Using data from a randomised experiment in Kenya, we estimate the causal effect of social networks on technology adoption. In this experiment, farmers were invited to information sessions about the use of Tissue Culture Banana (TCB), an in vitro banana cultivation technology. We find that an additional social connection with a treated farmer causes an untreated farmer to be 2.25 pp more likely to adopt TCB 6-18 months post-intervention, but not in the longer term. We provide evidence that the adoption of TCB by those social connections is the mechanism driving the effect; therefore, treated connections are significant because treated farmers are more likely to adopt. We also find that indirect social network effects, proxied for by eigenvector centrality, influence adoption at both the village level and the farmer level.