We released bobcats (Lynx rufus) on Cumberland Island, Georgia during 1988 and 1989 as part of a cooperative effort to restore an extirpated predator to the island. We collected data on prey use and prey abundance three times a year during 2 y following the initial bobcat releases. We tested four hypotheses concerning bobcat prey selection: (1) use of a prey species was closely associated with its abundance (functional relationships), (2) the number of prey species included in bobcat diets increased as the abundance of principal prey species decreased (diet optimization), (3) the number of species included in bobcat diets and diet diversity increased as population density increased (interference) and (4) increases in the proportion of males resulted in increases in the use of large prey and decreases in the use of small prey. Bobcats' use of marsh rabbits (Sylvilagus palustris) and cotton rats (Sigmodon hispidus) was positively correlated with each species' abundance as predicted by a hypothesis of functional relationships. The number of prey species in bobcat diets and diet diversity was negatively correlated with the abundance of marsh rabbits as predicted by a hypothesis of diet optimization. Changes in diet diversity in different regions of the island between years were not correlated with changes in bobcat density within each region, suggesting interference was not occuring at the bobcat densities observed in our study. Changes in the use of Eastern grey squirrels (Sciurus carolinensis) between years were negatively correlated with increases in the proportions of male bobcats in different regions of the island. Our results suggest that diet optimization and functional responses are both useful models for describing bobcat-prey relationships. The diet optimization model had greater power to explain use of prey species that were not used in all seasons.