Coordination problems are intrinsic to any economic setting where several actors must undertake complementary activities in pursuit of some common goal. Casual evidence from the field suggests that coordination success is common but not guaranteed. Evidence from the laboratory are also mixed. While some studies reveal surprisingly high levels of coordination, in experiments that have studied coordination in settings with Pareto-ranked equilibria, the standard finding is coordination failure. Previous research has shed important light on some of the determinants of coordination success – such as the number of players, the payoff structure and so on. In this paper we report a novel experiment designed to investigate another possible determinant: that is, the extent to which the closeness of pre-existing social relationships between game participants influences their ability to coordinate in a weak-link game. A novel feature of our design is a manipulation intended to control and vary the closeness of relationships that exist between the players in different groups. Another novelty is that we draw on concepts and tools from social psychology with a view to operationalising and measuring 'relationship closeness'. The results show clearly that relationship closeness does significantly enhance coordination success. Thus, in a given group, the higher the elicited relationship closeness, the more likely it is that a Pareto- efficient equilibrium prevails. But, given the richness of our data, it is possible to go much beyond this qualitative statement because the distribution of elicited levels of relationship closeness predicts the distribution of the coordination equilibria in a weak-link game.