Open Journal of Preventive Medicine, 2011, 1, 1-7
doi:10.4236/ojpm.2011.11001 Published Online May 2011 (http://www.SciRP.org/journal/ojpm/
OJPM
).
Published Online May 2011 in SciRes. http://www.scirp.org/journal/OJPM
Counting steps in research: A comparison of accelerometry
and pedometry
Melody Oliver1,2, Hannah M Badland1,3, Janine Shepherd1, Grant M Schofield1
1Centre for Physical Activity and Nutrition, Auckland University of Technology, Auckland, New Zealand
2School of Public Health and Psychosocial Studies, Auckland University of Technology, Auckland, New Zealand
3Division of Population Health, University College London, London, United Kingdom
Received 3 March 2011; revised 10 May 2011; accepted 12 May 2011.
ABSTRACT
The objective of this study was to assess the validity
of the step count functions in Actical accelerometers
and activPAL inclinometers, compared with pe-
dometer-derived step count data. Firstly, directly
observed step counts over 3 treadmill speeds were
compared with steps collected from 3 pedometers,
accelerometers, and inclinometers in 10 adults. Sec-
ondly, step count data were derived from 22 partici-
pants who wore a pedometer, accelerometer, and in-
clinometer over 48 hours. Agreement between meas-
urement tools was determined. All monitors appro-
priately measured steps in the laboratory conditions.
In free living conditions, the mean percentage differ-
ences with pedometer-determined step counts were
–7.3% and 7.0% for the Actical and activPAL moni-
tors, respectively. With the exception of slow walking
for the Actical units (ICC < 0.001), acceptable reli-
ability was found within units for all treadmill speeds,
and across units during the free living condition. The
95% prediction interval ranges were wide, ranging
from –68.8% to 54.2% for the Acticals, and from
–39.1% to 53.2% for the activPALs. Step counts
gathered from Actical and activPAL units should not
be used interchangeably with pedometer-derived step
count data.
Keywords: Physical Activity; Measurement; Pedometer;
Inclinometer; Accelerometer; Validation; Reliability
1. INTRODUCTION
Accurate quantification of free-living physical activity is
important to enable researchers to derive reliable infor-
mation on associations with health outcomes and de-
velop public health recommendations, identify determi-
nants of activity and individuals/population groups at
risk of insufficient activity for health, and to accurately
assess intervention effectiveness. Accelerometers are
increasingly being used for this purpose, due to the util-
ity of these monitors to provide objective and precise
information on activity intensity accumulated over sus-
tained periods. No best practice has yet been determined
for accelerometer data treatment; inconsistencies in ap-
proaches to data reduction have made it especially chal-
lenging to accurately track activity behaviors or compare
results across studies. Indeed, increasingly sensitive
measurement instruments may offer a decrease in speci-
ficity of measurement especially at the lower intensity
end of measurement. Moreover, understanding and in-
terpreting physical activity intensity concepts can be
challenging, making it difficult to translate and dissemi-
nate practical health promotion messages. In contrast,
the measurement of physical activity in the form of steps
is relatively straightforward to measure via pedometry,
and the data are (relatively) simple to analyze and dis-
seminate and make inferences to the general public.
Body composition-referenced step count recommenda-
tions have been developed [1] and pedometer-based
programs have been shown to be effective in increasing
physical activity (especially where step goals are in-
cluded) and improving health outcomes (e.g., reducing
body mass index and blood pressure) [2].
Accordingly there is opportunity to augment more
complex physical activity measures (accelerometry, in-
clinometry) with step counts. Two commercially avail-
able monitors exist that provide both step count and
more complex movement data: 1) Actical accelerometers
now offer the ability to assess steps accumulated in addi-
tion to existing accelerometry-based physical activity
assessment, and 2) the activPAL is also an accelerome-
ter-based monitor that includes an inclinometer; for ease
of differentiation between units these will be termed
inclinometers hereafter. These inclinometers provide

M. Oliver et al. / Open Journal of Preventive Medicine 1 (2011) 1-7 2
date and time stamped information on time spent sitting,
lying, standing, walking, and number of steps accumu-
lated. The Actical accelerometer has been shown to be
valid for estimating physical activity intensity in adults
[3]. Validity of the new step-count function in Actical
accelerometers has been assessed using laboratory-based
activities [4], however to our knowledge, no assessment
of the validity of this function in measuring steps accu-
mulated during free-living physical activity over sus-
tained periods of time (i.e., > 1 d) has been conducted.
