The ultra-structure of the developing notochord in urodele embryos, from the neurula to young tadpole stages, has been studied in thin sections. The first part of the paper is concerned with the intercellular membranes, the second with intracellular structures. In neurula stages the notochord cells are in rather loose contact, and gaps of considerable size occur between them. In tailbud stages, the cells become much more closely apposed, the surface of contact being usually thrown into slight waves or bumps; when sectioned normally it appears as two closely adherent profiles. In later tailbud stages the plasma membranes of the cells begin to fall apart again. The first sign of this is the appearance of small vesicles whose form suggests that fluid is being secreted into the intercellular spaces. These membrane vesicles increase considerably in numbers, but not in average dimensions (diameter about 500 to 700 angstrom). It is concluded that the increase in the closeness of association between contiguous cell membranes, which is seen during the early stages of chordagenesis, might provide the motive force which brings about the morphogenesis of the organ, as has been suggested earlier. The later separation of the cell membranes, with the appearance of membrane vesicles, is an unexpected phenomenon the significance of which is not clear. At the beginning of the period, the cells are of an undifferentiated embryonic type; by the end of it they have acquired a specific histological character, involving the appearance of large fluid-filled intracellular vacuoles, the formation of a notochordal sheath and other features. During the course of differentiation, two different types of ergastoplasm make their appearance one after another. The first is associated with the formation of the fluid-filled vacuoles; the second with the formation of the sheath; and an ergastoplasm resembling the second chordal type is also found in the mesenchyme cells which lie against the external surface of the sheath. All three ergastoplasms are continuous with the nuclear envelope at the time when they are rapidly increasing in size; and it seems probable that they are directly derived from the outer member of the nuclear envelope. Golgi elements, mitochondria and various other types of granule ('multi-vesiculate bodies') are also found. In the early stages the body of the nucleus is often penetrated by long cytoplasmic processes. It is suggested that these may arise when the new nuclear envelope is being formed at telophase. It is argued that the morphologically characteristic types of ergastoplasm found in different types of cell, and at different stages during the development of a given type of cell, are probably not merely consequences of the particular type of synthesis proceeding, since they appear before such synthesis can have got very far; it seems more probable that the ultra-microscopic morphology of the nuclear envelope and ergastoplasm is a visible expression of the nature of the synthetic machinery. The functions of these structures might either be to increase the efficiency of the nuclear control of cytoplasmic processes, or to contribute to the co-ordination between the various different synthetic processes which must be involved in differentiation.