Users of virtual reality (VR) technology, especially head-mounted displays (HMDs), often experience cybersickness, similar to motion sickness, with feelings of nausea, dizziness and sweatiness. Cybersickness typically increases with duration of wearing a HMD and is commonly evaluated in user studies with physiological measures, in-situ verbal reports and post session questionnaires. However, in addition to being time-consuming, user studies only provide insight into the specific configuration of the VR experience under study and can be limited to participant numbers, duration of VR exposure and the impact of cybersickness on VR experience dropout and completion rates. This paper presents a formal approach to modelling cybersickness. A Markov chain is used to define a general cybersickness model where probabilities represent changes in a user's state of cybersickness. The Markov chain can be populated with historical user study data and interrogated to gain further awareness of the VR experience under evaluation. The approach is exemplified with a custom Markov chain model generated from a public VR experience dataset. The resulting model is shown to be representative of the ground truth user experience from the source material. Examples are presented to demonstrate how the model can be explored to gain insight for (i) scaled up parameters, such as exposure duration and participant numbers, and (ii) acceptance thresholds for minimum/maximum cybersickness. Limitations on the generation of the model and its utility across different user populations and environment types are considered and discussed in the context of future work. INDEX TERMS Virtual reality, cybersickness, head-mounted display, Markov chain, prediction I. INTRODUCTION