The Champalimaud Foundation (CF) played an important role in this achievement. Not only was CF’s Systems Neuroscience Lab one of the 12 experimental labs where the large-scale recordings were carried out, but CF also hosted much of the engineering team that designed and maintained the infrastructure that made such an unprecedented dataset possible.
Science on a New Scale
“Traditionally, neuroscientists have studied brain regions one at a time”, says Zachary Mainen, Principal Investigator at CF and one of the founding members of IBL. “Here, we dramatically upped the ante. Across 12 labs in Europe and the US, we recorded from more than half a million neurons covering 279 brain areas, representing 95% of the mouse brain volume. Unlike fMRI, where each pixel may reflect the pooled activity of around a million neurons, our measurements captured the activity of single neurons at the level of individual spikes. That gives us a far more fine-grained picture of the brain and sets a new technical standard for the field”.
Launching in 2017 with support from Wellcome and the Simons Foundation, IBL set out to do something no single lab could achieve alone: measure the brain in action, not in fragments but as a whole. Coordinating 12 experimental sites on two continents was a logistical and scientific challenge of the highest order.
“A lot of the effort went into figuring out how to get different labs to literally do the same experiment”, Mainen explains. “Most collaborations divide tasks across labs, but here every site had to run identical protocols. That required building an entire infrastructure – hardware, software, training routines – so that you couldn’t tell whether a dataset came from Lisbon, London, or New York. That level of reproducibility was absolutely essential”.
The Lisbon Hub
At Champalimaud, IBL engineers and researchers were deeply involved in developing the nuts and bolts of the project. CF was one of the pilot sites where experimental rigs and training protocols were prototyped, tested, and refined.
Olivier Winter, Senior Scientific Programmer at IBL based at CF, recalls the scale of the challenge: “We had to ensure that a mouse in Lisbon was performing exactly the same task as a mouse in Cold Spring Harbor or Berkeley. That meant standardising everything – the rig design, the software, the data pipelines. Our engineers were crucial in making sure the data could be pre-processed, curated, and made available to the public in a reliable way”.
By splitting the experimental schedule across multiple labs, the team was able to build a dataset far beyond the capacity of any individual group, while also stress-testing the reproducibility of neuroscience on a global scale.
How Do Mice Decide?
At the heart of the project was a deceptively simple task. A mouse sits in front of a screen. An image flashes on the left or right. The animal turns a small wheel to indicate the direction and receives a drop of water as a reward for correct answers.
On some trials, however, the light is so faint that the mouse must guess. To make these guesses, mice rely on prior experience: if the light appeared on the left more often before, they lean left. This allowed researchers to study not only perception and action, but also how the brain uses internal expectations to guide choices.
Guido Meijer, now a postdoctoral researcher at the Donders Institute and formerly at CF, was among the first to join IBL. He performed neural recordings with state-of-the-art Neuropixels probes, which measure the activity of individual neurons across the brain. “At my first collaboration-wide meeting, we hadn’t even decided on which task to use”, he recalls. “But the goal was clear: to record the entire mouse brain at the cellular level during decision-making. Together, we designed and built the rigs, developed the behavioural training, and made sure it all worked reproducibly across every lab. I’m incredibly proud that we achieved our goal: to record the entire mouse brain during decision-making and release this massive dataset to the public”.
What the Brain Revealed
The first paper showed that decision-related signals are not confined to specialised regions but spread widely across the brain. Visual areas lit up when the stimulus appeared, motor regions when the mouse moved, and reward areas when the animal succeeded. But critically, signals linked to choice and reward appeared nearly everywhere, pointing to continuous communication across the brain.
The second paper found that the animals’ expectations – their internal guesses about what was likely to happen – were also represented across the brain. These expectation signals appeared not only in “high-level” cognitive areas but also in sensory and motor regions, even the thalamus, a gateway for visual input.
As Meijer puts it: “It’s not that information funnels into a small group of decision-making regions that dictate behaviour. Instead, information is highly distributed. Behaviour seems to emerge from the collective activity of the entire brain”.
Implications and Future Directions
The findings support the idea that the brain acts as a prediction machine, constantly updating its model of the world to guide behaviour. They also have potential relevance for psychiatric conditions such as autism and schizophrenia, which are thought to involve disruptions in how expectations are formed and updated.
For Mainen, the achievement is about more than scientific results: “Not every problem demands an effort on this scale. Individual labs will always be engines of discovery. But sometimes, to really move forward as a field, we need to join forces. If we’re not confident our data is reliable and reproducible, progress stalls. Other fields, like spatial navigation, managed this with simpler behaviours. But decision-making is much more complex, and that made the challenge greater. Projects like this show how neuroscience can rise to that challenge”.
Looking ahead, IBL plans to broaden its scope beyond decision-making, while continuing to share tools, datasets, and protocols openly with the scientific community. For CF, the project underscores the value of tackling scientific challenges collectively. “This was too big for any one lab”, Mainen reflects. “But by working together, we’ve created a resource the entire world can use”.
All data, tools, and protocols from this study are freely available via the International Brain Laboratory website.
Read the full papers in Nature:
- A brain-wide map of neural activity during complex behaviour
- Brain-wide representations of prior information in mouse decision-making
