We injected the anterograde fluorescent tracer 1,1;-dioleyl-3,3,3;, 3;-tetramethylindocarbocyanine methanesulfonate (DiI) into the nucleus ambiguus (NA) and used confocal microscopy to inventory NA fibers and axon terminals in whole-mounts of rat atrial tissues. Both the axons projecting to cardiac ganglia and the innervated principal neurons (PNs) were counted. Rats were injected unilaterally in the NA with DiI, either at four sites (between 600 microm rostral and 600 microm caudal to the obex) or at nine sites (1,600 microm rostral to 1,600 microm caudal). Fluoro-Gold was administered intraperitoneally to retrogradely label neurons of the dorsal motor nucleus of the vagus (DmnX), NA, and cardiac ganglia. To verify that the DiI-labeled fibers examined in the atria originated exclusively from the NA, neurons of the DmnX and the nodose ganglia were surveyed for DiI labeling. Our observations established that (1) NA fibers in the cardiac branches of the vagus numbered in the range of 82-151, left; or 60-122, right. (2) Both left and right NA supplied substantial numbers of fibers to each of the three major cardiac ganglionic plexuses. (3) NA axons terminated in dense basket, or calyx, endings around individual PNs. (4) By issuing divergent collaterals, individual NA fibers supplied numerous PNs with these calyx endings. (5) Labeled axons innervated 2,248 (left vagus) and 1,784 (right), or at least 56% and 45%, of the cardiac PNs. (6) Divergence (i.e., NA axons:PNs innervated) averaged between 1:27 (left vagus) and 1:30 (right vagus). Several features of these NA projections to cardiac ganglia contrasted sharply with those of DmnX projections that we have recently characterized with the same tracing protocol: (1) NA fibers did not innervate small intensely fluorescent cell clusters in cardiac ganglia, whereas DmnX axons did. (2) NA efferent fascicles contained more large fibers (presumably B-type), whereas the DmnX issued more fine caliber fibers (presumably C-type). (3) NA fibers diverged about three times as extensively as did DmnX axons. Taken together, our data strongly suggest that vagal control of the heart involves the convergence and integration of distinct NA and DmnX projections within the cardiac plexuses.