activPAL monitors have shown acceptable reliability
[5,6] and validity [7] for measuring time spent being
sedentary and active (including walking) under con-
trolled/laboratory conditions. Evidence suggests that the
accuracy of these monitors may be lowest when assess-
ing time spent walking [5]. The only studies to assess the
activPAL step-count function in measuring actual steps
taken have also been undertaken in laboratory/controlled
conditions [8,9]. Findings indicate these monitors are
reliable [8] and accurate to within 1.3% for treadmill and
outdoor walking [8,9].
Generally, laboratory-based testing has provided suf-
ficient evidence for the validity of both the Actical and
activPAL monitors to accurately assess steps in adults.
This research now needs to be extended to application in
free-living situations for durations greater than the 6
hours employed to date. Research aims were thus to as-
sess the reliability, validity, and agreement of step-count
data gathered from Actical accelerometers and activPAL
inclinometers with standard practice for step count
measurement (pedometry) in free-living conditions over
multiple days.
2. METHODS
2.1. Participants
Convenience samples of university employees in Auck-
land, New Zealand were recruited for each testing stage.
For the laboratory-based testing, 10 university staff were
invited to participate. Participants for the free-living
testing were recruited as part of a wider study investi-
gating measurement of sedentary behavior [10,11]. As
such, these participants were included only if they re-
ported spending the majority of their occupational time
sitting. No other inclusion/exclusion criteria were em-
ployed. Ethical approval to conduct the study was pro-
vided by Auckland University of Technology Ethics
Committee (AUTEC reference 09/253, granted 25 No-
vember 2009). Written informed consent was gathered
from all participants. Data were collected on weekdays
between December 2009 and March 2010.
2.2. Measurement Tools
Three Actical accelerometers (Mini-Mitter, Respironics
Inc Company, Bend, OR), three activPAL inclinometers
(PAL Technologies Ltd, Glasgow), and three Yamax
Digiwalker CW-700 pedometers (Yamax Corp., Kuma-
moto, Japan) were used in the current study. Yamax Di-
giwalker (e.g., SW-701, SW-200) pedometers are valid
and reliable for measuring steps in adults [12,13]. The
CW-700 improves over previous Digiwalker models by
including a multi-day-memory function, but otherwise
the mechanical properties of these units are identical to
earlier models. Performance of all units was first as-
sessed in laboratory conditions prior to application in
free-living conditions.
2.3. Procedures
2.3.1. Laboratory Testing
Tests were conducted on a Powerjog GX C200 motor-
ized treadmill (PowerSport International Inc., Bridgend,
UK). Participants underwent a 5 min familiarization
procedure, including practicing starting and stopping at
each speed. Units were then attached, with the three ac-
tivPAL units placed medially and vertically on the right
vastus medialis with Physiomed TheraFIX Underwrap
tape, and the three pedometers and accelerometers at-
tached to elastic belts and placed above the right iliac
crest. Participants then walked for 5 min at three speeds:
54 m·min–1, 107 m·min–1, and 80 m·min–1, with a 1 min
rest between each test. Walking speeds were identified as
being the lowest, highest, and median values used in
previous research [8,12]. Pedometers were set to 0 be-
fore each trial and readings from each pedometer were
taken immediately post each test. During the tests, one
research assistant counted down start and stop times,
recorded measurement times, and pedometer steps, and a
second research assistant counted observed steps using a
manual step counter.
