A practical probabilistic technique is presented for evaluating the distribution of project completion time. Upper and lower bounds on the completion-time probability are established and form the basis for the probabilistic network evaluation technique (PNET) algorithm. The mean durations and the correlations among the network paths are the main parameters used to generate a set of representative paths from which the probability distribution of completion time is evaluated. This technique overcomes the shortcomings implicit in PERT, although its application is almost as simple as the traditional PERT method. Several construction projects are presented as examples to illustrate the algorithm. The results obtained by PNET are shown to be in very good agreement with simulation runs, thus verifying its validity; computation times, however, are several order-of-magnitude less than those required by simulation.