Any rigorous approach to temporal phenomena in the social sciences involves a model of time, that is, a simplified representation of it, in order to keep what is seen as essential and to drop or minimise what is not.1 This holds true for any kind of method, be it narrative history, duration models, life course histories, process-tracing, Markov models, or auto-biography. My purpose in this paper is to explicate the way time is modelled by sequence analysis (SA), a new approach to longitudinal data brought to the social sciences in the 1980s (Blanchard 2011; LaCOSA 2012; Blanchard et al. forthcoming 2014). The importance of sequences has been long recognized and studied in various ways (Abbott 1995: 96–103), but SA has renewed and improved the way they are conceptualized and analyzed.