2.3.2. Free-Living Testing
Participants were provided with one activPAL incli-
nometer and tape, one Actical accelerometer, one Yamax
pedometer, and unit wearing instructions by a trained
research assistant. On delivery, pedometers were set to
zero, sealed with tamper-evident tape, then placed on an
elastic belt at the right hip above the iliac crest and se-
cured with an additional safety strap. Accelerometers
were attached to the elastic belts in line with the pe-
dometer. The inclinometer was attached medially on the
right vastus medialis, secured in place with Physiomed
TheraFIX Underwrap tape. Participants were asked to
wear all units for the following 48 hours and to remove
them only when bathing or sleeping. Participants also
completed a paper-based compliance diary for the 48 h
period they wore the measurement units (including re-
searcher-reported delivery and collection times and pe-
dometer steps accumulated at the time of pedometer
C
opyright © 2011 SciRes. OJPM

M. Oliver et al. / Open Journal of Preventive Medicine 1 (2011) 1-7 3
OJPM
collection), and were called once during the measure-
ment period to confirm monitor wear. Monitors were
collected by the research assistant exactly 48 h after they
were delivered. On collection, the researcher recorded
the time of unit removal and pedometer steps accumu-
lated for that day. Participants reported their height in
meters and weight in kilograms. Body mass index (BMI)
was calculated as kg·m–2, and thresholds of 25 kg·m–2
and 30 kg·m–2 were used to define overweight and obe-
sity, respectively.
2.4. Statistical Analyses
2.4.1. Laboratory Testing
Accelerometer and inclinometer data were downloaded
using Actical 2.04 and activPAL 5.8.3.5, respectively,
and combined by time in Microsoft Excel. Step count
data for the accelerometers and inclinometers were then
summed for each 5 min bout and matched with the pe-
dometer and observed step counts. Intraclass correlation
coefficients (ICC) for step counts from each unit at each
speed were calculated. Agreement between observed
steps (criterion) and step counts from pedometers, accel-
erometers, and inclinometers (comparison measures)
was assessed using the Bland-Altman method [14,15].
Initially, Bland-Altman plots and associated lowess
curves were produced and assessed for each of the mea-
surement units against the observed steps at each ambu-
latory speed to determine whether combined or sub-
grouped analyses were necessary. Average values of the
observed step counts and each comparison measure were
compared against the difference of the observed step
counts and steps from each comparison measure. Per-
centage differences between the observed steps and
comparison measures were then calculated and the cor-
responding mean differences and 95% limits of agree-
ment (±1.96 SE) calculated and plotted. Equality in va-
riance between percentage differences in step count
readings at each gait speed was formally determined for
each comparison measure using the Brown and Forsythe
test for equality [16]. Statistical analyses were under-
taken using Stata IC version 10 (StataCorp, TX) and
 =
0.05 was used to determine statistical significance.
2.4.2. Free-living Testing
Self-reported compliance in wearing the units was com-
pared with information gathered during the random
phone calls for accuracy. Participants not meeting mini-
mal compliance criteria of wearing the units for at least
one full (> 10 h on day 2) and one partial (afternoon of
day 1 or morning of day 3) day were excluded from fur-
ther analyses. Accelerometer and inclinometer data were
downloaded using the Actical and activPAL software,
respectively, and then combined by date and time in Ex-
cel. Data for the 48 h measurement period were manu-
ally extracted for each individual using the unit delivery
and collection times from the compliance diary, and step
counts for the measurement period were summed from
the extracted data. Researcher-reported pedometer step
counts for the day of collection (day 3) were obtained
from the compliance diary, and steps for previous days
(days 1 and 2) were gathered from the pedometer mem-
ory. Days where pedometer step counts were greater
than 30 000 or less than 1 000 (or proportionate values
for partial days) were considered outliers and removed
from analyses [17]. Only participants with at least one
full and one partial day of pedometer data remaining and
concurrent accelerometer and/or inclinometer data were
included in further analyses.
Remaining step count data were combined for each
measurement tool. Average pedometer steps·h–1 were
calculated and differences by sex, BMI status (dichoto-
mized as normal, overweight/obese), and age group (di-
chotomized as 20-43y, 44-58y) were determined using
independent t tests. Pedometer data were matched with
available accelerometer and inclinometer data by times
worn. Associations between the unit types for steps
counted (i.e., pedometer, Actical accelerometer, ac-
tivPAL inclinometer) was determined by ICC. Agree-
ment between step counts from the pedometer and 1)
Actical accelerometer step counts and 2) activPAL in-
clinometer step counts were assessed using the Bland-
Altman method [14,15] as per the laboratory testing
analyses above.
