Pramāṇa · Search · About · API

Elements of South-Indian Palæography, from the Fourth to the Seventeenth Century A.D. c10: Appendix A

Elements of South-Indian Palæography, from the Fourth to the Seventeenth Century A.D. c10

Language
English

Contents

Appendix A

c10.1

FOR the successful interpretation of the S. Indian inscriptions, as well as for extended researches into Dravidian Comparative Philology, it is now indispensable that a history of Dravidian phonetics should be drawn up. The materials that exist for this purpose are more extensive than might be supposed, and go back to perhaps nearly two thousand years. The earliest certain traces are a few words recorded in the Açoka inscriptions, and, later, a few more by the Greek geographers of the early centuries A. D.; secondly, some Tamiḻ words mentioned by Kumārilasvāmin (700 A. D.), and others in the Mahāvaṃso and in the travels of Chinese pilgrims; thirdly, the earlier inscriptions recording the campaigns of the Cālukyas and Coḷas; fourthly, the native grammarians of about the tenth century A. D. for the most part. Much help will also be gained from the earlier metrical compositions[^1). The Cera inscriptions show that the Canarese language had the peculiarities which now characterise it, already, in the 5th century A. D.; and Tamiḻ inscriptions of a date a few centuries later prove the same of that language. It is, therefore, almost certain that the three great Dravidian languages had, already, separated and assumed their characteristic forms some two thousand years ago[^2).

c10.2

An investigation of this nature is important from a palæographical point of view; but, at present, I can do no more than show with reference to the propositions I have advanced above (on p. 40):

c10.3

i. That the Tamiḻ alphabet has always been, and is still, a very imperfect system for expressing the Tamiḻ sounds, and that it is not adapted from a Sanskrit prototype.

c10.4

ii. That the Canarese and Telugu alphabets are adaptations of the Sanskrit alphabet, and are tolerably perfect expressions of the sounds found in those languages.

c10.5

The Dravidian languages naturally separate into two classes—the Telugu which stands by itself, and the Tamiḻic dialects which comprehend all the other languages of S. India. As far, however, as the history of the expression by alphabetic signs of the sounds used in these languages is concerned, the Tamiḻ and old Malayāḷam stand apart; the Canarese and Telugu must be classed together.

c10.6

[^1): Dravidian words adopted in Sanskrit, and they are many, are too much disfigured and of too uncertain source, to deserve a place in this list of materials for the phonetic history of these languages. [^2): The grants to the Israelites and Syrians and other inscriptions from the W. Coast prove that Tamiḻ and Malayāḷam were really the same language in the 8th century A. D. The dialogues in Varthema show that colloquial Malayāḷam was the same in 1503-8 as it is now. *Cfr.* also No. ii. of my "Specimens of S. Indian Dialects", preface.

§ 1. Tamil Phonetics

c10.s1.1

As the Tamiḻ alphabet now stands it is a very imperfect representation of the sounds to be met with in Tamiḻ.

c10.s1.2

There are at present vowel-marks for a, ā, i, ī, u, ū, ĕ, ē, ai, ŏ, ō and au; but of these in addition to the usual pronunciation of u and ai, these two letters have very commonly the value of ṳ, and this is noticed by the earliest grammarians¹. Again a, i, ī, and ū have distinct secondary² values in some cases, viz., they become ‘mixed’.

c10.s1.3

¹ Tŏlkāppiyam, i., 2, 24. Nannūl, ii., 6, etc.

c10.s1.4

² The cause of this I have been able to discover by means of Mr. Melville Bell's admirable book “Visible Speech”. These simple vowels are effected by the *following* consonant when it closes the syllable in certain cases. These consonants are ṭ, ḻ and ḷ, but at the end of a syllable they necessarily induce modification of the vowels. As Mr. Bell (p. 75) says: “The various positions of the tongue which produce ‘centre-aperture’ consonants, form vowels when the channel between the organs is sufficiently expanded and firm to allow the breath to pass without oral friction or sibilation. The vowel positions thus bear a definite relation to the consonant attitudes of the different parts of the tongue.”

c10.s1.5

These values occur in certain definite circumstances, but they are so numerous as to render the Tamiḻ alphabet very defective as far as the vowels are concerned.

c10.s1.6

The expression of the consonants is also defective³.

