A 7-year monitoring period of rare earth element (REE) concentrations and REE pattern shapes was carried out in well water samples from a 450 m long transect set up in the Kervidy/Coët-Dan experimental catchment, France. The new dataset confirms systematic, topography-related REE signatures and REE concentrations variability but challenges the validity of a groundwater mixing hypothesis. Most likely, this is due to REE preferential adsorption upon mixing. However, the coupled mixing-adsorption mechanism still fails to explain the strong spatial variation in negative Ce anomaly amplitude. A third mechanism-namely, the input into the aquifer of REE-rich, Ce anomaly-free, organic colloids-is required to account for this variation. Ultrafiltration results and speciation calculations made using Model VI agree with this interpretation. Indeed, the data reveal that Ce anomaly amplitude downslope decrease corresponds to REE speciation change, downhill groundwaters REE being mainly bound to organic colloids. Water table depth monitoring shows that the colloid source is located in the uppermost, organic-rich soil horizons, and that the colloid input occurs mainly when water table rises in response to rainfall events. It appears that the colloids amount that reaches groundwater increases downhill as the distance between soil organic-rich horizons and water table decreases. Topography is, therefore, the ultimate key factor that controls Ce anomaly spatial variability in these shallow groundwaters. Finally, the <0.2 µm REE fraction ultimately comes from two solid sources in these groundwaters: one located in the deep basement schist; another located in the upper, organic-rich soil horizon.