3. RESULTS
3.1. Laboratory Testing
Ten adults (9 females) completed the measurement pro-
tocol. Table 1 shows the mean (SD) of the observed step
counts and step counts from comparison measures at
each gait speed, the mean (SD) for the percentage dif-
ferences between observed steps and comparison meas-
ures, and the related ICC values for steps counted. With
the exception of the accelerometers at slow walking
speeds (ICC < 0.001), units demonstrated acceptable reli-
ability across all walking speeds (ICC = 0.335 - 0.998).
Unit reliability increased with walking speed for pe-
dometers and accelerometers, while activPAL perform-
ance was consistent across all pace conditions (ICC =
0.998). Visual analysis of Bland-Altman plots and asso-
ciated lowess curves indicated that the use of combined
analyses was appropriate. activPAL units performed to
98.2% accuracy across all pace conditions. Significant
heterogeneity was found in the variances of observed-
comparison measurements over gait speed for the pe-
dometers and accelerometers (Brown-Forsythe test p <
0.001) indicating that the accuracy of these measurement
tools was dependent on gait speed. Figure 1 shows the

M. Oliver et al. / Open Journal of Preventive Medicine 1 (2011) 1-7 4
Table 1. Median (SD) of steps observed and recorded from the Yamax Digiwalker pedometer, Actical accelerometer, and activPAL
inclinometer under three treadmill pace conditions, and relative percentage differences between observed steps and comparison units
in 10 adults.
OB Yamax Pedometer Actical accelerometer activPAL inclinometer
Treadmill
speeda Steps
mean (SD)
Steps
mean (SD)
% difference
from OB
mean (SD)
ICC Steps
mean (SD)
% difference
from OB
mean (SD)
ICC Steps
mean (SD)
% difference
from OB
mean (SD)
ICC
Slow 477 (21.6) 394 (109.4) –17.4 (23.1)0.387394 (93.8)–17.2(19.6)<0.001468 (21.6) –1.9 (0.3)0.998
Medium 568 (29.5) 557 (80.7) –2.1 (12.7) 0.369 546(63.1)–3.8(9.3)0.335 557 (28.8) –1.9 (0.4)0.998
Fast 627 (32.1) 631 (33.1) 0.5 (0.4)0.997621 (38.2)–1.0(3.8)0.590616 (31.5) –1.8 (0.4)0.998
Combined 557 (68.3) 527 (127.6) –6.3 (17.0)n/a 521(116.8)–7.3(14.4)n/a 547 (67.1) –1.8 (0.4)n/a
Notes: ICC = Intraclass Correlation Coefficient, n/a = not applicable, OB = Observed Steps, SD = Standard Deviation; a. Slow = 54 m·min–1, Medium = 80
m·min–1, Fast = 107 m·min
Figure 1. 95% limits of agreement for difference between
observed step counts and Yamax Pedometer step counts, Actical
accelerometer step counts, and activPAL inclinometer step
counts over 5 minutes treadmill walking, by ambulatory gait
speed.
95% limits of agreement for differences between the
observed step counts and each of the measurement tools,
by ambulatory gait speed. Performance of all units was
similar to previous validation studies [4,8,9,12] and so
these were considered acceptable for further testing in
free-living conditions.
3.2. Free-Living Testing
Twenty-five adults aged 41.6 ± 11.8 y participated in the
study, with an average BMI of 26.1 ± 5.0 kg·m–2. Self-
reported compliance matched that reported in the phone
calls, and participants wore the units for an average of
26.8 h (range 15.7 - 34.0 h) over the 48 h period. In total,
22 participants met the minimum pedometer compliance
criteria. Of these, corresponding accelerometer and in-
clinometer data were gathered from 22 and 19 partici-
pants, respectively. Table 2 displays the participant
characteristics and average step counts for participants
meeting the pedometer data inclusion criteria. Average
(minimum, maximum) wear times for pedometer and
accelerometer data included in analyses was 24.5 h (15.7
h, 31.3 h), and for inclinometer data was 22.3 h (13.3 h,
31.2 h). No significant differences in average pedometer
steps·h–1 were found for sex, age, or BMI status (p >
0.05). A high degree of correlation was found across all
units for the within-participant step counts (ICC = 0.850,
0.807, and 0.666 for pedometer-Actical, pedometer-
activPAL, and Actical-activPAL, respectively).
