This article analyses the role of social networks as facilitators of information flows and banana output increase. Based on a village census, full information is available on the socio-economic characteristics and banana production of farmers' kinship group members, neighbours and informal insurance group members. The census data enable us to use individual specific reference groups and include exogenous group controls to tackle standard difficulties related to identification and omitted variables bias when analysing social effects. For the survey village of Nyakatoke in Tanzania the results suggest that information flows exist within all types of groups analysed but output externalities are limited to kinship groups. Using networks may offer scope for effective information flows on agricultural techniques, but our evidence suggests that not just any local network will have a social externality impact, requiring a clear understanding of local social networks for maximum impact.