HRV
TL;DR
HRV (heart rate variability) is a wearable-derived, context-sensitive autonomic state proxy. In Vitals it is never used alone — always paired with resting HR and interpreted against the user’s own personal baseline. lnRMSSD is the primary backbone metric. LF/HF is not a valid autonomic balance construct. Single-day readings are nearly uninterpretable without 7–14 days of baseline.
Why it matters for Vitals
HRV is one of the most over-claimed biometrics in consumer wearables. Vitals needs it to be accurate because:
- HRV is the primary autonomic-correlate signal in the Vitals multi-metric panel for sleep, recovery, and readiness inference
- HRV + resting HR as a pair is the corroborating signal cluster for training load, overreaching, and illness-pattern detection
- Apple Watch lnRMSSD trend is the only HRV output that is detectably useful in Vitals product context — all other HRV metrics fail validity or availability thresholds
- The 8 most common HRV myths (vagal tone, LF/HF balance, higher-is-better, breathwork proof, etc.) are actively dangerous to Vitals product integrity if not corrected
- No-call conditions are load-bearing — arrhythmia, medication effects, and insufficient baseline must suppress coaching output, not generate it
Key facts
- Primary metric: lnRMSSD (natural log of RMSSD). Apple Watch HRV is in this domain.
- Always pair with resting HR — neither is interpretable alone
- Personal baseline required — 7 days minimum, 14 preferred, for any coaching framing
- Within-person trend only — population comparisons are invalid
- Measurement context determines interpretability — morning supine is gold standard; session-window HRV is not interpretable
- HRV does not measure vagal tone directly — it is a vagal-adjacent cardiac proxy
- LF/HF is not a valid autonomic balance metric — do not use
- Apple Watch cannot detect vagal tone, sympathetic activity, autonomic balance, or arrhythmia (it can flag irregular rhythm for clinical referral)
- Acute session-window HRV ≠ durable adaptation — minutes-to-hours shifts reflect state; weeks-to-months sustained trends may reflect adaptation but require stable conditions
- Arrhythmia or ectopic beat flag → immediate clinical referral — HRV metrics assume sinus rhythm
Metric hierarchy
| Metric | Signal domain | Apple Watch available | Vitals coaching validity |
|---|---|---|---|
| lnRMSSD | Vagal-adjacent cardiac modulation | Yes — primary output | Primary backbone — within-person trend over 7–14 days |
| RMSSD (raw) | Same as lnRMSSD | Yes | Use ln-transformed only |
| Resting HR | Autonomic output — heart rate | Yes (continuous/spot) | Always pair with HRV |
| pNN50 | Vagally-mediated HF beat-to-beat changes | Yes (derived) | Secondary; more sensitive to ectopy artifact than lnRMSSD |
| SDNN (short-window) | Total HRV | Yes (short-window) | Unreliable; 24h SDNN is clinical only |
| HF power | Respiratory-linked RSA; vagal efferent bursts | Not in Apple Watch output | Not usable from Apple Watch |
| LF power | Mayer waves; baroreflex oscillation | Not in Apple Watch output | Not usable; not a pure sympathetic marker |
| LF/HF ratio | Frequency-domain balance | Not in Apple Watch output | PRODUCT-UNSAFE — do not use |
Signal compartment taxonomy
Every HRV claim must be taggable to one of these compartments:
| Tag | What it captures | HRV validity for Vitals |
|---|---|---|
[NEURAL-VAGAL] | Vagal efferent burst at sinus node | Moderate — most specific available from wearables |
[RESPIRATORY-MECHANICAL] | RSA amplification via breathing pattern | High confound — assume uncontrolled in free-living |
[BAROREFLEX] | Blood pressure oscillation at ~0.1 Hz (Mayer waves) | Low specificity for wearables |
[AUTONOMIC OUTPUT] | Composite wearable signal — all above + state/load | Valid as within-person trend only |
[INDIRECT/SECONDARY] | HRV shifts driven by sleep, illness, training, alcohol, medications | Interpretable only with corroboration |
[WEARABLE PROXY] | Apple Watch PPG-RR algorithm output | Must never be equated with laboratory HRV |
[PATHOLOGY BOUNDARY] | Conditions requiring clinical referral, not coaching | Zero coaching tolerance |
Two interpretation modes that must never be conflated
- Within-person acute state tracking: How today’s HRV compares to this person’s own recent baseline. Safest and best-supported consumer use case for Apple Watch HRV.
