Science as we know it has been centered around individual labs: a principal investigator, a small team and a focused question. But today’s biggest scientific challenges require scale, coordination and a level of technical sophistication that no single lab can provide.

The International Brain Laboratory (IBL) was built to address that challenge. The Simons Foundation-supported endeavor began with a shared recognition among researchers that their central question, how the brain makes decisions, required a fundamentally different approach.

What followed was not just a scientific collaboration, but an experiment in how science itself can be organized. Recently, the IBL published a paper in Neuron outlining key lessons in conducting team science, revealing that when science scales up, everything changes, and the status quo must evolve.

A More Democratic Lab

It is tempting to think of large collaborations as simply bigger versions of individual labs. However, once dozens of labs, roles and priorities are involved, the challenge is no longer purely scientific — it’s organizational.

From the outset, the IBL tried to rethink how a scientific team should operate. Instead of a traditional hierarchy, it adopted a more collective governance model called sociocracy. This model uses structured, consent-based decision-making to ensure all members have a voice while maintaining effective and accountable leadership.

“The idea is that any stakeholder, anybody who’s in the organization, can have an idea for a proposal,” says Hannah Bayer, senior scientific officer of neuroscience collaborations at the Simons Foundation and former executive director of the IBL. “Good ideas come from anywhere, not just from the top.”

This governance model continued to evolve over time, and the IBL added additional measures to elevate the voices of everyone in the organization, including anonymous surveys and dedicated roles for junior members.

“We created roles such as staff and researcher representatives, each with voting rights — something quite different from the original structure,” says Gaëlle Chapuis, director of operations and outreach for the IBL. “We also tried to ensure that the working groups included people from all career stages, which helped foster more vertical discussion.”

While the IBL has demonstrated many ways to promote participation across all levels, there is still work to do to achieve true equity.

“It’s challenging to create a form of governance where everyone’s voice is truly heard,” says Anne Churchland, a professor of neurobiology at the University of California, Los Angeles and a founding member of the IBL. “We made some positive steps in that direction and chose a strong model, but more work is needed to ensure broad participation and to prevent a few loud, strongly opinionated voices from overshadowing others. There’s still a lot to learn.”

The Credit Quandary

Operating in a collaboration this large also raises concerns about credit and recognition. Traditional authorship models — first author, last author — don’t map neatly onto collaborations where dozens of people contribute in different ways. The IBL experimented with consortium-wide authorship alongside more focused individual projects, where graduate students and postdocs could still publish first-author papers.

“I think there needs to be greater recognition for what trainees gain from participating in team science,” says Bayer. “For many, first-author papers are still essential for success on the job market: That remains non-negotiable. So we worked to create those opportunities. But I felt like it was such a loss that there wasn’t as much recognition of the team‑science work.”

There are other options for assigning credit, such as contribution matrices that better map out each person’s responsibilities, but these aren’t yet the norm and can be easily overlooked.

“This contribution matrix that we include in every paper needs to be what people actually focus on rather than the author list itself,” says Dan Birman, a software engineer at the Allen Institute and a former postdoc in the IBL. “In hiring contexts, the first step should be to look at what each person did, not just scan the authorship order and make assumptions — like seeing someone in the middle of the list and not knowing their role. The information is already there; it just needs to be surfaced more prominently.”

The Engineers Behind the Science

Within its organizational ecosystem, the IBL also identified a critical role for a project this large to succeed: research software engineers (RSEs). RSEs work alongside scientists in everything from data scrutiny to tool development to bug reporting and documentation.

“Research software engineers are essential in these large-scale neuroscience projects because they really enable the project processes to be standardized, scalable and compatible across laboratories, and they contribute across every stage of the project,” says Chapuis. “There is a lot of value in specialization. For team science to work, having dedicated experts like the RSEs is necessary.”

The RSEs started as a group of three but have since tripled in number. These RSEs are solely dedicated to maintaining the IBL’s robust infrastructure and scaling up its tools, making them usable across many labs.

“Writing code that works across labs and researchers is far more complex than the typical approach in neuroscience, where individuals develop their own scripts,” says Bayer. “That model simply doesn’t scale to 10 or 11 labs collecting data. What’s needed instead are robust, standardized pipelines that can function reliably across labs and accommodate researchers with varying levels of experience, and that’s what the RSEs helped us build.”

Small Errors, Big Problems

The IBL quickly realized that the bedrock of the collaboration had to be standardization, and ensuring every experiment and analysis works in exactly the same way requires extensive quality control.

In smaller labs, quality control can tend to be informal. Issues are caught through experience, intuition or close supervision. But in a distributed collaboration, that doesn’t work. The IBL found that ensuring reliable data required validation at every stage, from hardware setup to final analysis. If mistakes go unnoticed, they can quickly proliferate, setting projects back by weeks.

To address this, the IBL built in dozens of automated inspections to flag when something could be amiss. However, human review is also crucial. In one case, a visual inspection revealed a subtle timing bug (every second, a few milliseconds disappeared from the data) that had gone unnoticed by automated systems.

“It really reinforced for us that the earlier you can share raw data — ideally showing it to the experimentalist as it’s being recorded, or immediately after acquisition and initial analysis — the more likely you are to catch major errors early,” says Birman.

Another key insight from a collaboration this large was that widely used tools cannot be assumed to work without testing.

“Software that works in one context may not translate well to another,” says Chapuis. “We found this when we applied tools as black boxes to the IBL’s data: The tools didn’t work on all the recordings. Those failures were visible only when running algorithms at a large scale. When that happens, we have to build software through a new approach: by building benchmarks encompassing the variety that the tools have to meet, and by designing modular algorithms whose performance can be assessed block-by-block.”

Impact Beyond the IBL

After years of rigorous testing and determining which standardization practices work and which don’t, the IBL has taken on a central role in defining guidelines for the field.

“We realized we had an opportunity to define a set of standards at a moment when they were urgently needed,” says Churchland. “The resulting framework is not prescriptive but intended as a starting point — a way for researchers to ask themselves whether they are truly generating high-quality results.”

This is just one aspect of the IBL that has become widespread outside of the collaboration itself.

“The technical developments we built, including the rigs, database and data processing pipelines, are now being reused beyond the IBL labs,” says Chapuis. “The data is also being used in a lot of teaching settings.”

“I hope the increased use of IBL tools demonstrates to other groups and funders how important those skills and kinds of contributions are,” Bayer adds.

The Future of Team Science

The IBL offers a glimpse of the future of science, and it is both exciting and challenging. Large-scale collaboration enables tackling problems that were once out of reach.

“I think the IBL showed that you can do large-scale, reproducible and effective neuroscience across institutions all around the world,” says Birman. “I really think that in five years, every project will be as large as the IBL, simply because the tools are improving and our ability to collect large-scale datasets keeps advancing.”

As the IBL enters its next phase, it will be sharing its resources with the broader neuroscience community, shifting from a single large collaboration to a central team that supports multiple projects and partners. The team recently announced its first round of new partners and projects for this “IBL 2.0.”

“I think this new technical core reflects a shift in how we support neuroscience research. The goal is to help individual labs by bringing in specialized expertise tailored to their specific project needs,” says Chapuis. “In that sense, it’s about professionalizing labs and helping them reach a higher standard. We’re not imposing ready-made solutions or requiring labs to use our rigs. Instead, we partner with them — offering guidance on best practices and, when needed, implementing new techniques. That approach is quite different from traditional cores.”