Research on the association between marijuana use and intimate partner violence (IPV) has generated inconsistent findings, and has been primarily based on cross-sectional data. We examined whether husbands' and wives' marijuana use predicted both husbands' and wives' IPV perpetration over the first 9 years of marriage (Wave 1, n = 634 couples). We also examined moderation by antisocial behavior, the spouse's marijuana use, and whether IPV was reported during the year before marriage. These predictive associations were calculated using a time-lagged multivariate generalized multilevel model, simultaneously estimating predictors of husband and wife IPV. In fully adjusted models, we found that more frequent marijuana use by husbands and wives predicted less frequent IPV perpetration by husbands. Husbands' marijuana use also predicted less frequent IPV perpetration by wives. Moderation analyses demonstrated that couples in which both spouses used marijuana frequently reported the least frequent IPV perpetration. There was a significant positive association between wives' marijuana use and wives' IPV perpetration, but only among wives who had already reported IPV perpetration during the year before marriage. These findings suggest there may be an overall inverse association between marijuana use and IPV perpetration in newly married couples, although use may be associated with greater risk of perpetration among women with a history of IPV perpetration.