Dams are often used for storing mine tailings. These structures present failure mechanisms that can lead to major risks for surrounding populations. In view to managing these risks, it is necessary to take a probabilistic approach in predicting their behaviour. The application of such approaches is limited by the difficulty of obtaining experimental data to estimate the variability of the parameters and conditioned by the relevance of the probability models chosen to represent this variability. This article proposes (1) a probabilistic modelling of the index properties of the mine tailings constituting these dams based on statistical analyses and (2) a method using dynamic penetration tests to estimate on site the mine tailings friction angle and its variability. This method, applied to chilean tailings dams proposes a single model for all tailings dams in order to associate a probability law to the effective friction angle (φ′). The procedure is illustrated on the probabilistic study of slope stability carried out at the global scale of a dam and also at the local scale of each of its constituent layers. © 2011 Springer Science+Business Media B.V.