The effects of yeast cells on membrane fouling by a protein mixture were studied in dead-end filtration. A 0.2 mu m cellulose acetate membrane was used with a 1 g/l protein mixture consisting of equal amounts of bovine serum albumin, lysozyme, and ovalbumin. Yeast cells were used either in suspension or as preformed yeast cakes on top of the membrane. A small concentration of 0.022 g/l yeast cells in suspension enhanced the permeate flux and maintained protein transmission at nearly 100%, compared with a 60% reduction in the protein concentration in the permeate obtained after 3 h for the protein mixture filtered alone. Higher suspended yeast concentrations of 0.043 and 0.18 g/l resulted in lower fluxes and intermediate values for the protein transmission. For the three different thicknesses of preformed yeast cakes studied (0.025, 0.05, and 0.10 cm), the cake with intermediate thickness resulted in protein transmission of nearly 100% and the highest permeate flux. The thinner yeast cake resulted in a lower permeate flux, but it maintained protein transmission at nearly 100%, whereas the thicker cake resulted in a reduction in both permeate flux and protein transmission. The mechanism proposed to explain the results is based on the formation of a secondary membrane by the yeast cells on top of the original membrane. This secondary membrane entraps protein aggregates, which would otherwise cause membrane fouling and reductions in permeate flux and protein transmission,