The primary focus of this research is to extend the principles of nanofiltration (NF) to non-aqueous systems using organic solvent nanofiltration (OSN) membranes. Solvent transport in organic solvent nanofiltration membranes has been studied in a lab-scale cross-flow nanofiltration rig over extended periods using the common solvents methanol, toluene, ethyl acetate and their mixtures. The organic solvent nanofiltration membranes STARMEMTM 122* (W.R Grace and Co.) and MPF-50 (Koch Membrane Systems) were investigated. Both solution-diffusion and pore-flow models can be used to predict permeation of solvent mixtures. For the solvents studied, it is possible for reasonable predictions of solvent mixture flux to be made over the whole concentration range, based on the data for pure solvents. To understand solute transport, flux and rejection performances under cross-flow, for concentrated (5-30 wt. %) methanol-dimethyl methylsuccinate (DMMS) solutions, were examined. The experimental flux/rejection data for the flat-sheet membranes was fitted using pore-flow and solution-diffusion models, coupled with film theory for liquid mass transfer effects. It was found that solution-diffusion gives a better description of some of the experimental trends. In the practical range of concentrations studied, rejection should be seen as a variable, dependent on the mass transfer characteristics of the nanofiltration system in use. Pilot plant OSN spiral-wound performance was investigated for highly concentrated solutions of a highly rejected solute. Flat-sheet determined model transport parameters were used, with success, for predictive modelling of pilot-plant spiral-wound experimental data. Lastly, this work presents a new modelling approach for solvent transport though OSN membranes. In this study an adsorption-diffusion membrane transport model, able to describe the experimental flux of a reasonable number of solvents permeating through a polyimide polymeric membrane, was derived. The main contribution of this modelling approach is its ability to incorporate all OSN flux determining factors in a predictive mathematical model.