Katz and Lazarsfeld’s Personal Influence introduced the world to the impact of networks on the dissemination of mass media. Their “two-step flow” model showed that broadcast signals reached most of the public by being filtered through well-connected people – “opinion leaders” – who were the primary receivers of media messages, and the primary vehicles through which those messages were then disseminated to everyone else. Scientific and industry attention soon shifted to the task of identifying who the opinion leaders were, and how they could be targeted to spread new content. I trace these intellectual developments through to the arrival of social media, which brought greater attention to the idea of “central” players – or “influencers” – in the social network, as the key leverage points for disseminating products, ideas, and political messages. I show how this scientific search for the sources of social influence eventually led to a paradox: the unlikely finding that many social contagions do not spread from the central players to the periphery, but rather from the network periphery to the center. To explain these startling findings, the distinction between simple contagions, like information and viruses, and complex contagions, like social innovations and political movements, shows how the spread of new ideas through social networks depends in counterintuitive ways on the complexity of the contagion and the structure of the social network.