Sanskrit Grammar/Chapter II
SYSTEM OF SOUNDS; PRONUNCIATION.
19. The a, i, and u-vowels. The Sanskrit has these three earliest and most universal vowels of Indo-European language, in both short and long form — अ a and आ ā, इ i and ई ī, उ u and ऊ ū. They are to be pronounced in the “Continental” or “Italian” manner — as in far or farther, pin and pique, pull and rule.
20. The a is the openest vowel, an utterance from the expanded throat, stands in no relation of kindred with any of the classes of consonantal sounds, and has no corresponding semivowel. Of the close vowels i and u, on the other hand, i is palatal, and shades through its semivowel y into the palatal and guttural consonant-classes; u is similarly related, through its semivowel v, to the labial class, as involving in its utterance a narrowing and rounding of the lips.
a. The Paninean scheme (commentary to Pāṇini’s grammar i. 1. 9) classes a as guttural, but apparently only to give that series as well as the rest a vowel; no one of the Prātiçākhyas puts a into one class with k etc. All of these authorities concur in calling the i- and u-vowels respectively palatal and labial.
21. The short a is not pronounced in India with the full openness of ā, as its corresponding short, but usually as the “neutral vowel” (English so called “short u”, of but, son, blood, etc.). This peculiarity appears very early, being acknowledged by Pāṇini and by two of the Prātiçākhyas (APr. i. 36; VPr. i. 72), which call the utterance saṁvṛta, covered up, dimmed. It is wont to be ignored by Western scholars, except those who have studied in India.
22. The a-vowels are the prevailing vowel-sounds of the language, being about twice as frequent as all the others (including diphthongs) taken together. The i-vowels, again, are about twice as numerous as the u-vowels. And, in each pair, the short vowel is more than twice (2½ to 3 times) as common as the long.
a. For more precise estimates of frequency, of these and of the other alphabetic elements, and for the way in which they were obtained, see below, 75.
23. The ṛ- and ḷ-vowels. To the three simple vowels already mentioned the Sanskrit adds two others, the ṛ-vowel and the ḷ-vowel, plainly generated by the abbreviation of syllables containing respectively a र् r or ल् l along with another vowel: the ऋ ṛ coming almost always (see 237, 241–3) from अर् ar or र ra, the ḷ from अल् al.
a. Some of the Hindu grammarians add to the alphabet also a long ḷ: but this is only for the sake of an artificial symmetry, since the sound does not occur in a single genuine word in the language.
24. The vowel ऋ ṛ is simply a smooth or untrilled r-sound, assuming a vocalic office in syllable-making — as, by a like abbreviation, it has done also in certain Slavonic languages. The vowel ऌ ḷ is an l-sound similarly uttered — like the English l-vowel in such words as able, angle, addle.
a. The modern Hindus pronounce these vowels as ri, rī, li (or even lri), having long lost the habit and the facility of giving a vowel value to the pure r- and l-sounds. Their example is widely followed by European scholars; and hence also the (distorting and altogether objectionable) transliterations ṛi, ṛī, ḷi. There is no real difficulty in the way of acquiring and practising the true utterance.
b. Some of the grammarians (see APr. i. 37, note) attempt to define more nearly the way in which, in these vowels, a real r- or l-element is combined with something else.
25. Like their corresponding semivowels, r and l, these vowels belong respectively in the general lingual and dental classes; the euphonic influence of ṛ and ṝ (189) shows this clearly. They are so ranked in the Paninean scheme; but the Prātiçākhyas in general strangely class them with the jihvāmūlīya sounds, our “gutturals” (39).
26. The short ṛ is found in every variety of word and of position, and is not rare, being just about as frequent as long ū. Long ṝ is very much more unusual, occurring only in certain plural cases of noun-stems in ṛ (371 b, d, 375). The ḷ is met with only in some of the forms and derivatives of a single not very common verbal root (kḷp).
27. The diphthongs. Of the four diphthongs, two, the ए e and the ओ o, are in great part original Indo-European sounds. In the Sanskrit, they wear the aspect of being products of the increment or strengthening of इ i and उ u respectively; and they are called the corresponding guṇa-vowels to the latter (see below, 235 ff.). The other two, ऐ āi and औ āu, are held to be of peculiar Sanskrit growth; they are also in general results of another and higher increment of इ i and उ u, to which they are called the corresponding vṛddhi-vowels (below, 235 ff.). But all are likewise sometimes generated by euphonic combination (127); and ओ o, especially, is common as a result of the alteration of a final अस् as (175).
28. The ए e and ओ o are, both in India and Europe, usually pronounced as they are transliterated — that is, as a long e- (English “long a”, or e in they) and o-sounds, without diphthongal character.
a. Such they apparently already were to the authors of the Prātiçākhyas, which, while ranking them as diphthongs (saṁdhyakṣara), give rules respecting their pronunciation in a manner implying them to be virtually unitary sounds. But their euphonic treatment (131–4) clearly shows them to have been still at the period when the euphonic laws established themselves, as they of course were at their origin, real diphthongs, ai (a + i) and au (a + u). From them, on the same evidence, the heavier or vṛddhi diphthongs were distinguished by the length of their a-element, as āi (ā + i) and āu (ā + u).
b. The recognizable distinctness of the two elements in the vṛddhi-diphthongs is noticed by the Prātiçākhyas (see APr. i. 40, note); but the relation of those elements is either defined as equal, or the a is made of less quantity than the i and u.
29. The lighter or guṇa-diphthongs are much more frequent (6 or 7 times) than the heavier or vṛddhi-diphthongs, and the e and āi than the o and āu (a half more). Both pairs are somewhat more than half as common as the simple i- and u-vowels.
30. The general name given by the Hindu grammarians to the vowels is svara tone; the simple vowels are called samānākṣara homogeneous syllable, and the diphthongs are called sandhyakṣara combination-syllable. The position of the organs in their utterance is defined to be one of openness, or of non-closure.a. As to quantity and accent, see below 76 ff., 80 ff.
31. The Hindu name for ‘consonant’ is vyañjana manifester. The consonants are divided by the grammarians into sparça contact or mute, antaḥsthā, intermediate or semivowel, and ūṣman spirant. They will here be taken up and described in this order.
