This paper was presented as an overview lecture summarizing the 1972 International Symposium on Fatigue at Elevated Temperatures held at Storrs, Conn. Starting with the observation of the diversity of subjects covered and lack of unanimity of approaches used, it becomes clear that there exists an urgent need for a unifying framework around which the many facets can be coherently structured. It is proposed that the strainrange partitioning concept has the potential of serving as such a framework. The method divides the imposed strain into four basic ranges involving time-dependent and time-independent components. It is shown that some of the results presented at the symposium can be better correlated on the basis of this concept than by alternative methods. It is also suggested that methods of data generation and analysis can be helpfully guided by this approach. Potential applicability of the concept to the treatment of frequency and holdtime effects, environmental influence, crack initiation and growth, thermal fatigue, and code specifications are then considered briefly. A required experimental program is outlined.