International guidelines for cardiopulmonary resuscitation (CPR) in adults advocate that cardiac arrest be recognized within 5-10 s, by the absence of a pulse in the carotid arteries. However, validation of first responders' assessment of the carotid pulse has begun only recently. We aimed (1) to develop a methodology to study diagnostic accuracy in detecting the presence or absence of the carotid pulse in unresponsive patients, and (2) to evaluate diagnostic accuracy and time required by first responders to assess the carotid pulse. In 16 patients undergoing coronary artery bypass grafting, four groups of first responders (EMT-1: 107 laypersons with basic life support (BLS) training; EMT-2: 16 emergency medical technicians (EMTs) in training; PM-1: 74 paramedics in training; PM-2: 9 certified paramedics) performed, single-blinded and randomly allocated, carotid pulse assessment either during spontaneous circulation, or during non-pulsatile cardiopulmonary bypass. Time to diagnosis of carotid pulse status, concurrent haemodynamics and diagnostic accuracy were recorded. In 10% (6/59), an absent carotid pulse was not recognized as pulselessness. In 45% (66/147), a pulse was not identified despite a carotid pulse with a systolic pressure > or = 80 mmHg. Thus, although sensitivity of all participants for central pulselessness approached 90%, specificity was only 55%. Both sensitivity and, to a lesser degree, specificity improved with increasing training; blood pressure or heart rate had no significant effect. The median diagnostic delay was 24 s (minimum 3 s). When no carotid pulse was found, delays were significantly longer (30 s: minimum 13 s), than when a carotid pulse was identified (15 s; minimum 3 s) (P < 0.0001). Of all participants, only 15% (31/206) produced correct diagnoses within 10 s. Only 1/59 (2%) identified pulselessness correctly within 10 s. Our cardiopulmonary bypass model of carotid pulse assessment proved to be feasible and realistic. We conclude that recognition of pulselessness by rescuers with basic CPR training is time-consuming and inaccurate. Both intensive retraining of professional rescuers and reconsideration of guidelines about carotid pulse assessment are warranted.