The splitting of one biological species into two, speciation, is a central process in evolutionary biology. Because species are defined as reproductive communities that are isolated from one another, speciation requires the evolution of reproductive isolation. A controversial idea known as ‘speciation by reinforcement’ posits that natural selection can favour increased reproductive isolation when hybridisation is costly, for example when genetic incompatibilities have accumulated between populations or when divergent selection has generated local adaptation. This idea has evolved since it was first suggested and the term ‘reinforcement’ now covers a range of more or less distinct processes. I wish to clarify the differences and similarities among these processes in order to underpin research aimed at determining the extent to which speciation relies on forms of reinforcement. I will consider various dimensions that characterise different forms of reinforcement, such as the nature of selection against hybrids or the genetic basis of the assortment that might increase isolation. Starting at a conceptual level, I will then relate my classification to existing theory and empirical examples. This may require more precise understanding of what is meant by reproductive isolation, a core idea in speciation research but one that is rarely given a formal definition.