This randomised controlled trial used a manikin model of cardiac arrest to compare skill performance in untrained lay persons randomised to receive either compression-only telephone CPR (Compression-only tel., n=29) or standard telephone CPR instructions (Standard tel., n=30). Performance was evaluated during standardised 10 min cardiac arrest simulations using a video recording and data from a laptop computer connected to the training manikin. A number of subjects in both groups did not open the airway. More than 75% in the Standard tel. group failed to deliver two effective initial rescue breaths, and only 17% provided an adequate inflation volume for subsequent breaths, delivering a median of only five inflations during the entire scenario. Most subjects in both groups gave chest compressions that were too shallow and at an inappropriately rapid rate. Hand position was also poor, but was worse in the group given simplified instructions. There was a significant delay to first compression in both groups, although this interval was shortened by over a minute when ventilations were eliminated from the telephone instruction algorithm (245 vs. 184 s, P<0.001). Over two-and-a-half times as many chest compressions were delivered during an average ambulance response time with compression-only telephone directions compared with standard CPR (461 vs. 186, P<0.001). These variables may be critical in predicting survival from out-of-hospital cardiac arrest. Further research is necessary to establish if modifications to scripted telephone instructions can remedy the identified performance deficiencies. Eliminating instructions for rescue breaths from scripted telephone directions will have little impact on the ventilation of most patients. Research is required to determine if the consequent reduction in the delay to starting chest compressions and the significant increase in the number of compressions delivered can increase survival from out-of-hospital cardiac arrest.