Buddhist Hybrid Sanskrit Ali, Kali as Grammatical Terms in Tibet

  • Published on

  • View

  • Download


  • Harvard-Yenching Institute

    Buddhist Hybrid Sanskrit li, Kli as Grammatical Terms in TibetAuthor(s): Roy Andrew MillerSource: Harvard Journal of Asiatic Studies, Vol. 26 (1966), pp. 125-147Published by: Harvard-Yenching InstituteStable URL: http://www.jstor.org/stable/2718462 .Accessed: 01/10/2014 17:31

    Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

    .JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact support@jstor.org.


    Harvard-Yenching Institute is collaborating with JSTOR to digitize, preserve and extend access to HarvardJournal of Asiatic Studies.


    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions




    ; E Luil-ston-pa rtsa-ba sum-cu-pa (hereafter: SCP) = Vye2yka- rana-mula-trimsat, the first of the two early Tibetan gram- matical treatises attributed to Thon-mi Sambhota,1 begins its

    description in sloka 1 :1-3 with the introduction of two untranslated Indic grammatical terms:

    yi ge 'ah li kah li giiis/ 'ah 1i gsal byed 'i sogs bzi / kah 1i sum cu tham paho / "The letters [of the Tibetan script consist of] two [varieties], 2li

    and kli; [if we] make clear the li [they consist of] four, 'i, etc., [and] the kli are thirty." At first glance the passage does not appear to be a particularly

    difficult one, but closer inspection soon shows it to be more carefully put together and more demanding in its interpretation than it might at first appear. These difficulties are most apparent in any attempt at

    1 Texts, translations, and commentaries in Jacques Bacot, Les slokas grammaticaux de Thonmi Sambhota, avec leurs commentaires, traduits du tibetain et annotes [= Annales du Musee Guimet, Bibliotheque d'etudes, tome 37] (Paris, 1928); Johannes Schubert, "Tibetische Nationalgrammatik, i. Teil: Das Sum-cu-pa und Rtags-kyi-'ajug-pa des Lama Dbyanis-can-grub-pai-rod-rje, Ein Kommentar zu den gleichnamigen Schriften Thon-mi Sam-bho-ta's. Ubersetzt und erkliirt," MSOS 31(1928).1-59; idem, tI. Teil," MSOS 32(1929) .1-54; Johannes Schubert, Tibetische Nationalgrammatik, Das Sum cu fia und rtags kyi 'ajug fia des Grosslamas von Peking Rol fiai rdo rje, Ein Kommentar zu den gleichnamigen Schriften Thon. mi Sambhota's auf Grund der Erk1drung des Lamas Chos skyon- bzaii -Po,Lo. tsa ba vonZha -lu [= Artibus Asiae, Supplementum Primum] (Leipzig, 1937); Inaba Shoju ri ft Chibetto-go koten bunpogaku 1- -Z '!1

    M (Kyoto, 1954), for which cf. the review by G. Morichini, East and West 6(1955). 172-175, and that by the author in Language 31(1955).477-485, as well as Nils Simons-

    son, Indo-tibetische Studien (Uppsala, 1957), p. 225, n. 2.


    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    rendering the second line of the sloka; the version above is simply an English equivalent which follows closely Inaba Shoju's Japanese ren- dering, shion wo [akiraka ni sureba] sanjui de aru, with brackets as in his translation.2 Here the Japanese scholar has taken 'ah li in the second line of the sloka as the direct object of gsal byed, which he treats as a subordinate verbal element in the sentence; here as in other instances it is more than worthwhile to inspect Inaba's meticulous renderings of the SCP slokas with some care, since even when it is eventually necessary to differ with his interpretation, he still always provides a solid basis for the discussion. Bacot, for example, in his translation of the sloka text has simply: "I1 y a deux sortes de lettres, les voyelles et les consonnes. / Les voyelles sont quatre, i, u, e, o. / Les consonnes sont trente,"3 which gives the over-all sense in a gen- eral way but completely obscures the details of the sloka's statement.

    The difficulties of the sloka cluster around two main problems: (1) the identification and origin of the Indic grammatical terms written in the Tibetan text as 'ah li and kaht li, i.e. the terms ali and kali. Chandra Das' Introduction to the Grammar of the Tibetan language. .. (Darjeeling, 1915), page 1, and the usual bilingual Tibetan lexical sources of the West following him, simply refer them to "Sanskrit ali, kali," but these are terms which are wholly unknown to the standard Sanskrit lexical tools, surely a surprising omission if they are indeed Sanskrit, and unknown to the Sanskrit grammarians as well. Most surprisingly (and significantly) they are unknown as grammatical terms to the Mahizvyutpatti (hereafter: MVP) ,4 which registers only ali as yur-phran, "a small ditch, water-course, conduit," with the later and slightly misleading Chinese gloss hsiao ch'i 4'I9 "small rivulet, creek" (No. 4177). They are also unknown to the Tibetan-Sanskrit dictionary of Tse-rifi dbafi-rgyal,5 who does however register a full repertory

    2Inaba, oP. cit., p. 325: JoS [RF 6; tb 11_ =] ~+Z C ) 7. Perhaps he meant to write "shion wo akiraka [ni sureba], etc."; this would more closely agree with the Tibetan.

    3 Bacot, oP. cit., p. 76. 4 Sakaki Ryosaburo I Honyaku myogi daishui ' [ = Kyoto

    teikoku daigaku bunka daigaku sosho Dai san -] (Kyoto, 1916); idem, [vol. ii, Sanskrit index] [= idem, Dai sanRfukan (Kyoto, 1925); Nishio Kyao , Honyaku myogi daishuF Chibettogo sakuin g,, p i [= Butten kenkyu, dai ichi 1 V- (Kyoto, 1936).

    5 Tse-ring-ouang-gyal (Che rifi dban rgyal), Dictionnaire tibetain-sanscrit, reproduction

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    of all possible equivalents for Tibetan gsal byed, including the most common vyanjana of the grammarians and Panini's sivasuitra abbrevi- ation hal (more properly, hal) 6 for "consonant." And the MVP knows gsal byed as a translation of vyanijana (No. 2013). Tse-rifi dbafl-rgyal also registers many equivalents for dbyaiis, including the grammarians' most usual svara, the less common aksara, and Panini's abbreviation ac, i.e. ac for "vowel"; the MVP is familiar with the dbyanis = svara equivalence (No. 248, among numerous instances), but again ignorant of zli in this connection. All of this confronts us with the other chief problem presented by the SCP passage as cited, (2) why does the second line of the sloka appear to equate, or at the very least bring together into some sort of relationship, the term cli, concerning which those sources which know it all agree on the fact that it has reference to "vowels," and gsal byed, which all sources agree has reference to "consonants"? It should also be possible, in the course of considering these two problems, to arrive at a more or less satisfactory interpretation of sloka 1 of the SCP, especially of this second line. The first step in approaching both problems lies in an inspection of what the more important Tibetan commentators on the grammatical literature have said about the passage.

