The Silt Density Index (SDI) is commonly applied as a measure for the fouling potential of particles in RO and NF feed waters. The accuracy and reproducibility of the SDI test is increasingly questioned. In this work, the influence of membrane properties on the SDI value is investigated. Eight commercial '0.45 mu m' membrane types made of different materials (PVDF, PTFE, acrylic copolymer, nitro cellulose, cellulose acetate, nylon 6,6, and polycarbonate) were used to measure the SDI. Three samples were randomly chosen from each membrane type (same lot), and several membrane properties were studied (pore size distribution, pore shape, surface and bulk porosity, thickness, surface charge, contact angle and surface roughness). SDI values for an artificial feed, composed of a solution of a alumina particles of 0.6 part diameter, were determined. The characterization of these membranes shows variation between the membranes used in this study (M1-M8), and within a batch of one membrane type. Substantial differences were found in the SDI values for the different types of membrane filters used. Pore size, porosity and thickness are the most important membrane properties and determine the membrane resistance. Using a membrane with high a membrane resistance results in a low SDI value. The variations in measured SDI values between batches and within a batch are large and explain, at least partly, the problems encountered in practice with unacceptable variations in SDI values. These observed differences make the test unreliable. The variations are attributed to differences in properties of the membranes used. In order to make the SDI a reliable fouling index, there is a very strong need for membrane filters with uniform and constant properties. (C) 2011 Elsevier B.V. All rights reserved.