Technological advancements have made it possible to create realistic virtual representations of the real world, although it is unclear in medical education whether high physical fidelity is required in virtual learning resources (VLRs). This study, therefore, aimed to compare the effectiveness of high-fidelity (HF) and low-fidelity (LF) VLRs for learning anatomy. For this study, HF and LF VLRs were developed for liver anatomy and participants were voluntarily recruited from two cohorts (cohorts 1 and 2). Knowledge outcomes were measured through pre- and post-tests, task outcomes including activity score and completion time were recorded and participants' perceptions of the VLRs were surveyed. A total of 333 participants (165 HF and 168 LF) took part in this study. Knowledge outcomes were higher for the HF activity in cohort 1 and for the LF activity in cohort 2, although not significantly. There were no significant differences in activity score within either cohort, although completion time was significantly longer for the HF activity in cohort 1 (P = 0.001). There were no significant differences within either cohort in perceptions of the VLRs regarding usefulness for reviewing conceptual knowledge, esthetics, quality, mental effort experienced, or future use, although the LF VLR was scored significantly higher regarding the value for understanding in cohort 1 (P = 0.027).This study suggests that high physical fidelity is not necessarily required for anatomy VLRs, although may potentially be valuable for improving knowledge outcomes. Also, level of prior knowledge may be an important factor when considering the physical fidelity of anatomy VLRs.