This paper aims at an empirical analysis of the influence of social interactions on scientists’ productivity. In the econometric analysis the authors investigate the aggregate importance of social interactions in science through the analysis of data on publications in four scientific fields of seven advanced countries. The paper builds a dynamic autoregressive model which provides long-run multipliers. The model is estimated with a panel fixed effects methodology. Social interactions among researchers have positive effects on a scientist’s productivity, and there is a U-shaped relation between the size of a scientific network and individual productivity. This result is interpreted as providing evidence for threshold externalities and increasing returns to scale.