After calibration sets an individual baseline, affect, fatigue, and time drift the baseline itself

What it is

Calibration measures an intercept from those twenty minutes of rest, not a permanent zero. Mood that day, sleep debt, caffeine, and the circadian cycle move the intercept. Drift happens in the baseline itself, so “change relative to calibration” mixes in components that have nothing to do with the current task.

Why it happens

The autonomic set point moves with circadian phase and homeostatic load: morning HRV is not afternoon HRV; sleep loss raises sympathetic tone. Affect is not only phasic: a sustained mood lifts or lowers tonic EDA and heart-rate level. Unlike hardware slip, the person is still in the same place and the physiology’s DC is walking. An adapter that subtracts the first calibration as a constant interprets new DC against an expired zero. Drift can be slow (hours of circadian) or a step after an argument. Non-stationarity is the default of physiological channels, not a fault.

Studying it

Repeat no-task rest at several points after calibration: morning versus afternoon the same day, before and after sleep restriction, before and after a drug. Outcomes: offset of rest level from first calibration, false “state change” positives computed on the old baseline. Contrast a mechanical re-don condition to separate device offset from physiological drift. Time-series models (drift-diffusion, sliding baseline) should report how window length moves false positives. A single laboratory session cannot see the circadian cycle.

Where it stops holding

On a very short task (five minutes of experiment right after calibration) drift can be ignored. Drug half-life, shift work, and jet lag can dwarf laboratory mood induction. Algorithms that continually update the baseline can absorb a true slow state change (a depressive episode) into the “new normal.” Drift direction differs across people; a group-mean curve is not a personal forecast.

Applying it

• Treat the baseline as an estimate with a half-life, not a snapshot; use a sliding window for slow drift and a change detector for steps.
• Offer an optional short calibration at known physiological kinks (waking, medication, timezone shift).
• Do not write “up relative to this morning’s calibration” as a definite affective event unless the window is short enough that drift cannot dominate.
• Verify by using morning calibration to interpret afternoon rest with no task, and counting how often it is labeled a state change; a high rate means calibration was treated as a frozen zero.

Related

• Same group: C9.10.1 Resting physiological baselines differ across people; absolute values cannot be compared across users · C9.10.3 Long-term use needs periodic recalibration; a one-shot calibration does not last · C9.10.4 Unrecognized baseline drift is misread as a state change
• Adjacent: C9.09 Wearable Fit and Signal Quality · C8.10 Gaze Drift and Recalibration
• Search: baseline drift · non-stationary physiology · circadian HRV