Seizures are resistant to treatment with currently available anticonvulsant drugs in about 1 out of 3 patients with epilepsy. Thus, there is a need for new, more effective anticonvulsant drugs for intractable epilepsy. Furthermore, because of the inadequacy of the currently available anticonvulsant armamentarium with respect to safety, newly developed drugs should be less toxic than existing drugs. Previous and current strategies for development of novel anticonvulsants with improved efficacy or safety are critically discussed in this review. 'Old drugs' (or 'first generation' drugs), which were developed and introduced between 1910 and 1970, are compared with new anticonvulsants both in terms of clinical efficacy and safety and in terms of mechanisms of action. The new drugs are referred to as 'second generation' drugs, i.e. anticonvulsants which have been introduced into clinical practice in recent years, or 'third generation' drugs, i.e. compounds in the pipeline of development. In spite of some 30 years of 'modern' neuroscientific epilepsy research, most novel, clinically effective second generation anticonvulsants have been found by screening (i.e. serendipity) or structural variation of known drugs and not by rational strategies based on knowledge of processes involved in generation of seizures or in development of epilepsy. An exception are only the GABA (gamma-aminobutyrate)-mimetic drugs vigabatrin and tiagabine and, to some extent, gabapentin, which have been developed by a rational strategy, i.e. the 'GABA hypothesis' of epilepsy. The fact that preclinical seizure models used for identification and development of novel drugs have been originally validated by old drugs, i.e. conventional anticonvulsants, may explain that several of the new drugs possess mechanisms which do not differ from those of the standard drugs. This may also explain that none of the new drugs seems to offer any marked advantage towards the old, first generation drugs with respect to the ultimate goal of drug treatment of epilepsy, i.e. complete control of seizures, although some of the second generation drugs may have benefits in terms of side effects and tolerability. It is to be hoped that the various novel currently used or planned strategies for drug development produce more effective and safe anticonvulsants than previous strategies. This goal can only be achieved by strengthening our understanding of the fundamental pathophysiology of seizure expression and epileptogenesis as theoretical substrates for new pharmacological strategies, and by devising and refining laboratory models for studying new agents obtained by such strategies.