The article investigates cognitive models of virtual reality space representation as a highly immersive computer-generated environment. The study aims to identify and describe the hierarchical structure of cognitive models underlying multimodal perception and comprehension of virtual reality space, as well as to establish general and modality-specific patterns of its representation. The relevance of the study is determined by the insufficient integration of verbal and visual representations into existing research. The methodological framework is based on the cognitive modeling of experimental data obtained from 20 participants who performed perception and representation tasks within a specially designed virtual reality environment. The verbal representations were analyzed with the multimodal annotation method, and with the tools of the Semograph IS, SciVi, and AntConc, while visual representations were processed using Creative Maps Studio followed by the Python-based analysis. The results revealed the multi-level system of cognitive models. The first level comprises verbal and illustrative models reflecting linear and configurational strategies of spatial representation. The higher level is a cognitive model, integrating both modalities and identifying stable cognitive patterns, including cyclic sequences, action coupling, and a transition from general to specific. In addition, a communicative-cognitive meta-level was identified, demonstrating the influence of interaction parameters of spatial experience organization. Thus, the virtual reality space representation is interpreted as a hierarchical multimodal and communicatively conditioned system.