I incline to accept the first explanation--that the custom of concubitancy has been evolved from an earlier system of group-marriage and uterine descent. I think that it dates from the remote period when there was indiscriminate intercourse between the members of two exogamous marrying classes, when it was impossible to say who was the actual father of the children born. Under such a system the reputed offspring of two brothers might in reality be the children of only one of them, and the children of two sisters might have a common father, and their union be incestuous. But the children of a brother and sister respectively could not possibly have a common parent, and their intercourse was therefore innocuous. For the same reason the children of concubitants who were not known to have cohabited were still held to be tabu to each other, for the male concubitant had a right of cohabitation with the female of which he might at any time have availed himself, and their offspring reputed to be by their other partners might in reality be half brother and sister without their knowledge. [Pageheader: CENSUS OF CONSANGUINEOUS MARRIAGES] Though the Fijian system of relationships is closely allied to those of the Tamils in India and the Two-mountain Iroquois, and the Wyandots in North America, none of these, except the Tamils, I believe, recognize the principle of concubitant cousinship. The custom must be regarded, I think, as being one of limited range, evolved from marriage laws of far wider application. It undoubtedly exercises upon the Fijians a marked influence in promoting consanguineous marriages--an influence from which the other races in the Pacific are comparatively free, if we except the inhabitants of the island of Tanna in the New Hebrides and possibly some other islands not yet systematically investigated. _Concubitancy in practice._--The fact of a race of men habitually marrying their first cousins promised to exhibit such remarkable features in vital statistics that we did not stop short at investigating the theory alone. We caused a census to be taken of twelve villages, not selected from one province, but chosen only for convenience of enumeration in the widely separated provinces of Rewa, Colo East, Serua, and Ba. I am indebted to the late Mr. James Stewart, C.M.G., for the analysis of the returns which follows:-- In the twelve villages there were 448 families. The couples forming the heads of these families have had born to them as children of the marriage 1317 children, an average of 2·94 to each marriage. But of these 1317 children, only 679 remain alive, 638 being dead. The heads of these families therefore do not replace themselves by surviving children, for only 51·5 per cent. survive, while 48·5 are lost. We divided the married couples into four classes--