Oxford Unwerstty Press, Amen House, London EC 4 GLASGOW NEW YORK TORONTO MELHOURNE WELLINGTON BOMBAY CALCUTTA MADRAS KARACHI CAFE TOWN IBADAN Geoffrey Cumberlege, Publisher to the Untwersity PRINTED IN CREAT BRITAnSPREFACE ‘Tus book 1s intended as a guide to the appreciation of the earliest phoneticians Whitney’s pioneer expositions of certain of our sources some eighty years ago are acknowledged 1n the introduc tory chapter, but a general reinterpretation has now long been overdue, and this fact zsinitself suggestive of the remarkable quality ofthe Indiantexts Foritimplies that they display alevel of phonetic discourse beyond the full comprehension of Whitney and his con- temporaries, such as only the advances of the late nineteenth and the twentieth centuries enable us to appreciate today The recognt tion that analyses so advanced in their technique should have been evolved at so early a date may well imspire a salutary scientific humility, and it would be at once arrogant and pessimistic not to expect that a reinterpretation will again be necessary in another eighty years—or even eight Iam grateful to Professor J R Furth for the encouraging interest he has shown at all stages in the progress of this work, and no less for his constructive suggestions, and I am happy to acknowledge the researches of Dr Sidgheshwar Varma, the stimulus of whose published work has been augmented for me by the background of his personal association with Professor Firth over twenty years ago—an association which has led, in the light of contemporary ingurstics, to a fuller realization of the wealth that lies in the ancient treatises My thanks are also due to Professor J Brough, who read the work in manuscript and made a number of helpful criticisms, and to Mr C A Rylands and Mr R H Robins, who generously undertook to read the whole of the proofs during my absence in India Finally I acknowledge the generosity shown by the authoritres of the School of Orrental and Afmcan Studies in providing a full subvention for the publication of this work Ws ALLEN DEPARTMENT OF PHONETICS AND LINGUISTICS SCHOOL OF ORIENTAL AND AFRICAN STUDIES: UNIVERSITY OF LONDON 71953CONTENTS PREFACE EDITIONS OF TECHNICAL WORKS REFERRED TO INTRODUCTION oo The Grammatical Achievement or The Indian Influence on Western Phonetics o2 The Sources 03 The Sansknt Alphabet 04 The Principles of Description 940 Word and Sentence 041 Phonetics and Phonology o42 Terminology 943 Onder of Analysis Chart of Sanskrit Alphabet ° PART I PROCESSES 10 Mental 11 “Physiological 110 Classification 111 Intra buccal 1110 Vowels and Consonants 1irr Fricatives 1112” Semsvowels ‘Liquids’ 1113 Retroflexion 112 Extra buccal 1120 Glottal 1321 Pulmonice 1122 Nasal Anusvara 113 The vargas 21 22 22 24 24 26 2 31 32 33 33 37 40 46vw CONTENTS PART II LETTERS 20 Consonants 200 Pulmonic and Glottal 201 Velar 202 Palatal 203 Retroflex 204 Dental 205 Labial 21 Vowels 210 a 211 1, 212 t1 213 @,0, at, au PART It] PROSODIES 30 Definition 31 Junction 3.10 Word and Morpheme yunctfon 3.11 Inttuality and finality 312 Letter junction 3.120 Consonant-|-stop (abkimdhana) 3 121 r-+consonant (svarabhakt:) 3.122 Stop-nasal (yama) 3123 Fnicative-+nasal 3.124 Stop+-fricative 3125 Gemunation 32 Syllable structure 3.20 Vowel and Consonant 321 Syllabic Division 322 Length and Duration 323 Quantity 3,24 Tone 33 Tempo INDEX 48 48 51 52 52 56 57 57 57 61 61 62 65 65 66 69 79° qt 23 75 78 78 79 79 79 8r 83 87 93 95EDITIONS OF TECHNICAL WORKS REFERRED TO N.B Where more than one edition ts noted, an asterisk signafies that the edition in question has been adopted for purposes of text and numbering (exceptions are duly indicated) + Apiak-Stksa (Ap S$) Ed Raghu Vira, JVS1 2, 1934, pp 225 ff Atharva-Pratisakhya (AP) (= Saunakiyd Caturadhyayska) Ed and trsi W D Whitney, AOS vu 333 Atharva-Pratiakijya (See p 5,7 1) *Ed and trsl S K Shastr: Lahore, 1939 Ed V B V Shastr: Lahore, 1923 Bharadvdja-Siksa Ed and trsl E Steg Berlin, 1892 *Ed V R R Dikshitar and P S S Ayyar Poona, 1938 Kauhahi-Siksa Ed Sadhu Ram, JVS u 1, 1935, pp 108 ff Mahabhasya (iMbh ) "Ed F Kiclhom Bomb&y, 1892-1909 With Prad:pa of Kasyata and Pradipoddyotana of Annambhatta, ed P P § Sastn (Part I, Ahmkas 1-4) Madras, 1948 Trsl V Trapp (AAmkas 1-5) Leipzig, 1933 Pamm (Pan ) *Ed and trsl O Bohthngk Leipzig, 1887 Ed and trs!) S Chandra Vasu Allahabad, 1891-7 Ed © Bohthngk Bonn, 1839-40 Trsl L Renou (Fase I, Adhyaya, tun} Pans, 1948 Panimya Stkga (PS) Ed andtrsl A Weber, Indische Studien, iv, 1858, pp 345-71 *With the Panimtyasiksapradipa and Staracaid:kaprakaranapankuvi- tarana, ed R P Sharma Benares, 1937 With the Siksé Pofipka and S:ksa Prakasa, ed and tral M Ghosh Calcutta, 1938 Pingala, Chandah Sutra Ed andtrs! A Weber, Indische Studien, vin, 1863, pp 269 ff *With the commentary of Haldyudha,ed Vo M Shastrn (Bibl Indica) Calcutta, 1874 Pratyfia Sutra A Weber, Uber etn sum werssen Yayus gehoniges phonetrsches Compen- dium, das Pratyita Satra(Abh d hin Ak d Wis = Berlin, 1871)x EDITIONS OF TECHNICAL WORKS REFERRED TO " Bk Pratssakhya (RP) ‘With the commentaty of Uvata. *Ed and trs) M D Shastri Allahabad, 1931, Lahore, 1937 With extracts from Uvata Ed and trsl F Max Muller = Rig-Veda, vol 1 pp 1-ccexev Lepag 1856 Ed and trs] Regnier, Fournal Anatique, V sérte, vols vu-xn Rktantravyakavana (RT) *Ed S K Shastri Lahore, 1933 Ed A C Burnell Mangalore, 1879 Sarkmreya-Sthsa Ed T Chowdhury, JVS u 2 1935 pp 197 ff Sarvasammata Siksd Ed andtrsl A O Franke Gédttngen, 1886 Siksd-Samgraha (SS) A Collectron of Stksas by Yajtiavalkya and others, ed Yugalakssora Vyasa (Benares Skt Sertes) Benares, 1893 Taittireya PratiSakhya (TP) With the Tnbhasyaraina *Ed and trsl W D Whitney, JAOS 1x, 1871 Ed Rayendralala Mitra (Bibtiotheca Indica), Calcutta 1872 With the Padakramasadana of Mahiseya Ed V V Sharma Madras, 1930 e ‘With the commentaries of Somayarya and Gargya Gopalayayvan Ed K. RangacharyaandR § Sastrs(Bibl Sansk 33) Mysore, 1906 Upalekha Ed and trsl W Pertsch Berlin, 1854 Vajasaney: Pratifakhya or Katyayanzya Prat (VP) With the commentaries of Uvata and Anantabhatta *Ed V V Sharma Madras, 1934 With extracts from the commentanes Ed. and trsl. A. Weber, Indische Studten, wv, 1858, pp 65 17%) 177— 331 . Varnapatalam The Parsustas of the Atharvaveda, xlvu, ed Bollmg and Negelem Leipzig 1969-10" Vyasa S:ksa (VS) “H Luders, Die Vyasa Siksa Géttingen, 1894 Ed V V Sharma Madras, 1929 Ed K V Sastn (Grantha text) Tiruvads, 1908INTRODUCTION . oo The Grammatical Achievement In the sphere of grammar it 15 a gratifying custom of present-day linguists to pay Inp-service to the greatest of descriptive gram- manians, the ancient Indian Pimim and it was an eloquent tribute to his achievement that one of the great linguists of our own time should write Indo-European comparative grammar had (and has) at its service only one camplete desceuntion of a Language, the grammar of Parnnt Far all, other Indo-European languages st had only the traditional grammars of Greek and Latin, wofully incomplete and unsystematic For no lan= guage of the past have we a record comparable to Panim’s record of his mother tongue, nor 1s xt likely that any language spoken today will be so perfectly recorded ? But an spite of the invaluable translations of his work by Bohtlingk and now by Renou, unless the Imguist 1s himself also a Sanskritist there are insuperable difficulties 1n the way of a full apprecration of Panim's achievement, and even for the Sanskntist a complete understanding 1s not easfly attamed—again to quote Bloomfield, Even with the many commentafies that we possess —_—several lifetimes of work will have ta be spent upon Panini before we have a conveniently usable exposition of the language which he recorded for all tme Tt 1s indeed in the extent of the interpretative material, some of which has stself attained to a canonical status, that we find stmking evidence of the honour accorded to the great grammanan in his own land? But this profusion of commentaries also bears witness to the difficulties of Panimi’s technique composed with analgebraic condensation, his work 1s a linguist s and not a language-teacher’s grammar, and for the more pedestrian purpose of teaching Sansknt rewriting was a practical necessity, thus giving nse to further grammatical hierarchies descending to a miscellany of school- grammars of recent date? It has been calculated that there are in existence over a thousand different Sanskrit works on Sansknt ? L Bloomfield Language, v 270 ff * In the Pradipoddyota of Nagoy Bhatta written some two thousand years after Panam s Asfadiyays, we have a sub commentary of no Jess than tife fourth degree Ramacandra’s Praknydkoumudt> Bhattoy Daksita s Siddhantakau- mrudi > Varadaraya s Laghukaumudt B230g, B2 PHONETICS IN ANCIENT INDIA grammar,! all inspired, directly or indirectly, by Panim’s model beside such a concourse the thousand manuscripts of Prscian’s Latin Grammar,? the pride of our western tradition, are but a drop m the grammatical ocean We have also to remember that Panini himself acknowledges a number of predecessors, whose work, except for fragmentary citations has been Jost to us—Burnell has \usted by name no less than sixty-eight of these pre Paninean grammanans,} well might the medieval philosopher Kuminla remark, We cannot think of any point of trme totally devoid of some work or other dealing with the grammatical rules treating of the different kinds of roots and affixes * But commendable as the cause may be, the non Sansknitist can hardly be expected to acquire the grammar of Sanskrit—for which the Indian tradition prescribes twelve years’ study—to the sole end that he may more fully appreciate the work of the ‘linguistic Homer’,’ or of later ‘grammatical saints’ ¢ In phonetics we all too rarely look back beyond the great names of the nineteenth century—Henry Sweet, A J Ellis, Alexander Melville Beli—except occasionally to honpur a few lonely and half- forgotten figures of the immediately preceding centuries? We justify some of our more grotesque and inadequate terminology {eg ‘fenuts’ and ‘media’)® by tracing it back to the Latin gram- marians sometimes as far as Dionysius Thrax or even Aristotle but generally speaking the expressions of anctent phonetic thought 1n the west have little to repay our attention or deserve our respect, whereas Indian sources as ancient and even more ancient are infinitely more rewarding And zn this field the linguist 1s fortu- nately in a more advantageous position to appreciate, the ancient achievement, in that the acquisition of a working knowledge of the Sanskrit sound system demands no very considerable labour, and in so far as there 1s a basis for general phonetic discussion which there 1s not for ‘general grammar’ ? * Belvalkar Systems of Sansknt Grammar p 1 * CE Sandys Hustory of Clasncal Scholarship 1 259 1 The Aindra School of Sanskrit Grammanans pp 32 . Paptravarttska trs! Ganganatha Jha p 306 5 P Thieme, Pamm and the Veda p 95 * Goldstiicker Panim p 52 (of Katyayana and Patanyali) 7 CE D Abercrombie Forgotten Phoneticians TPS'1948 pp 14 " See further 1 120 below ° CE Vendryts BSLxln 8fINTRODUCTION 3 o1 The Indian Influence on Western Phonetics Moreover the link between the ancient Indian and the modern Western schools of linguistics 1s considerably closer in phonetics than in grammar For whilst Painimean techniques are only just beginning to banish the incubus of Latin grammar, our phonetic categories and terminology owe more than 38 perhaps generally realized to the influence of the Sansknit phoneticians The impact of Sir Willram Jones's ‘discovery’ of Sanskrit on the course of Western linguistics 18 well known, but Jones, apart from his know- ledge of the Sanskrit language, was also acquainted with the tradi- tional statements of tts sound-system in his ‘Dissertation on the Orthography of Asiatic words in Roman Letters’ the whole order of treatment and descnptive technique ts clearly based on Indian models Ina paper on ‘The English School of Phonetics”? Pro fessor J R Firth has sard of this great onentalist, Wauthout the Indian grammanans and phoneticians whom he introduced and recommended to us it is difficult to amagine our nineteenth century school of phonetes ? The influence of thesIndian works on the phonetic views of William Dwight Whitney may be clearly seen in the discussions published in the Journal of the Amencan Oriental Soctety dunng the years 1862-6, subsequent upon the appearance of Lepsius’s Standard Alphabet 3 and we have the feeling that without their teaching Whitney might not have been ina position to express self- righteous indignation against that other country from which he had fearnt so much— Its really amazing how some of the most able physiologists and philo fogists of that nation (1¢ Germany) have blundered over the sumple and seemingly obsious distinction between an sandas anfandac apand ab etc* The ‘seemingly obvious’ distinction of voiced and voiceless here * TPS 1946 pp oz ® Teas remarkable that a German study of the English Schoo! should fail to make any reference whatever to the Indian influence (Hf Raudnitzky Die Bell Sreetsche Schule Fin Bettrag sur Geschichte der enghschen Phonetk piatburg ion) 2A Standard Alphabet for reduceng Ure-nitten Languages and Forngen Graphie Systems to a Uriform Orthography tn European Letters (London 1855) “7108 ex 313 Cf Trans ser Phat Assoc 1877 pp 41 ft4 PHONETICS IN ANCIENT INDIA referred to was subsequently recognized by Lepsius as ‘derived from the Sanskrit grammarians’ * In England the Indian influence 1s evident in the work of A J Elhs, especially m Part IV of his Early English Pronunciation, where frequent references ar¢ made to Whitney’s translations of the ancient works and also to his own observations on the speech of Jatter-day pandits He displays on occasion a fuller appreciation of the ancient statements than Whitney had done, and generously remarks on their descriptions of ‘voiced ’,” ‘The wonder 1s, not that they should be indistinct but that they should have been so much more distinct than the host of European grammanans and orthoepists who succeeded them As yet, hawever, the hnguzst cannot survey the Indian phonetic achievement without undertaking an extensive course of reading, of which only a certain proportion will be relevant to his purpose, and on the other hand, without having viewed the overall frame- work of the Indian analysis he can hardly assess individual desenip- tions The principal works have been translated and commented upon (so far as the phonetic climate of the translator’s environment permutted) by such Western scholars as Whitney, Weber, Regnier, and Max Muller, and more recently by Indian scholars amongst whom may be especially mentioned M D Shastn,S K Shastri, and M Ghosh An mteresting selection of special problems has , Deen discussed in detail by Siddheshwar Varma in his Critecal Studies in the Phonetic Observations of Indian Grammanans The ' present study aims at presenting a systematic account of Indian phonetic doctrine so far as 1t appears to possess more than purely Sanskritie interest Where Western antiquity provides any parti- cularly striking parallels or contrasts, some account of these has been given with a new to the comparative evaluation of the Indian statements Occasional discussions related to later Indo-Aryan developments have been inevitable m establishing a control for the pronunciations described im the treatises o2 The Sources Of the works themselves it will be sufficient to note that they fall info two main categories, the Pratisakhyas and the Siksas 1 FAOS vur 344 * See further 1 120 belowINTRODUCTION 5° The former are phonetic treatises relating to the pronunciation of the four Vedas, namely: , Re-Veda — Rk-Pratisakhya Sdama-Veda —— Rk-tantra-vyékarana Black Yayur-Veda — Tatttirtya-Pratuakhya White ” — Vaasaneyi- or Katyayanyya-Pratsakhya Atharva-Veda — — Atharva-Pratsakhya * The Siksis on the other hand are, with some exceptions, less specifically related to a particular Yeda, but in many cases supple- ment the teaching of the Pratrsakhyas ? Whilst it 1s likely that the Pratisakhyas are based on an early Siksa (such as that referred to in the Tattturya Aranyaka),? our extant texts of the latter appear to be of later date than the former the most important of them, the so-called Pammzya-Stksa, 1s sometimes claimed as the orginal Siksi and in consequence put back to a very early date but thts, as also its attribution to Pinim, ts highly doubtful + Varma places the Pratigakhyas in the period 500-150 Bc and the extinct Siksa literature between 800 aad soo BCS It 1s significant that one at least of the extant Siksds contains the admission, Jé Siked and Prtiakhya are found at variance, the Siksa 1s said to be the Jess authoritative, as the deer 1s