We have investigated a method of teaching community CPR in three stages instead of in a single session. These have been designated bronze, silver, and gold stages. The first involves only opening of the airway and chest compression with back blows for choking, the second adds ventilation in a ratio of compressions to breaths of 50:5, and the third is a conversion to conventional CPR. In a controlled randomised trial of 495 trainees we compared the performance in tests immediately after instruction of those who had received a conventional course and those who had had the simpler bronze level tuition. The tests were based on video recordings of simulated resuscitation scenarios and the readouts from recording manikins. Differences occurred as a direct consequence of ventilation being required in one group and not the other, some variation probably followed from unforeseen minor changes in the way that instruction was given, whilst others may have followed from the greater simplicity in the new method of training. A careful approach was followed by slightly more trainees in the conventional group whilst appreciably more in the bronze group remembered to shout for help (44% vs. 71%). A clear advantage was also seen for bronze level training in terms of those who opened the airway as taught (35% vs. 56%), for checking breathing (66% vs. 88%), and for mentioning the need to phone for an ambulance (21% vs. 32%). Little difference was observed in correct or acceptable hand position between the conventional group who were given detailed guidance and the bronze group who were instructed only to push on the centre of the chest. The biggest differences related to the number of compressions given. The mean delay to first compression was 63 s and 34 s, and the mean duration of pauses between compressions was 16 s and 9 s, respectively. Average performed rates were similar in the two groups, but more in the conventional group compressed too slowly whereas more in the bronze group compressed too rapidly. Observations were made for only three cycles of compression, but extrapolating these to the 8 min often considered a watershed for chances of survival for victims of cardiac arrest, an average of 308 compressions would be expected from those using conventional CPR compared with 675 for those using bronze level CPR. The implications of this difference are discussed.