An accurate assessment of physical activity is important for a variety of research applications in the exercise science field. Accelerometry-based motion-sensing devices (e.g., Tritrac, Computer Science and Applications, Inc. [CSA] monitor) have become popular assessment techniques, and a number of monitors have been validated under both laboratory (7,9,15,16) and field conditions (3,14,22). Although these monitors have strong application for a number of studies, the relatively high cost ($150–500) makes them cost prohibitive for many epidemiologic and intervention studies. Because much of a person’s daily activity involves locomotor movement such as walking, step-counting devices (pedometers) may offer a cost-effective measurement tool for field-based research applications. These devices cost approximately $20–25 and are small enough to be unobtrusive to study participants. The immediate feedback available from pedometers also makes them particularly useful as a behavior modification tool.
Similar to other accelerometry-based devices, pedometers respond to vertical accelerations of the body. The counting of steps is done by an internal trigger mechanism that records a “step” when the vertical acceleration exceeds a specific threshold value. Early research with these devices reported poor reliability across units and large errors in the estimations of steps and distance walked (10,21). Some of this error could be attributed to the technical limitations of these early monitors. Variability in spring tension, for example, causes considerable differences in sensitivity and leads to large inter- and intraindividual variability in the values recorded by these devices (1).
Recent advances in technology and quality control have led to improvements in the reliability and validity of pedometers. In a recent review of five contemporary electronic pedometers, Bassett et al. (1) reported acceptable results for most of the monitors, with exceptional reliability and validity exhibited by the Yamax Digi-Walker (Yamax Inc., Tokyo, Japan). The Digi-Walker model was more accurate in terms of step counts and distance walked for each of the walking paces tested. In a field-based evaluation along a 4.88-km sidewalk, the Digi-Walker measured the number of steps and distance to within 1% of actual values.
Because of these promising findings, several investigators have recently compared the Digi-Walker against other, more established assessments to examine convergent validity. Eston and colleagues (6) reported correlations of 0.92 between step counts and a scaled V̇O2 measure during unstructured, low-intensity activity in children. Among adults, we recentlyobserved average correlations of 0.76 between the Tritrac and a Digi-Walker across 7 d of monitoring (4). These two studies provide some preliminary evidence for the validity of the Digi-Walker step counter as an objective indicator of habitual physical activity.
Like all assessment techniques, the use of pedometers has limitations. One is that they are not sensitive to changes in speed. In fact, because stride length is greater when running compared with walking, a person would be expected to accumulate more steps walking a given distance compared with running. Another limitation is the inability to segment activity by time. Without a time-based indicator, it is not possible to determine the intensity or duration of the activity that was performed. Because a person can accumulate a large number of steps during normal activities of daily living, it is difficult to determine how many steps are needed to meet existing physical activity guidelines. Two studies in Japan have used the figure of 10,000 steps per day as a behavioral target, but studies have not attempted to empirically determine whether these guidelines are appropriate for other populations. Before research can systematically establish guidelines for daily step counts, it is important to further describe variability in step count data occurring during structured and free-living activity.
The purpose of this study was to evaluate the utility of the Digi-Walker for use in quantifying levels of physical activity in the population. Two separate studies were conducted as part of this overall evaluation. In study 1, we determined the number of steps required to complete a set distance at different speeds and under different conditions (treadmill vs track). Because stride length is likely to vary by size, gender, and pace, it is important to begin to understand the effect of these variables on step counter readings. In study 2, we sought to determine the utility of the Digi-Walker to assess activity under field conditions. Currently, adults are encouraged to accumulate at least 30 min of moderate intensity physical activity (approximately 150–200 kcal [627–837 K]) on most days of the week (17,20). Because pedometers do not include time-sampling capabilities, it is difficult to parse out steps accrued during moderate or vigorous activity from those accrued during the normal activities of daily living. By comparing step count data over 14 d with temporally matched data from the Physical Activity Recall, we hope to clarify the relationships between step counts and other measures of activity. Collectively, these studies will add to the accumulating knowledge base regarding the application of pedometers for public health research.
