One important limitation to the interpretation of study findings is that although we identified numerous predictors of teens adopting levels of substance use over time that were predicted by their friends initial levels of use, becoming more like one's friends in terms of substance use does not automatically equate to being directly influenced by them. For example, while direct influence processes may indeed operate, it could also be that teens select friends whose behaviors they would like to emulate, and then simply do so over time. For ease of description, we refer to both these direct and indirect processes by which teens adopt levels of substance use consistent with their friends' levels of use as peer influences. One finding in this study that supports this usage is that the predictive factors we identified were non-directional in nature (e.g., they predicted teens becoming more like their peers, whether that meant adopting higher or lower levels of substance use over time). This means that the predictive factors we identified were not simply indicators of a predilection toward higher substance use, but rather
