For approximately the first two weeks of life, dense serotonin immunoreactivity closely matches the pattern of thalamocortical axons innervating both the granular portion of the primary somatosensory cortex and area 17 in rodents [D'Amato et al. (1987) Proc. natn. Acad. Sci. 84, 4322-4326; Fujimiya et al. (1986) J. comp. Neurol. (1986) 246, 191-201; Rhoades et al. (1990) J. comp. Neurol. 293, 190-207]. This serotonin immunoreactivity is not contained in thalamocortical axons [Rhoades et al. (1990) 293, 190-207] but its source has never been demonstrated. In the present study, a variety of approaches were used to address this issue. The combination of electron microscopy and immunocytochemistry showed that all serotonin immunoreactivity in the developing cerebral cortex was contained in axons and that the terminals of many of these fibers made synapses with the dendrites of cortical cells. Treatment with fluoxetine, a specific inhibitor of serotonin uptake, did not result in a loss of the cortical pattern of serotonin immunoreactivity, indicating that immunoreactive fibers were not labeled solely as a result of serotonin uptake. The combination of retrograde tracing from the primary somatosensory cortex and area 17 with immunocytochemistry demonstrated numerous double-labeled cells in nucleus raphe dorsalis and the median raphe nucleus. Smaller numbers of double-labeled neurons were located in the B9 cell group and the region of the lateral midbrain tegmentum. Large electrolytic lesions that included most of the nucleus raphe dorsalis and median raphe nucleus, but which left the B9 group and more caudal serotoninergic cells undamaged, caused either a substantial reduction in density or complete disappearance of the serotonin pattern in both hemispheres. Unilateral electrolytic lesions of the medial forebrain bundle resulted in a loss of the pattern only on the side of the damage. Injection of the neurotoxin 5,7-dihydroxytryptamine directly into the mesencephalon either abolished or substantially reduced the density of the cortical serotonin immunoreactivity. Injections that produced substantial cell loss in the median raphe nucleus, but only minor cell loss in the nucleus raphe dorsalis had little effect upon the cortical pattern of serotonin immunoreactivity. These results indicate that the dense serotonin immunoreactivity which appears transiently in the visual and somatosensory cortices of perinatal rodents is contained in serotoninergic axons that arise from cells in the nucleus raphe dorsalis and perhaps also the median raphe nucleus.