This study was designed to characterize the relative contributions of the specialized conduction system and the myocardial architecture to the ventricular activation sequence. In animal experiments, the activation sequence within a 14 x 14-mm region on each surface of the pulmonary conus from isolated canine hearts was determined from electrograms recorded during ventricular drives applied at the periphery of the measurement region. Recordings were obtained simultaneously from electrode arrays mounted on the endocardium and epicardium. Activation sequences were determined before and after the right ventricular cavity was bathed with a dilute Lugol-normal Tyrode (LNT) solution that selectively inhibited excitation of Purkinje cells. Simulations of action potential propagation in three-dimensional models (14.4 mm long x 7.2 mm wide x 3.6 mm thick) that included the major features of the midwall architecture were performed to aid in the interpretation of the experimental findings. During endocardial pacing (7 animals, 43 total drives), LNT application markedly prolonged the endocardial (13.7 +/- 1.3 ms) and epicardial (5.7 +/- 1.0 ms) activation sequences. However, epicardial isochrone maps constructed with electrograms recorded before LNT application showed no signs of multiple breakthrough sites and, with the exception of overall timing, closely resembled isochrone maps constructed with electrograms recorded after LNT application. During epicardial pacing (9 animals, 55 total drives), LNT application prolonged the endocardial (3.7 +/- 0.6 ms) and epicardial (1.9 +/- 0.6 ms) activation sequences much less dramatically than during endocardial pacing, suggesting a primary contribution of myocardial architecture. However, in those instances where nonuniform anisotropy slowed epicardial expansion of the depolarization wavefront, the specialized conduction system contributed to the activation sequence to a greater extent.