    The so-called Za-ma tog, actually the Bod kyi brdahi bstan-bcos legs- Par bsad-pa rin-po chehi za-ma tog bkod-pa of Dharmapalabhadra (1441-1528), does not appear, at least in those portions of the text available to us in Berthold Laufer's "Studien zur Sprachwissenschaft der Tibeter. Zamatog,"7 to use the terms ali and kali; it has instead

    Phototypique publie'e par J. Bacot [ = Buddhica, Documents et travaux pour I'etude du bouddhisme, deuxieme serie: Documents-tome II] (Paris, 1930). For the preface to this work, see now J. Bacot, "Titres et colophons d'ouvrages non canoniques tibetains. Textes et traduction," BEFEO 44(1954).275-337.

    6 Betty Shefts, Grammatical Method in Panini: His Treatment of Sanskrit Present Stems [= American Oriental Series, Essay i ] (New Haven, 1961), p. 7. The Paninian siivasiutra enumeration of sounds was known to the Tibetan grammarians in, for example, the translation by Jetakarna and Ni-ma rgyal-mchan dpal-bzafi-po of the Candra-vyakarana- sultra, Tanjur 116, no. 3604, where following Candragomin the fifth and sixth of the 14 sivasu7tras are put together into a single one, hayavaralan; cf. Bruno Liebich, "Das Candra-Vyakarana," NAachrichten von der Konigl. Gesellschaft der Wissenschaften zu Got- tingen. Philologisch-historische Klasse, 1895, pp. 280-281.

    7 In Sitzungsberichte der Philosoplhisch-philologischen und der historischen Klasse der k. b. Akademie der Wissenschaften zu Miinchen, Jahrgang 1898, Erster Band, pp. 519-594.

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    the simple statement dbyaiis yig 'i 'u 'e 'o bzi / gsal byed ka sogs sum cu ... /, "die vier Vokalbuchstaben i, u, e, o und die dreissig Kon- sonanten k u.s.w." in Laufer's translation of the passage, which is in either the conclusion of part 7 or the beginning of part 8 of the work.8 In addition, the Za-ma tog employs the expression gsal byed tha ma at least twice; one instance is in the epitome of parts 1 through 6 of the work, in the following passage: re rehii gsal byed tha ma danl / 'i 'u 'e 'o ya ra la / wa yig rjes hjug bcu yis brgyan; this Laufer rendered "die einzelnen jeglich sind mit dem letzten Konsonanten, / mit i, u, e, o, y, r, 1, / w und den zehn Suffixbuchstaben geschmiicket," and added "unter dem letzten Konsonanten ist 'a yig go der Buchstabe 'a zu verstehen."9 The other instance is in a gloss inserted between lines 9 and 1o of the article from the Za-ma tog, part 1, on words in k-: ka yig la snion hjug mgo gsum med pa gsal byed tha ma daii dbyaiis bzi dani ya ra la wa danz rjes hjug gis phye bahi brda ste," which Laufer ren- dered, "der Buchstabe k ist ein Zeichen, das erlautert (weiter ausge- fiuhrt) wird durch den letzten Konsonanten ('a), die vier Vokale, y, r, 1, w und die Suffixbuchstaben; hier aber, wo es sich lediglich um den Buchstaben k handelt, kommen die drei Kopfprafixe (r, 1, s) nicht in betracht (da sie spater unter den Rubriken rk, 1k, sk abgehandelt werden) ."'0 In both these instances it is clear from the context and from the over-all meaning of the statements that gsal byed tha ma, though it appears literally to mean, as Laufer translated it, "the last consonant [i.e. of the Tibetan alphabet when given in its usual order, hence] 'a" and although the gloss to the passage in the epitome seems to state this explicitly, cannot be so understood here. The graph 'a is not used in the Tibetan script in any way that such an interpretation of these texts would imply, and hence the meaning is different. This is in fact the type of statement that Dharmapalabhadra employs in order to express the vowel a; for the same technique of statement in the translation of a Sanskrit grammatical work, see in the Tibetan transla- tion of the Cendrauna:di-vrtti, Tanjur No. 3726, by Thugs-rje dpal- bzaii-po = Krpa4ribhadra, rd/dha sdzdha yanz dag par grub pa laho / da yig dha yig dag gi 'a yig klog pahi don no /, "der Buchstabe a in da

    'Text in ibid., p. 546; translation, p. 548. 9 lbid., pp. 539, 540. 10 Ibid., PP. 574-575.

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    und dha steht der Aussprache wegen.""l The Za-ma tog epitome is stating here, in effect, that various ones from among the simple graphs, and hence the simple phonemes, of Tibetan may have added to them the vowels a (gsal byed tha ma), i, u, e, and o, or -y-, -r-, -1- or -w-, or one of the ten final consonants; the second Za-ma tog passage too yields to this interpretation. The difficulty arose because of the double use of the last graph of the writing system as both a consonant and a vowel symbol, and Dharmapalabhadra is here actually making the statement unambiguous by using gsal byed tha ma for a as a vowel only.

    The grammatical views of Dharmapalabhadra are also available through the Sum-cu-pa dan rtags-hjug-gi don inuin-inur bsad -pa blo-ldan dgah bskyed, which Rol-pahi rdo-rje based on Dharmapalabhadra's school teachings,12 but this text has only a list of the four vowels 'i 'u 'e 'o which it equates with ali = dbyafis, and the thirty consonants which it equates with kali = gsal byed.13

    The great commentary of the Mahapandita of Si-tu, completed in 174414 (ed. Sarat Chandra Das, op. cit., p. 7, 1. 14), glosses ali as 'a-phren and kali as ka-phreni, in the sense of "the 'a (and k) rows, lines," and equates these with gsal (scil. byed) and dbyanis respectively; the Mahapandita's long note on the second line of SCP ?loka 1 is best approached through its epitome in the Si-tu hi Zal-lun of Dharma- bhadra,15 who draws upon the Mahapandita in writing as follows:

    spyir yi ge la 'ah li ste 'a la sogs pahi 4ph rein ba dbyafisyig danz / kah li ste ka la sogs pahi 4phrein ba gsal byed kyiyi ge giiis suyod la / bye brag tu bod yig la ni legs sbyar gyi 'ah lihi bya ba gsal byar mtshon par byed

    11 See Bruno Liebich, op. cit. (in n. 6 above), pp. 278, 300, 301. 12 Inaba, Op. cit., p. 35. 13 J. Schubert, Tibetische Nationalgrammatik ... , pp. 29, 43, 44. 14 See my paper "The Si-tu Mahapandita on Tibetan Phonology," in iaaM Vl

    i:tgn-B?n9;t [A Collection of Papers Commemorating Dr. Hachiro Yuasa's Seventieth Anniversary] (Tokyo, 1962), pp. 921-933.

    15 Dharmabhadra has unfortunately had a bad press in the West, beginning with Bacot's remark that the somewhat inferior text which he translated "a semble au dge s'es Don grub l'emporter sur le merite d'un autre abrege de la grammaire de Situ, celui de Dharma bhadra" (op. cit., p. i); but Inaba's Bunpigaku, with its meticulous edition of the text and carefully annotated translation, even though unfortunately only covering those portions of Dharmabhadra which are on the SCP, now restores this commentary to its rightful place.

