Objective: To identify prospective predictors of mental health in Kabul, Afghanistan. Method: Using stratified random-sampling in schools, mental health and life events for 11- to 16-year-old students and their caregivers were assessed. In 2007, 1 year after baseline, the retention rate was 64% (n = 115 boys, 119 girls, 234 adults) with no evidence of selection bias. Self- and caregiver-rated child mental health (Strengths and Difficulties Questionnaire), depressive (Depression Self-Rating Scale), and posttraumatic stress (Child Revised Impact of Events Scale) symptoms and caregiver mental health (Self-Report Questionnaire) were assessed. Lifetime trauma and past-year traumatic, stressful, and protective experiences were assessed. Results: With the exception of posttraumatic stress, one-year trajectories for all mental health outcomes showed significant improvement (p less than 0.001). Family violence had a striking impact on the Strengths and Difficulties Questionnaire data, raising caregiver-rated scores by 3.14 points (confidence interval [CI] 2.21-4.08) or half a standard deviation, and self-rated scores by 1.26 points (CI 0.50-2.03); past-year traumatic beatings independently raised self-rated scores by 1.85 points (CI 0.03-3.66). A major family conflict raised depression scores by 2.75 points (CI 0.89-4.61), two thirds of a standard deviation, whereas improved family life had protective effects. Posttraumatic stress symptom scores, however, were solely contingent on lifetime trauma, with more than three events raising scores by 5.38 points (CI 1.76-9.00). Conclusions: Family violence predicted changes in mental health problems other than posttraumatic stress symptoms in a cohort that showed resilience to substantial socioeconomic and war-related stressors. The importance of prospectively identifying impacts of specific types of childhood adversities on mental health outcomes is highlighted to strengthen evidence on key modifiable factors for intervention in war-affected populations. (Contains 2 figures and 4 tables.)