Criteria for selecting patients for possible surgery in the management of intracerebral hemorrhage (ICH) are needed to plan a prospective therapeutic evaluation of surgical intervention. This study specifically addressed patients seen in the emergency room within a few hours of the ictus, many of whom were still awake, to identify and subsequently exclude from surgical procedure those expected to recover completely and those expected to die regardless of treatment. We retrospectively studied 75 patients evaluated at a mean time of 3 hours and 37 minutes after hemispheric ICH to determine factors that would predict both good and poor outcomes at the time of discharge. Eighty percent of our patients presented within 6 hours of symptom onset. These patients were younger and had more severe lesions than did those presenting later, yet most were still awake (mean admission Glasgow Coma Scale [GCS] score = 11.0). Using multivariate regression, we created two models. The first model predicts independent outcome, ie, Rankin 0 to 2, of all patients with a GCS score greater than 9 on admission who do not undergo surgery. The significant factors in this model were hemorrhage diameter, intraventricular extension, and age. The second model predicts poor outcome, ie, Rankin 5 and death, of all patients. GCS score, hemorrhage volume, age, and gender were the important factors in this model. We conclude that ICH patients presenting early to the emergency room have more severe lesions radiologically, although their initial clinical status may not be different from those seen late. Our models should identify and thus exclude those with very good and very poor prognoses from future randomized surgical trials.