Records on 1,297 people with witnessed out-of-hospital cardiac arrest, caused by heart disease and treated by both emergency medical technicians (EMTs) and paramedics, were examined to determine whether or not early cardiopulmonary resuscitation (CPR) initiated by bystanders independently improved survival. Bystanders initiated CPR for 579 patients (bystander CPR); for the remaining 718 patients, CPR was delayed until the arrival of EMTs (delayed CPR). Survival was significantly better (P less than 0.05) in the bystander-CPR group (32%) than in the delayed-CPR group (22%). Multivariate analysis revealed that the superior survival in the bystander-CPR group was due almost entirely to the much earlier initiation of CPR (1.9 minutes for the Bystander-CPR group and 5.7 minutes for the delayed-CPR group; P less than 0.001). There were significantly more people with ventricular fibrillation (VF) in the bystander-CPR group (80%) than in the delayed-CPR group (68%); and, for people in VF, the survival rate was significantly better if they had received bystander-CPR (37% versus 29%). The authors conclude that early initiation of CPR by bystanders significantly improves survival from out-of-hospital cardiac arrest, and they suggest that it may do so by prolonging the duration of VF after collapse and by increasing cardiac susceptibility to defibrillation. The benefit of this early CPR, however, appears to exist within a rather narrow window of effectiveness. It must be started within 4-6 minutes from the time of collapse and must be followed within 10-12 minutes of the collapse by advanced life support in order to be effective.