Automatic mode switching (AMS) allows patients with dual chamber pacemakers who develop paroxysmal AF to have a controlled ventricular rate. The aim of this study was to (1) compare the rate-controlled behavior of three AMS algorithms in response to AF, in terms of speed and stability of response and resynchronization to sinus rhythm, and (2) compare the influence of pacemaker programming on optimal mode switching. We studied 17 patients (12 men, 5 women; mean age 59 +/- 15 years) who developed AF during electrophysiological study. Unfiltered bipolar atrial electrograms during sinus rhythm and AF were recorded onto high fidelity tapes and replayed into the atrial port of three dual chamber pacemakers with different mode switching algorithms (Thera, Marathon, Meta). The Thera pacemaker uses rate smoothing, and mode switches occur when mean sensed atrial rate exceeds the predefined AMS rate (MR). Marathon mode switches after a programmable number of consecutive rapid atrial events (NR). Meta DDDR monitors the atrial rate by a counter for atrial cycles faster than the programmed AMS rate. It increases or decreases the counter if the atrial cycle length is shorter or longer than the programmed AMS interval, respectively. Mode switch occurs when the AF detection criteria are met (CR). A total of 260 rhythms were studied. NR was significantly faster than MR and CR (latency 2.5 +/- 3 s vs 26 +/- 7 s vs 15 +/- 22 s, respectively, P < 0.0001). During sustained AF, MR resulted in the most stable and regular ventricular rhythm compared to NR or CR. In CR, ventricular rate oscillated between AMS and atrial tracking (cycle length variations: 44 +/- 2 s vs 346 +/- 109 s vs 672 +/- 84 s, P < 0.05). At resumption of sinus rhythm, MR resynchronized after 143 +/- 22 s versus 3.4 +/- 0.7 s for NR and 5.9 +/- 1.1 s for CR, resulting in long periods of AV dissociation when a VVI/VVIR mode is used after AMS. Programming of atrial refractory periods did not affect AMS response, although the speed of AMS onset can be adjusted by programming of onset criteria in the Meta DDDR. AMS algorithms differ in their ability to handle recorded clinical atrial arrhythmias. The rapid-responding algorithm exhibits rate instability, whereas slow responding algorithm shows a long delay in response and risk of AV dissociation. Thus different instrumentation of AMS may have clinical implications in patients with dual chamber pacemakers who develop AF.