Social contagion is a subset of contagion which includes all social phenomena that can and do spread via social networks. The notion of how something becomes popular is very relevant to the concept of social contagion. Rumors, fads, and opinions can spread through social networks like wildfire, "infecting" individuals until they become the norm. This thesis investigates the dynamics of social contagion, employing a combination of formal analysis, simulation, and empirical data mining approaches to examine the processes whereby social contagion spreads throughout social networks. I introduce the concept of critical mass for a subclass of social contagion called complex contagion. This concept builds on earlier work to describe the nonlinear dynamics whereby most socially contagious phenomena infect very few people while a few become overwhelmingly popular. I also investigate socially contagious phenomena that arise when rational agents act under conditions of local information. Finally, I examine how my analytic work applies to a large dataset of empirical social contagion and draw implications for further research in the area.