Purpose: Temporal bone surgery requires excellent surgical skills and simulation-based training can aid novices’ skills acquisition. However, simulation-based training is challenged by early stagnation of performance after few performances. Structured self-assessment during practice might enhance learning by inducing reflection and engagement in the learning task. In this study, structured self-assessment was introduced during virtual reality (VR) simulation of mastoidectomy to investigate the effects on subsequent performance during cadaveric dissection. Methods: A prospective educational study with comparison with historical controls (reference cohort). At a temporal bone dissection course, eighteen participants performed structured self-assessment during 3 h of VR simulation mastoidectomy training before proceeding to cadaver dissection (intervention cohort). At a previous course, eighteen participants received identical VR simulation training but without the structured self-assessment (reference cohort). Final products from VR simulation and cadaveric dissection were recorded and assessed by two blinded raters using a 19-point modified Welling Scale. Results: The intervention cohort completed fewer procedures (average 4.2) during VR simulation training than the reference cohort (average 5.7). Nevertheless, the intervention cohort achieved a significantly higher average performance score both in VR simulation (11.1 points, 95% CI [10.6–11.5]) and subsequent cadaveric dissection (11.8 points, 95% CI [10.7–12.8]) compared with the reference cohort, who scored 9.1 points (95% CI [8.7–9.5]) during VR simulation and 5.8 points (95% CI [4.8–6.8]) during cadaveric dissection. Conclusions: Structured self-assessment is a valuable learning support during self-directed VR simulation training of mastoidectomy and the positive effect on performance transfers to subsequent cadaveric dissection performance.