The origin of comparative philology dates from the time when European scholars became accurately acquainted with the ancient language of India. Before that time classical scholars had been unable, through centuries of learned research, to determine the true relations between the then known languages of our stock. This fact alone shows the importance of Sanskrit for comparative research. Though its value in this respect has perhaps at times been overrated, it may still be considered as the eldest daughter of the old mother-tongue. Indeed, so far as direct documentary evidence goes, it may rather be said to be the only surviving daughter; for none of the other six principal members of the family have left any literary monuments, and their original features have to be reproduced, as best they can, from the materials supplied by their own daughter languages: such is the case as regards the Iranic, Hellenic, Italic, Celtic, Teutonic, and Letto-Slavic languages. To the Sanskrit the antiquity and extent of its literary documents, the transparency of its grammatical structure, the comparatively primitive state of its accent system, and the thorough grammatical treatment it has early received at the hand of native scholars must ever secure the foremost place in the comparative study of Indo-Germanic speech. The Sanskrit alphabet consists of the following sounds :—

(a ) Fourteen vowels, viz.:—

Ten simple vowels: a â, i î, u û, r r, l (l ); and

Four diphthongs: 6 di, 6 du.

(b ) Thirty-three consonants, viz.:—

Five series of mutes and nasals:

guttural: k kh g gh n palatal: c ch j jh ñ lingual: t th d dh n dental: t th d dh n labial: p ph b bh m;

Four semivowels: y r I v (w );

Three sibilants: palatal s, lingual sh , dental s ; and A soft aspirate: A.

(c ) Three unoriginal sounds, viz.:—

visarga (h ), a hard aspirate, standing mostly for original s or r ; and two nasal sounds of less close contact than the mute-nasals, viz., anusvâra (m ) and anumâsika (m).

As regards the vowels, a prominent feature of the language is the prevalence of a -sounds, these being about twice as frequent as all the others, including diphthongs, taken together (Whitney).

The absence of the short vowels ě and ǒ from the Sanskrit alphabet, and the fact that Sanskrit shows the a -vowel where other vowels appear in other languages,— e.g., bharantam = φίροντα, ferentem; janas — ytvos, genus,— were formerly considered as strong evidence in favour of the more primitive state of the Sanskrit vowel system as compared with that of the sister languages. Recent research has, however, shown pretty conclusively from certain indications in the Sanskrit language itself that the latter must at one time have possessed the same, or very nearly the same, three vowel-sounds, and that the differentiation of tho original «-sound must, therefore, have taken place before tho separation of the languages.

The vowels t and 6, though apparently simple sounds, are classed as diphthongs, being contracted from original di and du respectively, and liable to be treated as such in the phonetic modifications they have to undergo before any vowel except d.

As regards the consonants, two of the five series of mutes, the palatal and lingual series, are of secondary (the one of Indo-Iranian, the other of purely Indian) growth.

The palatals are, as a rule, derived from original gutturals, the modification being generally due to the influence of a neighbouring palatal sound i or y, or if (a): e.g., carati = Lat. currit; janu = yovu, genu, knee. The surd aspirate cA, in words of Indo-Germanic origin, almost invariably goes back to original sk : e.g., chid· (chind-)= scindo, σχίζω ; chaya = σκιά.

The palatal sibilant ś (pronounced sh) likewise originated from a guttural mute k, but one of somewhat different phonetic value from that represented by Sanskrit k or c. The latter, usually designated by A* (or q), is frequently liable to labialization (or dentalization) in Greek, probably owing to an original pronunciation kw (qu): e.g., katara = ποτϵρος, uter ; while the former (k 1 ) shows invariably κ in Greek, and a sibilant in the Letto-Slavic and the Indo-Iranian languages: e.g., svan (sun) = κύων (κνν), canis, Germ, hund; daian=SeKa, decern, Goth, taihun.

The non-original nature of the palatals betrays itself even in Sanskrit by their inability to occur at the end of a word,— e.g., acc. vacam = Lat. vocem, but nom. vdk = vox,— and by otherwise frequently reverting to the guttural state.

