1. The parametric relations of Larson and Miller or Manson and Haferd are basically more correct than those of Manson and Succop or Graham and Walles. However, since the coefficients are not constant, one must be very cautious in using these relations when changing from one stress to another [10]. 2. The temperature dependences log t-1/T or log t-T are linear over a large range of service lives and can be regarded as reliable for extrapolation to lower temperatures at a constant stress. 3. The above temperature relationships are practically identical (when the relative test temperature variation is about 10%); however, a t=f(1/T) relationship is physically sounder and can therefore, be recommended for practical use.