In the majority of organs which possess a lumen an epithelium is interposed between the parenchyma of the organ and the luminal space. Usually this epithelium lies upon a well-defined layer of fine reticulum fibres which constitutes the basement membrane. The developing lung of the embryo rabbit is no exception to this general rule, at any rate until the 24th day of embryonic life, when the complicated branching lumen is everywhere lined by tall, columnar epithelium supported by a reticular basement membrane. But histological examination of the lung of the adult rabbit shows no sign of epithelium or basement membrane. Indeed, the surface structure of the alveoli of the adult mammalian lung is one of the oldest of the unsolved problems in histology. A detailed study by several methods of investigation of the stages intervening between the embryo and adult lung shows that the luminal epithelium ceases to be visible after the 26th day of the 32-day gestation period. Before the 26th day, the total volume of epithelium increases but is unaccompanied by any evidence of cell division. Cell rupture is therefore imminent, and its results become apparent between the 24th and 26th days when degenerate nuclei are extruded into the distal part of the respiratory lumen from ruptured cell envelopes. The healthy epithelium of the more proximal parts of the lumen persists as bronchiolar epithelium, in which quantitative evidence of normal cell division is found. These facts explain the difficulty in interpreting the picture of cell outlines which is seen when silver nitrate impregnates the cement lines between epithelial cells. In the earlier days of embryonic life the impregnated cell outlines reveal the regular meshwork characteristic of a complete epithelium. In the adult lung no such clear and regular picture is seen, and a close study of the intervening stages discloses that this irregularity of cell outlines starts at the 24th day and progresses with the degenerative changes in the epithelium and extrusion of nuclei. When the lung starts to breathe such traces of impregnated epithelium as were present at term finally disappear. In the adult rabbit, counts of nuclei in the alveolar septa show that there are not enough cells to do more than invest the capillary plexus and to provide nuclei for a few alveolar phagocytes. Moreover, a method of investigation whereby the structure of the alveolar septum may be dissociated fails to reveal any trace of lining epithelium. On histological grounds, therefore, the presence of an alveolar epithelium in the lung of the adult rabbit seems to be ruled out. Criticism can, however, be levelled against this conclusion on the grounds of lung growth. It has been said that the presence of alveolar epithelium is required to account for further subdivision of the lung lumen during both pre- and post-natal life. New evidence is given in the second part of this investigation which suggests that the complexity of subdivision of the lung lumen is determined by purely physical factors. It is shown that the inequality of growth rates of total lung volume and of volume of the interstitial tissues is the fundamental factor which determines the complexity of lung architecture. The latter is the result of subdivision of the lumen by a complicated system of septa. The greater the number of septa, the more complex is the subdivision and the higher is the pitch of differentiation. By measuring numbers of septa in terms of the internal surface area of the lumen, a method has been found for quantitative estimation of differentiation. A linear relation is found to exist between this estimate of differentiation on the one hand and the ratio of total lung volume to interstitial volume on the other. The values of this ratio increase throughout embryonic life. Growth of interstitial tissue does not therefore keep pace with growth of total lung volume. The deficit in terms of volume is made good by the increasing volume of the lumen, but there is another deficit. For if interstitial tissue does not grow as rapidly as lung volume, then elastic fibres may be expected not to grow as rapidly. If this is so (and it is the heart of the problem), they will stretch as the lung expands with growth, and the ratio of total lung volume to interstitial volume will be a measure of the stretch to which elastic fibres are subjected in the growing lung. A linear relation is found between this ratio and numbers of septa, each of which contains a bundle of elastic fibres in its free edge. It is this linear relation which suggests a causal relationship between tension and the structural complexity of lung architecture. Support for this view in lungs at term and during the early stages of post-natal life is given by the results of artificial distension. The plan of the 5-day lung, the complexity of whose structure is much greater than that at term, can be reproduced by artificial distension of the dead lung at term. Hence there can be no question of any vital processes of growth nor of epithelial activity. Distension alters the lung architecture by altering fibre tension, especially the tension of elastic fibres. In the living lung, it is very probable that active contraction of plain muscle in the mouths of alveolar ducts and their main subdivisions is also involved. The fact that the structural results of respiration and 5 days' growth in vivo can be so completely reproduced by artificial distension, as it were in vitro, is good reason to believe that subdivision of the respiratory lumen in both cases depends upon the same factor, the tension in elastic fibres.