Summary: Scientists have produced the first brain-wide, single-cell map of decision-making in a mammal by recording activity from more than 600,000 neurons across 279 regions in mice. Using a standardized behavioral task and high-density Neuropixels recordings across an international consortium, the study shows that decision-related signals are far more widely distributed than previously recognized and that regions linked to movement also carry cognitive information.
This open, large-scale dataset challenges older, region-centered views of decision-making and provides a practical foundation for testing new theories about how distributed neural circuits guide choices and behavior.
Key Facts
- Massive dataset: 621,733 neurons recorded from 139 mice across multiple laboratories using 699 Neuropixels probes.
- Distributed signaling: Decision-related activity appeared in classical cognitive areas as well as in motor and sensory regions.
- Global collaboration: Data were collected and standardized across dozens of labs through the International Brain Laboratory, creating a shared resource for the field.
Source: Princeton
Mice turning tiny steering wheels to move shapes on a screen enabled researchers to build the first brain-wide, single-cell resolution map of decision-making in a mammal.
For decades, neuroscience experiments typically targeted small groups of neurons in isolated brain regions. While that approach yielded important discoveries, it misses how decisions emerge from interactions across many regions.
“The brain is constantly making decisions during everyday life, and we’ve come to realize that many brain regions, not just one or two, contribute to those decisions,” said Ilana Witten, Ph.D., professor of neuroscience at Princeton University and an investigator with the Howard Hughes Medical Institute.
Witten and colleagues organized a multi-lab effort within the International Brain Laboratory (IBL), a consortium that aligned experimental protocols, quality-control standards, and data-analysis pipelines so recordings from many labs could be combined into a single, comparable dataset.
Published in two companion papers in Nature, the work reports recordings from more than 600,000 neurons across 279 brain regions in 139 mice. By pooling standardized experiments and analyses, the team produced an unprecedented view of how distributed neural populations encode sensory stimuli, prior expectations, motor actions, and rewards during a single decision task.
A collaborative leap forward for neuroscience
Mapping an entire behaving brain at cellular resolution is technically and organizationally demanding. Alejandro Pan Vazquez, Ph.D., an associate research scholar in the Witten lab and a contributing author, emphasized the novelty of coordinating so many labs around a single experimental framework.
Three Princeton labs—led by Ilana Witten, Tatiana Engel, Ph.D., and Jonathan Pillow, Ph.D.—helped coordinate the collaboration by designing experimental parameters, implementing quality-control metrics, collecting data, and building standardized analysis pipelines so the diverse recordings could be merged into a coherent resource.
Steering wheels and flashing circles
The behavioral task was straightforward but informative. Mice sat before a screen where a black-and-white striped circle briefly appeared on the left or right. By turning a small steering wheel in the matching direction, a mouse could bring the circle to the center and earn a small reward. On difficult trials the stimulus was faint, forcing animals to rely on past experience and revealing how prior expectations shape choices.
While mice performed this task, researchers recorded brain activity using high-density Neuropixels probes that simultaneously sampled hundreds of neurons across many regions. Each participating lab focused on mapping particular areas, and the pooled effort resulted in recordings from 621,733 neurons across 279 regions in the forebrain, midbrain, hindbrain, and cerebellum.
Surprising and important findings
Analyses revealed that representations of sensory stimuli and decision variables are widely distributed. Visual responses first appeared in classical visual areas and then spread into midbrain and hindbrain regions where ramping activity reflected impending choices. Neural activity related to movement and anticipated actions was detectable across much of the brain, and responses tied to reward delivery and consumption were similarly widespread.
A complementary analysis showed that mice form and use a subjective prior about stimulus location. That prior was encoded across 20–30% of brain regions, spanning early sensory nuclei, diverse cortical areas, and motor regions. These findings support models in which prior information is represented broadly and maintained through loops between areas, rather than being confined to a single decision structure.
The publications highlight both the scientific conclusions and the dataset itself: a public, standardized resource intended to enable further analyses and new discoveries across laboratories worldwide.
A map made across the world
Beyond scientific insight, the project demonstrates the value of large-scale, coordinated neuroscience. “The brain-wide map is undoubtedly an impressive achievement, but it marks a beginning, not the grand finale,” said Tatiana Engel. The IBL’s effort shows how coordinated teams can reach territory no single lab could explore alone.
Funding: Research support came from the National Institutes of Health, National Science Foundation, Gatsby Charitable Foundation, Wellcome Trust, Simons Foundation, Max Planck Society, and the Humboldt Foundation.
About this brain mapping and decision making research news
Author: Daniel Vahaba
Source: Princeton
Contact: Daniel Vahaba – Princeton
Image: The image is credited to Neuroscience News
Original Research: Open access. “A brain-wide map of neural activity during complex behaviour” by Tatiana A. Engel et al. (Nature)
Open access. “Brain-wide representations of prior information in mouse decision-making” by Tatiana A. Engel et al. (Nature)
Abstract
A brain-wide map of neural activity during complex behaviour
Understanding how neurons across hundreds of interconnected regions combine sensory inputs and prior expectations to drive movements and decisions is a central challenge in neuroscience. To address this, the International Brain Laboratory collected a comprehensive set of Neuropixels recordings: 621,733 neurons recorded with 699 probes across 139 mice in 12 laboratories performing a standardized decision-making task.
The probes sampled 279 brain areas in the left forebrain and midbrain and the right hindbrain and cerebellum. Initial analyses show visual stimulus representations appearing in classical visual regions and then spreading into midbrain and hindbrain areas where activity ramped with choice. Signals related to upcoming motor actions were present throughout much of the brain, and reward-related responses were also widespread. This publicly available dataset is a resource for investigating how distributed computations across brain areas drive behavior.
Abstract
Brain-wide representations of prior information in mouse decision-making
To study how prior information is represented across the brain, researchers analyzed the IBL’s Neuropixels and widefield calcium imaging datasets while mice performed a task in which the probability of a visual stimulus appearing on one side or the other varied in blocks. Mice adapted their choices according to these priors and improved decision accuracy.
The subjective prior was encoded in a broad set of regions—roughly 20–30%—spanning early sensory nuclei, motor areas, and higher-order cortical regions. These results are consistent with a model in which prior information is maintained across loops of interconnected areas rather than being localized to a single decision-making center, underscoring the value of large-scale, standardized recordings for revealing brain-wide computations.