Anyone who’s worked in a laboratory probably feels that having key members of the group placed closer together – allowing more face-to-face interactions – makes for a better research project.
That’s certainly what Isaac Kohane thought when in 2005 he persuaded Harvard’s dean of administration, Richard Mills, that researchers at a new biomedical informatics centre should be given plenty of neighbouring space – even though they all worked on computers, so in theory could communicate by email from distant locations.
“I felt this viscerally, but there was no hard evidence,” says Kohane, who works at Harvard Medical School, Cambridge, Massachusetts. At Mills’ suggestion, he – and a small army of undergraduates – set out to test whether physical proximity had promoted higher-impact science at Harvard. They picked out 35,000 articles published between 1999 and 2003 in biomedical science, each with at least one Harvard author. It took them two years to puzzle out exactly where the Harvard investigators were studying when they produced their work – down to the level of individual offices and laboratories.
The results, published in PLoS One today, find that the closer the distance between first and last authors on a paper, the more highly-cited were their resulting research papers. First authors – who often bear the brunt of the work – and last authors – often the organizational lead – are probably just reliable proxies for key scientists in a research project, Kohane thinks. The effect wasn’t found for middle authors, who could be spread quite broadly without their pair-wise distances clearly affecting research impact.
The researchers studied physical distance at a number of scales – tens of metres, hundreds of metres, or kilometres across campus. At all of these scales, the distance between first and last author was highly correlated with the impact of an article, no matter how many authors appeared on the paper.
The team also looked at individual buildings on the four campuses over which Harvard life science research happens to be spread. The more the collaboration between researchers within a building, the higher the impact of publications from that building, they found. (The picture, taken from the research paper, shows buildings on the Longwood campus of Harvard Medical School. The height of the building shows the mean number of citations of publications originating from that building, while the colour reflects the proportion of publications in which both first and last authors work in the building, from grey=low to blue=high. The team has also published an interactive map).
Exactly why the associations hold isn’t quite clear, the team points out. It could be that physical proximity does make for better collaboration. That concept that won’t surprise anyone familiar with the importance of transmitting ‘tacit’ scientific knowledge: complex tasks, such as how to use equipment to run experiments, that are hard to glean from reading literature or emails and require a physical co-presence.
“It may also be that investigators have a strategic preference for keeping potentially high impact projects wholly within their own laboratory or close circle of research associates,” the authors write.
But, concludes Kohane, “if you want people to collaborate, these findings reinforce the need to create architectures and facilities that support frequent, physical interactions”.
The study is restricted to biomedical sciences and an organizational structure like that seen at Harvard, where life science researchers are connected by being affiliated to Harvard Medical School as well as their independent hospitals, the team points out – so it may not hold for other fields and across multiple institutions.
This seems to be the first empirical study of the connection between local distance and impact, says Anthony van Raan, an expert in using citation analyses to study scientific productivity and impact at Leiden University, the Netherlands. “I think their explanation of ‘strategic preference’ really makes sense,” he says. “All researchers like to – indeed must! – collaborate nationally and internationally. But I am convinced that scientists want to keep their really important breakthroughs within their own group.”
Most studies of the relationship between spatial distance and scientific impact (“spatial scientometrics”) have been done on a national and international scale, where it’s been demonstrated many times that international collaborations show greater impact than local collaborations. (Spatial scientometrics are also discussed in a Nature feature article, ‘Building the best cities for science’.)
“Combining the spatial scientometric literature and the PloS One article would in my view suggest that for producing high-impact science local spatial structures and global spatial structures – sometimes referred to as ‘glocal’ spatial structures – need to be combined,” comments Jarno Hoekman, a researcher at Eindhoven Unversity of Technology who studies the relationship between physical distance and research productivity.
Kohane speculates that the success of an international collaboration might depend on the physical locations of key players. This hasn’t yet been tested. “I think this effect would have a long half-life,” he adds: “if you were once co-located then you may have established a trust, collegiality, and ability to communicate effectively, even more remotely.”
The internet and email has helped a huge increase in international collaborations, but it doesn’t yet mean – at least up to 2003, according to this study – that key scientists can afford to avoid face-to-face interaction for high-impact science. “The message is that the sociology of science is still dominated by knowing each other in an intense daily environment, thus by physical proximity. All the modern web-based facilities are important and used very frequently, but are not of overriding importance,” says van Raan. As teleconferencing and video calling improves, the importance of physical distance might diminish, however.
Kohane practices what he preaches: he and first author Joon Lee (who coordinated the undergraduates’ fact-finding) work on the same floor. “When the study started we were on different floors, and Joon told me that I became a lot more helpful when I moved to his floor,” he notes.