Abstract We investigated the spatial variability of the heavy-mineral composition in the Early Pleistocene fluviatile Kedichem Formation in the Netherlands in order to meet the demand for more information about subsurface sediment composition. We first determined the spatial extension and thickness of the sediment body, then used Fuzzy clustering techniques on a database containing approx. 2000 heavy-mineral counts from the Kedichem Formation to map the spatial extension of the various sediment provenances within the formation. Three clusters could be discerned, one representing a combined Meuse-Scheldt source, the other two representing a mixed Rhine-Baltic source. We made slice maps at several depths through the formation, and plotted the cluster memberships. The maps show an overall dominance of the Meuse-Scheldt source in the south of the Netherlands, whereas the Rhine-Baltic source occurs mainly in the central Netherlands. The methods employed show that it is possible to map and study the 3-D variation in heavy-mineral composition and hence sediment provenance in the Dutch subsurface with the use of simple statistical and visualization techniques.