In this paper we present the technical issues of a temporal model for fault-tolerant parallel programs. We present successively the formal aspects of this temporal model, and an algorithm that we have developed to detect errors in parallel programs running on a parallel architecture with shared memory. The proposed algorithm takes as input the specification formulas of a given parallel program. Then, it compare the formula of each current action in the program with the specification formulas given in input. In the case, when an error is detected, the algorithm returns two sets containing the erroneous variables, and a set containing the faulty statement numbers. A simple example is given to illustrate the model and the algorithm.