Correlation effects in diffusion of CH4 and CF4 in MFI zeolite have been investigated with the help of molecular dynamics (MD) simulations and the Maxwell-Stefan (M-S) formulation. For single-component diffusion, the correlations are captured by the self-exchange coefficient D-ii(corr); in the published literature this coefficient has been assumed to be equal to the single-component M-S diffusivity, D-i. A detailed analysis of single-component diffusivity data from MD, along with published kinetic Monte Carlo (KMC) simulations, reveals that D-ii(corr)/D-i is a decreasing function of the molecular loading, depends on the guest-host combination, and is affected by intermolecular repulsion (attraction) forces. A comparison of published KMC simulations for diffusion of various molecules in MFI, with those of primitive square and cubic lattices, shows that the self-exchange coefficient increases with increasing connectivity. Correlations in CH4/CF4 binary mixtures are described by the binary exchange coefficient D-12(corr); this exchange coefficient has been examined using Onsager transport coefficients computed from MD simulations. Analysis of the MD data leads to the development of a logarithmic interpolation formula to relate D-12(corr) with the self- exchange coefficient D-ii(corr) of the constituents. The suggested procedure for estimation of D-12(corr) is validated by comparison with MD simulations of the Onsager and Fick transport coefficients for a variety of loadings and compositions. Our studies show that a combination of the M-S formulation and the ideal adsorbed solution theory allows good predictions of binary mixture transport on the basis of only pure component diffusion and sorption data.