had "no words representing abstract ideas; for each variety of gum-tree and wattle-tree, &c. &c., they had a name, but they had no equivalent for the expression 'a tree;' neither could they express abstract qualities, such as hard, soft, warm, cold, long, short, round." The lower races of men have excellent memories, but very poor reasoning powers; and the European child who acquires a vocabulary of three or four hundred words in a single year, but attaches all its words to individual objects of sense, reflects their condition very exactly. We may be sure that it was not "the ideas of prime importance" which primitive man struggled to represent, but those individual objects of which his senses were cognizant. As M. Bréal observes,' "It is not probable that in the ante-grammatical period there were as yet no words to denote the sun, the thunder, or the flame. But the day when these words came into contact with pronominal elements, and so became verbs, their sense also became more fluid, and they dissolved into roots which signified shining, thundering, or burning. We can understand how the old words which designated the (individual) objects, afterwards disappeared to make room. for words derived by the help of suffixes from these newly-created roots. We can better understand, too, the existence of numerous synonyms which signify going, shining, resounding; they are the abstracts or abstracta of former appellatives. The idea of shining, for instance, could be taken from the fire as well as from the sun, and so a considerable number of roots, from 1 See his excellent article: "La Langue indo-européenne," in the "Journal des Savants," Oct. 1876 (p. 17).

very different starting-points, have come to be united in a common term." An elementary work on French etymology groups words like rouler, roulement, roulage, roulier, rouleau, roulette, roulis, round a root, roul, with the general sense of "circular movement;" yet in this case we know that this imaginary root roul is nothing else than the Latin substantive rotula. The error of the Sanskritists. is really the same, though the loss of the parent-language prevents us from checking it with the same ease as when we are dealing with French. "Father" and "mother" must have had names in Aryan speech long before the suffix tar was attached to what we call the "roots" pa and ma, and Buschmann has shown that throughout the world these names are almost universally pa or ta and ma. Words like our door, the Latin fores, the Greek Suga, the Sanskrit dwâram (dur), cannot be traced to any root; that is to say, a group of cognate words has either never existed, or else been so utterly forgotten and lost, that we can no longer tell what common type they may have represented. "A word like [the French] car," remarks M. Van Eys, "could pass for a root if we did not know it's derivation."

Roots, then, are the phonetic and significant types discovered by the analysis of the comparative philologist as common to a group of allied words. They form, as it were, the ultimate elements of a language, the earliest starting-point to which we can reach, the reflections of the manifold languages framed by the childhood of our Each family of languages has its own stock of roots, and these roots are the best representatives 1 "Dictionnaire basque-français" (1873), p. v.

race.

we can obtain of the vocabulary of primitive man. Like grammar and structure, roots, too, embody the linguistic instinct and tendency of a race; they are the mirror whereon we can still trace the dim outlines of the thought and mental point of view which has shaped each particular family of tongues. What the language is, that also are its roots; the roots of Chinese or Polynesian are as distinctively and characteristically Chinese or Polynesian as the roots of Aryan are Aryan. We have to extract them from the existing records of speech, and like the individual sounds of which words are composed, the character they assume will be that of the particular speech itself. "Unpronounced," says Prof. Pott,' "they fluttered before the soul like small images, continually clothed in the mouth, now with this, now with that, form, and surrendered to the air to be drafted off in hundred fold cases and combinations." They are, in fact, the product of the unconscious working of analogy, that potent instrument in the creation of language. The name given to an individual object becomes. a type and centre of the ideas that cluster about it; sense and sound are mingled together in indissoluble union, and the instinct of speech transforms the combination into a root. Upon this root, or rather upon the analogy of the name that is the true source of the root, is built a new superstructure of words by the help of suffixes and other derivative elements. But the root and all the family of words that belong to it must remain the shadow and reflection of the original word from which it arose, and consequently display all the characteristics.

1 As quoted by Professor Max Müller, "Lectures," ii. p. 85.

of the words itself, and the language of which it forms part.

Hence it is that the roots of a family of languages have the characteristics of the languages to which they belong. Thus the roots of a Semitic tongue are triliteral, consisting, that is to say, of three consonants, while the roots of the Finno-Ugrian dialects exhibit the same vowelharmony as the developed dialects themselves.1 Hence, too, it is that the roots given by lexicographers merely represent the oldest forms of words of which we know, and do not exclude the possibility that these words are really compounds, or that phonetic decay has acted upon them in some other way long before the earliest period to which our analysis can reach back. In certain cases, indeed, we have good proof that such a possibility has been an actual fact. Thus the Arabic root 'âm, "to be orphaned," is a decayed form of an older 'alam,' and such co-existent Aryan roots as vridh and ridh, both signifying "growing," imply the loss of an initial letter, while it is only within the last few years that the labours of Dr. Edkins and M. de Rosny have given us any idea of the roots of Old Chinese. By the help of the old rhymes, of a comparison of the living dialects and of other similar sources of aid, Dr. Edkins has restored the pronunciation of Mandarin Chinese such as it was 2,000 or even 4,000 years ago.3 Thus yi, "one," was once tit; ta, "great," was dap;

I Donner in the "Z. d. D. M. G.," xxvii. 4 (1873).

2 Ewald: "Ausführliches Lehrbuch der Hebraischen Sprache" (8th edition), p. ix.

* See his "Introduction to the Study of the Chinese Characters (1876).

ye, "to throw," was tik. There are words in which we can trace a continuous process of change and phonetic decay, tsie, "a joint," for instance, being tsit in the classical poetry, and since in Chinese k changes to t, and not contrariwise, while there is evidence that the word once ended in a guttural, we are carried back to a period earlier than 1100 B.C. for the time when tsit was still tsik. But even tsik is not the oldest form to which we can trace it back. Tsik is developed out of tik, and to tik, therefore, we must look for a representation of the root to which it and other allied words have to be referred.

Wherever ancient monuments, or a sufficient number of kindred dialects are wanting, the roots we assign to a set of languages will represent only their latest stage. The further we can get back by the help of history and comparison, the older the forms of the words we compare, the better will be the chance we have that our roots will reflect an epoch of speech, not so very far removed, perhaps, from its first commencement. The so-called "root-period" of the primitive Aryan, really means the analysis of the most ancient Aryan vocabulary, which a comparison of the later dialects enables us to make. Behind that "root-period" lay another, of which obscure glimpses are given us by the roots we can still further decompose. A series of words, for instance, like the Greek vouin, and the Sanskrit yudhmas, presuppose a root yudh(a), but when we remember other sets of words presupposing the roots yu ("joining together") and dha ("placing”), we are carried back to a time when the word signifying "battle," which embodied, as it were,