Is HRV measured by a phone camera accurate?

For RMSSD measured at rest while still, camera photoplethysmography correlates above 0.9 with ECG, with a mean absolute percentage error of roughly 11–17%. This is sufficient for trend tracking. Somatic does not use the camera for frequency-domain metrics like LF/HF where accuracy is poor, and every scan passes a quality gate that rejects readings with motion artifact or low signal quality.

RMSSD (root mean square of successive differences) is a heart rate variability metric measured in milliseconds that reflects beat-to-beat changes driven primarily by parasympathetic activity. It is Somatic's primary HRV metric because it is the most reliable HRV measure for short, resting scans on consumer sensors. Somatic shows RMSSD relative to your personal rolling baseline.

Why doesn't Somatic show LF/HF or other frequency-domain HRV metrics?

Camera photoplethysmography is not reliable for frequency-domain HRV metrics like LF and HF, especially over short scans. Somatic surfaces RMSSD (primary) and SDNN (secondary) only — the metrics camera PPG can measure validly at rest.

Science — Somatic

Q: What is a good RMSSD value?

There is no universal good RMSSD value — healthy adults vary widely by age, fitness, and individual physiology. Somatic deliberately does not compare you to a population norm; every reading is shown against your own rolling 7-, 14-, and 30-day baseline range. What matters is whether your number is trending up relative to your own range.

Q: How is Somatic different from HeartMath Inner Balance?

HeartMath Inner Balance requires a hardware sensor and frames training around 'coherence.' Somatic uses your phone camera with no hardware, surfaces validated PSS-10 and GAD-7 stress and anxiety scores as the outcome metric, and shows HRV as a trend against your own personal baseline rather than as a coherence score. Both include real-time HRV biofeedback as a core practice.

Camera PPG

Good for time-domain HRV at rest. Not for everything HRV.

Smartphone-camera photoplethysmography (PPG) works by reading tiny color changes in your fingertip as blood pulses through it. Decades of validation studies — against ECG, against chest straps, against research-grade pulse oximeters — have established a clear pattern:

For time-domain HRV metrics (RMSSD, SDNN) measured at rest, still, under standardized conditions, smartphone camera PPG agrees closely with ECG. Correlations are frequently above 0.9. The mean absolute percentage error for RMSSD lands in the 11–17% range across studies.
For frequency-domain metrics (LF, HF, LF/HF ratio), camera PPG is unreliable. The math behind those metrics requires a clean continuous signal that camera-PPG short scans don't produce well.
For anything during motion, the signal degrades sharply. Walking, gesturing, talking on the phone while scanning — none of that produces trustworthy HRV.

What this means for what we show you

RMSSD is our primary metric. It's the most robust to breathing-rate variation and the most defensible from camera PPG at rest.
SDNN is secondary — useful as a broader autonomic snapshot, shown next to RMSSD.
LF, HF, LF/HF — never shown. Some competitor apps surface these as "stress" or "balance" gauges. From a camera scan, they aren't valid signals. We don't display them.
All scans run at rest, seated, still. No on-the-go scanning. No "walking HRV." Those aren't things a phone camera can do honestly.

RMSSD primary · SDNN secondary · never LF/HF

Why these metrics, and not the others.

Why RMSSD

RMSSD reflects beat-to-beat variation that's heavily influenced by the parasympathetic branch of your autonomic nervous system. It's the metric that holds up best when compared against ECG from a smartphone camera at rest. It's also the metric most responsive to slow-paced breathing — the strongest-evidence intervention in our library. That alignment matters.

Why SDNN as a secondary

SDNN is a broader picture of total HRV across the scan window. We show it next to RMSSD because some users want the larger view, but we treat RMSSD as the lead.

Why no LF/HF

LF/HF ratio is marketed in some apps as a "sympathetic vs. parasympathetic balance" score. That framing was always more popular than the science supports, and it's been further questioned in recent years. From a 90-second camera scan, the frequency-domain math is too noisy to be useful. Showing it would mean confidently displaying a number that isn't honest. We don't.

Evidence tiers

Every protocol in the library is tagged by how strong the evidence is.

We don't put every practice on equal footing. The library is ordered by strength of evidence, and each protocol shows its tier.

Tier 1 — Strongest evidence (meta-analytic support) Slow-paced breathing at ~6 breaths per minute (resonance frequency breathing). Multiple meta-analyses support acute increases in RMSSD/SDNN, reductions in state anxiety, and lowered blood pressure with consistent practice.
Tier 1 — Strongest evidence HRV biofeedback (HRV-B). Real-time pacer paired with a live HRV trace at the user's resonance frequency. One of the best-supported modalities in the entire nervous-system regulation literature.
Tier 2 — Good evidence Physiological sigh / extended-exhale patterns. Supported by parasympathetic activation literature; effective for acute calming.
Tier 2 — Good evidence 5-4-3-2-1 sensory grounding. Reasonable evidence for acute anxiety and dissociation moments.
Tier 2 — Good evidence Gentle somatic movement, shaking, body tapping. Effective in practice; framed as regulation skills with observed effects, not as mechanism claims.
Tier 3 — Offered, with honest caveats Humming, gargling, cold exposure ("vagus exercises"). Popular and often helpful. We do not brand them as clinical "vagus nerve stimulation" — that term properly belongs to electrical taVNS hardware in a clinical setting. We say what they actually are: practices that may support regulation, with weaker and more variable evidence than Tier 1.

You'll see the tier on every protocol card. Pick what you want; you'll always know what evidence you're picking on.