c10.s1.7

³ It is quite certain that the Tamiḻ alphabet was always limited in extent, for the Tŏlkāppiyam (i., 1, 1) and Nannūl (ii., 4) expressly put the number of *letters* at thirty. The Nannūl (ii., 8) says also: “Beginning with a, twice six are vowels beginning with k, (there) are thrice six consonants: thus say the learned.” It is remarkable that only in one system of writing and that just deciphered, we find the same character used to express both the sonant and surd consonants of the same class; I mean the mysterious Cypriote syllabary (*“Journal des Savants”*, Sept. 1877, p. 560). At the very first view it is easy to trace striking resemblances between some of the Cypriote characters and Vaṭṭeḻuttu letters (e. g. [Figure:] = **ϰε, ϰη; χε, χη; γε, γη**); but the syllabic nature of the former makes it difficult to suppose any real connection between the two in their forms as now known. Further research into the systems of writing once used in the Levant and in the Aramean country will, I have no doubt, eventually clear up the origin of the Vaṭṭeḻuttu.

c10.s1.8

Thus the following letters have distinct values:—

c10.s1.9

| *Letter* | *1. Initial* | *2. Medial* | *3. Medial (if doubled)* |
| k | = k | ɣ | k |
| č | = č | š, ǰ | ţ₈ |
| ṭ | = ṭ | ḍ | ṭ |
| t | = t | δ | t |
| p | = p | b | p |

c10.s1.10

According to the pronunciation of some places k following a nasal = g, and t following a nasal = d; but it is impossible to ascertain now if this was originally the case⁴.

c10.s1.11

⁴ In Canarese and Telugu, as spoken in some places, ద (d) has distinctly the value of ð; but not everywhere.

c10.s1.12

Now the earliest specimens of Tamiḻ that are to be found in foreign records show that the language then possessed these sounds for which there are no separate alphabetic characters, and which seem to have puzzled the Tamiḷ grammarians who leave them unnoticed[^1). These words are as follows:

c10.s1.13

In the second Girnar tablet of Açoka’s edict (ç. 250 B. C.) we find Pā(ṇ)ḍā as the name of a king; there can be no doubt that Pāṇḍiyan or the Madura king is here intended; and Pliny, Ptolemy and the Periplus also have Pandion.

c10.s1.14

The next traces we find are in Ptolemy and the Periplus of the Red Sea which may be put as representing Tamiḷ from the first to the third centuries A. D., and Kumārila Bhaṭṭa who lived in the 7th century. As regards the various powers of some of the vowels there is not much satisfactory evidence to be found[^2), but the evidence regarding the consonants is conclusive. It is as follows:

c10.s1.15

k, γ. 1) Sangara (=šaṅyāḍam) in Periplus Maris Eryth. § 60. 2) Sangamarta = Tam. šaṅga-maratta; (*i. e.* the town or camp by the *Monetia Barlerioides* trees; a station of the Nomad Sorae. Ptolemy vii., 1, § 68). 3) Bēttigō (*Ptolemy* vii., 1, § 68) which Dr. Caldwell has rightly identified with the Pōḍiyai mountain. 4) Μαγουρ (do:).

c10.s1.16

ṭ, ḍ. 1) Pandion = Pāṇḍiyan. (Periplus Maris Eryth. § 58. Ptolemy vii., 1, §§ 11 & 79. Pliny, vi., 105.) 2) Tundis, *i. e.* the Tam. tuṇḍi (Periplus Maris Eryth. § 54. Ptolemy, vii., 1, § 8). 3) Cottonara (Pliny, vi., 104); the last part is here evidently nāḍu (country) and the expression of ḍ by r is also found in the ‘sangara’ of the Periplus. 4) Kumārila has naḍer = naḍai[^3).

c10.s1.17

t, d. 1) Kolandiophōnta (Periplus Maris Eryth. § 60). The first part of this name for boats or ships is most probably the Tam. kuḷiṅda = hollowed; the last, ōḍam = boat[^4). 2) Modoura = Maḋurai. (Ptolemy, vii., 1, § 89. Pliny, vi., 105.) 3) Puḋu-paṭṭana (Cosmas)—πουδοπατανα. 4) ποδοπερουρα in Ptolemy.

c10.s1.18

p. b. 1) Kēprobotros = Keraputra (Periplus M. Er. § 54). The b here clearly shows the influence of the Tamiḷ pronunciation. Pliny (vi., 104) has Caelobothras. 2) Αρεμβουρ (Pt.) 3) Kumārila has pāṁb or pāṁp = pāmbu. The best MSS. I now find have pāṁb.

