The myth of global science collaboration—Collaboration patterns in epistemic communities

Scientific collaboration is often perceived as a joint global process that involves researchers worldwide, regardless of their place of work and residence. Globalization of science, in this respect, implies that collaboration among scientists takes place along the lines of common topics and irrespective of the spatial distances between the collaborators. The networks of collaborators, termed ‘epistemic communities’, should thus have a space-independent structure. This paper shows that such a notion of globalized scientific collaboration is not supported by empirical data. It introduces a novel approach of analyzing distance-dependent probabilities of collaboration. The results of the analysis of six distinct scientific fields reveal that intra-country collaboration is about 10–50 times more likely to occur than international collaboration. Moreover, strong dependencies exist between collaboration activity (measured in co-authorships) and spatial distance when confined to national borders. However, the fact that distance becomes irrelevant once collaboration is taken to the international scale suggests a globalized science system that is strongly influenced by the gravity of local science clusters. The similarity of the probability functions of the six science fields analyzed suggests a universal mode of spatial governance that is independent from the mode of knowledge creation in science.