Figure 2 shows the Bland Altman plot for agreement
and associated 95% limits of agreement for the Actical
and activPAL step counts relative to the pedometer- de-
termined step counts. The mean percentage differences
with pedometer-determined step counts were –7.3% and
7.0% for the Actical and activPAL monitors, respectively.
The 95% prediction interval ranges were wide for both
Table 2. Descriptive characteristics and step counts for free-liv-
ing study participants meeting pedometer data inclusion criteria.
Variable N (%)
Sex
Male 6 (27.3)
Female 16 (72.7)
Age (yr)
20 - 43 13 (59.1)
44 - 58 9 (40.9)
BMI classification
Normal/underweight
(<25 kg·m–2)
12 (54.5)
Overweight (25 - 29 kg·m–2)5 (22.7)
Obese (>30 kg·m–2) 5 (22.7)
Step counts Mean (minimum,
maximum)
Total step counts
Pedometera 14 207 (6335, 36290)
Accelerometer 13 118 (3859, 34113)
Pedometerb 12 814 (7846, 36290)
Inclinometer 13 272 (6820, 33232)
Average step counts (steps·h–1)
Pedometera 586 (277, 1788)
Accelerometer 544 (153, 1680)
Pedometerb 600 (277, 1831)
Inclinometer 617 (303, 1677)
aMatched for times worn with Actical accelerometer; bMatched for times
worn with activPAL inclinometer
C
opyright © 2011 SciRes. OJPM

M. Oliver et al. / Open Journal of Preventive Medicine 1 (2011) 1-7 5
comparison measures, however, ranging from –68.8% to
54.2% for the Actical accelerometer (width 122.9%), and
from –39.1% to 53.2% for the activPAL inclinometer
(width 92.3%). Within-individual inconsistencies in the
direction and scale of misclassification for the compari-
son units were also found; Figure 3 shows the percent-
age difference between the comparison measures and the
average steps gathered from the pedometer and each
comparison measure for each participant.
4. DISCUSSION
To our knowledge, this is the first study to investigate
the agreement between pedometer steps and steps gath-
ered from Actical accelerometers and activPAL incli-
nometers in free-living conditions over multiple days.
Our initial laboratory testing showed similar results to
previous research, with step count underestimation in
Yamax pedometers [8,9,12] and Actical accelerometers
[4] at slower speeds but acceptable accuracy and reli-
ability for medium speed and fast walking. For the ac-
tivPAL steps we found a high degree of accuracy and
reliability (>98%, ICC = 0.998) with observed steps for
all walking speeds [8,9]. In line with Esliger et al. [4] we
also noted considerable variability in within-individual
inconsistencies in the percentage differences for the Ac-
tical and activPAL monitors, indicating there was no
consistent pattern that could be related to the unit or in-
dividual participants.
Considering the laboratory-based results of the current
research and previous studies [8,9], it is likely that the
activPAL yielded the most accurate measure of steps ac-
cumulated in free living conditions, which are likely to
include walking at lower speeds. The only other study to
consider the accuracy of activPAL units in free living
conditions (albeit for approximately 6 h only) showed a
high degree of accuracy in classifying time spent stepping
[7]. A second-by second analysis of activPAL- derived
classification of time spent stepping, sitting/lying, and
standing showed an overall agreement of 95% with di-
rectly observed activities, however the limits of agreement
were widest for walking (range –16.1% to 12.1%) [5].
