In 1905, Albert Einstein published four papers that remade physics. Three of them had a single author. One of them, his doctoral dissertation, had no author at all. His famous 1915 paper on general relativity? One author. When Einstein won the Nobel Prize in 1921, the idea that a lone scientist could reshape the foundations of human knowledge was still the default in every laboratory on Earth.
A century later, that world has vanished.
A new analysis of the Nature Index, published July 23, 2026 in Nature, reveals just how thoroughly science has transformed itself. Between 2015 and 2024, the share of one- and two-author papers in the natural sciences fell from 10.8 percent to 7.1 percent. Papers with 11 to 50 authors are now the fastest-growing category. The share of papers with 3 to 50 authors climbed from 88.5 percent to 92.2 percent.
What we are witnessing is not merely a change in how scientists collaborate. It is the industrial revolution of knowledge production itself.
The rise of the scientific factory
For most of the 20th century, science operated like a system of cottage workshops. A principal investigator, a graduate student or two, maybe a postdoc. They worked in the same building, argued over the same data, wrote the paper together. The scale of the questions matched the scale of the team.
That model has been steadily replaced by something closer to a factory floor. Modern particle physics has long been the caricature of this trend: papers with thousands of authors, the byproduct of billion-dollar machines that no single university could host. But the Nature Index data shows that the factory model has now colonized every branch of the natural sciences.
The breakdown by discipline in 2024 is stark. In the biological sciences, only 3.8 percent of papers were one- or two-author works. In chemistry, the figure was 4.7 percent. Even in the physical sciences, where small-team traditions run deepest, the share was 9.6 percent.
The data by country complicates the narrative further. The United States saw its small-team share drop from 9.8 percent in 2015 to 6.0 percent in 2024. Germany and the United Kingdom fell from roughly 7 percent to roughly 4 percent. China, a nation that built much of its modern scientific capacity through large, state-coordinated programs, went from 3.5 percent to 2.4 percent.
But Japan is different. Japan’s share of small-team papers has stayed stable at around 8 percent, barely budging while every other major research nation marched toward mega-collaboration.
The Japan anomaly and what it reveals
Why has Japan resisted a global trend? Three factors appear to be at work, according to Yukie Sano of the University of Tsukuba, who studies the structure of scientific collaboration.
First, Japan’s weak yen has reduced the ability of Japanese researchers to travel abroad, inadvertently limiting the cross-border partnerships that drive team expansion. Second, a shift in Japanese funding policy away from stable institutional support toward competitive grants has created a heavier administrative burden on senior researchers, leaving them less time to coordinate sprawling international consortia. Third, cultural factors: Japanese laboratories have traditionally maintained smaller, more hierarchical structures, and that norm has proved resistant to change.
Japan is an instructive outlier precisely because it shows that the shift toward large teams is not an inevitable law of scientific progress. It is the product of specific incentives embedded in how science is funded, organized, and evaluated.
The machinery of incentives
Lingfei Wu of the University of Pittsburgh, who studies the dynamics of scientific collaboration, argues that the modern funding system has become an engine for team growth. The post-World War II model of centralized funding agencies such as the National Institutes of Health and the National Science Foundation created a structure where larger, more expensive projects attracted larger, more expensive grants. The logic was simple: a team of 30 scientists working on a coordinated project can produce more papers, in less time, than 30 scientists working independently.
But the incentives do not stop at scale. Interdisciplinary requirements, such as those embedded in European Union research grants, mandate cross-border collaboration. Grant applications now demand data management plans, ethics approvals, diversity statements, public engagement strategies. The administrative burden of modern science has become so heavy that individual researchers can no longer carry it alone. A team of 10 or 20 or 50 is necessary just to navigate the bureaucracy.
Chaoqun Ni of the University of Wisconsin-Madison, who has tracked authorship patterns across millions of papers, puts it plainly: the infrastructure of modern science demands collaboration. Shared telescopes, shared particle accelerators, shared biorepositories, shared computing clusters. These facilities do not belong to one lab. They belong to a community, and the community writes the papers together.
The paradox of disruption
And yet something curious happens when researchers look at which papers actually change the course of science.
A series of studies published in Nature in 2019, 2023, and 2025 found that small teams produce a disproportionately large share of “disruptive” science — work that opens new fields rather than refining existing ones. Small teams are more likely to cite older, less fashionable papers. They are more likely to combine ideas in ways that have not been combined before. They are more likely to be wrong, but also more likely to be profoundly right.
Large teams, by contrast, excel at consolidation. They produce reliable, incremental results. They confirm findings, extend them across populations, and build the slow architecture of settled knowledge. A 50-author clinical trial is not designed to overturn a paradigm. It is designed to produce a definitive answer to a narrow question.
Both modes are essential. Science needs the factory to build the sturdy infrastructure of verified results. And it needs the cottage workshop to produce the conceptual leaps that give the factory something to build.
The problem is that the funding system recognizes only the factory.
What the numbers miss
The Nature Index data counts authors. It does not count the intellectual distance between the first idea and the final paper. It does not count the moments when a single researcher, sitting alone in an office, realizes that the entire field has been asking the wrong question. It cannot measure disruption.
The shift from 10.8 percent to 7.1 percent in small-team papers over nine years is not, by itself, catastrophic. But it represents a trajectory. If current trends continue, the one- and two-author paper will become a scientific curiosity within a generation — a historical artifact, like Einstein’s solitary manuscripts, that tells future scientists about a world they can no longer inhabit.
Japan’s stability at 8 percent suggests there is no natural floor to this decline. The floor is set by policy. And policy can be changed.
The question facing research agencies, university administrators, and funding bodies is not whether the factory model is useful. It is. The question is whether the funding system has become so biased toward the factory that it is starving the workshops that produce the ideas the factory needs to process.
A scientific system that optimizes exclusively for reliable incremental results may produce more papers, more citations, more measurable output. It may even produce more knowledge. But it may also produce less understanding. And understanding, not output, is what science has always been for.
The funding agencies that reshaped science after World War II built a magnificent machine. It has delivered treatments, technologies, and data on a scale that Einstein could not have imagined. But machines have blind spots. They optimize for what they can measure. They starve what they cannot.
Japan’s anomaly proves that another path is possible. A deliberate funding strategy that preserves space for the small team, the heterodox idea, the unfashionable question. Not as a nostalgic gesture toward a romanticized past, but as a recognition that the factory and the workshop need each other.
The data are clear. The trajectory is set. The question is whether we have the wisdom to steer it.