With the ongoing trend of cannabis legalization in the United States and across the world, there is a critical need to better understand links between cannabis use, brain development, and cognitive-behavioral outcomes. Despite growing public belief that cannabis use is relatively benign, there remains a dearth of large longitudinal studies examining cannabis use and brain development in humans. We propose to leverage three large multimodal neuroimaging datasets, spanning preadolescence to late adulthood: ABCD (n=11,880; ages 9-20 followed longitudinally), IMAGEN (n=2,400; ages 14-23, followed longitudinally), and ENIGMA-Addiction (n=14,340; ages 12-80, from 118 studies). These large, rich datasets include measures of highly relevant behaviors (co-occurring alcohol and tobacco use, physical activity, psychopathology) while affording unparalleled statistical power. Leveraging these datasets, we will employ linear mixed-effects models to identify developmental windows in which areas and/or networks of the brain are most vulnerable to cannabis exposure. Similar analyses will be conducted for relevant behavioral trajectories (e.g., attention problems, internalizing symptomatology). Focusing on rigor and replicability, nonparametric permutation testing will be employed in imaging analyses, and we will explicitly test if findings in one dataset extend to others. We hypothesize that developmental windows with the greatest degree of prefrontal age-related change will be periods of greatest vulnerability to cannabis exposure. Using a range of methodologies (propensity score matching, cross-lagged panel design, discordant twin analyses, Bayesian causal networks), this proposal will move beyond associational analyses towards causal mechanisms. Such analyses are desperately needed to help bridge preclinical animal models with human neuroimaging findings. Based on our prior imaging work as well as existing rodent models, we hypothesize that early to middle adolescence, relative t