Our earlier results concerning bacterial transport of an adhesion-deficient strain Comamonas sp. (DA001) in intact sediment cores from near South Oyster, VA demonstrated that grain size is the principle factor controlling bacterial retention, and that Fe and Al hydroxide mineral coatings are of secondary importance. The experimentally determined collision efficiency (α) was in the range of 0.003-0.026 and did not correlate with the Fe and Al concentration. This study attempts to theoretically predict α, and identifies factors responsible for the observed low α. The modified Derjaguin-Landau-Verwey-Overbeek (DLVO) theory was used to calculate the total intersurface potential energy as a function of separation distance between bacterial and sediment surfaces and to provide insights into the relative importance of bacterial and sediment grain surface properties in controlling magnitude of α. Different models for calculating theoretical α were developed and compared. By comparing theoretical α values from different models with previously published experimental α values, it is possible to identify a suitable model for predicting α. When DA001 bacteria interact with quartz surfaces, the theoretical α best predicts experimental α when DA001 cells are reversibly attached to the secondary minimum of the energy interaction curve and α depends on the probability of escape from that energy well. No energy barrier opposes bacterial attachment to clean iron oxide surface of positive charge at sub-neutral pH, thus the model predicts α of unity. When the iron oxide is equilibrated with natural groundwater containing 5-10 ppm of dissolved organic carbon (DOC), its surface charge reverses, and the model predicts α to be on the order of 0.2. The theoretical α̇ for DA001 in the natural sediments from South Oyster, VA was estimated by representing the surface potential of the sediment as a patch-wise binary mixture of negatively charged quartz (ζ=-60 mV) and organic carbon coated Fe-Al hydroxides (ζ=-2 mV). Such a binary mixing approach generates α that closely matches the experimental α. This study demonstrates that it is possible to predict α from known bacterial and grain surface properties. Copyright © 2001 Elsevier Science B.V.