We address the problem of scheduling a flowshop plant with uncertain processing times described by discrete probability distributions. The objective is to find a sequence of batches that minimizes the expected makespan. To circumvent the problem of combinatorially explosive state spaces, we propose a novel and rigorous branch and bound algorithm that provides the optimal solution and is based on the result that a lower bound to the expected makespan can be obtained by evaluating over an aggregated probability model. Numerical results for a number of example problems show that the solution times for the proposed method are several orders of magnitude smaller than those for a multiperiod model. In addition, an important extension of this method is proposed for the case of continuous probability distributions of certain forms, using discretization schemes that give excellent approximations.