Overall, the fit of the model with alcohol use (Figure 1) was acceptable [χ2(69)=138.15, p<.001; CFI=.91, RMSEA=.05, SRMR=.05]. There was significant continuity in own alcohol use, with stronger coefficients at older ages, as well as significant continuity in friends' alcohol use in adulthood, although not between adolescence and emerging adulthood (age 17 to age 23). Friends' and own alcohol use were positively correlated within each wave. In addition, friends' alcohol use at age 17 predicted higher levels of own alcohol use at age 23, indicating a peer socialization effect, and own alcohol use at ages 23 and 29 predicted higher proportion of heavy drinking friends at the subsequent wave, indicating friend selection effects (see Figure 1). Multigroup modeling indicated significant model differences between low and high risk DRD4 genotype groups [Δχ2(16)=35.61, p<.01]. Follow up analyses using Bonferroni correction revealed differences in one path and one covariance. Having more heavy drinking friends at age 29 predicted greater alcohol use at age 33 for those with the high risk DRD4 genotype (β=.21, p<.0l) but not those with the low risk genotype (β=.06,
