Little is known about the microscopic operations of bicycle traffic. This lack of knowledge makes it difficult to design and to analyze facilities that are also used by cyclists, as well as to assess their efficiency and safety. This paper puts forward a novel composite headway model for bicycle flows. On the basis of a working definition of a bicycle headway, the model assumes that the headway is the sum of the so-called empty zone (i.e., the minimum headway that a cyclist has with respect to the bicycle he or she is following) and the free headway (i.e., the additional headway that cyclists maintain when they are not following). Both headways are stochastic variables, given the high degree of heterogeneity in cyclists’ behavior. Further, a distribution-free model estimation approach is put forward to make it possible to identify the distributions of the empty zone and the free headway without the need to specify their functional form first. The workings of the model and the estimation method are illustrated with the use of data collected at a busy intersection in the Netherlands.