In a multicenter registry conducted over 2 yr of patients with acute respiratory distress syndrome (ARDS), we enrolled 153 patients and collected data daily for 7 consecutive days and weekly thereafter until death or hospital discharge. The purposes of the registry were (1) to determine whether a more liberal definition of ARDS (PaO2/FIO2 < or = 250; bilateral pulmonary infiltrates within 7 days) than those commonly used would result in enrollment of patients earlier in their clinical course, and (2) to study the clinical course of the syndrome in survivors and nonsurvivors. The mortality rate was 54% and it was significantly greater in older versus younger patients (75% versus 37%) and in septic versus nonseptic patients (60% versus 43%). We found that the definition of ARDS used for the registry resulted in enrollment of patients 1 to 7 days earlier than was the case when other published definitions of ARDS were applied to the patient population. Fewer than 2% of the patients failed to meet one of the nonregistry definitions of ARDS within 7 days. The mortality rate was independent of the definition used to identify ARDS patients. Our results suggest that a more liberal definition of ARDS than those commonly used can result in identification of the same population of patients earlier in their clinical course.