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    pahiyi ge 'i 'u 'e 'o bzi dan / kah li mtshon par byed pa / ka kha ga iia / ca cha ja na / ta tha da na / pa pha ba ma / tsa tsha dza wa / za za ha ya / ra la sa sa / ha 'a / ste sum cu tham payod do /.16

    "eIn general, there are two [kinds of] letters, ali, the vowels (dbyafts) of the series 'a etc., and kali, the consonant letters (gsal byed kyiyi ge) of the series ka, etc.; in detail, in Tibetan letters, the letters which il- lustrate clearly the operation of the Sanskrit ali are the four 'i 'u 'e 'o, and those which illustrate kali are the thirty ... [as above in text] ...."

    Dharmabhadra's choice of language here from bye brag tu on is also determined in part by his wish to cite within his commentary as much of the original ?loka as possible, a common Tibetan commentator's technique, thus,17

    SCP ?loka 1:2 Dharmabhadra 'ah li gsal byed 'i sogs bzi 'ah lihi byed ba gsal bar mtshon

    par byed paki yi ge 'i 'u 'e 'o bzi... The Sum-rtag gzun-hchan legs bgad nor-buhi phreii-ba, by an

    anonymous pupil of Mkhas-grub dam-pa, uses the phreii terminology, and again employs the original SCP ?loka terminology as much as possible in presenting its commentary as an expanded paraphrase of the original text: yi ge 'ah li ni 'a phrein dain kah li ni kah phre7 giiis su dbyeho / 'ah li ste dbyains rnams gsal bar byed pa (hi) 'i sogs dbyains bzi daii / kah li ni kah phreni der sum cu tham payod paho /.18 Bacot trans- lated this, "les lettres se divisent en deux classes: les ali ou serie a et les kali ou s6rie ka. Dans la serie a, on distingue quatre voyelles, i et les autres (e, u, o). La serie ka compte trente consonnes."19 This ver- sion of the first and last lines of the text as cited cannot be questioned, but once again the second line, deriving from SCP ?loka 1:2, is diffi- cult. The commentator's text seems to have an extra dbyanis in the second line, and the (hi) of the citation is Bacot's emendation of his text; gsal bar byed pa (hi) appears to have become simply Ton dis-

    16 Text in Inaba, op. cit., pp. 8-9; translation ibid., pp. 324-325; now also available (but not to Inaba at the time of the preparation of his edition) in the text of Dharmabha- dra's commentary in Sum-rtsa-ba dahi dehi hgrel-pa Si-tuhi -zal-lwh, Chinese binder's title: Hsi-tsang wen-fa ssu-chung ho-pien Si ._


    tingue" in his translation, so that the dbyanis rnams immediately before this expression has completely disappeared. A more precise version of the second line here might be "the ali [series consists of] 'i and the other four vowels that distinguish the vowels, and ...." The text, redundant in the original, gains little or nothing in this redundant translation, and we are still left with a description in which the vowels are said to be carrying out some operation upon, and hence are in a sense being defined by, the vowels. Another possible version which would at least reduce the redundancy would be to break the line after dbyanzs rnams, which would result in "the ali [series consists of] the vowels which are 'i and the other four vowels which distinguish"; as we shall see below, this is not far from what eventually appears to be the most satisfactory version for the original SCP sloka 1 :2 passage.

    The SCP commentary by Dbyafis-can grub-pahi rdo-rje avoids the terms ali and kali and treats the subject succinctly in three lines of verse: dbyains kyi bya ba gsal po ru / byed pa 'i 'u 'e 'o bz1i / gsal byed ka sogs sum cu yin /; this Schubert translated "das Werk der Vokale (dbyanis) deutlich / Gemacht, ergibt vier: i, u, e, o. / Konsonanten (gsal-byed) gibt es dreiBig: k(a) usw./."20 Here byed pa is to be taken as kara; the equivalence is common in the MVP, as for example Nos. 1760, 1761, 2274, which citations however appear to have been omitted from the Tibetan index to the Sakaki edition by Nishio Kyao fiJ9A4 (Kyoto, 1936), page 204; cf. also Tse-rin dbyan-rgyal, op. cit., page 177r, which registers byed pa = kara. In Indic grammatical practice kara is used as a means for forming names of phonemes: "un phoneme suivi da kara est le nom dudit phoneme (et, avec a interpose, d'une consonne, . . . ex. ka-kara); . . . le nirdesa des phonemes [kmention indicative))] a lieu au moyen de kara."'21 The term bya ba in this text is to be taken as karana, cf. MVP Nos. 6624, 6892; karana is

    20 J. Schubert, MSOS 31(1928).30. This text is also now available in the Peking, 1956 volume cited in note 16 above, p. 1o6. Note that the reading cu for the MSOS text's bcu is from this Peking text. Translation, ibid., p. 41.

    21 Louis Renou, Terminologie grammaticale du sanskrit, troisieme partie [= Biblio- theque de l'eole des hautes etudes, Sciences historiques et philologiques, Deux cent quatre-vingt-deuxieme fascicule] (Paris, 1942), pp.52-53, 85. (The same work is also to be found as pp. [353]-541 in the identically titled one-volume version of Renou's Ter- minologie published in Paris, n.d. [1957], by the Librairie ancienne Honore Champion, which to add further to the confusion identifies itself on the spine not as the Terminologie but as the Terminaison grammaticale du sanskrit!)

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    ecmode ou organe producteur du son" .... tdefini yad upasa,mharati [


    "The translation equivalent (skad dod) for yi ge ["letter; phoneme"] is aksara ["the unvarying" mi hgyur ba], since by whomever the realization (rain bzin) of the form (or, "shape," akara) of the sounds (gdafis) of the letters is pronounced, the sounds of the voice (sgrahi gdanls) do not significantly vary (gian du mi hgyur ba), and hence they are yi ge [i.e. aksara]. Now, with reference to the language of the gods, there are two varieties, ali, the sixteen vowels (dbyanis), and kali, the thirty-three or thirty-four consonants (gsal byed).

    "eFrom among the sixteen ali vowels, in the Tibetan language, the vowel letters (dbyan-syig) which make clear the operation (bya ba = karana) of the vowels and which are grouped together under their closest head elements (iie bar mgo bahiyan lag), are the four, 'i, etc.; but since the sound of the Tibetan letters has length (rin cha can) the pronunciation of the two, [Sanskrit] i and a, is put together into one [Tibetan] 'i, so that the short and the long are similar or the same, [which is indicated in the Tibetan script by] the superscript i-graph (gug gu); and the pronunciation of the two, [Sanskrit] u and ui, are put together into one, [Tibetan] 'u, the subscript u-graph (iabs kyu); and the four, [Sanskrit] r, r, I and 1, and the two [Sanskrit] anusvara and visarga, together these six, are discarded, since in the simple (i.e. non-compounded) letters of the Tibetan language there are none under which these would align. The two [Sanskrit] e and ai are put together into one, [Tibetan] 'e, the superscript e-graph (hgrefn bu), and the two [Sanskrit] o and au are put together into one, [Tibetan] 'o, the superscript o-graph (na ro); and [all the above] is spelled by these symbols (brda chad), so that hence in the simple vowel letters of Tibet, 'i, 'u, 'e and 'o, there are four, consisting of the superscript i-graph and the other symbols.

