Background Gaze training is an effective way of training basic laparoscopic skills, resulting in faster acquisition periods and more robust subsequent performance under pressure. The current study is a randomized control trial which examines whether the performance benefits of gaze training stand the test of time (delayed retention) and transfer to more complex skills. Methods Thirty-six medical students were trained to proficiency (50 trials) on a one-handed laparoscopic task (picking and dropping balls) in either a discovery learning (DL) or gaze training (GT) group. Both groups performed the one-handed task in baseline, retention and delayed retention (1 month) tests. They also performed baseline, retention and delayed retention tests of a two-handed task (grasping and cutting). Performance (completion time) and gaze control (target locking) were assessed throughout. Results For the one-handed task, the GT group displayed superior performance at retention (p \ .001), underpinned by more expert-like gaze control (p \ .05). The GT group also displayed superior performance in the one-handed task at delayed retention (p \ .005), underpinned by more expert-like gaze control strategies (p \ .001). Although the DL group's performance fell to 84 % of performance at retention, the GT group maintained performance at 100 % of retention. There were no differences between the groups for the two-handed task at retention (p = .140); however, at delayed retention, the GT group outperformed the DL group (p \ .005) and displayed more expert-like gaze control (p \ .001). Conclusions Novices trained to adopt an expert-like gaze control strategy were able to attain higher levels of performance more quickly than novices who learned by discovery alone. Furthermore, these skills were more durable over time and were transferable to more complex skills. Gaze training is a beneficial intervention to aid the acquisition of the basic motor skills required for laparoscopy.