Herbicide resistance is selected for within a weed population through the recurrent use of an herbicide. Once the use of the herbicide is reduced or discontinued, the trait may persist in the population unless resistance endows a fitness penalty. Few studies have examined the long-term persistence of an herbicide resistance trait in a weed population once the selection pressure that led to its prevalence has been removed. The objective of this research was to re-survey the locations described by A.J.J. Smisek in her 1995 study of a paraquat resistant biotype of C. canadensis on the grounds of the Harrow Research and Development Centre. Results indicate that, ∼20 yr after paraquat use was significantly reduced, a glyphosate resistant biotype had replaced the paraquat resistant biotype at all but one of the original locations surveyed by Smisek. The sole location where the paraquat resistant biotype was observed was an old, un-treed orchard row currently managed by mowing. This location not only served as a refuge for the paraquat resistant biotype, ensuring its persistence in the face of strong selection for glyphosate resistance at all other locations, but it facilitated the co-occurrence of glyphosate and paraquat resistant biotypes. The observation of a multiple resistant biotype at this location alone, with resistance to both paraquat and glyphosate, suggests a role for refugia in the stacking of herbicide resistance traits. Given that this stacking occurred in the absence of either herbicide, we hypothesize that the multiple resistant biotype arose through pollen mediated gene flow among biotypes.