Considerable research has been done in order to schedule tasks on multiple processing systems. Some of the computation time of these tasks may, however, be imprecise due to the nature of the problem. In this paper, an imprecise task graph is used to model the problem where each node represents a task associated with its computation time. An algorithm, called rotation scheduling, is extended to handle the imprecise task scheduling and fuzzy arithmetic is used to estimate the size of a schedule. The goal of the algorithm is to give the optimized schedule as much as possible. Experiments showing the effectiveness of this approach are also presented.