To address the question of predictive validity, three separate analyses were conducted using the net overall change score from the lab session, and number of binge drinking occasions at baseline (i.e., month before enrollment), and during the two-year follow-up period (eight quarterly assessments of past month binge drinking). First, zero-order correlations between the net overall change scores measured by both the B-BAES and BAES with the number of binge drinking episodes were compared at the two-year follow up. Second, alcohol response as measured by B-BAES net overall change scores at baseline were used to predict future drinking trajectories from baseline through the eight quarterly assessments in the two-year follow-up period, and compared to results previously published using the BAES (King et al., 2011). To determine drinking trajectories, a discrete mixture modeling approach determined by the Bayesian Information Criterion for model selection was used, and as previously reported (King et al., 2011), four trajectory groups (gradual maturing, moderate-frequency binge, high-frequency binge, and exacerbating; see Figure 4 in King et al., 2011) were found to best describe the data. Third, to examine
