1. This is the first study to map the axonal projections of medullary neurons that are elements of the network responsible for the 10-Hz rhythm in sympathetic nerve discharge (SND) of urethan-anesthetized cats. Spike-triggered averaging and coherence analysis were used to identify caudal ventrolateral medullary (CVLM) and medullary raphe neurons with activity correlated to this component of SND. Spike-triggered averaging showed that CVLM neurons fired significantly earlier (17 ms on the average) than raphe neurons during the 10-Hz slow wave in inferior cardiac postganglionic SND. This observation raised the possibility that CVLM neurons are a source of the discharges of raphe neurons that are correlated to SND. 2. Nineteen of 47 CVLM neurons with activity correlated to the 10-Hz rhythm in SND were antidromically activated by micro-stimulation of the raphe. The longest onset latency of antidromic activation was 19.9 +/- 2.8 (SE) ms, a value comparable with the difference in firing times of CVLM and raphe neurons during the naturally occurring 10-Hz slow wave in inferior cardiac SND. In most cases the response likely reflected activation of an axonal branch of the CVLM neuron, because the onset latency of antidromic activation could be changed dramatically by moving the stimulating microelectrode as little as 0.2 mm within the raphe. Also, the onset latency of antidromic activation of nine CVLM neurons was significantly shortened (25.0 +/- 2.5 vs. 16.7 +/- 2.7 ms) when the stimulus intensity was raised above threshold. 3. The hypothesis that the axons of CVLM neurons with activity correlated to the 10-Hz rhythm in SND terminated on and excited raphe neurons was supported by the following observations. First, CVLM neurons could not be antidromically activated by stimuli applied to sites in tracks located 1.5-2 mm lateral to the midline, contralateral to the neuronal recording site; thus their axons did not cross the midline. Second, some CVLM neurons could be antidromically activated by stimuli applied to sites in only one of the tracks through the midline; thus it is unlikely that their axons were destined for more rostral or caudal portions of the brain stem. Third, 37% of the raphe neurons with activity correlated to the 10-Hz rhythm were synaptically activated by microstimulation of the CVLM, with a minimum onset latency of 18.1 +/- 2.6 ms. This value was not significantly different than the longest onset latency of antidromic activation of CVLM neurons by raphe stimulation. 4. CVLM neurons with activity correlated to the 10-Hz rhythm in SND could not be antidromically activated by microstimulation of the rostral ventrolateral medulla (RVLM) or thoracic spinal cord. Thus CVLM neurons are not a direct source of the 10-Hz discharges of RVLM or preganglionic sympathetic neurons. 5. Eight of 41 raphe neurons with activity correlated to the 10-Hz rhythm in SND were antidromically activated by microstimulation of the CVLM. The latency of the antidromic response of six raphe neurons was shortened from 15.2 +/- 3.1 to 11.9 +/- 3.1 ms by raising stimulus current above threshold, implying the existence of local axonal branching. The onset latency of antidromic activation of five raphe neurons was changed by moving the stimulating microelectrode within the CVLM. 6. The axons of at least some of these raphe neurons likely terminated in the CVLM, because higher current was required to antidromically activate these neurons from sites in a track located 0.5 mm further laterally, and they were not antidromically activated by microstimulation of the RVLM. Also 32% of the CVLM neurons were either excited or inhibited by microstimulation of the raphe. The minimum onset latency of synaptic activation (18.3 +/- 4.2 ms) or inhibition (10-20 ms) of CVLM neurons by raphe stimulation was similar to the longest onset latency of antidromic activation of raphe neurons by CVLM microstimulation. 7. These data are consistent with the view