A tiny fish's daily routine can predict when it will die

The African turquoise killifish lives fast and dies young. With a lifespan of just a few months, it compresses the full arc of vertebrate aging - adolescence, adulthood, decline, death - into a window short enough to watch from start to finish. That brevity has now made it the subject of one of the most complete behavioral portraits of aging ever assembled.

Claire Bedbrook and colleagues built a high-resolution recording platform that tracked killifish continuously, from roughly three to four weeks of age until death. Using machine learning and computer vision, they captured every swim, every burst of speed, every rest period. The result is what the team calls a "behaviorome" - a comprehensive behavioral dataset spanning the full adult life of each individual animal.

Fast movers, long lives

The data revealed something striking: fish destined to live longer were distinguishable from their shorter-lived counterparts well before any obvious signs of decline appeared. Long-lived killifish were more active, moved faster, and produced more vigorous bursts of movement early in adulthood. They also consolidated most of their sleep into nighttime hours.

Short-lived fish told a different story. They slept more during the day, displayed fragmented activity patterns, and moved with less vigor - behavioral signatures that emerged early and persisted. These were not fish that suddenly declined. They were fish that never quite hit their stride.

Building a behavioral clock

From these patterns, the team trained a machine learning model to estimate a fish's chronological age using only its daily movement and activity data. The model worked. But the researchers pushed further, asking whether the same behavioral measurements could predict something far more consequential: how long an individual would ultimately live.

Beginning in early adulthood, the behavioral patterns alone could reliably forecast whether a fish would have a relatively short or long lifespan. The behavioral clock did not require genetic data, blood markers, or anatomical measurements. It needed only the rhythms of daily life.

What behavior reveals about aging

Aging research has long relied on molecular and cellular markers - telomere length, gene expression, protein aggregation. Behavior, by contrast, offers a window into an animal's functional state, integrating signals from the nervous system, musculature, metabolism, and sensory organs simultaneously. A fish that swims less is not just moving less. Something systemic has shifted.

The challenge has always been observation at scale. Tracking behavior continuously across an organism's entire lifespan demands infrastructure that simply did not exist for most model organisms until recently. The killifish, with its compressed timeline, made the problem tractable.

But the implications reach beyond fish. Vertebrate aging shares deep conserved features across species. The behavioral architecture the team uncovered - distinct trajectories, early-life predictors, progressive fragmentation of activity cycles - may reflect principles that apply more broadly, including in humans, where disrupted sleep and declining physical activity are among the earliest signs of age-related decline.

A lifespan prediction from weeks of data

Perhaps the most provocative finding is the timeline. The behavioral signatures that predicted lifespan were visible in early adulthood, weeks before the fish showed any conventional signs of aging. This suggests that the trajectory of decline is not purely a late-life phenomenon but is instead shaped, or at least foreshadowed, by patterns established much earlier.

The study does not resolve whether these early behavioral differences cause longer or shorter lives, or whether they merely reflect underlying biological variation that independently drives both behavior and longevity. Correlation and causation remain entangled here.

Small sample, large ambition

There are important caveats. The killifish is a laboratory model with a deliberately simplified environment - no predators, no competition, no variation in food supply. Whether behavioral clocks built under such controlled conditions would function in natural settings remains untested. And while vertebrate aging shares conserved features, the leap from a fish that lives months to a human who lives decades involves enormous biological distance.

Still, the work represents a proof of concept with genuine potential. If behavioral patterns can predict biological aging trajectories, they could eventually complement molecular biomarkers in assessing health span - potentially offering a non-invasive, continuous, and relatively inexpensive approach to monitoring aging in other species.

For now, the killifish has delivered a message worth hearing: the way an animal moves through its days carries information about how many days it has left. We have been measuring aging by counting molecules. Maybe we should also be watching how things move.