- Between-person generalization: How one person’s HRV compares to population norms. Between-person variability is large enough (~7x range in healthy adults) that population comparisons are not appropriate for individual coaching.
Two recording contexts that must never be conflated
- Standardized morning HRV: Pre-activity, supine or seated, first 30 seconds–5 minutes post-wake. Only context with reasonable reproducibility for longitudinal tracking.
- Session-window HRV: Random timestamps during the day, variable activity, breathing, stress. Too noisy for any coaching claim.
Measurement windows
| Window | Valid for coaching? | Notes |
|---|---|---|
| Morning supine (gold standard) | Yes | First 30s–5min post-wake, before rising, before caffeine |
| Morning seated | Yes — acceptable | Consistent posture vs baseline required |
| Overnight / sleep HRV | Yes — useful trend context | Reflects nocturnal autonomic state |
| Post-exercise <60 min | No | Post-HRR dynamics dominate; exclude from trend |
| Midday / afternoon / evening random | No standalone claim | Behavioral logging only |
| Post-breathwork / post-intervention | No causal attribution | RSA confound dominates; state shift possible, adaptation claim unsupported |
Apple Watch HRV — what it can and cannot detect
What Apple Watch actually measures: RR intervals via wrist-based PPG (not ECG), processed to derive lnRMSSD. Single-lead optical sensor, ~1 Hz sampling. Not equivalent to clinical ECG-HRV.
What is likely detectable:
- Morning lnRMSSD within-person trend (7–14d) with consistent conditions
- Single-day morning HRV vs own baseline (large deviation >20%) with confirmation
- Resting HR trend (7d)
- Sustained illness-driven suppression (3+ days, corroborated with resting HR and symptoms)
What is not reliably detectable:
- Session-window HRV for stress/readiness
- LF/HF as autonomic balance score (not available from Apple Watch; invalid even if available)
- Between-person HRV comparison vs population norms
- Durable vagal adaptation from training (weeks-months; cannot isolate from confounders)
- Post-breathwork HRV increase as vagal activation proof (RSA mechanical artifact)
- HRV illness prediction before symptoms
- Arrhythmia / ectopy detection (hard no-call — immediate referral)
Myths and overmarketed claims
Myth 1: “HRV measures vagal tone directly”
lnRMSSD reflects vagal efferent bursts at the sinus node, but is a downstream proxy, not a direct assay of vagal nerve conduction. Respiratory mechanics, baroreflex state, and sympathetic overflow also shape lnRMSSD. → Use: “HRV is a vagal-adjacent cardiac proxy.”
Myth 2: “Apple Watch HRV measures vagal tone / autonomic balance / parasympathetic tone”
Apple Watch HRV is an optical-sensor RR-interval estimate. It cannot measure “balance” because LF/HF is not a valid construct and Apple Watch doesn’t expose LF/HF anyway. → Use: “Apple Watch HRV estimates lnRMSSD — a proxy, not a direct autonomic measurement.”
Myth 3: “LF/HF ratio shows sympathetic vs parasympathetic balance”
LF power (~0.1 Hz) is generated primarily by baroreflex-mediated blood pressure oscillations — NOT a direct measure of sympathetic tone. HF power (~0.25 Hz) reflects respiratory-linked RSA. The ratio is the most overclaimed metric in consumer HRV. Apple Watch does not expose it. → Use: “LF/HF is a frequency-domain ratio that shifts with multiple inputs. Do not interpret it as autonomic balance.”
Myth 4: “Higher HRV always means better recovery / health / fitness”
Optimal HRV is person-specific and context-dependent. Extremely sudden HRV elevation can indicate arrhythmia or ectopic beats. Low HRV is non-specific. → Use: “HRV values outside your personal range — higher or lower — may be worth noting and contextualizing.”
Myth 5: “Low HRV means you are stressed / overtrained / not recovered”
HRV suppression is non-specific. It can reflect training load, acute illness, alcohol, circadian shift, respiratory mechanics, emotional arousal, posture, caffeine, or measurement artifact. Interpreting illness-driven suppression as “overtraining” is dangerous. → Use: “Lower HRV than your baseline may be worth monitoring. Multiple factors can affect HRV.”
