In this paper we investigate the aggregate implications of social interactions for basic research and economic growth. In particular, we focus on the effects of both the size of the scientific community and the strenght of social exchange among researchers on science productivity and uneven growth. Basic research is modelled as a contest which awards a real prize to the winner and nothing to the losers. Agents are endowed with heterogeneous talent and discoveries are uncertain events which depend on the talent and effort of individuals and on their aggregate values. A CES index of the distribution of both talent and effort summarizes the features of the interactions of the scientific community from which increasing returns may derive. Social interactions among scientists cause multiple equilibria, among which a poverty trap with zero knowledge production and zero growth may emerge.