weaker than the lion * Apart from these specifically phonetre works, numerous state- ments on phonetic matters are to be found in the grammatical works, more especially in Panmi’s Asfadhyayi and Pataiyaly’s * As Whitney himself has sdmutted, the text whieh he has edited and trans lated under thys title 1s probably not the AP, and so should strictly be known by the title which it bears viz Saunakiyd Caturadhyaytha Nevertheless, the AP, if such it be (ed S K Shastn, V B V Shastri), contams almost nothing of general interest, and for present purposes the ttle AP may be retained without disadvantage to refer to Whitney’s edition * Cf Kielhorm, Ind Antig v 141-4, 193-200 (esp p 199) 2 vit un 1, om dhsam wyakhyasyamak The subjects of the Sikei are given as varnah (‘sound unit ), sarah (tone ), matrd (quantity ) balam sama samtanah Sayana interprets balam as ‘degree of buccal closure (cf 11x below) sama as ‘tempo’ and samtanch as junction’ Cf Sten Konow, Acta Onentalia, xX 1v, 1943, P 295 * Cf Thieme, op cit, p 86n a * Op cit, Introd © Sarvasammata Siksd, ed Franke 49. dtksd ca pratisakhyam ca virudhy ete parasparam fiksenva durbalety ahuh smhasy ava mrgi yathaINTRODUCTION 7 refer to each other’s opinions in a commendably objective manner. Certain pronunciations, however, are generally recogmzed as faulty, and lists of such faults (eg m chap xv of the RP) are hardly less interesting than the details of the approved pro- funciation The Pratiéakhyas have recerved the attentions of vartous later commentators In so far as these are the bearers of a continuous tradition, they are able to augment and elucidate the lacomc brevity of the aphorisms unfortunately, however, the main stream of the tradition seems im many cases to have been lost, and the commentaries that we possess have a habst of wrapping the obvious in obscurity instead of casting hight on the numerous difficulties Moreover, 1t 1s clear that the intellectual climate of phonetic study had undergone a marked detertoration between the time when the treatises were composed and the time of our commentaries In general we may say that Henry Sweet takes over where the Indian treatises leave off—though in some matters even Sweet could have learnt from them and a recent study of a modern Indo-Aryan language has successfully shown that many of the anctent descrip- tive techniques can still pe employed to advantage ' These early phoneticrans speak in fact to the twentieth century rather than to the Middle Ages or even the mid-nineteenth century, and many a statement which the commentators and even Whitney or Max Muller have failed to comprehend makes immediate sense to the phonetician today The one outstanding exception to the general mediocrity of the Indian commentators 1s Uvata, whose mterpreta- tions of the RP and of the VP reveal an enlightened and enlighten- ing approach to a variety of phonetic topics 03. The Sanskrit Alphabet Whilst the statements with which we shail be concerned are of wide phonetic interest, even the most general of them are of course based on the description of a particular language, namely Sansknt Qup water pe tendie pura howag he se sya th sound-units as generally assumed by our treatises certain diver- gences from this system will be considered in their appropriate place As regards the transcnption, two conventions have been ' BN Prasad, A Phonetic and Phonological Study of Bhajpurs (Thesis sub- mitted for the Ph D Degree of the University of London, 1950)8 PHONETICS IN ANCIENT INDIA emptoyed for purposes of textual quotation, the standard Roman transliteration of the Devanagari text 13 used (in italic type)—this will not generally concern the non-Sanskritist, as the texts will be translated and the orginal Sanskrit, unless it calls for special com- ment, relegated to footnotes Where, however, Sanskrit sounds and sound-sequences are made a subyect of phonetic discussion, a transcription 1s used which departs m some respects from the standard system, and which I have found converent in the teach- ing of Sanskrit phonetics such transcriptions are printed in heavy IPA type Inthe chart on p 20 the two conventions are shown | side by side Where narrower transcriptions are required, these’ are indicated by the use of square brackets It should be stressed at this point that, except for transcrip- tional purposes, the representation of a complex structure by category-labels based on a monosystemrc analysts 1s an unaccept- able procedure, which has nevertheless been adopted by the many modern linguists who favour an exclusively ‘phonemic’ approach the reason for settrng up such a system in our chart 1s that the Indians themselves have done so It ts true that the Devanagari method of writing 3s syllabic, but the analysts underlying :t and actually set out for example, in the varna-samamnaya or ‘alphabet’ at the beginning of Panim’s grammar, comes very near to that which a modern ‘phonemucist’ would evolve for Sansknt by a substitutional-distributional analysis of the word-1solates * How- ever, we can hardly criticize our predecessors of some two millennia ago for a procedure which only a few linguists in the last two decades have begun to reject as inadequate? and we shall see that the Indians, unlike many of their Western successors, appreciated that this technique was a means to a lumuted practical end, and by no means the ultimate analysis . 04 The Principles of Description We come now to a consideration of the fundamental principles of analysis and description as postulated and as observed by the authors of our treatises * Cf «M B Emeneau ‘The Nasal Phonemes of Sanskmt , Len; guage, xx 86" ,A H Try ‘A Phonemic Interpretation of Visarga’, Language xv 194 ff ad Cf J R. Fitth Phonological features of some Indian languages , Proc 2nd Int Cong Phon Sc pp 176 ff ,W F Twaddell On defining the Phoneme’, (Language Monograph xvi) especially PP s4fINTRODUCTION 9 040 Word and Sentence In early Indian linguistic discussion we find a full awareness of the view that the bastc hngutstic unit, upon which all other analyses must be founded, 1s the sentence, a famous couplet of Bhartrhan’s treatise on general linguistics, the Vakyapadiya, where the matter 1s debated at some length states the case in the following terms Within the sound unit the component features have no independent existence nor the sound unts within the word nor have the words any Separate existence apart from the sentence! For purposes of phonetic description the baste unit 1s also some- times stated as the ‘breath group (eka prana bha@va),? correspond- ing in the Vedic hymns to one line of verse The tendency to deny independence to the word 1s further stressed by the Sanskrit system of writing, which (unhke, eg Old Persian) takes no particular account of word division? Thus word sequences such as taan eva, tat punak are written together as taaneva, tatpunah, the sequences -ne- and -tpu- being represented as single graphic units (4, @) This elrmmatron of the inter-word spaces as indica~ tions of yunctron ts partially compensated by the graphic representa- tion of such phonetic junction features as the available symbols are capable of showing thus the junction of tat-+bhavatt 1s written as tadbhavati, tat+-Jrutvaa as tacchrutvaa, tat+bi as taddhi, maa+-udakath as modakaih, and so forth Certain other junction features are not generally indicated sequences of the type -h- k- or -h-+-p- being only sporadically written as -xk- or -p,* since x and are outside the phonemic and hence the general graphic system Even rarer 1s the indication of the Imking¥ presenbed Hy the phonetimans mm sequences such as taa’abruvan (for taah--abruvan) which 1s generally written with hiatus as taa abruvan 5 Elsewhere junction features may be nerther writ- ten nor prescribed, so that no distinction 1s recognized between for example, na tena lkhito lekhah, ‘he did not wnte the letter’, *1 93 Cf Rosetti Le Mot p 20 9 TP y x CE Sweet Primer of Phonetics § 93 Rousselot Prinapes de phonét que expérimentale p 972 ( Le mot n existe sans altération qu 4 1 état asolé Le groupe respiratoire posstde une andavidualité propre) "Shumb Handbuch des Sanskrit § 160 2 Cf Whitney Skt Gr § 98 Bloch Linde Arye pp 75f * CE Wackernagel 41 Gr § 226 5 Cf Wackernagel op ct § 285 b (8)BT) PHONETICS IN ANCIENT INDIA and natena likhito lehhah, ‘bowing, he wrote the letter’, this fact provides material for the construction of various types of word play and nddle based on alternative divistons of the piece ' and our treatises do not mention any phonetic criteria by which a distinction might have been indicated in utterance? Whether in fact there were subtle distinctions of prominence such asare capable in Enghsh of differentiatmg for example, a notion from an ocean’ we cannot tell—it would certainly be unwise te deny the possibility on the bass of an argumentum ex silentio—but it appears certain that in Sanskrit a delimitation of the word by purely phonetic enteria was even less of a possibility than in Enghsh The Vedic texts in fact come down to us in two principal forms the Samiuta or ‘compound text with the sentence or breath group as its basic unit, and the Pada or word text having the word~ isolate as its basis, the latter 1s generally recognized to be an arty ficial analysis devised by grammanians and others for purposes of anstruction that of the Recedais generally attributed to the ancient grammanan Sakalya * In the AP we find, 4 The study of the word isolates 1s designed to teach the beginnings and. ends of words and their correct form tond and meaning * to which the commentator adds, ° Without studying the word isolates one might make errors in the con tinuous text it2s for this reason that the study of the isolates 18 necessary ° Some statements of the relationship of Pada to Samiuta, however, seem to have left room for musinterpretation the RP makes the hughly ambiguous observation, ‘samhita padaprakrtth’,? which according to the interpretation of ‘pada prakyteh’ (where prakytih ==‘basis’) might mean either “The Samiuta 1g the basis of the Pada’ or "The Samhita has the Pada as its basis’, the term prakzi: 13 also regularly used of the word isolate im contradistinc- tron to exkdra ( modification’, ‘varrant ) the latter bemg apphed to } 9 CF Afbh 2 4 2 Chidlhorn 2 24) * "Though so long as the tonal system survived this must in many cases have provided a means of distinction 3? Cf D Jones “The word asa phonetic entity ALF jt field Iranguage pp 113 f 182 p % want PP 69 Bloom “Cf Liebich op at pp off * w 197 padddhyayanam antéd febda sardrtha gndndrtham a sPodddhydyt samhitdm vindlay et tasmdd ebluh kdramaw avatyadhyeyardINTRGDUCTION es the yunction-forms A convincing solution to the difficulty 1s pro- vided by the Vasd:habharana, a commentary on the 7P, which yoints out that as a result of statements such as the above ‘certain slow-witted persons have made the mustahe of thinking that the Veda ts constituted of the word-1solates’, whereas m fact ‘the word- solates are only treated as a basis for the purpose of facilitating instruction *! ‘ Here also should be mentioned the Krama-patha or ‘repetitive’ ‘ext, in the simplest form of which a word sequence I 2 3 4 518 tecited in patrs as follows~-I 2 23 34 45, with the realization of the appropriate yunction-features between the members of each gar? This device forms an instructional stepping stone between the Pada and Samhita texts , 1t appears to be held in no very high esteem, and the most that the RP can say for xt 18, The Krama ts of no use to one who knows both the Pada and the Samhita It does nexther good nor 1] and has no sacred tradition ? The recognized function of the Pratsakhyas appears to have been instruction firstly 1n the pronunciation of the word-1solates and secondly in the modesof their synthesis im the sentence The first ofsthese duties, however, mvolves the teacher in further analysis, below the word isolate level and since all analysis must be followed. by synthesis,t the 7'P aptly observes that there are various types of synthesis'—-of words, of syllables, and of sound- umts, to which 1s added as a fourth category, if we follow the com- mentator's mterpretation,® the remtegration of syllable structure 7 Whitney, failing to understand this passage, can only remark that ‘these four rules have no significance whatever, being a mere bit 'On uw 1 wbhakta rupasya tu prakptevam wyutpadana saukaryartham adfrryate 7 Cf RP x 44 > x1 66 kramena narthah pada samhitd wdah na codajapdya karo na ca $rutah * Cf Sweet op cit, §91 Sxsiv x 4 athacatasrah samhtah pada sam! staksara sam} ita varna sami | tafga samluta cet: nana pada samdhana samyogah pada samplutety abhidhiy ate pathd scam akjara samlutadindm apy evam © ‘This interpretation of aviga-samhzta rs based on xxt 1 (vyanjanam svarargam) cf also 3 20 below 7 For the fourth category ef especially Sweet Joc cit ‘Synthesis lastly, deals with the organic and acoustic grouping of sounds into syllables etc, and the divisions between these groups”12 PHONETICS IN ANCIENT INDIA of outside classification, in which someone has amused himself by indulging’ In one passage of the RP Uvata notes that strictly spealang the author 1s exceeding his duties by giving rules for tonal synthesis within compounds, the basis of synthesis, he claims, should be whole words as mnstitutionalized! units he 1s prepared to be indulgent, however, for just as a flower picker may also pick fruit and a wood gatherer may also gather honey such 1s the case.? 041 Phonetics and Phonology There 1s, on the other hand one respect in which our treatises do not fulfit the functions which they claim The first verse of the AP declares, Our subjects are the (phonetre) attributes in yunetion and im tsolation of the four word classes viz noun verb preposition and particle? This grammatical enumeration suggests that we may expect some thing like a phonological treatment, with grammar and phonetics integrated into a functronal whole, and Uvata, commenting on the VP, goes so far as to claim, ‘ ‘This treatise excels all other treatises an that st combines the two dis caplines of phonetics and grammar + ‘This however, 1s n the nature of a pious aspiration which regret tably does not see fulfilment for our text makes scant reference to any grammatical function 5 Particularly remarkable 1s the failure of the phonetictans to discuss one of the outstanding phonological processes of Sansknt, that of ‘vowel gradation’ for certain phono logical purposes it 1s convenient to recogmize a system of vocalic alternation of the type Grader i ut t » 2 @& o . ar » 3 ai au aar ' sddha lit established 2 On 26 nant paddnam laksanam ma kartavyam siddhesu In padesu samhtta prakpt h satyam eva yatha puspaharasya Phafakaranam darv aharosya madhe*aharanam evam etat * caturnam pada jatanam namdkhyatopasarga mpatanam sandhya padyau Bunau pratjnam *On1 169 * CE Litders Vyasa Siksd p 102 Goldstucker Pamm pp 195 ffINTRODUCTION 13 The working of this alternation 1s seen in verbal forms such as: (Jstu-, ‘praise’) (Jkr-, ‘make’) PP stuta PP. heta- Inf stétum Inf kartum Pr Ind stauti Pf Ind cahaara Though tgnored by the phoneticians, this alternation 1s duly noted by Panini (in bus opening aphonstms) and his followers, who treat Grade 1 (corresponding to the Indo-European ‘reduced’ grade) as basic, giving to Grade 2 (= IE ‘normal’ grade, or ‘Vollstufe’) the title of guna or ‘secondary quality’, and to Grade 3 the utle of vrddht or ‘increase’? A further phonological process which 1s similarly disregarded 1s that of samprasdrana (lit ‘extension’),? whereby a sequence of the type wa, te v-tsyllabicity, alternates with u,1e ‘syllabic v’ (cf Pr Ind svapiti: PP supta-, &c) Panin: uses the term both for the process and for the resultant vowel,? but we find neither the term nor any discussion of the process in the phonetic works Nor again do we find there any use of that great creation of Panini’s genius, the phonological zero + This mention of phonological omissions, however, 1s not to be taken as in any way detracting from the value of the treatises from a purely phonetic point of view 042 Termmnology : Before proceeding to the textual material some account must be given of certain terminological features which run through the whole system of desenption A particular problem 1s presented by the word varna, which can * Pan 111 ff vyddhirad ac ad en gunah tke gunatrddk ~The term and Process of gute are in fact first referred to by the carly etymologut Yaske (Nur x 17, deriving feva from fyyate) Only passing references are found in the phonetic works (guna in RP xt 10 erdditim VP v 29 AP (ed S K Shastri) a 113 43) CE Edgerton Skt Hist ‘Phonology (AOS Supp 5, 1946), §§ 118 fF ? The relevance of the term 1s not cleat The form praserana 1s also found, and 1s used by