A set of monthly streamflow records is generated and analyzed to give a set of synthetic critical periods. Some characteristics of these periods are derived, and the mean and standard deviation of their values are compared with the appropriate values for the historical critical period. This procedure is repeated with three characteristics for a number of different rivers by using a variety of generating techniques. A study of the results allows a comparison between the generating techniques in regard to both their cost in computer time and their ability to synthesize critical droughts comparable to those recorded historically. The results appear to be influenced by the relative aridity of the watershed and by the size of the reservoir considered. In addition, some general conclusions can be drawn regarding the advisability of using generated streamflow in studies of reservoir design and operation.