Heart Rate

Heart rate is the number of heartbeats per minute, usually written as bpm. Apple Health stores observations from watches, connected monitors and other apps. Each record has a value and a timestamp or time interval; it is not a list of every heartbeat.

Use this page to understand the shape of those records before interpreting a trend. Three questions make a large difference: which source recorded it, what was happening, and how often were readings saved?

Apple Watch's optical sensor uses light to detect changing blood flow at the wrist. Supported models also have an electrical sensor used through the Heart Rate or ECG apps. Fit, wrist blood flow, movement and some tattoos can affect optical measurements; an occasional extreme value can be an anomaly. Apple explains measurement modes and measurement limitations.

Recording behavior changes with hardware and settings. Apple's September 2026 documentation describes daylight measurements as often as every five seconds on Series 12 and Ultra 4 with continuous measurements enabled. That does not make a five-second schedule universal across Apple Health history. A stored timestamp span also does not describe the optical sensor's internal sampling frequency.

The figures use example Apple Health records, analyzed over a full reporting year with a 92-day focus. To avoid pooling different source groups, both charts select the source group with the most heart-rate records in that focus window. They show observations, not population targets.

The sleep subset spans only five observed days, while workout readings span 48 and other readings span all 92. Another source supplied much of the recorded sleep data. The chart therefore illustrates both contextual differences and uneven source coverage; it cannot estimate a typical person's sleeping heart rate.

Contexts use timestamp overlap with recorded workouts and asleep intervals. “Other or unknown” includes ordinary movement, rest and missing context. It is not a resting-heart-rate measurement. A workout overlap is a useful label, but does not prove that the workout produced the sample.

In the selected source, the median positive gap within recorded workout episodes was five seconds; the middle 90% was one to nine seconds. Within recorded sleep episodes, the median was about 4.1 minutes, with a middle 90% of roughly 43 seconds to 8.1 minutes. These are measured distributions for this example, not promised recording intervals.

The cadence figure shows the percentage of gaps at or below each duration. Workout and sleep curves exclude time between separate episodes. The other/unknown curve uses consecutive records in that context and can include both short bursts and long gaps. Comparing a single mean interval would hide these differences.

A day with readings does not establish continuous monitoring. A gap cannot by itself distinguish charging, non-wear, unavailable readings, recording policy or missing synchronization. Keep zero-length sample durations, zero timestamp gaps and missing intervals separate: they describe different things.

A mean across saved readings weights each record equally. When workouts generate many more records per minute, they contribute much more than their share of elapsed time.

In the full-year example, the mean across all 203,272 records was 81.6 bpm. Averaging each observed day's sample mean equally gave 75.6 bpm. Neither is a time-weighted all-day heart rate, and neither is Apple's dedicated resting-heart-rate estimate. The difference demonstrates how recording density and daily activity can change a summary.

Use the Resting Heart Rate entry for that separate estimate. For your own trends, compare like-for-like sources and contexts, and state the averaging method.

The full-year snapshot contained one stored zero. It was retained in the research and flagged as an invalid physiological heart-rate observation; silently deleting it would conceal a data-quality issue. Other low or high values were checked against neighboring readings, source and context. Some high readings appeared outside a recorded workout, showing why lack of workout overlap is not proof of rest.

A statistical outlier is a prompt to inspect a record, not a diagnosis or proof of sensor error. Check the same source's nearby readings, movement context and interval duration before deciding whether an observation belongs in a summary. Avoid turning a broad optical support range into a clinical normal range or a universal validation rule.

For adults who are calmly resting and feeling well, the American Heart Association describes 60–100 bpm as a usual resting range. Fitness, medication and sleep can produce lower values; exercise normally raises heart rate. Interpretation depends on the person and symptoms. AHA's heart-rate guidance advises urgent help for a sudden very high or low rate with symptoms such as chest pain, breathlessness, dizziness or fainting.

Heart-rate variability describes variation in beat intervals, not the number of beats per minute. Sleep-stage intervals, respiratory rate and workout timing can also help explain changing heart-rate patterns.

The research paired daily summaries with these types and compared workout phases and recorded sleep stages. These are exploratory comparisons within the example data. Different sampling schedules, medication, illness, movement and source changes can affect both measures. An association does not establish a mechanism; pooled all-day averages are especially vulnerable to recording-density effects.

Keep canonical beats-per-minute values, original sample boundaries, source revisions and raw metadata. Heart-rate motion context uses 0 = not set, 1 = sedentary, 2 = active. Active includes movement or working out; it is not an explicit workout link. Apple documents the motion-context metadata.

Some historical exports serialized this enum as booleans. This research retained the original representation and applied the database's narrowly documented legacy decoder for this key only. Do not extend that conversion to arbitrary boolean metadata or reinterpret “not set” as sedentary.