Communities and patterns of scientific collaboration

This is the author's accepted version of this article deposited at arXiv (arXiv:1006.1788v2 [physics.soc-ph]) and subsequently published in Scientometrics October 2011, Volume 89, Issue 1, pp 381-396. The final publication is available at link.springer.com http://link.springer.com/article/10.1007%2Fs11192-011-0439-1Author's note: 17 pages. To appear in special edition of Scientometrics. Abstract on arXiv meta-data a shorter version of abstract on actual paper (both in journal and arXiv full paper17 pages. To appear in special edition of Scientometrics. Abstract on arXiv meta-data a shorter version of abstract on actual paper (both in journal and arXiv full paper version)17 pages. To appear in special edition of Scientometrics. Abstract on arXiv meta-data a shorter version of abstract on actual paper (both in journal and arXiv full paper version)17 pages. To appear in special edition of Scientometrics. Abstract on arXiv meta-data a shorter version of abstract on actual paper (both in journal and arXiv full paper version)17 pages. To appear in special edition of Scientometrics. Abstract on arXiv meta-data a shorter version of abstract on actual paper (both in journal and arXiv full paper version)This paper investigates the role of homophily and focus constraint in shaping collaborative scientific research. First, homophily structures collaboration when scientists adhere to a norm of exclusivity in selecting similar partners at a higher rate than dissimilar ones. Two dimensions on which similarity between scientists can be assessed are their research specialties and status positions. Second, focus constraint shapes collaboration when connections among scientists depend on opportunities for social contact. Constraint comes in two forms, depending on whether it originates in institutional or geographic space. Institutional constraint refers to the tendency of scientists to select collaborators within rather than across institutional boundaries. Geographic constraint is the principle that, when collaborations span different institutions, they are more likely to involve scientists that are geographically co-located than dispersed. To study homophily and focus constraint, the paper will argue in favour of an idea of collaboration that moves beyond formal co-authorship to include also other forms of informal intellectual exchange that do not translate into the publication of joint work. A community-detection algorithm is applied to the co-authorship network of the scientists that submitted in Business and Management in the 2001 UK RAE. While results only partially support research-based homophily, they indicate that scientists use status positions for discriminating between potential partners by selecting collaborators from institutions with a rating similar to their own. Strong support is provided in favour of institutional and geographic constraints. Scientists tend to forge intra-institutional collaborations; yet, when they seek collaborators outside their own institutions, they tend to select those who are in geographic proximity