<p>Virtual Reality (VR) shows scientific phenomena with the representations digitalized in multimodal ways, and these multimodal representations play critical roles in facilitating inquiry with the affordances of each mode. This study aims to investigate the different patterns of multimodal representation and their epistemic features provided in VR-based inquiries depending on their reasoning types. The analysis focused on 31 VR simulations designed to support scientific inquiry in primary digital science textbooks in South Korea, examining their reasoning types and multimodal patterns. The results reveal intriguing patterns, showing how different combinations of multimodal representations are associated with distinct epistemic features, depending on their reasoning types. In phenomenon-based reasoning, succinct oral, written, and visual elements guided the inquiry, while auditory-visual aids represented phenomena. Relation-based reasoning combined oral, written, and visual elements to facilitate the connections between different variables in observed phenomena, while auditory and kinaesthetic elements provided technical support. Model-based reasoning used diverse combinations of multimodal patterns to help learners understand the structures of scientific models, with increased use of oral, written, and visual elements for explicit presentation of results. Finally, educational implications are discussed in terms of how to design and facilitate VR-based inquiry using multimodal patterns, their semiotic functions and epistemic features.</p>