Background: Virtual reality (VR) simulators and Web-based instructional videos are valuable supplemental training resources in surgical programs, but it is unclear how to optimally integrate them into minimally invasive surgical training. Methods: Medical students were randomized to proficiency-based training on VR laparoscopy and endoscopy simulators by two different methods: proctored training (automated simulator feedback plus human expert feedback) or independent training (simulator feedback alone). After achieving simulator proficiency, trainees performed a series of laparoscopic and endoscopic tasks in a live porcine model. Prior to their entry into the animal lab, all trainees watched an instructional video of the procedure and were randomly assigned to either observe or not observe the actual procedure before performing it themselves. The joint effects of VR training method and procedure observation on time to successful task completion were evaluated with Cox regression models. Results: Thirty-two students (16 proctored, 16 independent) completed VR training. Cox regression modeling with adjustment for relevant covariates demonstrated no significant difference in the likelihood of successful task completion for independent versus proctored training [Hazard Ratio (HR) 1.28; 95% Confidence Interval (CI) 0.96-1.72; p = 0.09]. Trainees who observed the actual procedure were more likely to be successful than those who watched the instructional video alone (HR 1.47; 95% CI 1.09-1.98; p = 0.01). Conclusions: Proctored VR training is no more effective than independent training with respect to surgical performance. Therefore, time-consuming human expert feedback during VR training may be unnecessary. Instructional videos, while useful, may not be adequate substitutes for actual observation when trainees are learning minimally invasive surgical procedures. © 2010 Société Internationale de Chirurgie.