Expert information is a valuable resource in developing robust spatial models to improve understanding and prediction of systems in health, environment, business and society. However, getting experts to reliably activate and encode what they know has proven to be an elusive goal. A root cause of this elusiveness is the fact that expert knowledge is largely tacit, i.e., experts struggle to describe what they know. The imperative to address this goal is increasing. There is a constant need for better models of expert knowledge in organizations, much observational data is sparse and inadequate for spatial modelling, and many domains have knowledgeable workers leaving in numbers. Interviews are often used for eliciting expert knowledge, due to their ease of implementation. However, there is evidence that a lack of appropriate stimuli reduces the quality of knowledge elicited. This paper explores the use of immersive 3D virtual worlds for improved knowledge elicitation, due to their priming effects on memory recall. A case study on habitat prediction for a rock wallaby species is presented, in which the new approach is trialed. This paper is one of the first that aims to combine new virtual spatial technology with expert elicitation for improved spatial statistical modelling.