The mechanisms responsible for atrioventricular (AV) delay remain unclear, in part due to the inability to map electrical activity by conventional microelectrode techniques. In this study, voltage-sensitive dyes and imaging techniques were refined to detect action potentials (APs) from the small cells comprising the AV node and to map activation from the "compact" node. Optical APs (124) were recorded from 5 x 5 mm (approximately 0.5-mm depth) AV zones of perfused rabbit hearts stained with a voltage-sensitive dye. Signals from the node exhibited a set of three spikes; the first and third (peaks I and III) were coincident with atrial (A) and ventricular (V) electrograms, respectively. The second spike (peak II) represented the firing of midnodal (N) and/or lower nodal (NH) cell APs as indicated by their small amplitude, propagation pattern, location determined from superimposition of activation maps and histological sections of the node region, dependence on depth of focus, and insensitivity to tetrodotoxin (TTX). AV delays consisted of tau 1 (49.5 +/- 6.59 ms, 300-ms cycle length), the interval between peaks I and II (perhaps AN to N cells), and tau 2 (57.57 +/- 5.15 ms), the interval between peaks II and III (N to V cells). The conductance time across the node was 10.33 +/- 3.21 ms, indicating an apparent conduction velocity (theta N) of 0.162 +/- 0.02 m/s (n = 9) that was insensitive to TTX. In contrast, tau 1 correlated with changes in AV node delays (measured with surface electrodes) caused by changes in heart rate or perfusion with acetylcholine. The data provide the first maps of activation across the AV node and demonstrate that theta N is faster than previously presumed. These findings are inconsistent with theories of decremental conduction and prove the existence of a conduction barrier between the atrium and the AV node that is an important determinant of AV node delay.