Stable isotope analysis has a long history in Caribbean archaeology. The Caribbean region, however, possesses a highly complex isotopic ecology, including both a large number of isotopically variable food sources, and a high degree of isotopic overlap between different food groups. As such, to date, most regional paleodietary studies have been limited to descriptive and qualitative conclusions concerning the relative contributions of different food sources. In this study we apply an iterative Bayesian multi-source mixing model (FRUITS) to skeletal stable isotope data from the prehistoric population of Tutu, St. Thomas, USVI, to test the feasibility of such models to generate quantitative and probabilistic individual paleodietary reconstructions. The isotope data set includes both bone collagen (δ13Cco and δ15Nco) and apatite (δ13Cap) data. The results of two different dietary models using four and five distinct food groupings, respectively, are compared and assessed relative to other relevant archaeological evidence pertaining to past diet at the site. We highlight the potentials and limitations of multi-source mixing models for regional paleodietary studies, and their relevance to ongoing debates within Caribbean archaeology concerning the relative importance of different food sources such as manioc, maize, and seafood.