While findings from the current study corroborate pre-
vious research findings, the reduced accuracy of the Ya-
max Digiwalker pedometers at lower speeds clearly lim-
its their accuracy in measuring actual steps taken in free-
living conditions, and therefore their utility as a crite-
rion/comparison measure in the current study. Pedometer
steps are the most common and widely referenced basic
unit of objectively measured physical activity. As such,
the inclusion of pedometers in the current study was still
deemed important, to determine whether steps gathered
from Actical and activPAL units could appropriately be
used interchangeably with pedometer steps as a measure
-100 -80-60-40-20020 40 6080100
Percentage difference in step counts (AC-YP)/YP
010000 2000030000 40000
Average of step counts ((YP+AC/2))
()
(a)
-100-80-60 -40-20020406080 100
Percentage difference in step counts (AP-YP)/YP
010000 200003000040000
Average of step counts ((YP+AP/2))
(b)
Figure 2. Bland Altman plots demonstrating agreement between
Yamax Digiwalker pedometer steps and Actical accelerometer
steps (a), and activPAL inclinometer steps (b).
-100 -50050
Percentage difference from pedometer steps
1 2345 6 7 8 910111213141516171819202122232425
Actical activPAL
Figure 3. Percentage difference between Yamax Digiwalker
pedometer steps and comparison measures (Actical accelerometer,
activPAL inclinometer) for each participant (ID 1.25) during
the free living condition.
C
opyright © 2011 SciRes. OJPM

M. Oliver et al. / Open Journal of Preventive Medicine 1 (2011) 1-7 6
of health-related physical activity.
The limits of agreement with pedometer steps for both
the Actical accelerometers and activPAL inclinometers
were wide, at 123% and 92%, respectively. As we may
expect daily step counts of between 7000 and 13 000 in
healthy adults [18], an example of the outcome of this is
provided using the mean of these values: Findings from
the current study can be interpreted as meaning that
when an observed step count of 10 000 is recorded by a
pedometer, in 95% of instances the Actical monitors will
record a step count between 3120 and 15 420 steps
(width 12 300 steps), and the activPAL between 6090
and 15 320 steps (width 9230 steps). Detecting changes
in daily step counts of 2500 steps as a result of an inter-
vention has been associated with improvements in health
outcomes [19] and generally equates to one ‘band’ of
physical activity level classification [20]. Such changes
in daily step counts may be considered clinically impor-
tant. The results from the present study clearly show that
the prediction widths observed for the Actical and ac-
tivPAL units were too great for these units to be consid-
ered as being in agreement with pedometer steps meas-
ured. Based on our findings we conclude that step count
data observed in Actical accelerometers and activPAL
inclinometers cannot be used interchangeably with pe-
dometer steps, nor should they be used to classify indi-
vidual activity levels based on step count criteria.
Taken together, the results of the current study indi-
cate that the activPAL monitor will provide the best rep-
resentation of actual steps taken in free-living situations.
This is not surprising given the placement of the unit on
a participant’s upper leg, where the detection of walking
at any velocity is based on large gross limb movement.
The hip-mounted pedometer and accelerometer must
rely on more subtle vertical displacements to detect a
step, and these displacements reduce when walking ve-
locity decreases. The Digiwalker is a spring-levered pe-
dometer whereby upon vertical displacement, an internal
spring-suspended arm moves vertically, opening and
closing an electrical circuit to register a step. There is
some evidence that piezoelectric pedometers (compris-
ing a piezoelectric crystal which is moved in response to
acceleration, e.g., New Lifestyles, Kenz Lifecorder) may
be more accurate at slower speeds than other pedometer
technologies, especially in overweight or obese indi-
viduals [12,21]. Accurate quantification of low velocity
physical activity may be a very important part of under-
standing the differences between lean and obese adults
[22]. Levine et al. [22] observed that lean subjects ac-
cumulated an extra 3.5 miles·d–1 of low velocity walking
(∼1 mile·h–1) than obese subjects. It is also plausible
that slower velocity walking provides a lower metabolic
benefit than walking at higher speeds and the under-
reading of pedometer steps at lower velocities provides a
reasonable proxy for such a difference in benefit.