The linguals differ in pronunciation from the dentals in their being uttered with the tip of the tongue turned up to the dome of the palate, while in the utterance of the dentals it is pressed against the upper teeth, not against the upper gums as is done in the English dentals, which to Hindus sound more like their own linguals. The latter, when occurring in words of Aryan origin, are, as a rule, modifications of original dentals, usually accompanied by the loss of an x or other adjoining consonant; but more commonly they occur in words of foreign, probably non-Aryan, origin, of regular occurrence in tho language, however, is the change of dental n into lingual n, and of dental s into lingual sh, when preceded in the same word by certain other letters.

The sonant aspirate h is likewise non-original, being usually derived from original sonant aspirated mutes, especially gh: e.g., hamsa = xhv (for χανς), anser, Germ, gans; aham = ϵγών, ego, Goth. ik.

The contact of final and initial letters of words in the same sentence is often attended in Sanskrit with considerable euphonic modifications; and we have no means of knowing how far the practice of the vernacular language may have corresponded to these phonetic theories. There can be no doubt, however, that a good deal in this respect has to be placed to the account of grammatical reflexion; and the very facilities which the primitive structure of the language offered for grammatical analysis and an insight into the principles of internal modification may have given the first impulse to external modifications of a similar kind.

None of the cognate languages exhibits in so transparent a manner as the Sanskrit the cardinal principle of Indo-Germanic word-formation by the addition of inflexional endings—either caseendings or personal terminations (themselves probably original roots)—to stems obtained, mainly by means of suffixes, from monosyllabic roots, with or without internal modifications.

There are in Sanskrit declension three numbers and seven cases, not counting the vocative, viz., nominative, accusative, instrumental, dative, ablative, genitive and locative. As a matter of fact, all these seven cases appear, however, only in the singular of «-stems and of the pronominal declension, other noun-stems have only one case-form for the ablative and genitive singular. In the plural, the ablative everywhere shares its form with the dative (except in tho personal pronoun, where it has the same ending as in the singular), whilst the dual shows only three different case-forms—one for the nominative and accusative, another for the instrumental, dative, and ablative, and a third for the genitive and locative.

The declension of a-stems, corresponding to the first and second Latin declensions, is of especial interest, not so much on account of its being predominant from the earliest time, and becoming more and more so with the development of the language, but because it presents the greatest number of alternative forms, which supply a kind of test for determining the age of literary productions, a test which indeed has already been applied to some extent by Professor Lanman, in his excellent Statistical Account of Noun Inflexion in the Veda. These alternative case-forms are :—

(1) dsas and ds for tho nominative plural masc. and fern.: e.g., asvasas and asvas = cqui (equae). The forms in dsas,— explained by Bopp as the sign of the plural as applied twice, and by Schleicher as the sign of the plural as added to the nominative singular,— occur to those in ds (i.e., the ordinary plural sign as added to the a-stem) in the Rigveda in the proportion of 1 to 2, and in the peculiar parts of the Atharvaveda in that of 1 to 25, whilst the ending ds alone remains in the later language.

(2) â and âni for the nominative and accusative plural of neuters: as yugd, yugdni = ζυγά, juga. The proportion of the former ending to the latter in the Rik is 11 to 7, in the Atharvan 2 to 3, whilst the classical Sanskrit knows only the second form.

(3) ebhis and dis for the instrumental plural masc. and neuter: e.g., devebhis, devais. In the Rik the former forms are to the latter in the proportion of 5 to 6, in the Atharvan of 1 to 5, while in the later language only the contracted form is used. The same contraction is found in other languages; but it is doubtful whether it did not originate independently in them.

(4) â and âu for the nominative and accusative dual masc.: e.g., ubhd, ubhdu = In the Rik forms in â outnumber those in âu more than eight times; whilst in tho Atharvan, on the contrary, those in âu (the only ending used in the classical language) occur five times as often as those in â.

(5) â and ena (enâ) for the instrumental singular masc. and neut.: as dânâ, dânena = dono. The ending ena is the one invariably used in the later language. It is likewise the usual form in the Veda; but in a number of cases it shows a final long vowel which, though it may be entirely due to metrical requirements, is more probably a relic of the normal instrumental ending â, preserved for prosodic reasons. For the simple ending â, as compared with that in ena, Prof. Lanman makes out a proportion of about 1 to 9 in the Rigveda (altogether 114 cases); while in the peculiar parts of the Atharvan he finds only 11 cases.

