The mRNA encoding vasopressin has recently been documented within the magnocellular hypothalamo-neurohypophyseal projections of the rat such as the median eminence (ME) and the posterior pituitary (PP), suggesting the possibility of its axonal transport. To address the origin of this mRNA and to investigate the functional significance of this unexpected axonal transport of mRNA, we have examined its subcellular localization within both magnocellular perikarya and their axonal projections. For this purpose, we have used nonradioactive in situ hybridization techniques in order to localize the vasopressin mRNA with precision at the ultrastructural level in magnocellular perikarya, dendrites, and axons from control, salt-loaded, and lactating rats. This approach permitted us to demonstrate directly the axonal localization of vasopressin mRNA. Moreover, we were able to obtain novel information concerning vasopressin mRNA compartmentation within both perikarya and axons. At both light and electron microscopic levels, we observed vasopressin mRNA-containing cells in the hypothalamic magnocellular cell body groups, but not in the ME or in the PP. When vasopressin mRNA was detected in medium-size dendrites, it was always associated with the rough endoplasmic reticulum (RER). Within the labeled magnocellular perikarya, the abundant vasopressin mRNA was mainly associated with discrete areas of the RER. However, vasopressin mRNA was never detected in the Golgi apparatus or in association with neurosecretory granules, in perikarya or axons. These data suggest that vasopressin mRNA translation is restricted to certain segments within the RER, and that axonal transport of vasopressin mRNA does not involve the classical neurosecretory pathway, via the Golgi apparatus and the neurosecretory granules, as has been proposed. Within the magnocellular neuron axons, vasopressin mRNA could be detected only in a subset of axonal swellings, all of which were confined to the internal layer of the ME and the PP. The mRNA-containing swellings were numerous in 7 d salt-loaded animals, less abundant in lactating animals, and almost undetectable in control animals. In all groups of animals, no vasopressin mRNA was detectable in any other region of the magnocellular neuron axons, including undilated axonal segments or varicose swellings. These results strongly suggest that, under physiological activation such as chronic salt loading, axonal vasopressin mRNA is increased and becomes aggregated in a selected subset of swellings of the ME and the PP. Furthermore, these data indicate that along the magnocellular neuron axons, the swellings may differ in their biochemical and functional features. Further analysis focused on the mRNA-accumulating swellings may illuminate the function of RNA within the axonal compartment.