Cybersickness is a measurable mismatch, and its standard measure is not commensurable

Cybersickness is the price the body pays for a display that moves the world without moving the person. The dominant explanation is a mismatch between the motion the senses report and the motion the brain predicts (Oman, 1990), But the field measures that feeling with a questionnaire built for flight simulators, and the score depends on the content, the system and the exposure length. The mechanism is shared; the numbers are not. This fifth of ten units takes the cost the body pays for those displays.

The mismatch is inherited from motion sickness research

Oman (1990) reviewed and evaluated the sensory conflict theory, in which sickness follows a disagreement between the motion the senses report and the motion the central nervous system expects. He states its weakness too. No conflict neuron or process had been identified physiologically, so the account is a framework rather than a proven circuit.

LaViola (2000) gave the category its founding statement, arguing that cybersickness is distinct from motion sickness because the user is typically stationary while the display moves. The mismatch can be produced with no physical motion at all, which is exactly the case a simulator questionnaire was never designed to fit.

The default instrument was built for flight simulators

Kennedy, Lane, Berbaum and Lilienthal (1993) developed the Simulator Sickness Questionnaire from the Pensacola motion sickness questionnaire by factor analysis, and it became the field's default. Chang, Kim and Yoo (2020) surveyed 77 studies and found 76 using a subjective measure, most often this questionnaire. Its 16 symptoms are rated from zero to three and combined into three subscales, and a total above about 33 marks high severity, a cut-off the review adopts rather than derives.

The instrument does not travel between systems

Stanney, Kennedy and Drexler (1997) factor-analysed a large set of simulator self-reports and found oculomotor symptoms dominating the simulator profile. Eight virtual environment experiments showed the profile reversed, with disorientation predominant and oculomotor the least, and total severity about three times greater. Their conclusion is that cybersickness is not simulator sickness.

If the profile changes with the system, a total summing the same items cannot mean the same thing across systems. Kim, Park, Choi and Choe (2018) rebuilt the instrument for head-mounted displays, retaining nine items in two components and dropping the nausea-weighted items the older total depends on. Fernandez-Canas and colleagues (2025) then appraised nine such instruments against COSMIN criteria and graded them: the Cybersickness in VR Questionnaire and the Virtual Reality Neuroscience Questionnaire were recommended, the simulator questionnaire and the VRSQ only acceptable, and two not recommended. A field with nine instruments and no dominant one has not settled what its scores mean.

Even inside one instrument, the score moves with the content

Saredakis and colleagues (2020) reviewed 55 studies of sickness after head-mounted display use, covering 3,016 participants, and pooled the scores. The pooled total across all content was 28.00 and gaming content scored highest at 34.26. Content type, visual stimulation, locomotion style and exposure length all moved the score, while the authors found no sex difference.

A quoted cybersickness rate without its content is therefore a range rather than a rate: the pooled figure sits below the high-severity line and the gaming figure above it, on the same instrument and the same scale.

Two blind spots: taken at the end, and silent about who left

Chang, Kim and Yoo (2020) name the limitation directly. Most questionnaires are administered after the experience, so they cannot describe discomfort in real time, and people differ in which symptom they treat as the threshold for reporting sickness. A post-hoc total flattens both the time course and the individual criterion, and a lower average can also mean those who felt worst stopped early and never answered.

Saredakis and colleagues (2020) pooled a mean dropout rate of 15.6 per cent across the 46 experiments that reported one, while 22 studies supplied no dropout figure at all, so the rate rests on the minority that said. A mean symptom score computed over whoever remained is not the mean over whoever was recruited.

The objective alternative is promising and unfinished

The obvious fix is to stop asking. Dennison, Wisti and D'Zmura (2016) measured physiological signals in seated participants who used a virtual environment first on a monitor and then through a head-mounted display. Sickness occurred only with the headset, and stomach activity, blinking behaviour and breathing predicted it, showing an objective channel recordable during exposure.

Weech, Varghese and Barnett-Cowan (2018) tested several candidate sensorimotor contributions in the same participants. A model built from them predicted 37 per cent of the variability, 16 per cent of it from balance control. Sway path length during vection was the strongest single predictor, inversely related at 7.5 per cent. Several processes matter, no single one decides, and most of the variance is unexplained.

Sickness moves the outcomes other units measure, and its cost is not shared equally

Weech, Kenny and Barnett-Cowan (2019) reviewed the relation between presence and cybersickness and concluded that the balance of evidence favours a negative relationship driven mainly by sensory integration. The two share causes, so a study that treats presence as an outcome while allowing sickness to vary is measuring both at once.

