Teeth are usually targeted for dating archaeological sites because they are less prone to dissolution, in comparison with bones. However, despite this apparent resistance, teeth do undergo diagenesis, which needs to be accounted for in order to obtain accurate ages. In particular, the uptake of trace elements such as uranium in dental tissues needs to be considered for dose rate determination when dated using electron spin resonance(ESR). Characterising the mineralogy and structural integrity of samples prior to dating may thus provide important information related to their state of preservation, especially in the case of teeth whose U content can significantly affect the dose rate. In this study, we dated five teeth of small-sized bovids using combined ESR/U-series dating. They were collected at the Middle Stone Age site of Lovedale, located in the central interior of South Africa. Micromorphology provided sedimentary context to the samples, which were recovered from a layer of gravel rich in faunal remains. Using cathodoluminescence, laser-induced fluorescence, Fourier transform infrared spectroscopy and Raman micro-spectroscopy we assessed the degree of preservation of the enamel. Results reveal that carbonate hydroxyapatite underwent post-depositional alteration, based on its molecular structure and elemental composition. Although the teeth all originate from the same layer and were sampled in the same 1-m square and at asimilar elevation, U-content in the enamel differs highly from one tooth to the other, with values ranging from 1.7 to 29.6 ppm. These values are correlated with equivalent doses (De) from 228 to 923 Gy and are consistent with variations in crystallinity determined with vibrational spectroscopy. We also investigated the possible saturation of the ESR signal, by repeating measurements with microwave power values from 1 to 20 mW.Despite such diversity in U-content, the ages calculated assuming an early uptake of U all fall within the same range, from 63 ± 8 ka to 68 ± 15 ka and may only represent a minimum estimate.< Réduire