Additionally, activPAL monitors are not necessarily
suitable for all types of physical activity measurement in
the same way pedometers, and to a certain extent accel-
erometers, may be. While they may be most accurate in
step count assessment, they are not able to assess subtle
movements such as shifting weight while standing, and
measurement is limited to the site/limb of placement
(e.g., right leg movement only). Notably, these limita-
tions also apply to pedometers and accelerometers (al-
beit some accelerometers are sensitive enough to capture
small movements such as weight shifts). It is also
worthwhile noting that this lack of sensitivity actually
improves accuracy in step counting; for example, Mad-
docks et al. [9] found that vibrations caused by car travel
influenced activity data collected by Digiwalker pe-
dometers and PALlite accelerometers, while no activity
was recorded by activPAL units under the same condi-
tions. In comparison to pedometers, activPAL monitors
are expensive (approximately $US400 per unit) and
mounting the unit with tape on a participant’s thigh is
difficult and more invasive than fitting belt-mounted
pedometers or accelerometers. When activity measure-
ment is over multiple days, monitor removal and re-
attachment needs to be managed by the participants
themselves, and this may present both compliance and
other measurement consistency problems. These issues
make larger population epidemiology, and work with
younger and older populations less feasible. Finally, it is
worth noting that the relatively small sample in this
study was predominantly female, employed, and likely
to be well educated, thus limiting the generalisability of
the results from the free-living conditions. Further re-
search would benefit from including a larger and more
representative sample in order to capture a broad variety
of daily activities across a wide range of population
groups.
5. CONCLUSION
The three types of units accurately measured step counts
under laboratory conditions, but when the step counts
derived from Actical accelerometers and activPAL in-
clinometers were compared with the pedometer values
under free living conditions, considerable disagreement
was evident. Therefore, these different devices are not
interchangeable when measuring step counts, and Acti-
cal and activPAL units should not be used to classify
physical activity levels based on existing step count
thresholds. Pedometers remain the most cost effective
and simple method of step count measurement within
free living settings, while activPAL inclinometers are the
most accurate for assessing steps gathered over a range
C
opyright © 2011 SciRes. OJPM

M. Oliver et al. / Open Journal of Preventive Medicine 1 (2011) 1-7 7
[10] Oliver, M., Badland, H.M., Schofield, G.M. and Shepherd,
J. (in press) Identification of non-wear time and sedentary
behavior using accelerometry. Research Quarterly for
Exercise and Sport.
of walking speeds.
6. ACKNOWLEDGEMENTS
At the time of writing, Melody Oliver and Hannah Badland were sup-
ported by National Heart Foundation of New Zealand research fellow-
ships. A Faculty of Health and Environmental Sciences Auckland
University of Technology Summer Studentship supported Janine
Shepherd.
[11] Oliver, M., Schofield, G.M., Badland, H.M. and Shepherd,
J. (2010) Utility of accelerometer thresholds for classifying
sitting in office workers. Preventive Medicine, 51, 357-
360. doi:10.1016/j.ypmed.2010.08.010
[12] Crouter, S.E., Schneider, P.L., Karabulut, M. and Bassett,
D.R., Jr. (2003) Validity of 10 electronic pedometers for
measuring steps, distance, and energy cost. Medicine &
Science in Sports & Exercise, 35, 1455-1460.
doi:10.1249/01.MSS.0000078932.61440.A2
REFERENCES
[1] Tudor-Locke, C., Ainsworth, B.E., Whitt, M.C., Thompson,
R.W., Addy, C.L. and Jones, D.A. (2001) The relationship
between pedometer-determined ambulatory activity and
body composition variables. International Journal of
Obesity, 25, 1571-1578.
doi:10.1038/sj.ijo.0801783
[13] Le Masurier, G.C., Lee, S.M. and Tudor-Locke, C. (2004)
Motion sensor accuracy under controlled and free-living
conditions. Medicine & Science in Sports & Exercise, 36,
905-910. doi:10.1249/01.MSS.0000126777.50188.73
[14] Bland, J.M. and Altman, D.G. (1999) Measuring
agreement in method comparison studies. Statistical
Methods in Medical Research, 8, 135-160.
doi:10.1191/096228099673819272
[2] Bravata, D.M., Smith-Spangler, C., Sundaram, V.,
Gienger, A.L., Lin, N., Lewis, R., et al. (2007) Using
pedometers to increase physical activity and improve
health. A systematic review. Journal of the American
Medical Association, 298, 2296-2304.
doi:10.1001/jama.298.19.2296
[15] Bland, J.M. and Altman, D.G. (2003) Applying the right
statistics: Analyses of measurement studies. Ultrasound in
Obstetrics and Gynecology, 22, 85-93.
doi:10.1002/uog.122
[3] Heil, D.P. (2006) Predicting activity energy expenditure
using the Actical® activity monitor. Research Quarterly
for Exercise and Sport, 77, 64-80.
