The Costs of Publish or Perish

Study suggests that publication pressures lead to more familiar -- and more likely to be published -- research, at the expense of innovation.

October 12, 2015

Shortly after being awarded the Nobel Prize in Physics in 2013, Peter Higgs, of Higgs boson fame, said he doubted he would have gotten a job, not to mention tenure, in today’s academic system. The professor emeritus at the University of Edinburgh said he simply wouldn’t have been “productive” enough, with academe’s premium on publication metrics. Conversely, said Higgs, working in today’s academic system probably wouldn’t have afforded him the opportunity to identify how subatomic material requires mass.

“It's difficult to imagine how I would ever have enough peace and quiet in the present sort of climate to do what I did in 1964,” he told The Guardian.

The statement resonated with many academic scientists running the funding-collaboration-publication treadmill. But while the negative consequences of the “publish or perish” paradigm, such as innovation costs and decreased attention to teaching and mentoring, are widely acknowledged, there’s been scant data to back them up. So a new study suggesting that publication pressures on scientists lead to more traditional, more likely to be published papers, at the expense of scientific breakthroughs, stands out.

“Pursuing innovation is a gamble, without enough payoff, on average, to justify the risk,” the study says. “Nevertheless, science benefits when individuals overcome the dispositions that orient them toward established islands of knowledge … in the expanding ocean of possible topics.”

The study, called “Tradition and Innovation in Scientists’ Research Strategies,” is in the current American Sociological Review. To begin, Jacob B. Foster, lead author and professor of sociology at the University of California at Los Angeles, and his co-authors created a database of more than 6.4 million biomedical and chemistry publications from 1934 to 2008.

They used chemical annotations from the National Library of Medicine to build a computer-modeled network of knowledge, and looked for chemicals that were linked, showing up in the same paper. They then sorted the links into two broad categories: those that built on past knowledge and those that were truly innovative, adding connections to the network.

The researchers looked at how many of each type of link appeared in a given year, and made inferences about scientists’ disposition to pursue tradition over innovation. This link classification allowed Foster and his team to classify papers to determine, via various regression analyses, whether papers with more innovative strategies were more frequently cited.

Finally, they built a database linking winners of some 137 major scholarly awards to their publications, to compare the mixture of links used by scientists with major achievement to the publication pool more generally.

Essentially, Foster and his co-authors created a map of which individual publications built on existing discoveries or created new connections. Then they correlated each of the research strategies to two different kinds of recognition -- citations and major awards.

Perhaps unsurprising, the work of prize-winning scientists involved significantly more innovation than the overall pool. And more than 60 percent of publications generally had no new connections, building on traditional research alone.

Foster and his co-authors, James Evans, an associate professor of sociology at the University of Chicago, and Andrey Rzhetsky, a professor of medicine and human genetics at Chicago, argue that researchers who focus on answering established questions are more likely to see their work published. But while researchers who pursue riskier academic work may not be published as frequently, if published, their work receives more citations.

Foster said in an email interview that what makes his study “distinctive is the scale.”

“We were able to study this tension at scale because of several intersecting trends: increasing availability of computer-readable information about science and scientific publications, increasing computer power, and the development of network-driven techniques for representing and analyzing knowledge,” he said. “It is this last development that allowed us to operationalize tradition and innovation in a reasonable way for large-scale analysis.”

The authors recommend various ways that colleges and universities can promote more innovation, such as not linking job security to productivity, in terms of easy metrics. They say that such a strategy, once proved successful at Bell Labs, where scientists could work on project for a year without being evaluated.

Other ideas include awarding research grants to researchers, not specific research proposals, or trying funding to a proposal’s inherent innovation.

Some universities have begun supporting riskier research goals, in the form of grand challenges-oriented research, and the National Institutes of Health and various private organizations have experimented with ways to support innovative research. But publication pressures persist. Foster said he was nonetheless “optimistic” about change.

Academe should resist “the temptation to outsource judgment of quality to easily countable quantities,” he said. “Top universities emphasize that they are not interested in counting publications or citations -- that colleagues ‘read the work’ when evaluating a case.”

Scholars approaching any milestone, from searching for a first job to going up for tenure, can feel “pulled toward something safe and decipherable,” Foster said. “At least they'll have the publications, right? And that's more what I hope we can keep in mind: the importance of creating and protecting space (or rather, time) to take real risks. That's what tenure is supposed to do, which is one of the reasons that attacks on the tenure system are so worrisome. It's a shortsighted and ultimately counterproductive trade-off.”

Of course, sometimes sticking with more traditional research has its value -- as a recent, massive study suggesting that most psychology study results cannot be successfully replicated indicates.

The lead author of that study, Brian Nosek, a professor of psychology at the University of Virginia, said he wasn’t familiar enough with Foster’s paper to critique its methodology, but said it sounded “intriguing.” In any case, he said, “innovation and accumulation are not mutually exclusive.”

“Innovation occurs when expectations are violated,” Nosek said. “Replication is actually a great way to spur innovation because, when replications are successful, they increase confidence and generalizability of existing claims, and when they are not successful, they spur innovation to try to understand why different results were observed.”

Foster said he agreed that building on existing knowledge was essential to science, but that he was interested in how much innovation should be mixed in -- what he called a "division of labor." 

"Too much innovation, and science would be incoherent," he said. "Too much tradition, and it would slow to a crawl."


Back to Top