A number of studies have investigated tracking of physical activity from childhood to adolescence and, in general, these studies have been based on methods with some degree of subjectivity (e.g., questionnaires). The aim of the present study was to evaluate tracking of physical activity from childhood to adolescence using accelerometry, taking into account major sources of variation in physical activity. Both a crude and an adjusted model was fitted, and, in the adjusted model, analyses were corrected for seasonal variation, within-week variation, activity registration during night time sleep, within instrumental measurement error, and day-to-day variation in physical activity. In all, 208 subjects were included in the crude analyses. Stability coefficients estimated from the crude model were low (i.e., 0.18 and 0.19 for boys and girls, respectively) and only borderline significant. However, in the adjusted model highly significant stability coefficients of 0.53 and 0.48 for boys and girls, respectively, were observed. It was concluded that physical activity behavior tends to track moderately from childhood to adolescence.