Previous work has shown that the growth of grafts of transplantable tumors can be in many cases prevented or retarded by underfeeding the new host or by putting it on a special diet. The effect of such treatment on large tumors has been little studied; and the effect on metastases and recurrences has not been studied at all. Apart from certain clinical observations nothing is known as to the influence on spontaneous tumors of alterations in the diet. Experiments with transplanted rat and mouse tumors along the lines thus suggested show that large growths of certain strains are checked in their development by underfeeding the host upon a special diet (Sweet's modification of one of Mendel and Osborne's foods) or in some cases by simple underfeeding. Two metastasizing mouse tumors are instances in point. They stopped growing or grew very slowly in hosts underfed on the special diet. The Flexner-Jobling rat carcinoma, on the other hand, was unaffected by the most rigorous underfeeding on a mixed diet when this was begun after the tumor had been growing for a short period. Experiments to test the influence of underfeeding upon recurrences of this tumor gave results that varied from series to series of animals. The findings strongly indicate that generalizations from work with transplanted tumors as regards the effects of diet on spontaneous growths are unwarranted. By underfeeding on Sweet's food mice with spontaneous tumors, beginning some days prior to operation, it has proved possible in most cases to delay for a relatively long period the development of recurrences and the growth of tumor bits (grafts) disseminated at the time of surgical interference. The treatment entailed great loss of weight. Tumor mice kept on ordinary diet previous to operation, but put thereafter on an abundant ration of Sweet's food, developed recurrences as early as the tumor mice on ordinary diet; whereas the growth of auto-implants was, relatively speaking, much delayed. These results seem attributable rather to a gradual malnutrition induced by the special food than to the circumstance that it lacked a growth principle. In none of the dieted mice was a definite cure obtained. Ordinarily a recurrence appeared and the grafts began to grow soon after the host, again on ordinary food, had regained weight. A few spontaneous tumors seem absolutely unaffected by the most rigorous dieting. Wounds heal with marked slowness in animals that have become thin as a result of dieting, and an inert foreign body (agar-agar) injected subcutaneously is very slowly encapsulated and organized. In these facts may be found a suggestion as to the method whereby dieting delays tumor growth. For it may well be that, with a lessened proliferative activity of the host tissue, the elaboration of a vascularizing and supporting stroma such as most tumors depend upon for their growth, at least indirectly, is much delayed. The rapid growth of tumors in emaciating individuals is not incompatible with the present findings. Such growth may be consequent upon a selection in the host of those cells most fit to cope with the increasingly difficult nutritive conditions. But experiments designed to demonstrate this have been unsuccessful. It is conceivable that recurrences of certain human tumors and the development of metastases may be delayed or prevented for a period by methods somewhat similar to those employed against spontaneous mouse tumors in the present investigation. But generally speaking only the more malignant human tumors would require such palliative measures, and these are precisely the ones that would prove,-if experience with mice is an index,-least amenable to alterations in the nutrition of the host.