The relationship between the research performance of scientists and their position in co-authorship networks in three fields

• Statistics shows a less connected and more fragmented network than the experimental fields.
• There is a relationship between the position of scientists in co-authorship networks and their g-index.
• Authors with a high number of collaborators and/or strong ties tend to show a higher g-index.
• Playing a bridging role is not associated to the g-index of scientists.

Research networks play a crucial role in the production of new knowledge since collaboration contributes to determine the cognitive and social structure of scientific fields and has a positive influence on research. This paper analyses the structure of co-authorship networks in three different fields (Nanoscience, Pharmacology and Statistics) in Spain over a three-year period (2006–2008) and explores the relationship between the research performance of scientists and their position in co-authorship networks. A denser co-authorship network is found in the two experimental fields than in Statistics, where the network is of a less connected and more fragmented nature. Using the g-index as a proxy for individual research performance, a Poisson regression model is used to explore how performance is related to different co-authorship network measures and to disclose interfield differences. The number of co-authors (degree centrality) and the strength of links show a positive relationship with the g-index in the three fields. Local cohesion presents a negative relationship with the g-index in the two experimental fields, where open networks and the diversity of co-authors seem to be beneficial. No clear advantages from intermediary positions (high betweenness) or from being linked to well-connected authors (high eigenvector) can be inferred from this analysis. In terms of g-index, the benefits derived by authors from their position in co-authorship networks are larger in the two experimental fields than in the theoretical one.