Studies report a decline in the psychomotor memory of cardiopulmonary resuscitation (CPR) providers within months of training, but they are prone to subject bias. We hypothesized that this degradation is faster and more prevalent in real world practice. The aims of our study were to 1. assess the quality of chest compressions (CC) delivered routinely by CPR-certified clinicians who are not primed by study conditions, and 2. investigate if psychomotor memory degrades if feedback devices are removed. Forty anaesthetists and intensivists participated in a voluntary, half-day, randomized crossover study using case-based simulation. Participants were paired and randomly assigned into 2 groups; each receiving automated feedback either in the first or second cycle of CPR. Two cycles of CC and defibrillation (ACLS protocol) were administered on a manikin. CC parameters including overall quality were measured by a feedback device. The median proportion of good quality CC was poor at baseline but improved with feedback; 38.2% (IQR 27.7, 58.7) to 57.7% (IQR 38.0, 68.7), P < .05. The median proportion of good quality CC fell after feedback withdrawal; 50.5% (IQR 24.5, 67.7) to 25.6% (9, 37.6), P < .05. No carryover effect was observed. Treatment effect and period effect were detected. Baseline quality of CC amongst frequent CPR providers is poor, and can be improved partly by feedback devices. As psychomotor memory of good quality CCs degrades rapidly after removal of feedback, a multimodal approach is required for CPR skill retention. Future research on the optimal frequency of CPR training, including the use of feedback devices in clinical practice should be explored.