The participants in this study were 31 adult employees (17 men, 14 women) of the Cooper Aerobics Center in Dallas, TX. The study population was made up of participants with different educational backgrounds and experiences including physicians, scientists, professional staff, research assistants, interns, and custodial staff. The mean ages for the study participants were 31.2 and 26.2 yr for the men and women, respectively. The overall study protocol was approved by the Institutional Review Board at the Cooper Institute, and written informed consent was obtained from all participants.
In study 1, participants completed three 1-mile (1.61-km) trials of walking/jogging under both track and treadmill conditions. Participants wore a Digi-Walker step counter on their right hip for both conditions and completed the distance at three matched speeds. The order of speeds was randomized, but all participants completed the outdoor track condition first to establish their personal target pace to replicate on the treadmill. Bassett (1) reported differences in pedometer output when monitors were worn on both hips, but this trend was not evident for the Digi-Walker unit. Therefore, one device was deemed appropriate to represent overall step counts in these trials.
During the track condition, participants were asked to walk a 15-min mile (6.44 km·h−1), jog a 10-min mile (9.66 km·h−1), and jog an 8-min mile (12.08 km·h−1). Measurement staff reported the current pace to the participant at each quarter-mile split to keep the participant close to the target speed. At the end of each mile, the actual time and number of steps on the Digi-Walker were recorded. Fifteen participants were not willing to do the run condition, so the sample size was reduced to 14 for this pace (7 men, 7 women).
For the treadmill condition, participants reported to the Clinical Applications Laboratory at the Cooper Institute. Various anthropometric measures were obtained to examine the potential factors influencing variability in step counts. Height and weight were measured with standard laboratory scales. Body composition was assessed with a three-site skin-fold procedure (12) using Lange calipers (Cambridge, MD). The measurements were obtained by a research assistant in the lab who was previously trained and highly experienced with skin-fold assessments. Leg length was measured from the anterior superior iliac spine to the medial malleolus with a standard tape measure. Stride length was measured by recording the steps required to walk and jog a standardized distance (37.4 m) in an indoor hallway. This controlled activity was also used to verify step recordings on the Digi-Walker. For this check, two measurement personnel independently counted the number of steps required for both the walking and jogging trial. Participants then completed the indoor walk/jog protocol on a Trackmaster TM310 treadmill in the laboratory. Measurement personnel programmed the treadmill speeds and timed the subjects to match the speed and time of the outdoor track condition. The number of steps taken was recorded for each test pace (walk, jog, and run).
Descriptive statistics were computed for all of the demographic and anthropometric variables. A one-way analysis of variance (ANOVA) was performed to test for potential gender differences in these variables. A three-way (gender × pace × site) repeated-measures ANOVA was used to examine differences in step counts under the different conditions and paces. Correlations with various anthropometric variables were used to examine potential sources of variability in the step count data between sites and for the different paces.
In study 2, the goal was to examine the utility of the Digi-Walker to assess activity under field conditions. Participants were instructed to wear the Digi-Walker on the right hip for a 1-wk period under two different conditions. In condition 1, participants wore the Digi-Walker during all waking hours (except for bathing and swimming). In condition 2, participants wore the Digi-Walker during the whole day but removed the device during structured moderate or vigorous physical activity. The order of conditions was randomized across participants and all participants were scheduled to complete both weeks of monitoring. Throughout the 2 wk period, the participants completed a daily log in which they recorded the times they woke up and went to bed and the times the Digi-Walker was put on and taken off. Participants also recorded the number of steps from the Digi-Walker at the end of each day. A subsample of eight participants wore a monitor on each hip to evaluate the intra-instrument reliability during bothconditions.
