Physicist Albert Einstein, widely regarded as one of the most prolific scientists of the past century, conducted much of his transformative work at the beginning of his career, before spending years defending his theories against the burgeoning field of quantum mechanics.
A new study shows that Einstein is not alone, and that most researchers begin their careers conducting their more disruptive work — overturning conventional wisdom and forging paths of their own — but as they age, they tend to abandon that groundbreaking energy. Instead, many become adept at connecting previously unlinked ideas. The paper, published Thursday in Science, helps offer an explanation of a trend that has increasingly worried scholars of science policy and innovation: that the pace of discovery has slowed in recent years.
“Decline in disruptive science is one of the most robust patterns in the science-of-science literature at this point. … What’s been harder to pin down is why,” said Russell Funk, a sociologist who first identified this trend, and has compiled over 100 studies that bear out the decline but was not affiliated with the new paper. “We pointed to a narrowing in how scientists use prior knowledge — they’re reading less and building on a smaller slice of what’s available. Other researchers have pointed to the growing burden of knowledge, the expansion of team sizes, and changes in funding incentives. This paper adds workforce aging to that list, and I think it belongs near the top.”
The new analysis reviewed the work of 12.5 million scientists who published at least three papers between 1960 and 2020, and tracked the ways those papers cited previous work and were then cited going forward. It found that as scientists age, they cite older and older work. In the field of medicine, the paper found that at the beginning of their careers, scientists were citing work that was, on average, 7.9 years old. By the end of their careers, 40 years later, that average was about 10.1 years old. This could be a sign that older researchers pursue lines of research they are more familiar with, while younger researchers are more in tune with fresh ideas.
“There is this very asymmetric difference in incentives; old scientists have the most to lose and the least to gain from new ideas, and young scientists have the most to gain and the least to lose,” said James Evans, a sociologist of science at the University of Chicago, and one of the authors of the new paper.
The effect of aging researchers is not limited to their papers. The study found that papers published in draft form as preprints cite more recent work in their preprint version when compared to the peer-reviewed version subsequently published in a journal. This effect was more pronounced in older fields — a sign that older peer reviewers could be influencing the work being referenced.
The study also found that within research teams, papers led by younger corresponding authors — who are usually listed as the last author on a paper and responsible for communication about the study — cited more recent work. This may mean that empowering younger researchers, or giving them their own funding — unmoored from more senior colleagues — could help spur more innovative work.
The change in individual practices is also having a larger effect: The paper found that countries with younger scientific workforces such as China and India are producing more disruptive work than countries like the U.S. and the United Kingdom, with more older researchers.
One limitation of the study was its use of citations as a metric of the disruptiveness of a paper, said Mikko Packalen, an economist at the University of Waterloo who has studied the impact of aging on biomedical research but was not affiliated with the new research.
“This paper is somewhat emblematic of the very dynamics it describes,” he said. “Much of the argument rests on citation-based measures. But while citations are useful for the study of the sociology of science, they have very limited potential as indicators of the actual content, direction, or originality of science.” He said he was disappointed the authors didn’t take advantage of recent advances in AI-based text analysis to examine the substance of scientific papers, adding, “The fact that the paper’s own citations are to a very narrow set of papers from select elite scientists further underscores this methodological point.”
Indeed, the new study found that social science papers written at the end of a researcher’s career cited papers that on average were 13 years old. In comparison, the average age of the papers cited by the authors was around 17 years.