Myth 6: “Post-breathwork HRV increase proves vagal activation”
Slow breathing mechanically amplifies HRV via RSA — a mechanical artifact, not a change in vagal neural firing. The effect dissipates within minutes to hours and does not constitute evidence of durable autonomic remodeling. → Use: “Post-session HRV was higher — possible state shift, not necessarily vagal activation in free-living conditions.”
Myth 7: “Apple Watch HRV is equivalent to clinical ECG-based HRV”
Apple Watch uses single-lead wrist PPG at ~1 Hz. Clinical ECG uses multiple leads at 1000 Hz. PPG cannot reliably resolve ectopic beats or fine waveform morphology. Apple Watch HRV is qualitatively different from laboratory ECG-HRV. → Use: “Apple Watch HRV provides a useful personal trend signal. It is not a clinical autonomic assay.”
Myth 8: “Session-window HRV across the workday shows your stress or readiness”
Random-timestamp HRV is dominated by posture, activity, respiratory pattern, caffeine, conversation, temperature, and emotional state. Noise exceeds signal. → Use: “Only morning HRV (first few minutes post-wake, before activity) is interpretable for trend tracking.”
Safe claims registry
Directly permissible
- “Your morning HRV today is [value], which is [above/below/within] your recent personal range.” — HIGH confidence; measurement conditions controlled
- “Your HRV trend over the past [7/14/28] days shows [direction] relative to the prior period. Trends are more reliable than single readings.” — ≥7 days consistent measurement
- “Many factors — including sleep quality, recent exercise, illness, alcohol, caffeine, and stress — can affect HRV on any given day.” — always true; user education
- “Your HRV and resting HR together suggest a [favorable/elevated] physiological state compared to your recent average.” — HIGH confidence both signals; directionally consistent
Use with required caveat
- “Lower HRV may reflect reduced recovery or elevated autonomic stress.” — required: “This is an association, not a diagnosis.”
- “Higher HRV may reflect improved parasympathetic tone or favorable recovery.” — required: “Higher HRV is not universally better.”
- “HRV can be used to monitor training adaptation over time.” — required: “Do not adjust training solely based on a single HRV reading.”
- “Deep breathing may transiently increase HRV.” — required: “This is a short-term response, not evidence of long-term adaptation.”
Do not use (prohibited)
- “HRV measures vagal tone / vagus nerve function directly”
- “Apple Watch measures vagal tone / autonomic balance”
- “Low HRV means you are stressed / overtrained / not recovered”
- “Higher HRV always means better health / better recovery”
- “LF/HF ratio shows sympathetic vs parasympathetic balance”
- “This HRV reading proves your protocol / supplement worked”
- “HRV predicts illness before symptoms appear”
- “Your HRV score = [specific number] / readiness = [specific score]”
- “Your HRV shows autonomic dysfunction or cardiovascular disease”
- “Regular HRV monitoring will improve your health outcomes”
- “Beta-blocker users with high HRV have excellent vagal tone”
- “Session-window HRV shows your stress level”
Confidence tiers
| Tier | Label | Conditions | Permissible interpretation |
|---|---|---|---|
| 4 | HIGH | Baseline ≥14 days; same time/posture/device; no illness/medication; reading consistent with 7-day trend | Single-reading interpretation permitted; directional trend framing allowed |
| 3 | MEDIUM | Baseline 7–13 days OR moderately consistent conditions OR partial context log | Trend-based only; single-reading requires explicit uncertainty qualifier |
| 2 | LOW | Baseline <7 days OR high day-to-day variance OR context log unavailable | Trend-only framing; label “preliminary — insufficient baseline for individual interpretation” |
| 1 | NO CALL | Arrhythmia; post-exercise <60 min; fever/illness; medication; <2 min data; extreme respiratory rate | No interpretation — suppress from user as meaningful |
No-call conditions
Hard no-call (suppress all coaching output)
- Arrhythmia or frequent ectopy — RR algorithms assume sinus rhythm; ectopic beats corrupt RMSSD unpredictably
- Active fever or acute infection — cytokine-CNS-autonomic pathway suppresses HRV; no training attribution
- Beta-blocker or autonomic-active medication — directly alters HRV via drug mechanism; suppress from recovery scoring
- Post-exercise measurement <60 min — post-HRR dynamics dominate; exclude from trend
- Measurement duration <2 min clean RR — RMSSD requires 60–120s minimum for stable estimate
- Insufficient baseline (<7 days) — cannot distinguish personal variability from noise
Soft no-call (display with mandatory disclaimer)
- Medication confounder suspected → “HRV may be affected by [medication class]. Consult your physician.”
