knowledge of the mere facts of language is interesting enough; nay, if you ask yourself what grammars really are those very Greek and Latin grammars which we hated so much in our schoolboy days - you will find that they are store-houses, richer than the richest museums of plants or minerals, more carefully classified and labeled than the productions of any of the great kingdoms of nature. Every form of declension and conjugation, every genitive and every so-called infinitive and gerund, is the result of a long succession of efforts, and of intelligent efforts. There is nothing accidental, nothing irregular, nothing without a purpose and meaning in any part of Greek or Latin grammar. No one who has once discovered this hidden life of language, no one who has once found out that what seemed to be merely anomalous and whimsical in language is but, as it were, a petrification of thought, of deep, curious, poetical, philosophical thought, will ever rest again till he has descended as far as he can descend into the ancient shafts of human speech, exploring level after level, and testing every successive foundation which supports the surface of each spoken language.

One of the great charms of this new science is that there is still so much to explore, so much to sift, so much to arrange. I shall not, therefore, be satisfied with merely lecturing on Comparative Philology, but I hope I shall be able to form a small philological society of more advanced students, who will come and work with me, and bring the results of their special studies as materials for the advancement of our science. If there are scholars here who have devoted their attention to the study of Homer, Comparative Philology will place in their hands a light

with which to explore the dark crypt on which the temple of the Homeric language was erected. If there are scholars who know their Plautus or Lucretius, Comparative Philology will give them a key to grammatical forms in ancient Latin, which, even if supported by an Ambrosian palimpsest, might still seem hazardous and problematical. As there is no field and no garden that has not its geological antecedents, there is no language and no dialect which does not receive light from a study of Comparative Philology, and reflect light in return on more general problems. As in geology again, so in Comparative Philology, no progress is possible without a division of labor, and without the most general coöperation. The most experienced geologist may learn something from a miner or from a ploughboy; the most experienced comparative philologist may learn something from a schoolboy or from a child.

I have thus explained to you what, if you will but assist me, I should like to do as the first occupant of this new chair of Comparative Philology. In my public lectures I must be satisfied with teaching. In my private lectures, I hope I shall not only teach, but also learn, and receive back as much as I have to give.

NOTES.

NOTE A.

ON THE FINAL Dental of THE PRONOMINAL STEM tad.

ONE or two instances may here suffice to show how compassless even the best comparative philologists find themselves if, without a knowledge of Sanskrit, they venture into the deep waters of grammatical research. What can be clearer at first sight than that the demonstrative pronoun that has the same base in Sanskrit, Greek, Latin, and German? Bopp places together (§ 349) the following forms of the neuter :

Sanskrit
tat

Zend
tad.

Greek
τό

Latin is-tud

Gothic

and he draws from them the following conclusions : —

thata

In the Sanskrit ta-t we have the same pronominal element repeated twice, and this repeated pronominal element became afterwards the general sign of the neuter after other pronominal stems, such as y a-t, ka - t.

Such a conclusion seems extremely probable, particularly when we compare the masculine form sa-s, the old nom. sing., instead of the ordinary sa. But the first question that has to be answered is, whether this is phonetically possible, and how.

If tat in Sanskrit is tata, then we expect in Gothic tha+ tha, instead of which we find tha+ta. We expect in Latin istut, not istud, illut, not illud, it, not id, for Latin represents final t in Sanskrit by t, not by d. The old Latin ablative in d is not a case in point, as we shall see afterwards.

