Evidence · Stressonika and CleverPoint 4
What we measured, and how well it repeats.
Stressonika is our 25-minute VR test recorded with the CleverPoint 4 module in a Meta Quest headset: paced breathing, attention and motor tasks, and rest breaks, read from the ECG and the facial channels. This page shows how its measures hold up.
We call a measure a trait only if the same person gets a similar value on another day. We checked the candidate measures on people who came back for a second session; the ones that repeat are reported as traits, and measures that repeat less are shown as observations of the day. Below are the data, the method and the numbers.
The data
3,488 Stressonika sessions from 3,176 people, recorded with CleverPoint 4 from 2021 to 2026 at 6 sites, all in the same protocol. 3,327 of them include every section the heart answers need.
About half of the people are teenagers aged 12–17; 1,557 are adults aged 18–65 (938 women, 619 men). The adult findings below are computed on adults.
People by age and sex, first session
| Age | Women | Men | All |
|---|---|---|---|
| 7–11 | 3 | 9 | 12 |
| 12–17 | 767 | 830 | 1,597 |
| 18–24 | 124 | 76 | 200 |
| 25–34 | 289 | 215 | 504 |
| 35–44 | 373 | 221 | 594 |
| 45–54 | 139 | 88 | 227 |
| 55–65 | 13 | 19 | 32 |
| over 65 | 1 | 4 | 5 |
3,171 people with a recorded date of birth. Sessions per year: 2021: 74 · 2022: 122 · 2023: 225 · 2024: 676 · 2025: 1,778 · 2026: 613.
How we tested repeatability
201 people took the test more than once. We compared their two sessions recorded on different days, typically about half a year apart (median 202 days); 171 of the 175 pairs are adults.
Agreement is measured with the intraclass correlation, ICC(2,1): 1 means the person gets the same value every time, 0 means the second session says nothing about the first. A measure counts as a trait when its ICC reaches 0.6 and the lower end of its 95 % interval stays above 0.4.
The last column of the table is the smallest real change: a difference between two sessions larger than this is a change in the person, not ordinary variation.
Months, not days
A retest a few days later is easy to pass: the person is in the same week of their life. Our pairs are months apart, so a measure that repeats here describes the person, not the week.
Within a single session the recovery measure repeats at 0.72–0.96.
ICC(2,1): two-way random effects, absolute agreement, single measurement; 95 % intervals from 2,000 bootstrap resamples of the pairs. Smallest real change: MDC95 = 1.96 × √2 × SEM, SEM = SD × √(1 − ICC).
Measures that repeat between days
| Measure | ICC [95 % interval] | Adults only | Pairs | Smallest real change |
|---|---|---|---|---|
| Heart | ||||
| Switching into load: rise of heart rate over the breathing level | 0.64 [0.54; 0.72] | 0.63 [0.52; 0.72] | 161 | on its relative scale* |
| Switching off: return of heart rate in the rest breaks | 0.70 [0.55; 0.81] | 0.70 [0.53; 0.80] | 162 | 10.7 bpm |
| Heart-rate range during paced breathing | 0.73 [0.65; 0.80] | 0.73 [0.64; 0.80] | 162 | 11.2 bpm |
| Reserve age, difference from calendar age | 0.65 [0.54; 0.73] | — | 129 | — |
| Cognitive and motor | ||||
| Schulte tables: search speed | 0.80 [0.74; 0.85] | 0.81 [0.73; 0.86] | 172 | 0.17 numbers/s |
| Tapping: tempo | 0.70 [0.59; 0.80] | 0.70 [0.58; 0.79] | 172 | 0.86 taps/s |
| Banana test: productivity | 0.68 [0.58; 0.76] | 0.68 [0.59; 0.76] | 172 | 18 percentage points |
| Stroop: holding the target while tracking | 0.66 [0.51; 0.77] | 0.67 [0.52; 0.79] | 172 | 19 % of the time |
| Head channels, shown in the report as observations | ||||
| Alpha peak with eyes closed | 0.57 [0.40; 0.70] | — | 104 | — |
| Blink rate in the first Stroop task | 0.61 [0.50; 0.72] | — | 131 | — |
* Switching into load is measured on a scale relative to the person's own breathing level; the report applies the same 95 % rule on that scale.
Trait or state of the day
Two day answers of the report, rest in the breaks and the cost of the day, repeat at 0.53–0.54, below the trait rule: partly stable, partly the day. The report names them as today's state rather than as a property of the person.
The heart answers to the load
The report's heart-state scale rates each stretch of a session from calm to high load, using heart rate and its variability. On 2,886 sessions it rates as high load 67.6 % of the windows under cognitive load, 31.2 % in the rest breaks and 9.5 % during paced breathing: highest under load, lowest when breathing is paced.
Age: what moves and what doesn't
In 1,545 adults aged 18–65 the breathing range falls steeply with age: a median of 25.6 bpm at 18–24 and 8.1 bpm after 55 (ρ −0.51, R² 0.27), in line with the known decline of heart rate variability with age. Switching into load and switching off hardly move with age (R² 0.01 and 0.03), so they describe the person rather than the birth year.
Cognitive speed and productivity decline gently with age (ρ −0.19 to −0.23). The reserve age, estimated from heart measures alone, tracks calendar age (r 0.55; mean absolute error 6.2 years against 7.6 for a guess of the average age, 1,501 adults), with a correction for regression to the mean. The report shows its difference from calendar age; that difference is what repeats between days. It is not a health age.
What we do not claim
No diagnoses. No risk scores. No promise that any measure predicts health or lifespan. Stressonika describes how one body met one standard load and how that compares with the database and with the person's last session.
The methods note with the full tables and the way the norms are built is available to customers and reviewers under a non-disclosure agreement.
Literature on the method
- Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86(2):420–428.
- Koo TK, Li MY. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J Chiropr Med. 2016;15(2):155–163.
- Laborde S, Mosley E, Thayer JF. Heart rate variability and cardiac vagal tone in psychophysiological research: recommendations for experiment planning, data analysis, and data reporting. Front Psychol. 2017;8:213.