Second, we did not measure supervision, which may have changed. However, to have an effect on injury rates, supervision would need to have increased in the intervention schools yet decreased
in the nonintervention schools, and we think this is unlikely.