the 4Pfed Shastri) ur 1 11 to refer to the replacement of -bh- by-p-mdlpsati Cf Twaddell op cit p 54 Merggs ‘in Paychologte du langage p 192, E Haugen First Grammatical Treatise (Supp to Lang 26 4) p 8 2 Cf Weber Ind St w 109 * Sarvasammata Sika 36 guts Sighs: hrah kamp: tatha lkh ta pdthakah enartha jno "pa kant} af ca gad ete pathakadhamah * Gf Konow Joe at ‘ PS 5 anusedro unargas ca sha spau caps pardtrayau 722 " ayogavaha vyneya dtraya sthana bhagenah 8 Mbh 11 2 on Pan SSu 3 (Kiethorn 1 28) ke punar ayogavahah wsar jamya phvamubyopadhmamydnusedrénundukya yamah Ratha punar ayogaINTRODUCTION Ww But ‘drawing unyohed" seems hardly to be a natural metaphor for ‘heard (though) excluded"! Uvata, in his commentary on the I’P, has an interesting alternative; he takes the inital a- of the term not as prvative but as referring to the fetter a and standing for the alphabet as a whole his explanation then reads, They are called a yore tdha because they draw. attain their realiza- tron, only when joired with a, &c.,2¢ with the letters of the alphabet * In view of this explanation, Weber reads simply sugat dha in both text and commentary ,! but appropriate as the term would be as applying to the contextually bound nature of the elements in ques- thon,* iss to be noted that the RT specifically distinguishes ayoga- vaha (= the contextually dependent elements) from yogardha (= the other letters) § An important terminological distinction underlying a large num- ber of the ancient desemptions 1s that of sthana and karana (lit ‘place’ and ‘organ’), which, generally speaking, denote the passive and active organs of articulation as the commentary to the AP explains, ‘The sthina 1s that whichis approached the karata that which ap proaches * The terms closely correspond to what Pike calls Point of \rticula wAASA yad ayubtd cohanty anuped sitfica fruyante There follows a discussion as to which pratydthdra the eyogordthar could be included under, end ng with the tuggetpon that they may belong under none (athardrifesenopadetah kar tavych) and haryyata accordingly assumes the title to refer to this lack of » Pratythdra (ayukidh pratydhsralaksanena) The explansnon of the PafyiAd (on PS foc. ct } 1s eater but hardly more acceptable (xa tudype ¢ porch somogo arpdutarena yepdm) Behe] ngk » inter pretatzon (Palm (t840) at 413) us fanciful—Trennang herverbnngend dhe Vovale von derComsonanten scheidend zwischen Beiden in der SI tte stehend Woackernsgel attributes the strangeness of the term tots having been orginally co ned for the teaching of children (4: Gr i p kext n 7 Dies und der hurnorutusche Charakter mancher Termini wie eyogantha weasen auf Herkunft aua dem Jugendunterncht} 3 Oni 28 chard nf carna-ramdmndyena sahitth santa ete tahanty dima- Aibham prdpmsvanty oyogavdhdh For the use of aktndd @ &c as = abe ef Sayapaon Tout de vitin Svar 23ff Cf on PS loc. cit Und St av 354) and on Pratyid Sutra 22 4 The greater appropriateness of this term was evidently felt by Canarese gtammanans who adopted it inatead of ayogatdha (1c PS S Sastn Lectures on Patanyali"s Vahdshina 1 143n) The Amareis S (SSp 121 $§ 50-51) has temyogariha Introd, ed Shatn p 2 *Qn1 19 25 jad upakramate fat sthdnam yenopakramya ¢ tat haronam Br c8 PHONETICS IN ANCIENT INDIA tion’ and ‘Articulator’! In a large majonty of cases the articulator 1s an area of the tongue, viz ‘root of the tongue’ (sehva-mula), ‘middle of the tongue’ (jzhvd madhya) and ‘tp of the tongue” (jthvagra), whilst the opposing points of articulation are ‘root of the yaw’ (hanu mula), 1e soft palate, ‘palate’ (talu) and ‘teeth? (danta) or teeth-roots’ (danta mula) The same classification 1s extended to the lips, so that in the articulation of the bilabials the AP? and the Tribkasyaratna? prescribe the lower hip as karana and the upper lip as sthana, and the AP goes so far as to apply it, sore what artificially, to infra buccal articulation in the case of the glottal sounds, for which the ‘lower part of the glottis’ 1s con- sidered as the karana+ The specification of minor distinctions of stharta in the alveolar area 1s sometimes not as clear as we could wish butin the absence, so far as we know, of palatographic atds § this 1s perhaps hardly surpnsing An attempt to apply the system to the feature of nasality can only lead to confusion the nose in such cases being stated by some treatises to be the articulator and by others the point of articulation Other terminological items of less vyde application will be dis cussed under their appropriate headings The reader isalso referred to the excellent glossary now available in vol in of Renou’s Ter munologie grammaticale du Sanskrit 043 Order of Analysis The treatment here adopted closely follows that of the Indian analytical procedure, which recognizes three mam stages : Analysis of the basic atticulatory ‘processes’ n Segmental analysis of the speech stream { letters’) un Synthesis ( prosodic features ) 7 * 1 Phonetics pp 120 ff 71 25 osthyanam adharaustham (sc haranam) On TP u 39° atrattarostha sthanam uttaratua samyad esam sthananam adharogthah karanam 44019 hatithyanam adhara kanthah * ‘The first recorded instance appears to be that of an Englishman J Oakley Coles who in 1871 in the cause of phonetic accuracy painted the roof of hus mast cath. a maxture of feat am momlape (al Roosselot Prmeper p 335 See also K C Chatters Techmeal Terms and Techque of Sanskrit Grammar pt 1(Calcutta 1948) 7 Cf J R Firth Sounds and Prosodies TPS 1948 pp 127f EJ A Henderson Prosodses m Stamese A Study mn Synthess Ana Mayor 1 u 8920 PHONETICS IN ANCIENT INDIA THE SANSKRIT ALPHABET (varna-samaraniya) Consonants ‘Unaspirated Glottal | Velar | Palatal |Retvoftes| Dental | Labsat | ¢ Aspirated Unaspirated Stops Voiced | Votceless + Aspirated Nasals! Semivowels Vorceless Fricatrves Voiced. Vowrs \| Short * Long . Diphthongs 2 Also ‘anust Gra’— mm ty ‘amunagha’—ih ~ Note ‘The order of letters as presented in Panunt’s Siva Sutra shows con- siderable divergences from the above this fact however as explicable by the Phonological as opposed to phonetic, approach there adopted Cf Thieme, Op cit, p ro¢ ‘The atrangement of Paminis list of sounds which at first looks rather disorderly, 18 explainable as due to the phonetic catalogue of sounds having been adapted to the practical requirements of the grammar, 12 which Paquni wanted to refer to certan groups of sounds by short expressions ”PARTI PROCESSES 10 Mental Tue Indian phoneticians spend but little time in discussing the mental or neural bases of speech The introductory stanzas of the PS are representative The soul, apprehending things with the intellect, inspires the mind with a desirc’to speak, the mind then excites the bodily fire, which an 1ts turn sumpels the breath The breath, circulating in the lungs, 1s forced upwards ant, umpinging upon the head, reaches the speech-organs and Bives rise to speech-sounds These are classified in five ways—by tone, by length, by place of articulation, by process of articulation and by scoondary features Thus the phoneticrans have spoken take careful eed * ‘The ‘secondary features’ here referred to (anupradana) are inters preted by the Siksd-Prakasa as ‘anundsikads’, ‘nasality, etc’ (see further 1 10 below) The musical treatises contain similar statements, though these are less closely related to the actusl speech organism The relevant passage of the Sampitaratnakara reads as follows The soul, desirous of expression, instigates the mind, the mind then excites the bodily fire, which in its tur umpels the breath This then Moves gradually upwards and produces sound im the navel, the heart the throat, the head and the mouth * Parallels to such statements are not far to seek in the west, notably in the doctrine of the Storcs Zeno 1s quoted as defining speech in terms of a stream of gir extending from the principal part of the soul to the throat and the tongue and the appropnate organs,? ‘ whilst Anstotle deseribed it as the striking against the so-called ‘artery’ (1 e trachea) of the air exhaled by the soul # § g-10 lezdmt widhdgah paicadha smytah svaratah kdlatah sthdndt prayatndnupraddnatah whvarga ridah préhur mpuzam tan nibodhata tras off ? Plutarch, De Place Phil w 21 go3c (rrevpa Sierarev and rod Hyeponnod péxpe dapuyyos xal phere wal rw olcalaw dpyarcur) . * De Arima uv 420° (} why} 708 aratreopdvon adpos wro Ty ey TovTOIS TONE poplors Puxys upos ray xalouzdry dprmptar dar} done)22 PHONETICS IN ANCIENT INDIA 1.1, Physiological A general statement such as that of the PS also introduces, 1n a rather haphazard manner, some mention of the basic articulatory processes (prayatna) which are more systematically presented by other treatises 1 10. Classification These processes are divided into two mam types, 4bhyantara, ‘internal’, and bahya, ‘external’ The first type comprises processes occurring within the buccal cavity (‘intra-buccal’) and the second those occurring elsewhere (‘extra-buccal’) For the first type both the VP! and Panin:? also use the term dsya-prayatna, ‘mouth- process’ this 1s interpreted by Patafyali as referring to the area from the lips to the ‘kakalaka’,? the latter being further identified by Kanyyata as the thyroid cartilage or ‘Adam’s Apple’* The Indran classification of the processes may be summarized as fol- lows 1 Intra-buccal processes (abhyantara-prayatna) (2) Closure —associated with the class of stops {5} Opemng— ” » » vowels (c) Constriction, of two degrees, associated with (1) the class of fricatrves,$ (a1) » oo» ©Semivowels 2 Extra-buccal processes (bahya-prayatna) (a) Glottal +—associated with voice and non-vouce (breath) © (8) Pulmonc— ,,—,,_—_ aspiration and non-aspiration 7 (ec) Nasal — ,, ” nasality and non-nasality Notall our statements adhere tgidly to this descriptrve framework, but 1t may be taken as a generalization of the varrous systems, and as set out in precisely the above terms by Pataiiyali® and by the Aptiah-Stksd? Departures from this system arise when, by the inversion already mentioned, the basic processes are considered as ™ 1 43 (glossed by Uvata as mukha prayatna) trig 3 Mbh t 1 4,0nPdn loc cit (Kielhorn 1 Gr) osthat prabhyt: prak kakalakdt * Qn Mbh, loc cit grivayam unnata pradedah > Cf Trubetzkoy’s “Annaherungskorrelation’ (loc ct ) © CE Trubetskoy s ‘Stummbeterigungskorrelation’ i Cf Trubetzkoy's ‘Exspirationsattkorrelatson’ : Mbh 11 4,0n Pan 1 1 9-10, Kaclhorn,1 61 ut iia tiPROCFSSES 23 distinctive features serving to differentiate one fetter from another A passage from the TP may be quoted in this connexion’ The distinction of letters 1s effected by secondary features, by combine- tion, by place of articulation, by the positon of the articulator, and by length! The meaning of some of these terms 1s made clearer by Uvata, who quotes this passage in his commentary on the RP,? as an instance of ‘secondary feature’ (anupradana) he mentions the vorce-process (2a3n the rbove summary), 28 examples of ‘combination’ (samsarga) he gives aspiration and nasality (2bc), and he interprets ‘position of the articulator’ (karana tinyaya) as referring to the intra-buccal processes of closure, opening and constriction (rade), which he exemplifies in the statement, Between letters having the same place of articulation and secondary features, eg 1,56 J, acoustic distinction 18 effected by the artaculator?® But it will be noted that in the TPs statement of ‘distinctive features’ these processes are treated on the same terms as the places of articulation and a prosodic feature such as length, to which other waters also add tone * This, however, 13 by no means only an ancient Indian failing ‘Twaddell, for instance, in his mono- graph On Defining the Phoneme, 1s prepared to admit as parallel “component terms of articulatory differences’ such vanous features as places of articulation, duration, and the processes of voice, aspration, closure, and constriction * A hist of five resonators (prdtefrutka) 1s also given by the TP,* "xa 2 anuprodindt somiorgdt ethdadt kovana vom aydt dy ate varna-ranieryan parimdindc ca parcamdt *am 13) Uvata refers to divtunctrve features by the term guna quality, which 11 also used by the fp S (iv 7)and the Wbh (hielhom 1 61) to refer pameularly to nasality Both Usata (on RP iu 2, Pos 130) and the Tribhdyyuratna(on TPs 1) also use in connenon with vowels the term dharma, Property, the vowel ateelf then being referred to as dhavmun possessor of a PRATT 1 tulya sthindevaradininim am UkSra-sakdra yakdriain harasa-kytah fruts orf * For tone vee Urataon MP an ayo cf Siddhdnta-Konensdi 12 Ite regected by PataAys (VBA 1 ag on Pda t 2 9 hucthom,1 62}0n the grounds of its Pond atinttivenessmobiedshs wddtisfayok pas “4: y tarya prdiruthins Bhoventy wrah Rawphah fro wubhon minke tte CTA pratient pratidioadh)24 PHONETICS IN ANCIENT INDIA of which the buccal, pharyngal, and nasal may be justified.’ the further mention of ‘chest’ and ‘head’ as resonators, however, 13 probably taken over from the subjective terminology used in India, as in the west, for the description of the various voice- registers 7 + ‘We will now exammme in detail, and an the order set out above, the statements on the individual processes rir. Intra-buccal Four degrees of closure between sthdna and karana are recog- mized Maximal closure 1s referred to as sprsta, ‘touching’, and minimal closure as evyta, ‘opened’ ri10 Vowels and Consonants The process of minimal closure, or ‘non-contact’ (aspysta) pro- vides the phonetic cmterton for the distinction of vowels (svara) from consonants (vyafzana) the TP expresses this in the following terms For the vowels the ‘place of articulatior® signifies the place to which @pproximation 15 made, and the ‘articulator’ refers to the organ which effects the approxumation For the rest the ‘place of articulation’ refers to the place where contact 1s made, and the ‘articulator’ refers to the organ which effects the contact * Maximal closure, on the other hand, provides the eriterion for the category of stops (sparsa) 5 ‘Thus far no problems anse But the intermediate degrees of * CE Joos Acoustic Phonetics, pp 58£, 96, Forchhammer Theorse und Techmk des Singens und Sprechens, pp 271 4f (Rachenzesonanz, Nasenresonanz, Mundresonanz) * See TPxxu 10, PS 36-37, VP1 10,30 Pike (op ct, pp 17 #f) gives a short eritique of the “mutation label technique’ used in singing classes one of the instructions quoted, viz to ‘place the tone between the eyes! finds a close parallel in the $hru madhya' of the relevant passage ofthe VP Cf Forchham- mer op eit, p 276 (¢ mut dern Begruf der Kopfresonanz veriassen wir das Gebset der akustisch-phystologischen Erschemungen und treten auf das Gebuet der Kérperempfindungen uber”), p 285 Die Brustresonanz muB wohl demnach, Senau wie die Kopfresonanz, in die Reshe der gesangtechnuschen Verurrungen verwiesen werden’) > PS 38 * 31-34 0 svardndm yatropasamhdras tat sthdnam yad upasamharatr tat karanam ony esdim tu yatra sparianam tat sthdnam yena spartayats tat karanam eg AP1 29 sprsfam sparlandm karanam.PROCESSES * 25 constriction are designated by various terms The Ap S' refers to the four intra-buccal processes as () Contact (sprsta) (u) Slight contact (isat-spysta) (ut) Slight openness (Isad vetyta) (1v) Openness (ciezia), a a classification which 1s reminiscent of our modern terminology for descnbing degrees of vowel closure The PS employs a rather different set of terms ? (1) Contact {11) Shght contact (u1) Half contact (nema sprsfa)! (av) Non contact The statement of the AP provides some difficulty of interpretation Like the Ap $ 1t mentions (1) contact, (11) shght contact, and (1v) openness, under (11) however, we find the words ‘and openness’ (etcyiam ca) + Patafiyall, who quotes this statement, ts probably right in saying that we must here understand ‘slight’ (Isat) from the preceding rule, thus bringing the statement mto line with that of the 4p $5 The AP commentator, however, suggests that the whole term isat sprsfa 1s to be understood ® so that (11) would then read ‘shght contact and openness’—a desenption which 1s more to the point than it might at first appear (see further 1 111 below) To (u) the RP gives the further title of duh sprsfa, ‘imperfect contact’? ~ Processes (11) and (11) Inke (1) and (1v) provide classificatory entersa, (11) for semvowels and (111) for frcatives. The application * ger 238 aco sprstd yanas ty yan nema spritdh dalak smytah Sesah sprsta halah proktd mbedhdmupraddinatak 3 Also ardha spritain Yajnavatkya S 209 f Varraratna prad paka S 39 4 1,29-32 sprstam spargandm karanam fat spystamt antahsthandm usmanam tnurtam ca svardnam ca * ffoh tr: 4 on Pant: ro(kselhorn 1 64) tieytam usmanam igad tty eva anucartate The VS s description of (iu) as open (294 Luders p 92) suggests a failure to observe this anweytti © Oni 31 7 xm 1026 PHONETICS IN ANCIENT INDIA of this descriptive framework may be exemplified by the palatal series as follows @) Contact —¢c (1) Half contact —y (m) Half openness—f [¢] (iv) Openness —i 1iil Frcatives One of our treatises gives a more detailed account of the articula- tron of the fricatives by process (in) The 7P, having remarked that ‘the fnicatives are articulated in the same places as the corre- sponding stops’,! goes on ta say, ‘But the centre of the articulator 1s open’,? a statement which lends some support to the view of the AP commentator quoted above Whitney, commenting on this doctrine, makes the enticism that, ‘This prescription of an unclosure of the muddle of the organ 13 rather an artificial device for saving the credit of the general prescription of actual contact in all the consonants Palatograms showing the articulation of the fricatives by modern Indian speakers would tend to support the 7'P’s observation as against Whitney’s uninformed scepticism 3 In the case of the retro- flex fricative the AP gives a rather more graphic descnption by referring to the tongue as ‘trough shaped"+(cf Grammont, on 8, Z, ‘ ta langue se dispose en forme de gouttiére et forme un canal trés dtroit sy The general term for the fricatives 1s usman, literally ‘hot, steam- ing’, perhaps because of their resemblance to the hiss of escaping steam itis glossed by Uvata as vayu, ‘wind § The terms applied not only to the letters [ gs but also to =) -x -h and hy and to the ‘a: 44 sparia sthanepiymana anupurvyena 31 45 karana madhyam tu vivptam bm Few Bhoypun vol n Patatograms Nos 2 63 69 (82 ph, a fa, ga 4323 sakdrasya dromka 5 Trawté de phonétique? p 69 (see also p 70 figs 81-82) Cf Pike op cit P rzt( = grooved asforasiblant *") Sievers Gr d Phonetsh* § 314( Nicht munder wichng ist aber wie es scheint da8 bei shrer Bildung die Zunge an thre Mutellime zu emer echmalen mehr oder wemger tefen Rinne emgekerbt wird '} * On RP: to See further K C Chatter: op cit PP 207 fF 7 Panini(Siea Su 5) appears to classify h also asa semuvowel but as the Mik pouts aut{t 4 2 on Pan loc et Kuethorn 1 27) thos 1s only for convenience in stating certain phonological rulesPROCESSES 27 breathy release of the aspirated stops (sosman). There is no special term corresponding to ‘sibilant’, though excessive sibilation is referred to by the RP as hsvédanam, ‘whistling’? 