If sickness were evenly spread it would still be a design constraint; because it is not, it is an equity problem. Stanney, Fidopiastis and Foster (2020) ran two experiments with twenty minute roller coaster exposures. They found that the primary driver of the gender difference in sickness was an interpupillary distance which could not be fitted to the headset, with motion sickness history second. When the fit was correct, women's responses resembled men's.

Howard and Van Zandt (2021) meta-analysed the relation between sickness and individual differences, finding significant associations with motion sickness susceptibility, gender, prior experience and other user attributes a designer cannot change. They frame the consequence as disparate treatment: a programme that makes some groups sicker is unfair before it is uncomfortable.

The contested question: one condition, or a family of conditions?

The literature divides over what is being measured. On the one-condition position, cybersickness is a single response to a single mismatch: the sensory conflict account supplies the mechanism (Oman, 1990), Kennedy and colleagues (1993) built one instrument to quantify it, and Saredakis and colleagues (2020) pool scores across content as if comparable. On the family position, cybersickness changes character with the system and the task: Stanney, Kennedy and Drexler (1997) report a different profile from simulator sickness, Kim and colleagues (2018) had to replace the item set, and Saredakis and colleagues' own subgroup means shift with content.

Ang and Quarles (2023) supply the current tally, and it does not favour convergence. They screened 1,055 papers and found 88 studies comparing sickness between two or more conditions, grouped the remedies into nine categories, and report that the evidence behind them is uneven: some approaches rest on a single study, and several on samples of fewer than twenty participants.

What would settle it is a measurement-invariance study: the same participants through several systems and content types, one instrument plus one objective marker in every condition. It would test whether the symptom structure holds across conditions before any totals are compared. If it is stable, the score is comparable once normalised. If it shifts, pooled totals are an artefact of the pooling. No study in this unit's list runs that design.

The unit's claim has a strong counter-argument. If the standard instrument were truly incommensurable, the pooled totals of Saredakis and colleagues (2020) would carry no meaning, yet they order content types consistently and the moderators the review reports replicate across studies. Kim and colleagues (2018) treat their new questionnaire as a revision of the same construct rather than a different one. The reply is that a replicated direction is compatible with a scale whose units are not shared, so the dispute is about whether a normed score exists.

What a designer should do while the dispute is unresolved. Report the instrument and its version, the content, the exposure length and the dropout alongside every figure, and compare totals only within one study rather than across studies. Take a mid-exposure measure where one is available, since a post-hoc total hides the time course. Fit the headset to the individual, and treat a susceptibility difference, whether it comes from an inner ear, an injury or the balance of an older user, as an accessibility constraint rather than a covariate to be averaged away.

Where this is heading

The direction is towards instruments taken during exposure and validated across conditions. Ang and Quarles (2023) map what has been tried and how thin the evidence is. Howard and Van Zandt (2021) make susceptibility a fairness question. Weech, Kenny and Barnett-Cowan (2019) show why sickness must be controlled in any study that reports presence. The need is not a better mechanism but a better instrument.

