Bacterial transport through porous media was modeled using detachment functions that incorporate the dependence of detachment rate on bacterial residence time on the collector. Model parameters and the relative merit of alternative forms for the detachment function were evaluated on the basis of comparisons between model simulations and experimentally derived bacterial breakthrough and elution curves. Only detachment functions that provided an initial period in which bacteria were rapidly released, followed by slow bacterial detachment, were able to reproduce the elution portion of the breakthrough curves. In optimal simulations, 90% of the bacteria that were captured by the porous medium detached within 1 min of attachment. Experiments involving saturated flow through columns packed with sand indicated that the time to achieve complete breakthrough was inversely related to the influent bacterial concentration. On this basis and because of the relatively slow approach to breakthrough that was typically observed in transport experiments, it was hypothesized that the experimental medium contained a number of preferred attachment sites that must be essentially filled before breakthrough is achieved. Only when such (irreversible) sorption sites were included in the model formulations was it possible to produce transport simulations that matched both the breakthrough and elution portions of the empirically derived curves. It is concluded that both a time-dependent detachment function and a degree of sorption site heterogeneity are required to describe bacterial attachment and detachment during transport as observed in our laboratory. ¿ American Geophysical Union 1995 ¿ American Geophysical Union 1995