important conclusions respecting them. Certain groups of characters have final m uniformly. Others have final p.

These final letters therefore

characters were first made.

Others have final k, and so on. were in existence when the This accounts for their having been retained as signs of words ending in these letters till they were lost in the upgrowth of the Mandarin tongue.

In the third chapter, containing a list of 1144 phonetics, will be found many lost finals restored. A considerable part of these have the restored final letter in a parenthesis to denote lack of certainty. In cases without the parenthesis, I have felt satisfied as to the propriety of restoring the lost letter, and usually the reasons are given. Whenever I could find the evidence, I have been careful to mark the authority of old works, and chiefly the Kwang yün, for the restoration of lost finals. This work has been for me the most prolific and valuable source of information on this point. In it the initials k, g, k', b, p, p', etc., are kept carefully distinct throughout.

When the old final and initial letters, or in other words the ancient phonetic values of the phonetics, are fixed, the determination of roots must follow. Phonetic characters are not roots. They are a key to the roots. Each widely extended root is written with several different phonetics. The knowledge of the phonetics will be followed by the discovery of the roots of which they were the signs.

If it be asked why had not each root a distinct phonetic, the reply must be that roots rapidly grew. Thus, many round things were in primitive times called lut or sut, both from an older dut. Four or five roots soon became ten or twenty. But it would happen that soon after a number

of round things had received this name, it would become polished, intensified, modified, abridged, and lengthened, in each instance after a fashion of its own. Then came the invention of writing. All the words were written on pictorial and phonetic grounds. The men who wrote them could only to a certain degree, while inventing signs for the various words, act under the impression that any of those words were etymologically connected. Thus round things with the sounds leu, lü, lu, sü, t'eu, teu, would come to be written at first with several phonetics. After the loss of final t, there was still greater confusion, for other phonetics which had lost k would be by some writers employed as signs for words which had lost t, while phonetics used as signs for roots anciently ending in p would be used by modern writers for roots once ending in k.

The best way to represent the Chinese roots would be, perhaps, that adopted by Pictet in his "Origines IndoEuropéennes." Philological studies should be perpetually associated with the life of the people and the objects embraced within the horizon of their knowledge. I cannot enter in the present work on so wide an enterprise.

After sketching the principles of formation in the characters, and the history of Chinese writing, I have described the sources for the history of the sounds and the letter changes. which have occurred in the language. As one among the means of gaining information on this point, reference has been made to the Japanese transcription.

Dr. J. C. Hepburn has been the first in his Dictionary to place the question of the old Chinese transcriptions in Japan in a correct and intelligible form. This he has done in his second edition. I have thus been aided in showing in the seventh chapter of this work how the Go on and Kan on

transcriptions may throw much useful light on the history of the Chinese language.

MM. Sarazin and de Rosny do not appear to have seen the new account given by Dr. Hepburn when they discussed the Japanese transcriptions at the Congress of Orientalists at Paris in 1873, as reported in the "Compte Rendu."

I am obliged to M. de Rosny for pointing out in the "Actes de la Société d'Ethnographie," 1871 to 1873, vol. vii., an error on the subject of the Japanese passive into which I had fallen in my "China's Place in Philology." It was an inadvertence, as was his when he represented me as seeking. to trace a path for the Chinese of the old ages to go in a pleasure train to admire the Tower of Babel. An amusing idea this, but it is not mine, for I was careful in my book to express the opinion that the Chinese must have gone away from western Asia before the time of the separation of languages to which the Hebrew and Babylonian document speaking of the Tower of Babel refers.

After reading M. de Rosny's opinion of etymology and of the comparison of words, I still think these comparisons may and ought to be made, and become eminently useful when under the guidance of a good philological method. Speaking of my book he book he says, "Les indianistes, les sémitistes et surtout les hellénistes n'auront qu'à ouvrir son livre pour se former une idée de la solidité de ses comparaisons." I know well that this habit of merely opening a book may lead to a strong condemnation of it. It is not, however, safe to form an unfavourable judgment after so brief an examination. These identities of Greek words, for example, and Greek formative syllables with those belonging to some Turanian languages, are too numerous to be accounted for as accidental. The

Mongol language has been so little studied by European savans that there is till the present time no Mongol dictionary or grammar of that language in English or French. The modern Hellenist believes that the Greeks came from the neighbourhood of the Caspian Sea, where they were near the Mongols, and that the languages of the two races are not connected. He would perhaps modify his view if he first examined the Mongol carefully in regard to roots and grammar and formed an unprejudiced decision, making fair allowance for the effect of geographical contiguity.

Probably M. de Rosny is one of those ethnologists who are opposed to comparisons of words because they seem to cast a doubt on the widely accepted opinion that the various families of human speech grew up separately like trees from the soil. But however appropriate this way of speaking may be, it should be remembered that each tree comes from a seed dropped from a similar tree. Whenever the metaphor of a tree is used of languages, of laws, of grammatical forms or of roots, their derivation in each case from a pre-existing tree of the same kind should be kept in view as a possibility.

Among the new methods in philology that are now coming into vogue is the use of the laws of position in determining the family relationship of languages. I rejoice to see that M. de Rosny has himself used this method in his "Affinités des Langues Finno-Japonaises." Professor Boller's method of proving the connexion of the Japanese and Tartar languages by comparison of words only, falls much short in force because he omits reference to the laws of position. Both these eminent philologists seem to me to limit their subject needlessly by passing over in silence the Dravidian languages. Nor does M. de Rosny notice in the Finnish the

circumstance that its geographical contiguity to Sclavonic and Teutonic peoples has caused a rough shaking in its syntactical system. It is indeed so free from that rigidity in the laws of position that marks the other languages of the group, that the combination Mongol-Japanese would be better as a name than that which M. de Rosny has chosen. But better still is the word Turanian. This may be made to include the Dravidian races, which it appears to me essential not to omit. I would keep the word Turanian, but not extend it to the monosyllabic languages. M. de Rosny has praised parts of "China's Place in Philology," and strongly condemned others. In a few years it will be seen whether he is right in lending encouragement to the hypothesis of mutual isolation between the families.

In giving prominence to the laws of position as valid proof of connexion or disconnexion in language, he cannot claim to be fighting under the "Sanscritist" banner. His studies lie in a more eastern region, and his intelligence compels him to the admission that a careful consideration of those laws is essential to complete the linguistic process which proves consanguinity. Let him carry the process a step further, and he will perhaps find himself driven to the conclusion, that Tartar processes of grammar and Tartar laws of position may be applied to elucidate the peculiarities of languages nearer home. His present position, as at the same time the writer of the Affinités and of the critique on my book, is not very tenable. Words are more easily borrowed by contiguous languages than grammatical features. If the close resemblance of grammatical features between Arian and Turanian languages can be proved by extending the method which M. de Rosny himself employs, then à fortiori the identity