Microfiltration is used in a wide range of municipal and industrial settings to remove particulate matter including pathogenic microorganisms such as bacteria and protozoa. As more water is filtered at constant pressure, the accumulation of retained particles on the membrane decreases the filtration rate; a process commonly referred to as fouling. Mathematical treatment of flux decline has proved to be a useful tool in diagnosing filtration data even though the mathematical underpinnings are not completely understood. In particular, little is known about the transition between fouling phenomena (e.g. pore blocking to cake filtration). Moreover, less is known about the effect of extracellular polymeric substances (EPS) production by bacteria when they accumulate on a membrane over extended durations. In this manuscript, we develop a novel approach to model bacterial microfiltration by considering the effects of both differential binding and exopolymer production. Spatial gradients in bacteria concentrations initially occur due to the non-uniform membrane surface porosity and differential deposition caused by the stochastic nature of microorganism adhesion. These heterogeneities in bacterial deposition and associated pore blocking result in variable secretion of extracellular polymers. Long-term fouling is quantified as the cumulative resistance posed by both bound bacteria and EPS. We compare numerical simulations quantitatively and qualitatively to previously published experimental data and investigate variations of microbial deposition patterns across the membrane. We find substantial agreement between the model and experimental observations. We are also able to conclude that fluid dynamics must be important if the dominant variability is in the membrane structure, rather than in bacterial adhesion. However, variation in bacterial adhesion alone can also induce substantial spatial heterogeneity. (C) 2009 Elsevier B.V. All rights reserved.