The purpose of the present study was to evaluate the cardiopulmonary resuscitation (CPR) skills of medical students after a 2-h basic life support class (n = 129) and 6 months later (n = 113). Mean +/- SD written test score decreased from 6.4 +/- 0.7 to 6.2 +/- 0.8 (P = 0.03). Mean +/- SD breaths delivered before CPR decreased from 2.9 +/- 0.6 to 2.2 +/- 1.2 (P = 0.0001), ventilation rate increased from 12.2 +/- 1.9 to 14.3 +/- 5.0 breaths/min (P = 0.0001), tidal volume increased from 0.75 +/- 0.2 to 0.8 +/- 0.31 (P = 0.11), minute ventilation from 9.1 +/- 2.6 to 10.8 +/- 3.61 (P = 0.0001), and stomach inflation from 13 +/- 22 to 18 +/- 27% of CPR breaths (P = 0.11). Mean +/- SD chest compression/min decreased from 56 +/- 9 to 54 +/- 12 (P = 0.34), depth of chest compression increased from 41 +/- 6 to 46 +/- 7 mm (P = 0.0001), hands held incorrectly on the thorax increased from 22 +/- 27 to 23 +/- 32% (P = 0.59), and leaning on the chest from 4 +/- 12 to 18 +/- 28% of compressions (P < 0.0001). In summary, ventilation skills were unpredictable; there was only a 5% chance that a given student would achieve the same mouth-to-mouth ventilation performance in both the BLS class and 6 months later. Despite the respiratory mechanics of the CPR manikin which prevented stomach inflation much better than an unconscious patient with an unprotected airway, stomach inflation occurred repeatedly. Teachers of basic life support classes need to consider the respiratory mechanics of the CPR manikin being used to assure clinically realistic and appropriate mouth-to-mouth ventilation skills.