c10.s1.19

It would be easy to add other words from the Greek geographers which point to this fact, but as their identification presents more or less difficulty, I shall omit them here.

c10.s1.20

The omission of the Tamiḷ grammarians to notice this fact that the consonants have

c10.s1.21

[^1): Except they intended to include them under vague statements of irregularities of pronunciation.—Nannūl, ii., 33, etc., copying Tōlk. i., 3, 6. [^2): Except in the words which occur in Kumārila Bhaṭṭa, and as these neglect the final u (as it is now written), it is safe to assume that it was then pronounced ŭ as is the case at present, and was therefore neglected in the Nāgari transcriptions as being a sound unknown to the Sanskrit alphabet, and almost imperceptible to foreigners. [^3): I have already discussed the passage where these words occur in the *Indian Antiquary*, vol. i., pp. 309-310. Dr. Caldwell (C. Gr. 2nd ed. p. 5 *note*) has misunderstood what Kumārila says about ‘āḷ’. [^4): It seems to be contrasted with the ‘šaṅyāḍam’ or raft.

c10.s1.22

double values (viz., as surds and sonants) is unaccountable except that they had to deal with a language already reduced to writing. Tamiḷ words, however, appear to have puzzled northern and Singalese authors, and they evidently were aware that the Tamiḷ and Sanskrit or Pāli t did not mark the same sounds. Thus the Pāli has Damila; the Sanskrit Dramila, just as Ziegenbalg in his Tamiḷ Grammar (1716) calls the language “Lingua Damulica,” though Baldæus (1672) being a Dutchman has T[^1). To show how the Dravidian sounds differ from the Sanskrit sounds indicated by the same letters would take too much space to be admissible here, and would need the use of special type. Since Mr. Melville Bell’s “Visible Speech” has been published, and the Prātiçākhyas have been edited by Profr. Whitney and others, (to which may be added the use of the phonograph), an enquiry of this kind need not present any special difficulties. At the present stage of philological research in S. India it is indispensable.

c10.s1.23

The Tamiḷ alphabet differs from the other Dravidian alphabets in using ṉ (ன்) which is simply a *final* n (*i. e.* of the syllable), and is therefore unnecessary according to the S. Indian system. It is here, however, a distinct letter as it was in the Vaṭṭeḷuttu, and in its original form not unlike the Sassanian [Figure:] generally read *man*.

c10.s1.24

It follows, then, that the pronunciation of Tamiḷ cannot have changed materially since the third century B. C.; but, as it is impossible to put the introduction of writing into the Tamiḷ country at so early a date, it is evident that the Tamiḷ alphabet is an imperfect expression of the phonetic system of that language from its origin, and that it cannot have become so by progress of phonetic decay[^2). As the alphabets used in

c10.s1.25

[^1): So the Peutingerian Map and the Ravenna geographer (ed. Parthey, pp. 14, 40, etc.) have Dimirice (*i. e.* Tamiḷ + ikē) which is the proper reading for the name, and not Limurikē as printed in the Periplus and Ptolemy. [^2): The utter uncertainty of S. Indian chronology renders it difficult to use the Tamiḷ literature for purposes of illustrating the history of Dravidian phonetics; but I can help, in a small way, to clear up the existing darkness. Buddhamittra (a Buddhist of the Coḷa country, and apparently, a native of Malakūṭa or Malaikkūṛṛam) wrote in the 11th century a Tamiḷ Grammar in verse, with a Commentary by himself, which he dedicated to the then reigning Coḷa king and called after him ‘Vīraśōḷiyam’. This C. cites a great number of Tamiḷ works current in the 11th century, and is therefore of much historical importance, for the approximate dates even of most Tamiḷ works are hardly known. He cites: Amṛitasāgaram; Avinayanār: Ārūrkōvai; Ĕliviruttam; Kapilar; Kamban; Kaviviruttam; Kākkaipāḍiniyār; Kātantra(s); Kānḍi; Kundalakēśiviruttam; Kuraḷ; Šaṅyaiauthors; Šintāmaṇi; Šoḷarājavariśai; Taṇḍi; Tiruśśiṛṛambalakkōvai; Tirumaṇṇivaḷaru; Tŏlkāppiyam; Nambi; Naḷavĕnbā; Nariviruttam; Nāḷaḍiyār; Nīyāyaśšūḍāmaṇi; Nēminādam; Pĕrundēvar’s Bhārata (Vĕnbā); Maṇippiravaḷam; Mayēśuranār; Vīraśōḷaṇmēṛkavi. This then represents the *old* Tamiḷ literature prior to the 11th century, and to it must be added the older Çaiva works. The above mentioned literature cannot be older than the 8th century, for in the 7th century Hiouen Thsang expressly states that the Tamiḷ people were then indifferent to literature. That this literature arose under N. Indian influences and copied N. Indian models, can hardly be disputed; but it is time now to assert that it is nothing more than an exact copy; if there be any *originality*, it is in some of the similes and turns of expression only. This was, long ago, remarked by Mr. Curzon (*J. R. As. Soc.*), and has since been emphatically asserted by M. J. Vinson. (“Le Verbe dans les langues Dravidiennes”, pp. viii-ix. and in Hovelacque’s “La Linguistique”.) Some have supposed, but without the least reason or evidence, that this Sanskritizing literature supplanted an older Tamiḷ literature of an indigenous growth.

