A crucial parameter deciding the clinical utility of new antiepileptic drugs (AEDs) is the therapeutic index expressing the margin between anticonvulsant and adverse effects. The latter is commonly quantified during preclinical testing in the rotarod test in normal, healthy rodents. However, the validity of using normal animals for adverse effect predictions in epilepsy patients is questionable. Limbic kindling of rodents induced by corneal kindling of mice and amygdala kindling of rats confirm that epileptic animals are more susceptible to the behavioral and cognitive alterations following acute administration of NMDA antagonists and certain established AEDs. This appears to represent a permanent reactivity specific for limbic kindling since it is absent in rats after chemical kindling with pentylenetetrazole. Animal species with inborn epilepsy, including audiogenic and photosensitive animals, are not revealing an enhanced susceptibility to the behavioral alterations induced by NMDA antagonists. In contrast, these induce severe adverse effects in genetic absence epilepsy rats where certain AEDs also are associated with a more marked deterioration of motor function than in normal animals. This appears in line with several complications with AED use in man being linked to an interaction with the dysfunction of the brain imposed by the epileptic condition. Thus, it is important to involve epileptic animals in preclinical adverse effect testing, in particular when evaluating new AED candidates with novel or unknown mechanisms. In that respect, limbic kindling appears to represent a sensitive and relevant approach.