ABSTRACT Prevention is key for combating obesity, but few programs have prevented future increases in BMI and onset of overweight/obesity, particularly during late adolescence when youth often assume responsibility for dietary intake and exercise choices. One exception is a brief 6-hr dissonance-based program (Project Health) wherein participants make small lasting incremental lifestyle changes to dietary intake and exercise to reach energy balance, and discuss costs of obesity, an unhealthy diet, and sedentary behavior, and benefits of leanness, a healthy diet, and exercise, which prompts them to align their attitudes with their publically displayed behavior. These activities promote internalization of health goals and executive control over lifestyle behaviors. Late adolescents randomized to Project Health showed less increases in BMI and a 41% and 43% reduction in overweight/obesity onset over 2-yr follow-up compared to a version of the program lacking dissonance- induction activities and an obesity education condition. Project Health appears to be the first program to produce these key obesity prevention effects relative to an alternative intervention, but it is critical to increase effects. A dissonance-based prevention program was more effective when implemented in single- versus mixed-sex groups, theoretically because it promoted greater participation in dissonance-inducing discussions. Aim 1a is to test whether the weight gain prevention effects will be larger when Project Health is implemented in single-sex groups; we will randomize 450 17-20 year olds to complete Project Health in female, male, or mixed-sex groups, assessing outcomes at pretest, posttest, and 6, 12, 24, and 36-month follow-ups. Aim 1b is to test whether greater participation in dissonance-inducing discussions and group cohesion mediate the effect of condition on any superior weight gain prevention effects. Adolescents who show greater fMRI-assessed reward and attention region responsivity to food images exhibit elevated future weight gain, implying that reducing this responsivity may reduce future weight gain. In a pilot trial, late adolescents who completed go/no- go, stop-signal, and respond-signal computer training in which they repeatedly inhibit responses to high-calorie foods and respond to low-calorie foods, and dot-probe and visual-search computer tasks that train attention away from high-calorie foods and to low-calorie foods, showed a greater reduction in reward and attention region responsivity to, palatability rating of, and willingness to pay for, high-calorie foods, suggesting reduced valuation and attentional bias, as well as greater fat loss over 1-yr follow-up versus controls who completed the training with non-food images. Aim 2a is to test whether adding food response and attention training to Project Health will produce larger weight gain prevention effects. Participants in the 3 conditions will be randomized to complete response and attention traini...