Two natural source waters containing natural organic matter (NOM) with different physical and chemical characteristics were dead-end filtered using five types of membranes having different material and geometric properties. In this study retained dissolved organic carbon (DOC) per unit membrane area is introduced as a better parameter compared to permeate volume, time, and delivered DOC to provide a reasonable comparison of NOM rejection and flux-decline trends. Retained DOC/NOM was calculated, which influences NOM concentration polarization at the membrane interface, and transport measurements of NOM rejection and flux decline were made. Molecular weight (MW) distribution measurements (by size exclusion chromatography) were used to calculate the average MW of the NOM. This persuasively demonstrated that the nominal molecular weight cut-off (MWCO) of a membrane is not the unique predictor of rejection characteristics for NOM composites. The charge density of NOM from the source waters was measured to estimate its effects on NOM rejection and flux decline during filtration. The contact angle of the membranes was used to determine hydrophobic interactions between NOM and membrane. All filtration measurements were performed at approximately the same permeate flow rate in order to minimize artifacts from mass transfer at the membrane interface. ESNA having a nominal MWCO of 200 Daltons showed NOM rejection greater than 95% and flux decline lower than 10% under a condition of a retained DOC of 0.5 mg C/cm(2) for the feed source waters. The other membranes having larger membrane pores (nominal MWCOs ranging from 8,000 to 20,000 Daltons) than the ESNA showed NOM rejection ranging from 68% to 86% and flux decline ranging from 5% to 17% at the same retained DOC for the waters.