Cardiopulmonary resuscitation guidelines are constantly optimized to increase survival. Keeping hands-off time brief is vital. Our hypothesis is that rhythm recognition is time-consuming during cardiopulmonary resuscitation. A Laerdal Sim-Man simulated three shockable and four nonshockable rhythms. Rhythms were presented to physicians who identified whether they were shockable and whether they would defibrillate. We measured time to stated decision. Thirty-five doctors participated, 32 had completed advanced life support training. The mean time to make a decision on whether to defibrillate or not was 3.4 s [95% confidence interval (CI): 2.8-3.9] for shockable and 4.4 s (95% CI: 3.6-5.3) for nonshockable rhythms (P<0.05). For all rhythms, the mean time was 4.0 s (95% CI: 3.5-4.5). Of all shockable rhythms, 95.2 % were correctly diagnosed as shockable, compared with 88.6 % of nonshockable rhythms being correctly diagnosed. Our simulation study indicates that doctors are able to correctly identify shockable rhythms within 4 s.