A person wakes, reaches for the phone, and meets a number before meeting their own body. The screen is the first light in the room. The sleep app has already rendered a verdict on a night the person has not yet had time to inhabit. A green morning and a red morning are read as two different days before either has happened.
Sleep used to be the one part of life that escaped scoring. You woke, took stock of your body, and got on with the day. Now a ring on a finger or a watch on a wrist has already filed its report. The number arrives before the feeling, and often replaces it.
For most healthy people, tracking sleep does more harm than good, because the pursuit of a perfect sleep score produces the very anxiety and insomnia it claims to cure. The number does not merely report the night. It can become the night. That claim is contestable, and I think it is true.
Start with what sleep actually is. Sleep is a state of the brain. In a sleep lab, a technologist glues electrodes to your scalp and measures electrical activity, eye movements, and muscle tone. That is polysomnography, the gold standard for observing sleep. A wearable never sees the brain. It sees a wrist that is still and a heart that is slow, and it guesses.
The guess is built in three stages. First, sensors: a light on the underside of the device shines into your skin and reads blood volume changes, which yields heart rate and heart rate variability. An accelerometer reads motion. Some devices add skin temperature and overnight oxygen. Second, a classifier, usually a machine learning model trained on a small group of healthy adults sleeping in a lab while wearing both the device and the electrodes. Third, a nightly summary: total sleep, efficiency, latency, time in each stage, and then a score, a single number with a threshold that turns the whole night into good or bad.
Each of those stages is a place where the score drifts from the sleep. The training group is small and healthy, and the model learns what sleep looks like in a young adult lying still in a lab. It is then applied to an older adult with insomnia in their own bed. The device is good at detecting sleep and poor at detecting wake. A still arm reads as asleep. That is a minor error for a healthy sleeper and a serious one for a person with insomnia, because insomnia is long stretches of motionless wakefulness. The device flatters the bad night and contradicts the sufferer's experience. Stage percentages, the light and deep and REM breakdowns that people agonize over, are the noisiest output of all. The boundary between light and deep sleep, and between REM and wake, is where wrist algorithms disagree most with the brain.
There is a second drift, less obvious. Where a device ships a single nightly score, it compresses several dimensions into one number with a threshold. Duration, efficiency, latency, stages, heart rate variability, temperature, all folded into a single grade. The weighting is proprietary and not auditable by the person being graded. Two people with the same felt night can receive different scores. The same person can receive different scores on nights that felt identical. The number is a new thing in the bed. It stands between the sleeper and the night, graded by a formula the sleeper cannot see.
The number is a new thing in the bed.
The harm begins where the number meets the night. The mechanism has a name in the literature: measurement reactivity. The act of measuring and judging sleep becomes itself a source of arousal. Draganich and Erdal showed in 2014, in the Journal of Experimental Psychology: Learning, Memory, and Cognition, how far this goes. Participants were given fake feedback about their REM sleep. Those told they had above average REM performed better on cognitive tests. Those told below average performed worse. The actual sleep was the same. The belief about the sleep moved the day. The number became the experience.
Conditioned arousal does the rest. In insomnia, the bed and bedtime already become cues for wakefulness and anxiety. A tracker adds a second cue: the score. Checking the number, anticipating it, trying to earn it, raises arousal at the exact moment arousal must fall. Sleep effort is the clinical term for trying to sleep, and trying to sleep is itself the obstacle. A tracker can turn sleep into a task with a grade, for a user who treats the number as a verdict. That is sleep effort in its purest form. The loop closes on itself: a bad score produces anxiety and next-day hypervigilance, which produces a worse night, which produces a worse score. The tracker did not create the insomnia. It built a feedback loop around it.
The 3 a.m. check is the same logic in miniature: a screen glowing in a dark room, the night being audited before it is over.
The documented case that named this is a 2017 paper by Kelly Glazer Baron and colleagues at Rush University, published in the Journal of Clinical Sleep Medicine, which coined the term orthosomnia. Orthosomnia means an unhealthy preoccupation with achieving correct or perfect sleep as measured by a device. The paper described patients who had become fixated on their wearable data. One patient insisted on achieving eight hours of sleep as measured by their tracker. They went to bed earlier and stayed in bed longer to catch the hours. The device never showed eight. Their insomnia and anxiety worsened, and the number became the judge of whether a night counted as good. The treatment included cognitive behavioral therapy for insomnia and stepping away from the tracker.
