The substrate-like 'canonical' inhibition by the 'small' serine proteinase inhibitors and the product-like inhibition by the carboxypeptidase inhibitor have provided the only atomic models of protein inhibitor--proteinase interactions for about 15 years. The recently published structures of cystatin/stefin--papain complexes and of hirudin--thrombin complexes reveal novel non-substrate-like interactions. In addition, the structure of pro-carboxypeptidase shows a model of inactivation which bears resemblance to proteinase/protein inhibitor systems. Considerable progress in understanding the transition between native and cleaved states of the serpins has also been made by several recent structural studies.