Problem:​ The diagnosis of epilepsy largely relies on the seizure description by a witness. Our aim was to assess the accuracy of seizure descriptions.

Methods:​ Twenty volunteers (10 medical students, 4 junior doctors working on a neurological ward and 6 non-medical students) viewed a video of a partial then secondary, i.e. generalised seizure, and were then asked to provide a written account of the event. The seizure had eight key features. Volunteers scored one mark for each described key feature. One mark was deducted for each false observation.

Results:​ The mean positive score was 3.5 (range 1 to 6). Unresponsiveness and lateralising features were often missed. The mean negative score was −0.8 (range 0 to −3). Erroneously described features included ‘patient rolled over’, ‘agitated’ or ‘arms flopped about’ as part of the tonic clonic seizure. Left and right were sometimes confused. The mean total score was 2.7 (range −2 to 6). A medical and a non-medical student achieved the highest scores, a doctor the lowest score.

Conclusions:​ The accuracy of seizure descriptions by witnesses was generally low and there were wide variations.