Rapid identification of stroke subtype is valuable for both practicing clinicians and the optimal design of clinical stroke trials. Mechanisms of ischemic injury might differ among different stroke subtypes. Certain subtypes might be clinically identified as suboptimal for future therapeutic or prophylactic stroke trials. Some subtypes might be so clinically distinct that extensive laboratory investigation is unwarranted. Investigators in the ongoing Trial of ORG 10172 in Acute Stroke Treatment are using criteria to categorize stroke etiology among enrolled patients into one of five subtypes: large-artery atherothromboembolic, cardioembolic, small-vessel thrombotic, other etiology, or undetermined etiology. As part of the study, physicians initially predict the most likely subtype of stroke based on clinical features and baseline CT. Three months after stroke, investigators use the criteria, which also incorporate results of diagnostic testing, to reclassify stroke subtype. Initial clinical impression of subtype agreed with final determination in 62% of patients, and the rate was similar for all stroke subtypes. No stroke subtype was more accurately diagnosed than others by initial assessment. No subtype was more commonly identified by diagnostic studies. Fifteen percent of patients remained without a clear etiologic subtype diagnosis at 3 months. We conclude that clinical trials in stroke should not attempt to restrict entry into trials based on presumed stroke subtype. A careful evaluation for etiology is justified in all patients presenting with stroke, regardless of presumed subtype.