Pollen records from fluvial deposits in The Netherlands have traditionally been used to establish continental stages for the Neogene in NW Europe. These stages are still widely used across Europe for stratigraphic correlations and palaeoclimatic studies. Despite their extensive use, the validity of this continental correlation scheme for the Plio- and Pleistocene has never been tested in the area where they were originally proposed: the Roer Valley Graben (RVG). The fluvial and deltaic depositional setting in this area is controlled by a complex pattern of tectonic and climatic processes, which strongly complicates the definition of stages and sub-stages of regional or supra-regional significance. Here, the original palynological data used to define the Praetiglian, Reuverian, Brunssumian, Susterian, and 'Linne' (sub-)stages are assessed and compared to new borehole data. With a multivariate (Correspondence Analysis) approach we show that most of these stages cannot be distinguished by numerical analysis of the pollen assemblages. A tripartition of pollen assemblages is possible, which results in a cooler Pleistocene cluster, warm-temperate to sub-tropical Pliocene cluster, and a warm and relatively dry Late Miocene to Early Pliocene cluster. However, no further stages, sub-stages or clear climatic phases can be consistently recognized in these data. A detailed lithostratigraphic framework of the study area allows a correlation of the sections independent of the pollen data. Results confirm the numerical analyses and furthermore point to a strong lithological (depositional) bias on the pollen assemblages. Variations in the abundance of Sequoia and Taxodium, originally used to define the Brunssumian/Reuverian transition and Brunssumian A-C sub-stages, are shown to represent mostly local changes. The number and absolute percentage of the fluctuations in Sequoia and Taxodium abundances depend largely on the sample resolution and the site location within the RVG. Our data assessment indicates that long-distance chronostratigraphic correlations based on the original continental Neogene stages likely are invalid. Furthermore, the palaeoclimatic reconstructions that have been based on the RVG successions are largely over-interpreted. The present palaeoclimate reconstructions from the RVG reflect variations in the highly-dynamic fluvial depositional setting, taphonomic effects and environmental conditions, including climate. A sequence-stratigraphic approach combined with the analysis of interfingering marine sediments is potentially better suited to establish absolute ages and associated climatic conditions of the Neogene deposits within the RVG. © 2007 Elsevier B.V. All rights reserved.