This conclusion comes from a new study supported by the Knight Initiative for Brain Resilience at Stanford’s Wu Tsai Neurosciences Institute. Researchers continuously monitored dozens of short-lived fish throughout their lives to better understand how behavior connects to aging.

Even though the fish shared similar genetics and lived in the same controlled conditions, they aged in very different ways. By early adulthood, those differences were already visible in how they swam and rested. These patterns were strong enough to predict whether a fish would ultimately have a shorter or longer lifespan.

Although the study focused on fish, the findings suggest that tracking subtle daily behaviors such as movement and sleep, now commonly recorded by wearable devices, could provide insight into how aging progresses in humans.

The research, published in Science on March 12, 2026, was led by Wu Tsai Neuro postdoctoral scholars Claire Bedbrook and Ravi Nath. It grew out of a collaboration supported by the Knight Initiative between the Stanford labs of geneticist Anne Brunet and bioengineer Karl Deisseroth, the study’s senior authors.

Tracking Aging in Real Time

Most aging research compares young animals to older ones. While useful, this approach can miss how aging unfolds within individuals over time and how differences between individuals develop.

Bedbrook and Nath wanted to follow aging continuously across an entire lifespan. Even animals raised under nearly identical conditions can age differently and live for very different lengths of time. The team aimed to determine whether natural behavior could reveal when those differences begin.

To do this, they used the African turquoise killifish, a species with a lifespan of just four to eight months. Despite its short life, it shares important biological features with humans, including a complex brain, making it a valuable model for aging research.

The Brunet lab has played a leading role in establishing the killifish as a model organism. This study was the first to track individual vertebrates continuously, day and night, throughout their entire adult lives.

The researchers designed an automated system where each fish lived in its own tank under constant camera surveillance. Similar to a real-life version of The Truman Show, the setup recorded every moment of each animal’s life. In total, the team followed 81 fish and collected billions of video frames.

From this massive dataset, they analyzed posture, speed, rest, and movement. They identified 100 distinct “behavioral syllables,” which are short, repeating actions that form the basic elements of how the fish move and rest.

“Behavior is a wonderfully integrated readout, reflecting what’s happening across the brain and body,” said Brunet, the Michele and Timothy Barakett Professor of Genetics at Stanford Medicine. “Molecular markers are essential, but they capture only slices of biology. With behavior, you see the whole organism, continuously and non-invasively.”

With this detailed record, the researchers began asking new questions: When do individuals start aging differently? What early traits define those paths? And can behavior alone predict lifespan?

Early Behavioral Signals of Longevity

One of the most striking discoveries was how early aging paths begin to diverge. After tracking each fish for its entire life, the team grouped them by lifespan and then looked back to identify when behavioral differences first appeared. They found that by early midlife (70 to 100 days of age), fish that would later live longer or shorter lives were already behaving differently.

Sleep patterns stood out as a key factor. Fish that eventually had shorter lifespans tended to sleep not only at night but increasingly during the day. In contrast, fish that lived longer mostly slept at night.

Activity levels also played a role. Fish on longer lifespan trajectories swam more vigorously and reached higher speeds when moving around the tank. They were also more active during daylight hours. This type of spontaneous movement has been linked to longevity in other species as well.

Importantly, these behavioral differences were predictive, not just descriptive. Using machine learning models, the researchers showed that only a few days of behavioral data from middle-aged fish were enough to estimate lifespan. “Behavioral changes pretty early on in life are telling us about future health and future lifespan,” said Bedbrook.

Aging Happens in Distinct Stages

The study also revealed that aging does not progress in a slow, steady way. Instead, most fish experienced two to six rapid shifts in behavior, each lasting only a few days. These transitions were followed by longer periods of stability that lasted weeks. Fish generally moved through these stages in sequence rather than switching back and forth.

“We expected aging to be a slow, gradual process,” said Bedbrook. “Instead, animals stay stable for long periods and then transition very quickly into a new stage. Seeing this staged architecture appear from continuous behavior alone was one of the most exciting discoveries.”

