We consider a partnership game with two roles in which a large population of firms interact to carry out, say, R&D joint ventures. The partners have to build a common-property asset through a sequence of costly investments. Firms have access to a monitoring technology whose cost depends on the actual diffusion of cooperative behaviour across the economy. The paper explores the relationship between the (endogenous) level of transaction (viz., monitoring) costs and the emergence of social norms of trust. We find that in certain cases, i.e. when the partnership is very rewarding in absolute terms and the risk of exploitation from defector partners is relatively low, pure trust may emerge as a social standard of behaviour in a straightforward way. Vice versa, when the return from the partnership is less large and risk of exploitation is substantial, a two-phase transition to pure trust may emerge: in the first phase, ‘cautious’ co-operators who monitor the partner spread over, to be suddenly substituted by ‘pure’ co-operators who do not monitor the partner (and thus economise on transaction costs) once the other type of co-operators has wiped out defector opponents.