This work aims to test the reliability of calcined bones for radiocarbon dating of the Paleolithic. Fifty-five calcined bone samples coming from Aurignacian and Gravettian layers at Abri Pataud (Dordogne, France) were selected based on their macroscopic features. For each sample, the heating state was estimated on the basis of bone crystallinity (splitting factor [SF] using FTIR) and δ13C value. Twenty-seven bone samples (3 unburnt and 24 calcined) from 5 different levels were prepared for 14C dating. The majority (15/24) of the calcined samples had to undergo a sulfix treatment prior to graphitization, probably due to the presence of cyanamide ion in these samples. The comparison between our results and recently published dates on bone collagen for the same levels shows that unburned bone apatite is systematically too young, while a third of the calcined bones fall within or very near the range of expected age. No clear correlation was found between 14C age offset and δ13C value or SF. Most of the sulfixed samples (14/16) yielded ages that were too young, while almost all of the non-sulfixed samples (8/9) gave ages similar or <0.2 pMC from the expected minimum age. Although preliminary, these results suggest that sulfix should be avoided if possible and that clean CO2 gas from well-calcined Paleolithic bones can provide reliable 14C ages.