32. Mutes. The mutes, sparça, are so called as involving a complete closure or contact (sparça), and not an approximation only, of the mouth-organs by which they are produced. They are divided into five classes or series (varga), according to the organs and parts of organs by which the contact is made; and each series is composed of five members, differing according to the accompaniments of the contact.
33. The five mute-series are called respectively guttural, palatal, lingual (or cerebral), dental, and labial; and they are arranged in the order as just mentioned, beginning with the contact made furthest back in the mouth, coming forward from point to point, and ending with the frontmost contact.
34. In each series there are two surd members, two sonant, and one nasal (which is also a sonant): for example, in the labial series प् p and फ् ph, ब् b and भ् bh, and म् m.
a. The members are by the Hindu grammarians called respectively first, second, third, fourth, and last or fifth.
b. The surd consonants are known as aghoṣa toneless, and the sonants as ghoṣavant having tone; and the descriptions of the grammarians are in accordance with these terms. All alike recognize a difference of tone, and not in any manner a difference of force, whether of contact or of expulsion, as separating the two great classes in question. That the difference depends on vivāra opening or saṁvāra closure (of the glottis), is also recognized by them.
35. The first and third members of each series are the ordinary corresponding surd and sonant mutes of European languages: thus, क् k and ग् g, त् t and द् d, प् p and ब् b.36. Nor is the character of the nasal any more doubtful. What म m is to प् p and ब् b, or न् n to त् t and द् d, that is also each other nasal to its own series of mutes: a sonant expulsion into and through the nose, while the mouth-organs are in the mute-contact.
a. The Hindu grammarians give distinctly this definition. The nasal (anunāsika passing through the nose) sounds are declared to be formed by mouth and nose together; or their nasality (ānunāsikya) to be given them by unclosure of the nose.
37. The second and fourth of each series are aspirates: thus, beside the surd mute क् k we have the corresponding surd aspirate ख् kh, and beside the sonant ग् g, the corresponding sonant aspirate घ् gh. Of these, the precise character is more obscure and difficult to determine.
a. That the aspirates, all of them, are real mutes or contact sounds, and not fricative (like European th and ph and ch, etc.), is beyond question.
b. It is also not doubtful in what way the surd th, for example, differs from the unaspirated t: such aspirates are found in many Asiatic languages, and even in some European; they involve the slipping-out of an audible bit of flatus or aspiration between the breach of mute-closure and the following sound, whatever it may be. They are accurately enough represented by the th etc., with which, in imitation of the Latin treatment of the similar ancient Greek aspirates, we are accustomed to write them.
c. The sonant aspirates are generally understood and described as made in a similar way, with a perceptible h-sound after the breach of sonant mute-closure But there are great theoretical difficulties in the way of accepting this explanation; and some of the best phonetic observers deny that the modern Hindu pronunciation is of such a character, and define the element following the mute as a “glottal buzz”, rather, or as an emphasized utterance of the beginning of the succeeding sound. The question is one of great difficulty, and upon it the opinions of the highest authorities are much at variance. Sonant aspirates are still in use in India, in the pronunciation of the vernacular as well as of the learned language.
d. By the Prātiçākhyas, the aspirates of both classes are called soṣman: which might mean either accompanied by a rush of breath (taking ūṣman in its more etymological sense), or accompanied by a spirant (below, 59). And some native authorities define the surd aspirate as made by the combination of each surd non-aspirate with its own corresponding surd spirant, the h-sound (below, 65). But this would make the two classes of aspirates of quite diverse character, and would also make th the same as ts, ṭh as ṭṣ, ch as cç — which is in any measure plausible only of the last. Pāṇini has no name for aspirates; the scheme given in his comment (to i. 1. 9) attributes to them mahāprāṇa great expiration, and to the non-aspirates alpaprāṇa small expiration.
e. It is usual among European scholars to pronounce both classes of aspirates as the corresponding non-aspirates with a following h: for example थ् th nearly as in English boathook, फ् ph as in haphazard, ध् dh as in madhouse, भ् bh as in abhor, and so on. This is (as we have seen above) strictly accurate only as regards the surd aspirates.
38. The sonant aspirates are (in the opinion of most), or at least represent, original Indo-European sounds, while the surd aspirates are a special Indian development. The former are more than twice as common as the latter. The unaspirated (non-nasal) mutes are very much more frequent (5 times) than the aspirates (for the special frequency of bh and the original gh, see 50 and 66); and among them the surds are more numerous (2½ times) than the sonants. The nasals (chiefly n and m) are nearly as frequent as the surd non-aspirates.
We take up now the several mute series.
39. Guttural series: क् k, ख् kh, ग् g, घ् gh, ङ् n̄. These are the ordinary European k and g-sounds, with their corresponding aspirates and nasal (the last, like English ng in singing).
a. The gutturals are defined by the Prātiçākhyas as made by contact of the base of the tongue with the base of the jaw, and they are called, from the former organ, jihvāmūlīya tongue-root sounds. The Paninean scheme describes them simply as made in the throat (kaṇṭha). From the euphonic influence of a k on a following s (below, 180), we may perhaps infer that in their utterance the tongue was well drawn back in the mouth.
40. The k is by far the commonest of the guttural series, occurring considerably more often than all the other four taken together. The nasal, except as standing before one of the others of the same series, is found only as final (after the loss of a following k: 386, 407) in a very small number of words, and as product of the assimilation of final k to a following nasal (161).
41. The Sanskrit guttural series represents only a minority of Indo-European gutturals; these last have suffered more and more general corruption than any other class of consonants. By processes of alteration which began in the Indo-European period, the palatal mutes, the palatant sibilant ç, and the aspiration h, have come from gutturals. See these various sounds below.
