The directional relationships suggested in this study do not completely rule out other types of relationships between the 2 behaviors. First, a cross-lagged model based on retrospective measures may not be able to capture a short-term disinhibitory effect of alcohol on antisocial behavior. Evidence from experimental studies indicates that alcohol can increase aggression and violence (Bushman and Cooper, 1990; Cherek et al., 1985; Miczek et al., 2004). Highly significant residual correlations between alcohol use and antisocial behavior may also indicate such short-term or cross-sectional relationships between the 2 behaviors, but the directional relationship between residuals could not be included in the model without fixing other parameters to a constant.
