But, without wishing to deny that very remarkable things have been accomplished, all these experiments prove one fact beyond question, namely, that the languages mentioned do not even approximately, and cannot indeed possibly, satisfy the requirements which science must demand of the artificial auxiliary language. Science could not, therefore, have chosen any of these languages as the artificial auxiliary language even had she wished, nor could she do so in the future without experiencing failure. An examination of the reasons for this state of affairs will enable us to arrive at the relation between science and the international auxiliary language. It can be shown what the nature of this relationship must be, and it follows therefrom whether any particular system will or will not be serviceable to science. There are two necessary criteria, namely, internationality of vocabulary and logical precision of expression. One might be inclined to emphasise the importance of the second criterion without paying any attention whatever to the first, and to regard a system constructed on a purely logical basis as alone worthy of science. But this would be a retrograde step, for indeed the question of artificial language originated with the idea of a so-called philosophical language in the mind of Leibnitz and afterwards. If one takes the point of view that the scientific auxiliary language should be constructed on an ideographic basis (that is to say, a system of correlation between symbols and ideas, which, however, as it is a language, must be capable of being spoken), one arrives at an _à priori_ system, as it is called in the theory of universal language. Thanks to the laborious and self-sacrificing work of the thousands who during the last twenty years have devoted and still devote themselves to the great experiments in language, we are able nowadays to refer this question to the test of experience. The latter has shown with absolute certainty that _à priori systems cannot be spoken_. The learning of any natural language, with all its irregularities, peculiarities, and anomalies, is child's play compared to the learning of an _à priori_ system. All experiments in this direction have failed and need no longer be seriously considered. But even when an artificial language has not been constructed _à priori_ another error, producing much the same effect, may very greatly injure its facility in practice. An otherwise so successful system as Volapük came finally to grief through an error of this sort. Although Volapük was constructed by a man of whom it is said that he was master of, or at least acquainted with, fifty-five living languages, and although, according to its whole nature, it appeared to be modelled very closely on natural languages, nevertheless the abbreviations which Schleyer introduced so often into the words he took over (for example, _vol_ for _world_, _pük_ for _speak_, _Melop_ for _America_) produced the same psychological effect as