Nanofiltration (NF) membranes have been shown to effectively remove low-molecular-weight compounds, but predicting their rejection has posed challenges due to the fact that multiple mechanisms can be involved (e.g. adsorption, steric-hindrance effects, and electrostatic effects) that concurrently control removal. The objective of this study was to systematically evaluate which factors were responsible for steady-state rejection of selected antibiotics such as sulfamethoxazole (SMX) and carbadox (CDX) for three NF membranes of varying tightness through a unique modeling approach which accounts for the fact that these compounds bear distinctive acid-base speciation properties under varying pH conditions. Both the steric-hindrance/pore-diffusion model for neutral solutes along with the Donnan steric partitioning model for charged solutes were integrated together to predict overall rejection. This is the first known study to evaluate compound rejection of compounds that carry such unique speciation properties as a function of pH by using such a combined approach. Antibiotic rejection was found to vary with both pH and membrane tightness. Rejection controlled by charge repulsion effects was accurately predicted for both negatively charged solutes such as SMX- and CDX- and positively charged solutes such as SMX+. However, rejection controlled by steric-hindrance effects for neutral solutes either overestimated (e.g. SMX) or underestimated (e.g. CDX) rejection by up to 23%. These findings improve the understanding of which transport mechanisms influence rejection of both neutral and charged low-molecular-weight compounds by NF membranes so that removal can be better predicted. (C) 2011 Elsevier B.V. All rights reserved.