A substantial complication to using the oxygen isotope composition (δ¹⁸O) of vertebrate bioapatite in paleoclimate studies is the need to distinguish variation due to temporal changes in the δ¹⁸O of surface waters from that due to temperature-dependent fractionation during biomineralization. One solution is multiple-taxon comparisons using data from coexisting homeothermic and heterothermic animals. Fossil emydid turtles have been suggested to be potentially useful as functional homeotherms because (1) modern emydids employ behaviors, such as basking, to restrict skeletal growth to a narrow temperature range; (2) their aquatic habitat constrains the isotopic variability of dietary inputs; and (3) emydids have a dense fossil record. But because turtles lack teeth and therefore tooth enamel, sampling must focus on bone, which is potentially more susceptible to diagenetic alteration. This study examines the δ¹⁸O of carbonate (δ¹⁸O_c) and phosphate (δ¹⁸O_p) in hydroxylapatite from co-occurring emydids and heterotherms (crocodilians and gars) from the Paleocene–Eocene of the Clarks Fork Basin, Wyoming. Previous isotopic studies of this area provide an extensive data set for comparison with the results of this study. Bone and enamel δ¹⁸O_c values measured here exhibit a greater range (16‰–32‰ Vienna Standard Mean Ocean Water) than previously observed, suggesting alteration, while the range of δ¹⁸O_p values (9‰–15‰) is within that predicted by presumably unaltered mammalian tooth enamel δ¹⁸O_c. While high crystallinity indices (0.28–0.55) and a lack of covariation between δ¹⁸O_c and δ¹⁸O_p suggest alteration of one or both of these constituents, a strong correlation between crocodilian enamel and bone δ¹⁸O_p suggests bone phosphate may be reliable.¹