[16] Brown, M.B. and Forsythe, A.B. (1974) Robust tests for
equality of variances. Journal of the American Statistical
Association, 69, 364-367. doi:10.2307/2285659
[4] Esliger, D.W., Probert, A., Gorber, S.C., Bryan, S.,
Laviolette, M. and Tremblay, M.S. (2007) Validity of the
Actical accelerometer step-count function. Medicine &
Science in Sports & Exercise, 39, 1200-1204.
doi:10.1249/mss.0b013e3804ec4e9
[17] Tudor-Locke, C., Giles-Corti, B., Knuiman, M. and
McCormack, G. (2008) Tracking of pedometer-determined
physical activity in adults who relocate: Results from
RESIDE. International Journal of Behavioral Nutrition
and Physical Activity, 5, doi:10.1186/1479-5868-5-39
[5] Grant, P.M., Ryan, C.G., Tigbe, W.W. and Granat, M.H.
(2006) The validation of a novel activity monitor in the
measurement of posture and motion during everyday
activities British Journal of Sports Medicine, 40, 992-
997. doi:10.1136/bjsm.2006.030262
[18] Tudor-Locke, C.E. and Myers, A.M. (2001) Methodological
considerations for researchers and practitioners using
pedometers to measure physical (ambulatory) activity.
Research Quarterly for Exercise and Sport, 72, 1-12.
[19] Tudor-Locke, C., Myers, A.M., Bell, R.C., Harris, S.B.
and Rodger, N.W. (2002) Preliminary outcome evaluation
of the First Step Program: a daily physical activity
intervention for individuals with type 2 diabetes. Patient
Education and Counseling, 47, 23-28.
doi:10.1016/S0738-3991(01)00169-0
[6] Busse, M.E., van Deursen, R.W. and Wiles, C.M. (2009)
Real- life step and activity measurement: Reliability and
validity. Journal of Medical Engineering & Technology,
33, 33-41. doi:10.1080/03091900701682606
[7] Godfrey, A., Culhane, K.M. and Lyons, G.M. (2007)
Comparison of the performance of the activPAL™
Professional physical activity logger to a discrete
accelerometer-based activity monitor. Medical Engineering
& Physics, 29, 930-934.
doi:10.1016/j.medengphy.2006.10.001
[20] Tudor-Locke, C. and Bassett Jr, D.R. (2004) How many
steps/ day are enough? Preliminary pedometer indices for
public health. Sports Medicine, 34, 1-8.
doi:10.2165/00007256-200434010-00001
[21] Crouter, S.E., Schneider, P.L. and Bassett, D.R. (2005)
Spring-Levered versus Piezo-Electric Pedometer Accuracy
in Overweight and Obese Adults. Medicine & Science in
Sports & Exercise, 37, 1673-1679.
doi:10.1249/01.mss.0000181677.36658.a8
[8] Ryan, C.G., Grant, P.M., Tigbe, W.W. and Granat, M.H.
(2006) The validity and reliability of a novel activity
monitor as a measure of walking. British Journal of
Sports Medicine, 40, 779-784.
doi:10.1136/bjsm.2006.027276 [22] Levine, J.A., McCrady, S.K., Lanningham-Foster, L.M.,
Kane, P.H., Foster, R.C. and Manohar, C.U. (2008) The
role of free-living daily walking in human weight gain
and obesity. Diabetes, 57, 548-554.
doi:10.2337/db07-0815
[9] Maddocks, M., Petrou, A., Skipper, L. and Wilcock, A.
(2008) Validity of three accelerometers during treadmill
walking and motor vehicle travel. British Journal of
Sports Medicine, Online First,
doi:10.1136/bjsm.2008.051128
OJPM