This study was carried out to test the suggestion that close interactional and linguistic examination of the communication between neurologists and patients during a first encounter can contribute to the differential diagnosis of epilepsy or psychogenic nonepileptic seizures. Twenty unselected patients admitted for video/EEG telemetry because of diagnostic uncertainty were included. Two linguists blinded to all medical data independently studied video recordings and transcripts of 25- to 35-minute interactions. They attempted to predict the medical diagnosis on the basis of qualitative assessments addressing 17 separate observations. They also used a diagnostic scoring aid (DSA) to convert their qualitative assessments into a simple numeric score. Using qualitative assessment, both linguists predicted 17 of 20 (85%) diagnoses (kappa=0.59). With the DSA, diagnoses were predicted with a sensitivity of 85.7% (71.4%) and a specificity of 84.6% (92.3%). This blinded, prospective multirater study confirms the diagnostic value of linguistic and interactional observations in the seizure clinic setting.