42. Palatal series: च् c, छ् ch, ज् j, झ् jh, ञ् ñ.
The whole palatal series is derivative, being generated by the corruption of original gutturals. The c comes from an original k — as does also, by another degree of alteration, the palatal sibilant ç (see below, 64). The j, in like manner, comes from a g; but the Sanskrit j includes in itself two degrees of alteration, one corresponding to the alteration of k to c, the other to that of k to ç (see below, 219). The c is somewhat more common than the j (almost as four to three). The aspirate ch is very much less frequent (a tenth of c), and comes from the original group sk. The sonant aspirate jh is excessively rare (occurring but once in RV., not once in AV., and hardly half-a-dozen times in the whole other language); where found, it is either onomatopoetic or of anomalous or not Indo-European origin. The nasal, ñ, never occurs except immediately before — or, in a small number of words, also after (201) — one of the others of the same series.
43. Hence, in the euphonic processes of the language, the treatment of the palatals is in many respects peculiar. In some situations, the original unaltered guttural shows itself — or, as it appears from the point of view of the Sanskrit, the palatal reverts to its original guttural. No palatal ever occurs as a final. The j is differently treated, according as it represents the one or the other degree of alteration. And c and j (except artificially, in the algebraic rules of the grammarians) do not interchange, as corresponding surd and sonant.
44. The palatal mutes are by European scholars, as by the modern Hindus also, pronounced with the compound sounds of English ch and j (in church and judge).
a. Their description by the old Hindu grammarians, however, gives them a not less absolutely simple character than belongs to the other mutes. They are called tālavya palatal, and declared to be formed against the palate by the middle of the tongue. They seem to have been, then, brought forward in the mouth from the guttural point, and made against the hard palate at a point not far from the lingual one (below, 45), but with the upper flat surface of the tongue instead of its point. Such sounds, in all languages, pass easily into the (English) ch- and j- sounds. The value of the ch as making the preceding vowel “long by position” (227), and its frequent origination from t + ç (203), lead to the suspicion that it, at least, may have had this character from the beginning: compare 37 d, above.
45. Lingual series: ट् ṭ, ठ् ṭh, ड् ḍ, ढ् ḍh, ण् ṇ. The lingual mutes are by all the native authorities defined as uttered with the tip of the tongue turned up and drawn back into the dome of the palate (somewhat as the usual English smooth r is pronounced). They are called by the grammarians mūrdhanya, literally head-sounds, capitals, cephalics; which term is in many European grammars rendered by ‘cerebrals’. In practice, among European Sanskritists, no attempt is made to distinguish them from the dentals: ट् ṭ is pronounced like त् t, ड् ḍ like द d, and so on with the rest.
46. The linguals are another non-original series of sounds, coming mainly from the phonetic alteration of the next series, the dentals, but also in part occurring in words that have no traceable Indo-European connection, and are perhaps derived from the aboriginal languages of India. The tendency to lingualization is a positive one in the history of the language: dentals easily pass into linguals under the influence of contiguous or neighboring lingual sounds, but not the contrary; and all the sounds of the class become markedly more frequent in the later literature. The conditions of their ordinary occurrence are briefly these: 1. ṣ comes from s, much more rarely from ç, j, kṣ, in euphonic circumstances stated below (180, 218 ff.); 2. a dental mute following ṣ is assimilated into it, becoming lingual (ṭ, ṭh, ṇ: 197); 3. n is often changed to ṇ after a lingual vowel or semivowel or sibilant in the same word (189 ff.); 4. ḍh, which is of very rare occurrence, comes from assimilation of a dental after ṣ (198 a) or h (222); 5. ṭ and ḍ come occasionally by substitution for some other sound which is not allowed to stand as final (142, 145–7). When originated in these ways, the lingual letters may be regarded as normal; in any other cases of their occurrence, they are either products of abnormal corruption, or signs of the non-Indo-European character of the words in which they appear.
a. In a certain number of passages numerically examined (below, 75), the abnormal occurrences of lingual mutes were less than half of the whole number (74 out of 159), and most of them (43) were of ṇ: all were found more frequent in the later passages. In the Rig-Veda, only 15 words have an abnormal ṭ; only 6, such a ṭh; only 1, such a ḍh; about 20 (including 9 roots, nearly all of which have derivatives) show an abnormal ḍ, besides 9 that have ṇḍ; and 30 (including 1 root) show an ṇ.
b. Taken all together, the linguals are by far the rarest class of mutes (about 1½ per cent. of the alphabet) — hardly half as frequent even as the palatals.
47. Dental series: त् t, थ् th, द् d, ध् dh, न् n. These are called by the Hindus also dantya dental, and are described as formed at the teeth (or at the roots of the teeth), by the tip of the tongue. They are practically the equivalents of our European t, d, n.
a. But the modern Hindus are said to pronounce their dentals with the tip of the tongue thrust well forward against the upper teeth, so that these sounds get a slight tinge of the quality belonging to the English and Modern Greek th-sounds. The absence of that quality in the European (especially the English) dentals is doubtless the reason why to the ear of a Hindu the latter appear more analogous with his linguals, and he is apt to use the linguals in writing European words.
48. The dentals are one of the Indo-European original mute-classes. In their occurrence in Sanskrit they are just about as frequent as all the other four classes taken together.
49. Labial series: प् p, फ् ph, ब् b, भ् bh, म् m. These sounds are called oṣṭhya labial by the Hindu grammarians also. They are, of course, the equivalents of our p, b, m.
50. The numerical relation of the labials are a little peculiar. Owing to the absence (or almost entire absence) of b in Indo-European, the Sanskrit b also is greatly exceeded in frequency by bh, which is the most common of all the sonant aspirates, as ph is the least common of the surd. The nasal m (notwithstanding its frequent euphonic mutations when final: 212 ff.) occurs just about as often as all the other four members of the series together.
a. From an early period in the history of the language, but increasingly later, b and v exchange with one another, or fail to be distinguished in the manuscripts. Thus, the double root-forms bṛh and vṛh, bādh and vadh, and so on. In the Bengal manuscripts, v is widely written instead of more original b.