At the end of each week, participants completed a 7-d physical activity recall (PAR) to provide a comparison measure of physical activity over the course of the week. Two trained interviewers who completed the standardized training procedures at the Cooper Institute conducted the PAR assessments. During the interview, participants were asked to report on time spent sleeping, sitting, and performing moderate, hard and very hard physical activity over the past week. The interview protocol utilizes a segmented recall format that allows an individual to recall morning, afternoon and evening activities separately. Examples of activities that fit each activity category are provided on a sheet and described by the interviewer. Following each interview, the total time spent in each category was recorded for each day along with the corresponding step count data. Previous studies have supported the reliability and validity of the PAR as a measure of physical activity for adults (5,13,19). A major advantage of the PAR is its ability to record detailed information about different levels of activity across a full week.
Data from the Digi-Walker and the PAR were processed on a daily basis and averaged across 7 d to reflect typical daily activity levels for the week. Mean steps per day and mean minutes in each of the PAR categories were computed for each of the participants for both conditions. Energy expenditure estimates from the PAR were also calculated for each day using a modified version of the traditional PAR scoring algorithm (2). Typically, all time that is not spent sleeping or in moderate, hard, or very hard activity is presumed to be “light” and is scored as 1.5 METs. A limitation of this approach is that it does not consider the varying levels of occupational or home tasks that are performed in the course of the day. A revised algorithm that codes sitting time as “rest” has recently been developed and preliminary work suggests that it may yield a more accurate assessment (23). Because sitting time would be likely to be inversely proportional to daily step counts, we used the alternative scoring system for the present study to better incorporate inactivity during the day. Standardized energy expenditure estimates (kcal·kg−1·d−1) were used to avoid introducing bias in results due to body weight.
Descriptive statistics were computed to reveal the activity patterns of participants in the study. Correlations were computed between average daily step counts and several activity indices from the PAR to examine relationships occurring under real-life conditions. Relationships with condition 1 reflect total activity patterns since the pedometers were kept on during all activities. Because the participants in condition 2 removed the monitor during all structured vigorous activity, relationships among measures from this condition reflect only light and moderate levels of activity that are accumulated throughout the day.
Study 1—Treadmill Track Comparison
The purpose of study 1 was to determine the number of steps required to complete a standardized distance under different conditions. To examine the reliability and validity of the monitors for recording steps, we first compared measured step counts with those recorded by two independent observers over an indoor hallway (37.4 m long). Across all participants, the mean step counts during this pilot test were within 3–5% of the recorded values. For the walk condition, the step count was within 10% of the observed steps for 81% of the trials (26 of 31), and the intraclass reliability was moderate (R = 0.56). For the jog condition, the step count was within 10% of the observed count for 90% of the trials (28 of 31), and the intraclass reliability coefficient was high (R = 0.89).
The descriptive characteristics of the study participants are provided in Table 1. The relatively low BMI and percentage fat levels among study participants suggest that they are leaner and probably more fit than the general population. Men were taller, heavier, and leaner than the women in the study (P < 0.05). Leg length was also significantly greater in men than in women (P < 0.05). This likely led to the differences in stride length (P < 0.05) observed between men and women for the walking pace. Interestingly, there was no difference in stride length observed under the jogging condition.
The three-way repeated-measures ANOVA revealed nonsignificant differences with site (i.e., no interactions or main effects), suggesting that similar results are obtained under both treadmill and track conditions. As expected, there were differences in step counts by pace, but the gender-by-pace interaction also was significant [F (2,41) = 6.37, P < 0.05]. Step counts decreased with increasing pace for both genders, but the effect was slightly steeper in women (Fig. 1). The mean number of steps observed for men was 1875, 1605, and 1307 for the walk, jog, and run paces, respectively. The corresponding values for women were 1996, 1662, and 1330. Simple effects for pace were significant for both men [F (2,36) = 94.1, P < 0.001] and females [F (2,34) = 251.7, P < 0.001], with significant univariate differences (Tukey post hoc tests) observed for all pairwise comparisons (P < 0.05). Gender differences were observed for both the walk and jog paces, but not for the running pace.
Step counts were related to a variety of anthropometric variables (Table 2). Step counts for all paces were negatively related to height, weight, leg length, and stride length and were positively related to body fatness. Stronger correlations with anthropometric variables were evident for walking compared with jogging paces but there were no major differences in correlations between the treadmill and field conditions.
