Material came from 70 tench, 35 dogs and 5 cats fixed in Bouin''s or 10% formalin, embedded in paraffin-celloidin, and stained by Gomori''s chromhematoxylin-phloxin, azan, or hematoxylin-eosin. Pap''s reticulum fiber impregnation was used. Ganglion cells and non-medullated axones of the preoptic nucleus of the tench and of the supraoptic and paraventricular nuclei of dogs and cats are specifically stained by chromhematoxylin-phloxin due to the neurosecretion granules which are located in the cytoplasm, the neuroplasm of the fibers ("Herring bodies") and possibly on the fiber surfaces. These granules arise in the cytoplasm at the expense of Nissl substance, and their probable path of transmission can be traced from the preoptic and supraoptic nuclei along and within the axons through the pituitary stalk into the neurohypophysis where they accumulate, especially in cats, at the border between the neurohypophysis and pars intermedia. There is no evidence for in situ formation of this granular secretion in the posterior lobe, or for its origin from secretion that has passed into the area from the anterior lobe. There is no sign of the ascent of secretion from the posterior lobe to the diencephalon. There is similar evidence in the paraventricular nuclei of the dog and cat, and that the secretion is led along the tracts to the supraoptic nuclei, thence to the posterior lobe. No evidence is presented as to the function of these secretions, but it is suggested that they may be related to the antidiuretic factor. In tench the typical pattern of secretion formation and distr. occurred only in specimens over 6 cm. long. Secretion was most active in summer and almost nil in winter. In the dog and cat activity also increased with maturity. Cells of the nucleus tuberis cannot be stained by the Gomori technique, but colloid particles form in their nuclei and are conducted to the ventricle. A similar situation apparently occurs in the nucleus preopticus also.