The literature on social innovations focuses on alternative approaches to organising the provision of (and the valuation of) products and services that meet societal needs that are not met through conventional channels. Alternatively, institutional innovations are imagined in terms of institutional change: "institutional change[i]s a difference in form, quality, or state over time in an institution. (…) If the change is a novel or unprecedented departure from the past, then it represents an institutional innovation."(Hargrave and Van de Ven 2006: 866) In this paper, I examine when and how social innovations might be characterised as institutional innovations, that is, when do social innovations become institutionalised and how might we understand the boundaries of social innovations from an institutional perspective? I do this by exploring social innovations in agriculture. I analyse three examples of social innovation from three different countries, each at a different stage of institutionalisation. First, I explore participatory guarantee systems in Bolivia, which are currently institutionalised into national organic policy. Second, I explore an integrated production and training centre in Benin developed by a religious NGO, which is only now beginning to gain both market and public sector recognition. Finally, I explore Rainforest Alliance certified tea production in Tanzania, which has become institutionalised within the private sector, but has been facilitated by collaboration and reorganization of public sector organizations. Through these examples I focus on the questions of the boundaries of social innovations in terms of scale, knowledge, actors and market dynamics.