To evaluate the influence of quality of bystander cardiopulmonary resuscitation (CPR) on outcome in prehospital cardiac arrest we consecutively included patients with prehospital cardiac arrest treated by paramedics in a community run ambulance system in Oslo, Norway from 1985 to 1989. Good CPR was defined as palpable carotid or femoral pulse and intermittent chest expansion with inflation attempts. Outcome measure was hospital discharge rate. One hundred and forty-nine of 334 patients (45%) received bystander CPR. The discharge rate after good BCPR (23%) was higher than after no good BCPR (1%, P < 0.0005) or after no BCPR (6%, P < 0.0005). There was no difference between no good and no BCPR (P = 0.1114). There were no differences in paramedic response interval between the groups, but the mean interval from start of unconsciousness to initiation of CPR (arrest-CPR interval) was significantly shorter in the group receiving good bystander CPR (2.5 min, 95% confidence interval (CI): 1.7-3.3 min) than no good CPR (6.6 min, CI: 5.2-8.0 min) or no bystander CPR (7.8 min, CI: 7.2-8.4 min). Bystanders started CPR more frequently in public than in the patient's home (58 vs. 34%, P < 0.0005). Good bystander CPR was associated with a shorter arrest-CPR interval and improved hospital discharge rate as compared to no good BCPR or no BCPR.