, Goerger BM. Reliability and interpretation of single leg
during exercise on the bicycle ergometer. Scand J Rehabil Med stance and maximum voluntary isometric contraction methods of
1986;16(Suppl.):1–43. electromyography normalization. J Electromyogr Kinesiol 2010;20(3):420–5.
Ericson MO, Nisell R, Arborelius UP, Ekholm J. Muscular activity during ergometer Portney LG, Watkins MP. Foundations of clinical research: applications to
cycling. Scand J Rehabil Med 1985;17(2):53–61. practice. New Jersey: Prentice Hall Health; 1990.
Ericson MO, Nisell R, Ekholm J. Quantified electromyography of lower-limb muscles Potvin JR. Effects of muscle kinematics on surface EMG amplitude and frequency
during level walking. Scand J Rehabil Med 1986;18(4):159–63. during fatiguing dynamic contractions. J Appl Physiol 1997;82(1):144–51.
Farina D, Macaluso A, Ferguson RA, De Vito G. Effect of power, pedal rate, and force Rainoldi A, Bullock-Saxton JE, Cavarretta F, Hogan N. Repeatability of maximal
on average muscle fiber conduction velocity during cycling. J Appl Physiol voluntary force and of surface EMG variables during voluntary isometric
2004a;97(6):2035–41. contraction of quadriceps muscles in healthy subjects. J Electromyogr Kinesiol
Farina D, Merletti R, Enoka RM. The extraction of neural strategies from the surface 2001;11(6):425–38.
EMG. J Appl Physiol 2004b;96(4):1486–95. Ricard MD, Hills-Meyer P, Miller MG, Michael TJ. The effects on bicycle frame
Felici F. Applications in exercise physiology. In: Merletti R, Parker PA, editors. geometry on muscle activation and power during a Wingate Anaerobic test. J
Electromyography: physiology, engineering, and noninvasive applications. New Sports Sci Med 2006;5(1):25–32.
Jersey: John Wiley and Sons Inc.; 2004. p. 365–79. Rouffet DM, Hautier CA. EMG normalization to study muscle activation in cycling. J
Felici F. Neuromuscular responses to exercise investigated through surface EMG. J Electromyogr Kinesiol 2008;18(5):866–78.
Electromyogr Kinesiol 2006;16(6):578–85. Sleivert GG, Wenger HA. Reliability of measuring isometric and isokinetic peak
Ferndández-Peña E, Lucertini F, Ditroilo M. A maximal isokinetic pedalling exercise torque, rate of torque development, integrated electromyography, and tibial
for EMG normalization in cycling. J Electromyogr Kinesiol nerve conduction velocity. Arch Phys Med Rehabil 1994;75(12):1315–21.
2009;19(3):e162–170. Taylor AD, Bronks R. Reproducibility and validity of the quadriceps muscle
Gamet D, Duchene J, Goubel F. Reproducibility of kinetics of electromyogram integrated electromyogram threshold during incremental cycle ergometry.
spectrum parameters during dynamic exercise. Eur J Appl Physiol Occup Physiol Eur J Appl Physiol Occup Physiol 1995;70(3):252–7.
1996;74(6):504–10. Winter DA. The biomechanics and motor control of human gait: normal, elderly and
Gandevia SC. Spinal and supraspinal factors in human muscle fatigue. Physiol Rev pathological. Waterloo: University of Waterloo Press; 1991.
2001;81(4):1725–89. Winter DA, Yack HJ. EMG profiles during normal human walking: stride-to-stride
Goto S, Toyoshima S, Hoshikawa T. Study of the integrated EMG of leg muscles and inter-subject variability. Electroencephal Clin Neurophysiol
during pedalling at various loads, frequency, and equivalent power. In: Komi 1987;67:402–11.
PV, editor. Biomechanics V-1A. Baltimore: University Park Press; 1976. p. Yang JF, Winter DA. Electromyography reliability in maximal and submaximal
246–52. isometric contractions. Arch Phys Med Rehabil 1983;64(9):417–20.
Gregor RJ, Cavanagh PR, LaFortune M. Knee flexor moments during propulsion in Yang JF, Winter DA. Electromyographic amplitude normalization methods:
cycling – a creative solution to Lombard’s Paradox. J Biomech improving their sensitivity as diagnostic tools in gait analysis. Arch Phys Med
1985;18(5):307–16. Rehabil 1984;65(9):517–21.
 Y. Albertus-Kajee et al. / Journal of Electromyography and Kinesiology 20 (2010) 1036–1043 1043

Yumna Albertus-Kajee is a Post Doctoral Fellow in the Wayne Derman is a Professor in the UCT/MRC Exercise
UCT/MRC Exercise Science and Sports Medicine Unit at Science and Sports Medicine Unit at the University of
the University of Cape Town. She graduated with her Cape Town and is Medical Co-Director of the Chronic
PhD from the University of Cape Town in 2008, where Disease Rehabilitation Program at the Sports Science
she focused her thesis on normalization techniques Institute of South Africa. He graduated MBChB from the
during dynamic activity in healthy and diseased popu- University of Pretoria in 1986, and obtained his BSc
lation. Her research interests include muscular activity (Med)(Hons) in sport science and his PhD in sports
in patients with chronic disease and the effects of medicine in 1988 and 1993, respectively. His current
supervised exercise programs on muscle activity. As research interests include functional capacity in chronic
well as neuromuscular factors contributing to fatigue disease, sports injury prevention, psychoneurocardiol-
and pacing strategies in the athletic population. ogy, chronic fatigue and underperformance in athletes,
cardiorespiratory conditions in athletes and mind-body
interaction.

Ross Tucker is a research scientist at the University of Mike Lambert is a Professor in the UCT/MRC Exercise
Cape Town’s Research Unit for Exercise Science and Science and Sports Medicine Unit at the University of
Sports Medicine. He obtained a doctorate in 2006, Cape Town. He has a research interest in muscle damage
studying fatigue, pacing strategy and the brain. Areas of and regeneration and monitoring performance and
research include thermoregulation, altitude and per- training status of high performance athletes.
formance analysis. He is a consultant to various sporting
teams, and has a particular interest in the application of
sports science to performance.