This study explored the effect of differing levels of interactivity in a VR language learning system on learner oral task performance. 30 participants were divided into 15 pairs which undertook two spot-the-difference tasks. One in a static-VR system, the other in an interactive-VR system which required them to move objects into the correct position in a 3D environment. Oral performance was measured along complexity, accuracy and fluency dimensions. Results suggested that the static-VR system pushed learners to produce more fluent and lexically complex output than the interactive-VR system. There was no significant difference in output in terms of accuracy. Reasons for these results are explored and implications for teachers provided. The paper concludes with recommendations for future research in this area.