- Alcohol within prior 24h → “This reading may be affected by recent alcohol consumption.”
- Postural inconsistency from baseline → “Posture differs from your baseline — interpret with caution.”
- Recent travel across >3 time zones → “Recent travel may affect HRV — re-baseline after 3–5 days.”
User pattern classification
| Pattern | HRV signature | Key triage action |
|---|---|---|
| Baseline-builder | <7 days data, high variance | No coaching output; baseline-building message only |
| Stable-normal | Within ±1 SD of rolling mean; low variance | No action; informational display only |
| Acute suppression | Single-day >20% drop; not confirmed | Flag; require 2-of-3 confirmation before action |
| Training-load pattern | HRV down + resting HR up + sleep disruption 3–7 days | Context-conditional coaching; prioritize sleep |
| Accumulating load / early overreaching | HRV declining + resting HR rising sustained 10–14 days | Monitor closely; proactive recovery nudge; not a diagnosis |
| Illness-pattern | HRV suppressed 30%+ over 3+ days; resting HR elevated 5+ bpm; symptoms | HARD STOP: prioritize rest; no training prescription |
| Favorable adaptation | HRV trending up + resting HR trending down over 14+ days; sleep stable | Positive signal; continue current approach |
| Arrhythmia-flagged | Irregular rhythm notification or known arrhythmia | HARD NO-CALL; immediate clinical referral |
Resting HR + HRV interpretation pairs
| HRV direction | Resting HR direction | Pattern | Interpretation |
|---|---|---|---|
| Down | Up | Discordant elevated concern | Possible elevated load or sympathetic dominance — flag |
| Up | Down | Corroborated favorable | Possible favorable autonomic state — corroborate before positive framing |
| Both declining | — | Both declining | Possible overreaching or illness — no-call gate |
| Unchanged | HRV declining | Isolated HRV decline | Possible acute stressor or training load — full context required |
| Unchanged | HRV rising | Isolated HRV rise | Possible favorable state or artifact — corroborate before framing positive |
Evidence boundaries
[EVIDENCE-BASED]
- Within-person lnRMSSD trend direction over 7–14 days is detectable with Apple Watch in standardized morning conditions
- Resting HR is more stable than HRV and should always be interpreted alongside HRV
- Respiratory pattern materially affects short-window HRV interpretation
- Apple Watch HRV is a useful proxy for short-window HRV trend tracking in healthy subjects under controlled conditions
- LF/HF is not a valid measure of sympathetic/parasympathetic balance
- Between-person HRV variance is enormous (~7x range in healthy adults)
- Session-window HRV (random timestamps) is not interpretable for readiness or recovery inference
[EXTRAPOLATED]
- Durable (weeks-months) HRV trend changes may reflect vagal or training adaptation but cannot be disentangled without controlled conditions
- HRV-guided training decisions may be useful in tightly monitored athletic contexts; generalization to consumer populations is limited
[HYPOTHESIS ONLY — Product-unsafe]
- LF/HF as autonomic balance score
- Post-breathwork HRV increase as evidence of vagal activation
- HRV predicting illness before symptoms appear
- “Vagal tone” as a unitary, HRV-measurable construct in consumer wearables
- HRV composite scalar scores as calibrated metrics
- Single-session HRV change as proof of protocol efficacy
[CONTESTED]
- Degree to which short-window LF power reflects sympathetic vs combined autonomic activity
- Clinical utility of HRV-guided training outside tightly monitored athletic settings
Related notes
Mechanisms
- Respiratory Sinus Arrhythmia — primary mechanical confound on HRV; RSA amplification
- Mayer Waves — LF power source; baroreflex oscillation at ~0.1 Hz
Biometrics
- Cardiovascular signatures — cardiovascular effects of substances; broader HR/BP context
- REM suppression — other autonomic-correlate biometric
MOCs
- Vitals Knowledge Map — Biometrics section