A number of neuronal functions governed by the hypothalamic paraventricular nucleus are influenced by serotonin, and it is generally believed that the moderate density of serotonin-immunoreactive fibres and terminals within the paraventricular nucleus originates from the midbrain dorsal and median raphe nuclei. To further evaluate the intricate anatomy of projections from brain stem raphe nuclei of the rat, a combination of retrograde and anterograde tracing experiments were conducted to determine the medullary raphe nuclei projection to the paraventricular nucleus. Rhodamine-labelled latex microspheres, Cholera toxin subunit B and FluoroGold we used as retrograde tracers. Intracerebroventricular injections into the third ventricle of all retrograde tracers labelled a distinct population of neurons in the dorsal raphe situated in the subependymal stratum adjacent to the cerebral aqueduct indicating that these cells take up the tracer from the cerebrospinal fluid. Very few retrogradely labelled neurons were seen in the median raphe after i.c.v. administration of the tracers. Retrograde tracers delivered into the medial part of the paraventricular nucleus labelled no further cells in the midbrain dorsal and median raphe nuclei, whereas a substantial number of retrogradely labelled cells emerged in the pontine raphe magnus. However, when the retrograde tracers were delivered into the lateral part of the paraventricular nucleus, avoiding leakage of the tracer into the ventricle, very few labelled neurons were seen in the dorsal and median raphe, whereas the prominent labelling of raphe magnus neurons persisted. The anatomical organization of nerve fibres terminating in the area of the paraventricular nucleus originating from midbrain raphe nuclei was studied in a series of anterograde tracing experiments using the plant lectin Phaseolus vulgaris leucoagglutinin. Injections delivered into the dorsal raphe or median raphe labelled but a few fibres in the paraventricular nucleus proper. A high number of fine calibered nerve fibres overlying the ependyma adjacent to the paraventricular nucleus was, however, seen after the injections into the subependymal rostral part of the dorsal raphe. Injections delivered into the raphe magnus gave rise to a dense plexus of terminating fibres in the parvicellular parts of the paraventricular nucleus and moderately innervated the posterior magnocellular part of the paraventricular nucleus as well as the magnocellular supraoptic nucleus. Concomitant visualization of serotonin-immunoreactive neurons and retrograde FluoroGold-tracing from the paraventricular nucleus revealed that none of the serotonergic neurons of the raphe magnus projects to this nucleus, while a few of the neurons putatively projecting to the paraventricular nucleus from the median raphe are serotonergic. The current observations suggest that the raphe magnus constitute by far the largest raphe input to the paraventricular nucleus and strongly questions the earlier held view that most raphe fibres innervating the paraventricular nucleus are derived from the midbrain dorsal and median raphe. However, the source of serotonergic innervation of the paraventricular nucleus remains elusive.