    "Now the meaning of what in the SCP is expressed as 'ah li gsal byed 'i sogs bzi is that since the 'a letter, the last of the thirty kali, has its own sound in pronunciation, it should be put together with the variety of sound which it resembles, along with ka, etc.; 'i, etc. exist in order to make clear the distinctions in significance in pronunciation."

    On kah li sum cu tham pahi tha ma 'ayig, see above under the dis- cussion of gsal byed tha ma in the Za-ma tog.

    In effect, what Blo-bzafi tshul-khrims is doing here is attempting to derive Tibetan, as a language, from Sanskrit, the "language of the gods" (Ihahi skad; perhaps he intended this for devanagari?), and to

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    explain the Tibetan vowel system and script at once as the leavings over and above the richer Indic system. The equivalence aksara = mi hgyur ba which he suggests appears to be original with him; the MVP consistently hasyi ge = aksara, as for example No. 2014, and in the MVP the equivalent of mi hgyur ba is acanicala, No. 501. The commentator's remarks on aksara are based on the generally accepted Indic views on the word: "les a? sont ainsi appeles parce qu'ils ne s'ecoulent pas (na ksaranti): s'ecouler, c'est se mouvoir en tant que faisant partie integrante d'autre chose."23 For rain bin = prakrti, cf. MVP No. 945. etc.; "plus gen. Po se dit du sabda comme ((base)) des phonemes."24 On sgra, cf. Nils Simonsson, Indo-tibetische Studien, Die Methoden der tibetischen Ubersetzer, untersucht im Hinblick auf die Bedeutung ihrer (6bersetzungen fur die Sanskrit philologie, I (Uppsala, 1957), who translates it as "Laut [komplex]" (p. 250) and "lautlich Gestalt" (p. 265). On the ordering of anusvara and visarga in the text, cf. J. Kirste, "Die alphabetische Einordnung von Anusvara und Visarga," Sitzungsb. phil.-hist. Kiasse kgl. Akad. Wissens. (Wien) 133:8 (1895), 1 ff.; the numbers in the text for Sanskrit are for the graphs, not the sounds, on which see ibid., page 2, and H. Jacobi, "Kultur-, Sprach- und Literarhistorisches aus dem Kautiliya," Sitzungsb. kg. preus. Akad. Wissens. (Berlin) 1911. 965 ff. The term rin cha is rare, but is registered in Chos-kyi-grags-pa, Brtsams-pahi brda-dag mini-tshig gsal-ba (Peking, 1957), with Chinese glosses by Fa Tsun aft, Chang K'o-ch'iang C-AB, etc., page 831, where it is defined in Tibetan as sgra dbyanis kyi cha rin Po lahfb, and in Chinese as yin ch'ang-che (ju ch'ang a-tzu yin *AK. (flS1-), ["a long sound, (like the sound a) "].

    Blo-bzafi tshul-khrims ends his extended comments on the ?loka with a passage so deeply rooted in his own basic confusion between language and script, which he shares with many of the later com- mentators, as to be almost untranslatable. The difficulty here is that the final letter of the Tibetan script is both a vowel symbol, for a, and a consonant symbol, for the initial glottal stop ' ; this is what he has reference to when he says that "the 'a letter . . . has its own sound," i.e. its own vowel sound, along with the initial glottal stop phoneme, which means that as a graph it may, indeed should, be classified both under the consonants and under the vowels.

    23 Renou, op. cit., p. 2. Ibid., p. 99.

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    The several instances in the texts reviewed above in which the gsal byed of the original ?loka text has been reworked into some more meaningful grammatical construction, with various translations above, can be understood best in the context of the traditional Indic glosses to vyanijana "consonant": "le terme est glose par vyanjayanti ( = prakatan kurvanty arthan . . ; par uparisthayina tena vyanigam (lire vyaizgyam) ((ce qui se laisse manifester par la (voyelle) situee au-dessus> . . . ; cf. aussi sparsosmabhir vyajyaman ... , dit du langage (vac) ((qui se manifeste au moyen des occlusives et des spirantes ).)25

    The Smra sgo of Smrtijfianakirti26 may also be cited for a valuable instance of the use of gsal . . .byed expanded into a larger syntactic unit and used in grammatical commentary; the passage, from a gloss to the SCP's statement on stete-de, sloka 13, may be found cited in the commentary of the Mahapan.dita of Si-tu, ed. Das, page 20, lines i iL- 12: gan zizg gi ni snza ma la / gsal bar byed pa dam bcah ba, "dam bcah ["definition"; Bacot: "promesse," Inaba: risshlu i1:w] is [that which,] in respect to something, makes [it] clear antecedently [Inaba: aru koto ni tsuite, saki ni / akiraka ni suru mono ga "risshui" de aru]." On dam bcah ba, var. bcas pa -=pratijiia, MVP No. 4464; cf. MVP No. 7121, dam hchah ba = pratijanite; the citation in Das' edition of the Mahapan.dita of Si-tu has once dam bcas, once dam hchah; Inaba has normalized them both to dam bcah, which I have followed here. The examples in the commentary of the Mahapandita make it clear that what the Smra sgo passage refers to is "definition," in particular the process by which a syntactic structure with ste4te de is used to identify the topic, which comes first (snza ma la), and which is then defined (gsal bar byed pa) by what follows. The element bcah4 bcas4hchah in the term dam bcah is most likely to be referred to bcas pa "together with, connected with, having, possessing, containing, . . with dan or termin." (H. Jaschke, Tibetan-English Dictionary [reprinted London, 19491 s.v.) and originally here had nothing to do with the bca, bcas of

    25Ibid., p. 148. 26 Kyoto Tibetan Tripitaka reprint, No. 5784; correct Inaba, op. ti., p. 31. Cf. his

    paper, - -t 1 ; Qd 0 _ Q --D6fi ("An As- pect of Grammatical Study in the oeuvre of the Sa-skya Pandita"), Otani shigaku }kZ?

    ,T 8(1961).7 and G. N. Roerich, The Blue Annals (Calcutta, 1953), I, pp. 16o, 165, i66. Professor Inaba drew our attention to these facts in the course of his remarks at a meeting of the Nihon Chibetto Gakkai in Kyoto on September 28, 1958.

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    dam bca(s) "promise, solemn vow." The SCP sloka 13 :4 has de ni lIhag (var. Ihag ma) danz bcas paho //, and probably the dam of the com- mentators' dam bcah is a result of an assimilation of the final -ni of the original sloka text's danz to the initial b- of the following word.

