Is scientific progress waning? Too many new papers may mean novel ideas rarely rack up citations

There are so many papers coming out in the largest fields of science that new ideas can’t get a foothold, according to a recent study in PNAS. Researchers analyzed citation trends for 90 million papers and found that in very large fields such as molecular biology or electrical engineering, older canonical works tend to be the most frequently cited for decades, while a deluge of newer papers, with potentially groundbreaking ideas, rarely become highly cited.

“We’ve set up a system now where new ideas can’t unseat existing dominant ideas,” says lead author Johan Chu, a strategy and organizations scholar at Northwestern University in Evanston, Illinois. “We are inundated with choices,” he says, but “more is not better.”

In a series of studies over the last few years on competition and consumer markets—scrutinizing everything from mutual funds to Korean pop songs—Chu developed a theory he calls “durable dominance.” The theory posits that consumers with too many choices will tend to become overwhelmed and react by choosing a familiar option. Second, it holds that when many of the options on the market are newcomers, the newbies compete so intensely amongst themselves that they can’t compete with the established dominant choice—in the case of research, that often means papers that had become the popular option before the market was so competitive.

Thanks to the explosion of new research papers in the last few decades, Chu thought the scientific literature was “a perfect testbed” for his durable dominance theory. So he and his coauthor James Evans analyzed 1.8 billion citations from 90 million papers, across 241 subjects on Web of Science. They found that the most popular subjects on the search database, such as molecular biology, grow by more than 100,000 new papers every year. And in these mega-fields, the top 1% of papers get the vast majority of citations. What’s more, even the rankings of these highly cited papers rarely change; the top-cited article in molecular biology came out in 1976 and has been number one since 1982.

Chu and Evans also found that the odds of a paper reaching the top 0.1% of highly cited studies in a given year declines precipitously with field size. When the rare paper does break through, it usually does so in less than 12 months, suggesting that popularity comes from social media, news coverage, or via existing networks of people who are already well-connected in the subject area—rather than from citations in other work. Even iterative papers that seem to be written with the intent to eclipse older studies are not being cited in the largest fields, Chu says, based on his analysis of how effectively such papers introduce new ideas that disrupt the existing literature. The results imply that even if every new study was groundbreaking, there are just too many new papers now for readers to spend meaningful time on them.

But should we even expect that there is some healthy rate at which new papers overturn the canon? Many researchers assume turnover drives progress across fields, but that’s an assumption worth testing, says sociologist Russell Funk at the University of Minnesota in Minneapolis. Funk, who was not involved in this study—and calls it a “really cool, well done paper”—expects that replacing canon does indeed help fields move forward. But it’s worth asking by what measure, he says. And Funk wonders whether a stagnant canon can be useful in some cases—for example, in interdisciplinary fields where researchers specialize in different subjects such that shared touchstone papers provide common ground for nascent ideas. Chu isn’t sure what an optimum turnover rate would be—he didn’t test it in this study—but says he’d expect it to vary by field size, as well as, among other factors, field maturity and the timescales involved in new research studies.

Several recent studies do hint that innovation is slowing down in research, Funk notes. “The advancement of scientific knowledge is kind of becoming a victim of its own success.” If that is the case, solutions seem hard to come by. Perhaps graduate students and their advisors could focus on non-canonical experiments. But even so, those potentially groundbreaking papers would be harder to publish and less well-read, Chu notes, creating a catch-22 for any junior researcher trying to secure an academic post. One useful next step, he suggests, would be to investigate different journals’ review processes, to see if reviewers push the authors of new papers to cite older canon. This could reveal one way that older papers are staying entrenched.

Ironically, the theory of durable dominance had been kicking around the literature for a few years, itself an example of a new idea that wasn’t unseating older works—that is, perhaps, until Chu’s PNAS paper went viral on Twitter, largely courtesy of scientists. As of this writing, it’s been tweeted more than 1,800 times.

Source