πατὴρ μόνος, οὐδέ νιν θνατὰ φύσις ἀνέρων ἔτικτεν. These laws are to account for such peculiarities as give to each language its own distinctive character.

This science of Comparative Philology, however, very soon assumed three different aspects, and was cultivated in three schools, which may be called (1) the Etymological, or genealogical, (2) the Analogical, and (3) the Psychological.

Etymological School.

In comparing such languages as Latin, Greek, and Sanskrit, it was soon found that they were really varieties only of one and the same historical prototype, that they pointed to a common origin, and that all their differences must be accounted for either by Phonetic Corruption, or by Dialectic Growth. The comparative study of these languages became therefore genealogical, or, in grammatical phraseology, etymological.

Starting from a certain number of radical and formal elements (the latter being frequently radical elements of an earlier period), the principal object of the genealogical or etymological school has always been to discover the system according to which these elements were combined into words, and to determine the laws which regulate the phonetic changes of words, either in the same or in different languages. When these laws are sometimes treated as natural laws, this means in reality no more than that they admit of no exception, except such as can again be accounted for by new laws.

Analogical School.

The next school, the analogical, or as it might also

be called, the dialectic, tries to discover what in the same or in different languages is not identical, but yet analogous. While the genealogical school looks upon all cognate languages as dialects developed from one ideal κown, the dialectic school looks upon each language as the result of a previous independent growth, and is thus able to account for freedom and variety in single languages as well as in whole families of speech, as against the iron laws of phonetic change, established by the etymological school.

It would be impossible, for instance, or at allevents undesirable 1, to treat say the Ionic dialect as a corruption of the Aeolic, or the Aeolic as a corruption of the Ionic. The same applies to High German and Low German, to Sanskrit and Prâkrit, to Cymric and Goidhelic. These are all independent streams of language, which it is as hopeless to trace back to one common source, as it is to discover the one small source of the Nile, or even of the Thames. They spring indeed from the same geological stratum, and they follow parallel courses under similar conditions, but they are not yet one stream of water or of speech, kept in by the same shores and moving on in the same bed. Even after their confluence, the peculiar colours of what I call Dialectic Growth remain, and help us to account, whether by true or by false analogy, for that want of uniformity or regularity which the etymological school postulates with unyielding severity.

Thus dvau in Sanskrit, dúw in Greek, duo in Latin are phonetic varieties of one and the same type. They are identical in origin, and their differences can be

1 See Lectures on the Science of Language, vol. i. p. 55 seq.

accounted for by phonetic laws. But Sk. dvitîya, the second, and Greek devrepos are not identical in origin. They are dialectic forms, sprung from the same etymological stratum, not the products of one and the same creative act.

Nevertheless, it is in cognate languages only that we could account for such words as Sk. prathama, the first, Greek πрŵтоs, Latin primus, and Gothic fruma. These are all analogous formations, only they must not be treated as varieties of one common prototype. Their differences are not due to the influence of phonetic modification, which can be reduced to a law, but to the freedom of dialectic growth, which must be accepted as a fact.

In

I go even further. We can hardly doubt, for instance, that the words for twenty were formed by a composition of words meaning two and ten. Chinese shi is ten, eúl is two, therefore cúl-shi, twenty. Our own twenty comes from Anglo-Saxon tuen-tig, which corresponds to the Gothic tvai tigjus, and to the modern German zwanzig.

In Sanskrit we should expect a form like dvi-dasa, and in Latin duo-decem. But instead of this we find in Sanskrit vimsati, in Latin viginti, in Greek ekoσt, and the older form Feikari. According to strict phonetic laws, these forms are all irregular. Dvi does never lose the initial d in Sanskrit, nor does dvi in Latin become vi, or in Greek el. In Sanskrit dvi ought to have remained; in Greek dvi ought to have become d, in Latin di or bi. Yet the fact remains that in one of the ancient Aryan dialects dvi was replaced by vi, for thus and thus only can we explain vi in vimsati, el in Greek in ekoσɩ, vî in Latin in

viginti. The stem for ten or decad, namely dasa or dasat, was shortened to sat1, which is likewise without any phonetic excuse or analogy.

Here then we see what I call dialectic influence, as different from the independent working of phonetic laws. Vimsati is not a phonetic corruption of *dvi+dasati, nor viginti of *dvî+decinti, nor FEIKATI of *dFi+deкаTI, but they are dialectic forms in which some old compound of twice-ten was fixed and retained, and was afterwards modified according to the peculiar phonetic instincts of Sanskrit, Greek, and Latin.

I cannot enter more fully into this subject at present, but I may remark that it is the disregard of this distinction between phonetic modification and dialectic growth which at the present moment seems to me to have led to a series of misunderstandings between the most prominent representatives of Comparative Philology 2.

Psychological School.

The comparison of various languages, after it had led to the discovery of the great families of human speech and settled the principles according to which cognate languages should be analysed and explained, opened in the end a still wider prospect and disclosed before our eyes, not only what was common to Greek and Latin, to Hebrew and Arabic, to Finnish and Hungarian, but what was common and essential to all languages, what constitutes in fact the nature of language in general, and indirectly the nature of thought. 1 See Science of Language, i. p. 47.

2 G. Curtius, Zur Kritik der neuesten Sprachforschung, 1885; Delbrück, Die Neueste Sprachforschung, 1885; Brugmann, Zum heutigen Stand der Sprachwissenschaft, 1885.

This kind of study, comparative in the widest sense, though it aimed at the discovery of the highest philosophical truth, does not depend for that discovery on abstract reasoning, but, differing thereby from all former attempts to construct a science of general grammar and of logic, it takes its materials entirely from the facts supplied by that infinite number of languages in which the power of language and thought has become realised. It matters little whether we call this branch of Comparative Philology psychological or ethno-psychological, as long as we see clearly that it aims at explaining that intellectual development which has its outward form in language, and that it derives its materials entirely from a careful study of the different types of human speech, so far as they are still accessible to the student of the present day. To me, that branch of the Science of Language seems almost to transcend the powers of the present generation, and to belong to the future of our race. But I look to it as the final consummation of all that has ever claimed the name of philosophy, as the solution of all psychological, logical, and metaphysical problems, and in the end as the only true key to our knowledge of the Self.

Comparative Mythology.

What applies to Comparative Philology, applies mutatis mutandis to Comparative Mythology. I do not mean to say that the science of Comparative Mythology is as yet so firmly established as the Science of Comparative Philology. There is an honest difference of opinion with regard to many minor points, but the fundamental principles of Com