While successful termination by pacing of organized atrial tachycardias has been observed in patients, rapid pacing of AF can induce a local capture of the atrial tissue but in general no termination. The purpose of this study was to perform a systematic evaluation of the ability to capture AF by rapid pacing in a biophysical model of the atria with different dynamics in terms of conduction velocity (CV) and action potential duration (APD). Rapid pacing was applied during 30 s at five locations on the atria, for pacing cycle lengths in the range 60-110% of the mean AF cycle length (AFCL(mean)). Local AF capture could be achieved using rapid pacing at pacing sites located distal to major anatomical obstacles. Optimal pacing cycle lengths were found in the range 74-80% AFCL(mean) (capture window width: 14.6 +/- A 3% AFCL(mean)). An increase/decrease in CV or APD led to a significant shrinking/stretching of the capture window. Capture did not depend on AFCL, but did depend on the atrial substrate as characterized by an estimate of its wavelength, a better capture being achieved at shorter wavelengths. This model-based study suggests that a proper selection of the pacing site and cycle length can influence local capture results and that atrial tissue properties (CV and APD) are determinants of the response to rapid pacing.