This stepwise pattern aligns with findings from human studies, which suggest that molecular changes in aging occur in waves, particularly during midlife and later years. The killifish results provide a behavioral perspective on this phenomenon.

The researchers propose that aging may involve long periods of relative stability interrupted by brief, rapid changes. They compare it to a Jenga tower, where many blocks can be removed with little effect until one critical change triggers a sudden shift.

To explore the biology behind these patterns, the team examined gene activity in eight organs at a stage when behavior could reliably predict lifespan. Instead of focusing on single genes, they looked at coordinated changes across groups of genes involved in shared processes.

The most noticeable differences appeared in the liver. Genes related to protein production and cellular maintenance were more active in fish with shorter lifespans. This suggests that internal biological changes occur alongside behavioral differences as aging progresses.

Behavior Offers a Window Into Aging

“Behavior turns out to be an incredibly sensitive readout of aging,” said Nath. “You can look at two animals of the same chronological age and see from their behavior alone that they’re aging very differently.”

This sensitivity is evident in many aspects of daily life, especially sleep. In humans, sleep quality and sleep-wake cycles often decline with age, and these changes are linked to cognitive decline and neurodegenerative diseases. Nath plans to investigate whether improving sleep could support healthier aging and whether early interventions could shift aging trajectories.

The researchers also plan to explore whether aging paths can be altered through targeted strategies, including dietary changes and genetic interventions that may influence the pace of aging.

For Bedbrook, the findings raise broader questions about what drives transitions between aging stages and whether those shifts can be delayed or reversed. She is also interested in moving toward more natural environments, where animals can interact socially and experience more realistic conditions.

“We now have the tools to map aging continuously in a vertebrate,” she said. “With the rise of wearables and long-term tracking in humans, I’m excited to see whether the same principles — early predictors, staged aging, divergent trajectories — hold true in people.”

Another key area of research involves the brain. Deisseroth’s lab is developing tools to monitor neural activity continuously over long periods, which could reveal how brain changes align with aging in the rest of the body or potentially influence its pace.

Bedbrook and Nath will continue this work as they establish their own laboratories at Princeton University this July, building on the tools and insights developed at Stanford.

Ultimately, this research aims to explain why aging varies so widely and to uncover new ways to support healthier, longer lives.

Publication Details Research Team

Study authors were Claire Bedbrook from the Department of Bioengineering at Stanford Medicine and Stanford Engineering; Ravi Nath from the Department of Genetics at Stanford Medicine; Libby Zhang from the Department of Electrical Engineering at Stanford at Stanford Engineering; Scott Linderman from the Department of Statistics in Stanford Humanities and Sciences, the Knight Initiative for Brain Resilience and the Wu Tsai Neurosciences Institute; Anne Brunet from the Department of Genetics at Stanford Medicine, Wu Tsai Neurosciences Institute, Knight Initiative for Brain Resilience, and the Glenn Center for Biology of Aging; and, Karl Deisseroth, the D.H. Chen Professor, from Departments of Bioengineering at Stanford Medicine and Stanford Engineering and of Psychiatry and Behavioral Sciences at Stanford Medicine, Knight Initiative for Brain Resilience, and the Howard Hughes Medical Institute at Stanford University.

Research Support

The research was funded by the National Institutes of Health (R01AG063418 and K99AG07687901), a Knight Initiative for Brain Resilience Catalyst Award and Brain Resilience Scholar Award, the Keck Foundation, the ARIA Foundation, the Glenn Foundation for Medical Research, the Simons Foundation, the Chan Zuckerberg Biohub — San Francisco, a NOMIS Distinguished Scientist and Scholar Award, the Helen Hay Whitney Foundation, the Wu Tsai Neurosciences Institute Interdisciplinary Scholar Award, and the Iqbal Farrukh & Asad Jamal Center for Cognitive Health in Aging.

Competing Interests

Karl Diesseroth is a cofounder and a scientific advisory board member of Stellaromics and Maplight Therapeutics, and advises RedTree and Modulight.bio. Anne Brunet is a scientific advisory board member of Calico. All other authors declare no conflicts of interest.