Nonribosomal peptide synthetases (NRPSs) are giant multi-domain enzymes that catalyse the biosynthesis of many commercially important peptides produced by bacteria and fungi. Several studies over the last decade have shown that many of the individual domains within NRPSs exhibit significant substrate selectivity, which impacts on our ability to engineer NRPSs to produce new bioactive microbial peptides. Adenylation domains appear to be the primary determinants of substrate selectivity in NRPSs. Much progress has been made towards an empirical understanding of substrate selection by these domains over the last 5 years, but the molecular basis of substrate selectivity in these domains is not yet well understood. Perhaps surprisingly, condensation domains have also been reported to exhibit moderate to high substrate selectivity, although the generality of this observation and its potential impact on engineered biosynthesis experiments has yet to be fully elucidated. The situation is less clear for the thioesterase domains, which seem in certain cases to be dedicated to the hydrolysis/cyclization of their natural substrate, whereas in other cases they are largely permissive.