Sources

• Oman, C. M. (1990). Motion sickness: a synthesis and evaluation of the sensory conflict theory. Canadian Journal of Physiology and Pharmacology. doi:10.1139/y90-044 Reviewed and evaluated the sensory conflict theory, in which sickness follows a mismatch between sensed and predicted motion. It reports that no conflict neuron or process had been physiologically identified, so the account remains a framework rather than a proven circuit.
• LaViola, J. J. (2000). A discussion of cybersickness in virtual environments. ACM SIGCHI Bulletin. doi:10.1145/333329.333344 Surveyed the causes of cybersickness and argued it is distinct from motion sickness because the user is often stationary while the display moves. It establishes the category whose measurement this unit questions.
• Kennedy, R. S., Lane, N. E., Berbaum, K. S. and Lilienthal, M. G. (1993). Simulator Sickness Questionnaire: An Enhanced Method for Quantifying Simulator Sickness. The International Journal of Aviation Psychology. doi:10.1207/s15327108ijap0303_3 Developed the Simulator Sickness Questionnaire from the Pensacola motion sickness questionnaire by factor analysis. It rates sixteen symptoms from zero to three within nausea, oculomotor and disorientation subscales, and became the field's default measure.
• Stanney, K. M., Kennedy, R. S. and Drexler, J. M. (1997). Cybersickness is Not Simulator Sickness. Proceedings of the Human Factors and Ergonomics Society Annual Meeting. doi:10.1177/107118139704100292 Factor-analysed simulator self-reports and eight virtual environment experiments, finding the symptom profile reversed: disorientation dominated cybersickness where oculomotor had dominated simulator sickness, and total severity was about three times greater.
• Dennison, M. S., Wisti, A. Z. and D'Zmura, M. (2016). Use of physiological signals to predict cybersickness. Displays. doi:10.1016/j.displa.2016.07.002 Measured physiological signals in seated participants who used a virtual environment first on a monitor and then through a head-mounted display. Sickness occurred only with the head-mounted display, and stomach activity, blinking behaviour and breathing predicted it.
• Kim, H. K., Park, J., Choi, Y. and Choe, M. (2018). Virtual reality sickness questionnaire (VRSQ): Motion sickness measurement index in a virtual reality environment. Applied Ergonomics. doi:10.1016/j.apergo.2017.12.016 Derived the Virtual Reality Sickness Questionnaire by factor analysis, retaining nine items in two components, oculomotor and disorientation, and dropping the nausea-weighted items. A different item set is evidence that the scores are not on one scale.
• Weech, S., Varghese, J. P. and Barnett-Cowan, M. (2018). Estimating the sensorimotor components of cybersickness. Journal of Neurophysiology. doi:10.1152/jn.00477.2018 Ran balance, vection and vestibular-sensitivity tests in one session and modelled them against measured cybersickness. The model predicted 37 per cent of the variability, 16 per cent of it from a balance-control component, with sway path length during vection strongest at 7.5 per cent.
• Weech, S., Kenny, S. and Barnett-Cowan, M. (2019). Presence and Cybersickness in Virtual Reality Are Negatively Related: A Review. Frontiers in Psychology. doi:10.3389/fpsyg.2019.00158 Reviewed studies measuring presence and cybersickness together and concluded the balance of evidence favours a negative relation driven by sensory integration. Because the two share causes, a presence result from a session that also induced sickness is not clean.
• Saredakis, D., Szpak, A., Birckhead, B., Keage, H. A. D., Rizzo, A. and Loetscher, T. (2020). Factors Associated With Virtual Reality Sickness in Head-Mounted Displays: A Systematic Review and Meta-Analysis. Frontiers in Human Neuroscience. doi:10.3389/fnhum.2020.00096 Systematically reviewed 55 head-mounted display studies covering 3,016 participants and pooled questionnaire scores. The pooled total was 28.00 and gaming content scored 34.26, with content, visual stimulation, locomotion and exposure length all moving the score.
• Chang, E., Kim, H. T. and Yoo, B. (2020). Virtual Reality Sickness: A Review of Causes and Measurements. International Journal of Human-Computer Interaction. doi:10.1080/10447318.2020.1778351 Surveyed causes and measurements across 77 experimental studies. Seventy six used a subjective measure, most often the simulator questionnaire, a total above about 33 marks high severity, and post-hoc administration cannot capture discomfort in real time.
• Stanney, K., Fidopiastis, C. and Foster, L. (2020). Virtual Reality Is Sexist: But It Does Not Have to Be. Frontiers in Robotics and AI. doi:10.3389/frobt.2020.00004 Ran two experiments with twenty minute roller coaster exposures to identify what drives gender differences in cybersickness. An interpupillary distance that could not be fitted to the headset was the primary driver, and correct fitting removed most of the difference.
• Howard, M. C. and Van Zandt, E. C. (2021). A meta-analysis of the virtual reality problem: Unequal effects of virtual reality sickness across individual differences. Virtual Reality. doi:10.1007/s10055-021-00524-3 Meta-analysed the relation between sickness and individual differences, finding significant associations with motion sickness susceptibility, gender, prior experience, neurological disorder and phobia, and framing unequal susceptibility as a fairness problem.
• Ang, S. and Quarles, J. (2023). Reduction of cybersickness in head mounted displays use: A systematic review and taxonomy of current strategies. Frontiers in Virtual Reality. doi:10.3389/frvir.2023.1027552 Screened 1,055 papers and selected 88 studies comparing sickness between two or more head-mounted display conditions. The strategies fall into nine categories, and the strength of evidence within them is uneven: within several categories only a minority of studies found a significant reduction, some approaches rest on a single study, and several studies used samples of fewer than twenty participants.
• Fernandez-Canas, M., Ortiz-Gutierrez, R. M., Martin-Casas, P., Estrada-Barranco, C., Marcos-Anton, S. and Blazquez-Fernandez, A. (2025). Cybersickness Evaluation in Immersive Virtual Environments: A Systematic Review with Implications for Neurological Rehabilitation. Journal of Clinical Medicine. doi:10.3390/jcm15010046 Appraised nine cybersickness instruments against COSMIN criteria and graded them by GRADE: the Cybersickness in VR Questionnaire and the Virtual Reality Neuroscience Questionnaire were recommended, the simulator questionnaire and the VRSQ only acceptable, and two measures not recommended without complementary data.