Sleep Interpretation Library · Revised September 24, 2026

Why Is My Sleep Score Stable But My REM Dropping?

By Mike Ye & Ella

A composite sleep score can hold roughly steady while a wearable’s REM estimate changes. The devices infer stages rather than measuring them with a clinical sleep study, and the score’s exact weighting is proprietary. A difference between the score and one estimate is worth investigating; it does not prove physiological strain.

The observed divergence

In the historical Week 3 report, average REM was 0.67 hours versus 1.10 hours in Week 2, and deep sleep averaged 0.99 hours while the overall score stayed relatively stable. HRV was largely unrecorded that week, so there was no complete autonomic series to confirm a broader response. The earlier report called this “latent strain”; the measured finding is narrower: stable composite score, weaker stage estimates and missing HRV.

Questions before interpreting it

Check total sleep and bedtime first: a shorter or earlier-ended night can reduce the opportunity for later REM. Then check device consistency, missing nights, alcohol and illness, training dose, travel and other known conditions. One week’s Garmin stage estimate may also reflect classification error. Neither an unchanged activity label nor a lower REM estimate rules these possibilities out.

What to track next

Compare repeated nights with similar sleep opportunity and well-documented preceding days. In the active food-to-sleep log, the exposure day and the next-night response are linked, and unknown values stay blank. A stable score with repeatedly lower estimated REM is a question for further observation; it is not an early-warning prediction or a diagnosis. See the revised term and research method.