1.112. Semivowels As regards process (i1), with which 1s associated the class of semi- vowels, the validity of the analysis is not entirely beyond question, involving as 1t does the postulation of a greater degree of contact for this class than for the fricatives. In the case of the lateral ] and the rolled r the classification might be justified; but the case for y and v [w]? 1s less clear. We should expect the criteria for setting up a category of semivowels to be phonological,‘and related to the fact that they do not function‘as sonants* in the structure of the syllable; from the phonetic point of view y and w might be de- scribed with the close vowels i and u,® and we may suspect that in erecting a separate phonetic category for them the Indians have been misled by their system of letters. As Pike points out, Syllabic contextual function 1s reflected in phonetic alphabets. Sounds which are described by the same procedure but which are used differently in phonemic systems as syllAbics 2n contrast to non-syllabics are given different symbols, and at times are given names such as ‘semavowel’ and the like ® The apparent failure of the Indians to recognize the phonological {as opposed to phonetic) basis of this category of letters has the result that whereas i 1s regularly classed as ‘open’ or ‘lacking con- tact’, the corresponding semivowel y is described as having ‘slight contact’. Regarding the nature of this contact a more specific state- ment is found inthe TP: For y conf&ct is made on the palate by the edges of the muddle of the tongue.” The accuracy of this particular statement would in fact be sup- * xiv, 20 (Uvata adhiko vartasya sariipe divamh) A further fault in their Pronunciation is given the name of fomafya, lat. ‘shagginess’ , 1t 16 interesting that the same metaphor 13 used in the general Greek term for the aspirates, viz Baovs 7 On the alternative labio-dental articulation see 2 05 below. + Sée further observations on p 67,n 2 below. * Cf ¥ R Firth, ‘The Semanties of Linguistic Science’, Lingua, 1 4, 1948, P 402; Bloomfield, Language, pp tozn, 121 3 Cf Pike, op. cat, p 143 * Op. cit, p. 76, ¢f Trager, Language, xvi 220 Tu 40 tdlau jhud-madhyantdbhyam 4 akare.28 PHONETICS IN ANCIENT INDIA ported by palatographic evidence'—it 1s the description of the ‘vowels that 1s really at fault, no distinction being generally made between open and close qualities ,? 1t should be mentioned, how- ever, that the 7'P, in discussing the e-vowel prescmbes some degree of contact, an observation which could again be supported * it 1s only strange that 1t should make no such statement with regard to i, where the contact 1s considerably greater+ The TP also mentions ‘approximation of the lips’ for the articulation of the lip-rounded vowel u$ The tradition of the 7'P 1s followed by the VS, which refers to ip protrusion in the case of u$ and, discussing the process of openness generally associated with the class of vowels, points out that this does not apply in the case of i and uy” whulst the next rule goes on to mention actual contact * It 1s pre- sumably to isolated statements such as these that the AP 1s refer- ring when it gives as the opinion of some sources that contact 1s mvolved in the vowels an oprmion which Whitney impatiently dismisses as‘ too obviously and grossly incorrect, one would think, to be worth quoting’ Against the foregoing criticisms of the Indian analysis it may be argued that im certain contexts, more especially as initials, y and V were more tensely articulated than elsewhere and involved greater contact than in the case of 1 and u, for this we have the specific statements of a number of the Siksis, some of which even pre scribe for y a pronunciation as y in such cases''—an observation which 1s significant with regard to later developments '* Only on ' CE Prasad op cit Palatogram No 89 (ma ya) 2 This shortcoming provides the Latin phoneticians with one of ther few triumphs (more especially Terentianus Maurus cf Sturtevant Pronuncanon of Pre and Jeatint §§ rr1 ff) ‘rasad op cit Palatograms Nos 100 (al + Tbd Nos 96 97d cima my © (abe beba Dy Sun 24° osthopasamhdra uvarne 6 284 uvarga prakyter osthau drghau 7 294 (Linders P 94) * 295 (Lilders p 92 n 2) 1 32-33 suaranam case witytam) eke sbrstam 10 Quoted in detail by Varna op cit 26 Ww “eg Yay aalkya 2 yo PP See futher 3 x0 belo dadads ca padadau ea sam ogavagraheru ca Jah fabda tt uys eyo yo nyah a Cf Praty & Sutra 9-13 #090 mah saya ah smrtah 1 have observed this pronunciation of ¥- in recitation of the Suki di . lay ajureeda {Vayaraney: Madhyandina) as also a pecul ar tense stop realization of ¥ in certain cases (cf the atssamsprstaprayatna of the gloss on kay S 158) eg Skt sata>Pkt java (>Hind: Jaw) Sc Cf also the developmentPROCESSES 29 such grounds could the doctrine of a special degree of closure for the sermvowels be sustified ," the earlier treatises, however, quote no such evidence in their defence The Sanskrit term for the category of semivowels 1s auta(h)stha, lit ‘standing between’ It 1s tempting, and has tempted modern commentators, to interpret this term as referring to the postulated ‘intermediate’ degree of contact discussed above,? or, Irke our term ‘semivowel’, to their phonological alternation 3 The ending -stha, ‘standing’, however, 1s more readily applicable to the place which these Ietters occupy in the alphabet, viz between the stops and the fncatives ,* and it 1s doubtful whether the ancient sources provide evidence for any other interpretation A comparison with the ancrent western classification 1s here of some interest It will first be necessary, however, to mention that the Indians do not set up their vowel consonant distinction on exclusively phonetic grounds it has also a phonological basts 1n the structure of the syllable (see further 3 20 below), from this point of view the vowel 1s defined by its ability to function as a sonant or syllabic nucleus’—as the RP observes A vowel with a consonant ‘or even by itself forms a syllable § and it 1s significant that Patafijal etymologizes the word svara (‘vowel ) as <*svayam rajate = ‘is autonomous 7 Tn Greece also both types of criteria were employed Plato men- tions the classes of dwvnevra (lit ‘having voice ) and ddwva {lit ‘lacking voice’),® these categories, exemplified by Greek vowels and consonants respectively, appear to be set up on a phonological basis, and might be rendered by ‘sonant’ and non sonant’ Anstotle goes on to relate this phonological distinction to the phonetic criftria of ‘non contact’ (avev mpooPoArjs , cf Skt aspysta) and ‘contact’ (zeta mpooPodijs, cf Skt spysfa) Plato further men- w>b in Skt tana>Pkt vana>Hindiben &c See further S K Chatter ‘Ongin and Development of the Bengal: Language 1 § 133 1 Cf alse palatagam.afy iin Ee your) hewde thatafig.en by Gammont op cit p 97 fig 95 2 eg Whitney on AP: 30 3C£ Renouw Gr Sanser § 5 4 CE Uvataon RP1 9 sparfosmanam antarmadhye tisthantaty antahst] dh 5 Cf Pike op cit pp 66f * xvin 32 satj3anjanah sanusvdrah fuddho vapi svara ksaram 7 yu x on Pan 1 1 29-30 (Kielhorn 1 206) svayant rajante svara amag bhavatityananam The word 1s in fact to be related tothe root:vy sound ® Cf also Euripides Frog 578 (Palamedes 2) 1f30 PHONETICS IN ANCIENT INDIA tions a sub-category of consonants which have ‘noise (#dégos)' but no voice’ or ‘mo voice but some sound (¢8dyyos)’, and which he elsewhere calls ‘intermediate’ (uéoa) Aristotle refers to this class as jpipwra, ‘half-sonant’, and proceeds to define them by a com- bination of phonological and phonetic criteria, the ‘sonants’, he says, are ‘wathout contact and independently pronounceable’, the ‘non-sonants’ are ‘with contact and not independently pronounce- able’, whilst the ‘half-sonants’ are ‘with contact and independently pronounceable’? ‘The only actual example of these ‘half-sonants’ gtven by Plato 1s—rather surprisingly—s to this Aristotle adds r, and a full list 1s given by Dionystus Thrax, followed by Dionysius of Halicarnassus, viz S, %, 1,4, m,n? Thus the Greek ‘half- sonants’ turn out to be the fricatives, the liquids and the nasals in the absence of 1 or r vowels and (in Attic-Ionic) of y or w glides, the question of a phonological category of ‘semivowels’ (the usual translation of sjufduva}! does not arise We are here in fact dealing not with semivowels but with ‘continuants’ of various types, some of which may have quasi-syllabic function outside the Greek phonological system—as Dionystus Thrax expresses 1t, ‘They are called ‘half sonant’ in that, when used in murmurings and hissings, they are only less sonorous (e¢wva) than the ‘sonants’ * And it 1s noteworthy that the Latin grammarians generally include amongst their 'semrvocales’ the Latin fricative f,5 but not the semt- vowels y and w (j, 0}? Some Greek sources seem also to have classified as ‘half-sonant’ the h-element of the voiceless aspirates ph, th, kh (4, 8, x) ® the inclusion of the aspirate h- 1s reyected by * Cf the German use of the term ‘Gerauschlaut (eg Dieth Vademekum der Phonetik §§ 200 £1) see also Bloomfield Language p 95 1 The relevant passages from Plato and Aristotle are Platu Crat 424¢, Pil 18af , Theact 2030 Anstotle Poet xx 1.456%, 14554, Hist An 1v 9 5354 * Ed Uhlig pp wf = Bekker, p 631 jplpuva per owrw { (= 08 or 80) $(= #0) 6(= 70) Ae yp @ CED Hal, De Comp 72 78 4 Cf Marouzeau Lexque de la termmologie lineustique p 192 # Loe eit aplguva 8 Myeras En mepoco Frrov ray garnevrav ehfwra xadeorquey dy te Tous poypous mal otypots * Seee g Donatus, Keil, 1v 367, Prseian expressly disagrees with thes miclur sion (Keil, 1 9 22) 7 Ibid. 13 © CE Sextus Empincus Adv Gramm (Mazh 1) 102. Its also to be noted that in the list of Diogenes Babylonius the aspirates are nat included emongst the stops(cf Diog Laert vai §7) Attempts have been made to explain this classifica- tion by assuming an affricate or fricative realization (cf Sturtevant, op cit ,§ 90a,PROCESSES ” Pnscian,! but appears again in the Old Icelandic grammatical treatises, which also include the Icelandic dental fricatives 2 The Greco-Roman tradition of the ‘semvocahs’ still finds expres- ston in the work of Grammont Les sem: voyelles sont encore éminemment des spirantes et aussi: bien des fncatayes ct des constnetves > There 2s in fact little common ground between the Indian ap- proach to the antahsthd and the Greek approach to the fpiganov The only mention in our Indian sources of a contrast between instantaneous and continuous articulation 1s that of the RP For the stops there 1s momentary contact for the vowels and fricatives there 13 continuous non-contact + and the only Western statement of a special degree of contact 18 that of Marius Victorinus Semvocales m enuntiatone propria ore semicluso strepunt.* Our own term ‘semrvowel’ has its origin in the Greek jyidwror, through the medsum of the Latin semrcocals, whilst 1ts employ- ment, though not its justification, generally corresponds more closely to that of the Sansknit antekstha It1s further to be noted that our term ‘hquid’, a word more con- venient than descriptive, owes its ongin to the west rather than the east The Greek term typds, It ‘monst’, ‘fluid’ (translated by the Laun lgurdus) 1s first used mn a phonetic sense by Dionysius Thrax, who apples it to the Greek 1, r, m, 2 ® most of his com- mentators interpret the word as meaning ‘slippery’, 1¢ ‘unstable’, with reference to the metrical effect of these sounds as second members of a group stop-+-liquid, where a preceding $3 lable con- taming a short vowels of ‘doubtful’ quantity, a state also referred to 28 dypds7 Terentianus Maurus, however, explains the term 2s referring to their ‘lubrica natura’, in that they may funcuon either Blass, Pron of Ancient Greek trsl. Parton, pp tor ff), but such a pronuncuen cannot be supported so early as second century Bc. * Loc at. * CF Codex Upsahenns, ed Dablerup & Jdnsson, pp 61, 65 ? Op ext. p 97 * ung 1 sfytien astltam svardmudropmanan axpystans sthatam. ® hell, v.32 * Ed. Ubhg p 14 = Bekker, p 632 7 Hilgerd, pp 46, 342, Bekker, pp. S16 ff Cf Pnsaan, Kel, u. 9, Mex. Victonous Keil, ve 21632 PHONETICS IN ANCIENT INDIA as vowels or as consonants'—a remarkable interpretation as applied to the phonological systems of Greek or Latm? Atihus Fortu- natianus sees in the term a reference to lack of tenseness (quae manus virtum habeant) 2 and other interpreters of Dionysius Thrax refer simply to their ‘smooth and even articulation’ ¢ ‘Liquid’ 1s in fact one of those terms of which Grammont has said, ‘Elles sont consacrées par un long emplo1, grace auquel le lecteur sat immédiatement de quo} on veut parler, des appellations nouvelles pour- raent tre plus adequates sans offmr le méme avantage ’ 1.113 Retroflexion Amongst the intra-buccal articulatory processes we might have expected the Indians to have mentioned one further feature, namely, retroflexion This, however, 1s generally discussed by them m connexion with the places of articulation (see 2 03), and also in relation to its prosodic function (see 3 10) To consider the retro- flex articulations on the same terms as the velars, palatals, dentals, or Jabuals ts, even from the pomt of view of the Indian descriptive framework, not entirely justufied In the 7'P we find a prescnption regarding the position of the articulators in their quiescent or ‘neutral’ state (a close parallel to Snevers’ ‘Ruhelage’® or ‘Indifferenzlage’) 7 5 . the tongue 1s extended and depressed, and the lps are in the position oT a.’ * Keil, ws 350 Graeeus udas nominat lubniea est natura vi lis namque et alternus vigor mune emm votalts usumt nunc munistrat consonae 2 More justifiable 1a Macdonell s use of the term (Skt Gr, f 11,§17,B 1, on Skt 4, uy 7, 1}—Vowels which are lable to be changed into semivowels liquid vowels’ Cf Renou Gr Santer ,§5 7 Keil, 1 279 * Hilgard,p 46, Bekker, p 817 cf Psellusap Boissonade Anec Gr m 213, and also ps Anstotle, De Audib 803> Cf also Jakobson, ‘Observations sur le classement phonologique des consonnes’ (Proe 3rd Int Gong Phon Sc) p 40 Il semble que c est le fait du ghssement qui est décisif pour I impression acaus tape des consonnes en question? jp cat, p 7r ® Gr d Phonetik', § 55 7 Gr d@ Lauwphynologe, p 15, cf also Sweets ‘Organic wes (Pame, & 384 #) and Vietor s ‘Artukulationsbasis’ (Elem d Phon 5, §§ 128) M 20-21, anadese pramyasia pla akaratad osthau, on which Tnibh ,yatra sseanadefastatra = hoa. tusnim bhutdB " stud n der Mundane isnt utg Bhavats (cf Sievers, Die Zunge leegtPROCESSES 33 In this condition the velar, palatal, dental, and labial articulators are approumately opposite their respective places of articulation, and the utterance of these series 1s effected sumply by means of the closure-processes already discussed? This, however, 1s not the case with the retroflex series, which 1s articulated, as our treatises recognize, ‘by rolling bach the tip of the tongue’*—that 1s to say, the place of articulation 1s not automatically determined by the application of the closure-processes to the apical articulator 3 there 1s need of a further prayatna, ‘arnculatory effort’, which might with consistency have been included at this point ¢ 112, Extra-buccal 1.120 Glottal In their recognition of the voicing process the Indian phoneti- cians make one of their greatest single contributions The term for ‘voiced’ (ghosavat) 1s, as we have already seen, found 1m early non- technical hiterature, and the specialist discovery 1s hkely to have been of even earlier date To designate the glottis the Indians use either the word kantha, which in non technical usage means simply ‘throat’, or more specifically khah (or é:lam) kanthasya, ‘aperture of the throat’ In the Indian musical literature we also find the picturesque term Sarr: vind, ‘bodily lute’,5 which some authorities have interpreted as referring to the vocal cords,® in a recent paper, however, Dr A A Bake has pointed out that this interpretation 1s unfounded, and that ‘strange to say, there 1s no trace of the know- ledge of the existence of the vacal cords 1n the texts on the theory of music’? The following are typical of the phonetic statements The ar, respiration, or pulmoruc emission, at times of vocal activity, . * Cf Sievers Phon,§57 ‘Die Ruhelage des Sprachorgans 1st die naturliche Bass fiir die einzelnen Articulations bewegungen welche zur Bildung von Sprachlauten fuhren * * ee TPu 37 phvagrena pratiwegtya murdham tavarge APi 22 murdhan yanar phe agram pratvestitam VPs 78 murdhanyah pratsvestyagram 2 The VS slone suggests this basre distinction (228-9), Ravargadisu jihtads madhyantogthena copart favarge waktra madhyena phvagrena yatha spriet CF Luders pp 92 94 * Note ‘Trubetzkoy s inclusion under ‘Ergentongegensatze’ (FCLP w 103 £) + eg Samgiedarpana : 48 Cf also Ait Ay m2 § § Cf J Grossetin Lavignae Encyclopédie de la munque 1 285 7 In a paper “The Anatormeal Background of Indian Music’, read at the Leyden Congress of Orientalists June 1950 Brs98 D34 PHONETICS IN ANCIENT INDIA becomes breath (fvasa) or voice (vada) according as the glottss 18 open or closed * ‘When the glottrs 1s closed voice 1s produced, when it 1s open breath? The words used to refer to the two poles of this artrculatory process are samuyta, ‘contracted, closed’, and vizta, ‘opened’ 1t will be remembered that the latter term was also used in the description of the intra-buccal processes As regards the relationship of the voicing process to the various letters, the AP states, Breath 1s emitted in the case of the voiceless consonants, and voice in the case of the vaiced consonants and the vowels $ The Mth also notes the effect of the voicing process on inter- vocalic ‘voiceless’ stops, saying of the ¢ in 2 word such as pacati, It 1s overlard by the young & the preceding and following vowels,* a feature whichis further supported by the statements of the Prakrit grammanans $ Apart from the two poles of ‘breath’ and ‘vorce’, a number of treatises introduce a further factor into their descriptions, the RP observes that, When the glottis 1s in an intermediate condition (between clased and open) both breath and voice are produced,® and goes on to relate this feature to the individual letters as follows, Breath 1s emutted for the voiceless sounds and vorce for the others except for the vorced fricative (lh) and the voiced aspirates, where both breath and yaice are emitted ? * RP xur 1-2 cayuh pranch kosthyam anupradanam kantuasya khe wreyte samegte ta apadyate fvasatam nadatam va vaktrihajam * TP u 4-5 samute kanthe nadah kriyate mute fvdsah Whilst ghosavat aghosa are generally used for voiced, voiceless’, the usual term for voice 13 nada and not ghosa this fact leads to misunderstanding on the part of later writers. who list nada and ghosa as separate processes—e g Siddh Kaum 12 : 4 12-13 sudso ’ghosesu anupradanah nado ghosavat svaresut hav 4 on Pam t1v rog(Kielhom 1 355) purva parayor hradena pracchad yate, f& Kayyata dvayor akdrayor ghogavator madhye eakaro ghosavan wa lakpyata sty arthah * Nee Dole, Les Grammainens Praknts, pp 1g1 ff ,ef Bloch L'Indo Aryen PPT * xu 2 ubhay am wintarobhau nda mut 4-6 fvdso"ghoranam wtaresam tu nadah sosmosymanam ghopném hdsa lauPROCESSES 35 This 1s supported by the statements of the 7P, viz When the glottis ts in an intermediate condition, ‘h sound? is produced," For vowels and voiced (unaspirated) consonants the emission 18 voice, for voiceless consonants breath, and for h and the voiced aspirates ‘h sound’? Further, the RP condemns as a fault in the pronunciation of h ‘excessive bteath or similanty to a voiceless sound’? Regarding this third category of glottal ‘half-closure’ or ‘h- sound’, Max Muller remarks, Dies ist eine indische Vorstellung, welche wohl nicht 2u rechtfertigen 1st,* and Whitney in a series of unsympathetic comments, T confess myself unable to derrve any distinct idea from this desenption, knowing no intermediate utterance between breath and sound (The RP) declares both breath and sound to be present in the sonant aspirates and in hk, which could not possibly be true of the latter, unless at were composed, like the former, of two separate parts esonant and asurd, and this is smpossible 5 The attempt to establish this distinction 1s forced and futile ‘That sntonated and unintonated breath should be emitted from the same throat at once 1s physically impossible © Needless to say, the two western scholars were wrong The modern Indo-Aryan languages bear ample evidence, if evidence were needed, that the aspiration of the voiced aspirates (6h, jh, &c )as vorced aspiration,” and there are strong historical and phono- lopical reasons for believing the Sanskrit h to have been ‘voiced h’ [fi] 8 the possibility of such an articulation 1s no longer 2 matter of doubt—to quote one of many available descriptions A-voiced ean be made For this sound the vocal cords vibrate along "6 madhye hakdrah 2 8-10 nddo nupraddnam svara ghosavatsu hakdro ha caturthesu aghoseyu Stdsah Cf y2-14 0 uyma essaryaniya prathama-deitiya aghosth na hakdrah waijana tego ghosavdn xv 28 feaso ghoga nibhatd td hakdre * On RP 710 * On AP’ 13 © On TPu 6 7 Seeeg J R Pithin Harley, Collegual Hindustam p xx eg he Phin dima (beside Av zima &c) Within Sansknt ef alternations such as ghnantifhant: dhd jhsta &c , and yunctions of the type tat thr = taddht < Thumb, Handbuch des Sanskrit § 542, Edgerton Skt Hist Phonology, asf36 PHONETICS IN ANCIENT INDIA a considerable part of ther length, while a triangular opening allows the air to escape with some friction * ‘The failure of early western phonetics to take note of the voicing process has already been suggested Anstotle, in a passage of his Historia Anmmaltum,? m fact comes nearer to 2ts discovery than ts generally recogmuzed in making the distinction between vowels and consonants he says that whereas the latter are produced by the tongue and fips, the former are produced ‘by the voice and larynx’ But the matter 1s not further pursued by him or his successors, and the western tradition 1s really that which begins with Dionysius Thrax, the latter distinguishes the three classes of Greek stops (voiceless, vorced, votceless aspirate) by their degrees of ‘aspira- tion’, viz as ‘smooth’, ‘medium’, and ‘rough’ respectively,) the voiced stops being considered from this point of view as inter- mediate between the voiceless unaspirated and the vorcelessaspirated stops It 1s difficult to see how this classification can have been justified Sturtevant has now nghtly abandoned an earher theory that the statement could refer to a lens/fortss distmction,* and another hypothesis, which would to some extent fit Dionysius’ classification, ‘namely, that the Greek B, 8, y represented vorced asprrates, 1s supported by no positive evidence whatever 5 But whether or not this description was ever applicable to Greek, rt was evidently not in the case of Latin, and with one notable exception 1s not taken over by the otherwise ovine Latin gram- manans The responsibility for transmitting to us the stil familiar Latin translation of the Greek terms—tenuzs, media, and asprrata® -—must be bome by Priscian, who takes over the Greek classifica- tron mits entirety Undaunted by the fact that Latin possesses no asptratae, he apphes the term to the Latin frcative £, which he * Westermann and Ward Practical Phonetics for Students of African Lan- oe Char ; Pome na a) oi wmevra 1p Guar} Ker d Aapvyé adqaw, ra Bt ddwve if p ref = Bekker, oT Te Bape $x phon B vera sla Bo eee ee ee eri juetek cpa tuw BR Baaduv puddrepa Cf Dion Hal, De Comp 83, Anistides Quintlianus, pp 29 54 Jahn ster? pagieiidn distinction 1s an fact found eather in Ps -Anstotle, De Aud + Op cit,p 86 n 89 5 Ibid ® The Latin translations fenuis and asptrata (for Gk. filf Bt are remark we we should expect fems and aspera (ef spinitus lentsfasper for wrevpa pdde]PROCESSES 7 identifies with the Greek ¢# (the latter having by his time probably developed its present fricative value). The ambiguity which the term thus developed was ultimately to provide Jacob Grimm with a deceptive symmetry in his famous statement of the Leutrer- “sehiebung? As regards the nature of the distinction between the voiced and voiceless Latin staps, the ancient writers seem to have had only the vaguest impressions. With t and d there is the suggestion of a different place of articulation’—a method of differentiation that we find perpetuated in Ben Jonson's English Grammar, ‘The learn- ing of the distinction is recommended by Quintihan as an essentzal item in a boy's cducation,* but he guardedly omits to discuss the matter in detail. Ina description by Terentuanus Maurus, dealing with the distinction between b and p, ¢ and K, there is a sugges tion, though obscurely expressed, that the author had recognized the extremely important denis{fortis opposition (which Whitney was rather too ready to dismiss);§ and Mars Victorinus, para- phrasing this statement, does in fact use the former term ® But any good that may have resided in these descriptions quickly perished; the medieval grammarian Hugutio distinguishes alquando from aliquanto by the position of the stress, heet enim det ¢sint diverse htterac, habent tamen adeo affine sonum, quod ex sono non posset perpend: aliqua differentia,’ and in the seventeenth century John Wallis states the distinctive feature to be nasality.! Only in the latter part of the nineteenth century, under the influence of Indian teaching, does the recogni- tion of the voicing process make headway. nr2t, Pulmonic 2 2 . . . Tt will be convenient to consider next the process of aspiration, in ag much as one of our treatises, the PS, links this with the voicing process: hand the voiced aspirates are voiced, the sermvowels and vorced stops * Keil ik 20+ Inter ¢ aine aspuratrone et cum aspiratione est g, inter Equoque et th est d, et inter pet pk sive fest & Sunt imtur hae tres, hoc est 6, d, ¢, mediae, quee nec penitua catent aspiratione nec eam plenam postutent * * For discussion cf Jespersen, Language, p 44 eg Mar, Vict, Keil.vs 33 tow 06 * CF ‘On the Relsvon of Surd end Sonant’, Trans Am PAT Aun 1837, Pp. 4qut * For texts see Sturtevant, op cit. § 1890 1 Haase, De Medi Aect Studus Philology, p 34. ® Geemmatica Panguce Anghcanse, pp ist38 PHONETICS IN ANCIENT INDIA are partly verced , the voiceless aspirates are breathed, the voiceless stops are partly breathed This is the Jaw of speech * In other words, h and the voiced aspirates are considered as more fully voiced than the non-aspirates, and the voiceless aspirates more fully breathed than the non-aspirates In the case of the voiceless patr the statement makes good sense when we consider vasa, ‘breath’, as referring to ‘force of voiceless breath’ rather than simply ‘voicelessness’ Thuis interpretation 1s supported by the statement of the 7'P that, More breath 1s emitted in the other vorceless consonants (1¢ the aspirated stops and the fneatives)? than in the unaspirated stops,* and 1s in accordance with the grammatical and later phonetic terminology of the distinction between aspirate and non-aspirate, namely ‘maka-prana’, ht ‘big-breath’, and ‘alpa-prana’, it ‘lutle- breath’ ¢ If we now turn to the PS s statement regarding the voiced pair, similar considerations apply nada, ‘voice’, being interpreted as “force of voiced breath’, the statement implies greater breath-force on the release of the aspirates than of the non-aspirates The Justification for such a statement 1s clearly reflected in ky mographic tracings, where the voiced breath correlates with a particularly high amplitude in the vacahic wave forms,} from this point of view ht and the release elements of the voiced aspirates may be considered as an ‘overblowing’ of the following vowel® (cf alsa 2 00 below) ‘The Indian treatment of the aspiration-process provides httle else for discussion, with the exception of a statement in the RP that, Some say that the aspiration of the aspirates consists of a homorganic fricative,? . * 39-40 nddino ha shagah smptdh Jian nddd yon yata$ ca fidrinas tu kha phadayah : Fyac-chi ddim caro eidyad gor di dmautat pracaksate . CE Plato, Crat 427a(— ~ rou der nal rou fee wal vou ciyya war rou fyTo Sn wrevparudy 7a ypoppara) 2 at 11 dbisydn prathamebhyo "myers ‘The WS (280 282 Liders p 93) gues to thay fPeceal Scaree of breathiness the ttle of arka eg hte 4 omnPdn is o(hiethom c Sen Proc and Int Cong Phon soft i Sidy Sw alt COA ‘ CE Prasad op ct, Kymograms Nos 85, 92 (tho ghar, bhabhut} ‘ Cf Pike op at, pp 31 fF, Doke Comp Study sn Shona Phonetics p 92 ant 16 sopmatdn ca ropmandm usymandhuh sasthdnenaPROCESSES x” a view that as repeated for the voiceless aspirates by the Ap St The breathy release of an aspirated stop inevitably has, as Sweet observed, ‘something of the character of the preceding consonant’ ;? and, in the sorceless senes at Icast, varying degrees of affrication are to be heard from speakers of some modern Indo-Aryan lan- guages Dhalectal pronunertions of this type are likely to have eusted in the case of Sanskmt; the later development of the aspirates, however, hardly supports Uvata’s assumption of a strongly affmeated pronunciation (p>, ts, hx) such as ts attested in, for example, High German 1.122, Nasal ‘The nasalazation process may occur in combination with vanous intra-buceal processes, and the ancient statements regarding the mode of combination for the most part present ne preat difficulty of interpretation The nasal consonants are referred to either a3 nasrky a, ‘nasal’, or anundnka, ‘having a nasal component’ Of the mechanism of the process the TP saya simply that, ' Nasatity ts produced by opening the nasal cavity * Qur phonetic treatises, 23 also Panins, reahze that both nose and mouth are involved,‘ and the TP further points out that the articu- lator 13 as for the corresponding oral consonants * Applied to the stop series (sparia, sprz{a) this process gives mse to the nasal con- sonants 1), ft, IL,,m nasalized forms of three of the semuowles, ¥, 1, €, are also attested as junctional features in Vedic Sanskrit and duly noted by the phonetscians ® In connetion with the vowels the working of the process 1s aumilarly quite clear, Here again the term anundska 13 regularly Vay g sartRlnena dettiyth * Pamer,p 59,cf Pike, op at, p 383 Jan g2 ndnkd-trwarand § dnundnkyert “ AP’ 27 anundakdnuin mutha mdkam VPs 73) mukha ndnkd kavana nundnkah (cf 1 89) Pda 1 4B mubha ndnkd-vacano ‘nundnksh Fn gt cargatae congu * Ch TP 28 antasthd paral ca sat artam anundsikam eg saSyudht for sam+yudhi suvarpal lokam for suvargam+lokam yajnat vagpy for yajpam+vaspu Onty Tas found in Classical Sanskat (and only when = phonological -n) Fs not attested (ser, houes er, Konow, op ot, p 309), this ss 1n accordance with the non-occurrence of geminate r{ef junctions of the type S-lg +r-> Ml Fe for expected "sirre)4° PHONETICS IN ANCIENT INDIA used, as opposed to the fuddha or ‘pure’ non-nasalized vowels,t Another term, however, is also used by some of the treatises, namely rakta, ‘coloured’, nasalization being referred to as rdga or ranga, ie. (nasal) colour’.