Our results and those of others (Table I) suggest that both anatomic and electrogram (potential) approaches are highly successful in eliminating AVNRT. The use of slow-pathway potentials appears to minimize lesion delivery and to be associated with a very small likelihood of complete AV block. Approaches aimed directly at the midseptum also appear to reduce lesion delivery. It is important, however, to understand that the fast and slow AV-nodal pathways are not always confined to anterosuperior (fast) and posteroinferior (slow) locations (at least as they are determined fluoroscopically). On occasion, the slow pathway may be ablated anteriorly and the fast pathway posteriorly. Our three inadvertent successful fast-pathway ablations support these findings. We prefer to conceptualize the AV node as having three ablation zones. Ablation in the anterosuperior zone most often affects fast-pathway conduction; ablation in the posteroinferior zone most often affects slow pathway conduction; and ablation in the midseptal region predominantly affects slow-pathway conduction. Lesions applied to the midseptum do, however, appear more likely to affect inadvertently the fast (or both) pathway(s), probably because of the anatomic convergence of the posteroinferior and anterosuperior AV-nodal approaches in this region. A preliminary report by Wu et al. supports this three-zone concept. The subsequent larger series reported by this group has raised concern that midseptal approaches may be associated with too great a risk of complete AV block. On the other hand, approaches guided exclusively by potentials may be associated with much longer procedure times. Controversy exists over the acceptable end point for ablation procedures. We have not found it necessary routinely to eliminate dual-nodal conduction to maintain a low (3.2%) overall recurrence rate. Aggressive attempts to eliminate all evidence of slow-pathway conduction must be balanced against the risk of inadvertent complete AV block. In conclusion, cumulative data and our clinical experience with ablation of AVNRT suggest that it is possible to be both pragmatic and highly successful. The key components of our approach are (1) an anatomically based, systematic, time-limited search for potentials; (2) elimination of unnecessary lesions that are too atrial or too ventricular to involve the reentrant circuit; (3) a caudocephalad approach that avoids excessively anterior initial lesions, which may result in inadvertent complete AV block; and (4) avoidance of unnecessary lesions in the most inferoposterior sector, which results in patient discomfort and low clinical efficacy. This approach is safe (with minimal risk of AV block), reproducible, and efficacious.