The finite difference time domain method (FDTD) has been extended for application to frequency-dependent (dispersive) media such as the ionospheric plasma. This formulation of FDTD can be applied to the study of electromagnetic propagation in the plasma of the ionosphere. In the current work, we consider propagation in randomly structured ionization. The results of numerical FDTD computations of HF fields propagated through realizations of ionospheric structure are compared to the predictions of the standard random media propagation theory, which is based on the parabolic wave equation. The motivation for this study is to expose the way in which the approximate theory breaks down when applied outside of its regime of strict applicability. An earlier published analysis by the authors was limited to spatial signal decorrelation as expressed in ℓ0, the signal correlation length. In the current work, this spatial decorrelation analysis is supplemented with a similar analysis of signal frequency decorrelation (or delay spread) as expressed in ƒ0, the channel bandwidth. We find that the standard propagation theory for the calculation of ƒ0 predicts larger values than those produced in the FDTD computation. This appears to be due to a polarization coupling effect that is ignored in the standard propagation theory. FDTD-computed correlation lengths are found to be in substantial agreement with the propagation theory at higher frequencies. Agreement degrades in a graceful and understandable way at lower frequencies where the limitations of the approximations used in the theory are violated.