Contagion has been of interest to social scientists for several decades, but until recently empirical data on the spread of information was limited to population-level counts and generally only available for "successful" events. In this talk I describe how the increasing availability of digital data at the individual level and for much larger populations of events have allowed computational social science to ask new questions, such as: How much contagion is "viral?" What do we mean by viral anyway? And how accurately can we predict the spread of information on large networks? I then conclude with some discussion about how the answers to these questions shed light on theories of contagion and on social scientific explanations more generally.