Modeling the chaos in the behavior of larval zebrafish

Zebrafish have been used as models for human development due to their neural and motor system similarities to humans. Since their larvae are transparent, researchers can track their neural dynamics during development through observation of their behaviors.

Han et al. set out to extract hidden behavioral states of zebrafish larvae by monitoring and recording their postural dynamics — how the larvae move in space based on sensory information.

To determine the behavioral states, they observed the larvae and extracted seven behavioral factors. Then, they could group them into three behavioral states: turning, routine swimming, and slow swimming.

Then, using a dynamical reconstruction framework, they simplified the high-dimensional, complex behaviors of the larvae into a low-dimensional state space. Building on this, the researchers achieved short-term prediction of future behavioral states, “providing a unified quantitative framework for describing behavioral organization, transitions, and predictability,” said author Peng Ji.

“This suggests that different motor states may rely on shared neural circuits or motor-control modules, potentially including central pattern generators,” he said.

In response to external stimuli, the behavioral dimensionality of the larvae was reduced. Ji believes that this type of reorganization is caused by modality-specific sensory and motor pathways that precisely control motor states, something they want to explore in future studies.

“We plan to further develop interpretable parametric models that quantitatively describe the rich behavioral repertoire of zebrafish and support more realistic behavioral simulation in virtual environments,” he said. “We also hope to combine behavioral reconstruction with whole-brain neural recordings to investigate the neural mechanisms that maintain different behavioral states and drive transitions between them.”

Source: “Latent factors capturing evolution of behavior in larval zebrafish,” by Changhong Han, Yuchen Gong, Le Sun, Jiulin Du, and Peng Ji, Chaos (2026). The article can be accessed at https://doi.org/10.1063/5.0354035 .