Previous research has shown that predictors of activity in adults depend upon the method of measurement. This study is designed to assess the predictors of activity in a sample of 59 obese children. Activity was measured using self-reported and TriTrac accelerometer METs. Self-report and TriTrac accelerometer measures were moderately correlated, r = 0.46, with the self-reported activity (2.3 METs) significantly greater than TriTrac (1.6 METs). Hierarchical regression analysis examined the influence of socioeconomic level, body composition, fitness, hedonics of child and adult activity behaviors, and decisional balance on self-reported and accelerometer-measured activity, controlling for child and parent psychopathology. Child and parent psychological symptoms accounted for 8.3% and 3.4% of the variance in accelerometer and self-reported METs, respectively. The model for accelerometer-measured activity showed socioeconomic level and parent self-report of activity accounted for 14.8% of the incremental variance in child activity. The model for self-report of child activity found that child fitness accounted for 23.5% of the incremental variance in child activity. These results suggest that the predictors of activity level are different based upon the method of measurement, consistent with research in adults.