When it is introduced in a new location, an infectious disease goes through a phase, which we call here the stochastic phase, during which case prevalence and incidence remain small. In a population not subject to additional case importations, the outcome of this phase is random: infections can keep taking place until such time that the case count increases exponentially, or chains of propagation can cease leading to the extinction of the disease. However, populations are almost never isolated (otherwise they would not see case introductions in the first place). I will present work with Evan Milliken (U of Louisville) in which we consider a very simple two patch stochastic metapopulation SIS model to investigate the effect of connections between locations on the duration of the stochastic phase of an epidemic. The specificity of the model resides in our considering not only zero cases as an absorbing state, but also an upper threshold above which we assume that the model enters a quasi-deterministic exponential growth phase. The duration of the stochastic phase is then the time that the system spends between these two absorbing states. Through extensive numerical simulation, we investigate the duration of this phase, both in the one and the two patch cases.