Physical activity levels of an individual can vary from day to day. Such variability can make estimation of habitual physical activity difficult. The variability can be associated with behavioral (natural behavior change from one measurement period to the next) and analytic (inconsistent measurement outputs under the same set of circumstances) factors. To date, both sources of variability have been examined separately, but they can occur simultaneously while assessing daily physical activity. Therefore, the primary purpose of this study was to examine sources of variability in the measurement of physical activity levels for adolescents using generalizability theory. The secondary purpose was to investigate the minimum number of days required for monitoring physical activity levels of adolescents to determine their habitual physical activity when using a pedometer or an accelerometer. Thirty one (M = 12.3 years, SD = 0.5 years) middle school students wore two pedometers and two accelerometers during 5 weekdays and 4 weekend days. A two random facet (instrument and day) completely crossed design was conducted to examine seven sources of variability for pedometers and accelerometers separately. A total of three separate analyses were employed for pedometers across weekdays (W), weekends (WK), and weekdays and weekend days combined (WWK). The replicated analyses were employed for accelerometers. During W and WK, the primary sources of variability for pedometers were the variance components of individuals (49.13% and 32.33%, respectively) and the individual by day interaction (48.07% and 66.21%, respectively). Using accelerometers, large amounts of variability were related to individuals (53.90% and 25.92%, respectively), and the individual by day interaction component (29.96% and 57.37%, respectively) across W and WK. During WWK, a relatively large percentage of variability was associated with residuals for pedometers (35.23%) and accelerometers (35.59%). Relatively low variance was associated with instruments, and the individual by instrument and the day by instrument interaction components. Using one accelerometer, the minimum numbers of days required for monitoring daily physical activity levels of adolescents to estimate their habitual physical activity were 5 and 9 days during W and WK, respectively. Using one accelerometer, 4 and 14 days of monitoring physical activity were needed during W and WWK, respectively. Estimation of habitual physical activity levels during WWK using one pedometer and during WK using one accelerometer was problematic within reasonable length of time. These findings indicate that sources of variability in habitual physical activity are associated with behavioral variability rather than analytic variability. Keyword(s): assessment, exercise/fitness/physical activity, measurement/evaluation