(6) âin and ânám for the genitive plural: e.g., (asvâm), asvânâm = Ticiruv, equum (cquorum). The form with inserted nasal (doubtless for andm, as in Zend aspanam), which is exclusively used in the later language, is also the prevailing one in tho Rik. There are, however, a few genitives of «-stems in original dm (for a-dm), which also appear in Zend, Prof. Lanman enumerating a dozen instances, some of which are, however, doubtful, while others are merely conjectural.

The Sanskrit verb system resembles that of the Greek in variety and completeness. While the Greek excels in nicety and definiteness of modal distinction, the Sanskrit surpasses it in primitiveness and transparency of formation. In this part of the grammatical system there is, however, an even greater difference than in the noun inflexion between the Vedic and the classical Sanskrit. While the former shows, upon the whole, the full complement of modal forms exhibited by the Greek, the later language has practically discarded the subjunctive mood. The Indo-Aryans never succeeded in working out a clear formative distinction between the subjunctive and indicative moods.; and, their syntactic requirements becoming more and more limited, they at last contented themselves, for modal expression, with a present optative and imperative, in addition to the indicative tense-forms, and a little-used aorist optative with a special “precative” or “benedictive” meaning attached to it.

Another part of the verb in which the later language differs widely from Vedic usage is the infinitive. The language of the old hymns shows a considerable variety of case-forms of verbal abstract nouns with the function of infinitives, a certain number of which can still be traced back to the parent language, as, for instance, such dative forms as jiv-ase = viv-ere; sah-adhyai = έχεσϴαι; dâ - mane = δόμεναι; dâ-vane = δόύναι. Further, ji-she, “to conquer,” for ji-si, apparently an aorist infinitive with the dative ending (parallel to the radical forms, such as yudh-i, “to fight,” dris'-e, “to see”), thus corresponding to the Greek aorist infinitive λυσαι (but cf. also Latin da-re, for dase, es-se, &c.). The classical Sanskrit, on the other hand, practically uses only one infinitive form, viz., the accusative of a verbal noun in tu, e.g., sthatum, etum, corresponding to the Latin supinum datum, Hum. But, as in Latin another case, the ablative (datu), of the same abstract noun is utilized for a similar purpose, so the Vedic language makes two other cases do duty as infinitives, viz., the dative in lave (e.g., datave, and the anomalous é tavai), and the gen.-abl. in tos (datos). A prominent featuro of the later Sanskrit syntax is the so-called gerund or indeclinable participle in tvâ, apparently the instrumental of a stem in tva (probably a derivative from that in tu), as well as the gerund in ya (or tya after a final short radical vowel) made from compound verbs. The old language knows not only such gerunds in tvA, using them, however, very sparingly, but also corresponding dative forms in tvâya (yuktvdya), and the curious contracted forms in tvî (kritvi, “to do ”). And, besides those in ya and tya, it frequently uses forms with a final long vowel, as bhid-yd, i-tyd, thus showing the former to be shortened instrumentals of abstract nouns in i and ti.

The Sanskrit verb, like the Greek, has two voices, active and middle, called, after their pamary functions, parasmâi-pacla, “word for another,” and âtmane-pada, “word for one’s self.” While in Greek the middle forms have to do duty also for the passive in all tenses except the aorist and future, the Sanskrit, on the other hand, has developed for the passive a special present-stem in ya, the other tenses being supplied by the corresponding middle forms, with the exception of the third person singular aorist, for which a special form in i is usually assigned to the passive.

The present-stem system is by far the most important part of the whole verb system, both on account of frequency of actual occurrence and of its excellent state of preservation. It is with regard to the different ways of present-stem formation that the entire stock of assumed roots has been grouped by the native grammarians under ten different classes. These classes again naturally fall under two divisions or “conjugations,” with this characteristic difference that the one (the second) retains the same stem (ending in a) throughout the present and imperfect, only lengthening the final vowel before terminations beginning with v or m (not final); while the other shows two different forms of the stem, a strong and a weak form, according as the accent falls on the stem-syllable or on the personal ending: e.g., 3 sing, bhára-ti, ϕϵρϵι— 2 pl. bhára-tha, ϕϵρϵτϵ ; but e -ti, eicri—i-tha, ire (for ίτέ); 1 sing, strno-mi, στόρ-νυμι— 1 pl. strnu-mas (στόρνυϵς ).

