Importance: Alcohol abuse correlates with gray matter development in adolescents, but the directionality of this association remains unknown. Objective: To investigate the directionality of the association between gray matter development and increase in frequency of drunkenness among adolescents. Design, Setting, and Participants: This cohort study analyzed participants of IMAGEN, a multicenter brain imaging study of healthy adolescents in 8 European sites in Germany (Mannheim, Dresden, Berlin, and Hamburg), the United Kingdom (London and Nottingham), Ireland (Dublin), and France (Paris). Data from the second follow-up used in the present study were acquired from January 1, 2013, to December 31, 2016, and these data were analyzed from January 1, 2016, to March 31, 2018. Analyses were controlled for sex, site, socioeconomic status, family history of alcohol dependency, puberty score, negative life events, personality, cognition, and polygenic risk scores. Personality and frequency of drunkenness were assessed at age 14 years (baseline), 16 years (first follow-up), and 19 years (second follow-up). Structural brain imaging scans were acquired at baseline and second follow-up time points. Main Outcomes and Measures: Increases in drunkenness frequency were measured by latent growth modeling, a voxelwise hierarchical linear model was used to observe gray matter volume, and tensor-based morphometry was used for gray matter development. The hypotheses were formulated before the data analyses. Results: A total of 726 adolescents (mean [SD] age at baseline, 14.4 [0.38] years; 418 [58%] female) were included. The increase in drunkenness frequency was associated with accelerated gray matter atrophy in the left posterior temporal cortex (peak: t_1,710 = -5.8; familywise error (FWE)-corrected P = 7.2 × 10^-5; cluster: 6297 voxels; P = 2.7 × 10^-5), right posterior temporal cortex (cluster: 2070 voxels; FWE-corrected P =.01), and left prefrontal cortex (peak: t_1,710 = -5.2; FWE-corrected P = 2 × 10^-3; cluster: 10624 voxels; P = 1.9 × 10^-7). According to causal bayesian network analyses, 73% of the networks showed directionality from gray matter development to drunkenness increase as confirmed by accelerated gray matter atrophy in late bingers compared with sober controls (n = 20 vs 60; β = 1.25; 95% CI, -2.15 to -0.46; t_1,70 = 0.3; P =.004), the association of drunkenness increase with gray matter volume at age 14 years (β = 0.23; 95% CI, 0.01-0.46; t_1,584 = 2; P =.04), the association between gray matter atrophy and alcohol drinking units (β = -0.0033; 95% CI, -6 × 10^-3 to -5 × 10^-4; t_1,509 = -2.4; P =.02) and drunkenness frequency at age 23 years (β = -0.16; 95% CI, -0.28 to -0.03; t_1,533 = -2.5; P =.01), and the linear exposure-response curve stratified by gray matter atrophy and not by increase in frequency of drunkenness. Conclusions and Relevance: This study found that gray matter development and impulsivity were associated with increased frequency of drunkenness by sex. These results suggest that neurotoxicity-related gray matter atrophy should be interpreted with caution.