The Paks loess-paleosol sequence is one of the most important terrestrial records of Middle and Late Pleistocene environmental changes in East Central Europe, spanning the last ca. 0.8 Ma. While geochemical proxies demonstrate a general decreasing chemical weathering trend over the last 0.8 Ma in the Carpathian Basin, mineralogy and derived indices reflect intensifying physical erosion. In theory, the observed chemical weathering trend can be accounted for both by enhanced input of relatively unweathered material and by climate deterioration during the Quaternary, as the used proxies like CIA are not capable of distinguishing between pre- and post-depositional weathering. Enhanced physical erosion of the source areas, driven by tectonism, and resulting increased sedimentation of fresh mineral dust at the depositional site are demonstrated by increasing dolomite, illite and chlorite contents and sme/ill, sme/(ill+chl) ratios from older to younger sediments in the profile, together with increasing thickness of loess layers towards the youngest part of the sequence. At the same time, constant smectite contents (30-40%) in paleosols appear to disprove progressive aridization of interglacials through time and suggest that the duration of pedogenesis played an important role in determining soil types. Further, the increasing proportion of inherited phyllosilicates (illite and chlorite) would, in theory, raise the possibility that the decreasing values of chemical weathering indices are just artifacts of enhanced physical erosion and resulting increased dust deposition by a dilution effect. The above findings highlight the fact that our general view on chemical weathering is oversimplistic, as its ’equation’ includes two basic variables, tectonism and time beyond climate and the interplay of these equally important factors will eventually determine its final value. To get a better grasp of these processes one needs further data (more age control in loess profiles, data on uplift in and around sedimentary basins) and more sophisticated proxies, as the mineralogical data presented here can be considered only semi-quantitative. Regarding the provenance of sediments in the Paks profile, geochemical data demonstrate that felsic rocks dominated the source areas and there have been only very little variations in provenance over the last ca. 0.8 Ma. Significant contributions from mafic/ultramafic rocks to the sediments can be ruled out as revealed by lower abundances of ferromagnesian trace elements. The appearance of amphiboles and high dolomite contents suggest that loess material was at least partly sourced from local rocks and geochemical data reveal a genetic link between floodplain sediments and loess deposits.