The objective of this article was to look for evidence of nonrandom be
havior during atrial fibrillation by examining long (> 15 minutes) rec
ordings. We have previously reported transient ''linking'' of atrial a
ctivation during atrial fibrillation, and showed that activation was n
ot entirely random. Over the few episodes of linking seen during 1 min
ute, activation directions apparently repeated, indicating a possible
anatomical or physiological constraint. In the present study, we exami
ned atrial fibrillation over longer time periods to see if this consta
ncy of direction was stable. Endocardial recordings were made from 12
patients with atrial fibrillation using a catheter with three orthogon
al bipoles, allowing measurements of local activation directions in th
ree dimensions. The direction was calculated using Pipberger's half-ar
ea method, and episodes of transient linking were identified. An avera
ge direction for each episode of linking was calculated and plotted in
two dimensions using spherical coordinates (altitude and azimuth). In
addition, the nature of initiation and termination of linking was exa
mined. Of the twelve patients, 611 episodes of linking (range 1 to 169
per patient, mean 51) were identified. The episodes for most patients
clustered closely in direction. In contrast, directions measured for
all activations (i.e., linked and not linked) filled up the entire ava
ilable range. Linking in most cases subjectively appeared to initiate
and terminate suddenly. The results indicate that the local anatomy, p
athology, or physiology of the atrium has a strong constraining effect
on the electrical activations occurring during atrial fibrillation, a
nd revises our perception of activation during atrial fibrillation as
''random.'' The demonstration that local properties greatly influence
conduction during fibrillation has important implications for ablation
or pacing therapy.