This paper deals with nominal time-series. The spectral analysis of an operator which sums up time dependences provides a decomposition of the nominal process into a sequence a time-invariant uncorrelated random variables and a set of non-random time tunctions which give changing scores to the States of the process. These results extend scalar harmonic analysi's to nominal processes, multiple correspondence analysis to a continuons set of nominal vari ables and generalized canonical analysis to mu1tidimensional stochastic pro cesses. Numerical computation comes clown to a particular kind of correspon dence analysis .