IT is doubtful whether we are getting more net work from metals today in dynamically loaded parts than was obtainable 25 years ago, and no super-strength-alloy discoveries seem imminent; however, much can be done to increase the fatigue strength of many machine parts made from ordinary structural materials by merely extending processes already known to be satisfactory, and avoiding practices that reduce fatigue strength.We have today new concepts of fatigue failure: Fatigue failures result only from tension stresses, never from compressive stresses. Any surface, no matter how smoothly finished, is a stress-raiser.Structural materials are not rigid. Many fatigue failures can be traced to elastic deflection for which no allowance was made in design.From experience with practical machine parts, we can only conclude that stress calculations by textbook methods are wholly inadequate unless we generously temper our calculations with experience. The accuracy of stress data from photo-elasticity, brittle lacquers, extensometers, and similar methods is usually greater than by mathematical analysis, but these are far from reliable.As a working hypothesis, it seems reasonable to assume, except possibly for very ductile metals, that:The slope of the fatigue curve, as measured on a log-log plot, is a measure of effective stress; and fatigue curves for varying stress concentrations converge toward a point near the tensile strength of the material at some considerable number of stress cycles.Fully 90% of all fatigue failures occurring in service or during laboratory and road tests are traceable to design and production defects, and only the remaining 10% are primarily the responsibility of the metallurgist as defects in material, material specifications or heat-treatment.Study of fatigue of materials is the joint duty of the metallurgical, engineering, and production departments. There is no definite line between mechanical and metallurgical factors that contributes to fatigue. This overlapping of responsibility is not sufficiently understood. Until more time is devoted to searching for mechanical causes rather than metallurgical ones, we cannot make full use of our materials.