Abstract Dramatic progress was seen in 14C-dating with the introduction of accelerator mass spectroscopy (AMS) which made possible the direct dating of prehistoric artworks painted or drawn with charcoal. However, in the case of engravings and red paintings, only indirect methods can be used that allow us to date deposits that have covered the works over time (TL, U/Th, oxalates, etc.). The uranium/thorium dating method gives reliable and relatively precise results in the case of massive speleothems, because the sampling is carried out at the heart of the material where the hypothesis of a closed system (that is, no exchange with the outside environment) is justified in most cases. Unfortunately, the situation is quite different in the case of thin layers of calcite that overlie Palaeolithic cave drawings. The conditions under which calcite forms depend largely on the hydrologic activity, which has greatly varied over the course of the Upper Palaeolithic and Holocene. In many cases, we can see that the growth of speleothems stopped during much of the Upper Palaeolithic. Consequently the ages obtained are minimum ages (terminus ante quem) which are frequently much younger than the real ages of the underlying artworks. Moreover, a much more serious but rarely considered source of error contradicts the assumption of a closed system. In thin layers of carbonate deposits and in damp media, the uranium incorporated into the calcite during its crystallization may be partially eliminated because of its solubility in water. Uranium leaching causes an artificial increase of the age that may reach considerable proportions (e.g. a negative hand in a cave in Borneo was dated to 27,000 years by U/Th whereas its 14C age was only 8–10,000 cal BP; Plagnes et al., 2003). Due to these two contradictory sources of error, the dates given by the U/Th method may prove to be younger or older, with deviations that are much larger than the standard deviations given by laboratories. As a result it is nearly impossible and very dangerous to base archaeological reasoning on U/Th ages of Palaeolithic artworks, so long as the dates are not confirmed by an independent method, dating the carbonates in the same samples by 14C being the best means of detecting anomalies. The application of the U/Th method for the dating of prehistoric rock art is still experimental. Technical improvements (for less damageable sampling) and fundamental research on the causes of errors are needed.