The Neurotoxicity Scale was devised as a patient-based report scale to assess the adverse effects of antiepileptic drugs on cognitive function. In a previous report we reported the clinical validity of the scale, tested in a double-blind randomized study, using a benzodiazepine in normal volunteers. In the present study, the clinical sensitivity, construct validity and reliability of the scale was tested in patients with epilepsy. Patients (n = 189), selected from both participating centres, representative for the patients with chronic epilepsy were included in the study. Reliability was tested with Cronbachs alpha and yields an almost maximal score (.95). Clinical sensitivity was compared with the previous normal volunteer study and was evaluated as satisfactory. Construct validity showed a five-factor structure, explaining 66.5% of the variance, with 'fatigue and slowing' as the dominant factor. In line with the assumptions for this scale and with the results obtained in normal volunteers, the scale appears to be unsuitable for differential assessment of type or severity of drug-induced impairment. The most valid primary outcome measure is the overall score that renders a global ('all or nothing') evaluation indicating that a subject experiences cognitive impairment and associates this with the antiepileptic treatment. Other factors that may impair cognitive function, such as seizure frequency do not influence this score. The scale has therefore maximal applicability as a screening instrument in outpatient practice and in early (phase II, IIIa) drug trials.