In recent years, the theme of actors and agency has made a notable reappearance in the neo-institutionalist literature, in relation in particular to what has been called institutional entrepreneurship and institutional entrepreneurs. Two recent books by M. Granovetter on the one hand, Neil Fligstein and Doug McAdam on the other, can be seen as significant examples of this “return of the actor” in American sociology. The troubling conceptual inconsistencies they contain, however, also document the incomplete integration of an action perspective into what basically remains a neo-institutionalist framework, giving epistemological priority to structure over action. This paper aims to highlight the most important of these inconsistencies and sets out to interpret them as a sign of how sociologists position themselves and their discipline in the wider field of social science, and in particular in relation to economics. It concludes by suggesting that in order to go beyond such methodological and conceptual confusion, we need to get away from a substantialist, decontextualized view of the actors’ identities and rationalities, and replace it with a relational conception of the actors’ identities and rationalities, in which their behavior would be considered to be attributes not of the actors, but of the local relational configurations and the stable patterns of transactions maintained by them.