51. Semivowels: य् y, र् r, ल् l, व् v.
a. The name given to this class of sounds by the Hindu grammarians is antaḥsthā standing between — either from their character as utterances intermediate between vowel and consonant, or (more probably) from the circumstance of their being placed between the mutes and spirants in the arrangement of the consonants.
b. The semivowels are clearly akin with the several mute series in their physical character, and they are classified along with those series — though not without some discordances of view — by the Hindu grammarians. They are said to be produced with the organs slightly in contact (īṣatspṛṣṭa), or in imperfect contact (duḥspṛṣṭa).52. The र् r is clearly shown by its influence in the euphonic processes of the language to be a lingual sound, or one made with the tip of the tongue turned up into the dome of the palate. It thus resembles the English smooth r, and, like this, seems to have been untrilled.
a. The Paninean scheme reckons r as a lingual. None of the Prātiçākhyas, however, does so; nor are they entirely consistent with one another in its description. For the most part, they define it as made at “the roots of the teeth”. This would give it a position like that of the vibrated r; but no authority hints at a vibration as belonging to it.
b. In point of frequency, r stands very high on the list of consonants; it is nearly equal with v, n, m, and y, and only exceeded by t.
53. The ल् l is a sound of dental position, and is so defined and classed by all the native authorities.
a. The peculiar character of an l-sound, as involving expulsion at the side of the tongue along with contact at its tip, is not noticed by any Hindu phonetist.
b. The semivowels r and l are very widely interchangeable in Sanskrit, both in roots and in suffixes, and even in prefixes: there are few roots containing a l which do not show also forms with r; words written with the one letter are found in other texts, or in other parts of the same text, written with the other. In the later period of the language they are more separated, and the l becomes decidedly more frequent, though always much rarer than the r (only as 1 to 7 or 8 or 10).
54. Some of the Vedic texts have another l-sound, written with a slightly different character (it is given at the end of the alphabet, 5), which is substituted for a lingual ḍ (as also the same followed by h for a ḍh) when occurring between two vowels. It is, then, doubtless a lingual l, one made by breach (at the side of the tongue) of the lingual instead of the dental mute-closure.
a. Examples are ईळे īḻe, for ईडॆ īḍe, but ईड्य iḍya; मीळ्हुषॆ mīḻhuṣe, for मीढुषे mīḍhuṣe, but मीढ्वान् mīḍhvān. It is especially in the Rig-Veda and its auxiliary literature that this substitution is usual.
55. The य् y in Sanskrit, as in other languages generally, stands at the closest relationship with the vowel इ i (short or long); the two exchange with one another in cases innumerable.a. And in the Veda (as the metre shows) an i is very often to be read where, in conformity with the rules of the later Sanskrit euphony, a y is written. Thus the final i-vowel of a word remains i before an initial vowel; that of a stem maintains itself unchanged before an ending; and an ending of derivation — as ya, tya — has i instead of y. Such cases will be noticed in more detail later. The constancy of the phenomenon in certain words and classes of words shows that this was no merely optional interchange. Very probably, the Sanskrit y had everywhere more of an i-character than belongs to the corresponding European sound.
56. The y is by its physical character a palatal utterance; and it is classed as a palatal semivowel by the Hindu phonetists. It is one of the most common of Sanskrit sounds.
57. The व् v is pronounced as English or French v (German w) by the modern Hindus — except when preceded by a consonant in the same syllable, in which case it has rather the sound of English w; and European scholars follow the same practice (with or without the same exception).
a. By its whole treatment in the euphony of the language, however, the v stands related to an u-vowel precisely as y to an i-vowel. It is, then, a v only according the original Roman value of that letter — that is to say, a w-sound in the English sense; though (as was stated above for the y) it may well have been less markedly separated from u than English w, or more like French ou in oui etc. But, as the original w has in most European languages been changed to v (English), so also in India, and that from a very early time: the Paninean scheme and two of the Prātiçākhyas (VPr. and TPr.) distinctly define the sound as made between the upper teeth and the lower lip — which, of course, identifies it with the ordinarily modern v-sound. As a matter of practice, the usual pronunciation need not be seriously objected to; yet the student should not fail to note that the rules of Sanskrit euphony and the name of “semivowel” have no application except to a w-sound in the English sense: a v-sound (German w) is no semivowel, but a spirant, standing on the same articulate stage with the English th-sounds and the f.
58. The v is classed as a labial semivowel by the Hindu phonetical authorities. It has a somewhat greater frequency than the y.
a. In the Veda, under the same circumstances as the y (above, 55 a), v is to be read as a vowel, u.
b. As to the interchange of v and b, see above, 50 a.
59. Spirants. Under the name ūṣman (literally head, steam, flatus), which is usually and well represented by spirant, some of the Hindu authorities include all remaining sounds of the alphabet; others apply the term only to the three sibilants and the aspiration — to which it will here also be restricted.
a. The term is not found in the Paninean scheme; by different treatises the guttural and labial breathings, these and the visarga, or all these and anusvāra, are also (in addition to the sibilants and h) called ūṣman (see APr. i., 31 note). The organs of utterance are described as being in the position of the mute-series to which each spirant belongs respectively, but unclosed, or unclosed in the middle.
60. The स् s. Of the three sibilants, or surd spirants, this is the one of plainest and least questioned character: it is the ordinary European s — a hiss expelled between the tongue and the roof of the mouth directly behind the upper front teeth.
a. It is, then, dental, as it is classed by all the Hindu authorities. Notwithstanding the great losses which it suffers in Sanskrit euphony; by conversion to the other sibilants, to r, to visarga, etc., it is still very high among the consonants in the order of frequency, or considerably more common than both the other two sibilants together.
61. The ष् ṣ. As to the character of this sibilant, also, there is no ground for real question: it is the one produced in the lingual position, or with the tip of the tongue reverted into the dome of the palate. It is, then, a kind of sh-sound; and by European Sanskritists it is pronounced as an ordinary sh (French ch, German sch), no attempt being made (any more than in the case of the other lingual sounds: 45) to give it its proper lingual quality.
a. Its lingual character is shown by its whole euphonic influence, and it is described and classed as lingual by all the Hindu authorities (the APr. adds, i. 23, that the tongue in its utterance is trough-shaped). In its audible quality, it is a sh-sound rather than a s-sound; and, in the considerable variety of sibilant-utterance, even in the same community, it may coincide with the sh of some among ourselves. Yet the general and normal sh is palatal (see below, 63); and therefore the sign ṣ, marked in accordance with the other lingual letters, is the only unexceptionable transliteration for the Hindu character.
b. In modern pronunciation in India, ṣ is much confounded with kh; and the manuscripts are apt to exchange the characters. Some later grammatical treatises, too, take note of the relationship.
