ResXR: infrastructure for quality-controlled, reproducible studies of human behavior in Extended Reality

Abstract

Extended Reality (XR) combines experimental control and ecological validity, yet behavioral XR research lacks shared infrastructure: building immersive experiments demands specialized engineering, and custom tools yield data in custom formats that other laboratories cannot readily reanalyze. We present ResXR (Research with XR), an open-source toolkit providing a path from immersive experiment to standardized dataset and quality report, running on standalone Meta Quest headsets. A Unity template records head, hand, eye, and face tracking with per-sample software development kit (SDK) timestamps and status flags; an independent Python pipeline checks data quality, can mask flagged intervals, and exports raw and derivative datasets in Motion-BIDS, the motion extension of the Brain Imaging Data Structure (BIDS), with self-contained quality reports. Three ready-to-run paradigms, covering discrete choice, spatial navigation, and free visual exploration, let researchers collect and process data without programming. ResXR grounds its default schema and the signals its quality checks read in stress tests rather than in vendor documentation: when we induced tracking failures on Meta Quest Pro headsets, several status flags kept marking the data as valid. With its code, stress-test protocols and example datasets openly available, ResXR provides a transparent, community-extensible foundation for reproducible XR experimentation.