It may fairly be claimed that up to the last decade no antiepileptic drug (AED) had undergone rigorous testing. The development programmes of the new AEDs registered in recent years have necessarily been innovative, and methods of AED testing are still undergoing rapid evolutionary change. Clinical evaluation of AEDs is both difficult and complex, due mainly to two factors: (1) intermittence of clinical events, which means that dosing for periods of several weeks is generally necessary, leading to problems of poor compliance and inaccurate reporting of events by carers and patients; and (2) therapeutic necessity, which means that it is, in general, unacceptable to withhold effective treatment from a person with epilepsy. Consequently monotherapy, either with a trial drug or with placebo, can rarely be justified. In consequence most phase II trials use add-on therapy which in turn causes various problems. Conventional phase II AED trials are usually placebo-controlled add-on studies employing either a parallel or crossover design. The latter is subject to a number of practical and theoretical objections, notably on grounds of carry-over and order effects. Increasing attention has recently been directed to ethically acceptable monotherapy designs. One approach first exploited in the development of felbamate is the performance of monotherapy trials in patients whose AEDs have been withdrawn as part of a preoperative assessment protocol. Other possibilities for achieving monotherapy are also discussed.