The efficacy of bystander CPR in resuscitation from cardiac arrest when defibrillation is available within five to six minutes has been questioned. Epidemiologic studies from different cities have shown conflicting results. We conducted a study to determine the effect of early CPR versus no CPR on resuscitability, 24-hour survival, and neurologic deficit in an animal model of cardiac arrest. Twenty-two mongrel dogs were subjected to five minutes of electrically induced ventricular fibrillation. In 11 dogs, closed-chest massage and ventilation with room air was begun immediately and was continued for five minutes. The other 11 dogs received no CPR. At five minutes defibrillation was attempted and advanced cardiac life support (ACLS) protocols were followed until the animal was resuscitated or died. No statistical difference in resuscitability or 24-hour survival between the two groups was demonstrated. Eight of 11 "early CPR" animals were resuscitated and survived 24 hours; six of 11 "no CPR" dogs were resuscitated, and five lived for 24 hours. A significant difference was demonstrated by the Student t test in neurologic deficit and ease of resuscitation. "Early CPR" dogs had no neurologic deficit, while "no CPR" dogs had a 41% deficit (P less than .01). "Early CPR" dogs were resuscitated in significantly less time once ACLS was started (29 versus 317 seconds), and required less electrical energy (100 versus 560 J), fewer countershocks (1.3 versus 4.0), and less epinephrine (0.1 versus 1.7 mg) than did "no CPR" animals. In this animal model of cardiac arrest, early CPR was shown to be beneficial to neurologic function and ease of resuscitation, even when ACLS was provided within five minutes.