Defatted whey was obtained by aggregating residual fat to calcium phosphate precipitates and separating the precipitate by membrane microfiltration (pore diameter 0·2 μm). When ultrafiltering this defatted whey the performance of an inorganic membrane (molecular mass cut-off, 10 kDa) was limited by the large concentration of Ca and phosphates. Consequently, the influence of the aggregation pH (either decreasing or constant) on membrane fouling has been studied for ultrafiltration (UF) of defatted sweet whey and defatted whey UF retentates (protein content up to 30g l–1). In all experiments protein rejection was 100%. When pH was kept constant during the pretreatment, membrane fouling was significantly lowered. Hydraulic resistances ascribed to irreversible fouling were in good agreement with fouled membrane analyses performed by i.r. and X-ray photoelectron spectroscopies. They showed that provided a low Ca and phosphate content was maintained in the microfiltrate, which was achieved at constant pH, no apatite was detected within the membrane, and proteins were less fouling. On the other hand, the amount of fouling material depended on the transmembrane pressure gradient along the hydraulic path. On the membrane surface, the higher the pressure, the higher the fouling. In the membrane bulk, the fouling heterogeneity depended on the ability of the defatted whey to precipitate apatite. If it did, the higher the pressure, the higher the calcium phosphate and the protein fouling. With other phosphate structures, the bulk fouling depended on the barrier formed by surface fouling layers and the protein concentration polarization layer, which were more resistant to solute and solvent transfer under higher pressure, where they were thicker.