The purpose of this study was to examine tracking of physical activity at home and recess in young children. Three hundred fifty-one Mexican-American and Anglo-American children entered the study (mean age = 4.4 yr), and 83% completed 2 yr of measurement. Physical activity was directly observed on 10 d over 2 yr. Measurement waves occurred every 6 months, and each wave consisted of 2 d of observation within 1 wk. Children were observed for 60 min at home on a weekday evening and up to 30 min during recess at preschool or school. Maximum likelihood procedures using a linear mixed-effects model indicated that most of the variance in home and recess physical activity was accounted for by short-term/weekly factors. Tracking of physical activity accounted for 15% of the total variance at home and 8% at recess. Pearson correlations for physical activity over time were higher at home than at recess. Tracking of home physical activity was r = 0.15 when single days were correlated and r = 0.36 when the means of 4 d were correlated. There was a small, but detectable, stable component of physical activity in young children, at least at home.