Wearables in sleep and stress research: what the studies actually show

A University of Texas at Austin study, published in Nature Scientific Reports, put wrist-worn trackers on 82 young adults and followed them for months rather than a single night. The finding that made headlines: physical activity lengthened REM latency, the time it takes to enter REM sleep. The university's news service called it the most reliable validation to date of the link between activity, sleep quality and psychological health.

The methodology is the interesting part.

Participants wore a tracker that recorded movement and heart rate. From those two signals, the researchers inferred periods of deep (NREM) sleep, REM sleep and physical activity. A separate smartphone app collected self-reported well-being data. The study grew out of a pilot run under Whole Communities–Whole Health, a research programme at UT Austin.

According to the university, the results replicated lab findings. Low-intensity activity and moderate-to-vigorous activity were both linked to deeper, more restorative sleep, and better sleep in turn went with more energy and less stress the following morning. The novelty was the setting: continuous monitoring at home, at work and during daily activities over weeks and months, instead of a single night wired up in a clinic.

"You can learn a lot from lab studies, but obviously there are limitations to studying the sleep patterns of individual participants in just one night," said Benjamin Baird, a research assistant professor of psychology and one of the authors. "It's an unfamiliar, clinical-type setting, which can be stressful. And you can't really look over time, either. So, there are always questions about generalizability from that kind of design."

David M. Schnyer, co-author and chair of the Department of Psychology, went further. He said the team had shown that a standard Fitbit, not an expensive scientific device, was sensitive to sleep architecture measures and produced predictive results. Baird described sleep architecture as the structure of each 90 to 120 minute cycle: three stages of non-REM sleep plus REM, which makes up roughly the final 25% of each cycle.

From the bedroom to the population

The UT Austin study is small by consumer-tech standards. Sleep Cycle, a Swedish app, has about 2 million active nightly users. Its head of science, Michael Gradisar, told ABC News in September 2024 that the company's data scientists spotted an increase in night-time coughing in New York roughly a week before confirmed Omicron cases rose.

"It's almost like it's an early warning system. We had all of these phones at night collecting information," Dr Gradisar said.

The same ABC report, by technology reporter James Purtill, described how Sleep Cycle detected an above-average number of people briefly waking when an earthquake hit New South Wales earlier that month. Data scientists in Sweden picked it up. Less than 10 years after the first sleep-tracking Apple Watch, Purtill wrote, the world is wired for the study of sleep.

David Walsh built the first Apple Watch sleep-tracking app, AutoSleep, released in early 2016. He told the ABC it now has more than 5 million users. His method was the same inference trick the UT Austin team relied on: movement and heart rate, not electrodes.

Not everyone welcomed it. Gabriel Pires, a sleep researcher in Brazil, said he and colleagues started publishing papers warning that the new technologies did not work well. Gradisar, then a professor of sleep science at Flinders University, was also sceptical before he moved to industry.

Regulators have since moved. In early 2024 Samsung's smart watch sleep apnoea detection system received approval from the US health regulator after an eight-month review, the first consumer smart watch to clear that hurdle. Apple's smart watch received the same approval in September 2024.

Danny Eckert, director of the Adelaide Institute for Sleep Health at Flinders University, told the ABC that consumer devices could end up better at diagnosing sleep apnoea than the current gold standard. A 2022 study he co-authored found that up to half of single-night diagnostic studies for sleep apnoea misclassify severity. "The single-night tests are hugely noisy and inaccurate," he said.

The stress problem

Stress tracking has had a rougher ride. In August 2025, Wareable published a critique of news coverage around a study that tracked stress, fatigue and sleep for three months in 800 young adults wearing Garmin watches. Participants checked in four times a day on how stressed, fatigued or sleepy they felt, and the answers were cross-referenced with tracker data.

The study found "basically zero" correlation between a device's stress score and a user's self-reported feelings. The Guardian summarised it as smartwatches being unable to measure stress accurately. Other outlets went further, claiming watches confuse stress with exercise.

Wareable's objection was partly about hardware. The benchmark device was the Garmin Vivosmart 4, released in 2018. Testing a seven-year-old tracker to draw conclusions about the category, the piece argued, is like testing a 2018 Kia and pronouncing on electric vehicles. The critique also attacked the biology behind the "confuses stress with exercise" line. Exercise is acute physiological stress, which is how training adaptation works. A device that flags a workout as a high-stress event is doing its job.

The study did produce awkward numbers. No participants saw their tracker's stress scores meet the baseline for significant change when they recorded feeling stressed. For a quarter of participants, the device said stressed or unstressed when the participant reported the opposite.

Wareable's reading of that result is contrarian: a lack of correlation may say more about how badly people judge their own physiological state than about the sensor. The publication is not neutral. It reviews stress trackers and has criticised Garmin's data presentation separately. But the point about interpretation is hard to dismiss, given how readily users blame a device for a reading they do not like.

What the evidence supports

Three things survive contact with the sources. First, wearables are now good enough for longitudinal, in-the-field sleep research: UT Austin used an off-the-shelf Fitbit and published in a peer-reviewed journal. Second, scale changes what questions can be asked. Population-level cough monitoring, earthquake wake events and sleep apnoea screening every night for weeks are not lab experiments. The ABC reporting shows researchers treating them as a new kind of data rather than a replacement for clinical diagnosis.

Third, stress remains the weak spot, and the weakness is not only technical. The Garmin study's own design, one algorithm, one old device, one narrow set of self-reported metrics, makes it a poor basis for condemning the category. The headlines built on it were worse than the paper. Neither the UT Austin research nor the ABC reporting claims consumer trackers can diagnose anything on their own. Samsung says its apnoea feature is not meant to replace a sleep clinic.

The gap between what these devices measure and what people think they measure is where most of the argument lives. That gap is not closing on its own.