Virtual reality (VR) simulators may enhance surgical resident colonoscopy skills, but the duration of skill retention and the effects of different simulator training methods are unknown. Medical students participating in a randomized trial of independent (automated simulator feedback only) versus proctored (human expert feedback plus simulator feedback) simulator training performed a standardized VR colonoscopy scenario at baseline, at the end of training (posttraining), and after a median 4.5 months without practice (retention). Performances were scored on a 10-point scale based on expert proficiency criteria and compared for the independent and proctored groups. Thirteen trainees (8 proctored, 5 independent) were included. Performance at retention testing was significantly better than baseline (median score 10 vs. 5, P < 0.0001), and no different from posttraining (median score 10 vs. 10, P = 0.19). Score changes from baseline to retention and from posttraining to retention were no different for the proctored and independent groups. Overinsufflation and excessive force were the most common reasons for nonproficiency at retention. After proficiency-based VR simulator training, colonoscopy skills are retained for several months, regardless of whether an independent or proctored approach is used. Error avoidance skills may not be retained as well as speed and efficiency skills.