Gum arabic is a natural product used widely in the food industry as an emulsifier and stabilising agent. The gum contains 3 main fractions: an arabinogalactan (~80 wt%; AG) fraction, a glycoprotein (~ 2 wt%; GP) fraction and an arabinogalactan-protein complex (~18 wt%; AGP). This AGP fraction is largely responsible for the functional properties of gum arabic and, due to natural variation, the proportion of AGP within a gum arabic batch varies enormously. There is industrial interest, therefore, in fractionating the gum arabic to allow creation of a more homogenous product, as well as new products for the food industry. The aim of this work was to investigate the feasibility of using membrane technology to fractionate gum arabic. Polymeric membranes were used initially and showed success at rejection of AGP by size exclusion. Polysulfone membranes of 0.1, 0.5 and 0.8 μm nominal pore size were employed and the rejection of AGP was seen to decrease with increasing pore size, but the overall transmission of solids was seen to increase. Beneficial fouling was observed with the larger two pore sizes, which allowed greater fractionation after a fouling layer had developed. It was hypothesised that the 0.1 μm PS membrane was fouled by mainly cake formation, whereas the 0.5 and 0.8 μm PS membranes were subject to more in pore fouling. The critical flux of gum arabic was measured for these three pore sized membranes and was found to be highest (27 L m-2 h-1) for the 0.1 μm PS membrane. This was attributed to the lack of in pore fouling, which the 0.5 and 0.8 μm membranes suffered and resulted in fouling occurring at lower fluxes for these membranes; the critical fluxes for these membranes was found to be 15 and 22 L m-2 h-1, respectively Increasing the crossflow velocity (CFV) from 0.18 to 0.67 m s-1 was found to increase the critical flux. Filtration experiments above and below the critical flux for each pore size demonstrated the efficiency of operating below the critical flux, as operation could be sustained for much longer periods (up to 4h was tested) without the need for cleaning cycles. It also demonstrated the beneficial effect of the fouling layer with the 0.5 iii and 0.8 μm membranes, which showed little or no fractionation during operation below the critical flux. Finally, filtration studies were carried out with 3 different membrane materials and detailed surface analysis was performed to explain the differences in performance observed. Both polysulfone and fluoropolymer membranes were fairly hydrophobic, with contact angles of between 70 and 90º, and showed very high overall rejection of solids. High transmission of solids is required together with good rejection of AGP for an effective fractionation process. Hydrophilic cellulose acetate, however, showed very high transmission of gum arabic (~ 75%), but no rejection of AGP. Overall, the work has shown that fractionation of gum arabic with membranes is feasible using polysulfone membranes, but that further work is needed to optimise the separation. Higher transmission of the GP and AG are required whilst maintaining rejection of the AGP.