A rational approach to the design of antiherpetic nucleoside analogues is based in part on the broad specificity of virus-coded thymidine kinases. Herpes virus thymidine kinase 'activates' many 5-substituted 2'-deoxyuridines, analogues of thymidine (e.g., idoxuridine, trifluridine, edoxudine, brivudine), 5-substituted arabinofuranosyluracil derivatives (e.g., 5-Et-Ara-U, BV-Ara-U, Cl-Ara-U), acyclonucleosides of guanine (e.g., aciclovir, ganciclovir, penciclovir), and purine nucleosides with the pentafuranosyl ring replaced by a cyclobutane ring (e.g., cyclobut-G, cyclobut-A). Activation involves selective, and frequently regiospecific, phosphorylation of these analogues to the 5'-monophosphates. These are further phosphorylated by cellular enzymes to the 5'-triphosphates, which are usually competitive inhibitors of the viral-coded DNA polymerases. Some analogues are also incorporated into viral, and to a lesser extent cellular, DNA. A recent, unusual, exception is human cytomegalovirus, which does not code for a thymidine kinase, but for a protein with the sequence characteristics of protein kinase and which phosphorylates ganciclovir to its 5'-monophosphate. The interaction of the analogues with cellular catabolic enzymes such as uridine and thymidine nucleoside phosphorylases is also discussed, as is the relationship between physicochemical properties (configuration, conformation, electronic and hydrophobic parameters) and antiviral activities, with particular reference to those drugs that are licensed, or under consideration, for clinical use.