Several three-wave, four-variable models are estimated that specify causal links between measures of religious attitude and predispositions (religious commitment and an act-specific religious belief), involvement with marijuana-using friends, and self-reported use. The results provide evidence that the impact of religion is indirect through its influence on peer associations. The findings also show the emergence of an effect of the act-specific belief on behavior over time. This effect is interpreted to be more of a function of within-group attitude-behavior similarity due to social selection than to socialization to peer group standards. These findings extend rather than refute previous research that fails to control for the effects of peer influences. 3 figures, 2 tables, and 29 references. (Author abstract modified)