This paper describes a new method for the elucidation of the disulphide bonding pattern in a protein from an initial set of unrefined nuclear magnetic resonance solution structures. The use of both local and global proton-proton nuclear Overhauser enhancement (NOE) distance information for the identification of the disulphide bridge network in cysteine-rich polypeptides was investigated by statistical analysis of the crystal structures of a selected group of proteins. There are six different types of inter-cysteine proton-proton distances which can potentially be used for the prediction of disulphide links. The uniqueness and the extent to which disulphide bonds could be identified by these distances was evaluated. Only NOEs between C beta H/C beta H and C alpha H/C beta H were shown to have positive predictive values for the characterization of disulphide links. Contrarily, the observation of an NOE between C alpha H and NH is a strong indication for the absence of a disulphide bridge between the two residues. The global analysis of the nuclear magnetic resonance data starts with the calculation of an initial set of conformers. First, pairing weights wij were assigned to all putative cysteine pairs in the protein according to a Gaussian-type distribution function from the C beta-C beta interatomic distances. In a second step, all conceivable disulphide patterns were formed by an exhaustive combinatorial enumeration. Statistical weights were then assigned to all patterns from the weights of the participating cysteine pairs. This method was validated with protein crystal structures deposited in the Brookhaven Protein Data Bank having three or more cysteine residues. It was then used to determine the previously unknown disulphide bonding pattern of the 12 cysteine residues of flavoridin.