All patients (1,328) suspected of having transient ischemic attacks (TIA) who came to six institutions over a 21-months period were identified. Each case was reviewed by a neurological investigator, and a definite diagnosis of TIA was supported in 39% and ruled out in 30%. The reliability of the neurologist's review diagnosis was assessed by three methods, and close agreement (84% to 93%) was obtained. Factors demonstrated to effect the diagnosis were historical information, neurological training of the examiners, type of symptom complex, presence of carotid artery bruits, and examination during an attack.