A statistical analysis of the precision of the estimate of endurance limit as determined by the Prot method is carried out. This analysis shows that the Prot method cannot achieve a great saving in the number of specimens required, for its statistical efficiency is found to be even less than that of its other two competitors, the probit and staircase methods. This fact, coupled with serious objections to its assumptions, renders its usefulness highly questionable. A secondary result of the analysis is that the optimum efficiency of the Prot method is achieved when only two rates of increase of stress, as widely separated as possible, are used. A derivation of a general formula, of which the Prot method is a particular case and which makes explicit all assumptions necessary for the valid derivation of the Prot method, is also given.