Study 2: Field Monitoring of Activity
In study 2, we examined the relationships between daily step counts and various activity indices provided by the PAR during two weeks of monitoring. Participants reported keeping the monitors on for an average of 15:04 h in condition 1 and 15:14 h in condition 2. The descriptive statistics for the two conditions are provided in Table 3. There was no significant difference in total energy expenditure or levels of physical activity reported on the PAR between the weeks and the intraclass correlation between weeks was high (R = 0.71).
In condition 1, participants kept the Digi-Walker on for all activities, so the values reflect average daily activity levels over a full week. Participants reported a mean of 49.2 min of physical activity a day (sum of moderate, hard, and very hard). The energy expenditure values estimated from the PAR were approximately 32.86 kcal·kg−1·d−1 (137.5 kJ·kg−1·d−1), with approximately 4.64 kcal·kg−1·d−1 (19.41 kJ·kg−1·d−1) attributable to physical activity. The mean step counts were 11,603 d−1 but there was wide variability among the participants. The coefficient of variation (SD/Mean) was 0.44. Correlations between average daily step counts and average daily energy expenditure were low (r = 0.34). Correlations were slightly higher (r = 0.39) for comparisons with total minutes of physical activity (Fig. 2).
In condition 2, participants kept the Digi-Walker on during their normal daily tasks but removed the Digi-Walker during all structured moderate to vigorous activity. By subtracting off the contribution of hard and very hard activity reported on the PAR, this condition allows the relationships between step counts to be examined specifically for light and moderate activity. To check compliance with this aspect of the study protocol, we computed correlations between averaged measures of physical activity recorded on the PAR and average step counts. The correlations were near zero for both total minutes of activity (r = 0.06) and calories due to total physical activity (r = −.07). The lack of a relationship here supports the premise that individuals did remove the pedometers during their structured bouts of physical activity.
To examine the relationship between step counts and moderate and lower intensity activities, we recalculated the energy expenditure values without the contributions from hard and very hard activity. No correction was made for the missing time, so the data reflect a time span of less than 24 h for some participants. The correlation between the average step counts and average energy expenditure was r = 0.49 (Fig. 3). Although the correlation is higher than that observed for total activity, there is still considerable variability in the measures, even when the data are averaged over a 7-d period. A major factor contributing to this variability may be the differences in job tasks and lifestyle activity patterns of the participants. Job descriptions were not obtained in the study, but each participant provided an estimate of their sitting time during the PAR interview. The correlation between average weekday step counts and average weekday sitting time was r = −0.38 (Fig. 3).
Walking is often recommended to the public as a means of satisfying current physical activity guidelines. Pedometers provide an objective and inexpensive way to monitor walking (and other forms of activity) and therefore have potential for public health research on physical activity. In this project, we sought to examine the utility of using pedometers to objectively monitor physical activity. Study 1 was designed to test the Digi-Walker in a controlled setting, and study 2 was designed to test the device under field conditions.
The results for study 1 indicate that walking or running a mile (1.61 km) requires about 1300–2000 steps (depending on pace and anthropometric characteristics). As expected, higher step counts were found for walking paces compared with jogging or running because average stride length is shorter. Although we observed a significant gender × pace interaction, the differences between genders were small and probably not practically significant. The interindividual variability in step counts for the different paces was related to several anthropometric variables, but body weight, length, and stride length are all highly correlated with height. Therefore, the differences between genders may simply reflect differences in height between men and women.
The results were consistent with the Bassett et al. study (1) on several accounts. Both studies found no difference in step counts for walking or jogging on different surfaces. Both studies also found reliability to be worse for slower speeds. In the reliability check conducted in our study, we found higher intraclass reliability and closer correspondence in step counts for the jogging pace compared with the walking pace. Bassett et al. suggested that the errors for slow walking paces are caused by less pronounced accelerations at the hip. As described, we observed moderate relationships between step counts and various anthropometric variables but no clear trends revealing any major differences between walking and running. One observation that may be noteworthy is that the participant’s predetermined walking stride length was more highly correlated with steps for the walking pace and the predetermined jogging stride length was more highly correlated with the steps for the jogging/running paces.
