The heuristic-direct QSAR (quantitative structure-activity relationships) approach has been applied to a series of 34 muscarinic receptor ligands, including antagonists, weak partial agonists, partial agonists and full agonists, interacting with the human ml-muscarinic receptor subtype. The first step of this procedure consists of the computer-aided 3D-model building of the receptor. The second step involves docking simulations with selected ligands, maximizing the complementarity between ligand and receptor. In the third step, a detailed and extensive correlation analysis between the computed interaction energies, their components and the experimental pharmacological affinity and action is accomplished in order to evaluate the consistency of the QSAR model proposed and to provide a quantitative tool for comparisons among the different complexes considered. In this context, good linear correlations have been obtained between ad hoc theoretical intermolecular interaction descriptors and the pharmacological action, which allow one to classify quantitatively and predict the pharmacological action of new ligands. Finally, according to the ml-receptor model proposed, it has been possible to speculate on the amino acid residues which are mainly involved in the interaction with the ligands, and on the nature of the prevailing intermolecular interactions which are responsible for the different behaviour of antagonists, weak partial agonists, partial agonists and full agonists.