The interobserver agreement for the diagnosis of a cerebral transient ischemic attack (TIA) was investigated in a pool of eight senior and interested neurologists from the same department. They interviewed 56 patients in alternating pairs. The diagnosis was based on internationally accepted criteria. The agreement rates were corrected for chance (kappa statistics). Both neurologists agreed that 36 patients had a TIA and 12 had not, but they disagreed about 8 patients (kappa = 0.65; for perfect agreement kappa would be 1.0). The vascular territory (carotid or vertebrobasilar) was agreed upon in only 24 of the 36 patients in whom both diagnosed TIA's (kappa = 0.31). We concluded that currently the diagnosis of a TIA, made by a single neurologist, is a poorly defined entity.