Techniques for assessment of the colloidal fouling propensity of RO membrane feed waters are usually based on filtration tests with less tight membranes. In an effort towards improving such techniques, prototype experiments are performed under constant pressure, employing the same feed waters, with both RO and UF membranes. Fouling species representative of two typical categories of RO membrane foulants are employed, i.e. humic acids and polysaccharides (organic) as well as colloidal iron oxide (inorganic), in the concentration range 0.75-20 mg/L, at various fluid salinities. The RO fouling experiments confirm that there exists a linear dependence of fouling rates on foulant concentration, which permits extrapolation of the new results to lower concentrations, usually encountered in practice. The fouling behavior is analyzed in an effort to relate RO and UF test results. It is found that the specific cake resistances of deposited colloidal iron are much smaller in dead-end UF tests compared to those obtained in RO cross-flow experiments. This difference is smaller in the case of humic acids, whereas UF and RO specific cake resistances are in close agreement for alginate foulants, when corrected for incomplete species rejection by UF and deposition on RO membranes. Four different factors to account for the observed differences, namely pressure, UF membrane rejection, fluid shear, and the effect of cake-enhanced concentration polarization, are examined using additional targeted experiments and theoretical tools. It is concluded that, in the range of conditions tested, the first three factors are important and need to be carefully considered in a systematic way in order to obtain reliable predictions of RO fouling on the basis of UF filtration data. (c) 2009 Elsevier B.V. All rights reserved.