Index of everything on this site: https://ai.hiina.space/index.md
The evidence base is thin. Most studies have 15 to 30 participants, use one task and one session, and none tests Gorilla Tag-style hand walking, climbing or grapples. The cleanest small-N results all share one property: the player controls the velocity profile and can stop instantly. TriggerWalking (controller trigger "steps") raised SSQ by 17.6 versus about 143 for gamepad and keypad (omnibus p = 0.005; versus gamepad p = 0.009; versus keypad only marginal; N = 15) (Sarupuri et al.). PushPull world-drag had fewer symptoms than smooth joystick, but only in the smooth condition, and its usability was worse (SUS 66.0 versus about 75) (arXiv). Head-lean flight left SSQ unchanged where gamepad raised it significantly, and 16 of 22 preferred it (Adhikari et al. 2021).
My reading, which these studies do not test directly, is that the benefit comes from control over velocity and easy stopping, not from embodiment as such. A counterexample supports that reading: the Cybershoes foot-slider produced the most sickness in one maze study (abstract only, not re-verified) (DOAJ). Leaning is task-specific. It beat the gamepad for flying, yet a joystick beat every leaning seat, swivel chair included, for ground travel (Kitson et al. 2017 via nweon, secondary). So "leaning is better" is not a safe claim without naming the task.
| Rule | Bears on | Evidence | Strength |
|---|---|---|---|
| R1. Velocity is a direct, stoppable function of effort: impulse plus friction, a still hand or still pedals means a quick stop | Chair, rim, crank, climbing, Gorilla core | TriggerWalking, PushPull, Lone Echo's near-instant ramps | Weak to moderate; mechanism untested |
| R2. Treat artificial yaw as the main hazard. Real yaw (swivel chair) is free; otherwise snap at 22.5° or smooth at 15 to 35°/s | Wheelchair, paddling, mounts, monowheel | Snap rotation SSQ 29.8 vs 48.1 (N=28) (Farmani & Teather); Phantom and Wheelchair Basketball VR fixes | Moderate (small seated samples plus field reports) |
| R3. Stabilise viewpoint elevation and pitch; drive the camera from a low-passed anchor, not a bouncing bone | Mounts, monowheel, wheelies, piggyback | Meta guidance, qualitative (Meta) | Weak; the mount application is inference |
| R4. Use a world- or vehicle-fixed frame (cockpit, chair frame, horse neck), not a head-fixed one | All vehicles and mounts | A head-fixed nose gave no benefit (N=32, preregistered) (Springer); cockpits cut optic flow (Meta) | Moderate against a head-fixed frame; cockpit benefit unverified |
| R5. Prefer a physical-analogue steering input (handlebar, rim, reins) to gaze steering for ground vehicles | Bike, handcar, wheelchair, mounts | Handlebar steering gave the least sickness (N=18) (nweon, secondary) | Weak to moderate; secondary summary |
| R6. Blow air with a fan synced to speed | Bike, handcar, mounts | Airflow was the best countermeasure (N=24) (same, secondary) | Weak |
| R7. Make vignettes optional, not default-forced | All smooth motion | Helped in N=30 (ScienceDaily, secondary); 11 of 15 got sicker under 3× yaw gain, the other 4 showed no effect (Norouzi et al.) | Mixed |
| R8. Ship at least two comfort modes and ramp the first sessions | Everything | A "telesick" subgroup got sicker with teleport (Clifton & Palmisano); habituation is game-specific (Wollongong) | Moderate |
| R9. Budget arm fatigue: keep hands low, allow coasting and gliding, keep sessions short | Rim, crank, flapping, climbing | Drag x Drive arm strain (secondary); Flappy Arms 15 to 20 minute sessions; Consumed Endurance metric (Manitoba) | Weak |
| R10. Keep latency and jitter low; smooth on the remote side, not in the local camera | Trackers, co-op | 116 vs 246 ms pose latency changed SSQ (N=34, walking) (Sick Moves) | Moderate, from a different task |
Two caveats change how to read this table. Sickness varies widely by sex, susceptibility and gaming experience (Howard & Van Zandt 2021), so a solo developer's own tolerance is a poor test. Gorilla Tag's audience is also not an adult-lab population, a point we could not source. Presence and sickness correlate negatively in most small studies, but positive correlations exist too, and only two studies were adequately powered (Weech et al. 2019). That is no license to trade comfort for "immersion".
Text licensed CC BY 4.0 (attribution: Hiina, ai.hiina.space). Source: https://github.com/hiinaspace/ai-hiina-space