In this paper, a study concerning the problem of parameter estimation from one-component chromatographic experiments when linear models can be adopted is presented. The question is to know exactly which parameters can be theoretically estimated from an experiment when time-domain analysis is used. This question is independent of the measurement noise and depends only on the structure of the model used for the estimation procedure. To illustrate the approach, the identifiability concept is used here in the case of two examples of linear models: the N CSTRSs in series model and the plug flow with axial dispersion model. In the first case, the adsorbant concentration is supposed to be uniform, leading to a lumped parameter model. In the second case, Fickian diffusion is taken into account in the solid, leading to a distributed parameter model. The transfer function approach is used to test the parameter identifiability. A change of state variable is proposed in order to formulate the state-models according to a form suited to the parameter estimation purpose. This analysis is numerically illustrated in the case of the second model by using the data of Hufton and Danner, A.I.Ch.E. Journal, 39(6) (1993) 962-974, 954-961. A thermodynamical interpretation of the result is also given.