must be equally true in all. This is not so, and cannot be so.

Different treatment for different stages of language.

Though language is governed by intelligible principles throughout the whole of its growth, its progress is not so uniform as to repeat exactly the same phenomena at every stage. As the geologist looks

for different characteristics when he has to deal with London clay, with Oxford clay, or with old red sandstone, the student of language, too, must be prepared for different formations, even though he confines himself to one stage only in the history of language, the inflectional. And if he steps beyond this, the most modern stage, then to apply indiscriminately to the lower stages of human speech, to the agglutinative and radical, the same tests which have proved successful in the inflectional, would be like ignoring the difference between aqueous, igneous, and metamorphic rocks. There are scholars who, as it would seem, are incapable of appreciating more than one kind of evidence.

If languages were all of one and the same texture, they might be unravelled, no doubt, with the same tools. But as they are not-and this is admitted by all-it is surely mere waste of valuable time to attempt to test the relationship of Tungusic, Mongolic, Turkic, Samoyedic, and Finnic dialects by the same criteria by which the common descent of Greek and Latin is established; or to try to discover Sanskrit in the Malay dialects, or Greek in the idioms of the Caucasian mountaineers. The whole crust of the earth

is not made of lias, swarming with Ammonites and Plesiosauri, nor is all language made of Sanskrit, teeming with Supines and Paulo-pluperfects.

Phonetic Laws.

Up to a certain point the method by which so great results have been achieved in classifying the Aryan languages may be applicable to other clusters of speech. Phonetic laws are always useful, but they are not the only tools which the student of language must learn to handle. If we compare the extreme members of the Polynesian dialects, we find but little agreement in what may be called their grammar, and many of their words seem totally distinct. But if we compare their numerals we clearly see that these are common property; we perceive similarity, though at the same time great diversity: 1

1

1 Hale, United States Exploring Expedition, vol. vii. p. 246.

When we look at such lists of words, what we have to do first is to note the phonetic changes which have taken place in one and the same numeral, as pronounced by different islanders. We thus arrive at phonetic rules, and these, in their turn, serve to remove the apparent dissimilarity in other words which at first seemed totally irreconcilable. Let those who are inclined to speak disparagingly of the strict observance of phonetic rules in tracing the history of Aryan words, and who consider it mere pedantry to be restrained by Grimm's Law from identifying such words as Latin cura and care, Greek kalein and to call, Latin peto and to bid, Latin corvus and crow, look at the progress that has been made by African and Polynesian philologists in checking the wild spirit of etymology even when they have to deal with dialects never reduced as yet to a fixed standard by the influence of a national literature, never written down at all, and never analysed before by grammatical science. The whole of the first volume of Dr. Bleek's 'Comparative Grammar of the South African Languages treats of Phonology, of the vowels and consonants peculiar to each dialect, and of the changes to which

each letter is liable in its passage from one dialect into another (see page 82, seq.). And Mr. Hale, in the seventh volume of the 'United States Exploring Expedition' (p. 232), has not only given a table of the regular changes which words common to the numerous Polynesian languages undergo, but he has likewise noted those permutations which take place sporadically only. On the strength of these phonetic laws once established, words which have hardly one single letter in common have been traced back with perfect certainty to one and the same source.

Dialectic Regeneration.

At the same time, mere phonetic change or decay will not account for the differences between the Polynesian dialects. We must admit another process also, that of dialectic regeneration. It will hardly be believed, for instance, that since the time of Cook five of the ten simple numerals in the language of Tahiti have been thrown off and replaced by new ones?

Two was rua; it is now piti.
Four was ha; it is now maha.
Five was rima; it is now pae.
Six was ono; it is now fene.

Eight was varu; it is now vau.1

Such changes are very different from those which we observe in the Romanic dialects in their divergence from Latin, or in the ancient Aryan languages in their divergence from a common source. In the Romanic dialects, however violent the changes which made 1 United States Exploring Expedition under the command of Charles Wilkes. Ethnography and Philology,' by H. Hale, vol. vii. p. 289.

1

Portuguese words to differ from French, there always remain a few fibres by which they hang together. It might be difficult to recognise the French plier, to fold, to turn, in the Portuguese chegar, to arrive, yet we trace plier back to plicare, and chegar to the Spanish llegar, the old Spanish plegar, the Latin plicare, here used in the sense of plying or turning towards a place, arriving at a place. It is very different when we have to deal with languages which do not shrink from dropping some of their commonest words and replacing these by new words, generally taken from parallel dialects. Successive changes, taking place in the same language or in the same dialects, may be reduced to phonetic laws, but changes produced by a mixture of dialects are of a totally different character.

Thus, when we have to deal with dialects of Chinese, everything that could possibly hold them together seems hopelessly gone. The language, for instance, now spoken in Cochin-China is a dialect of Chinese, at least as much as NormanFrench was a dialect of French, though spoken by Saxons at a Norman court. There was a native language of Cochin-China, the Annamitic,2 which forms, as it were, the Saxon of that country on which the Chinese, like the Norman, was grafted. This engrafted Chinese, then, is a dialect of the Chinese, and it is most nearly related to the spoken dialect of Canton.3 Yet few Chinese scholars would

1 Diez, Lexicon, s. v. llegar; Grammar, i. p. 379.

2 On the native residuum in Cochin-Chinese, see Léon de Rosny, Tableau de la Cochinchine, p. 138.

In the island of Hai-nan there is a distinct approach to the form