All the experimental evidence which was cogent in leading us to discard these hypotheses three years ago, still retains its cogency. Ribbert has considerably modified his hypothesis that carcinoma is due to dislocation of epithelium from its normal association with surrounding epithelium in consequence of chronic inflammatory changes in the connective tissues, whereby they are prepared beforehand to become a soil suitable for the growth of epithelium. Ribbert's own papers and the careful contributions of Victor Bonney in this country are mere reiteration of the observations upon which the hypothesis was originally based. All the additional and new experimental observations made during the past three years, point to the connective-tissue modifications. described by Ribbert and Bonney, e. g., around early epitheliomata, as being secondary and not primary fundamental changes.

Much of the pessimism with which the future of the investigation of cancer is still regarded is due to the persistence of the widely disseminated but ill-defined idea, that malignant new growths, as a whole, are of "congenital origin." This idea arises from a generalisation of the fact that certain forms of cancer are undoubtedly associated with congenital abnormalities.

In the Second Scientific Report attention was called to the association of cancer with peculiar and very different forms of irritation, and to the impossibility of reconciling this fact with a congenital origin in all cases. As the data have increased from year to year the mediate relation between various irritants and cancer in particular sites of the body, has become more and more significant.

The study of the incidence of cancer as determined by irritants in man, demonstrates absolutely that the generalisation of the idea of a congenital or embryonic origin is incorrect and this conclusion agrees

1 Beiträge zur Entstehung der Geschwülste, i. ii. iii. Ergänzung zur "Geschwulstlehre," Bonn, 1906-8.

2 The connective tissue in carcinoma and in certain inflammatory states that precede its onset. Lancet,' May 23rd, 1908.

with the results of experiments, and notably with the experimental production of sarcoma referred to below.

Neglecting cancer in animals-although generalisations to carry any weight at all must be extended to them-and to take but three examples in the case of man, cancer of the skin of the abdomen is practically unknown in Europe, yet it is most extraordinarily frequent in Kashmir; this is not due to a distribution of congenital "germs in the skin of the abdomen of Europeans and natives of India other than that in Kashmiris, but to the fact that the latter irritate the abdominal wall by wearing a charcoal oven round the waist and the former do not. Cancer of the floor of the mouth is rare in European women, although not uncommon in men; but in Ceylon and India generally, the women suffer to a high degree from cancer of the inside. of the mouth. Again, this is not due to developmental differences between women in England and those in India, but to the fact that Indian women chew betel-nut and sleep with the plug in the cheek at the exact spot where cancer starts. In needle-women melanotic sarcoma often develops in the fingers at the site of frequent puncturing by the sewing-needle.

Instances for other sites of the body need not be quoted; suffice it to say that if all forms of cancer are to be explained by such speculations, and the intervention of congenital "germs" is to hold good for all cases, then it must also be assumed that such germs are as uniformly distributed, e. g., over the surface of the body of all vertebrates as the skin is itself, and thus the explanation becomes no explanation at all.

Analogies between infective neoplasms and malignant new growths would appear to be more easy to establish for the sarcomata than for the carcinomata. Whereas in the Second Scientific Report our experimental observations were restricted to carcinoma, all of them that are of fundamental importance have now been extended to sarcomata.

of varied histological structure. The parallel nature of the phenomena for the epithelial and connective tissue new growths is most striking, e. g., experimental transference is effected only by implantation of living cells, and is limited to animals of the same species as that in which the primary growth started; the amount of growth under artificial propagation is unlimited and alternates in its rate; the age of the animals into which implantation is made has an important influence, young animals being much more favourable for growth; the specific nature of the phenomena of immunity to inoculation reveals no evidence in favour of a virus common to sarcoma in different species, indeed, it appears to exclude such a possibility. These, and many other points of agreement in the behaviour of epithelial and connective tissue new growths under experimental conditions, make it as certain for the true sarcomata as for the carcinomata, that no analogy exists with any known form of infective disease.

The question will doubtless be asked "if infection and congenital germs cannot be made responsible, is it not possible that hereditary predisposition plays a part in determining the frequency of cancer"? The most recent returns of the Registrar-General show, that in 1906, out of a total of 141,241 deaths of males above 35 years of age, 12,695 died of cancer, and out of a total of 140,607 deaths of females over 35 years of age, 17,671 also died of cancer. Thus the chance that a man over 35 years will die of cancer is one in eleven, and the chance for a woman above the same age is one in eight. On the basis of a similar approximation for 1905 the following table given in the Fifth Annual Report shows how often, taking the proportions as 1 : 12 and 18 no death, or one, two, three, etc., deaths from cancer may be expected to be recorded in 100 families, half the members of which are men and half women, no hereditary tendency being assumed, and excluding all persons dying under 35 :

The proportion of deaths from cancer was calculated in a similar way by Mr. Harrison Cripps in the 14th Vol. of St. Bartholomew's Hospital Reports, and similar tables were given by Dr. Ogle in the Report of the Registrar General for 1889.

The frequency of cancer as a cause of death is so great that few families of large size escape.

The great frequency of cancer as a cause of death as revealed by the above figures requires to be further analysed. From the national mortality figures above quoted, in one of two cases either a parent or a grandparent will, on an average, have died of cancer, supposing such parents and grandparents to have died after 35 years of age. Suppose a man and wife both of whom died of cancer 60 years ago; further suppose, that of their children, three males and three females, all survived and married, and, two of them, one male and one female, married children of parents who died of cancer like their own parents, while the others married into families with no history of cancer. Were it possible to follow the fate of all descendants of these six families, the comparative frequency of cancer in those of double cancerous heredity and in those of single cancerous heredity would show whether a tendency to the disease is transmitted. It is evident that such observations would have to be continued for many years until the descendants should become numerous enough to exclude the fallacies inherent in small numbers, or, that they should be multiplied by investigations on a large number of families with similar histories. In any case the observations could not be completed within the life-time of one generation of investigators.

If such detailed analyses of the incidence of cancer in a large number of families were practicable, and if they showed great variations above and below the average given in the preceding table, the possibility of the existence of a family susceptibility would be enhanced. In man such an analysis is impracticable because of the length of life, and low fecundity, and because of the progressive alteration in the value to be attached to records of the occurrence of cancer. Resort to experiments is necessary in order to define more accurately the circumstances associated with the spontaneous appearance of cancer.

The difficulty referred to can be obviated in the case of short-lived animals. In the mouse the question is now in a fair way to be definitely settled by means of breeding and in-breeding experiments on a large scale. As in man, so in the mouse, the total number of cases of cancer occurring in different strains appears to vary. The disease has been so frequent as to lead some observers to assert the occurrence of epidemics in certain cages. Our prolonged observations on the occurrence of cancer in the many thousands of mice bred for our investigations have given no support to this interpretation of the greater frequency of cancer in some communities of mice as compared with others, or in the same breeding establishments at different times. The frequency of cancer in mice obeys the general law of age-incidence *.

The surgical removal of spontaneously occurring mammary tumours has enabled us to prolong the life of many mice and to breed from them. In this way we have already obtained upwards of 1000 mice of known cancerous parentage. By successively crossing other spontaneously affected animals with the offspring of cancerous parents, strains are being obtained in which the cancerous heredity is,, or 18, and even higher. The concentration of a hypothetical hereditary factor in a known amount, and in large numbers of animals of known age, should in the course of a few years definitely settle, whether there is a family, or only an individual, liability to the disease. As yet we have

Cf. "Age-incidence of Cancer." The Statistical Investigation of Cancer. Second Scientific Report of the Imperial Cancer Research Fund, part i, March 1905.