Researchers tracked African turquoise killifish — a species that lives only a few months — from birth to death, recording movement and sleep patterns continuously. Published in Science (Bedbrook, Nath, et al., Stanford, March 2026), the study found that by early midlife (70–100 days), behavioral differences already predicted which fish would live longer. Fish on longer trajectories moved more vigorously, reached higher speeds, and mostly slept at night. Fish on shorter trajectories increasingly slept during the day.
Machine-learning models using just a few days of midlife behavioral data could forecast lifespan. The signal wasn't a single biomarker. It was a pattern — the relationship between activity and rest across the day, the timing of sleep onset, the peak velocity during active periods. No single measurement distinguished short-lived from long-lived fish. The distinction was in the temporal structure of behavior.
The aging itself wasn't gradual. Most fish experienced two to six rapid behavioral shifts — transitions lasting only days — separated by longer periods of stability lasting weeks. Aging looked like a series of discrete steps, not a continuous decline. Each step locked in a new behavioral regime. The midlife signal was an early indicator of which step sequence a particular fish was on.
The killifish result is interesting for what it implies about measurement. Lifespan prediction from behavior doesn't require understanding mechanism. The model doesn't know why daytime sleep predicts shorter lifespan — whether it's a cause, a consequence, or a correlate of some underlying physiological state. It just knows the association is reliable. The prediction works without explanation.
For humans, the parallel is obvious: wearable devices already record movement and sleep continuously. The question isn't whether human behavior contains lifespan signals — given the killifish result, it almost certainly does. The question is whether anyone would want to know. Prediction without intervention is just prognosis. And prognosis without treatment is a number you have to live with.