Does Your Wearable's Stress Score Actually Measure Stress?

This article covers what electrodermal activity (EDA) sensors in consumer wearables actually detect and what the underlying research says about EDA as a marker of sympathetic arousal. It does not cover heart rate variability-based stress scores, cortisol, or clinical stress diagnosis.

Electrodermal activity is a genuine, sensitive readout of sympathetic nervous system arousal, that part is well established. But arousal is not the same thing as stress: the same EDA spike can come from heat, physical movement, cognitive effort, pain, or an emotional reaction, and a stress score built on it has to guess which one it's looking at.

The spike is real. The story around it is a guess

You get the notification: elevated stress detected. Maybe you were annoyed. Maybe you were just walking fast in the heat, or concentrating hard on something that felt fine. The number on the screen doesn't know the difference, and that gap between what the sensor measured and what the app tells you it means is where most of the confusion about these features actually lives.

Electrodermal activity is one of the older tools in psychophysiology, tracking tiny changes in skin conductance driven by sweat gland activity, which is controlled by the sympathetic nervous system. That link to sympathetic arousal isn't in dispute. What's less settled is whether a single number pulled from a wrist sensor, worn all day, in the real world, can reliably be labeled 'stress' rather than 'arousal of some kind.'

What people keep asking about the number on the screen

The tension shows up less as skepticism about the sensor itself and more as confusion about what the resulting score is actually sorting.

Arousal has more than one source

Part of what complicates a wearable stress score is that EDA doesn't only rise during emotionally stressful moments. Research comparing EDA to skin sympathetic nerve activity during cognitive stress and pain tasks found the two signals tracked each other closely once timing differences were accounted for, which supports EDA as a genuine sympathetic marker. But the same studies used deliberately distinct stimuli, Stroop tasks, a pain test, a breathing maneuver, precisely because different kinds of sympathetic activation can look similar in the raw signal.

That ambiguity shows up in unexpected places. One study examining electrodermal activity during a haptic exploration task found that adult participants' EDA was actually lower when using a robotic assistance system, even though the setup was meant to test for uncertainty or stress. It's the kind of result that cuts against a simple 'higher EDA equals more stress' reading.

Motion is another confound. Wearables get used during ordinary movement, not just seated lab sessions, so researchers have had to build machine learning methods specifically to detect and strip out motion artifacts from ambulatory EDA recordings. That work exists because the raw signal picked up on a wrist during daily life is noisier than the signal from an electrode in a controlled room.

Where the signal has shown something clinically real

None of this means EDA is meaningless outside the lab. A large community study using passively collected EDA from a wrist sensor over several weeks found differences in diurnal EDA patterns, along with temperature and heart rate, associated with mental health status in free-living conditions, a step toward showing these signals carry real information outside a lab session.

Separately, a systematic review of EDA research concluded that hypoactive electrodermal response is a fairly consistent feature in patients with depression, and found preliminary evidence that EDA monitoring might help distinguish phases of mood disorders and, in some studies, differentiate acutely suicidal patients from non-suicidal depressed patients. I find that review's own framing worth sitting with: it calls EDA a valid, sensitive marker while also treating the differentiation findings as preliminary, not settled.

A separate machine learning study using EDA recorded across baseline, stress, and recovery phases was able to screen for major depressive disorder with a decision tree classifier at meaningfully better than chance accuracy. The study population, though, was limited to a small group of diagnosed patients and healthy controls rather than a general wearable-wearing public.

Why the same reading can mean different things for different people

A framework applying generalizability theory to psychophysiological measures, including EDA, makes a point that matters for anyone staring at a daily stress trend line: reliability isn't a fixed property of a signal, it depends on how much of the variation comes from real trait-level differences versus noise across trials, tasks, and sessions. Put plainly, a single stress score on a single day is a much shakier read than a pattern across many days.

Formal guidelines for publishing EDA research also lay out just how many variables affect the raw signal, electrode placement, skin temperature, hydration, recording technique. Researchers who study this signal for a living treat it with more caution in write-ups than a consumer app treats it on a wrist.

Common questions