Core body temperature is the gold-standard biomarker for circadian timing. Its 24-hour oscillation — peaking in late afternoon, troughing in early morning — tracks the master clock in the suprachiasmatic nucleus more directly than any behavioral measure. Specialized sensors that measure core temperature should therefore be the best wearable tool for circadian assessment.
A comparative study of commercial wearables (arXiv:2404.03408) finds the opposite. Simple accelerometer-based activity measurement correlates with reference circadian timing at r=0.96 — higher than heart rate and substantially higher than core body temperature measured from consumer devices. The cheapest, simplest sensor wins.
Note (2026-11-15; read against the paper's full text): the reference for the activity measure is itself an activity measure, the acrophase derived from Actigraph activity, so the 0.96 correlation compares two activity-based estimates. Against the chronotype questionnaire the ordering differs: heart rate has the strongest relation (r = -0.73), surpassing the reference Actigraph (-0.71) and the watch activity (-0.66), with temperature at -0.61, and heart rate's R2 for predicting the questionnaire score (0.53) is comparable to the Actigraph's (0.51). Combining the core body temperature sensor with the smartwatch gave the best prediction (R2 = 0.64). The temperature data came from a dedicated core body temperature monitor (the CORE sensor, which combines skin temperature and heat flux) alongside a smartwatch, so the essay's account of noisy wrist-worn consumer temperature, and its conclusion that the simplest sensor wins, are not what the paper reports.
The reason: consumer temperature sensors introduce measurement noise that degrades the circadian signal. Wrist-worn temperature is confounded by ambient conditions, clothing, and peripheral vasoconstriction. The theoretical superiority of temperature as a biomarker is overwhelmed by the practical inferiority of measuring it from a watch. Meanwhile, activity measurement — just detecting whether the arm is moving — is trivially accurate from an accelerometer and captures the dominant circadian-driven behavior: the sleep-wake cycle itself.
This creates a hierarchy inversion. In clinical settings with rectal temperature probes and controlled environments, temperature dominates. In real-world settings with consumer devices, motion dominates. The best measurement depends on the measurement context, not on the biology.
The broader pattern: when choosing between a theoretically ideal measurement with practical noise and a theoretically inferior measurement with practical precision, the latter often wins. Signal-to-noise ratio matters more than signal amplitude. The motion sensor doesn't track the circadian clock — it tracks the behavior the clock produces. But it tracks that behavior with so much less noise that the indirect measurement outperforms the direct one.