This article presents a flexible, data-analytic procedure for calculating spectral-temporal indexes by exploiting the duality property between frequency and time-domain procedures. The approach consists of capturing the interrelationships between the subordinate series across a range of cycles using a principal-components decomposition in the frequency domain and transforming this information into the time domain to construct a temporal index. The approach is applied to the calculation of an index of U.S. short and long interest rates. The main result of the empirical analysis is that, for a broad range of cycles, a single index can be used to summarize and synthesize the information contained in a range of interest rates differing in terms to maturity.