The availability of new, highly selective antagonists, in the field of opioid peptides and of other pain peptides, is important both for a better understanding of the interaction of the receptors with their ligands and for their practical relevance. The design of antagonists is not obvious even when the essential features of agonists are well known. In this review we have examined the main aspects of the problem using, as leading criteria two theoretical models of antagonism and the subdivision of opioid peptides into two functional domains. The main causes of antagonism have been integrated in two very general models: one, referred to as the participation model, attributes antagonism to the lack, with respect to the parent agonist, of an essential group, whereas another model, attributes antagonism to the misfit of the molecule inside the receptor. The second criterion is the division of the structure of peptide hormones, originally put forward by Robert Schwyzer, in two functional domains, the message domain, which is responsible of the larger part of the binding affinity of opioid agonists, and an address domain, which dictates most of the peptide specificity. The most significant achievements in the design of opioid antagonists are classified according to the relative importance of chemical constitution, conformation and chirality.