The step count data from study 1 provide useful benchmarks for quantifying activity levels with the Digi-Walker. For example, using the step count data for the walking pace (average of 1935 steps per mile), one could conclude that approximately 3800–4000 steps would be sufficient to satisfy the current activity guidelines of 30 min of moderate intensity activity or two miles or walking at 4 mph (17,20). This figure cannot be used however to categorize an active person if monitoring is done over the course of a whole day. The reason for this is that most people will accumulate a large number of steps going through their normal activities of daily living. Because much of this activity is probably of light intensity, it may not provide meaningful health benefits. Therefore, if the Digi-Walker is used to assess daily activity patterns, the target values for daily steps must be set higher.
Two studies in Japan (11,24) have used the level of 10,000 as a behavioral target for intervention research; however, it is not clear how this figure was determined. Despite the apparent lack of empirical evidence, this guideline has been promoted in various activity campaigns and has gained general acceptance as a realistic behavior target. Although the present study was not designed to test this guideline, exceeding the level of 10,000 steps did appear to correspond with reaching the traditional public health activity guidelines (e.g., > 30 min of moderate activity and/or 20 min of vigorous activity), but there was considerable variability in the classifications. For example, when the data from study 1 were examined separately across days, participants who performed more than 45 min of total activity in a day reached the 10,000 Figure 77% of the time. The percentage was similar (73%) if the criteria of 30 min of activity was used as the criteria. Conversely, 29% of the days with less than 30 min of self-reported activity were still above 10,000 steps.
The lack of a strong correspondence between the daily step count totals and measures from the PAR is also evident in the modest correlations observed between the different measures. Low to moderate correlations were observed between the averaged daily step counts and average energy expenditure values for both condition 1 (r = 0.34) and condition 2 (r = 0.49). The fact that the correlations were somewhat higher for condition 2 was expected and reflects the fact that there is less error (for pedometers) associated with assessing moderate intensity activity than for vigorous activity. For example, runners would accrue fewer steps than walkers for the same distance. Pedometers would also tend to underpredict activity levels for other activities that are commonly performed as vigorous aerobic activity in fitness centers (e.g., indoor ski machines, stair stepping machines, and bicycling). This type of intensity bias must be considered when interpreting data from pedometers.
Variability in step count data can also be attributed to differences in occupational work tasks or activities of daily living. Participants in the present study had a variety of different occupations and steps during typical job tasks appeared to influence the relationships with the PAR. The range of sitting time per weekday ranged from 1:36 to 12:00 h a day (mean = 6:13/SD = 2:36) when averaged over the 5 weekdays. Using data from condition 2, we obtained moderate correlations between average daily step counts and average sitting time (work days only). The magnitude of this correlation (r = −.38) was about the same magnitude as that observed for the correlation with min of activity (r = 0.39). Therefore, in the present study, step counts were as highly correlated with inactivity as they were with activity. The fact that sitting time would relate this closely with step counts suggests that future work with pedometers should consider levels of inactivity and/or stratify the sample by types of occupational activity.
Although the nature of activity patterns can explain some of the results, it also must be acknowledged that the PAR has some limitations as a criterion measure for this type of study. Because the PAR includes only bouts of activity longer than 10 min, it is possible for an individual to accumulate a lot of daily physical activity but still report little or no moderate physical activity. This is a likely problem for individuals that are highly active as part of their normal work tasks. As described, these individuals could accumulate a large number of steps doing their normal job tasks but still be essentially inactive according to the PAR. This may have occurred to a considerable extent in our sample as the mean step count recorded on days with no reported physical activity was 7439 ± 3797 (condition 1). Although this effect certainly influenced our results, we believe it is more reflective of the inherent measurement challenges than the choice of comparison measure. Similar problems would likely occur if pedometers were compared against any other criterion measure.