+ Some of our authors give picturesque descriptions of the quality of these vowels: the account in the Sarvasammata- Siksa is as follows 3 4 ‘The nasal colour should anse from the heart,* bells: ust as the mulkmands of Suristra cry ‘takri should the nasality be realized The nasalized vowels are not of frequent occurrence. They appéar in certain types of yunctton (e g. trite ekaada faa iha for triin-+ ekaadafaan-t iha),5 and as features of finality n the sentence or breath-group Apart from such cases there was a tendency, cen- sured by the RP but general in the modern Indo-Aryan languages, for vowels to take on some degree of nasal ‘colour’ in contact with nasal consonants.” But apart from the above, the accounts given by our treatises, as also the system of writing, present us with a third sub-category of the nasalization process. The name which this third feature bears 18 ‘anusvara’ (yt, 1y), which might be literally translated either as ‘after-sound’ or ‘subordinate-sound’, The contexts in which it may occur are clearly defined. It is restricted to post- vocalic position, and its primary context is before the fricatives J, §,8(andalsoh), in cases where historical and phonological evidence point to an alteration with m or (medially) n;8 at an early date it also made its appearance under certain conditions before r,? and in Classical Sansknit replaces the Vedie § and,¥ before y and Vv 1 AP w. 121, Varnapatalam wi 5 ; ¢ & RP: 36 rakta-samyiio "nundnkah 4 asound like that of 4? (‘buttermilk!’), so ‘ hamsy a-dhvant-samam raigem hrdaydd utthitam bhatet Jathd saurdstrike nari takrds sty ablabhdsate etam rangdh prayoktauyah oo, Cf Varma, op at, pp 149f ‘ ‘The Cards aniya-Siksd has, more realistically, ndsdd utpadyate raigah .. Cf also the so-called emunankopadha weit (RP u 67), eg sacha indrah for sacaa+indrah, * CE RP. 63 astdy adyan ov asdne "pragrhydn dedryd dhur anundnkdn seardn Sen arammata-Stkpd 46 pluto'earnak padinta-stho ndskyo rakga-samyiakak i RP xv 56 raktai rdgah tamavdye seardndm Cf also xiv. 9 Cf tan) sarvam beside tam api, &, hamsa besde Germ Gans, Gk aie &e (ghans-) ed. SaMraat bende SaMsat, Ke su; s the posternonty of the sequence -Myre (cf. Wackernagel, op cit, § 283¢) eee postenonty *PROCESSES ‘“" respectively (see above) Its optional use is further extended, even as early as Panim, to word final positron preceding a stop,! where previous phonetic teaching had prescribed 2 homorganic nasal ,? later treatises extend this practice to morpheme junctions within the word, and even to intra morphemic position,? being followed in this last extension by the graphic practice of manuscripts and of some printers In Praknit it 1s further extended to the position before an initial vowel + The phonetic value of this feature, however, has provided a problem for phoneticians and commentators whether ancient, medieval, or modern Some confusion may perhaps be avo:ded by first considering the evidence for 1ts pronuncration in the earlier and more hmuted contexts, namely, before the frcatives Sequences of the type -VimS- (where V = any vowel and S = any fricative) were phonologically parallel to sequences of the type - VLL- (where L = any semivowel except r} or -VNT- (where T' = any stop and N = homorganic nasal) sang skrta, for example 1s parallel to sal-laya and to san-taana, sam-paata, & Thus the nasality in the types -VEL- and -VNT- conforms to both the sthana and the abhyantara prayatna of the following consonant, by sts paral lelism with these sequences one might also make the theoretical supposition that -VmS- = -VZS- (where Z 1s a nasalized frica- tive), a form of realization that may be heard, for example, in Modern Icelandic, where in a phrase such as soln skin fegurst the word junctions are realized with some overlapping of the nasaliza- tion and friction processes (in a segmental representation -Z s-, -¥ f-)5 In only one Indian statement, however, 1s there any implication that my 1s to be considered as a fricative ® Ina number of languages the tendency 1s in fact for the nasality in such contexts to be realized in the preceding vowel,’ and fer the * yur av 5 2 Cf TPy 27 makarah sparia paras tatya tasthanam anunankam 2 CF Sarvasammata § 32(eg sary Jnaanam, damkgru) * CE Pischet § 348 (Hemacandra 2 24) 5 Cf Stefan Emarsson Icelandic Grammar pp 19 29 Bestrage zur Phonenk der Islandischen Sprache p 18 Note also Petrovic: De la nasahite en Rouman Ce nest pas seulement Je heu d articulat on qui devient identique 4 celur de Ia consonne suivante mats aussi le mode d articulatron * RP1 ro with Uvatascomment On the general rareness of nasal fricatrves cf Grammont Traté p 95 Martmet TCLP vu 282 and on the mcom patibility of friction and vorce in Skt cf p 44 2 4 7 CE especially the development in Avestan (Bartholomae Gr § 62) eg dqstugm bende Skt daysak mafram beside Skt mantram42 PHONETICS IN ANCIENT INDIA syllabic quantitative pattern to be maintained by a Jengthenmng of the vowel, as Sweet long ago pointed out with regard to Latin, m before the hisses and semivowels represented a nasal lengthening of the preceding vowel? The same development 1s postulated for *-ns- in Insh,? and 1s further attested in Old Lithuanian by alternations of the type kandu(pres.) kqsu (fut) > Amongst the modern languages parallel alternations may be quoted from Polish* and Spanish 5 On the basis of analogies rn other languages it 18 therefore tempt- yng to assume that the value of m was a nasalization and lengthen ing of the vowel (if not already long), this view was adopted by Whitney,® and has support in the fact that the TP speaks of anus- tara as having precisely this value,’ whilst the AP makes no mention of it apart from anundska ® Whitney further supports his interpretation by the fact that the anuszara symbol 1s written over the vowel-symbol {eg Hy or Aa for amja) The lengthemng of the vowel is amplied i the statement of the JP that metrically a syllable which 1s nasalized is equivalent to a syllable containing a long vowel 9 But on the other hand some of our authonties quite certainly distinguished the terms anunanka and anusvara, as the following passage from the VP indicates According to Aupagivi anundstha of a vowel occurs only before a fol- lowing vowel (Usata, ‘eg maha indrah ), whilst before a following consonarit there as an insertion of anundra (Uvata ‘eg gavayaars tvagtre ) 1° A similar view 1s taken even by the 77S, which in most respects 1s * Prot Phil Soc 1882-4 p xv Cf Sturtevant op cit § 174 , Cf Petersen, Keltusche Gr 1 $$ 53.10 943 P fame Leskien Litowsches Lesebuck § 25 2, note also present tense formations of the type seqlu ‘The nasalized vowels of Old Lith (q &c ) are now pronounced as non nasal long vowels Cf Broch, Slaviche Phonetth § 124, Sokolnicka Izdebska Etude Expért mentale des Comonnes Nasales en Polonats pp 38 ff, also L hoschmesder 7 f Aa! Sprachforichung 69 3/4 1951, pp 219 ‘ Cf Navarro Tornas, Promincianén Espaitola pp 311 f £ On TPs 30,¢f. Grammont, op at, p 365 3 Lec cit anusrdrottamd anundnkdh withxy 1 purea vara ‘nundnkah : ct Sarma oP sit Pp 14847, Renou se xu 14 modem developments such as Hind: ba s-N'S- (cf the Sansknt sandh: mahaantsan for mahaan-+san, &c)® The extension of anusvara in Classical Sanskrit to cases where the following consonant 1s a semivowel provides no fresh difficulty But, as already mentioned, Pani: allows the optional use of my in certain cases where earlier treatises prescribe 2 homorgantc nasal It has been suggested that this extension was purely graphic, and based on the convenience of the simple symbol to represent nasality ,° this seems less than fair to Panin1’s iinguistic competence, though the possibility of a phonological rather than phonetic basis for the extension must not be overlooked Graphic considerations may well underlie the later extenston of * mun 4t (hrawamisvara vy atrsangavat) * alfau are an fact contrasted with efo, where the qualities of a snd ifu are “fused (xin. 40 samsargad) cf 2 13 below 4 Cf Bourciez Eléments de iinguistique Romane, §333e,E B Willams From Latin to Portuguese, § 157, A R Gongalves Vianna Portugais Phondtque et Phonologie §§ 29-32 * Of interest, though of doubtful historical significance, xs thé Marathi pro- nunciation of Skt tatsamas,e g aW§? (< amia ) -maWs (< mdmsa ) ohiivsa (< ahimsd) sitvh? (< sha) If this does in fact reflect an historical feature + the realization of anusedra mught be simply stated in terms of homorganic articulstion for every case including the freatives—e g tam-+s- > t-aZe-, fmetion and voice being mutually incompatible, the heavy quantity of the syllable, af at contains a short vowel, 13 maintained by a W type vowel closure (cf on ovarga sandhi p 68 n.1), thus ® af would > a s-, & Cf Varma, op at, p 15, and VS 170 (Litders p 87) Note also the Presenteday Bengali Pronunciation of Skt mas fe however the 8 reference Wht Smuliyanusvara hanu mulena), ef pS : any of unspecified type(s 83 7 uu a2 * Cf Bloch op ait, p 88 * S K, Shastn, RT Notes, P 54 cf Bloch, op cat p 40PROCESSES 45 anusvara to replace homorganic nasals within a morpheme! Of such extensions Whitney has said To write the anusvdra signin the mterior of a word for a nasal mute which 1s equally radical or thematic with the succeeding non nasal seems an indefensible practice and one wholly to be disapproved and rejected 7 These strictures are directed against Max Muller’s support of the practice referred to,? but there 1s much to be said on the other side For although thts ‘slovenly and undestrable habit 4 may well have grown out of mere graphic convenience, it mcidentally recognizes an important phonological pnnciple namely, that the n or mm in a sequence Vnt or Vmp is a very different functional unit from that an VnV or VmV, for whereas 1n the latter case n and m are mutually contrastive, this 1s not so in the former case’ The homorganic nasals form a single phonological unit, and a phono logical transcription will recognize this fact ¢ Thus 1s particularly the case in those modern Indo Aryan languages where the only purpose of certain nasal symbols taken over from Sansknt (velar, palatal 7 retroflex) 1s to represent homorganic nasality before con- sonants of the appropriate series In some dialects, moreover, we find alternative pronunciations of the type VNT/V Ty? 1e an alternation of homorganic nasal (para savarna) with nasality and length of vowel (anunastka), in such cases the convenience and Phonological appropnateness of a single symbol for the two pos 1 Lithuanian here provides another parallel im that the indigenous grammars extend the nasal vowel symbol (which 1s phonetically justified before the frica tives) as a graphic device for writing the homorganic nasal before stops (eg pjts for pint) cf Leskien op at § 26 Note also that the Sarvasammata Siksa i presenbing this extension (32) acknowledges that amusvara here lacks sts peculiar properties ( sty atrdnus tdro_p: e:dharmakah ) 2 FAOS x égn ? Hitopadefa Introd pp x1 “Whitney Ske Gr § 73) $Cf£J R Firth Proc 2nd Int Cong Phon Sc p 180 © Note also the use of a special eymbol (#) for the homorgamc nasal in Avestan where there can hardly be any quest on of graphic convemence(® = 14 D=3m-=—6) ef Bartholomae Gr $268 53 ? Even in Sanskrit the palatal pi cannot be justified on a distinctive basis (cf Emeneau Lang xxn 89 ff) * Cf Prasad op cit pp 467ff (eg lamba fl ba) Sumuilar alternations ate hustoncally suggested by doublets an Marathi (cf Bloch La Formation de ta fangue marathe p 82)46 PHONETICS IN ANCIENT INDIA sible realizations are obvious ! Possibly simular alternations were prevalent in more ancient times, and pave rise to at least some of the confuston which besets the early phoneticians ‘The problem of anusvara has been considered at some length; for whilst at 1s 1n itself disappointingly unrewarding, :t Serves to demonstrate how little we mrght know if our sources had been equally imprecise on other points In view of their generally high standard of competence 1t seems fair to assume that the phonetic problem in question was a particularly difficult one, complicated perhaps by multiple contextual dialectal, and personal fuctua- tions If we were to seek an attested feature of a modern language such as might give nse to uncertainties of a similar order, 1f not of type, 1t might perhaps be found in the Japanese so called ‘syllabre nasal’ (n)? which has so strangely received special recognition amongst the otherwise general phonetic categories of the Inter- national Phonetic Alphabet 113 The Vargas With regard to the three extra buccal articulatory processes dis- cussed above(glottal pulmonic, nasal) st should finally be remarked that their combination with the intra buccal process of maximal closure gives rise to a set of five sparfa letters for each of the five * Cf Bloch toc cit ‘Ia regularité de cette alternance fait que le seul signe de | anusedra suff t & noter ces deux cas dans le cas of 1a voyelle est bréve sf teprésente la nasale de méme ordre que la consonne qui suit, si 1a voyelle est longue, | amundra a la m&me valeur que | anundstka* * Before condemning the Indians for their disagreements and obscunties in the desenption of amurdra it may be salutary to compare the diversity of modem desenptions of the Japanese feature in question. the following may be referred to Hi Fre: Bull de Iq Masson Franco-Japonaue vit 1 1: . HE Pater The Prnoptn f here port ” D jous The Phoneme p 88 n.3 wards Etude phonétque de la langue saponante 1 MG Mon The Pronunciation of Japanese, ‘45 18 Pps POM Suski The Phonetics of Japanese Language pp 71 ft B Bloch "Studies tn Colloquial Japanese IV (Phonemics), Language xxvt 1950 p 102 S E Martin Morphophonemses of Stondar (Language Dutertation No 47) pp ” f t 4 Coltoqenal Japanese ( “ Whilst not implying adverse criticism of all these statements we may wonder whether linguists in @ distant future, reading such vanous accounts—=ranging from semu-consonne ou semi voyelle to voiced frictionless mediovelar spirant oil have any clearer an idea regarding the phonetic value of the Japanese eyllabic nasal than we have reparding that of the Sansknt anusedraPART II LETTERS 20 Consonants Tris the Indian practice to describe the places of articulation in the reverse order to that of the IPA Quite logically they begin with those which are nearest to the ongin of the arr-stream and work progressively upwards and forwards towards the lips < 200 Pulmonte and Glottal Thus the first organs to be considered are the lungs, which are treated as the place of articulation for the vorced h [fiJ and vorce- less -h. This treatment, however, 1s optional, most of the treatises also allow these sounds to be classed as ‘glottal fricatrves'—a term which 1s still commonly accepted today, though in need of revision (the Greek term ‘breathing’ (mveiia) might be more appropnate) ! The following statements allustrate the alternative prescriptions The fricatives b and -h are glottal (kanya), or, a8 some say, pulmonic (urasya) * , hb and -h are glottal, the latter may alternatively be considered as pulmonic 7 Certain of our authors allow the pulmome alternative only in the case of h followed by nasals or semivowels h before nasals and semuvowels 3s to be considered a3 pulmonic, other wise 1t 18 glottal + * Particularly in view of ats frequent presodic function (the Greek wveipa is a spoop'la), cf J R. Firth TPS, 1938 p 131 * RP i 39-40 prathame paiicamau ca ded usmandm {sc Ranjhyau) kecd td uras au ‘ Rt 23 Aah kanthe uran tuaramyo td hakéram paficamar yuktam antahithdblat a samyutam auratyam tam wydmydt kanthyam Shur aramyutam Cf Sarvasammata-S 42 hakdram auratam exdy&d antasthdea bards ca utlamesu poresu ecam ‘The baus of this disunction 1s problematic, but it 1s to be noted that for one form of yuncton innal groups of the type h-+-nasal or seruvowel are optionally treated as if the h were phonematically irrelevant (kin hnute, kif hyah, &c.