Discomfort must be sampled continuously in-experience with comparable scales

What it is

Sickness is a climb, not a score that appears at the end. The Simulator Sickness Questionnaire (SSQ) is filled after doffing, collapsing a session into three subscales. It is fit for between-group contrasts and blind to which minute the slope started. To see the slope, ask one comparable question at a fixed interval during the experience — Fast Motion Sickness Scale (FMS, 0–20 spoken) or a short VRSQ. Same instruction, same anchors, same interval, or another lab and another build cannot be lined up. A single post-hoc “were you sick” is both late and incomparable.

The topic is how to measure discomfort as a time series, not why optic flow or stance sickens.

Why it happens

Symptoms lag and accumulate: the peak can sit a minute or two after the strongest stimulus, and can still be rising after the stimulus stops. End-only sampling writes the same total for “already bad mid-way, finished on willpower” and “flat until the last ten minutes”. Design decisions need the first or the second — cutscene, accel, teleport — and without time there is no attribution.

Scales also do not convert on a whim. SSQ’s nausea, oculomotor and disorientation weights come from flight simulators; in VR the ocular items are easily inflated by the headset itself, and treating totals as one stick across studies warps. FMS is a single spoken item, low intrusion, coarse resolution, good for a curve, not a diagnosis. VRSQ reworks the structure for VR; its items are still not item-wise identical to SSQ.

“Comparable” means three alignments: wording, tick marks, time grid. Align only one, and the curves will not stack.

Studying it

In one session, run in parallel: FMS (or 0–10) every 60 or 120 s, SSQ / VRSQ at the end, and if needed a delayed post (10–15 min after doffing). Plot individual curves; do not report only the end mean.

Independent variables: sampling interval, spoken versus panel prompt, whether mid-session dropout is allowed. Dependent variables: slope and time-to-peak of the curve, end-scale scores, dropout rate, how often the prompt itself broke presence.

The method failure mode: papers all say SSQ, with different instructions, different fill times (just-doffed / after a splash of water), different stop-anytime policies. Replication aligns the time grid first, not the questionnaire’s name.

Where it stops holding

Experiences shorter than a minute or two make continuous sampling denser than the stimulus; the prompt becomes a new load, and a pre/post pair is enough. Children, other languages, both hands occupied: a spoken single item is feasible where a multi-page form is not, but norms do not carry over from SSQ. Medical or safety claims cannot rest on FMS alone; that is an engineering curve. A lab that pauses the world for every rating measures an interrupted experience, which is not the same curve as an unpaused spoken prompt in a product.

Applying it

• Write discomfort sampling into the playtest protocol: fixed interval, fixed wording, permission to stop now. Do not wait until the session ends to send a long form.
• Keep the same stick and the same grid across versions. Switching FMS for VRSQ voids a claim of “20% better than last build”.
• Decide from the curve: whichever action the peak sits on is the one to change. Do not retune the whole locomotion system from an end total.
• How to check: a playtest must produce a time series. End score with no time-to-peak is a sampling failure. Two curves with different wording or interval must not be subtracted.

Related

• Same group: N1.13.1 The area and speed of optic flow determine sickness intensity · N1.13.2 Self-initiated motion causes less sickness than imposed motion · N1.13.3 Standing is more prone to imbalance and sickness than sitting
• Nearby: N1.06 Cybersickness · N4.07 Continuous Locomotion and Discomfort
• Search terms: SSQ · FMS · VRSQ