express descent, or coming from, such as âyana, ayani, etc., should have nothing to do with ǎyana, coming? Is it a mere accident that the suffix bha generally forms names of animals?

1

I have not been afraid to assert that the Latin suffix tas, tatis, the Sanskrit tâti, might have been formed from the root TAN, to stretch, and that it meant originally the same as tantu, a string, then a series, a class; and I am not aware of any objections against this theory, beyond vague expressions of incredulity. If then tâti may come from TAN, why not the adjectival suffix tnu, which expresses habit, custom, ability, such as kri-tnu, able to work, clever; ha-tnu, able to kill, i. e. a weapon; giga-tnu, quick; pîya-tnu, hating? And if tnu comes from tan, then tna and other suffixes would follow suit.

By the side of tnu, there is another suffix, snu, as in kari-shnu, poshayi-shnu, etc. This has sometimes been explained as a phonetic corruption of tnu, but as tnu comes from TAN, snu may have come from SAN, to achieve, to gain, from which we have si-snu (sishnu), wishing to gain. We find the root SAN at the end of compounds in such words as

Go-sán (RV. iv. 32, 22), cow-gaining;

Go-sâh (RV. ix. 2, 10), cow-gaining ;

Go-sánih (RV. vi. 53, 10), cow-gaining.

Why should we hesitate to connect the primary suffix snu with SAN, while we do not doubt for a moment that such forms as si-snu (sishnu) or go-san, nay, even go-sa, are connected with that root?

''Mind,' July 1876; see Note on p. 248.

This is not the place to work out the question of the origin of suffixes in full detail. It is sufficient for my purpose to have shown that some suffixes, both primary and secondary, may with some effort be traced back to predicative roots, and that the greater facility in putting them all down as 'somehow pronominal' does not prove the greater correctness of that theory. But even if many of these suffixes are purely pronominal, if sarit, river, for instance, was originally no more than SA R, running, and it, here, even then we have to admit an act of synthesis, which is intelligible, instead of the process of demonstrative excrescence, which, to my mind at least, conveys no meaning at all.

And this argument, whatever its value may be, applies with the same force to terminations Terminations. as to suffixes. Bopp and Curtius may have gone too far in identifying the terminations of nouns and verbs with the actual pronominal bases which we find in the Aryan languages. The termination of the nom. sing. in s need not be identified with the Sk. pronoun sa, he, but both must have proceeded from the same source; they certainly are demonstrative, not predicative. The personal terminations in Sanskrit, mi, si, ti, e, se, te, m, s, t, i, thâs, ta, need not be derived straight from the pronominal bases mad, tvad, tad, and svad, nor need we have recourse to such compounds as dâ-mâ-tvi, i. e. give-I-thou, in order to account for Latin damus, we give1. But though

1 I was the first to remonstrate against these derivations proposed by Curtius, at a time when to remonstrate against him was still considered heresy; see 'On Curtius' Chronology of the IndoGermanic Languages,' in Selected Essays, vol. i. p. 93.

all phonetic laws would cry out against such proceedings, and could hardly be quieted with the assurance that they had no right to exist in proethnic periods, the broad fact remains that these terminations and their corresponding pronominal bases claim in the end a common origin. The elements of such terminations belong by their very nature to a period of language long anterior to that in which they could be used as dead, empty, and formal exponents, and whoever knows the wide influence both of phonetic change and of dialectic variety in the most ancient periods of language, will not be surprised at the widest dialectic divergences between pronominal roots and verbal and nominal terminations which in the beginning of all things are supposed to have been identical. It is a false method to attempt to prove minutely what from the nature of the case can only be proved broadly.

It is generally admitted that the Ugric languages, to say nothing of other members of the

Terminations in Ugric Languages.

Turanian class, form their declension and conjugation by means of agglutination. They are called agglutinative, not because they are entirely free from inflection, but because there is in their grammar a decided preponderance of agglutinative forms. In those, however, which have received considerable literary cultivation, the principle of inflection has made so much progress that my friend Kelgren, for instance, treated Finnish as an inflectional language. If, then, we examine these languages more carefully, we shall find the same discrepancies between their terminations and their pronominal roots as in Sanskrit, Greek, and Latin. If any one will only glance at the comparative tables, printed at the end

of my 'Letter on the Turanian Languages' (1853), he will see in Finnish the pronoun of the first person minä, but the terminations n and ni; the pronoun of the second person sinä, but the terminations t and si; the pronoun of the third person hän, but the terminations hn, pi, wi, nsa, sa. Again, in the declension of Finnish the terminations expressive of every possible kind of locality have been reduced to four elements, of which, if we adopt Professor Ludwig's principles, it would be very difficult to decide whether they should be treated as words originally independent and significative, or as pronominal elements, or as mere excrescences. All scholars, however, seem to be agreed that the four elements, na, he (si), ke, and ta, with which the thirteen cases in Finnish are built up, had originally their own meaning, and that they are related,

1. he (si) to se, he, into, sisä, the interior;

2. ke

3. ta

4. na

to ke, thither, ki, without;

to ta, away, táv, distant;

to n, ne, na, near, with, na, near space1.

If out of such elements Ugrian scholars succeed in constructing the whole framework of the Finnish declension, Aryan scholars need not be afraid to recognise similar constituent elements in the terminations of the Aryan noun. A few such demonstrative elements, or whatever else we like to call them, would originally have been quite sufficient to express every case of our declensions, leaving the noun, without any such distinction, to do service for the nominative. Nor do

1 Boller, Die Declination der Finnischen Sprachen, Sitzungsberichte der K. K. Akademie in Wien, Philos. Hist. Classe, xi. Band, 1853, p. 973.

I see any difficulty in following the process through which, either by phonetic change or by imitation, a few such compounds of nominal bases with local adverbs could be differentiated and adapted to all purposes of declension. During a period of language in which Professor Ludwig1 admits a possible transition of svi into bhi, because in Greek, but in Greek only, ope becomes Fe, and Curtius the transition of tvi-tva both into thas and sai, the difficulty of deriving all nominal terminations from a few pronominal elements such as sa, ta, na, bha would not be very great. But it stands to reason that when we manipulate such vague elements and under such lax phonetic conditions, we can hardly expect unanimity as to the exact process by which our analysis ought to be conducted. Hence the divergence between the results arrived at by Bopp, Pott, Benfey, Grassman, Schleicher, Curtius and others3; hence also the unwillingness of other scholars to add one more to the many guesses as to how svi becomes abhi, and abhi ai, and ai e, and e i. Why will not scholars learn that there are certain subjects which from their very nature do not admit of accurate treatment, and that across a quagmire you may indeed throw a high bridge, but you cannot make a footpath? There is surely a via media between the extreme opinion to which Ludwig would lead us, that what we call the grammatical elements of language had never any independent existence at all, and the opposite opinion of M. de Saussure and his followers, that we

1 Enstehung der A. Declination, p. 181.

2 Selected Essays, vol. i. p. 94.

3 See Hübschmann, Casuslehre (1875), P. 93.