62. This sibilant (as was noticed above, 46, and will be more particularly explained below, 180 ff.) is no original sound, but a product of the lingualization of s under certain euphonic conditions. The exceptions are extremely few (9 out of 145 noted occurrences: 75), and of a purely sporadic character. The Rig-Veda has (apart from √ sah, 182 b) only twelve words which show a ṣ under other conditions.
a. The final ṣ of a root has in some cases attained a more independent value, and does not revert to s when the euphonic conditions are removed, but shows anomalous forms (225–6).
63. The श् ç. This sibilant is by all the native authorities classed and described as palatal, nor is there anything in its history or euphonic treatment to cast doubt on its character as such. It is, then, made with the flat of the tongue against the forward part of the palatal arch — that is to say, it is the usual and normal sh-sound. By European scholars it is variously pronounced — more often, perhaps, as s than as sh.
a. The two sh-sounds, ṣ and ç, are made in the same part of the mouth (the ṣ probably rather further back), but with a different part of the tongue; and they are doubtless not more unlike than, for example, the two t-sounds, written ṭ and t; and it would be not less proper to pronounce them both as one sh than to pronounce the linguals and dentals alike. To neglect the difference of s and ç is much less to be approved. The very near relationship of ṣ and ç is attested by their euphonic treatment, which is to a considerable extent the same, and by their not infrequent confusion by the writers of manuscripts.
64. As was mentioned above (41), the ç, like c, comes from the corruption of an original k-sound, by loss of mute-contact as well as forward shift of the point of production. In virtue of this derivation, it sometimes (though less often than c) “reverts” to k — that is, the original k appears instead of it (43); while, on the other hand, as a sh-sound, it is to a certain extent convertible to ṣ. In point of frequency, it slightly exceeds the latter.
65. The remaining spirant, ह् h, is ordinarily pronounced like the usual European surd aspiration h.
a. This is not, however, its real character. It is defined by all the native authorities as not a surd element, but a sonant (or else an utterance intermediate between the two); and its whole value in the euphony of the language is that of a sonant: but what is its precise value is very hard to say. The Paninean scheme ranks it as guttural, as it does also a: this means nothing. The Prātiçākhyas bring it into no relation with the guttural class; one of them quotes the opinion of some authorities that “it has the same position with the beginning of the following vowel” (TPr. ii. 47) — which so far identifies it with our h. There is nothing in its euphonic influence to mark it as retaining any trace of gutturally articulated character. By some of the native phonetists it is identified with the aspiration of the sonant aspirates — with the element by which, for example, gh differs from g. This view is supported by the derivation of h from the aspirates (next paragraph), by that of ḷ + h from ḍh (54), and by the treatment of initial h after a final mute (163).
66. The h, as already noticed, is not an original sound, but comes in nearly all cases from an older gh (for the few instances of its derivation from dh and bh, see below, 223 g). It is a vastly more frequent sound than the unchanged gh (namely, as 7 to 1): more frequent, indeed, than any of the guttural mutes except k. It appears, like j (219), to include in itself two stages of corruption of gh: one corresponding with that of k to c, the other with that of k to ç; see below, 223, for the roots belonging to the two classes respectively. Like the other sounds of guttural derivation, it sometimes exhibits “reversion” (43) to its original.
67. The ः ḥ, or visarga (visarjanīya, as it is uniformly called by the Prātiçākhyas and by Pāṇini, probably as belonging to the end of a syllable), appears to be merely a surd breathing, a final h-sound (in the European sense of h), uttered in the articulating position of the preceding vowel.
a. One Prātiçākhya (TPr. ii. 48) gives just this last description of it. It is by various authorities classed with h, or with h and a: all of them are alike sounds in whose utterance the mouth-organs have no definite shaping action.
68. The visarga is not original, but always only a substitute for final s or r, neither of which is allowed to maintain itself unchanged (170 ff.). It is a comparatively recent member of the alphabetic system; the other euphonic changes of final s and r have not passed through visarga as an intermediate stage. And the Hindu authorities are considerably discordant with one another as to how far ḥ is a necessary substitute, and how far a permitted one, alternative with a sibilant, before a following initial surd.69. Before a surd guttural or labial,respectively, some of the native authorities permit, while others require, conversion of final s or r into the so-called jihvāmūlīya and upadhmānīya spirants. It may be fairly questioned, perhaps, whether these two sounds are not pure grammatical abstractions, devised (like the long ḷ-vowel: 23 a) in order to round out the alphabet to greater symmetry. At any rate, both manuscripts and printed texts in general make no account of them. Whatever individual character they may have must be, it would seem, in the direction of the (German) ch- and f-sounds. When written at all, they are wont to be transliterated by χ and φ.
70. The ं anusvāra, ṅ or ṁ, is a nasal sound lacking that closure of the organs which is required to make a nasal mute or contact-sound (36); in its utterance there is nasal resonance along with some degree of openness of the mouth.
71. There is discordance of opinion among both the Hindu phonetists and their modern European successors respecting the real character of this element; hence a little detail is necessary here with regard to its occurrence and their views of it.
a. Certain nasals in Sanskrit are of servile character, always to be assimilated to a following consonant, of whatever character that may be. Such are final m in sentence-combination (213), the penultimate nasal of a root, and a nasal of increment (255) in general. If one of these nasals stands before a contact-letter or mute, it becomes a nasal mute corresponding the the latter — that is, a nasal utterance in the same position of the mouth-organs which gives the succeeding mute. If, on the other hand, the following consonant does not involve a contact (being a semivowel or spirant), the nasal element is also without contact: it is a nasal utterance with unclosed mouth-organs. The question is, now, whether this nasal utterance becomes merely a nasal infection of the preceding vowel, turning it into a nasal vowel (as in French on, en, un, etc., by reason of a similar loss of a nasal mute); or whether it is an element of more individual character, having place between the vowel and the consonant; or, once more, whether it is sometimes the one thing and sometimes the other. The opinions of the Prātiçākhyas and Pāṇini are briefly as follows:
b. The Atharva-Prātiçākhya holds that the result is everywhere a nasalized vowel, except when n or m is assimilated to a following l; in that case, the n or m becomes a nasal l: that is, the nasal utterance is made in the l-position, and has a perceptible l-character.
