In 1994, all emergency medical services (EMS) ambulance officers in Singapore were trained to perform pre-hospital defibrillation with semi-automated external defibrillators (AED). All non-traumatic cardiac arrest patients over 10 years old were included, excluding those who were obviously dead and children below 36 kg. The data were collected by the ambulance officers according to the Utstein guidelines. From 1 February 1994 to 31 January 1999; resuscitation was attempted in 968 non-trauma cardiac arrests. Fifteen percent of the cases were of non-cardiac origin. The overall survival rate was 40/968 (4.1%, 95% CI 2.9-5.6%). Of 968 patients, 22/136 (16.2%, 95% CI 10.4-23.5%), 18/622 (2.9%, 95% CI 1.7-4.5%) and 0/210 (0%, 95% CI 0-1.7%) survived in the EMS witnessed, bystander witnessed and un-witnessed groups, respectively (P < 0.001). Within the EMS witnessed group, those with an initial rhythm of VF/VT had a higher survival rate (30.6%) than those without VF/VT (4.1%). P < 0.001, OR = 10.3, 95% CI 2.9-36.9. Similarly, the VF/VT survival rate in the bystander witnessed group (4.5%) was higher than the non-VF/VT (1.0%) (P = 0.011, OR = 4.4, 95% CI 1.3-15.4). The survival rate of patients with bystander witnessed VF/VT arrest who received bystander CPR was 9.4% compared to 1.0% in those who did not (P = 0.037, OR = 4.4, 95% CI 1.01-20.1). Our survival rate of bystander witnessed VF/VT arrest is comparable to large metropolitan cities in the USA. The determinants of survival include EMS witnessed arrest and VF/VT arrest. Increased quantity and quality of bystander CPR rate may improve the outcome in bystander witnessed cardiac arrest.