Optical Heart-Rate Sensing (PPG) Explained

What it is

Most wrist and finger wearables read heart rate with photoplethysmography, or PPG. Heart rate from wearables is derived from PPG, an optical method for measuring changes in blood volume under the skin. [1] A light shines into the skin, a sensor reads how much comes back, and software turns the rhythm of those changes into a pulse rate. The same light-based approach underlies fingertip pulse oximeters, which estimate oxygen saturation and pulse rate. [2]

Direct versus derived:

What the evidence shows

Resting and steady heart rate is usually close. A Duke University study compared consumer- and research-grade wearables across activities. Devices were reasonably accurate at rest and at prolonged elevated heart rate, but absolute error during activity was on average 30% higher than at rest, and devices differed in how they followed changes in activity. [1] Label: mixed.

Model matters. In 60 healthy young adults, with a Polar H10 chest strap as the heart-rate reference, the Apple Watch 6 had a coefficient of variation under 5% across sitting, walking, running, resistance exercise and cycling, while the Polar Vantage V and Fitbit Sense ranged from high to poor accuracy depending on the activity. All three had poor accuracy for energy expenditure (calorie) estimates. [3] Results apply to those models and that group, not to newer versions or other people.

Wearable pulse rate can have wide error in individual readings. A meta-analysis of four wearable pulse-rate studies (176 people) found average biases near zero, but 95% limits of agreement of about -16 to +14 bpm for light skin pigmentation, -19 to +17 for medium and -34 to +33 for dark. The authors found no statistically or clinically significant average bias by skin pigmentation, though the small number of studies limits conclusions. [4] The Duke study found no statistically significant difference in heart-rate accuracy across skin tones, while noting that no earlier study had systematically explored the full range of skin tones. [1]

Known sources of error. The Duke authors named three: diverse skin types, motion artifacts and signal crossover. [1] For how fit, placement and skin affect readings, see fit, placement and skin characteristics.

Pulse rate variability is not heart-rate variability

Heart-rate variability (HRV) is normally calculated from the electrical signal of an ECG (see electrical heart-rate measurement). A PPG device measures pulse timing, so its version is called pulse rate variability (PRV). In 103 college football players, heart rate and pulse rate agreed closely, but PPG-based PRV values were lower than ECG-based HRV (rMSSD 80.9 ms versus 103.9 ms), and PRV caught fewer of the changes ECG-HRV identified, with an average delay of 1.8 to 5.5 days. The authors concluded PRV should not be treated as an equivalent substitute. [5] A 2026 meta-analysis of 10 studies found small pooled differences (absolute standardized error 0.188 for rMSSD) based mostly on resting or controlled conditions, and warned against generalizing to sleep, exercise, stress or everyday life. [6] Label for using PRV as a stand-in for HRV in daily life: insufficient evidence. For what HRV means, see heart-rate variability explained.

Accuracy is not benefit

Accurate pulse-rate readings do not show that checking them improves health. Raise any worry about your heart rate with a clinician instead of relying on a watch trend; see resting heart rate.

Who should be careful

People who notice palpitations, fainting, chest pain or breathlessness should seek care rather than relying on a watch; in the US, call 911 for severe symptoms. Do not change medicines because of a wearable reading. The FDA lists poor circulation, skin thickness, skin temperature and nail polish among factors that can affect light-based oximeter readings, so a poor signal can happen in some situations. [7]

How to evaluate a claim

Look for the exact device model, the reference (ECG chest strap or 12-lead), the activity tested, and whether limits of agreement (not just averages) are reported. See smartwatches: what they measure.