Mostly, fractionation in laboratory experiments have been carried out with single or binary model protein solutions. The question has arisen whether these experiments can be representative for industrial and natural biological solution fractionations. In this study, a comparison of single, ternary, and natural egg-white solutions is made. Thus, fractionation of ternary mixtures of ovalbumin, conalbumin, and lysozyme and natural egg-white protein solutions at different pH and two ionic strengths was studied with unmodified and UV modified polysulfone ultrafiltration membranes. The modified membranes had an increased initial water flux and their zeta potentials were more negative than those of the unmodified membranes. The UV modified membranes became more hydrophilic due to the formation of carboxylate and sulphonate groups. In ultrafiltration of single protein solutions the highest flux reduction and the lowest protein retention were obtained with ovalbumin at its isoelectric point (pH 4.8). At this pH lysozyme and conalbumin were positively charged and highly retained because of effective size exclusion due to charge repulsion. Also, in fractionation of ternary mixtures and natural egg white solutions, ovalbumin was the major protein that could permeate the membranes at pH 4.8. Ovalbumin was highly retained due to charge repulsion at all other tested pH values. Retention of ovalbumin and transmission of lysozyme increased in ultrafiltration of egg-white solutions in the presence of salt at pH 4.8. It could be seen clearly that the behaviour of ovalbumin being the most abundant protein, mainly determined the fractionation properties of the mixtures. Comparison of the results from fractionation of solutions of ternary model proteins and natural egg-whites showed that retention was lower and flux reduction smaller in UF of the natural egg-white solutions. This was probably due to interaction of proteins or/and salts presents in the natural egg-white solutions, which could not be modelled by the main protein components. On the whole it seemed as the proteins studied behaved much in the same way in mixtures as separately.