Purpose To evaluate the variability of performance among novice ophthalmic trainees in a range of repeated tasks using the Eyesi virtual reality (VR) simulator. Methods Eighteen subjects undertook three attempts of five cataract specific and generic three-dimensional tasks: continuous curvilinear capsulorhexis, cracking and chopping, cataract navigation, bimanual cataract training, anti-tremor. Scores for each attempt were out of a maximum of 100 points. A non-parametric test was used to analyse the data, where a P-value of o0.05 was considered statistically significant. Results Highly significant differences were found between the scores achieved in the first attempt and that during the second (Po0.0001) and third (Po0.0001) but not between the second and third attempt (P ¼ 0.65). There was no significant variability in the overall score between the users (P ¼ 0.1104) or in the difference between their highest and lowest score (P ¼ 0.3878). Highly significant differences between tasks were shown both in the overall score (P ¼ 0.0001) and in the difference between highest and lowest score (P ¼ 0.003). Conclusion This study, which is the first to quantify reproducibility of performance in entry level trainees using a VR tool, demonstrated significant intra-novice variability. The cohort of subjects performed equally overall in the range of tasks (no inter-novice variability) but each showed that performance varies significantly with the complexity of the task when using this high-fidelity instrument.