What actually occurred during the two last links in the 'chain of survival': defibrillation and advanced life support (ALS), was studied in 156 patients with cardiac arrest of cardiac aetiology using the computer recording of the defibrillator and the Utstein-style data record. Ten patients (6%) survived. The ECG artefacts caused by chest compressions enabled a detailed analysis of compression rates (median 108 min(-1)) and duration of important compression free periods. The time from initiation of monitoring during asystole until chest compressions were initiated was median 29 s, significantly shorter than during electromechanical dissociation (EMD, 109 s; P < 0.001). These times were both significantly longer than the median time from initiation of monitoring until the first shock was given in cases with VF (19 s; P < 0.001). A total of 883 shocks (median six shocks) were administered to 110 patients with a significant difference in number of shocks between survivors and non-survivors, one versus seven, respectively. The success rate for the first shock and all shocks defined as non-VT/VF 5 s after the shock, was 75 and 63%, respectively. However, just 10% of all shocks resulted in a rhythm with a pulse and only 4% resulted in sustained return of spontaneous circulation (ROSC). An isoelectric period followed 38% of the shocks, and in 27% this lasted more than 20 s, with five patients obtaining electrical activity with a pulse after more than 30 s of isoelectric ECG. Thoracic impedance did not affect the shock efficacy. The method of analysing resuscitation we describe may be useful for quality improvement.