Abstract Statistical analysis of bed thickness was performed for sampled turbidite successions from well-documented architectural elements of the Gres d' Annot Formation to characterize confined deep-water mini-basins of the Tertiary foreland basin of SE France. The purpose was to use advanced statistical processing techniques in order to evaluate whether a discrimination of different architectural elements is feasible through observed statistical signatures of bed thickness. Statistical methods were focused on: i) fitting of widely used non-normal theoretical distribution models using robust non-parametric goodness-of-fit statistical tests, and ii) detecting the possible presence of non-random bed thickness clustering using existing and new clustering estimation methods. Results indicate that the bed thickness data are best characterized by a multi-modal lognormal distribution model which probably reflects a background sedimentological process. Several datasets exhibit power law as well as exponential thick-bedded tails. The data also exhibit non-random clustering of bed thickness. Discrimination of architectural elements in this confined turbidite succession seems to be feasible based on the characteristics of the observed composite lognormal distributions such as number and variability of the detected components. The estimation of the degree of facies clustering has potential for the discrimination of architectural elements in confined basin settings if used in conjunction with alternative estimation methods (such as periodogram estimation). This methodology may now be applied to other confined turbidite successions, be they outcrops with less certain architecture, or subsurface datasets with borehole imaging.