Collectively, the results of this study reveal the inherent challenges associated with monitoring daily activity patterns with pedometers. Observations from several other studies also reveal the complexities of collecting and interpreting step count data. In a recent weight loss trial (8), the mean steps per day ranged from 3916 to 15,383 (mean: 9155/SD: 2990) for a group of overweight middle-aged women. The number of min of reported activity greater than 5 METs ranged from 0 to 56 min (mean: 19.6/SD: 17.7). The authors did not stratify the data to look at different levels of activity, but it is apparent that even inactive individuals would have had almost 4,000 steps a day. The mean for the group was also almost 10,000 for a fairly modest amount of moderate activity.
A large population study of occupational activity (18) reported average step counts ranging from 6700 to 11,900 for working adults (N = 493). Stratification by occupational status yielded mean steps for inactive occupations ranging from 6700 to 7300 for women and men, respectively. Corresponding values for heavy labor were 9800 and 10,800. When the data were stratified by leisure activity level, individuals pursuing “fitness training” were found to have mean step counts of 10,200–10,500 regardless of occupational activity.
An interesting observation from this study was that individuals pursuing some form of physical activity program outside of their work appeared more likely to accumulate greater than 10,000 steps a day, whereas those reporting only occupational activity were less likely to reach 10,000 steps a day. We observed similar results in our study as the mean step count in condition 2 was less than 10,000 steps per day when structured activity was not included. A useful analysis that was not performed in the Sequeira et al. study (18) would have been to stratify the data by both occupational and leisure time activity. This comparison would provide information on the additional increment in steps that would occur from structured activity. This would also provide an indicator of the degree of “compensation” in leisure time activity from active or inactive occupations.
In summary, although pedometers can be used to accurately measure steps taken and distance walked (2), there are a number of factors that complicate the interpretation of step count data for field research. We observed variability due to different intensities of activity and variability associated with differences in amount of sitting during the day. Although pedometers offer considerable promise for assessing daily physical activity patterns, additional work is needed to clarify daily step patterns across different ages, genders and occupational classes. To facilitate the development of step count guidelines for public health research it would also be important to better understand the independent contributions of work and leisure-time physical activity on daily step counts. Although research has supported the health benefits of accumulating moderate intensity activity throughout the day, the importance of accumulating a lot of light activity through the day has not been established. The use of time-based activity monitors may be particularly useful in examining daily step patterns in more detail.
This study was completed while the lead author was employed at the Cooper Institute. The authors thank the staff and interns who assisted with the data collection for the study, especially J. Abate, A. Arola, C. Fuchs, and S. Symington. We thank the staff and interns of the Cooper Institute and the Cooper Fitness Center for their willingness to be participants in the study.
This work was supported by the International Life Sciences Institute Center for Health Promotion (ILSI CHP). The use of trade names and commercial sources in this document is for purposes of identification only and does not imply endorsement by ILSI CHP. In addition, the views expressed herein are those of the individual authors and/or their organizations and do not necessarily reflect those of ILSI CHP.
1. Bassett, D. R., Jr., B. E. Ainsworth, S. R. Leggett, et al. Accuracy of five electronic pedometers for measuring distance walked. Med. Sci. Sports Exerc. 28: 1071–1077, 1996.
2. Blair, S. N. How to assess exercise habits and physical fitness. In:Behavioral Health: A Handbook of Health Enhancement and Disease Prevention,
D. Matarazzo, N. E. Miller, and S. M. Weiss (Eds.) New York: John Wiley & Sons, 1984, pp. 424–447.
3. Bouten, C. V. C., W. P. H. G. Verboeket-Van De Veene, K. R. Westerterp, M. Verduin, and J. D. Janssen. Daily physical activity assessment: comparison between movement registration and doubly labeled water. J. Appl. Physiol. 81: 1019–1026, 1996.
4. Differding, J., G. J. Welk, P. Hart, J. Abate, and S. Symington. The use of the Digi-Walker step counter to measure levels of physical activity (Abstract). Med. Sci. Sports Exerc. 30: S11, 1998.
5. Dishman, R. K., and M. Steinhardt. Reliability and concurrent validity for a 7-d recall of physical activity in college students. Med. Sci. Sports Exerc. 20: 14–25, 1988.