Both Gothic tha-ta, therefore, and Latin istud, postulate a Sanskrit tad, while Zend and Greek at all events do not conflict with an original final media. Everything therefore depends on what was the original form in Sanskrit; and here no Sanskrit scholar would hesitate for one moment between tat and tad. Whatever the origin of tat may have been, it is quite

certain that Sanskrit knows only of tad, never of tat. There are various ways of testing the original surd or sonant nature of final consonants in Sanskrit. One of the safest seems to me to see how those consonants behave before t add hita or secondary suffixes, which require no change in the final consonant of the base. Thus before the suffix iya (called kha by Pânini) the final consonant is never changed, yet we find tad-îya, like mad-iya, tvad-îya, asma d-i y a, yushmad-î y a, etc. Again, before the possessive suffix v at final consonants of nominal bases suffer no change. This is distinctly stated by Pânini, I. 4, 19. Hence we have vidyut-vân, from vidyut, lightning, from the root dyut; we have udasvit - vân, from uda-svi-t. In both cases the original final tenuis remains unchanged. Hence, if we find tad-vân, kad - vân, our test shows us again that the final consonant in tad and kad is a media, and that the d of these words is not a modification of t.

Taking our stand therefore on the undoubted facts of Sanskrit grammar, we cannot recognize t as the termination of the neuter of pronominal stems, but only d;1 nor can we accept Bopp's explanation of tad as a compound of ta+t, unless the transition of an original t into a Sanskrit and Latin d can be established by sufficient evidence. Even then that transition would have to be referred to a time before Sanskrit and Gothic became distinct languages, for the Gothic tha-ta is the counterpart of the Sanskrit tad, and not of tat.

Bopp endeavors to defend the transition of an original t into Latin d by the termination of the old ablatives, such as gnaivod, etc. But here again it is certain that the original termination was d, and not t. It is so in Latin, it may be so in Zend, where, as Justi points out, the d of the ablative is probably a media.a In Sanskrit it is certainly a media in such forms as mad, tvad, as mad, which Bopp considers as old ablatives, and which in madîya, etc., show the original media. In other cases it is impossible in Sanskrit to test the nature of the final dental in

1 Dr. Kielhorn in his grammar gives correctly tad as base, tat as nom. and acc. sing., because in the latter case phonetic rules either require or allow the change of d into t. Boehtlingk, Roth, and Benfey also give the right forms. Curtius, like Bopp, gives y a t, Schleicher ta t, which he supposes to have been changed at an early time into tad (§ 203).

2 Weich ist es ( oder d) wohl im abl. sing. gafnât (gafnâdha). Justi, Handbuch der Zendsprache, p. 362.

the ablative, because d is always determined by its position in a sentence. But under no circumstances could we appeal to Latin gnaivod in order to prove a transition of an original t into d; while on the contrary all the evidence at present is in favor of a media, as the final letter both of the ablative and of the neuter bases of pronouns, such as tad and y ad.

These may seem minutia, but the whole of Comparative Grammar is made up of minutia, which, nevertheless, if carefully joined together and cemented, lead to conclusions of unexpected magnitude.

NOTE B.

DID FEMININE BASES IN â TAKE S IN THE NOMINATIVE SINGULAR?

I ADD one other instance to show how a more accurate knowledge of Sanskrit would have guarded comparative philologists against rash conclusions. With regard to the nominative singular of feminine bases ending in derivative â, the question arose, whether words like bona in Latin, ayatá in Greek, sivâ in Sanskrit, had originally an s as the sign of the nom. sing., which was afterwards lost, or whether they never took that termination. Bopp (§ 136), Schleicher (§ 246), and others seem to believe in the loss of the s, chiefly, it would seem, because the s is added to feminine bases ending in î and û. Benfey 1 takes the opposite view, viz. that feminines in â never took the s of the nom. sing. But he adds one exception, the Vedic gnâ-s. This remark has caused much mischief. Without verifying Benfey's statements, Schleicher (1. c.) quotes the same exception, though cautiously referring to the Sanskrit dictionary of Boehtlingk and Roth as his authority. Later writers, for instance Merguet,2 leave out all restrictions, simply appealing to this Vedic form gnâ-s in support of the theory that feminine bases in â too took originally s as sign of the nom. sing. and afterwards dropped it. Even so careful a scholar as Büchler 3

speaks of the s as lost.

There is, first of all, no reason whatever why the s should

1 Orient und Occident, vol. i. p. 298.

2 Entwickelung der Lateinischen Formenlehre, 1870, p. 20.

8 Grundriss der Lateinischen Declination, 1866, p. 9.