    This syntactic structure is illustrated by the Mahapan.dita with an example which has most recently been discussed by Simonsson, sanls rgyas te ma rig pahi ginid sains pa daii ses bya la blo gros rgyas paho, which he renders "Buddha sein bedeutet aus dem Schlaf des Nicht- Wissens erwacht . . . sein und in bezug auf das Wissbare . . . einen aufgebliihten . . . Sinn besitzen."27 Inaba independently sought to bring out the force of the te here in his translation with the Japanese hotoke ni shite ["as for the Buddha"], which he amplified with hotoke to wa sunawachi ["that (which is called) the Buddha, i.e."], in brackets in his translation; the example, it should be noted, derives ultimately in content if not in syntactic structure from the statement in the sec- tion on buddha = sains rgyas in the Sgra sbyor bam-po giiis-pa. Inaba illustrates the construction with several citations from Tibetan trans- lations of Buddhist Hybrid Sanskrit, including the following from the Tibetan Abhidharma-kosa, nzag gi ni tshig smra bar bya ba la dbanz po inid mayin te / slob pahi khyad par la Itos pahi phyir ro /-Abhidharma- ko?a vyakhya vacas tu nendriyatvam vacane / Siksa-viseesapeksatvat.28

    It is interesting in this connection to read the oldest European account of the Tibetan writing system, in the Alphabetum Tangutanum sive Tibetanum, published by the Sacred Congregation for the Propa- gation of the Faith in Rome, 1773, for its remarks on gsal byed and dbyaizs: "Consonantes litterae habent Tibetani omnino triginta, easque appellant gsal byed [in Tibetan script in original-RAM] Selce, idest Clarificantes, forsan quod originem, & naturam vocabulorum osten- dant, & ideo clarificantes dicantur, quia simili fungantur munere, quo funguntur radicales in Hebraica, Syrica, aliisque Orientalibus linguis;

    . . Vocales a Grammaticis Tibetanis vocantur dbyaiis [in Tibetan script-RAM] jang, & etiam yan lag [in Tibetan script-RAM] jen lah, idest soni, & cantus; eaeque quatuor tantum numerantur, scilicet I, U, E, O.... Hae vocales sive breves sint, sive longae, semper eadem nota signantur" (pp. 12, i8, 19). The association of dbyanzs and yan lag here must have been due to some knowledge, on the part of the

    27 Simonsson, of. Cit., p. 266. 28 Inaba, of. cit., p. 213.

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    authors of this passage or their ultimate Tibetan informants, of MVP No. 6390, sas(tyanga-sahasropeta-svarah = dbyans yan lag drug stonl danz Idan pa, for which the Chinese version has chii liu-ch'ien chihyin XAf:aWi (rd. yin chih W.#) "possessing 6,ooo varieties of sound." At any rate, in the light of these passages, and in the sense of the equivalence gsal byed = vyanijana, SCPs'loka 1 :2 with its juxtaposition of ali with gsal byed becomes intelligible; 'i sogs bzi, "the four, 'i etc." are the vowel graphs which give concrete expression to, manifest, de- fine, make "clear" (gsal byed) the members of the set ali, "a and the other vowels, i.e. the set of vowels." The number given here in the SCP sloka text is "four" partly because there are only four overt vowel graphs in the Tibetan script, but also because (and these two facts are closely interrelated) the a vowel has already just been men- tioned in the sloka in the word ali itself, which is a plus ali, and hence now need not be repeated; for this reason the statement is a plus the set of four beginning with i, for a total of five. The author of the SCP was no mean student of the Indic ideal of economy in grammatical statement, though subsequent confusion of script and language on the part of his commentators has tended to obscure the elegance of his statement here.

    The total and striking absence of ali and kali as grammatical terms in the MVP is important for the further clarification of the relationship of the SCP and its author to the MVP. It is clear that the SCP sloka was unknown to the compilers of the MVP; or else we must assume that they knew it but that they chose to neglect its Indic grammatical terminology, which seems the less likely circumstance. It does appear, on the other hand, that the author of the SCP ?loka under considera- tion was familiar with the MVP, and at least he is seen here to be using two canonical MVP equivalents, yi ge (this, to be sure, not a very critical choice), and gsal byed, both in the sense of their Indic equiva- lents as registered in the MVP. Elsewhere we have considered the evidence for the historicity and dating of the two early Tibetan gram- matical treatises commonly attributed to Thon-mi Sambhota,29 and there is no need to repeat it here, except to note that this indication that SCP sloka 1 is to be dated after the MVP, since it supposes a prior knowledge of the MVP and its equivalents, fits in well with the general

    29 See my paper, "Thon-mi Sambhota and His Grammatical Treatises," JAOS


    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    conclusions about the SCP reached on the basis of other evidence. But what of the terms ali and kali themselves, which all the com-

    mentators agree are Indic, and which Dharmabhadra dubs "Sanskrit" (legs sbyar) ? Both are as unknown to the conventional Sanskrit lexical sources as they are to the MVP; they are to be found registered only in Franklin Edgerton's Buddhist Hybrid Sanskrit Grammar and Diction- ary, Vol. ii, Dictionary (New Haven, 1953), s.v., where he lists for ali, in addition to the MVP equivalence yur phran, a second meaning which is more to the point here: "(2) a-series (i.e. a plus ali), name for a series of syllables (chiefly vowels and combinations of a or a with semivowels), used as a magic formula in Sadh." So also for kali, which he registers as "(ka plus ali), ka-serie5, name for a series of syllables beginning with ka,"' again with the reference to the Sadhanamala. This text, ed. Benoytosh Bhattacharyya,30 offers (see its Index, s.v.) at least twelve citations of the two terms as "vowel" and Tconsonant," in various sections of this huge collection; only one citation, however, is to be found in a datable text, No. 251 by Advayavajra, f. cir. 978- 1030 A.D.

    The two terms ali and kali are found to be generously employed in all parts of the Sadhanamala collection as part of the mnemonic jargon of the Vajrayanists, who used them in the characteristic state- ments of their mantras, for examples of which see, in the Bhattachary- ya edition, numbers 73, io6, i8o, 210,i225,228,239,244,246 (Edger- ton's citation for both terms, since it defines, i.e. lists the members of both ali and kali), 250,251, and 252.

    But further consideration of the Sadhanamala's mzntra-statements at once brings us back to the phonological and morphological state- ments of the SCP. Bhattacharyya translates (Vol. ii, p. lxii; text, Vol. I, p. 335) the Bija-mantra of Sarasvati from the Sazhdanamala as fol- lows: "It stands on the second syllable of the seventh; and is the fourth of the eighth, it is accompanied by the fourth of the first and decorated with the spot." He adds, "The explanation seems to be: the second syllable of the seventh class (Antahstha) is R, fourth of eighth (Usma) is H, fourth of the first (Svara) is I, the spot is M and therefore the resultant B-ja is HRIM, . . the B-ja of SarasvatL." Another example of the same technique, loc. cit., page lxxiii: "It ends in H placed on Fire, is pierced by the fourth vowel and is accompanied

    30 [- Gaekwad's Oriental Series, Vol. I (No. xxvi) (Baroda, 1928) 1.