6. Eston, R. G., A. V. Rowlands, and D. K. Ingledew. Validity of heart rate, pedometry, and accelerometry for predicting the energy cost of children’s activities. J. Appl. Physiol. 84: 362–371, 1998.
7. Fehling, P. C., D. L. Smith, S. E. Warner, and G. P. Dalsky. Comparison of accelerometers with oxygen consumption in older adults during exercise. Med. Sci. Sports Exerc. 31: 171–175, 1999.
8. Fogelholm, M., H. Hilloskorpi, P. O. Laukkanen, W. V. M. Lichtenbelt,and K. Westerterp. Assessment of energy expenditure in overweight women. Med. Sci. Sports Exerc. 30: 1191–1197, 1998.
9. Freedson, P., E. Melanson, and J. Sirard. Calibration of the Computer Science and Applications, Inc. accelerometer. Med. Sci. Sports Exerc. 30: 777–781, 1998.
10. Gayle, R., H. J. Montoye, and J. Philpot. Accuracy of pedometers for measuring distance walked. Res. Q. 48: 632–636, 1977.
11. Hatano, Y. Use of the pedometer for promoting daily walking exercise. Int. Council Health, Phys. Educ. Recr. 29: 4–28, 1993.
12. Jackson, A. S., and M. L. Pollock. Practical assessment of body composition. Physician Sportsmed. 13: 76–86, 1985.
13. Jacobs, D. R. Jr., B. E. Ainsworth, T. J. Hartman, and A. S. Leon. A simultaneous evaluation of 10 commonly used physical activity questionnaires. Med. Sci. Sports Exerc. 25: 81–91, 1993.
14. Matthews, C. E., and P. S. Freedson. Field trial of a three-dimensional activity monitor: comparison with self report. Med. Sci. Sports Exerc. 27: 1071–1078, 1995.
15. Nichols, J. F., C. G. Morgan, J. A. Sarkin, J. F. Sallis, and K. J. Calfas. Validity, reliability, and calibration of the Tritrac accelerometer as a measure of physical activity. Med. Sci. Sports Exerc. 31: 908–912, 1999.
16. Pambianco, G., R. R. Wing, and R. Robertson. Accuracy and reliability of the Caltrac accelerometer for estimating energy expenditure. Med. Sci. Sports Exerc. 22: 858–862, 1990.
17. Pate, R. R., M. Pratt, S. N. Blair, et al. Physical activity and public health: a recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA 273: 402–407, 1995.
18. Sequeira, M., M. Rickenbach, V. Wietlisbach, B. Tullen, and Y. Schutz. Physical activity assessment using a pedometer and its comparison with a questionnaire in a large population survey. Am. J. Epidemiol. 142: 989–999, 1995.
19. Taylor, C. B., T. Coffey, K. Berra, R. Iaffaldano, K. Casey, and W. L. Haskell. Seven-day activity and self-report compared to a direct measure of physical activity. Am. J. Epidemiol. 120: 818–824, 1984.
20. U. S. Department of Health and Human Services. Physical Activity and Health: A Report of the Surgeon General.
Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, 1996.
21. Washburn, R., M. K. Chin, and H. J. Montoye. Accuracy of pedometers in walking and running. Res. Q. 51: 695–702, 1980.
22. Welk, G. J., and C. B. Corbin. The validity of the Tritrac-R3D activity monitor for the assessment of physical activity in children. Res. Q. Exerc. Sport 66: 202–209, 1995.
23. Welk, G. J., R. W. Thompson, D. I. Galper, and A. D. Dunn. The validity of the 7-Day Physical Activity Recall against the Tritrac-R3D Activity Monitor. (Abstract). Med. Sci. Sports Exerc. 29: S44, 1997.
24. Yamanouchi, K., K. Shinozaki, and K. Chikada. Daily walking combined with diet therapy is a useful means for obese NIDDM patients not only to reduce body weight but also to improve insulin sensitivity. Diabetes Care 18: 775–778, 1995.