In the first place, for the sociologist as for the historian, social facts vary with the social system of which they form a part; they cannot be understood when detached from it. This is why two facts which come from two different societies cannot be profitably compared merely because they seem to resemble each other; it is necessary that these societies themselves resemble each other, that is to say, that they be only varieties of the same species. The comparative method would be impossible, if social types did not exist, and it cannot be usefully applied except within a single type. What errors have not been committed for having neglected this precept! It is thus that facts have been unduly connected with each other which, in spite of exterior resemblances, really have neither the same sense nor the same importance: the primitive democracy and that of to-day, the collectivism of inferior societies and actual socialistic tendencies, the monogamy which is frequent in Australian tribes and that sanctioned by our laws, etc. Even in the work of Frazer such confusions are found. It frequently happens that he assimilates simple rites of wild-animal-worship to practices that are really totemic, though the distance, sometimes very great, which separates the two social systems would exclude all idea of assimilation. Then if we do not wish to fall into these same errors, instead of scattering our researches over all the societies possible, we must concentrate them upon one clearly determined type. It is even necessary that this concentration be as close as possible. One cannot usefully compare facts with which he is not perfectly well acquainted. But when he undertakes to include all sorts of societies and civilizations, one cannot know any of them with the necessary thoroughness; when he assembles facts from every country in order to compare them, he is obliged to take them hastily, without having either the means or the time to carefully criticize them. Tumultuous and summary comparisons result, which discredit the comparative method with many intelligent persons. It can give serious results only when it is applied to so limited a number of societies that each of them can be studied with sufficient precision. The essential thing is to choose those where investigations have the greatest chance to be fruitful.