Bigger Is Not Always Better for Team Science

Using currently available metrics, it is well established that bigger research teams tend to produce higher-impact papers as measured by the number of their citations. But a well-cited paper isn’t necessarily a breakthrough paper, says James Evans of the University of Chicago who led the research. Review articles, for example, tend to be highly cited but rarely shift scientific paradigms.

To investigate how groundbreaking or “disruptive” a piece of research may be, Evans and his team employed a metric (previously devised by Owen-Smith and management strategist Russell Funk of the University of Minnesota) that determines the way in which a piece of research is cited, rather than merely how many times.

Using this nuanced approach, a paper’s disruptiveness is determined as follows: if subsequent research articles cite the given paper together with many of that paper’s own citations, it indicates the paper has built upon an existing body of work. If, however, subsequent articles cite the given paper alone, it suggests the work was transformative, or “a jumping-off point” for a new field of research, explains Owen-Smith.

Owen-Smith and Funk’s own work using the metric showed patents from universities tend to be more disruptive than those from companies.

Now, Evans’s team has applied the algorithm to a much larger and diverse dataset consisting of 42 million scientific publications listed on Web of Science (published between 1954 and 2014), 5 million patents granted from the US Patent and Trademark Office (from 1976 to 2014) and 16 million software products uploaded to GitHub, a web-based hosting service for software developers, from 2011 to 2014.

The analysis showed that the most disruptive papers, patents, and software products tend to be produced by small groups, and that, as team size grows, disruptiveness declines. This trend held true even after controlling for scientific discipline, year of publication, and author. Indeed, an author’s disruptiveness was found to drop as the number of their coauthors increased.

“It’s really carefully done with a lot of alternative explanations explored,” says Anita Woolley, who studies organizational behavior and teamwork at Carnegie Mellon University’s Tepper School of Business. “That the effect remains [after controlling for potential confounders] is pretty compelling.”

The team also found that Nobel Prize–winning papers tended to fall within the top 2 percent of the most disruptive papers, while review articles fell within the bottom 46 percent.

“A lot of big ideas come from disruption,” says Woolley, who did not participate in the research. But, funding agencies often “push us in the direction of having bigger and bigger teams,” she adds. “This [research] is really calling that into question.”

The paper is not suggesting big-team science isn’t valuable, says Owen-Smith. After all, building upon existing theories is an essential research endeavor. It merely reveals the type of science that differently sized teams tend to engage in, he says, and highlights the importance of valuing small collaborations.

Ultimately, Owen-Smith says, if we want science to “push forward human knowledge and also to provide a social insurance policy against an uncertain future,” we need a combination of big teams developing and strengthening existing ideas and small teams performing high-risk, high-reward, innovative projects. The benefit of this study on disruption and group size, he adds, is that it “allows us to start asking questions about how we can tune the system to get the mix we need.”

L. Wu et al., “Large teams develop and small teams disrupt science and technology,” Nature, doi:10.1038/s41586-019-0941-9, 2019.