Epilepsy represents the most common serious neurological disorder, with a prevalence of 0.4 - 1%. Approximately 30% of patients are resistant to currently available drugs. New anti-epileptic drugs are needed to treat refractory epilepsy, improve upon current therapies, improve the prognosis of epilepsy and to prevent the epileptogenic process. Designing compounds with specific physiological targets would seem the most rational method of anti-epileptic drug development, but results from this approach have been disappointing; the widespread screening of compounds in animal models has been much more fruitful. Older methods of animal screening have used acute seizure models, which bear scant relationship to the human condition. More modern methods have included the development of animal models of chronic epilepsy; although more expensive, it is likely that these models will be more sensitive and more specific in determining anti-epileptic efficacy. In this review, we consider the possible physiological targets for anti-epileptic drugs, the animal models of epilepsy, problems with clinical trials and ten promising anti-epileptic drugs in development (AWD 131-138, DP16 (DP-VPA), ganaxolone, levetiracetam, losigamone, pregabalin, remacemide, retigabine, rufinamide and soretolide). Perhaps the most important advances will come about from the realisation that epilepsy is a symptom, not a disease. Preclinical testing should be used to determine the spectrum of epilepsies that a drug can treat, and to direct later clinical trials, which need to select patients based on carefully defined epilepsy syndromes and aetiologies. Not only will such an approach improve the sensitivity of clinical trials, but also will lead to a more rational basis on which to treat.