There are several limitations to this study. First, despite causal assumptions, its cross-sectional nature precludes determining the direction of causality. Longitudinal research is needed to replicate our findings in a prospective context, and to test competing models that provide similar patterns of results. For example, parents may have modified their parenting behavior in response to adolescent substance use. Second, since our measure of parental monitoring only contained three items and did not examine this construct for mothers and fathers separately, our results may have differed had we assessed monitoring individually for each parent. Our measure of parental closeness contained only two items, and future studies could employ better measures. Third, some of our subgroup analyses may have been underpowered, which could have limited our ability to detect significant effects.
