While engineering methods employed to predict mass transport rates across carrier fluids are often limited to either ideal gas mixtures or constant density liquids, we deal here with species mass transport across nonisothermal compressed gas "films", with special reference to hydrocarbon fuel vapor transport across nonisothermal N-2 or H2O boundary layers at pressures up to ca. 300 atm, and at temperatures. often above 1000 K. We show that because of the pressure sensitivity of the Soret factor (which quantifies the relative importance of mass transfer due to temperature gradients compared to that due to concentration gradients) mass transfer coefficients become far more sensitive to pressure level than would have been anticipated from the pressure sensitivity of rho D-12(infinity). Motivated, in part, by combustion applications, we first examine the expected pressure dependence of the binary Soret factor, alpha(T,12) for each of the n-alkanes (CH4 to C20H42) dilute in compressed N-2 or H2O, exploiting a rational formulation for "correcting" Chapman-Enskog-derived Soret factors (alpha(0)(T,12)) to higher pressures based on the Thermodynamics of Irreversible Processes (TIP) combined with a Virial Equation of State (VES). Our TIP-VES-predicted Soret factors are used to demonstrate the pressure sensitivity of expected "Soret-modified" mass transfer coefficients (Sherwood numbers)for the illustrative case of C12H26(g) transport across N-2(g) at temperature ratios between 0.3 ("cold"-wall) and 2.0 ("hot"-wall) at pressures up to 300 atm. Because of the growing importance of compact, high-pressure systems infields beside combustion, including supercritical fluid extraction and even distillation, our present results suggest that reliable mass transfer rate predictions in nonisothermal dense vapor systems of engineering importance will generally require systematic inclusion of non-Fickian molecular mass transport. (c) 2007 American Institute of Chemical Engineers.