Cognitive scientists increasingly use virtual reality scenarios to address spatial perception, orientation, and navigation. If based on desktops rather than mobile immersive environments, this involves a discrepancy between the physically experienced static position and the visually perceived dynamic scene, leading to cognitive challenges that users of virtual worlds may or may not be aware of. The frequently reported loss of orientation and worse performance in point-to-origin tasks relate to the difficulty of establishing a consistent reference system on an allocentric or egocentric basis. We address the verbalisability of spatial concepts relevant in this regard, along with the conscious strategies reported by participants. Behavioural and verbal data were collected using a perceptually sparse virtual tunnel scenario that has frequently been used to differentiate between humans' preferred reference systems. Surprisingly, the linguistic data we collected relate to reference system verbalisations known from the earlier literature only to a limited extent, but instead reveal complex cognitive mechanisms and strategies. Orientation in desktop VR appears to pose considerable challenges, which participants react to by conceptualising the task in individual ways that do not systematically relate to the generic concepts of egocentric and allocentric reference frames.