Objective: To evaluate support for three hypotheses about the etiology of adolescent-onset ADHD symptoms: (1) a "cool" cognitive load hypothesis, (2) a "hot" rewards processing hypothesis, and (3) a trauma exposure hypothesis. Method: Participants (N=50) were drawn from two public high schools in a culturally diverse metropolitan area. A detailed procedure for identifying and confirming late-onset ADHD cases is described. Adolescents with late-onset ADHD (n=15) were identified and compared to childhood-onset (n = 17) and non-ADHD classmates (n = 18). Adolescents and parents completed measures of neurocognition, rewards processing, clinical profile, and environmental demands. Results: Late-onset cases were clinically and neurocognitively indistinguishable from childhood-onset cases; however, they experienced higher demands from parents (d=1.09). Compared to the non-ADHD group, late-onset cases showed significant deficits in metacognition (d=1.25) and academic motivation (d=0.80), as well as a pronounced history of multiple trauma exposure (OR=11.82). At one-year follow-up, ADHD persisted in 67.7% of late-onset cases. Late-onset cases (26.7%) were more likely than childhood-onset cases (0.0%) to transfer to alternative schools (OR=1.36) by one-year follow-up. Conclusions: Multiple factors may contribute to adolescent-onset ADHD. Adolescents with metacognition and motivation deficits may be at greatest risk for the late-onset ADHD phenotype, particularly in highly demanding environments. Exposure to traumatic stress may play a key role in the exacerbation of existing deficits or onset of new symptoms. Late-onset ADHD was persistent in most cases and associated with higher risk for school disengagement than childhood-onset ADHD. Further work is needed to better understand the etiologies of late-onset ADHD symptoms. [This paper will be published in "European Child & Adolescent Psychiatry."]