The Computer Science Application (CSA) accelerometer uses integrated circuitry and memory to provide a continuous recording of minute-by-minute movement counts. It has been previously validated as an objective monitor of children's physical activity in field and laboratory settings. Our purpose was to derive accelerometry summary variables reflective of different physical activity intensity levels, evaluate the stability of these summary variables, and define the number of days needed to adequately measure usual physical activity. A secondary study purpose was to compare three self-report questionnaires to accelerometry. Thirty children (7-15 yr) wore accelerometers for 12 h.d-1 for 6 d. Daily summary variables of average movement count (total physical activity) and daily frequency of sedentary through vigorous activity were constructed. Intraclass correlation coefficients (R) and 95% confidence intervals (CI) were used to analyze the data. Accelerometry stability using 1 monitored day to represent usual physical activity was R = 0.42-0.47. When 6 d were used, stability increased to R = 0.81-0.84. Acceptable intraclass correlations and CI were achieved with 4 d of monitoring (R = 0.75-0.78, CI = 0.60-0.88). The self-report questionnaires were poorly to moderately correlated to accelerometry variables (r = -0.03-0.51). Data indicate that in field settings: 1) accelerometry can be used to assess the intensity of children's activity and 2) 4 or more days of activity monitoring are needed to achieve satisfactory reliability.