What Makes Something Feel like it Fits in Mixed Reality

Abstract

Mixed Reality (MR) allows digital objects to coexist in the physical world, but many MR experiences fail to achieve coherence, leaving virtual elements that feel out of place. This thesis investigates what visual and environmental factors a developer can modify to make virtual objects feel like they “fit” within a MR environment. To explore this, an interactive MR application was developed using Unity and the Meta Quest 3 SDK. In the experience, participants evaluated how well virtual objects fit within the real-world surroundings across four categories: model appearance (shaders, textures, and geometry), adding other virtual objects to the environment (vegetation density), visual effects (VFX), and lighting adjustments to the passthrough video. Twenty participants provided 1640 ratings on a 1 - 6 scale in addition to a qualitative survey and observational data from a short gameplay experience. Results show that coherence with the real-world environment is more important than photorealism. Furthermore, physically plausible shaders and desaturated color palettes consistently improved perceived fit, while oversaturation and inconsistent lighting reduced plausibility. Adding context to the environment such as vegetation also significantly increased ratings. Stylized effects performed well when they were internally consistent, while ambiguous or partially realistic elements often produced an uncanny effect. Lighting adjustments were highly context-dependent, with small, restrained changes outperforming extreme modifications. These findings suggest two effective design strategies for MR: either maintain strong coherence with the physical environment through subtle, consistent visual choices, or fully commit to stylization to avoid uncanny effects. Across both approaches, avoiding contradictions with user expectations is critical. This work contributes practical guidelines for designing MR experiences that better integrate with the real world and highlights the importance of context-aware and adaptive systems in future MR development.