Archaeologists interested in radiocarbon dating shell midden sites express concern regarding the accuracy of shell dates and how such determinations should be interpreted. This article discusses the problem of dating shells from sites in the southeastern United States. New results are presented comparing shell, bone, and soil-charcoal age determinations from the Crystal River site, located along the west-central Gulf Coast of Florida. Crystal River is a large multimound site whose occupants engaged in long-distance exchange throughout eastern North America during the Woodland period (∼1000 BC to AD 1050). In the summer of 2012, test units were excavated in several contexts at the site, including both mounds and occupation areas. Samples were collected for 14C dating, which were then processed at the University of Georgia Center for Applied Isotope Studies. This article focuses on samples from the stratified shell midden, from which it was hoped to construct a local correction for marine shell that could be used to date other contexts. The soil-charcoal and bone collagen from these samples have very similar ages (bone samples ranging from about 100 cal BC to cal AD 530 and soil-charcoal from cal AD 345 to 560); however, the shell samples collected from the same stratigraphic units are significantly older than the terrestrial dates (ranging from 1300 to 390 cal BC). The difference in calibrated ages between organic materials and the shells ranges between 560 to 1140 yr. This phenomenon cannot be explained solely by the marine reservoir effect. It appears that all the shell samples formed in mixed marine (∼50–60%) contexts, as indicated by the stable isotope ratios and the amount of atmospheric carbon remaining in the samples. The age of the shell samples cannot be used to date archaeological events as they are influenced not only by the marine reservoir effect, but also the local hardwater effect, which makes them significantly older.