Rendered plainly, that is the whole argument in one person. A person who decided that eight hours measured by a sensor was the definition of a good night, and then spent their nights performing sleep for the sensor. The sensor could not see their sleep. It could only see their stillness. The target was unreachable by construction, and the chasing made the sleep worse. That is what happens when a felt experience becomes a number to optimize.
The harm is not evenly distributed. It concentrates where the user is the judge and there is no interpreter. Anxious sleepers, perfectionists, and people with a history of insomnia are the most vulnerable, because they are already primed to monitor and catastrophize. Even calmer users can slide into reactivity once the nightly check becomes a ritual. The transition from useful feedback to performance review is a slope, not a wall, and no personality type has been shown to be fully immune.
I am not claiming that every healthy sleeper is harmed. The performance framing is the default condition of a culture that sells sleep as a project to optimize. The slope from feedback to performance review is the normal path, and the risk rises with anxiety and with how much weight the user gives the number.
There is also a version of this that is not about anxiety at all but about fairness. Nighttime care is not shared evenly. A tracker on a mother's wrist can turn a baby's waking into a scored failure. The score has no column for keeping a child alive. Meanwhile the partner in the same bed can wake to a green number. The number is not neutral. It is blind to caregiving labor. It measures one person's night and ignores the other's labor. Sleep tracking sold as self optimization can land hardest on people whose sleep is already constrained by care or shift work, because they cannot fix a score they never controlled.
Trackers do help, and the help is real in specific cases. People with undiagnosed sleep apnea may get a first warning from overnight oxygen dips: in 2024 Apple announced sleep apnea notifications for select Apple Watch models, developed with and validated against polysomnography, and Samsung's Galaxy Watch sleep apnea feature received FDA authorization in February 2024. Both are screening prompts, not diagnoses. Athletes use recovery scores as permission to rest, and sleep extension has been shown to improve sprint times and shooting accuracy in collegiate players. Clinicians use tracker data to correct sleep state misperception, showing a patient who insists they never sleep that they actually get six hours a night. Shift workers and people with circadian rhythm disorders benefit from longitudinal timing data. And a cheap continuous device reaches people who will never get a sleep study, which is a real equity argument. If a tracker nudges a chronic five hour sleeper toward seven, even imperfectly, it pushes in a direction associated in population studies with lower mortality and cardiometabolic risk. Those studies show association, not proof that changing sleep duration changes risk, and a tracker cannot tell voluntary short sleep from insomnia or apnea.
These cases are real. They are also specific. The genuinely helpful uses require either a specific condition, a clinician as interpreter, or a user who treats the score as a governor rather than a grade. The evidence that tracking improves sleep for the average healthy adult is thin: small studies, short time frames, self selected participants, and self reported outcomes. There is no large randomized trial showing that wearing a consumer sleep tracker improves sleep for healthy adults. There is also no large trial showing that it harms most of them. The harm evidence is case reports, clinical anecdotes, and user experience. So the honest position is not that tracking is proven harmful. It is that the net effect on the average healthy sleeper is unmeasured, and the risk of harm rises with the user's anxiety and with how much weight the user gives the number. Marketing implies certainty in both directions. The evidence offers a gradient.
What remains is the question the essay has been circling. A device that estimates your sleep does not know your sleep. The number is an inference from a still wrist, and it arrives before you have had a chance to ask your own body how the night went. For a person with apnea, that inference can be a lifeline. For a person who is anxious about sleep but otherwise healthy, it can become the thing it measures: the score replaces the sleep, and the night that used to end when morning came now leaves a number behind to be negotiated with.
Does knowing more about your body always help? I do not think the answer is yes. I think the answer depends on whether the knowing is in service of the body or in service of the number. The two look identical from the outside. They feel different from the inside, in the dark, at three in the morning, when the screen glows and the number is already there before you have asked yourself how you feel.