— ace further 3 122 below) 1¢ bis considered asa Prosodie, non knear feature of\ ‘ LETTERS 49 We have already suggested that Sanskrit h [8] might be con- sidered as an ‘overblowing’ of the following vowel, the close relationship of both h and -h to their vocalic context 1s mentioned by the TP , For h and -h the glottis 1s the place of articulation, but in the opmmion of.some authonties h ys homorganic with the beginning of the following vowel, and -h is homorgame with the end of the preceding vowel * Whuitney’s observations are for once entirely sympathetic, with reference to the pulmonic nature of these sounds he says,* The authonty who called the asprrations chest sounds may also be commended for his acuteness, since in thezr production it may even be satd that the throat has no part at 1s only, like the mouth the avenue by which the breath expelled from the chest finds exit 3 elsewhere he shows himself to be m agreement with the view that they are homorgamic with their vocalic context Why, then, shall we pronounce the larynx the ‘charactenstc place of production’ of 4, any more than, of the vowels? An ius asound which 1s produced in any one of these sarne positions of the mouth-organs (sc as for the vowels), but wath the vocal corda mm the larynx only slightly approached + « The RP rather surprisingly hsts this homorganic realization amongst the ‘faults’ in chap xuv,5 but Uvata quotes another com- mentator who considers this statement to be out of place, and who prefers ta treat it as a rule rather than a prohibition—‘for not even the gods could pronounce it im any other manner’ | ‘This will be a convenient point to give some account of various breathiness (cf BSO.AS xu: 944£) the realization of which may occur simul- taneously witls that of the phonematic units It 1s perhaps this peculanty that Our treatises intend to indicate "46-8 Ragtha sthdnau hakara vssarjamyau udaya svaradi sasthanohakdra ekesam piirvanta sasthdno visarjanyah As the Tribhazyaratna expresses it, they have no articulator of their own (anayok karanabhavah) CF Sweet, NEG I, §237,D Jones, Outhne of English Phonetics’, §§ 977 £f , Broch, Slavische Phone- tk, §§2, 51 7 On AP: 29 7 T cannot agree with Fry (Lang xvn 199) when he states “The use of the Adjective eurasa does not appear to be more than a hazy attempt to localize the open spyrants representing # un pausa’ * FAOS vi 350° cf D Jones, op sit, p 23 n 1, H Abrahams, Studer Phondtsques sur les Tendances Evolutwes des Occlunves Germamques, p 102 XIV 30. Sanya ithane —devarr aps na Sakya wecarayitum Brava EPa PHONETICS IN ANCIENT INDIA problems connected with -h This voiceless breathing primarily occurs only in final position at pausa, where historically st replaces #5 (or less frequently *7}! Corresponding to -h m pausa, there appeared in yunction with amtial voiceless consonants the appro- priate homorgante fricatives (viz =X, -J, -6, “5+ ->)? Three of these fricatives, J, 8, 8, occur also m initial and medzal positions, where they are in parallel distribution and qualify as separate phonematie umts, which consequently find their place amongst the other ‘letters of the alphabet’ -x and -$, however, as also ~h, are (a)yogavaha’ (cf 0 42 above),1¢ they are bound to final position, and are in complementary distribution (-x before velars, -« before labuals, and -h inpausa) "Thisalternation1s understandable enough, and it 1s impossible to agree with Whitney’s evaluation of -x and -: It may be fairly questioned, perhaps, whether these two sounds are not pure grammatical abstractions > Since these variants are not included in the alphabet, special names are devised for them, viz vtsarjantya (or later visarga) for -hy uhvamultya for -x, and upadhmantya for -. The last two terms provide no difficulty shoamiliya, lit ‘formed at the root of the tongue’, 1s the general term for ‘velar’, and upadhmaniya means literally ‘blowing upon’—the consecrated description of the vorce- less bilabial fricative + The term for -h 1s not so readily explicable —a fact which 1s reflected in Monier-Willrams’s dictionary Its called Visarnaniya either from its lability to be ‘rejected’ or from its being pronounced with a full ‘emussion’ of breath, or from sts usually appearing at the ‘end of a word or sentence The verb from which the word 1s derived (e2-37)-) has meantngs of the type translatable by ‘to discharge, relax, cast off’, & We shall ; Cf eg gharmah beside Gk @eppds antah beside Lat inter AP ui 40 e1sarjaniyasya para sasthdno 'ghoze RP iw 31-2 aghose copmanam sparia uttare tatsthanam fam evog manam ugman RT 177 usma sthdnam (Comm upmd ea para sasthanam dpadyate) TP x 2 aghoja-paras tasya sasthanam usmanam VP in 6-12 2 Skt Gr,§69 Cf on AP us 40, The division of this indistinct and im definite sound into three kinds of indefiniteness savors strongly of over refine. ment of analysis" For other sceptics see note by Fry, op cit, p 19. eg D Jones, op cit, § 685, ‘One form of $ 2s the sound made mn blowing eutacandie’ GE Sapuw Language: 37 &LETTERS st perhaps be giving the most direct and phonetically appropriate translation 1f we render it by ‘off-ghde’, as referring to the breathy transition from the vowel to silence In later, though still ancient, times there appears to have been @ tendency for -h to extend its usage to contexts other than im pausa The earliest of these extensions was to the position before the initial fricatives f-, §-, S-, where it replaced the homorganic final -[, -g, -s (indraf Juurah>indrah fuurah, &c)' This practice was then extended to the position before the velar and labial voiceless stops mm connexion with this mnovation we find Mentioned the names of Agnivedya, Valmiki, Sakalya, and the Madhyandma school, whulst the ancient grammanan Sikatdyana? is quoted as holding to the more conservative practice ? These changes have been generally accepted so far as the writing of Sanskrit 1s concerned, and AH. Fry im his article ‘A Phonemic Interpretation of Visarga’ has suggested that the spread of -f was due to the writers of Classical Sansknit ‘operating with a phonemic orthography’ Though the term ‘orthography’ once again begs the vexed question of writing, 1t 1s possible that this extension had a phonological rather than a phonetic basis, but in this matter we are faced with similar uncertainties to those which enshroud the extension of anusvdra at the expense of the homorgamic nasals zor Velar The velar series (ka-warga) 18 most generally described as being produced at the jrhva mila, ‘root of the tongue’, which 1s, strictly speaking, an articulator and not a place of articulation the sfhana of this sertes 18 1n fact the Aanu milla’ or ‘root of the (upper) jaw’ —a rather madegquate though intelligible designation of the soft palate— In the & series contact 1s made by the root of the tongue at the root of the yaw & Amongst the velar consonants 1s also mentioned the velar fricatrve X (jthuamultya),7 to which reference has already been made "Cf TPx 5, VPm 10, Pan vir uw 36 ? Not to be confused with the nunth-century author of the Sakatayana Vya karana CF TP ix 4, VP in rr-1z 4 Longuage xvu 1p4 iF * Cf APs 20 hvamuliyanam hanu mulam ° Pn 35 kann mule jshea mulena ka varge sparsayan Cf ¥P1 83 7 eg VPi 65 RP: 41 RI4 Forthe general term velar t es Jhtsa Ingual instead of the usual shvamulya oorre§2 PHONETICS IN ANCIENT INDIA ‘The Indian term ‘root of the tongue’ has found favour with a number of Western phoneticians, amongst them Sweet! and Pike? It has to be mentioned that im the later Pininean scheme, as reflected, for example, in the S:ddhanta-Kaumudt, the pulmonic breathungs (alias ‘glottal fricatives’) are classed with the k-senes,? the whole group bemg referred to as ‘glottal’ (kanfhya)* ‘The ‘unequivocal name of the fricative pphodmilltya preserves it from this confusion 5 202 Palatal No difficulty 1s provided by the descriptions of the c-series, which at the period described by our treatrses appear still to have been true palatal plosives rather than prepalatal affricates such as are general in modern Indian pronunciations © They are described as being articulated ‘at the palate (¢alu)’ 7 more specifically, In the c-series contact 1s made with the middle of the tongue upon the palate 203 Retroflex | We have seen that the retroflex series involves a special process | {rather than a place of articulation Since, however, the Indian scheme treats this settes as parallel to the other vargas, and nett im order after the palatals, 1t will be appropriate to consider it at this pot ‘Though the term for ‘retroflexed’ (pratiesfita) 13 well attested in the descriptions given by our treatises, the general term for the retroflex series, employed by both grammarians and phoneticians, 1s rirdhanya, an adjective derived from mirdhan, ‘head’—e g For the murdhanyas the artreulator 1s the tp of the tongue retroflexed ? In the t-senes contact 1s made with the tip of the tongue rolled back in the murdhan © T Promer, §9 * Phonetics, pp 1z0f 1 SK 10 a ku ha usarjanyandm kanthah Cf Ap $17 ° “ In this imprecise usage kanthya 13 perhaps best rendered by the equally amprecase guttural ‘ sx ° phvamuliyasya pha mulam ee also Grierson FRAS, 1913 pp 391 7eg VPi 66 RTS : TP 36 talau pha madhyena ca varge CL APi 21 VPs 79 2 AP 1 22 mmurdhanyénam ysheagram pratweshtam Cf VP1 78 TP u 97° gshtagrena pratwwestva murdham fa vargeLETTERS 53 Commenting on the latter statement, the Tribhasyaratna says, By the word mirdhan 1s meant the upper part of the buccal cavity," But there as no evidence that the word was ever used 1n this special sense, and comparisons by modern commentators with Greeh ‘ odpavds? (Lit ‘(vault of) heaven’, thence applied to ‘roof of the mouth’) are hardly relevant Mdrdhan means simply ‘head’ or ‘summit’,} and the Indian terminology 1s reflected in the still not entirely obsolete terms ‘cerebral’ and ‘eacuminal’ + The term:s m fact unusually imprecise, and Whitney 1s probably night in sug- gesting that it represents a traditional title surviving from a period when phonetic science was less well developed’ (cf also the term uisman for the fricatives—1 111 above) From the hustortcal stand- point the retroflex sounds are relatve late-comers into Indo-Aryan and they consequently occupy a peculiar place in the phonological system ,° they are thus iikely to have attracted attention even at a period when specialist phonetic analysts was unknown, and the terminology, like that of Latin in the west, 1s likely to have per- sisted into a period of more precise description In connexion with the role of the tongue in the retroflex series, the Ap 5 makes the remarkably acute observation that the contact 1s made not with the tp but ‘with the part next to the tip, or the under-side of the tip’ Functioning phonologically as a member of the retroflex series we have also the semivowel r, on the phonetic value of thus letter, however, widely diverse accounts are given,® ultimately depending perhaps on dialectal variation The same applies to the vowel f, which will therefore be most convemently considered in connexion with the semnivowel . The retroflex pronunciation of both semivowel and vowel 1s in * murdha §abdena voktra vivaropan: bhago tmvakgyate * Cf Max Mulleron RP 44 > "The PS (13) im a list of the sthanas, also uses the term Jeras in place of murdhan * Cf Pike, op eit, p 3123 * On APs 22 TPu 37 ® CF also Jakobson, Proc 3rd Int Cong Phon Se,p 40 7 1 6-7 jthuopagrena murdhanyandm ychvagradhahkaranam ta Cf J R Firth in Harley, op cit, p xx The Indian ¢ 1s not made with the up 101 the English manner but with the very edge or rim of the up, which 1s slightly curled back to make this possible’ See also BSOAS xu 859 "Cf Varma, op at, pp 654 PHONETICS IN ANCIENT INDIA fact preseribed by the PS,! but 1s exceptional elsewhere? The Pritisikhyas generally require an alveolar artrculation? (which agrees with the present pronunciation of Sanskrit and the general practice of the modern Indo-Aryan languages) The AP, VP, and RT refer to the alveolar position by the term danta mula, ‘toot(s) of the teeth’, a name which has beenemployed also by Sweet 5 A slight difficulty 1s caused by the fact that some of the treatises refer to the dental series by this same term , in such cases, however, the reference 1s to the yunction of the teeth with the gums (Sweet s ‘r1m”),6 and the alveolar position of r 1s then clearly distingmshed by a further description, e g For r contact 1s made by the centre of the tongue up behind (pratyak) the roots of the teeth? the word pratyak being further interpreted by the Tribhasyaratna as meaning ‘within and above’ ® Certain authorities quoted by the RP also refer to r as ‘vartsya’,° a hapax glossed by Uvata as denot- ing ‘the projection behind the roots of the teeth’,!°1e the alveolar arch Tt “ The prescription of alveolar articulation corresponds well with the name repha interpreted as ‘tearing sound’ (see 0 42 above), 10 that a rolled r such as this seems to smply could hardly be retro flex," excessive rolling however, ss hsted as a fault by the RP, and Uvata refers to this type of pronunciation as ‘indelicate’ * Two treatises, the RP and the RT, treat r as dental, but mention the alveolar pronunciation as an alternative 'S "397 syur murdhanya 7 fu ra sah * Ap 133 7 fu va $a murdhanyah but 14 ro danta mula sthanam ekesam 2 Cf APs 28 TPu 41 VPs 68 + AP 28 vephasya danta mulam VP. 68 ro danta mule RTS 5 Primer p 8 e ; Ibid (Sweet s termunology distinguishes thus mm from the edges ) TP u 41 rephe phvdgra madhy ena pratyag danta muléBhy ah © pratyag sty abhyantara upant bhaga tty arthah ° 3 46 repham vartsyam eke Q vartsa Sabdena danta mulad uparistad ucchunah pradesa ucy ate If a special term 13 required to translate vartsya gingival might be appro prate—cf Pike op ct p 122 alveolar arch (which might with more Justice be called the gingival one since the contact 13 made agamst the gum, not the bone } Bloomfield Language p 98 ? ‘The fricative nature of the retroflex ris clearly indicated by ste equivalence to [4] in the sandha “sarvaig+punalh = saryair gurzaib, & " xiv 26 atspario barbarata ca rephe ™ barbaratapy asaukumaryam eva * RP1 44-46 RT 7-8 (dante lah repho mule td}LETTERS ss ‘The disagreements on the pronunciation of r are duly noted by Uvata Some schools pronounce ¥ as a ‘cerebral , some as an alveolar * As regards the vowel f, an alveolar pronunciation 1s suggested by the 7P mn a passage which reads Infandry the tip of the tongue is approumated tothe bersas * The Tribhdsjaratna wnterprets the ‘barsvas’ as refernng to ‘the elevations behind the row of teeth’, which 1» remimscent of its comment on the semivowelr(sceabove) Other treatises, however, agree in allotting f to the velar class* this prescription 1s prob- lematic, and 1s apphed by at least one author also tof S The latter appears only in the single root hJp-, and it has been suggested that in such a phonetic context | 1s hkely to have been articulated with ‘dark’ resonance, and that it ss this which has caused it to be classed as velar6(ef on consonantall,2 04 below) But no such arguments apply to r, and indeed the Middle Indian developments pornt rather to a palatal resonance for both vowels 7 It is in any case difficult to believe that the Indians would have classified these sounds by their secondary rather than their primary articulations Tt 1s gust possible that in connexion with r we should interpret pioamiiliya as ‘uvular’ rather than ‘velar’, 1t 18 only strange thit we have no such deserption of the semrvow el r, except in so far as it 1s mentioned amongst a lst of alternatives by the Varga- patalam 8 In the retroflex series there remains only a peculanty connected * On RP1 1 kasyam éakhayam repho murd! anyah hasycim dantamulsy ats * 38 yihvagram phararkaralkdresa barstesupasamharaty 1 barseesu 1 danta pankter upanstad ucca pradesesu tty arthah 4 CE VP: 65 rhkauzind mule RT4 quod mule kr RP 3 4t $ RP1 41 ykaralkarav atha sas{i a usmad gthy amuliy dh prathamad ca cargah * Whitney on AP: 20 Teg kilitta tasya prayatna akarah ewytdsya prayatna stare svarah) Jura ror ahah savarne dirghah coament the brevity of the orginal might be preserved by a rendering such as ‘acs 331 2,00 Pan, Siw Su 1 (Kielho: kai = . nae ¢ m 1: 35) akdraya ruptopadela dkdra-LETTERS 59 heading, viz. as Ranthya, ‘glottal’—a term which has already been used in connexion with the voiced and voiceless’ ‘breathings’— ass glottal* a and h are glottal* a,h, and -h are formed at the glottis * To class the open vowels as ‘glottal’ appears at first sight an indefensible procedure. It becomes less so when we perceive the conceptual framework underlying these statements. Tt will be remembered that the TP referred to a ‘neutral’ position of the articulatory organs, in which the tongue 1s extended and depressed, and the lips are 1n the position for a4 The classification of a as glottal begins to make sense if we assume that it was viewed asa ‘neutral’ vowel in the sense of involving no special intra-buccal articulatory efforts Such an assumption 18 fully supported by a statement in the Mahabhasya The place of articulation of the a-vowels 1s extra-buccal or, as some would have it, 1t1s the whole mouth ® In other words a has no specific intra~-buccal sthana or karana; as with h and -h, it is a case of Aaranabhava (cf. p. 49,9 3)- From this recognition we may proceed to the peculiar doctrine mentioned by the RP, in itself inexplicable,’ that all the vowels are to be pronounced with the ‘articulatory condition’ (karanavastha) of a.8 This statement also, becomes phonetically meanmgful if a is interpreted as ‘vocalic neutrality’ or “unmodified vorce’, on which are superimposed the vowel-articulations involving various degrees of tongue-raising.? 