The evidence
The quantitative case, with sources
MoodAir brings together well-studied signals. This page shows the measured dose-response of CO₂ on cognition and sleep, the coefficients linking CO₂ to cardiac autonomic activity, and the inference chain the product relies on, with every figure traced to a study and its limits stated.
1 · Dose-response of indoor CO₂
| CO₂ level | Measured effect | Source | Type |
|---|---|---|---|
| ~945 ppm | Cognitive-function scores about 15% below a ~550 ppm baseline; each +400 ppm associated with roughly a 21% decline across domains. | Allen et al. 2016 | controlled |
| ~1000 ppm | Decision-making performance significantly reduced versus 600 ppm. | Satish et al. 2012 | controlled |
| 1000 to 1500 ppm | Complex cognitive-task performance declines significantly (pooled across studies). | Fan & Cao 2023, meta-analysis | meta |
| ~1400 ppm | Cognitive scores about 50% below the ~550 ppm baseline. | Allen et al. 2016 | controlled |
| 1000 ppm, sleep | Sleep efficiency reduced and time awake increased versus 750 ppm. | Xu et al. 2023 | field-lab |
| 1300 ppm, sleep | Deep-sleep duration decreased, morning salivary cortisol increased, and the HRV LF/HF balance shifted toward sympathetic activity. | Xu et al. 2023 | field-lab |
The WHO and ASHRAE both recommend keeping indoor CO₂ at or below 1000 ppm. Thresholds here flag air state; they do not diagnose anyone.
2 · CO₂ and cardiac autonomic activity
Heart rate variability reflects autonomic balance. Most real-room studies measure heart rate, which is inversely related to HRV, so a CO₂-driven rise in heart rate implies a fall in HRV. The measured effects:
| Relationship | Quantified effect | Source |
|---|---|---|
| CO₂ → sleeping heart rate | +2.3 bpm / 1000 ppm | MacNaughton et al. 2016 |
| CO₂ vs sleeping heart rate, pooled | r = 0.28, P<0.01 | Oota et al. 2024 (positive in all 9 subjects) |
| CO₂ regression coefficient | β = 0.0044 bpm/ppm, std β = 0.19, power 0.82 | Oota et al. 2024 |
| Mean heart rate vs HRV (SDNN) | inverse, r about 0.91 | Sacha & Pluta 2008; Oota 2024 |
| Acute 35% CO₂ inhalation → HRV | significant rise in high / very-high frequency | Psychiatry Research 2020 |
3 · The inference chain
CO₂ ↑ ⇒ heart rate ↑ (β ≈ +2.3 bpm per 1000 ppm; r ≈ 0.28)heart rate and HRV move in opposite directions (r about 0.91)
∴
CO₂ ↑ ⇒ HRV ↓ the autonomic stress direction.MoodAir estimates this relationship per person, controlling for sleep stage, temperature and noise.
Honest limits: most bedroom evidence measures heart rate as an HRV proxy, not RMSSD directly. Samples are small (Oota n=9), observational, and confounded by sleep stage and temperature, which is exactly why the model controls for them. Acute inhalation studies use high-concentration CO₂ and are not the same as ambient room air; they are shown only as directional support.
4 · What existing data does, and does not, give us
Several studies have measured CO₂ and a cardiac signal together, and they matter. They are small, offline, and after the fact. What is missing is a live, per-person, continuous read, which is what MoodAir builds.
| Source | Indoor CO₂ | Individual HRV | Live & per-person |
|---|---|---|---|
| UCI Occupancy (Candanedo 2016) | yes | no | no |
| DigitalExposome (Johnson 2025) | no | yes | no |
| Bedroom field study (Oota 2024) | yes | HR proxy | per person, n=9, offline |
| MoodAir | yes | yes | live, per person |
5 · Expert engagement
Prof. Yuming Guo
Encouraged the focus on individual-level exposure and response.
Dr. Sam McKay, Turner Institute
Advised controlling confounds and pairing HRV with a validated self-report; recommended an exploratory study at this stage.
Dr. Prerna Varma
Pointed to the bedroom as the cleanest first setting, with CO₂ ideally below 800 to 1000 ppm.
Measured, not proven
Advice shaped the design. It is not endorsement. MoodAir measures. It does not diagnose.