The aim of this study is to evaluate the accuracy of three-step diagnosis in discriminating subtypes of acute ischemic stroke. A total of 120 consecutive patients with first-ever ischemic stroke, admitted to one general hospital, were prospectively studied. In the first step (within 24 hours of clinical onset), the first diagnosis was made according to clinical symptoms and signs, and patients were subdivided into four groups according to the classification of Oxfordshire Community Stroke Project: lacunar infarcts (LACI), total anterior circulation infarcts (TACI), partial anterior circulation infarcts (PACI), and posterior circulation infarcts (POCI). In the second step (24 hours to 72 hours from the onset), neuroimaging diagnosis was performed by CT and/or MRI. Four lesion sites were classified: 1) small subcortical infarction < or = 1.5 cm in diameter in the perforating artery territory (SSI), 2) supratentorial cortical or striatocapsular infarction (CI), 3) low-flow infarction (LFI) which includes centrum semiovale infarct and internal junctional infarct, and 4) posterior circulation infarction other than SSI (PI). In the third step, etiological diagnosis was made by examination including trans-thoracic echocardiography and MRA (3-D, PC). In accordance with the TOAST Study, the presumed stroke mechanism was categorized as either small-vessel occlusion (lacune), cardioembolism (CE), large-artery atherosclerosis (LAA), or others. The majority of patients with TACI, PACI or POCI showed the corresponding lesions on CT or MRI, while only 69% of LACI patients demonstrated SSI. Seventy-five percent of patients with TACI were categorized as CE in the third diagnosis, while the etiology of the patients with PACI was either CE or LAA in equal numbers. Only 60% of LACI patients were classified as lacune and 21% of them as LAA. Patients with LACI but classified as LAA usually had atypical clinical symptoms (e.g. monoparesis) and lesions other than SSI. The positive predictive value (PPV) of lacune in the combination of LACI and SSI was 0.75. Eighty-two percent of patients with CE had atrial fibrillation (af), which was the most frequent cardioembolic source. When patients with TACI or PACI had af, the PPV of CE was 0.93, but when they did not, the PPV of LAA was only 0.68. The etiology of POCI was variable. In conclusion, the agreement of the three-step diagnosis is considerable, but more rigorous clinical examination is needed for some clinical groups (POCI and LACI) and the etiological diagnosis of LAA.