1 RPs 38. Ranthyo *kdrah 2 PS xy Ranthydv a-hau i vp 1°71. a-ha-visarjantyah kanthe Cf comm on AP1 19 ee 1.113 5 CE Sievers, Gr d Lautphysiologie, p 38 ‘Bem ast der Mundcanal durch+ gehends mabig gedfinet; die Zunge entfernt sich nicht viel sus threr Tadaffer- tnzlage , eat 4, on Pan 1 1 9 (Kiclhom, 1 61) bahyam hy gsyat sthdnam avarnaiye* tarva-mukha-sthanam avaryam eka techants ce. Ap 230 : 7 Cf. Max Muller (on 823),* . sehr michtssagend 20 sein schemen ' Liv. 65 akaratya karandvatthaydnydn svardn braydt CF 66 4 © "The statement ie of course nonsensical f karanavasthd 1s simply meerp= as referring to the tongue-position,, but the use of the term avasthd (not thant), a word otherwise unattested mn the phonetic literatute, 18 some guarantee of the special nature of the reference Cf MI Wallaser, ZIDy 19360 PHONETICS IN ANCIENT INDIA We are now mm a position to understand a third problematic doctrine referred to the RP, viz Some say that the voice of the voiced consonants consists of a! Thus last staternent enables us to trace a consistent thread running through the senes of apparently eccentric aphorisms, and to relate them precisely to the descriptive framework of the other ‘glottal’ articulations (a) -h 1s considered as ‘pure breath’, lrable to modification by the close vowels,? and capable either of independent function (= uisaryantya) or of providing the appropriate air-stream for the voiceless consonants 3 (8) has considered as ‘breath-+ voice’,t hable to modification by the close vowels,? and capable either of mdependent func- tion (= hakara) or of providing the appropriate air-stream for the voiced aspirates 3 (c) 218 considered as ‘pure vorce’, lable to modification by the close vowels, and capable either of mdependent function (= evarna) or of providing the appropriate air-stream for the voiced consonants Artificial as such a descriptive basts may appear, it 1s in fact not so very remote from some statements of the most recent branch of phonetic analysis, ‘acoustic phonetics’; the following may be quoted for comparison We therefore discuss vowel production on the hypothesis that the glottis emits a spectrum that 1s independent of supra glottal articulation and that the filtering which determines the ultimate vowel spectrum 1s independent of the glottal adjustment that 19 the orginal production and the articulatory modification of the glottal tone are entirely indepen- dent of each other The spectrum of the vowel as it exist im th open air 38 to be reckoned, then, as the glottal spectrum multiplied for each frequency by the transmission percentage of the articulatory filter ° ‘Two thousand years and more before the sound-spectrograph, @ sound’ was not an unreasonable substitute for the fiction of a pure ‘glottal spectrum’ ¢ * xn 15 ahur ghosatp ghosqvatam akaram eke 2 Cf TP 47-48 (see 2 00 above) 4. Gf APs 1a RP au 1-6, TP 9-10 (see 1 20 above) ‘ Bee r 20 above *M Joos Acoustic Phonetics p 39 ‘or discussion of the concept of a. as the natural vowel or princepstocaltum” see Sievers, Phon ,§§ 197 fF ef also Jakobson in Trubetzkoy Prineipes p 376LETTERS 6x With regard to a xt remains only to mention that in later treatises, owing to the extension of the term kanthya (see 2 0x above), a (lke h and -h) 1s grouped with the velar series, thus adding con- siderably to the symmetry of the varna-samamnaya at the expense of phonetic precision 211 isu . The close front quality i 1s appropriately classified as ‘palatal’ (talavya), and the T'P says more specifically, For i-quality the middle of the tongue 1s approximated to the palate * The close back quality w 1s classified by the lip- rather than the tongue-position, viz as ‘labial’ (osfhya);* the shape of the lips 1s variously referred to as ‘approximated’, 1¢ rounded,$ or as ‘long’, 1e protruded © 212 Fy] As to the pronunciation of the vocalic F and J, the ancient state- ments are perhaps not as clear as we could wish, but theit general trend 1s easily followed Their places of articulation have already been discussed above In distinction from the other vowels they are referred to as ‘mixed’,? 1¢ combing features of vowel and Brondal Proc jrd Int Cong Phon Se pp 49 TCLP ws 6zff In the Anthropos Lautschrift a 1s the vocalts indifferens (see M Heepe Lautzerchen P 6) CE further Millet s observations on the role of the pharyngal resonator and its tmbre—L Articulation des vayelles p 3 Nous considérons la bouche comme le générateur efficace Je résonateur déterminant de la résonance vocalique 1a cavité pharyngienne donne son timbre & Ja vor seulement qui enveloppe celur de la voyelle and Etude expérimentale de la formation des voyelles p 68" Il nest pas de timbre de yoyelle qui ne sort accompagnée du tunbre de la vore' Allowing for the fact that the Indians seem not to have differentiated pharynx and larynx we may say that their conception of a was im Millet s termunology that ofa pharyngal timbre without buccal determunation We may here note that a similar device was adopted in India with reference to nasality anwsvara being treated as pure nasality, forming the basis of all nasal sounds (RP xut 15 anusvdram anunastkanam) TEg Siddh Kaum 10 (cf 201 above) Some suthonties even gave to it the title jzhuya which means spetifically velar (see Ap S 1 10) 2 PS iy s-cu ya fas talavydh Cf Sweet Primer $33 un 22 tdlau yshua madhyam twarne ; CE PS 17 osthaydv w pu ‘ TP u 24 osthopasamhara uvarye VS 284 wuvarna prakyter osthau dirghau TVS ix62 PHONETICS IN ANCIENT INDIA consonant {r/l),? for this reason some writers even refused to admut them to the vowel-system 7 There 1s general agreement that their phonetic structureisof thetype consonantal element—vocakic element—consonantal element, f¥ contains r (as also does the first half of rr) and the rss in the muddle 3 The characteristic of T 1s that it 1s compounded of four segments, of these the first and Jast are vocalic, whilst the central parr are consonantal viz. particles of r* In this connexion tt 1s of interest to compare such Avestan parallels as porafz beside Skt prthu, karap- beside k]p~ 5 Asto the quality of the vocaltc element, the VP states that p and ] consist of r and blended into one unt with the vowel a & Regarding the method of combining the vocalic and consonantal elements we may note as an example of picturesqueness rather than illumination the statements quoted by the commentary on the AP, which declare that they are connected like a nail on the finger, or a pearl on a string, or a worm in the grass 7 . 213 ¢,0; ai, au It will be convenient to constder in conjunction the guna and vrddhe* vowels efo and aijau. The latter, as the transcmption sug- gests, are diphthongs and are regularly so described; the former also were historically diphthongal? and continued to function as such for certain phonological purposes (eg vijnaaya--idam = vijnaayedam) But there are indications that whereas the phono- "Cf AP 1 37 39 samsprspa repham poarnam salakaram Jvarnam Ap S$ 1 26 sarepha pvarnah 7 Cf Katyyataon Mbh 11 4 on Pan 119 ange te ssat sprita karanatvad anayor rkara [harayos ca vitztatuat tabhyam tayor agrakanad anacteam ahuh 2 RP xn 34 repho 'sty pare ca parasya cardhe purce madhye sah (cf AP 1 38 dirgha plutayoh purva matrdy * Sarvasammata Stksa 19 tharasya x arupart ht shspart pada eatusfayam padesu tesu vyneyav adav ante svaratmakan ; anu rephasya madhye tu vyneyau vyaiyandtmakau Cf also Pkt Avitta < kipta, and Oldenburg ZDMG ly: 8435 Saw 348 7] varne repha lakarau samfhstav a fruts dhardv eka yarnau For the quality of the vowel in Middle Indian developments see 2.03 above , On AP 1 37 syathaagulya nakham tatha sutre mame wety eke tne krimr auett ca 8 See o 41 above ° CE GE ola, Av vaeda beside Skt veda, &cLETTERS 63 logical value of efo was a-+-i/u, that of aifau was once aa-+i/u- ths distinction may be illustrated by junctions of the type nagare-+ha = nagara iha beside striyait-uktam = striyaa uktam.' From the phonetic standpoint e/o are represented at a still com- paratively early period by simple long vowels mtermediate in qualty between aa and u/uu. To consider now the ancient descriptions the term for the diph- thongs (including e/o) 1s samdhy-aksara, ‘compound vowel’, in contrast to samdndksara, ‘simple vowel’ ai and au are designated respectively ‘glotto-palatal’ and ‘glotto-labral’—as the VP says, Inaiand au the first mora1s glottal and the second palatal or Jabal ,* both the AP# and the VPS point out, however, that Although diphthongs are combinations of vowels they aze treated as single letters As regards e and 0, the PS seems to preserve the tradition of a diphthongal pronunciation (distinct from that of ai and au) Ine and 0 the glottal element has a length of 4 mora and in al and au 1mora,é the passage continues with the words ‘tayor anvpta-samortam’, literally ‘in them there 1s openness and closeness’, which Ghosh interprets as referring to the fact that, 1n ai and au, “thew first half or the a element 1s open and the second half or s- and u-element 1s close’, but at 1s more probable that the words refer to the open a4 which forms the first element of aifau and the closer a which forms the first element of the narrower diphthongs e/o7 The * Note however, that the attested sandh: of both a and aa+tfaas efo (eg baalaa4tikgate — baaleksate) 2 PS18 ¢ astukantha talavyao aukanphosthajau smptax Cf Ap S 1 12-13 21093 akarauburayoh Ranthya pun.a matra talv osthayor uttora CE RP xan, 38-39 samdhyam samdhy akgarany dhur eke des sthanatasteyu tatho- Bhayeru samdhyeso akaro 'rdham tkara uttaram yuor ukara 11 dakapayanah $1 40. sandhy-akparam samsprita tarnany eka varnavad vrtth aw rag * Ghosh Reconstructed text 13 ardha matra tu kanthyasya ekaraukarayor Bhavet athdraukarayor matra fh 7 For the distinction of the two vaneties of diphthong one may compare the Nepali falling type (alfau) where the first element 1s conderably the mote Prominent and the narrower ay/aw ({[2e]/[a0]) with closer and relatively less Prominent starting point—e g bhalle beside mayle ‘There 1s also somets PHONETICS IN ANCIENT INDIA latter interpretation is supported by a passage in the Mfakabadrya, where the semryta a of efo is specifically contrasted with the more open a2 of alfau.! The monophthongal pronunciation of efo seems to be indicated by the RP when it says that they are not, like ai/au, heard as 2 distinct sequence, because of the coalescence (samsarga) of their parts? as Uvata goes on to explain, One does not observe where the 9 ends and the f or u begins, because the two coalesce like mulk and water,? a type of combination which Karyy ata, continuing the traditional sumule, contrasts with the mixture of sand and water ¢ There as ttle in the way of detailed desenptions of the monoph- thongal articulation of efo. The 7P, however, mentions that fore the lips are more spread and for o more rounded than in the case of 233 and the intermedhate degree of closure for e (between 2 and {I} ss stressed by the Tnibhdsy aratna: Ine the raising of the middle of the tongue towards the palate is Jest than in the case of f, owing to the fact that the former tt mixed with a.* Tt will be noted that even where the monophthongal value of efo ts phonetically established, the feeling for its phonological equiva- lence to o+-S/u still prevails, and the basis of description is still provided by the simple framework i u alternahon of the second type with « monophthongd realusticn b—e g dhere, ‘much, anys very’, bende sabay, ‘all’, where the final «@ and + ey are morpho: lopcally comparible . "hi g,0n Pin t Lo (Rother. | 63) praliytdernte of” *i enytas Earde arms ston (oe, att} af 8 xn 49 peird-somaeedd avere ‘pritah-fnatl t 1 ¥ geare parte ¢ 6 sly ete mSird-tamiavgdt: mitravek . somuisret ma sttyate kedrany-mid bro + On aft bg. (parmidtoren), Su rgerg ehdvecs epthaw tipasanket * Ong. 2) farar paths ple d-malivegasondbive na seen ences fev avthah> Autah> * etd hfe meePART III PROSODIES 30 Definition In the technique of letter-abstraction various features of the larger units of utterance are left unaccounted for It 1s the reintegration of these features that forms one of the tasks of synthesis, and it 1s to them that the title of ‘prosody’ 1s here applied! The ancient accounts of these prosodic features will be considered under the following headings 1, Features of punction (sandhi) * 2 Features of syilable-structure 3.1. Function The nature of our material makes it convement to work with the following sub-divisions (a) Word- and morpheme-junction (5) Letter-junction The treatment of word-junction and morpheme-yunction under the same heading 1s sustified by the close parallelism of the two classes of prosodies in Sanskrit,3 as also by the stated principles of our treatises, ¢ g Unless directed to the contrary, one should treat the parts of a word as words + Morphological analysis must observe the same mules of finality 23 apply to word isolstes * In both (a) and (4) certain of the prosodic features are relatable to the basze processes considered in Part 1, this 1s only to be expected in view of the fact that these processes had been arbitrarily seg- mented by the letter-analysis, and have to be restored in the syn- thests here considered 7 See further J R. Firth Sounds and Prosocies’, TPS 194% pp 127 7 Cf VP wm 3 padanta padadyoh sandhth 3 For divergences cf Whitney, Skt Gr, § 109, Thumb, Hdd des Sanskrit, § 268 4 RP1.61 apratyanmaye padatac ea padsan t 5 VP 1 1§3 avagrahah padantavat Bes08 F64 PHONETICS IN ANCIENT INDIA latter interpretation 1s supported by a passage in the Mah&bhasya, where the samvuyta a of efo 1s specifically contrasted with the more open aa of ai/au * The monophthongal pronunciation of efo seems to be indicated by the RP when it says that they are not, like aifau, heard as a distinct sequence, because of the coalescence (samsarga) of their parts * as Uvata goes on to explain, One does not observe where the a ends and the i or u begins because the two coalesce like mulk and water,3 a type of combination which Karyyata, continuing the trad:tional simile, contrasts with the mixture of sand and water ¢ ‘There 1s little in the way of detailed deserptions of the monoph- thongal articulation ofefo The TP, however, mentions that for e the lips are more spread and for o more rounded than in the case of a § and the mtermeduate degree of closure for e (between a and 4) 18 stressed by the Trsbhasy aratna In e the raising of the middle of the tongue towards the palate 18 less than in the case of , owing to the fact that the former 1s mixed with a © It will be noted that even where the monophthongat value of e/o 1s phonetically established, the feeling for its phonological equiva- lence to a-+-i/u still prevails, and the basis of descmption ss still Provided by the simple framework i u a « alternation of the second type with a monophthongal realization e—e g chere, much many, very’, beside Sabay, ‘al] where the final ¢ and -2y are morpho- lomeally comparable Tu 4 on Pan tt 9(Kielhom 1 62) pratlista efautse ef} tavzta fardvarndo etau (se, arc) ) prallutacarnde ‘ > ee * xu 40 mdtrd semsargad avare *prthak fruts ) avare parce € 0 ity ete mdtrd samsargdt métrayoh samayoh Rprodakatat somrargat na jndyate didvarna mdtrd kea vee arnoearnayor tte * On Moh t 1 4 (pamnedckarat) : 41 13-15 okdre ca osthau tuparamhptatarau tyat prakpstdv ekdre Onu 23 sarge yathd phea madhyopasamhara na khalv evam ekare kim tu tato myuna tty arthah kutah akdra mifntatedd ehdraryaPART III PROSODIES 30 Definition In the technique of letter abstraction various features of the larger units of utterance are left unaccounted for Its the reintegration of these features that forms one of the tasks of synthesis and it is to them that the title of ‘prosody’ is here applied! The ancient accounts of these prosodic features will be considered under the following headings 1 Features of yunetion (sandhi) 2 2 Teatures of syllable structure 31 Junction The nature of our material makes 1t convenient to work with the following sub divisions (a) Word and morpheme junction (8) Letter yunction The treatment of word junction and morpheme junction under the same heading 1s justified by the close parallehsm of the two classes of prosodies in Gansknit 3 as also by the stated principles of our treatises ¢g Unless directed to the contrary one should treat the parts of a word as words + Morphological analysis must observe the same rules of finality as apply to word ssolgtes 5 In both (a) and (5) certain of the prosodic features are relatable to the basic processes considered in Part I, this ts only to be expected in view of the fact that these processes had been arbitrarily seg mented by the letter analysis and have to be restored in the syn thesis here considered ' See further J R Frth Sounds and Prosodes TPS 1948 pp 127 *Cf VPin 4 paddnta padadyoh sandia? 5 oo divergences ef Whimey Ske Gr § 109 Thumb Hdb des Sonskr t I * RP: 61 apratydmnaye padavac ca padyan * VP1 153 avagrahah padantavat Bates F