In pervaporation, membrane selectivity is defined as the ratio of the ratios of two components of a liquid:liquid mixture in the permeate and in the feed. In a thermodynamically-controlled system, this value will be the same as the difference between vapour and liquid composition in the absence of a membrane. We demonstrate this to be the case for water:ethanol pervaporation at high water:ethanol ratios both for an unmodified polydimethylsiloxane membrane and for a membrane modified by grafting a layer of hydrophilic polyacrylamide. We observe transient kinetic deviations toward greater selectivity on addition of salts which push the thermodynamic vapour pressure equilibrium towards ethanol, and to a more significant degree with grafting of polyacrylamide to the hydrophobic membrane, suggesting that these modifications retard the flow of water through the membrane. The physical plausibility of the chemical potential gradient used in interpretation of pervaporation data by the solution-diffusion model is critiqued.