analyse every form into the most minute component elements. It is wonderful what brilliant results have now and then been obtained by this minute phonetic analysis. One hardly trusts one's eyes when one sees the phonetic analogies running through languages so long dissociated as Sanskrit and English. But here too there is the danger of going too far, and of forgetting that though language is taught by grammarians, it was not made by grammarians. Language is made by analogy, but often by an instinctive, not by a reasoning analogy, and what to the earliest framers of language seemed analogous, is often by us perceived to be really anomalous.

We may speak, for instance, of a,s as the termination of the nom. plur. after stems in a But I doubt whether before the invention of grammar, and before the conception of any such thing as grammar, people were conscious of a,s as an independent termination. I doubt even whether this as was ever an independent element; at all events there is more than one way of accounting for it. The oldest Vedic form of the nom. plur. is as vâsas. This may be looked upon as an abbreviation of a proethnic form asvas and (asv) as, and one such form would have become the prototype of millions. We must not suppose, however, that after such a typical form as as vâsas had once arisen, it was submitted to an analysis, and that all future plurals were the result of a conscious synthesis. Far from it. A man who knew that asvas is one horse, and as vâsas many horses, would form by an unconscious process of analogy vrikas, one wolf, and vrikâsas, many wolves.

This form as vâsas, which still occurs in the Veda,

may in course of time have been shortened to asvâs, and who could tell how much of âs belonged to the stem, how much to the termination? Still M. de Saussure, guided by the strictest phonetic principles, thinks that he can analyse this â s into two independent elements1, which, even during the earliest historical period, were separated by a perceptible hiatus. The termination of the nom. plur. is supposed to have been as, the final vowel of the stem a. The original synthesis therefore would have been akwa, a,s, pronounced ekwoes, contracted ekwôs. I do not say that this is impossible, I only wish to keep the door open to other possibilities, such as the phonetic contraction of â sas into âs, and I should certainly want stronger confirmation for a phonetic rule that a,+a1 becomes â than this analysis of a svâs.

2

=

Again, if I am told that in order to explain yuktes as the genitive of yuktis, we must suppose that the termination -As changed the i of the stem to y, and that ya became î, which î, preceded by a1, would become ê, I cannot help asking myself whether this complicated process, however creditable to the analytical ingenuity of the grammarian, ever took place in the mind of an illiterate pro-ethnic Aryan, and whether he would really have been aware of any difference between ka,twA-ár-a‚s and katwáАr—as3.

However, these minutiae belong to comparative grammarians, and though I am afraid they may be carried too far, I am the last man to refuse them the full credit which they deserve. All I stand up for is that there must be both in the synthesis and analysis

1 De Saussure, Système Primitif des Voyelles, p. 91.

of grammatical forms broad and intelligible principles. Now synthesis is only intelligible when we have two independent elements that can be put together; hence, I hold, that in the beginning suffixes and terminations were either demonstrative or predicative elements, though by phonetic corruption and false analogy their original character was often obscured and they dwindled down to unmeaning signs. Many of them we must accept as such, without being able to reduce them to their primitive phonetic form. Where phonetic analysis can restore their original form, its results will always be welcome, but in many cases an ignoramus will here too be a greater proof of wisdom than novimus. In the Science of Thought we stand on the broad principle that nothing can exist in language which had not originally a purport: in the Science of Language we must be satisfied to explain all we can, though conscious all the time that we cannot explain all we wish.

One result, at all events, seems to me firmly established with regard to the Aryan case- Case-termiterminations, namely that all of them, not nations local. excepting those of the nominative, accusative, and genitive, were in the beginning local. An exception is generally made in favour of the nominative and accusative, as being from the beginning intended to convey more general or purely logical relations. But though it may be true that in these two cases their local meaning was obscured and forgotten at a much earlier time than that of the other cases, yet in the beginning the nominative through its demonstrative termination expressed clearly the here and there in space, and the accusative the hither and thither,

or the object towards which the action of a transitive verb was conceived to tend1. As to the genitive, it was either predicative and adjectival, conveying the genus to which a subject belonged 2, or it was an ablative, expressing origin and removal. The originally local character of the other cases, the locative, ablative, instrumental, and dative, is no longer questioned, and recent researches in comparative syntax have shown how clear are the transitions from one case to another, and how untenable are the views of all philosophical grammarians when they attempted to determine the sphere of each case by one fundamental relation assigned to each. Here, as elsewhere, historical research must form the foundation of philosophic reasoning, not vice versa. We may follow the ways of language and understand them by following them, but we can never trace beforehand the ways which language is to follow, except in the most general way. Certain it is that the abstract meanings, whether modal or causal, which were formerly considered as the fundamental meanings of the caseterminations have been proved to be everywhere of later date than the local and temporal meanings, a fact which is in full harmony with all that the Science of Language has taught us of the growth of human language and human thought 3.

1 Chap. vi. p. 319.

2 Lectures on the Science of Language, vol. i. p. 122, as quoted by Hübschmann, Casuslehre, p. 104. I had proposed the same explanation in my Letter on the Turanian Languages, 1853, p. 41, and, earlier still, in my paper on Bengâli, 1847, in the Transactions of the British Association, p. 41.

On this point I differ much from Noiré's views as expressed in his Logos,' pp. 246 seq.

There are languages, such as Chinese, in which there is as yet no outward difference How roots between what we call a root, and a noun nal and verbal or a verb. Remnants of that phase in the

became nomi

bases.

growth of language we can detect even in so highly developed a language as Sanskrit. But it was one of the characteristic features of Sanskrit and the other Aryan languages that they tried to distinguish the various applications of a root by means of what I have called demonstrative roots or elements. If they wished to distinguish the mat, as the product of their handiwork, from the handiwork itself, they would say 'Platting-there;' if they wished to encourage the work they would say 'Platting-they, or you, or we.' We found that what we call demonstrative roots or elements must be considered as remnants of the earliest and almost pantomimic phase of language in which language was hardly as yet what we mean by language, namely logos, a gathering, but only a pointing. How some of these elements came in time to be restricted to certain meanings, such as here, there, he, thou, I, it, etc., we cannot tell. All we can say is, that we find these elements as adverbs, local and temporal, as prepositions, as pronouns, as suffixes, and as terminations of declension and conjugation, and that in their skilful employment consists the power and the charm of Aryan speech.

Take, for instance, the root YUDH, to fight. As a root, that is, as the type from which YUDH and both verbs and nouns are derived, we may its derivatives. call it a mere abstraction. the same shape, only with nominal and verbal base.

R

But it exists in exactly a new purpose, both as a Yudh as a noun, means