Anticipating that patients with large ischemic lesions exceeding one-third of the middle cerebral artery (MCA) territory and detected on CT scans within 6 hours of stroke onset will not benefit from thrombolysis, we trained each participant of the second European Cooperative Acute Stroke Study (ECASS II) in the recognition of early ischemic lesions. Participants (n=532; neurologists, radiologists, neuroradiologists) were tested before and after each 4-hour CT reading training course. We asked the participants to estimate the extent of acute ischemic lesions on 10 CT scans, which we presented for 90 seconds without clinical information. Two sets of 10 CT scans each (A and B) were alternatively presented to each group, so that 254 participants evaluated set A before the training and 278 participants evaluated set B. We compared the numbers of correct estimates, underestimations, and overestimations before and after the course for each participant. The person who taught all courses (RvK) provided the reference estimates. We found that training significantly increased the number of correct estimates (p < 0.0001). Subsequently, we studied the incidences of large infarctions and parenchymal hemorrhages in the ECASS II population. In comparison with ECASS I investigators, the local investigators of ECASS II reduced the number of falsely included patients to an extent similar to that of the training courses. More remarkably, among the ECASS II patients, the proportion of patients with large infarctions or parenchymal hemorrhages was reduced to 50%. Careful CT reading may have contributed to this result.