The purpose of this study was to determine ascending and descending afferents to a medullary reticular formation (MRF) site that, when electrically stimulated, evoked EMG activity in lumbar deep back muscles. In anesthetized female rats, the MRF was explored with electrical stimulation, using currents less than 50 microA, while EMG activity was recorded from the ipsilateral lateral longissimus (LL) and medial longissimus (ML). MRF sites that evoked muscle activity were located in the gigantocellular nucleus (Gi). At the effective stimulation site, the retrograde fluorescent tracer, Fluoro-Gold (FG), was deposited via a cannula attached to the stimulating electrode. In matched-pair control experiments, FG was deposited at MRF sites that were ineffective in producing EMG activity in LL and ML, for comparison of afferent projections to effective versus ineffective sites. Labeled cells rostral to FG deposition at effective MRF sites were located in the preoptic area, hypothalamus, limbic forebrain and midbrain, with particularly high numbers in the ipsilateral midbrain central gray, tegmentum, paraventricular nucleus and amygdala. At medullary levels, there was a heavy projection from the contralateral Gi. FG labeled cells were also located in the contralateral parvocellular reticular nucleus, and lateral, medial and spinal vestibular nuclei. Labeled cells with ascending projections were observed in greatest number in the rostral cervical spinal cord, with fewer cells at mid cervical levels and even fewer in the lumbar spinal cord. These labeled cells were located primarily in lamina V, VII, VIII and X. Locations of labeled cells following FG deposition at ineffective MRF sites were similar. However, there was a striking difference in the number of cells retrogradely labeled from the effective MRF sites compared to ineffective MRF sites. Significantly greater numbers of labeled cells were observed in the contralateral MRF, the midbrain, and the cervical spinal cord from the FG deposition at effective stimulation sites. These results suggest that one characteristic of MRF sites that activate epaxial muscles is a larger amount of afferent input, from the midbrain central gray and from contralateral Gi, compared to ineffective MRF sites. Ascending and descending inputs converge at the effective MRF sites, and the larger number of descending projections suggests a more powerful contribution of these afferents to deep lumbar back muscle activation.