In this paper, we present a theoretical method to describe molecular structures in terms of hierarchically related substructures. The approach is based on the location of local maxima (peaks) in promolecular electron density distributions (EDD) established at continuously varying resolution levels. For each of the so-calculated EDD, the local maxima are determined by using a hierarchical clustering algorithm wherein peaks obtained at a given resolution are used as starting points for discovering peaks at the next lower resolution level through gradient trajectories of the EDD. The use of such an approach allows assignment of molecular fragments or chemical groups to peaks, at any resolution level. Results, obtained for a set of four benzodiazepine-related molecules and three thrombin inhibitors, are presented in terms of dendrograms wherein each node corresponds to a well-defined molecular substructure.