PSS-10 and GAD-7

Two short, well-established self-report measures. Not scores we invented.

The category is full of internal "stress scores" — proprietary numbers an app made up, calculated from inputs that may or may not predict anything outside the app. We don't use those.

We use:

PSS-10 — Perceived Stress Scale (10-item). Cohen et al., 1983. One of the most widely used self-report measures of perceived stress in research. Ten items, takes about two minutes.
GAD-7 — Generalized Anxiety Disorder 7-item. Spitzer et al., 2006. Standard primary-care screen for generalized anxiety; widely cited and used.

We re-administer them every 7–14 days. Frequent enough to see a trend; spaced enough to avoid fatigue and prevent the score from being gamed by repetition. Your line plots over time. That line is the outcome proof.

To be explicit: PSS-10 and GAD-7 are self-report instruments. They measure your perception of stress and anxiety. They're not clinical diagnoses, and a score change in the app is not a medical claim. They're directional, validated, and far more honest than an app-internal score.

What we don't claim

The shortcuts we won't take.

We do not claim Somatic itself is clinically proven. There are no Somatic-specific clinical trials yet. We lean on category-level evidence for the interventions, and on validated instruments for the outcomes.
We do not use "vagus nerve stimulation" as a clinical descriptor for humming or cold exposure. That language belongs to electrical taVNS hardware studied in clinical settings.
We do not use the "6 nervous system states" map sometimes drawn from polyvagal theory. The theory's specific predictions remain under scientific challenge, and we don't want our product hanging on a frame that may not hold.
We do not call ourselves a medical device. Somatic isn't a diagnostic tool, a treatment, or a substitute for clinical care.
We do not call our app-internal numbers "coherence." That term is owned by a specific competitor's framework and we don't claim its outcomes.

Honest limitations

Where the measurement is weakest.

Short scans have wider error bars. A 20-second pre/post window is a directional signal, not a precise reading. A 12% delta might genuinely reflect a shift, or it might reflect normal scan-to-scan variation. We report it because the trend across many sessions becomes meaningful even when each single delta is noisy.
Resting standardization matters. If you scan walking from a meeting, after coffee, with a cold finger, or while distracted, the reading is less interpretable. The morning standardized scan exists exactly to control for this.
Camera PPG is not a chest strap or ECG. For research-grade precision, neither is your watch. For tracking your own trend against your own baseline, the science supports the approach we're using.
HRV varies for many reasons. Sleep, caffeine, alcohol, illness, training load, menstrual cycle, breathing pattern. We never tell you your number "means" something singular. The point is the trend over time and the practice deltas — not interpretation of any one reading.

We'd rather under-promise on the precision and let the loop do the proving than wave around a number with more confidence than the science supports.

Common questions

What people ask about the science.

For RMSSD measured at rest while still, camera photoplethysmography (PPG) correlates above 0.9 with ECG, with a mean absolute percentage error of roughly 11–17%. This is sufficient for trend tracking, which is how Somatic uses it. Somatic does not use the camera for frequency-domain metrics like LF/HF where accuracy is poor, and every scan passes a quality gate that rejects readings with motion artifact or low signal quality.

RMSSD (root mean square of successive differences) is a heart rate variability metric measured in milliseconds that reflects beat-to-beat changes driven primarily by parasympathetic (vagal) activity. It is Somatic's primary HRV metric because it is the most reliable HRV measure for short, resting scans on consumer sensors. Somatic shows RMSSD relative to your personal rolling baseline, never as a universal "good" or "bad" number.

There is no universal "good" RMSSD value — healthy adults vary widely by age, fitness, and individual physiology. Somatic deliberately does not compare you to a population norm; every reading is shown against your own rolling 7-, 14-, and 30-day baseline range. What matters is whether your number is trending up relative to your own range.

Camera photoplethysmography is not reliable for frequency-domain HRV metrics like LF and HF, especially over the short scans Somatic uses. Showing those numbers would be measurement dishonesty. Somatic surfaces RMSSD (primary) and SDNN (secondary) only — the metrics camera PPG can measure validly at rest.

Every protocol in Somatic is tagged Tier 1, Tier 2, or Tier 3 based on the strength of supporting research. Tier 1 covers practices with meta-analytic support — slow-paced breathing near 6 breaths per minute and HRV biofeedback. Tier 2 covers practices with good supporting evidence such as extended-exhale and grounding. Tier 3 covers preliminary practices like humming, which we caveat clearly and do not brand as clinical vagus-nerve stimulation.

PSS-10 (Perceived Stress Scale, 10 items) and GAD-7 (Generalized Anxiety Disorder, 7 items) are validated, widely-used self-report questionnaires for stress and anxiety. Somatic re-administers them every 7–14 days so you can see whether your stress and anxiety scores are actually moving over weeks — not just whether individual sessions felt nice.

HeartMath Inner Balance requires a $159 hardware sensor and frames training around "coherence." Somatic uses your phone camera with no hardware, surfaces validated PSS-10 and GAD-7 stress and anxiety scores as the outcome metric, and shows HRV as a trend against your own personal baseline rather than as a coherence score. Both include real-time HRV biofeedback as a core practice.

Calm and Headspace are meditation content libraries; they do not measure your physiology and do not show whether a given session affected your nervous system. Somatic measures HRV before and after every short practice with the iPhone camera and re-administers PSS-10 and GAD-7 every 7–14 days to prove whether stress is moving. Somatic is a measurement-and-practice loop with a small evidence-anchored library, not a content catalog.

What's measurable, what isn't, and how we tell the difference.