Self-regulated learning is optimized when instructional supports are provided. We evaluated three supports for self-regulated simulation-based training: practice schedules, normative comparisons, and learning goals. Participants practiced 5 endoscopy tasks on a physical simulator, then completed 4 repetitions on a virtual reality simulator. Study A compared two practice schedules: sequential (master each task in assigned order) versus unstructured (trainee-defined). Study B compared normative comparisons framed as success (10% of trainees were successful) versus failure (90% of trainees were unsuccessful). Study C compared a time-only goal (go 1 min faster) versus time + quality goal (go 1 min faster with better visualization and scope manipulation). Participants (18 surgery interns, 17 research fellows, 5 medical/college students) were randomly assigned to groups. In Study A, the sequential group had higher task completion (10/19 vs. 1/21; P < 0.001), longer persistence attempting an ultimately incomplete task (20.0 vs. 15.9 min; P = 0.03), and higher efficiency (43% vs. 27%; P = 0.02), but task time was similar between groups (20.0 vs. 22.6 min; P = 0.23). In Study B, the success orientation group had higher task completion (10/16 vs. 1/24; P < 0.001) and longer persistence (21.2 vs. 14.6 min; P = 0.001), but efficiency was similar (33% vs. 35%; P = 0.84). In Study C, the time-only group had greater efficiency than time + quality (56% vs. 41%; P = 0.03), but task time did not differ significantly (172 vs. 208 s; P = 0.07). In this complex motor task, a sequential (vs. unstructured) schedule, success (vs. failure) orientation, and time-only (vs. time + quality) goal improved some (but not all) performance outcomes.