As several of the personal endings show a decided similarity to personal or demonstrative pronouns, it is highly probable that, as might indeed be a priori expected, all or most of them are of pronominal origin,—though, owing to their exposed position and consequent decay, their original form and identity cannot now be determined with certainty. The active singular terminations, with the exception of the second person of the imperative, are unaccented and of comparatively light appearance; while those of the dual and plural, as well as the middle terminations, have the accent, being apparently too heavy to be supported by the stem-accent, either because, as Schleicher supposed, they are composed of two different pronominal elements, or otherwise. The treatment of the personal endings in the first, and presumably older, conjugation may thus be said somewhat to resemble that of enclitics in Greek.

In the imperfect, the present-stem is increased by the augment, consisting of a prefixed a. Here, as in the other tenses in which it appears, it has invariably the accent, as being the distinctive element (originally probably an independent demonstrative adverb “then ”) for the expression of past time. This shifting of the word-accent seems to have contributed to the further reduction of the personal endings, and thus caused the formation of a new, or secondary, set of terminations which came to be appropriated for secondary tenses and moods generally. As in Greek poetry, the augment is frequently omitted in Sanskrit.

The mood-sign of the subjunctive is a, added to (the strong form of) the tense-stem. If the stem ends already in a, the latter becomes lengthened. As regard the personal terminations, some persons take the primary, others the secondary forms, while others again may take either the one or the other. The first singular active, however, takes ni instead of mi, to distinguish it from the indicative. But besides these forms, showing the mood-sign a, the subjunctive (both present and aorist) may take another form, without any distinctive modal sign, and with the secondary endings, being thus identical with the augmentless form of the preterite.

The optative invariably takes the secondary endings, with some peculiar variations. In the active of the first conjugation, its mood-sign is yA, affixed to the weak form of the stem: e.g., root as,— syâm = Lat. siem, sîm; while in the second conjugation and throughout the middle it is f, probably a contraction of yâ: e.g., bhares = ϕepois.

Besides the ordinary perfect, made from a reduplicated stem, with distinction between strong (active singular) and weak forms, and a partly peculiar set of endings, the later language makes large use of a periphrastic perfect, consisting of the accusative of a feminine abstract noun in A (-Am) with the reduplicated perfect forms of the auxiliary verbs kar, “to do,” or as (and occasionally bhû), “to be.” Though more particularly resorted to for the derivative forms of conjugation—viz., the causative (including the so-called tenth conjugational class), the desiderative, intensive, and denominative—this perfect-form is also commonly used with roots beginning with prosodically long vowels, as well as with a few other isolated roots. In the Rigveda this formation is quite unknown, and the Atharvan offers a single instance of it, from a causative verb, with the auxiliary kar. In the Vedic prose, on the other hand, it is rather frequent, and it is quite common in the later language.

In addition to the ordinary participles, active and middle, of the reduplicated perfect,— e.g., jajan-van, yeyov-ds; bubudh-âná, πεπυσ-μένο,— there is a secondary participial formation, obtained by affixing the possessive suffix vat (rant) to the passive past participle: e.g., krita-vant, lit. “having (that which is) done.” A secondary participle of this kind occurs once in the Atharvaveda, and it is occasionally met with in the Brahmanas. In the later language, however, it not only is of rather frequent occurrence, but has assumed quite a new function, viz., that of a finite perfect-form; thus kritavdn, kritavantas, without any auxiliary verb, mean, not “having done, but “he has done,” “they have done.”

The original Indo-Germanic future-stem formation in sya, with primary endings,— e.g., dasyati = δώσει (for δώσετι),— is the ordinary tense-form both in Vedic and classical Sanskrit,—a preterite of it, with a conditional force attached to it (adasyat), being also common to all periods of the language.

Side by side with this future, however, an analytic tense-form makes its appearance in the Brahmanas, obtaining wider currency in the later language. This periphrastic future is made by means of the nominative singular of a nomen agentis in tar (datar, nom. data=Lat....