What 'stress' really measures.

• It's a number, not a feeling

  The watch builds its 'stress' score from heart-rate variability: the tiny timing differences between consecutive beats. From those it estimates how far my nervous system was tilted toward 'activated' (the sympathetic side) versus 'at rest' (the parasympathetic side), scored against my own recent baseline. Garmin prints the word 'stress' on it. That label does a lot of quiet damage.

• It can't tell why I was activated

  A hard walk, a short night, a virus coming on, a glass of wine, caffeine, even digestion all push the same number up. Only some of that is anything a person would call stress in the everyday sense. The watch reads the arousal; it is blind to the cause. So a high 'stress' day is not evidence of a stressful day; it's evidence of an activated body, from a source the watch can't name.

• And the raw signal isn't even kept

  On this watch the underlying HRV is never stored; the score is computed on the wrist and the raw variability discarded. So even this proxy rests on a stand-in (an overnight-stress measure), which is why every stress-based finding here is treated as resting on a proxy, and said so plainly. See what this watch can't see.

So the honest gloss is "how activated my nervous system was," not "how stressed I felt." That's why the findings here group these channels as the autonomic-load family rather than as a stress meter, and why, when the watch's daily 'stress' and how I actually felt don't line up, it isn't the watch failing. It's two different things being held against each other: an arousal proxy and a feeling. And because the number is built from heart-rate variability, its relationship to anything we feel is curved by construction; a straight-line comparison was never the right tool.

Where that mismatch plays out: not a straight line. How the number gets built in the first place: from watch to number.

The full sensor audit, in the research repo: garmin_indicators_audit.md ↗.