c10.s1.26

the Açoka inscriptions prove, the Sanskrit grammarians had already extended the alphabet to suit their marvellously accurate discrimination between the different sounds of the language in the 3rd century B. C.; it is impossible, therefore, to suppose that the Tamil alphabet is to be attributed to them. Besides their treatment of the Canarese and Telugu phonetics is totally different, as I shall now show, though the Canarese grammar was formed on the same model as the Tamil.

§ 2. Canarese Phonetics

c10.s2.1

The Hindu civilization of the Canarese country is quite as old as that of the Tamil people, but the earliest traces we find of writing are in a modified form of the Açoka character, and the orthography, with a few unimportant exceptions and allowing for the obsolete form of the letters, is just what we find now. About the tenth century A. D. Canarese grammar was treated on the principles of the Sanskrit grammarians of the Aindra school[^1), and with steady reference to Sanskrit phonetics; the author of the Canarese Grammar “Çabdamaṇidarpaṇa” evidently considered the alphabet he used as a mere adaptation from the Sanskrit, and he was perfectly right in doing so. His account is as follows[^2):

c10.s2.2

There are fourteen Sanskrit-Canarese vowels (a, ā, i, ī, u, ū, ṛi, ṝi, lṛi, lṝi, e, ai, o and au) and in Canarese e and o have both long and short forms. There are 34 Sanskrit-Canarese consonants classed (vargāxara) and unclassed (avargāxara), that is to say, the ordinary Sanskrit alphabet with xa; but of these aspirates are not used in Canarese, except in some peculiar cases. To these are added the peculiarly Canarese letters r̥, ḷ̥ and ḷ. The author then states (p. 44) that there are only 47 letters in pure Canarese—a, ā, i, ī, u, ū, ĕ, ē, ai, ŏ, ō, au; k, kh, g, gh, ṅ, c, ch, j, jh, ń, ṭ, ṭh, ḍ, ḍh, ṇ, t, th, d, dh, n, p, ph, b, bh, m, y, r, l, v, s, h, x, r̥, ḷ̥, ḷ. The Sanskrit prepossessions of the author have induced him to include erroneously the aspirates and x; h is the modern representative of p. Rejecting these letters, therefore, the remainder represent very nearly the sounds we find really to exist in Tamil.

c10.s2.3

This Canarese Grammar is, like the Tamil Tölkāppiyam and Nannūl, a very complete work, and is really what it professes to be.

c10.s2.4

[^1): As to what is to be understood by the Aindra Grammar see my book on this subject, 1875. [^2): Kittel’s “Çabdamaṇidarpaṇa”, pp. 13-45.

§ 3. Telugu Phonetics

c10.s3.1

Here again the grammar has been formed on Sanskrit models, but the pattern is either Pāṇini's or Hemacandra's treatise, and the terminology, chiefly, that of Pāṇini¹.

c10.s3.2

¹ The dates of Nannaya Bhaṭṭa and Ātharvaṇācārya can easily be fixed. Nannaya Bhaṭṭa translated the first part of the Mahābhārata into Telugu for Vishṇuvardhana who was Rāmānujācārya's chief convert, and therefore lived in the middle of the 11th century ["Cyclic Table" by C. P. Brown; *Madras Journal*, x., p. 52; Brown's "Telugu Grammar" (2nd ed.), p. i.]. Ātharvaṇācārya is generally supposed to have preceded Nannaya; but this cannot be the case, as he twice cites Hemacandra by name ("Triliṅgaçabdānuçāsana", i., 5; iii., 13 of the Madras MS.). Hemacandra was probably born in 1088 A. D. and died in 1172 A. D. (*Bombay Journal*, x., p. 224); Ātharvaṇācārya must, therefore, have written about fifty years later than Nannaya, and was probably a Jain rival of this Brahman.