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    by the head of a spot and the half moon. This first B-ja is a great Bija. Now hear of the second which ends in T with Fire, is pierced by the same and is accompanied by Nadabindu. The third also I state care- fully which ends in H trampling on the sixth vowel and is accom- panied by Neadabindu. This B-ja is the most powerful and is able to set the three worlds on fire. I state now as was done before by the Buddha, the fourth syllable which ends in PHA and gives all kinds of perfec- tions. In order to complete the Mantra hear the half syllable ending in T which is devoid of A and by mere utterance saves all." Bhat- tacharyya adds, "The first B1ja consists of H, R (fire), I (fourth vowel), H (spots) and M and the resultant syllable is H R I M H .... The second B-ja includes T, R, (fire), I and M which together makes T R I M .... The third has H, U (sixth vowel), and M, which to- gether give H U M .... The fourth is PHA while the fifth letter is T which is deprived of its A and therefore considered as a half letter. The last two will give P H A T. So the whole Mantra stands as ... Hriihs Triih Hum Phat ... ."; As Bhattacharyya wrote, in what if anything is rather a surprising understatement, "These verses are extremely curious and give practically no meaning to ordinary read- ers" (loc. cit., p. lxxi). Curious though they are, they do provide a valuable parallel to the equally curious form in which many phono- logical and morphological statements of the SCP are cast. Inaba has sought in vain3l in the texts attributed to Thon-mi Sambhota for any trace of the sivasiitra systems, and as already pointed out above, their absence from these early texts must be regarded as significant; later of course the translations of Candragomin-school texts made the en- tire Paninian system familiar to the Tibetan grammarians. But what is to be found in the two early Tibetan grammatical texts is instead of the sivasiitra rather a reworking toward grammatical ends of the mnemonic phonological jargon of the Vajrayanists.

    The SCP arranges the consonant phonemes of classical Tibetan into seven and a half vargas of four each (sloka 3), and thereafter identifies these phonemes, in morphological statements, as "the last two of the 1st, 3rd and 4th [vargas]," (sloka 4), i.e. g, R, d, n, b, m; "the third of the 6th (ibid.), i.e. h; "the 7th [varga] minus s" (ibid.) i.e. r, 1, s; "the first of the 3rd" (sloka 13), i.e. t. The vowels too are

    31 Inaba, op. cit., p. 12. My own comment on this point in Language 31.3(1955).479 was wide of the mark.

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    numbered; o is 4 (sloka 7), u is 2 (sloka 8), and e is 3 (sloka 13). Most strikingly reminiscent of the Vajrayanist texts are the SCP's morpho- logical process statements, especially when they involve this number- ing system; one example, SCP sloka 8, will suffice:

    8.1 rjes hjug yi ge bcu po las 2 gain minf mthah na bcu pa gnas 3 de la 'ah ii giiis pa sbyar 4 gant minz mthah na brgyad pa gnas 5 de la giiis /a 'u yafz sbyar 6 gail min- mthah na gsum pa gnas 61 de bzin min giihi bdun pahnk gnas 7 de la 'ah li ginis pa sbyar 8 bzi pa dgu pa dnzos kyanz ste 9 las daii ched dan rten gnas danz

    1o de inid tshe skabs la sgra yin 1 "Among the ten affixed letters, 2 when the tenth [s] is placed at the end of any word, 3 to this the second vowel [u] is added; 4 when the eighth [r] is placed at the end of any word, 5 to this the second vowel u is added; 6 when at the end of any word the third [d] and 61 in this way too when the seventh radical consonant [t]

    is placed at the end of a word, 7 to these the second vowel [u] is added; 8 the fourth [n] and the ninth [1], however, are used

    by themselves, 9 for accusative, dative, locative of place,

    10 cognative object, and locative of time, [these comprising] the morpheme la."32

    Other passages which could be cited in this same connection in- clude SCP sloka 12 (kyi, hi, yi - i + ii = kyaii, haii, yain), sloka 13 (su - u + t + e = ste), sloka 15 (na, la + s = nas, las), sloka i8 (n + i = ni), sloka 19 (da + nt = daii), sloka 20 (d + e = de), and sloka 21 (ga + in = ganz).

    With this evidence it is no longer necessary to trouble ourselves 32 See Bacot and Inaba for the text and other versions; line 61 is, however, attested

    only in the text as printed in the Peking, 1957 edition of Blo-bzafl tshul-khrims cited above.

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    unduly over the problem of why the two grammatical treatises attrib- uted to Thon-mi Sambhota do not employ the sivasuitra or other Indic grammatical techniques for the identification and citation of phonemes and sets of phonemes: their Indic models on this score were ultimately found not among the texts of the Indic grammarians but among the mnemonic devices by which the Vajrayanists trans- mitted their mantras. Such important Tantric associations of the early Tibetan grammatical texts and their supposed author should, of course, occasion no particular surprise, in view of the over-all Tantric-Vajrayana character both of Tibetan Buddhism and of Indic culture as known in Tibet. In this connection it is interesting also to cite from the Blue Annals what at first appears to be a chance associa- tion of the Thon-mi Sambhota tradition with the Tantrics, but in view of what has been said above is probably of more than routine significance: ". . . when the religious council of Sambho-ra was held at Gfial it was attended by 1,200 priests who possessed copies of the text of the Nag-pohi rgyud = Krsayamaritantraraja, 8oo Tantrics possessing copies of the same book, in all 2,000 disciples."33 Sambho- ra, var. Sa-hbo-ra, is the name of a mountain in GCial, one of the places held by some to have been the birthplace of Thon-mi Sambhota.34

    The only additional text known to me from which ali, kali may be cited is the manuscript from the Tun-huang finds identified as Biblio- theque Nationale Dons 4502, Collection Pelliot No. 3531, published by Joseph Hackin under the title Formulaire sanscrit-tibe'tain du Xe siecle [= Mission Pelliot en Asie Centrale, Serie Petit in-Octavo, Tome ii] (Paris, 1924); beginning at Feuillet vi, line 94 through line 103 this text reads as follows:

    'a sid ti byi hdza na / ka kha ga hga na / tsa tsha dza hdzah iia / ta tha hda da nha / ta tha hda tha na / pa pha hba ba ma / hia ra la hbha sa sa sah khyk / > / ka ka / ki ki / ku ku / ke kahi / ko kahu / kafn ka / ((dio / mu ka sa ra / kaka swa ra / kum ba swa ra / mye ga swa ra / myirrgaswara/ >/sati/rastu/''aa'a/'i/'u'u / ri ri / li li / 'e 'ahi / 'o 'ahu /'am 'a sa ra / 'a nag khya ra / ka ya ba ga tsid ta ka ka sum cu rtsa bzi dafn // na stag bcu giiis dafn // sfnags iii s-u dafn / dbyaiis lIna dani //yi ge myi btub pa drug / dani // sku gsum thugs dai /

    33 Roerich, The Blue Annals, I, 376. 34 Cf. Alfonsa Ferrari, Mk'yen brtse's Guide to the Holy Places of Central Tibet [ = Serie

    Orientale Roma, xvil (Rome, 1958), p. 126, n. 258.