c. The other Prātiçākhyas teach a similar conversion into a nasal counterpart to the semivowel, or a nasal semivowel, before y and l and v (not before r also). In most of the other cases where the Atharva-Prātiçākhya acknowledges a nasal vowel — namely, before r and the spirants — the others teach the intervention after the vowel of a distinct nasal element, called the anusvāra after-tone.
d. Of the nature of this nasal afterpiece to the vowel no intelligibly clear account is given. It is said (RPr.) to be either vowel or consonant; it is declared (RPr., VPr.) to be made with the nose alone, or (TPr.) to be nasal like the nasal mutes; it is held by some (RPr.) to be the sonant tone of the nasal mutes; in its formation, as in that of the vowel and spirant, there is (RPr.) no contact. As to its quantity, see further on.
e. There are, however, certain cases and classes of cases where these other authorities also acknowledge a nasal vowel. So, especially, wherever a final n is treated (208–9) as if it were ns (its historically older form); and also in a small number of specified words. They also mention the doctrine of nasal vowel instead of anusvāra as held by some (and TPr. is uncertain and inconsistent in its choice between the one and the other).
f. In Pāṇini, finally, the prevailing doctrine is that of anusvāra everywhere, and it is even allowed in many cases where the Prātiçākhyas prescribe only a nasal mute. But a nasal semivowel is also allowed instead before a semivowel, and a nasal vowel is allowed in the cases (mentioned above) where some of the Prātiçākhyas require it by exception.
g. It is evidently a fair question whether this discordance and uncertainty of the Hindu phonetists is owing to a real difference of utterance in different classes of cases and in different localities, or whether to a different scholastic analysis of what is really everywhere the same utterance. If anusvāra is a nasal element following the vowel, it cannot well be any thing but either a prolongation of the same vowel-sound with nasality added, or a nasalized bit of neutral-vowel sound (in the latter case, however, the altering influence of an i or u-vowel on a following s ought to be prevented, which is not the case: see 183).
72. The assimilated nasal element, whether viewed as nasalized vowel, nasal semivowel, or independent anusvāra, has the value of something added, in making a heavy syllable, or length by position (79).
a. The Prātiçākhyas (VPr., RPr.) give determinations of the quantity of the anusvāra combining with a short and with a long vowel respectively to make a long syllable.
73. a. Two different signs, ं and ँ, are found in the manuscripts, indicating the nasal sound here treated of. Usually they are written above the syllable, and there they seem most naturally to imply a nasal affection of the vowel of the syllable, a nasal (anunāsika) vowel. Hence some texts (Sāma- and Yajur-Vedas), when they mean a real anusvāra, bring one of the signs down into the ordinary consonant-place; but the usage is not general. As between the two signs, some manuscripts employ, or tend to employ, the ँ where a nasalized (anunāsika) vowel is to be recognized, and elsewhere the ं; and this distinction is consistently observed in many European printed texts; and the former is called the anunāsika sign: but the two are doubtless originally and properly equivalent.b. It is a very common custom of the manuscripts to write an anusvāra-sign for any nasal following the vowel of a syllable, either before another consonant or as final (not before a vowel), without any reference to whether it is to be pronounced as nasal mute, nasal semivowel, or anusvāra. Some printed texts follow this slovenly and undesirable habit; but most write a nasal mute whenever it is to be pronounced — excepting where it is an assimilated m (213).
c. It is convenient also in transliteration to distinguish the assimilated m by a special sign, ṁ, from the anusvāra of more independent origin, ṅ; and this method will be followed in the present work.
74. This is the whole system of sounds recognized by the written character; for certain other transitional sounds, more or less widely recognized in the theories of the Hindu phonetists, see below, 230.
75. The whole spoken alphabet, then, may be arranged in the following manner, in order to show, so far as is possible in a single scheme, the relations and important classifications of its various members:
76. The Hindu grammarians take the pains to define the quantity of a consonant (without distinction among consonants of different classes) as half that of a short vowel.
77. They also define the quantity of a long (dīrgha) vowel or diphthong as twice that of a short (hrasva) vowel — making no distinction in this respect between the guṇa- and the vṛddhi-diphthongs.
78. Besides these two vowel-quantities, the Hindus acknowledge a third, called pluta (literally swimming), or protracted, and having three moras or three times the quantity of a short vowel. A protracted vowel is marked by a following figure 3: thus, आ३ ā3.
a. The protracted vowels are practically of rare occurrence (in RV., three cases; in AV., fifteen; in the Brāhmaṇa literature, decidedly more frequent). They are used in cases of questioning, especially of a balancing between two alternatives, and also of calling to a distance or urgently. The protraction is of the last syllable in a word, or in a whole phrase; and the protracted syllable has usually the acute tone, in addition to any other accent the word may have; sometimes it takes also anusvāra, or is made nasal.
b. Examples are: adháḥ svid āsī́3d upári svid āsī3t (RV.) was it, forsooth, below? was it, forsooth, above> idám bhū́yā́3 idā́3m íti (AV.) saying, is this more, or is that? ágnā́3i pátnīvā́3ḥ sómam piba (TS.) O Agni! thou with thy spouse! drink the soma.
c. A diphthong is protracted by prolongation of its first or a-element: thus, e to ā3i, o to ā3u.
d. The sign of protraction is also sometimes written as the result of accentual combination, when so-called kampa occurs: see below, 87 d.
79. For metrical purposes, syllables (not vowels) are distinguished by the grammarians as heavy (guru) or light (laghu). A syllable is heavy if its vowel is long, or short and followed by more than one consonant (“long by position”). Anusvāra and visarga count as full consonants in making heavy syllable. The last syllable of a pāda (primary division of a verse) is reckoned as either heavy or light.
a. The distinction in terms between the difference of long and short in vowel-sound and that of heavy and light in syllable-construction is valuable and should be observed.
