I do not think that this question can be correctly answered in more than one way. We have first to find out the causes of the social phenomena; then, from the prevalence of the causes, we may infer the prevalence of the phenomena themselves, if the former must be assumed to have operated without being checked by other causes. If, then, historical researches based on ethnography are to be crowned with success, the first condition is that there shall be a rich material. It is only by comparing a large number of facts that we may hope to find the cause or causes on which a social phenomenon is dependent. And a rich material is all the more indispensable, as the trustworthiness of ethnographical statements is not always beyond dispute. Without a thorough knowledge of a people it is impossible to give an exact account of its habits and customs, and therefore it often happens that the statements of a traveller cannot, as regards trustworthiness, come up to the evidences of history. As the sociologist is in many cases unable to distinguish falsehood from truth, he must be prepared to admit the inaccuracy of some of the statements he quotes. What is wanting in quality must be made up for in quantity; and he who does not give himself the trouble to read through a voluminous literature of ethnography should never enter into speculations on the origin and early development of human civilization. Often, no doubt, it is extremely difficult to make out the causes of social phenomena. There are, for instance, among savage peoples many customs which it seems almost impossible to explain. Still, the statistical ‘method of investigating the development of institutions,’ admirably set forth in the paper which Dr. Tylor recently read before ‘The Anthropological Institute of Great Britain and Ireland,’[6] will throw light upon many mysterious points. Dr. Tylor has there shown that causal relations among social facts may be discovered by way of tabulation and classification. The particular rules of the different peoples are to be scheduled out into tables, so as to indicate the “adhesions,” or relations of coexistence of each custom, showing which peoples have the same custom, and what other customs accompany it or lie apart from it. If, then, starting with any two customs, the number of their “adhesions” is found to be much greater than the number of times they would coexist according to the ordinary law of chance-distribution—which number is calculated from the total number of peoples classified and the number of occurrences of each custom—we may infer that there is some causal connection between the two customs. Further on, I shall mention some few of the inferences Dr. Tylor has already drawn by means of this method.