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    de ni pe byad / bzaii Po brgyad cu ho // / dah ba tini sad da lag sa na / da ra gan ti /// mtsan bzaii Po sum cu rtsa gnis ni / 'a li ka li /

    If we rewrite this curious text slightly, chiefly by substituting normal Sanskrit forms following Hackin for "les formes batardes du texte" (op. cit., p. vi) and "correcting" the Tibetan version of the Sanskrit alphabet to the usual values of the letters, we have:

    asitivyanijana / k kh ggh n / cch jh i / I dh 1/ t th d dh n / p p/h b bh m /y r 1 v s' s s h ks / -di? / ka kaz / ki ki / ku kiu / ke kai / ko kau / kam kah / ((di? / mukhas[v]ara / kzkasvara / kumbhasvara / meghasvara / mrgasvara / -di)) / siddhii/ rastu / a / ia/ u u / r f /

    / e ai / o au / am ah / sada / naksara / kaya vzk citta/ [...] dvatrimsratlaksana [.. .]ali kali /

    Unfortunately Hackin in his 1924 edition provided virtually no commentary for this strange mixture of lists and bilingual glosses, except for his "Observations relatives a un essai de transcription de l'alphabet sanscrit," pages 86-89, which contents itself with bringing some order into the exceedingly loose version of the Sanskrit alphabet which this section of the Formulaire has preserved. Without attempt- ing a complete commentary, certain items in this fragment of the Formulaire require some short annotation if we are to control the text sufficiently to study the notice of azli, kezli with which it concludes.

    Hackin remarks correctly that "le gi-gu inverse caracterise 1 'i, le gi-gu normal la longue." (This reversed i-graph is indicated in the transcription above by i.) For Sanscrity the Formulaire has the strange writing hia, of which Hackin writes simply "noter la transcription"; but it is more than likely that what actually stood in the Tun-huang manuscript was simply an unusually expansive writing of the Tibetan ya-graph, which looked to Hackin like an ha-graph followed by a ia- graph. Otherwise the prefixed Tibetan h- in this version of the San- scrit alphabet list follows more or less consistently modern Tibetan orthographic practice, indicating the voiced pronunciation for the ga-graph as against the voiceless with a low tone when without the prefix. Probably a line like ka kha ga hga ina was pronounced [ka kha (both high tones) kha (low tone) ga (low tone) ina (low tone)]. The total lack of any attempt at distinction among s / s / s is surprising, but sah is simply an error (or misreading in the edition) for sa ha. On asativyanijana, see Hackin's note, page 8o to line 87 of the Formu- laire, where it appears earlier. The mysterious


    or what appear to be quotation marks, both in Hackin's transcription of the text and in his type-set edition, is ignored in his commentary and translation; probably it is, with Tibetan d for t as is common in the Formulaire, an attempt at the grammarians' iti marking the end of enumerations. The set of five, mukhas[v]ara, kekasvara, kumbhasvara, meghasvara, and mrgasvara, seems otherwise not attested; it is prob- ably this set of five to which the dbyafns lina in the text below has reference (and hence dbyaiis lina here is not "five vowels".!), and is a list of tones of the voice or ways of pronunciation. Monier-Williams knows only kzkasvara "a shrill tone," and the MVP only meghasvara, No. 99, = hbrug sgra, Chinese lung-yin ' E} "dragon sound," where it is one of the names of the Tathagata (correct Edgerton, op. cit., s.v., who has "n. of a former Buddha"). The others seem to fit such an explanation: mukhasvara "mouth sound," kumbhasvara "pot sound," mrgasvara "wild deer sound," though it is impossible even to specu- late on what kinds of articulation these may have been intended to distinguish. The sadanaksara is explained later on in this same Formulaire fragment byyi ge myi btub pa drug; this must be a reference to the om mani padme hum formula, also called the sad-aksari; anak- sara is "unfit to be uttered, unable to articulate a syllable" in Monier- Williams, but better here "?ineffable." The kaya vak citta is later on glossed by sku gsum thugs.

    Of greater linguistic interest is the kaka sum cu rtsa bii, which Hackin translated "les trente-quatre alphabets" (p. 38); the reference here clearly is to the thirty-four consonant graphs of the Sanskrit alphabet, counting ks as the final one (this is why Blo-bzafi tshul- khrims [above] states that Sanskrit has thirty-three or thirty-four). Hackin was unable to translate na stag bcu giiis, for which he put simply "les douze . . . ," (p. 38-39), but there can be little doubt that it is a reference to the twelve Sanskrit phonemes a 2 i i u Ft e ai o au m and h, as listed in the Formulaire before the second -di-. (The second list, following the last -div, is the usual and complete one, including r etc. in their proper places following Ft.) The term na stag itself re- mains obscure, but probably we are to read [s]na stag. The resulting "nose + tiger" is to be understood in terms of the fanciful names for the vowel graphs used, for example, in the fragment of one of the grammatical texts of Bsod-nams rtse-mo (1142-1182) cited in the commentary of Blo-bzail tshul-khrims, page 52, lines 20-22 in the

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    Peking, 1957 edition: senl ge hdra daii brgya byin mig / klaii sna hod ma hgreni hdra dani / thani gug / ['a] which is 'like unto a lion,' and 'Indra's eye' [i], 'elephant trunk' [u], 'like a light not erect' [e], and 'bent down flat' [o] ... ." The Tibetan-Sanskrit dictionary of Tse-rin dbafi-rgyal (ed. Bacot, p. 92b) registers under the Tibetan-Sanskrit entry na ro = dan.da another Tibetan gloss, sna ru had (read: hod) zer, which must be a reference to the same terminology; sna ru, var. na ro are the ordinary terms for the Tibetan superscript o-graph. The "twenty mantra" are otherwise unknown to me, but here seem to be without linguistic interest. The "five dbyafns" as explained above probably here are not the five vowels of Tibetan but the intercalated list in the Formulaire beginning with mukhas[v]ara. For pe byad etc. read dpe byad bzafn Po brgyad cu, the title of chapter xviii in the MVP, Nos. 268 ff., asity-anuvyanijanani = Chinese pa-shih chung hao ming- hao Ait^w~iMS . For the dvatrimsat laksana = mtsan bzani po sum cu rtsa giiis, see Hackin's commentary page 8o to line 87 of the Formulaire, and the elaborate and definitive treatment in Edgerton, op. cit., pages 458a-460a; the list corresponds to MVP chapter xvII No. 235 ff. The da ra gan ti of the text remains obscure; one might suggest that it is intended for *tadagan[i]ti, but this is no particular help. Since the reversed i-graph is used for short i, and both Sanskrit a and a are indifferently written in this fragment with Tibetan a, the writing 'a la ka 1i can be taken as dli, kali.

    In other words, ali, kali in the tenth-century Tun-huang Formu- laire are simply the last two items given there in a miscellaneous and more than a little confused list; they are not otherwise defined in this list and have no organic relationship to any other part of the section of the Formulaire text to which they form the conclusion. Their occurrence in this unusual and datable document is of great impor- tance, however, and it is only regrettable that their appearance there is not distinguished by the introduction of any further information on them.

