We developed a computer model of the interaction of impulse propagation with anatomic barriers in uniformly anisotropic tissue. Its predictions were confirmed experimentally by using an in vitro cut to create a 6 x 1-mm anatomic barrier in 12 canine epicardial strips. The model predicted that long, thin barriers located parallel to the direction of impulse propagation would have little effect in delaying conduction regardless of the arrangement of cardiac fibers. In this situation, the mean experimental ratio of postcut to control conduction times across the barrier was 1.05:1.00 in 10 tissues. When impulses were proceeding perpendicular to an anatomic barrier, significant distal conduction delay was predicted and found to occur only when the conduction from pacing to recording sites was initially longitudinal to fiber orientation (mean experimental ratio, 2.34:1.00 in five tissues) but not transverse to fiber orientation (ratio, 1.08:1.00 in five tissues). We conclude that the direction of initial impulse propagation and the orientation of myocardial fibers have large effects in the degree to which anatomic barriers delay activation in cardiac tissue. These findings may have implications for the participation of anatomic barriers in reentrant circuits.