Scientists are most likely to collaborate when their research interests overlap to a small or moderate degree, an analysis of co-authorships at a US university has suggested.

Researchers at the University of Florida analysed the topics covered by thousands of publications from the institution ¨C involving 3,400 authors and 12,500 collaborative ties ¨C to establish when academics were most likely to work together.

Similar to previous theories that academics may be less likely to work together if their areas of interest were too closely or too disparately aligned, they found that there was a ¡°sweet spot¡± in?the middle ground where they would seek collaboration.

¡°We identify an inverted U-curve relationship signifying that collaboration is most likely at low to moderate levels of interest overlap between colleagues,¡± say the authors of?, published in the?Journal of Informetrics.

³ÉÈËVRÊÓƵ

One of the most interesting aspects of the results, according to the authors, was that researchers sought to work with people with interests that were ¡°substantially¡± different to the topics covered by their own papers.

¡°In other words, they look for collaborators whose expertise is outside the topical range spanned by their own research,¡± the paper says.

³ÉÈËVRÊÓƵ

The researchers say their findings held for most of the disciplines and types of collaboration they analysed. However, the relationship between topic overlap and collaboration in engineering seemed to be different, with co-authorship more likely when researchers shared a higher degree of similar interests.

The paper concludes that ¡°teams that match scientists with low-to-moderate overlap in research topics may be seen by individual scientists as the most appealing, promising and acceptable, as opposed to collaborations with colleagues who work on very similar topics (with whom competition may be stronger) or very different areas (with whom collaborating may be too costly or risky)¡±.

Co-author Thomas Smith, of Florida¡¯s department of sociology and criminology and law, said that using tools such as the language analysis methods set out in the paper could ¡°inform localised interventions that promote team science¡±.

¡°Finding an opportunity to approach a fellow researcher and propose a collaboration is difficult ¨C especially if they are established in their field,¡± he said.

³ÉÈËVRÊÓƵ

¡°The tools we apply in this paper can be used to help create opportunities and design incentives which encourage collaborations that are fruitful¡± for researchers and other interested parties, such as funders, he added.

simon.baker@timeshighereducation.com