In view of the evolving indications for device therapy in atrial arrhythmia the accurate detection of high rate atrial events is a necessity. In a sheep model of atrial fibrillation (AF) we observed a contradictory behavior of the Thera DR pacemaker. The pacemakers were programmed to deliver burst pacing on detection of sinus rhythm (SR). Paradoxically, progressively more inappropriate bursts were delivered during AF, at a higher sensitivity. This implied that the pacemaker interpreted AF as SR. We assessed the atrial detection of the Thera DR, Diamond, Saphir, and Marathon pacemakers during AF, in a sheep model and in vitro using a waveform generator. By counting the annotated atrial-sensed events reported by the pacemaker we charted the behavior of the pacemakers at different sensitivities. At a higher sensitivity both the Thera DR and the Diamond paradoxically reported fewer atrial events during AF. This behavior led to inappropriate mode switching and incorrect diagnostic data collection. It could be reproduced in vitro. The Marathon did not show this paradoxical undersensing in vivo or in vitro. This paradoxical undersensing at a high sensitivity is dependent on the amplitude of the input signal and could not be explained by an overlap of programmable timing parameters. It is caused by ringing of the atrial sensing amplifier. At high atrial rates and a high sensitivity the ringing of the repetitive atrial signal input results in blanking of the atrial amplifier. This observation may be relevant in the device management of patients with paroxysmal AF.