Potential and limitations of virtual reality for the language sciences

Virtual reality (VR) is often described as a technology that increases realism, but its deeper significance for the language sciences is methodological and epistemological: bodies, contexts, and social contingencies become programmable variables, thereby reshaping what counts as a language experiment and what counts as language data. It is argued that VR has altered the conceptual landscape of psycholinguistics and language learning by enabling meaning to be studied as enacted behaviour—coordinated perception, movement, attention, and interaction—rather than as an internal product inferred from decontextualised reaction times and accuracy alone. The ways in which VR enables controlled coupling between language and action are discussed, including motor embodiment, gesture, effector-specific virtual actions, and manipulations of agency, body ownership, and self-location that are difficult or impossible to implement outside immersive environments. Limitations, risks, and biases are also examined, including hardware/software heterogeneity, cybersickness and selective attrition, novelty and workload confounds, commensurate baselines, and sampling and accessibility biases. Methodological guidelines are outlined to support cumulative progress, emphasizing decomposable experimental logic, construct-to-measure mappings, rigorous control conditions, multimodal synchronisation, and transparent reporting of technical parameters and tolerability. Finally, a ten-year roadmap is proposed across L1/L2 acquisition, language neuroscience, clinical rehabilitation (including aphasia), and human–computer interaction, highlighting the role of shared reference tasks, transfer-oriented outcomes, and responsible integration of AI-driven embodied conversational agents.