80. The phenomena of accent are, by the Hindu grammarians of all ages alike, described and treated as depending on a variation of tone or pitch; of any difference of stress involved, they make no account.
81. The primary tones (svara) or accent-pitches are two: a higher (udātta raised), or acute; and a lower (anudātta not raised), or grave. A third (called svarita: a term of doubtful meaning) is always of secondary origin, being (when not enclitic: see below, 85) the result of actual combination of an acute vowel and a following grave vowel into one syllable. It is also uniformly defined as compound in pitch, a union of higher and lower tone within the limits of a single syllable. It is thus identical in physical character with the Greek and Latin circumflex, and fully entitled to be called by the same name.
82. Strictly, therefore, there is but one distinction of tone in the Sanskrit accentual system, as described by the native grammarians and marked in the written texts: the accented syllable is raised in tone above the unaccented; while then further, in certain cases of the fusion of an accented and an unaccented element into one syllable, that syllable retains the compounded tone of both elements.
83. The svarita or circumflex is only rarely found on a pure long vowel or diphthong, but almost always on a syllable in which a vowel, short or long, is preceded by a y or v representing an originally acute i- or u-vowel.a. In transliteration, in this work, the udātta or acute will be marked with the ordinary sign of the acute, and the svarita or circumflex (as being a downward slide of the voice forward) with what is usually called the grave accent: thus, á, acute, yà or và, circumflex.
84. The Prātiçākhyas distinguish and name separately the circumflexed tones arising by different processes of combination: thus, the circumflex is called
a. Kṣāipra (quick), when an acute i- or u-vowel (short or long) is converted into y or v before a dissimilar vowel of grave tone: thus, vyā̀pta from ví-āpta, apsvàntár from apsú antár.
b. Jātya (native) or nitya (own), when the same combination lies further back, in the make-up of a stem or form, and so is constant, or belongs to the word in all circumstances of its occurrence: thus, kvà (from kúa), svàr (súar), nyàk (níak), budhnyà (budhnía), kanyā̀ (kaníā), nadyàs (nadī́-as), tanvā̀ (tanū́-ā).
c. The words of both the above classes are in the Veda, in the great majority of cases, to be read with restoration of the acute vowel as a separate syllable; thus, apsú antár, súar, nadī́as, etc. In some texts, part of them are written correspondingly, thus, súvar, tanúvā, budhníya.
d. Praçliṣṭa, when the acute and grave vowels are of such character that they are fused into a long vowel or diphthong (128 c): thus, divī̀ ’va (RV. AV. etc.), from diví iva; sū̀dgātā (TS.), from sú-udgātā; nāì ’vā̀ ’çnīyāt (ÇB.), from ná evá açnīyāt.
e. Abhinihita, when an initial grave a is absorbed by a final acute é or ó (135 a): thus, tè ’bruvan, from té abruvan; sò ’bravīt, from só abravīt.
85. But further, the Hindu grammarians agree in declaring the (naturally grave) syllable following an acute, whether in the same or in another word, to be svarita or circumflex — unless, indeed, it be itself followed by an acute or circumflex; in which case it retains its grave tone. This is called by European scholars the enclitic or dependent circumflex.
a. Thus, in téna and té ca, the syllable na and word ca are regarded and marked as circumflex; but in téna té and té ca svàr they are grave.b. This seems to mean that the voice, which is borne up at the higher pitch to the end of the acute syllable, does not ordinarily drop to grave pitch by an instantaneous movement, but descends by a more or less perceptible slide in the course of the following syllable. No Hindu authority suggests the theory of a middle or intermediate tone for the enclitic, any more than for the independent circumflex. For the most part, the two are identified with one another, in treatment and designation. The enclitic circumflex is likewise divided into a number of sub-varieties, with different names: they are of too little consequence to be worth reporting.
86. The essential difference of the two kinds of circumflex is shown clearly enough by these facts: 1. the independent circumflex takes the place of the acute as the proper accent of a word, while the enclitic is the mere shadow following an acute, and following it in another word precisely as in the same word; 2. the independent circumflex maintains its character in all situations, while the enclitic before a following circumflex or acute loses its circumflex character, and becomes grave; moreover, 3. in many of the systems of marking accent (below, 88), the two are quite differently indicated.
87. The accentuation is marked in manuscripts only of the older literature: namely, in the primary Vedic texts, or saṁhitās, in two of the Brāhmaṇas (Tāittirīya and Çatapatha), in the Tāittirīya-Āraṇyaka, in certain passages of the Āitareya-Āraṇyaka, and in the Suparṇādhyāya. There are a number of methods of writing accent, more or less different from one another; the one found in manuscripts of the Rig-Veda, which is most widely known, and of which most of the others are only slight modifications, is as follows.
a. The acute syllable is left unmarked; the circumflex, whether independent or enclitic, has a short perpendicular stroke above; and the grave next preceding an acute or (independent) circumflex has a short horizontal stroke below. Thus,
b. But the introductory grave stroke below cannot be given if an acute syllable is initial; hence an unmarked syllable at the beginning of a word is to be understood as acute; and hence also, if several grave syllables precede an acute at the beginning of a sentence, they must all alike have the grave sign. Thus,
c. All the grave syllables, however, which follow a marked circumflex are left unmarked, until the occurrence of another accented syllable causes the one which precedes it to take the preparatory stroke below. Thus,
d. If an independent circumflex be followed by an acute (or by another independent circumflex), a figure 1 is set after the former circumflexed vowel if it be short, or a figure 3 if it be long, and signs of accent are applied as in the following examples:
The rationale of this mode of designation is not well understood; the Prātiçākhyas give no account of it. In the scholastic utterance of the syllable so designated is made a peculiar quaver or roulade of the voice, called kampa or vikampana.
e. The accent-marks are written with red ink in the manuscripts, being added after the text is written, and perhaps often by another hand.
