There is good evidence for a genetic causation of a significant proportion of the human epilepsies. These genetic epilepsies include a large number of rare mendelian syndromes in which epilepsy usually occurs as part of a more complex neurological phenotype, and the more common familial epilepsies in which seizures occur in isolation but inheritance is not mendelian. The molecular basis of these epilepsies is entirely unknown. The current revolution in genetics and neuroscience is, however, providing methods for the molecular genetic analysis of epilepsy. Two principal strategies exist. The first, positional cloning, requires localization of the disease genes by linkage analysis followed by analysis of transcripts within the disease-gene region. The second, candidate gene analysis, comprises direct screening of cloned genes which may on theoretical grounds have a role in epilepsy. At present, two epilepsy genes have been localized by linkage analysis: that for benign familial neonatal convulsions to chromosome 20, and that for progressive myoclonic epilepsy of Unverricht and Lundborg to chromosome 21. Linkage analysis of non-mendelian epilepsies is more difficult, but may be feasible with the generation of high-resolution linkage maps of the human genome. A number of plausible candidate genes for the epilepsies have now been cloned, including in particular the ligand-gated and voltage-gated ion channels. An understanding of the genetic epilepsies at a molecular level will provide new precision in diagnosis and genetic counselling, and may allow new strategies for pharmacological control of seizures.