Despite Graduated Driver Licensing (GDL), motor vehicle crashes remain a leading cause of death and injury for US adolescents. GDL typically applies to those <18 and restricts initial driving experience to lower risk conditions; however, these drivers <18 years enter licensure underprepared and have the highest crash rates. New evidence from our R21 showed an association between decreased crash rates and man- dated driver education and training in the context of GDL in Ohio, but few states require such training. In Ohio, which requires GDL and professional behind-the-wheel (BTW) training for licensure before 18 (those >18, exempt from GDL and training), those licensed <18 years demonstrate lower crash rates, higher licensing exam pass rates, and safer performance on a validated virtual driving assessment (VDA) than those licensed at 18. By contrast, Pennsylvania and most other states have no professional training requirements for young drivers, which can largely be attributed to the negative results of a randomized controlled trial (RCT) of pre- licensure training in DeKalb in 1983. Our Ohio results and major safety advances since DeKalb (GDL, stronger national driver training standards, and online training to address novice driver errors) moti- vated our proposed RCT of training (the 1st RCT in 3 decades). We propose a Phase III 3-arm RCT involv- ing 1000 adolescent novice drivers balanced by ages 16-18 years. With usual care in PA (GDL) as control, we will test 2 training interventions: Arm 2, ACCEL, an online training in hazard anticipation/response and atten- tion maintenance and Arm 3, 8 hours of state-of-the-art BTW training. We will examine 3 outcomes: (1) Smartphone-monitored driving from start of intervention period through 6 months post-licensure, (2) PA license exam result; and (3) VDA performance at licensure. Participants will be recruited from 5 Philadelphia-area practice sites (2 urban, 2 suburban, 1 rural) as part of routine adolescent care. All participants will undergo baseline assessment of driving performance (VDA) and a neurocognitive and personality battery. Aim 1: De- termine the effect of ACCEL or BTW training versus usual care on driving crash risk (e.g., rates of hard brak- ing) during the first 6 months post-licensure. Hypotheses: (1) Both ACCEL and BTW will reduce early licensure crash risk. (2) ACCEL will produce greater reduction in crash risk than BTW training because ACCEL targets skills that reduce crash risk. Aim 2: Determine the effect of ACCEL or BTW training versus usual care on skill acquisition. Hypotheses: (1) Both ACCEL and BTW will improve skill acquisition. (2) ACCEL will produce fewer errors on the VDA measure of skills than BTW; however, BTW may produce better license exam performance (e.g., parallel parking). Aim 3: Identify age-related risk factors for crashes that are resistant to training. Hypoth- eses: ACCEL and BTW will mitigate cognitive skills but will be less successful for impulse/personality...