88. a. Nearly accordant with this, the Rig-Veda method of designating accent, are the methods employed in the manuscripts of the Atharva-Veda, of the Vājasaneyi-Saṁhitā, and of the Tāittirīya-Saṁhitā, Brāhmaṇa, and Āraṇyaka. Their differences from it are of trifling importance, consisting mainly in peculiar ways of marking the circumflex that precedes an acute (87 d). In some manuscripts of the Atharva-Veda, the accent-marks are dots instead of strokes, and that for the circumflex is made within the syllable instead of above it.
b. In most manuscripts of the Māitrāyaṇī-Saṁhitā, the acute syllable itself, besides its surroundings, is marked — namely, by a perpendicular stroke above the syllable (like that of the ordinary circumflex in the RV. method). The independent circumflex has a hook beneath the syllable, and the circumflex before an acute (87 d) is denoted simply by a figure 3, standing before instead of after the circumflexed syllable.
c. The Çatapatha-Brāhmaṇa uses only a single accent-sign, the horizontal stroke beneath the syllable (like the mark for grave in RV.). This is put under an acute, or, if two or more acutes immediately follow one another, only under the preceding syllable. To mark an independent circumflex, it is put under the preceding syllable. The method is an imperfect one, allowing many ambiguities.
d. The Sāma-Veda method is the most intricate of all. It has a dozen different signs, consisting of figures, or of figures and letters combined, all placed above the syllables, and varying according to both the accentual character of the syllable and to its surroundings. Its origin is obscure; if anything more is indicated by it then by the other simpler systems, the fact has not been demonstrated.
89. In this work, as everything given in the devanāgarī characters is also given in transliteration, it will generally be unnecessary to mark the accent except in the transliterated form; where, however, the case is otherwise, there will be adopted the method of marking only the really accented syllables, the acute and the independent circumflex: the latter by the usually svarita-sign, the former by a small u (for udātta) above the syllable: thus,
90. The theory of the Sanskrit accent, as here given (a consistent and intelligible body of phenomena), has been overlaid by the Hindu theorists, especially of the Prātiçākhyas, with a number of added features, of a much more questionable character. Thus:
a. The unmarked grave syllables following a circumflex (either at the end of a sentence, or till the near aproach of another acute) are declared to have the same high tone with the (also unmarked) acute. They are called pracaya or pracita (accumulated: because liable to occur in an indefinite series of successive syllables).
b. The circumflex, whether independent or enclitic, is declared to begin on a higher pitch than acute, and to descend to acute pitch in ordinary cases: the concluding instant of it being brought down to grave pitch, however, in the case of an independent circumflex which is immediately followed by another ascent of the voice to higher pitch, in acute or independent circumflex (a kampa syllable: 87 d).
c. Pāṇini gives the ambiguous name of ekaçruti (monotone) to the pracita syllables, and says nothing of the uplifting of the circumflex to a higher plane; he teaches, however, a depression below the grave pitch for the marked grave syllable before acute or circumflex, calling it sannatara (otherwise anudāttatara).
91. The system of accentuation as marked in the Vedic texts appears to have assumed in the traditional recitation of the Brahmanic schools a peculiar and artificial form, in which the designated syllables, grave and circumflex (equally the enclitic and the independent circumflex), have acquired a conspicuous value, while the undesignated, the acute, has sunk into insignificance.
92. The Sanskrit accent taught in the native grammars and represented by the accentuated texts is essentially a system of word-accent only. No general attempt is made (any more than in the Greek system) to define or mark a sentence-accent, the effect of the emphasis and modulation of the sentence in modifying the independent accent of individual words. The only approach to it is seen in the treatment of vocatives and personal verb-forms.
a. A vocative is usually without accent except at the beginning of a sentence: for further details, see 314.
b. A personal verb-form is usually accentless in an independent-clause, except when standing at the beginning of the clause: for further details, see 591 ff.
93. Certain other words also are, usually or always, without accent.a. The particles ca, vā, u, sma, iva, cid, svid, ha, and the Vedic kam (or kám), gha, bhala, samaha, īm, sīm, are always without accent; also yathā in RV. (sometimes also elsewhere) in the sense of iva, at the end of a pāda or verse-division.
b. The same is true of certain pronouns and pronominal stems: mā, me, nāu, nas, tvā, te, vām, vas (491 b), ena (500), tva (503 b), sama (513 c).
c. The cases of the pronominal stem a are sometimes accented and sometimes accentless (502).
d. An accentless word is not allowed to stand at the beginning of a sentence; also not of a pāda or primary division of a verse; a pāda is, in all matters relating to accentuation, treated like an independent sentence.
94. Some words have more than a single accented syllable. Such are:
a. Certain dual copulative compounds in the Veda (see 1255), as mitrā́váruṇā, dyā́vāpṛthivī́. Also, a few other Vedic compounds (see 1267 d), as bṛ́haspáti, tánūnápāt.
b. In a few cases, the further compounds and derivatives of such compounds, as dyā́vāpṛthivī́vant, bṛ́haspátipraṇutta.
c. Infinitive datives in tavāí (see 972 a), as étavāí, ápabhartavāí.
d. A word naturally barytone, but having its final syllable protracted (see 78 a).
e. The particle vā́vá (in the Brāhmaṇas).
95. On the place of the accented syllable in a Sanskrit word there is no restriction whatever depending upon either the number or the quantity of the preceding or following syllables. The accent rests where the rules of inflection or derivation or composition place it, without regard to any thing else.
a. Thus, índre, agnāú, índreṇa, agnínā, agninā́m, bāhúcyuta, ánapacyuta, parjányajinvita, abhimātiṣāhá, ánabhimlātavarṇa, abhiçasticā́tana, híraṇyavāçīmattama, cátuçcatvāriṅçadakṣara.96. Since the accent is marked only in the older literature, and the statements of the grammarians, with the deduced rules of accentuation, are far from being sufficient to settle in all cases, the place of the stress of voice for a considerable part of the vocabulary is undetermined. Hence it is a general habit with European scholars to pronounce Sanskrit words according to the rules of the Latin accent.
97. In this work, the accent of each word and form will in general be marked, so far as there is authority determining its place and character. Where specific words and forms are quoted, they will only be so far accentuated as they are found with accent in accentuated texts.