How do environmental and headphone noise relate to heart rate variability?

SDNN as a one-minute readout of autonomic regulation

Heart rate variability (HRV) describes how much the interval between consecutive heartbeats changes. A healthy heart adjusts beat-to-beat timing through the autonomic nervous system, so HRV is used as a non-invasive indicator of that regulation. The paper focuses on SDNN, the standard deviation of normal-to-normal intervals, which summarizes overall variation in milliseconds rather than separating sympathetic and parasympathetic contributions.

The measurement here is a 1-minute SDNN estimate produced by a smartwatch, paired with sound levels from smartwatches and smartphones. That choice matters for interpretation: the outcome is a short-window variability statistic, not a clinical Holter recording, and the paper notes that consumer-grade wearables were uncalibrated. The authors also report that noise exposure alone explained little of the variance in SDNN, with marginal R² values of 0.007 to 0.018 across the four models, while individual differences explained much more.

Four cohorts split by sound source and exposure window

The analysis drew on consented adults in the Apple Hearing Study, a nationwide prospective cohort, using data collected from November 2019 to December 2024. Eligible participants had shared headphone audio levels or environmental sound levels plus HRV through their iPhone, contributed at least 20 consecutive SDNN measurements, and completed a baseline survey on demographics, self-rated hearing, and tinnitus within one week of enrollment.

The authors built four analytical cohorts by crossing sound source (environmental versus headphone) with exposure window (short-term, 20 minutes, versus long-term, 160 minutes). Because of demographic representativeness and model computation limits, they randomly sampled 20 participants from each unique age-sex-ethnicity stratum, using a fixed seed. Strata with fewer than 20 eligible participants were excluded, which mainly affected older participants. The resulting samples ranged from 980 to 1,180 participants across analyses, drawn from pools as large as 49,983 eligible people for short-term environmental noise and as small as 10,726 for long-term headphone sound.

Exposures were equivalent sound levels in A-weighted decibels, and models adjusted for time of day and physical activity. Distributed lag non-linear hierarchical Bayesian models estimated lag-specific and overall SDNN changes, with random effects for individual variation; a second-stage linear regression summarized average changes by age, sex, BMI, stress, self-rated hearing, and tinnitus. Median sound levels differed by source: 52.3 and 62.0 dBA for environmental noise in the short- and long-term cohorts, versus 66.4 and 68.4 dBA for headphone sound.

A brief rise in SDNN, then a sustained decline

The lag-specific predictions show a consistent shape. When environmental noise rose from the 40 dBA reference to 50 dBA, SDNN increased by a median 0.2–0.7% at each 1-minute lag during the first 6 minutes, showed no meaningful change around 7–8 minutes, then fell by 0.2–1.4% from 8 to 20 minutes. Headphone sound produced a similar but shorter early bump, a 0.1–0.2% increase over the first 3 minutes, followed by a 0.1–0.9% reduction after 5 minutes.

The long-term models looked different in timing. For environmental noise at 50 dBA, reductions of 1.4–2.9% appeared across the 160-minute lag window, with smaller contributions from earlier exposures. For headphone sound, the initial reduction was 1.4% but flattened toward zero, at –0.5%, in the last 20-minute lag. The authors interpret this pattern as physiological stress that can persist from minutes to hours, reflected in lower HRV after elevated noise exposure.

Environmental noise reduction grows with the exposure window

The headline comparison is between windows. A 10-decibel increase in environmental noise was associated with an overall SDNN reduction of 6.8% in the short-term model and 16.0% in the long-term model, with 95% credible intervals of 5.1–8.4% and 13.5–18.3%. At 60 dBA, the long-term predicted reduction was 27.4%, more than twice the short-term figure.

The short-term environmental curve was not monotonic: it was slightly U-shaped, with the largest reduction, 10.4%, around 60 dBA. The authors note that this turning point lies close to a 65 dBA cut-off reported elsewhere and suggest the attenuated reduction at high short-term levels may relate to fewer data points at the upper end of the exposure range. That is a data-density explanation, not evidence that loud short-term noise is safer.

Headphone sound stays flatter across windows

Headphone sound showed a much smaller gap between windows. A 10-decibel increase was associated with a 7.1% overall SDNN reduction in the short-term model and 7.4% in the long-term model. At 60 dBA, the predicted reductions were 14.3% and 14.6%, respectively, close to each other and well below the long-term environmental figure.

The authors point out that headphone exposure is more likely to be voluntary, desirable, or used for stress relief than environmental noise, so the two sources may not act through the same pathway. The headphone sampling pool was also less than half the size of the environmental pool, partly because headphone use occupies less of a day. These differences in context and sample size qualify any direct ranking of the two sources.

Older adults and people with tinnitus show larger reductions

Subgroup comparisons used individual-specific posterior estimates and did not include significance testing, because uncertainty from the relative-change estimates was not carried into the individual-level model; the reported 95% confidence intervals may therefore be too narrow. With that caveat, older adults showed greater average SDNN reductions than younger participants in the long-term environmental model, and people reporting tinnitus showed greater reductions than those who never had it. At 60 dBA, participants with tinnitus a few times a year or a few times a day had reductions of about 28%, compared with about 25.6% for those who never had tinnitus.

Self-rated hearing split the two sources in opposite directions. For environmental noise, people who reported excellent hearing showed greater reductions at loud levels. For headphone sound, participants with poorer self-rated hearing stood out: at 70 dBA their average change was –25.6%, versus –22.3% for excellent-hearing participants, and at 80 dBA the gap was –32.5% versus –26.9%. Perceived stress also mattered for headphones, with the most stressed group showing –18.8% at 60 dBA compared with –13.1% for the less stressed.

The authors report that the included variables explained less than 20% of between-subject variation in the environmental models and up to 23% in the headphone models, with short-term models explaining less. Age, stress, hearing, and tinnitus therefore mark groups with different average responses, but they do not account for most of the person-to-person differences.

Where the findings apply and where they stop

The study's strength is scale and realism: it characterizes short- and long-term sound exposure and HRV using free-living consumer-wearable data, which laboratory studies of brief controlled noise cannot capture. The distributed lag model also avoids assuming that exposures further back in time contribute equally to the response, and it allows the exposure-response relationship to vary over time.

The boundaries are equally specific. The cohort is a subset of Apple Hearing Study participants who consented to share audio and HRV data and met the measurement requirements, and the stratified sampling excluded strata with fewer than 20 eligible participants, mainly older ones. Sound levels and SDNN came from uncalibrated consumer-grade devices, and the outcome was 1-minute SDNN rather than a clinical HRV measure. The authors also note that the analysis is observational, so the associations should not be read as established causal effects of noise on HRV.

For readers weighing the practical message, the most decision-relevant split is not noise versus silence but source and window. A 10-decibel rise in environmental noise carried a reduction more than twice as large in the 160-minute window as in the 20-minute window, while headphone sound stayed near 7% in both. Within that pattern, the groups the paper flags for closer attention are older adults and people with tinnitus, plus people with poor self-rated hearing for headphone exposure. Those are the strata where the paper's own stratified results diverge most from the cohort average, and where the authors argue risk assessment should be concentrated.

Sources used in this answer