    An otherwise unknown term for "consonant" appears to be attested in Johannes Schubert's translation (MSOS 31 [1928].6, footnote lb) of the passage from the Dpag bsam Ijon bzanz, ed. S. C. Das (Calcutta, 1908), 2.167, paragraph 2, in which the activities of Thon-mi Sam- bhota are described. He translates "er stellte die acht Lautklassen fest und brachte, nachdem er die 30 Zeichen zum Schreiben der

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    Buchstaben gebildet hatte, die 30 Gsal-byed oder Ai bar phyi mo (Konsonanten), die mit k(a) beginnen, die vier Dbyafis (Vokale) i usw., die acht Rnam-dbye (Kasus) sowie im Hinblick auf die Anwend- ung des Rtags (Geschlechts) die Frahigkeit (thob-than-, eigentlich Gabe?) des Geschlechts von Pho., Mo, Ma-nifn bei Praffixen, Affixen und Basisbuchstaben in ein System."' But the text for this passage, loc. cit., line 15 ff., reads as follows: sgrahi skor brgyad mdzad par grags pa las sum cu par rgyahiyi ge la bri ba stan byas te ka nasya [sic] hi bar p/hyi mohm gsal byed sum cu dani 'i sogs dbyanls bzi danz / rnam dbye brgyad daz rtags hjug tu snlon rjes milz gzihi pho mo ma nini gi nags kyi thob thaln gtan la phab paho //, and there is no term "ai -bar 'phyi -mo (Konsonanten)" attested here. The yahi of Das' edition is almost surely to be emended (as indeed Schubert has in a sense done, despite his translation) to 'ahi; bar is the "interstice, interval" from k (ka nas) to ', i.e. the consonant graphs of the Tibetan alphabet; phyi mo is to be taken here in the sense of rtsa ba "root, origin, basis, fundament," and is so defined in the Tibetan-Chinese dictionary cited above, page 532b, where it is also glossed by Chinese ken-pen fJR. It is this term, phyi mo in sense of rtsa ba, which the text gives as an alternate (hence h[a]m) for gsal byed, no doubt because rtsa ba is part of the full Tibetan title of the SCP; hence we must translate "the thirty bases (phyi mo) or consonants (gsal byed) [of the] interval between k and ' ." The correction of Schubert's translation of this portion of the passage in question would take us too far afield, but even if only in brief it must be noted that the passage names and describes the two early Tibetan grammatical treatises, and once this is understood most of its difficulties are cleared up. Thus, sum cu par here is a reference to the SCP, which the text states was written in Indic letters (rgyahi yi ge); what immediately follows is a kind of epitome of the contents of the SCP, which covers the thirty con- sonants, the vowels and the eight cases (the / should be moved from after bz-i dani to after rnam dbye brgyad dani); rtags hjug is a reference to the Rtags kyi hjug pa, the second of the grammatical treatises at- tributed to Thon-mi Sambhota, and what follows from here to the end of the passage cited is an epitome of the contents of this second text.

    As far as the two Indic terms in question themselves are concerned, Edgerton's Dictionary suggests, both for ali and for kali, a and k + ali "ditch, row," reminding one of the Tibetan equivalents 'a-phreli

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    and ka-phreii, both with Tibetan (4)phreii "thread, string, cord"; perhaps too there has been some confusion here with the MVP's Tibetan equivalent yur phran for ali "small ditch." At any rate, the idea seems to be based on the concept of a class of phonemes as a "line," "row" or "column (in a table)," straight and regular as, for example, a string or a small irrigation ditch, and named after and with the phoneme at the head of the "line" or "row". A. H. Francke long ago pointed out that as early as the eighth century the Tibetan alpha- bet was being taught and written in virtually the same order as today,35 and the Central Asian alphabet practice-pages which he edited show how old in Tibet the Indic concept of arranging letters in "lines" and "crows" is. Another etymology for Indic ali and kali is possible, how- ever, which would remove ali "ditch, row" from the question entirely and is probably to be preferred. This is to associate it with Skt. adi "beginning, commencement; beginning with; beginning with C,"86 which would make ali < *a + azdi and kali < *k + adi. The phonology here presents no particular problems. Tibetan at the period of the texts in question did not distinguish between the dental and the retro- flex series, so that there was no d/4 contrast in Tibetan, and later Sanskrit dialect mixtures leading to replacements of intervocalic -1- for -d- as well as -d-, of the type resulting in sets of doublets such as gola "a ball," gula, guda "a globe, ball; raw sugar," gudika, gutika, gulika, gudaka, (all) "a ball (as a missile) ," cf. Buddhist Hybrid Skt. golika, gaulika "a shop for selling treacle" < Skt. .guda + ika, are well documented in the handbooks witl multiple examples.37 As a result of an interdialectical borrowing process the d/l doublets are especially reminiscent of such sets as Skt. (Panini) lipi- "Schrift" from Old Persian dipi- and Classical Skt. khola- "Regenhut" from Avestan xao5a "Hut."83

    This line of conjecture may be substantiated by only one citation, but this is an important one which lends a considerable degree of

    35 A. H. Francke, "Aus der tibetischen Schule Alt-Turkestans," MSOS 32(1929).149. 36 I am indebted to my colleague Warren C. Cowgill, who first pointed out to me the

    relevant glosses in B6htlingk's Kiirzerer Fassung, which had until then escaped my attention.

    37 Albert Thumb, Handbuch des Sanskrit, I (Heidelberg, 1905), P. 91, ?122.3; Louis Renou, Grammaire sanscrite, I (Paris, 1930), p. 59.

    38Jakob Wackernagel, Altindische Grammatik, i. Lautlehre (Gottingen, 1896), p. 222, ?194.b.

    This content downloaded from on Wed, 1 Oct 2014 17:31:06 PMAll use subject to JSTOR Terms and Conditions



    plausibility to the developments suggested. In his Terminologie gram- maticale du sanskrit, troisieme partie (Paris, 1942),3 where he deals exclusively with grammatical terms attested in the technical treatises of the Vedic literature, Louis Renou registers under the rubric vyanjana "consonne" (pp. 147-148) the entry "[les vyanjana sont] decrites par l'expression khdi 4k et les phonemes qui suivent>> V[ajasaneyi Prati&akhya] I 47." The passage itself reads vyanjanam kadi, which Albrecht Weber translated "k u. s. w. heissen vyanjana" in his article "Das Vajasaneyi-Pratikakhyam," in the Indische Studien 4 (1858), p. 113. The Vajasane~yi Prati&akhya lists the consonants in VIII 14 if., Weber, op. cit., p. 324, but here as atha vyanijanani, kiti khiti giti ghiti niti kavarga .. ., "Nunmehr die Consonanten. / die Gutturalen. / . . . ," etc. For the eight simple vowels this same text interestingly enough employs the fictive designation sim, which Renou cites from no other source exceptthe VajasaneyiPratisakhya, I 43, Weber, page 111, and IV 50, Weber, page 227, and concerning which he simply writes "(fait sur sama, sima?) ."4I This text appears to be the only evidence for the form which can be cited, and *adi, "*a et les phonemes qui suivent" is absent both from Renou's Terminologie and from the Vajasaneyi Pratigakhya; but this clue does indeed reassure us in our postulation of earlier Indic *adi