We propose an original physically justified method for the evaluation of the spread of values of the critical temperature of brittleness. Its principal distinction from the standard method and advantage over this method is that the spread of data is determined according to the source data used for the evaluation of the critical temperature of brittleness but not according to the set of values of this temperature. We introduce quantitative characteristics of a series of impact-toughness tests aimed at finding the critical temperature of brittleness: the effective number of experimental points and the weight (degree of influence) of each of these points on the critical temperature of brittleness. A criterion of sufficiency of the number of values of impact toughness for the adequate determination of the critical temperature of brittleness according to these values is also proposed. We describe numerous experiments that reveal good correlation between the spread of values of the critical temperature of brittleness obtained by the proposed method and the corresponding standard deviations.