The aim of this study was to report the authors' experience of radiofrequency ablation of accessory atrioventricular pathways over a 10 year period (01-91 to 10-00), and the effect of the "learning curve" on the results. The data of 400 patients admitted to primo-ablation of a bundle of Kent was analysed retrospectively. A total of 481 ablations were performed (1.20 per patient). The cumulative global success in the 414 accessory pathways treated was 90.6%. The primary success rate increased from the 1st to the 4th quartile from 68 to 97% (p = 0.0001). The mean duration of fluoroscopy and number of ablation sites decreased from the 1st to the 4th quartile respectively from 47 +/- 27 to 25 +/- 18 minutes (p = 0.0001) and from 8.5 +/- 7.8 to 4.5 +/- 3.8 minutes (p = 0.0001). The average recurrence rate over the four quartiles was 3.6. The overall complication rate was 1.44%. The improved primary success rate from 1991 to 2000 and, in parallel, the reduction of the number of inappropriate ablation sites and fluoroscopy duration are explained not only by the "learning curve" of our centre but also by the benefits of the application of scientific acquisitions (unipolar recordings, criteria for ablation site localisation...) and technical progress (ablation with temperature monitoring...) over this period.