In daily life, recent experience shapes future choices. When foraging or avoiding risks, animals adjust their next movement based on recently acquired cues. This bias is widely observed across species and is thought to help the brain exploit temporal continuity in the environment to improve behavioral efficiency. Brain regions have been identified to carry historical information, but how such information is stably maintained, flexibly updated, and transformed into future behavior remains unclear at the whole-brain level.

A study published in Nature and conducted by MU Yu's team from the Center for Excellence in Brain Science and Intelligence Technology (CEBSIT) of the Chinese Academy of Sciences and WU Si's team from Peking University, using brain-wide single-cell-resolution calcium imaging in larval zebrafish, closed-loop virtual-reality behavior, optogenetic manipulation, and neural computational modeling, revealed how the brain coordinates multiple computational modules to maintain recent experience and flexibly guide future decisions.

The researchers established a closed-loop virtual-reality obstacle-avoidance system for larval zebrafish. They found that avoidance responses to a current obstacle were significantly influenced by prior experience: When two consecutive obstacles appeared on the same side, zebrafish showed stronger avoidance responses. These findings demonstrate that zebrafish can retain recent experience over tens of seconds and use it to regulate subsequent behavior.

A key technical strength of this study is the ability to record brain-wide neural activity at single-cell resolution and precisely register these signals to a standard zebrafish whole-brain atlas established by CEBSIT researchers. Brain-wide recording combined with atlas-based registration enabled the researchers to trace, within an intact brain network, the full computational pathway from sensory input, memory maintenance, and cross-regional integration to behavioral output.

Systematic screening across brain regions identified the dorsal thalamus as the area most stably and persistently encoding the side of the most recent obstacle through persistent activity. Optogenetic experiments confirmed its causal role: Inhibiting dorsal thalamic activity abolished history-dependent behavior, while unilateral activation could artificially instill a "past experience" signal, altering the animal's next choice.

Further analysis showed that the dorsal thalamus stored the most recent experience as a stable discrete state, and downstream neuronal populations in the brainstem combined thalamic history signals with current sensory input to generate a continuous signal reflecting multiple recent experiences and ultimately driving behavioral output. Through this cross-regional division of labor, the brain transformed a transient sensory event into an internal state that can be sustained, updated, and used to guide future actions.

Moreover, the researchers constructed a brain-atlas-constrained whole-brain computational model incorporating real biological information, including cell numbers, neuronal types, and projection patterns, from CEBSIT atlas. The model connected a sensory input layer, a thalamic attractor network, and a brainstem integrator, reproducing both sequential decision behavior and associated neural dynamics. The findings reveal the importance of inhibitory-neuron heterogeneity for stable memory maintenance and flexible state switching.

This study proposes a whole-brain attractor-integrator architecture in which stable memory and flexible updating are jointly achieved through division of labor across brain regions, neuronal types, and dynamical modules. The researchers noted that the whole-brain atlas served not merely as an anatomical resource but as an important bridge connecting real brain structure and computational principles.