In 2023, INSEAD Assistant Professor of Organisational Behaviour Michael Park, together with Erin Leahey of the University of Arizona and Russell J. Funk of the University of Minnesota, published a landmark study in Nature showing that science and technology are becoming less disruptive. The paper became one of the 100 most-discussed scientific papers of that year, reported by more than 300 media outlets worldwide and informing policy discussions from United States Congressional hearings to White House speeches.
Their central finding: Papers and patents are increasingly less likely to break with the past in ways that push science and technology in new directions. Breakthroughs still happen, but they make up a shrinking share of rapidly growing scientific literature. This shift carries implications for how research is funded and organised.
But the finding has now been challenged by researchers at Vrije Universiteit Brussel and other critics. In a comment published in Nature on Aug 12, the critics argued that the measured decline partly reflects database entries with no listed references—artefacts of the data, they suggested, rather than a real change in science.
Replying in Nature the same day, Park, Leahey and Funk published a peer-reviewed analysis that put their own research to the test, using the critics’ own model. The decline remains, and it’s large and statistically significant.
"The strongest test you can give a finding is your critics’ own tools," says Park. "We reran the analysis using their dataset, their metric, their exclusion criteria and their statistical model. The decline is still there and it is large, equal to the gap between an average paper and one by a Nobel laureate."
The reply by Park and his coauthors also documents severe quality problems in the critics’ own dataset, problems that magnified the very bias the critics set out to correct. One in five works in their sample is not research at all: millions of editorials, obituaries, comments, book reviews and children’s books, which cite nothing because of what they are, not because a database recorded them incorrectly. Standard practice in the field, followed in the 2023 study, excludes such content for exactly this reason; the critics’ preprocessing did not.
Meanwhile, the 2023 finding no longer stands alone. Evidence of declining disruptiveness has now been documented in more than 100 independent studies spanning different databases, different metrics and measures that use no citation data at all.
"We have tested this finding across databases, across measures and across methods," says Funk. "It holds every time. The interesting question now is not whether disruptiveness is declining but why, and what that means for how we organise science."
