Non-random patterns of aggregate youth smoking rates and trajectories across communities suggest that individual-level characteristics cannot account fully for the behavior in question. Instead, at least part of the explanation must lie somewhere within the community context. Such community-level contextual effects can impact directly both group and individual-level behavior (e.g. main effects), and they can also condition the effects of individual-level factors on individual behaviors (e.g. moderating effects). This paper reviews previous research examining community-level contextual effects regarding smoking and substance use more generally and identifies important limitations of this extant work, thus defining an agenda for future empirical studies. Next, the (in)compatibility of previous empirical findings with current theoretical models is discussed. In offering an alternative to these existing models, the paper concludes with presentation and discussion of a multi-level, integrated model of adolescent smoking trajectories. In this model, community/institutional forces are presumed to impact smoking above and beyond individual-level main effects. These posited community-level forces are broad and varied, representing school characteristics, neighborhood demographic characteristics, religious culture, media influence, economic context, health services and so on. In addition to exhibiting contextual main effects, the effects of community in the proposed multilevel model can be mediated by community-level processes, including the processes of control and socialization discussed herein. Also, community-level characteristics may interact in producing certain tobacco-use outcomes and, perhaps most importantly, they may moderate or condition the effects of interindividual differences on smoking.