Temporal variational structures of long-term trends, recursive variations and seasonal variations were investigated, as to the incidence data of infectious diseases obtained from Infectious Diseases Surveillance Program in Japan by a newly devised procedure of time series analysis. In this procedure, three-dimensional spectral array indicates temporal variations of periodic structures of the incidence data, and is useful for investigating the long-term trend of the incidence data. Dominant spectral lines of the whole incidence data for each disease were assigned as the linear combinations of the fundamental modes. The optimum least squares fitting curve calculated with the fundamental modes essentially reproduces the recursive variation of the incidence data and an extension of the curve could be used to predict the incidence of the disease. Seasonal periodic modes of the incidences were estimated. The infectious diseases were classified into three basic groups (measles-, chickenpox- and mumps-patterns) and others, according to the resemblance of temporal variational structures of the incidence data.