The data mining/groundwater modeling methodology developed in McDade et al. (2013) was performed to determine if matrix diffusion is a plausible explanation for the lower-concentration but persistent chlorinated solvent plumes in the groundwater-bearing units at three different pump-and-treat systems. Capture-zone maps were evaluated, and eight wells were identified that did not draw water from any of the historical source areas but captured water from the sides of the plume. Two groundwater models were applied to study the persistence of the plumes in the absence of contributions from the historical source zones. In the wells modeled, the observed mass discharge generally decreased by about one order of magnitude or less over 4 to 10 years of pumping, and 1.8 to 17 pore volumes were extracted. In five of the eight wells, the matrix diffusion model fit the data much better than the advection dispersion retardation model, indicating that matrix diffusion better explains the persistent plume. In the three other wells, confounding factors, such as a changing capture zone over time (caused by changes in pumping rates in adjacent extraction wells); potential interference from a high-concentration unremediated source zone; and limited number of pore volumes removed made it difficult to confirm that matrix diffusion processes were active in these areas. Overall, the results from the five wells indicate that mass discharge rates from the pumping wells will continue to show a characteristic “long tail'' of mass removal from zones affected by active matrix diffusion processes. Future site management activities should include matrix diffusion processes in the conceptual site models for these three sites. © 2013 Wiley Periodicals, Inc.