c10.s3.3

The earliest of the two grammars is by Nannaya; he begins by saying that Sanskrit has fifty letters, Prakrit ten less, but the Telugu has thirty-six, as the other letters only occur in Sanskrit words which have been adopted in that language. These letters he says are: a, ā, i, ī, u, ū, ĕ, ē, ai, ŏ, ō, au, two anusvāras (o and c ), k, g, two č (č and ţ), two ǰ (ǰ and ḍ), ṭ, ḍ, ṇ, t, d, n, p, b, m, y, r, l, v, s, ṟ, ḷ².

c10.s3.4

² "Āndhraçabdacintāmaṇi", i., 14-18 and 23.

c10.s3.5

Ātharvaṇācārya is by no means so precise, but as he is later than Nannaya what he says is of little importance. He mentions seven or (excluding ai and au) five vowels (*i. e.* a, i, u, e, o) which might be short, long or *pluta*³. He does not specifically enumerate the consonants.

c10.s3.6

³ "Triliṅgaçabdānuçāsana", i., 8-11. "Prāṇāḥ sapta svarūpeṇa" (8) "*vacam* (*read* aucam) vinā svarāḥ pañca hrasvadīrghaplutais tridhā" (9).

c10.s3.7

Thus two Telugu grammarians, not of the Aindra school, have treated the Telugu alphabet far more completely than was done by Aindra grammarians in respect of the Tamiḷ, though the Telugu grammarians hold the strange theory that the Telugu language is a "Vikṛiti" of Sanskrit⁴, and treat the grammar as a mere appendix to Sanskrit and Prakrit grammar.

c10.s3.8

⁴ "Āndhraçabdacintāmaṇi", i., 12. iii., 8. 43. 59. 83. iv., 2. 11. 23. 28. 42. 46. Thefirst of these sūtras is: "Ādyaprakṛitiḥ prakṛitiç cā 'dye, eshā tayor bhaved vikṛitiḥ." Ahobala says on this: " 'Ādyaprakṛitiḥ' iti sarvabhāshāmūlakatvena Āndhrabhāshāhetutvena cā 'dyā Saṁskṛitabhāshā."—" 'Eshā' Āndhrabhāshā."

c10.s3.9

This theory is an important one in considering references to foreign words in Sanskrit grammatical works, and has been, as yet, quite misunderstood. The meaning of the term vikṛiti, as thus used, is as follows: The grammarians (as is required in Hindu cosmogony⁵) considered all languages to be eventually derived from the Sanskrit, much as in Europe, in the Middle Ages, Hebrew was supposed to be the source of all languages then known; they also considered merely the *external* forms of words and not the *meaning*[^1). It was thus easy to find a plausible explanation of any foreign words by means of Sanskrit. The Mīmāmsists contended against this doctrine, as they attached more importance to the *meaning* than to the *form*[^2). In considering foreign words mentioned by Sanskrit grammarians it is necessary to keep the nature of this theory in view.

c10.s3.10

⁵ *See* Muir's "Sanskrit Texts" (i., pp. 480 ffg.) where several passages are to be found in which it is asserted that peoples of quite different races, e. g. Oḍras, Draviḍas, Kāmbojas, Yavanas and Cīnas (Manu, x., 43-4); Yavanas, Cīnas, Pahlavas, Āndhras and Kāmbojas (Çāntiparvan); Çakas, Yavanas, Kāmbojas, Colas and Keralas (Harivaṁça) were originally Xattriyas. This notion is found already in the Brāhmaṇas.

c10.s3.11

Comparing the Telugu-Canarese alphabets with the Tamiḷ it is, then, impossible to suppose that the last is the work of Sanskrit grammarians; for had they been the authors of it, it would have been far more perfect[^3), and would have shown signs of adaptation which are wanting in it. Add to this that the Tamiḷ letters ḷ, ḻ[^4) and ṟ are totally distinct from the Telugu-Canarese corresponding letters and ṉ superfluous, and the amount of proof that the Vaṭṭeḻuttu is of independent origin, and not derived from the S. Açoka character, appears to be conclusive[^5).

Linked texts (7)